Race, Skin Color, and Economic Outcomes in Early Twentieth-Century America Statistics

(a) The Spencers in 1910 (b) Isaac and John Spencer in 1940 Figure 1: The Spencer brothers in 1910 and 1940 Figure 2: Geographic distribution of BMS families, by county Note: BMS are households with at least two sons aged 3–18 in 1910 where at least one classified “Black” and at least one classified “Mulatto.” .8 .6 .4 .2 0 Proportion of sons classified as "Mulatto" <1 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 Age Sample is all "black" or "mulatto" sons 18 years old or younger that live in a household with at least one "mulatto" and one "black" son in 1910 (a) Proportion of “Mulattoes” in households with both “black” and “mulatto” sons, by age (b) Older siblings who are Black more likely to attend school Figure 3: Evidence consistent with skin-color based classifications 0 .1 estimated density .2 .3 .4 .5 Annual income from salary by race, 1940 5 50 Annual income White 500 5000 Black Income truncated at $5,000 in census Kernel-density estimates using Epanechnikov kernel and .25 bandwidth Males with positive incomes Source: IPUMS 1940 (a) Income distribution 0 500 Annual income 1,000 1,500 2,000 2,500 Income differences by race for selected occupations, 1940 Laborers Janitors Carpenters Truck drivers SalespeopleTeachers Doctors White Black Income truncated at $5,000 in census Males with positive incomes in selected occupations Source: IPUMS 1940 (b) Income from labor by occupation and race Figure 4: Income from labor by race, 1940 Table 1: Racial compositions of African-American couples and parents in Black-and-Mulattosons (BMS) families (a) In population, proportion Black children in mixed families consistent with color coding ,ĞĂĚ ůĂĐŬ ůĂĐŬ DƵůĂƚƚŽ DƵůĂƚƚŽ WĂƌĞŶƚƐΖĐŽůŽƌ tŝĨĞ йŽĨĐŽƵƉůĞƐ ůĂĐŬ ϳϰй DƵůĂƚƚŽ ϴй ůĂĐŬ ϰй DƵůĂƚƚŽ ϭϰй ŚŝůĚƌĞŶΖƐĐŽŵƉŽƐŝƚŝŽŶ EĐŚŝůĚƌĞŶ йůĂĐŬ й&ĞŵĂůĞ Ϯ͘ϭϭ ϵϵ͘Ϯй ϱϬ͘Ϭй Ϯ͘ϬϮ Ϯϳ͘ϱй ϰϵ͘ϳй Ϯ͘ϭϳ ϰϰ͘Ϯй ϰϵ͘ϳй Ϯ͘ϭϮ ϭ͘ϲй ϰϵ͘ϴй (b) Families with both Black and Mulatto sons, by parents’ race (percent of 6,047 families) &ĂƚŚĞƌŝƐ 17 ůĂĐŬ DƵůĂƚƚŽ tŚŝƚĞ KƚŚĞƌ ďƐĞŶƚ ůĂĐŬ ϭϱ͘ϳϳ ϭϴ͘ϲϵ Ϭ͘ϭ DƵůĂƚƚŽ ϯϮ͘Ϯϭ ϵ͘ϭ Ϭ͘Ϭϳ DŽƚŚĞƌŝƐ tŚŝƚĞ Ϭ͘Ϭϴ Ϭ͘Ϭϯ Ϭ͘Ϭϱ ϭϯ͘Ϭϭ ϰϳ͘ϱϳ ϲ͘ϲϯ ϰϴ͘Ϭϭ Ϭ͘ϭ Ϭ͘Ϯϳ KƚŚĞƌ Ϭ͘ϬϮ Ϭ͘Ϭϯ Ϭ͘Ϭϱ ďƐĞŶƚ Ϯ͘ϰϳ ϭ͘ϲ Ϭ͘ϬϮ dŽƚĂů ϱϬ͘ϱϱ Ϯϵ͘ϰϳ Ϭ͘Ϯϯ ϰ͘Ϭϵ ϭϵ͘ϳϱ ϭϬϬ Notes: 1. Numbers in 1a are reported for the whole population of couples where the head and his wife were classified as either Black or Mulatto in 1910. 2. Numbers in 1b are percentages of the 6,047 households where some sons were classified as Mulatto and the other sons were classified as Black. 3. Blank cells contain no households. 