¹ Smith, James P., “Race and Human Capital,” American Economic Review, 74 (09 1984), table 12, p. 695.Google Scholar

² Ibid.table 4, p. 688.Google Scholar

³ See Darity, William A., “The Human Capital Approach To Black-White Earnings Inequality: Some Unsettled Questions,” Journal of Human Resources, 18 (Winter 1982), pp. 72–93;Google ScholarKiefer, David and Phillips, Peter, “Race and Human Capital: An Institutionalist Response” (unpublished manuscript, University of Utah, 1985).Google Scholar

⁴ The income ratios are from Smith, “Race,” table 12, p. 695. Smith assigns fixed race-specific weights to the census occupation distributions and ignores changes over time in the occupational wage structure. I use Smith's estimates uncritically, though a similar procedure yields a greatly distorted picture of male-female wage ratios over time.Google Scholar See Claudia Goldin, “The Earnings Gap in Historical Perspective” (unpublished manuscript, University of Pennsylvania, 1985).Google Scholar

⁵ The 1890 and 1900 censuses report the distribution of months attended in the following manner: less than 1 month, 2 to 3 months, 4 to 5 months, and 6 months and over. Average months attended is estimated by multiplying the share of individuals in each interval by the interval's midpoint. The midpoint of the interval “6 months and over” is assumed to be 6.5 months.Google Scholar

⁶ National data on average days attended per enrolled pupil are given in U.S. Bureau of the Census, Historical Statistics (Washington, D.C., 1975), p. 375. Comparisons with the 1890 and 1900 censuses suggests the days attended series is biased downwards by 10 to 15 percent. Average, months attended in 1910 is estimated by (days attended per enrolled pupil in 1910 × 1.10)/(20 school days per month). The black-white ratio of average months attended in 1910 was assumed to be the same as in 1900. Within race, the female-male ratio of average months attended in 1910 in each age group was assumed to be the same as in 1900.Google Scholar

⁷ The 1910 census gives information on p_ij for single years of age. The sum Σ_j = 5²⁰p_ij estimates years of schooling. The estimate is unchanged if data for age groups are used instead of data for single years of age and suggests estimating lifetime months using data for age groups introduces no serious bias.Google Scholar

⁸ The difference between these two figures, 27.7 months, can be divided into the effect of race differences in school attendance rates and race differences in months attended. Had black and white school attendance rates been equal at every age, the racial gap in lifetime months would have been 7.5 months. Thus race differences in school attendance rates account for 74 percent (20.2/ 27.7) of the racial gap in lifetime months.Google Scholar

⁹ The regression is: where BL is the black literacy rate, M is an estimate of total months of school from age 5 to age 14, and D is a dummy variable for 1910 observations. Absolute values of t-statistics are shown in parentheses. The construction of M at the state level differs from the aggregate estimates in Table 2 by relying on a single year of data rather than interpolating between census years. M is lagged ten years behind BL; for example, M in Alabama in 1900 is related to BL in Alabama in 1910. Thus Mis a proxy for time spent in school prior to the reporting of census literacy. The estimate of the constant term is peculiar but may reflect upward biases in the census literacy data for blacks (alternatively, measurement error in the M variable because of interstate migration). See Margo, Robert A., “Educational Achievement in Segregated School Systems: The Effects of “Separate-ButEqual”, “ National Bureau of Economic Research Working Paper No. 1620 (Cambridge, Mass., 1985).Google Scholar

¹⁰ The native white literacy rate in 1910, ages 15 to 24, was 98 percent, see U.S. Bureau of the Census, Negro Population in the United Suites, 1790–1915 (Washington, D.C., 1918), p. 423. According to the regression in fn. 9, a 98 percent ilteracy rate would require 33 months, compared to an actual level of 50.1 months for whites born between 1886 and 1890 (see Table 2).Google Scholar

¹¹ See Margo, Robert A., Disfranchisement, School Finance, and the Economics of Segregated Schools in the United States South, 1890–1910 (New York, 1985).Google Scholar

¹² Margo, “Educational Achievement.”Google Scholar

¹³ Welch, Finis, “Educational and Racial Discrimination,” in Ashenfelter, Orley, ed., Discriminalion in Labor Markets (Princeton, 1973), p. 59.Google Scholar