4. Single-parent households in1b show the missing parent as “Absent” Table 2: Attributes of African-American households (1910) ŝĨĨĞƌĞŶĐĞƐŝŶďĂĐŬŐƌŽƵŶĚĐŚĂƌĂĐƚĞƌŝƐƚŝĐƐďĞƚǁĞĞŶƐĂŵƉůĞƐ͕ϭϵϭϬ ZĂŶĚŽŵ EĞŝŐŚďŽƌƐ ^ĂŵƉůĞ͗ ^ŝďůŝŶŐƐ ,ŽƵƐĞŚŽůĚĐŽůŽƌ͗ ůĂĐŬ DƵůĂƚƚŽ ŽƚŚ ůĂĐŬ DƵůĂƚƚŽ ;ϭͿ ;ϮͿ ;ϯͿ ;ϰͿ ;ϱͿ ^ŽŶƐŝŶŚŽƵƐĞŚŽůĚ ϯ͘ϱϰ ϯ͘ϱϰ ф ϯ͘ϴϮ х Ϯ͘ϴϬ Ϯ͘ϴϵ ŐĞ ϵ͘ϵϭ ϭϬ͘ϭϭ ф ϭϬ͘ϭϵ ďůĞƚŽǁƌŝƚĞ Ϭ͘ϲϮϮ Ϭ͘ϳϭϴ х ,ĞĂĚŝƐůŝƚĞƌĂƚĞ Ϭ͘ϱϲϮ Ϭ͘ϲϲϰ ,ĞĂĚΖƐŽĐĐƵƉĂƚŝŽŶ͗ Ͳ&ĂƌŵůĂďŽƌĞƌ Ϭ͘Ϭϰϲ Ͳ&ĂƌŵĞƌ ϵ͘ϵϭ ϵ͘ϵϬ Ϭ͘ϱϵϱ Ϭ͘ϱϰϵ Ϭ͘ϲϳϲ  Ϭ͘ϱϲϴ Ϭ͘ϱϭϬ Ϭ͘ϲϯϲ Ϭ͘Ϭϯϴ ф Ϭ͘Ϭϰϵ Ϭ͘Ϭϱϴ Ϭ͘ϬϰϮ Ϭ͘ϱϵϬ Ϭ͘ϱϭϯ ф Ϭ͘ϲϭϵ Ϭ͘ϲϭϵ Ϭ͘ϱϴϬ Ͳ>ĂďŽƌĞƌ;ŶŽƚĨĂƌŵͿ Ϭ͘ϭϲϯ Ϭ͘ϭϱϭ х Ϭ͘ϭϯϱ ф Ϭ͘ϭϰϴ Ϭ͘ϭϯϱ Ͳ/ŶŚŽŵĞƐĞƌǀŝĐĞƐ Ϭ͘Ϭϰϳ Ϭ͘Ϭϳϭ  Ϭ͘Ϭϲϲ х Ϭ͘Ϭϱϭ Ϭ͘Ϭϱϵ ,ĞĂĚŽǁŶƐŚŽŵĞ Ϭ͘Ϯϰϰ Ϭ͘ϯϴϬ  Ϭ͘Ϯϳϴ Ϭ͘Ϯϰϳ Ϭ͘ϯϲϳ ŽƵŶƚLJƵƌďĂŶ Ϭ͘ϭϵϭ Ϭ͘Ϯϯϲ х Ϭ͘ϭϲϭ Ϭ͘ϭϱϮ Ϭ͘ϭϲϭ ŽƵŶƚLJĨƌ͘ŵĞƌŝĐĂŶƐ Ϭ͘ϰϲϬ Ϭ͘ϰϭϱ ф Ϭ͘ϰϴϳ Ϭ͘ϱϯϱ Ϭ͘ϱϭϭ KďƐĞƌǀĂƚŝŽŶƐ ϴ͕ϵϯϵ ϳ͕ϯϰϱ ϭϲ͕ϭϲϱ ϭϬ͕ϮϮϱ ϭϵ͕ϵϭϬ х ф Note: Mean characteristics are reported for the sample of random sons in households with at least two sons all Black or all Mulatto (columns (1) and (2)), sons in households with at least one Black and one Mulatto son (column (3)), and sons that are neighbors of sons in the Black-and-Mulatto sons households. Table 3: Characteristics of matched vs. all searched individuals ŝĨĨĞƌĞŶĐĞƐŝŶϭϵϭϬĐŚĂƌΖƐďĞƚǁĞĞŶƐĞĂƌĐŚĞĚƐĂŵƉůĞĂŶĚŵĂƚĐŚĞĚƐƵďƐĂŵƉůĞ͕ƉŽŽůĞĚƐĂŵƉůĞƐΎ DĂƚĐŚϭ hŶǁĞŝŐŚƚĞĚ tĞŝŐŚƚĞĚ DĞĂŶ ŝĨĨĞƌĞŶĐĞ DĞĂŶ ŝĨĨĞƌĞŶĐĞ ;ϮͿ ;ϯͿ ;ϰͿ ;ϱͿ ͲͲͲϮϬ͕ϱϳϰͲͲͲ ͲͲͲϭϰ͕ϴϲϳͲͲͲ ͲͲͲϮ͕ϵϮϱͲͲͲ DĂƚĐŚϮ;ƐƚƌŝĐƚĞƌͿ hŶǁĞŝŐŚƚĞĚ tĞŝŐŚƚĞĚ DĞĂŶ ŝĨĨĞƌĞŶĐĞ DĞĂŶ ŝĨĨĞƌĞŶĐĞ ;ϲͿ ;ϳͿ ;ϴͿ ;ϵͿ ͲͲͲϲ͕ϲϭϯͲͲͲ ͲͲͲϯ͕ϮϯϰͲͲͲ ͲͲͲϰϬϵͲͲͲ Ϭ͘ϱϵϳ Ϭ͘ϱϴϰ ϭϬ͘Ϭ ϭϬ͘Ϭ Ϭ͘ϲϭϰ Ϭ͘ϲϯϭ Ϭ͘ϬϮϳ ΎΎΎ ;Ϭ͘ϬϬϳͿ Ϭ͘ϲϮϰ Ϭ͘ϬϮϬ ΎΎΎ ;Ϭ͘ϬϬϳͿ Ϭ͘ϲϳϮ Ϭ͘Ϭϲϱ ΎΎΎ ;Ϭ͘ϬϭϬͿ Ϭ͘ϲϱϮ ϯ͘Ϯϲ ϯ͘Ϯϵ Ϭ͘ϬϱϮ ΎΎΎ ;Ϭ͘ϬϭϰͿ ϯ͘Ϯϳ Ϭ͘Ϭϯϯ ΎΎ ;Ϭ͘ϬϭϰͿ ϯ͘ϯϰ Ϭ͘Ϭϴϴ ΎΎΎ ;Ϭ͘ϬϮϯͿ ϯ͘Ϯϴ Ϭ͘Ϭϯϭ ;Ϭ͘ϬϮϮͿ ,ĞĂĚŝƐůŝƚĞƌĂƚĞ Ϭ͘ϱϳϯ Ϭ͘ϱϴϲ Ϭ͘ϱϳϯ Ϭ͘ϬϬϳ ;Ϭ͘ϬϬϱͿ Ϭ͘ϲϮϳ Ϭ͘ϬϲϬ ΎΎΎ ;Ϭ͘ϬϬϴͿ Ϭ͘ϱϳϯ ,ĞĂĚΖƐŽĐĐƵƉĂƚŝŽŶ͗ Ͳ&ĂƌŵůĂďŽƌĞƌ Ϭ͘ϬϮϬ ΎΎΎ ;Ϭ͘ϬϬϱͿ Ϭ͘ϬϬϳ ;Ϭ͘ϬϬϴͿ Ϭ͘Ϭϰϴ Ϭ͘Ϭϰϲ ͲϬ͘ϬϬϯ ;Ϭ͘ϬϬϮͿ Ϭ͘Ϭϰϳ ͲϬ͘ϬϬϭ ;Ϭ͘ϬϬϮͿ Ϭ͘Ϭϰϯ ͲϬ͘ϬϬϱ Ύ ;Ϭ͘ϬϬϯͿ Ϭ͘Ϭϰϳ ͲϬ͘ϬϬϮ ;Ϭ͘ϬϬϰͿ Ͳ&ĂƌŵĞƌ Ϭ͘ϱϵϭ Ϭ͘ϱϵϬ ͲϬ͘ϬϬϮ ;Ϭ͘ϬϬϱͿ Ϭ͘ϱϵϮ Ϭ͘ϬϬϬ ;Ϭ͘ϬϬϱͿ Ϭ͘ϱϴϴ ͲϬ͘ϬϬϰ ;Ϭ͘ϬϬϴͿ Ϭ͘ϱϵϱ Ϭ͘ϬϬϯ ;Ϭ͘ϬϬϴͿ Ͳ>ĂďŽƌĞƌ;ŶŽƚĨĂƌŵͿ Ϭ͘ϭϰϳ Ϭ͘ϭϰϲ ͲϬ͘ϬϬϭ ;Ϭ͘ϬϬϰͿ Ϭ͘ϭϰϳ ͲϬ͘ϬϬϭ ;Ϭ͘ϬϬϰͿ Ϭ͘ϭϯϵ ͲϬ͘ϬϬϵ Ύ ;Ϭ͘ϬϬϱͿ Ϭ͘ϭϰϳ ͲϬ͘ϬϬϭ ;Ϭ͘ϬϬϲͿ Ͳ/ŶŚŽŵĞƐĞƌǀŝĐĞƐ Ϭ͘Ϭϲϭ Ϭ͘Ϭϱϴ ͲϬ͘ϬϬϱ ΎΎ ;Ϭ͘ϬϬϮͿ Ϭ͘Ϭϲϭ ͲϬ͘ϬϬϮ ;Ϭ͘ϬϬϮͿ Ϭ͘Ϭϰϴ ͲϬ͘Ϭϭϱ ΎΎΎ ;Ϭ͘ϬϬϯͿ Ϭ͘Ϭϱϵ ͲϬ͘ϬϬϯ ;Ϭ͘ϬϬϰͿ ,ĞĂĚŽǁŶƐŚŽŵĞ Ϭ͘Ϯϴϳ Ϭ͘ϯϭϮ Ϭ͘Ϭϯϳ ΎΎΎ ;Ϭ͘ϬϬϱͿ Ϭ͘Ϯϴϳ Ϭ͘ϬϭϮ ΎΎΎ ;Ϭ͘ϬϬϱͿ Ϭ͘ϯϰϱ Ϭ͘Ϭϲϲ ΎΎΎ ;Ϭ͘ϬϬϴͿ Ϭ͘Ϯϴϲ Ϭ͘ϬϬϲ ;Ϭ͘ϬϬϳͿ ŽƵŶƚLJΖƐƐŚĂƌĞƵƌďĂŶ Ϭ͘ϭϳϯ Ϭ͘ϭϳϯ Ϭ͘ϬϬϭ ;Ϭ͘ϬϬϮͿ Ϭ͘ϭϳϯ Ϭ͘ϬϬϭ ;Ϭ͘ϬϬϮͿ Ϭ͘ϭϴϮ Ϭ͘Ϭϭϭ ΎΎΎ ;Ϭ͘ϬϬϰͿ Ϭ͘ϭϳϱ Ϭ͘ϬϬϰ ;Ϭ͘ϬϬϰͿ ůůŽďƐ KďƐĞƌǀĂƚŝŽŶƐ;ĐŽƵŶƚͿ ŽƚŚĐŽůŽƌƐŝŶĚŝƐƚƌŝĐƚ ŽƚŚĐŽůŽƌƐŝŶĨĂŵŝůLJ ŚĂƌĂĐƚĞƌŝƐƚŝĐƐŝŶϭϵϭϬ ůĂĐŬ ŐĞ ďůĞƚŽǁƌŝƚĞ ηƐŽŶƐŝŶŚŽƵƐĞŚŽůĚ DĞĂŶ ;ϭͿ ϲϭ͕Ϭϱϱ ϱϭ͕ϱϭϭ ϮϬ͕Ϭϲϭ ͲϬ͘Ϭϭϴ ΎΎΎ ͲϬ͘ϬϬϱ ;Ϭ͘ϬϬϰͿ ͲϬ͘ϬϬϲ ;Ϭ͘ϬϬϰͿ Ϭ͘ϱϲϬ Ϭ͘ϬϮϮ ;Ϭ͘ϬϯϵͿ Ϭ͘ϬϮϭ ;Ϭ͘ϬϰϬͿ ϭϬ͘ϭ ϭϬ͘Ϭ ͲϬ͘Ϭϰϭ ΎΎΎ ;Ϭ͘ϬϬϳͿ Ϭ͘Ϭϱϰ ;Ϭ͘ϬϲϬͿ Ϭ͘ϱϵϳ ͲϬ͘ϬϬϯ ;Ϭ͘ϬϬϳͿ ϭϬ͘ϭ Ϭ͘Ϭϲϯ ;Ϭ͘ϬϲϯͿ Ϭ͘Ϭϰϱ ΎΎΎ ;Ϭ͘ϬϭϬͿ Notes: 1. Column (1) reports statistics of background characteristics for our samples in 1910. Columns (2)-(5) reports the characteristics for sons matched to the 1940 census using one matching rule (Match 1) and columns (6)-(9) reports them for sons matched to the 1940 using a second matching rule (Match 2). 2. The top panel of the table reports the number of observations in the baseline 1910 sample and in the matched samples. The first row reports the overall number while the following rows report the number of observations that are in a district (row 2) or households (row 3) where both Black and Mulatto sons are found. 3. Columns (4) and (5) repeat the results of columns (2) and (3) re-weighting matched observations according to how likely they were to be matched. See footnote 22 for the list of variables used to calculate these weights. Similarly, columns (8) and (9) show re-weighted statistics for columns (6) and (7). Each pair of columns reports the mean characteristic in the matched sample and the result of a t-test for the difference in the characteristic between the matched and unmatched groups. 4. Match 1 requires each observation’s best potential match to have a link score of at least 0.5 and the second best potential match to have a score that is lower by at least 0.1. Match 2 requires the best match’s link score to be at least 0.75 and the second best match to be at least 0.3 lower or to be at least 0.15 lower if only the best match’s first and last names’ Soundex codes match. See Appendix A for further details. - * p<0.10, ** p<0.05, *** p<0.01 Table 4: Literacy higher for children classified as Mulatto in the population; family background differences drive most differences in population Dependent variable: Literacy, 1910 Comparing between: General population (1) (2) Black (vs. Mulatto) -0.093 *** (0.010) -0.039 *** (0.010) Boys Brothers (3) -0.013 (0.008) Neighbors of brothers (4) (5) -0.094 *** (0.017) -0.069 *** (0.017) General population (6) (7) Girls Sisters (8) Neighbors of sisters (9) (10) -0.128 *** (0.009) -0.022 *** (0.008) -0.116 *** (0.025) -0.079 *** (0.009) -0.087 *** (0.026) Head is literate 0.238 *** (0.010) 0.141 *** (0.019) 0.178 *** (0.009) 0.146 *** (0.035) Family owns home 0.098 *** (0.011) 0.148 *** (0.023) 0.116 *** (0.010) 0.130 *** (0.042) -0.263 *** (0.027) -0.076 (0.057) -0.149 *** (0.026) -0.209 ** (0.106) Farmer -0.206 *** (0.019) -0.005 (0.041) -0.120 *** (0.017) -0.100 (0.084) Laborer -0.110 *** (0.018) -0.029 (0.041) -0.054 *** (0.016) -0.118 (0.073) Home services -0.042 * (0.023) -0.011 (0.045) -0.022 (0.020) -0.103 (0.071) Head is employer 0.006 (0.018) 0.003 (0.037) 0.003 (0.017) -0.057 (0.061) Head works on own account 0.008 (0.018) -0.032 (0.036) 0.029 * (0.016) -0.023 (0.058) Household head occupation: Farm laborer Main fixed effects: None None &ĂŵŝůŝĞƐ ŶƵŵ͘ ĚŝƐƚƌŝĐƚƐ ŶƵŵ͘ ĚŝƐƚƌŝĐƚƐ None None &ĂŵŝůŝĞƐ ŶƵŵ͘ ĚŝƐƚƌŝĐƚƐ ŶƵŵ͘ ĚŝƐƚƌŝĐƚƐ Observations R-squared 8,480 0.042 8,078 0.163 6,875 0.094 4,545 0.058 4,290 0.105 8,256 0.070 7,919 0.153 7,220 0.118 1,777 0.077 1,684 0.131 Mean[Literate | Mulatto] Mean[Literate | Black] 0.718 0.622 0.718 0.622 0.614 0.605 0.667 0.555 0.667 0.555 0.832 0.702 0.832 0.702 0.752 0.706 0.732 0.563 0.732 0.563 Notes: 1. Literacy is taken from the question on whether an individual can write. Enumerators were instructed to ask each household member over 10 years old whether they could read or write. 