¹⁴ Perlmann, Joel, “Black Patterns of Schooling in a Northern City: Providence, RI., 1880–1925” (unpublished manuscript, Harvard University, 1985).Google Scholar

¹⁵ The census bureau believed that school attendance rates in 1900 may have been understated; if so, my estimates of lifetime months would be too low. Generous allowances for bias, however, do not change the substantive results. For example, increasing black attendance rates by 20 percent in 1900 (this implies no change in the black-white ratio of attendance rates from 1900 to 1910, which is inconsistent with other evidence) still implies that months per grade was less for blacks than for whites.Google Scholar

¹⁶ See Kitagawa, Evelyn M. and Hauser, Phillip H., Differential Mortality in the United States (Cambridge, Mass., 1973);CrossRefGoogle ScholarRosen, Sherwin and Taubman, Paul, “Changes in the Impact of Education and Income on Mortality in the United States,” in Del Bene, Linda and Scheuren, Fritz, eds., Statistical Uses of Administrative Records With Emphasis on Mortality and Disability Research (Washington, D.C., 1979).Google Scholar

¹⁷ Denison, Edward, The Sources of Economic Growth in the United States and the Alternatives Before Us (New York, 1962), p. 70. Another possibility is adult education, although school attendance rates after age 25 are too low to explain the bias.Google Scholar See Denison, p. 79, fn. 11.Google Scholar

¹⁸ This estimate is an unweighted average of the annual school terms given in U.S. Bureau of the Census, Historical Statistics, p. 375.Google Scholar

¹⁹ One way to gauge the size of mortality and creep bias is to compare cohort-specific schooling levels across census years. Median years of schooling among blacks born from 1911 to 1920 (30–39 years of age in 1950) rose 0.35 years between 1950 and 1960: the corresponding increase for whites was 0.05 years. This is the maximum increase observed for any age group, excluding 25 to 29 year olds (some individuals in this age group were still in school). I will assume these figures describe the average decade-to-decade mortality and creep bias for the 1886 to 1890 cohorts. Multiplying 0.4 by 3 (the number of decades from 1910 to 1940), subtracting from 4.9 (mean black attainment, 1886 to 1890 cohorts) gives 3.7 years of schooling. The calculation for whites gives 7.7 years. The new estimates of months per grade are 5.8 months for blacks and 6.5 months for whites. These adjustments for mortality and creep bias are too high; census attainment data reveal no evidence of creep or mortality bias between 1950 and 1960 among late nineteenth-century cohorts.Google Scholar

²⁰ Bond, Horace Mann, The Education of the Negro in the American Social Order (New York, 1934), pp. 339–44. Unadjusted for age, the achievement gap was nearly three grades. In other words, black pupils were generally overage for their grade, which is consistent with the shorter black school year.Google Scholar

²¹ U.S. Bureau of the Census, Sixteenth Census of the United States: 1940, Population, Volume IV: Characteristics By Age, Part I: United States Summary (Washington, D.C., 1943), p. 178 (emphasis added).Google Scholar

²² Calculated from data in Report of the Superintendent of Public instruction, State of Texas (Austin, 1900), pp. 71–85.Google Scholar

²³ This estimate of the southern black school year is a weighted average of state-level estimates, with the weight equal to the state's share of total southern black enrollment. The data necessary for the calculation are given in Margo, Disfranchisement;Google ScholarDu Bois, W. E. B. and Dill, A. G., The Common School and the Negro American (Atlanta, 1911); and U.S. Bureau of the Census, Negro Population.Google Scholar

²⁴ U.S. Office of Education, Report of the Commissioner of Education (Washington, D.C., 1912), p. 11;Google ScholarU.S. Bureau ofthe Census, Negro Population, p. 381.Google Scholar

²⁵ Welch, “Education and Racial Discrimination,” p. 58. Precise dating of the shift to graded schools is difficult. However, many southern states began reporting enrollment statistics by grade levels around 1910.Google Scholar

²⁶ Smith, “Race,” table 4, p. 688.Google Scholar

²⁷ The estimate of 6.4 additional years is based on a 7.2-month school year as a standard for passing a grade. The true increase was larger since black pupils took longer than a single school year to complete a grade.Google Scholar

²⁸ The racial schooling gap, 4 years, is biased downwards because it assumes blacks could complete a grade in the same number of months as whites; see fn. 27.Google Scholar