2. All samples taken from the 1910 Census. “General population” columns ((1),(2),(6),(7)) report results for a 1% random sample of children classified as either Black or Mulatto aged 3–18 in families that had at least two sons aged 3–18 classified in the same color. Siblings columns ((3),(8)) report results for the universe of either sons or daughters aged 3–18 in households that have at least one more son or daughter color-coded differently. Neighbors of siblings columns ((4),(5),(9),(10)) report results for children close in age and of equal gender who also live in the same enumeration district in 1910. 3. In columns (3)-(5) and (8)-(10) only children that have other children of the opposite color from the same group (family or enumeration district) with non-missing literacy are included. 4. The omitted category for household head occupation is all other occupations. The omitted category for the employment type (head employer or works on own account) is those employed by others. All regressions control for a set of age and birth order (among sons) dummies. 5. Standard errors are clustered at the county level for the general population and enumeration district level for the brothers and neighbors regression. - * p<0.10, ** p<0.05, *** p<0.01 Table 5: Lighter-skinned attain more years in school in population and between neighbors; less so within families Dependent variable: Years of schooling, 1940 General population Brothers (1) (2) Black (vs. Mulatto), 1910 Observations Mean[Education | Mulatto] Mean[Education | Black] -1.115 *** (0.175) -0.214 (0.320) Neighbors (3) -1.270 *** (0.447) 1,743 374 281 7.02 5.92 6.03 5.87 6.50 5.43 Notes: 1. The dependent variables are the eventual number of years in school as reported in the 1940 census. The variable of interest, Black, is taken from the 1910 census classification. 2. “General population” in column (1) reports difference in schooling for the of children aged 3–18 in 1910 who had siblings only of the same color classification and were in families randomly selected from the population. The sample is further restricted to those who were successfully matched to 1940 records. 3. “Brothers” in column (2) reports difference in schooling for the universe of sons aged 3–18 in households that have at least one more son color-coded differently with at least one son of each color classification successfully matched to 1940. 4. “Neighbors” in columns (3) report difference in schooling for sons who, in 1910, were close in age and lived in the same enumeration district as children from the “Brothers” sample . Only neighbors who were successfully matched with at least one other neighbor of the opposite color classification successfully matched entered the sample. 5. All regressions also control for age and birth order. Standard errors are clustered at the county level for the general population and enumeration district level for the brothers and neighbors regression. - * p<0.10, ** p<0.05, *** p<0.01 Table 6: Income premium for light skin explained by differences in educational attainment Comparing between: (1) Black (vs. Mulatto) Dependent variable: weekly labor income in 1939 (natural log) General population Brothers Neighbors (2) (3) (4) (5) (6) (7) (8) -0.116 ** (0.047) Years of schooling -0.028 (0.045) -0.157 * (0.095) 0.087 *** (0.006) 0.079 *** (0.008) Years of schooling X Black 0.046 (0.074) 0.222 (0.158) 0.091 *** (0.020) 0.106 *** (0.021) 0.019 (0.013) Main fixed effects: Observations R-squared 0.035 (0.087) Mean[wkly labor income | mulatto, >0] Mean[wkly labor income | black, >0] 1,173 0.148 -0.670 * (0.376) 0.092 *** (0.021) 0.047 (0.030) 206 0.012 203 0.216 0.083 (0.052) ŶƵŵĞƌĂƚŝŽŶĚŝƐƚƌŝĐƚƐ Families 1,173 0.150 -0.107 (0.119) -0.029 (0.024) None 1,188 0.008 -0.168 (0.130) (9) 203 0.225 111 0.063 110 0.279 $19.7 $15.4 $17.8 $17.6 $15.7 $16.2 110 0.307 Notes: 1. The dependent variable, ln(weekly salary), was computed as the natural log of the reported annual income from salary divided by the reported number of weeks worked. Only individuals with positive weekly income enter the analysis but no differences by color were found in whether income was positive or zero. 2. “General population” in columns (1)-(3) report results for the of children aged 3–18 in 1910 who had siblings only of the same color classification and were in families randomly selected from the population. The sample is further restricted to those who were successfully matched to 1940 records. 3. “Brothers” in columns (4)-(6) report results for the universe of either sons aged 3–18 in households that have at least one more son color-coded differently with at least one son of each color classification successfully matched to 1940. 4. “Neighbors of siblings” in columns (7)-(9) report difference in schooling for sons who, in 1910, were close in age and lived in the same enumeration district as children from the “Brothers” sample . Only neighbors who were successfully matched with at least one other neighbor of the opposite color classification successfully matched entered the sample. 5. Regressions also control for age (quadratic) and birth order among sons in the childhood households. 6. Standard errors are clustered at the county level for the general population and enumeration district level for the brothers and neighbors regression. - * p<0.10, ** p<0.05, *** p<0.01 Table 7: Income premiums robust to the imputation of non-wage income for occupations where wage income is relatively small Comparing between: (1) Black (vs. Mulatto) General population (2) -0.092 ** (0.038) Years of schooling -0.014 (0.036) -0.115 (0.072) 0.077 *** (0.005) 0.070 *** (0.006) Years of schooling X Black 0.032 (0.071) 0.073 (0.061) 0.054 (0.138) 0.092 *** (0.015) 0.090 *** (0.019) 0.016 (0.010) Main fixed effects: Observations R-squared Dependent variable: weekly income1 in 1939 (natural log) Brothers (3) (4) (5) (6) (7) Mean[wkly income | mulatto, >0] Mean[wkly income | black, >0] 1,519 0.147 -0.000 (0.090) -0.077 (0.235) 0.082 *** (0.016) 0.076 *** (0.024) 318 0.004 313 0.200 0.013 (0.037) ŶƵŵĞƌĂƚŝŽŶĚŝƐƚƌŝĐƚƐ Families 1,519 0.149 (9) 0.003 (0.022) None 1,542 0.009 -0.108 (0.092) Neighbors (8) 313 0.200 241 0.032 238 0.197 $17.0 $13.9 $14.2 $15.3 $13.8 $13.7 238 0.198 Notes: 1. The dependent variable – weekly income (in natural log) – is the natural log of the sum of the weekly income from labor as reported directly in the census and the median non-labor income for the occupation and region (South or non-South) of the individual. Median incomes calculated from the IPUMS sample of the 1950 census, and adjusted for 1940 prices (multiplied by 0.4). 2. The dependent variable, ln(weekly salary), was computed as the natural log of the reported annual income from salary divided by the reported number of weeks worked. Only individuals with positive weekly income enter the analysis but no differences by color were found in whether income was positive or zero. 3. “General population” in columns (1)-(3) report results for the of children aged 3–18 in 1910 who had siblings only of the same color classification and were in families randomly selected from the population. The sample is further restricted to those who were successfully matched to 1940 records. 4. “Brothers” in columns (4)-(6) report results for the universe of either sons aged 3–18 in households that have at least one more son color-coded differently with at least one son of each color classification successfully matched to 1940. 5. “Neighbors of siblings” in columns (7)-(9) report difference in schooling for sons who, in 1910, were close in age and lived in the same enumeration district as children from the “Brothers” sample . Only neighbors who were successfully matched with at least one other neighbor of the opposite color classification successfully matched entered the sample. 6. Regressions also control for age (quadratic) and birth order among sons in the childhood households. 7. Standard errors are clustered at the county level for the general population and enumeration district level for the brothers and neighbors regression. - * p<0.10, ** p<0.05, *** p<0.01 Table 8: Lighter-skinned more likely to migrate and leave South by 1940; Education gap can explain differences Skin color and migration 1910-1940 General population (1) (2) Black (vs. Mulatto) Gen. pop, within county Siblings Neighbors (3) (4) (5) (6) (7) (8) Dependent variable: Whether state in 1940 different than state in 1910 0.007 -0.075 * -0.014 0.035 0.101 -0.116 *** -0.147 (0.054) (0.040) (0.078) (0.042) (0.097) (0.031) (0.140) -0.023 (0.027) Years of schooling 0.023 *** (0.005) 0.029 *** (0.007) 0.019 * (0.011) 0.015 (0.015) Black X schooling -0.000 (0.007) -0.004 (0.010) -0.008 (0.014) 0.008 (0.021) 0.399 0.376 Mean[dep. | Mulatto] Mean[dep. | Black] Black (vs. Mulatto) 0.112 *** (0.028) 0.368 0.392 0.389 0.367 -0.024 (0.046) 0.436 0.347 Dependent variable: Lives in southern state in 1940 0.093 *** -0.022 -0.041 -0.049 (0.034) (0.062) (0.036) (0.082) 0.085 *** (0.027) -0.045 (0.113) Years of schooling -0.041 *** (0.005) -0.025 *** (0.006) -0.011 (0.011) -0.041 (0.011) Black X schooling 0.015 ** (0.007) 0.015 * (0.009) -0.001 (0.012) 0.013 (0.017) 0.575 0.669 Mean[dep. | Mulatto] Mean[dep. | Black] Observations 1,779 0.656 0.716 1,741 951 0.670 0.652 932 392 0.656 0.729 381 1,285 493 Notes: 1. The dependent variable in the top panel is an indicator for whether an individual’s residence in 1940 was in a different state than his residence in 1910. The dependent variable in the bottom panel is an indicator for whether a person resided in the South in 1940 where the South is defined as Texas, Louisiana, Mississippi, Alabama, Georgia, South Carolina, Florida, North Carolina, Tennessee, Virginia, Maryland, Arkansas, and the District of Columbia. 2. “General population” in columns (1)-(4) report results for the of children aged 3–18 in 1910 who had siblings only of the same color classification and were in families randomly selected from the population. The sample is further restricted to those who were successfully matched to 1940 records. Columns (3) and (4) adds county fixed-effects to avoid differences in initial location that are correlated with color to drive differences in 1940 location. 3. “Siblings” in columns (5)-(6) report results for the universe of sons aged 3–18 in households that have at least one more son color-coded differently with at least one son of each color successfully matched to 1940. 4. “Neighbors of siblings” in columns (7)-(8) report results for children who, in 1910, were close in age and of the same gender who also live in the same enumeration district as children from the “Siblings” sample . Only neighbors who were successfully matched with at least one other neighbor of the opposite color classification successfully matched entered the sample. - * p<0.10, ** p<0.05, *** p<0.01 Table 9: When samples restricted to siblings with higher likelihood to be share same parents, differences between siblings are unchanged Dependet variable: Literacy, 1910 Comparing between: Black (vs. Mulatto) Observations Black (vs. Mulatto) Observations All siblings (1) In hhods w. both parents Same last name as head Born current marriage Born to moms 30+ (2) (3) (4) (5) -0.013 (0.008) -0.015 (0.009) 6,875 5,143 -0.022 *** (0.008) 7,220 Boys -0.014 * (0.008) -0.018 ** (0.009) 6,211 Girls -0.025 *** (0.008) 5,278 6,393 -0.012 (0.011) -0.010 (0.016) 3,757 1,692 -0.010 (0.010) -0.008 (0.016) 3,906 1,598 Notes: 1. The dependent variable is literacy (being able to write), as recorded for individuals aged 10 years or older. 2. Column (1) has the full sample of siblings, and matches columns 3 (boys) and 8 (girls) in Table 4. 3. Column (2) limits the sample only to siblings in families where both parents were present. 4. Column (3) limits the sample to siblings who had the same last name as the household head. 5. Column (4) limits the sample to siblings whose age is strictly smaller than the length of current marriage of the head of household and the spouse. 6. Column (5) limits the sample to siblings born after the top female in the household (head or spouse) reached the age of 30. - * p<0.10, ** p<0.05, *** p<0.01 Table 10: Returns to passing-for-White in are high; even between siblings (1) Differences in income, 1939 (natural log) All observations Between siblings (2) (3) (4) (5) (6) 1 White in 1940 (passed) 0.273 *** (0.040) Years of schooling 0.185 *** (0.037) A. Weekly income -0.041 0.191 * (0.079) (0.097) 0.054 (0.089) -0.442 ** (0.216) 0.068 *** (0.004) 0.064 *** (0.004) 0.098 *** (0.016) 0.065 *** (0.022) 0.029 *** (0.009) White 1940 X years of scho ---- 3,948 ------- 15.0 ------- 22.3 ---- Observations 6 Mean[wkly income | black] 6 Mean[wkly income | white] White in 1940 (passed) 0.351 *** (0.045) Years of schooling 0.269 *** (0.042) 0.066 *** (0.006) ---- 293 ------- 17.0 ------- 23.5 ---B. Weekly labor income -0.063 0.130 (0.096) (0.125) 0.061 *** (0.006) 0.001 (0.116) -0.772 ** (0.313) 0.120 *** (0.020) 0.063 ** (0.027) 0.042 *** (0.010) White 1940 X years of scho Observations 6 Mean[wkly labor income | black] 6 Mean[wkly labor income | white] 0.065 ** (0.026) ---- 3,104 ------- 16.9 ------- 26.3 ---- 0.100 *** (0.032) ---- 184 ------- 19.6 ------- 26.0 ---- Notes: 1. Weekly income in Panel A is the sum of the weekly income from labor as reported directly in the census and the median non-labor income for the occupation and region (South or non-South) of the individual. Median incomes calculated from the IPUMS sample of the 1950 census, and adjusted for 1940 prices (multiplied by 0.4). 2. Columns (1) through (3) report results for all Mulatto children who have brothers, all aged 3-18, in 1910, matched to 1940. Columns (4)-(6) report results for children classified in 1910 as Mulatto who were classified as White in 1940 and their siblings who were classified as “Negro” in 1940. The latter category represented what would be classified in 1910 as either Black or Mulatto. Fixed effects at the household level were used in columns (4)-(6). 3. All regressions additionally control for age, squared age, and birth order among sons in the household in 1910. 4. The omitted category for White in 1940 has individuals classified as “Negro” in the 1940 census. All were classified as Mulatto in 1910. 5. All standard errors are clustered at the enumeration-district level. 6. Means of the income levels for the samples of columns (1) and (4) are presented at the bottom of each panel. Means are calculated only for strictly positive incomes. - * p<0.10, ** p<0.05, *** p<0.01