² McFeely, Mary D., Women's Work in Britain and America from the Nineties to World War I (Boston, 1982), P. xvii.Google ScholarOf course, the amount of time required before full proficiency could be achieved varied according to the job. For example, skilled buttonhole makers might have to work for three months before they could make buttonholes of high quality, and it might take a year before attaining maximum speed (Willett, Employment of Women, p. 83). In the production of smoking and chewing tobacco, on the other hand, “the occupation of the women may be learned, as a rule, in from a week to two months, according to the worker's intelligence.Google Scholar After that speed may be gained but there is little opportunity of advance.” Hutchinson, Emile J., Women's Wages (New York, 1919), pp. 53–55.Google Scholar

³ Thus, in the garment trades a boy ultimately might rise from basting and shaping to the status of a master tailor. Women's opportunities were limited in comparison: in the women's coat and suit trade, for example, a girl might be promoted from basting to sewing and then to cutting, but no further. In a department store it was possible to advance only from cash girl to stock girl, or from selling in the basement to selling on one of the upper floors. Smuts, Robert W., Women and Work in America (New York, 1959), p. 84;Google Scholar see also Beardsley, Elizabeth, Sales-Women in Mercantile Stores (New York, 1912).Google Scholar Early work on the emergence of hierarchically structured labor markets includes Ulman, Lloyd, The Rise of the National Trade Union (Cambridge, Massachusetts, 1955).Google Scholar

⁴ The differential by gender in the time path of earnings was well known to contemporary observers. Men's earnings were known to rise until at least the onset of middle age, but women's earnings were thought to peak considerably earlier. For example, among the women studied for the 1911 Report on the condition of women and child wage earners, earnings were observed to rise steadily through the age of 20. Between the ages of 21 and 24, there was some leveling off and an increasing dispersion of wage growth; in the words of the Report's authors, “this orderly sequence [of wage growth] is frequently interrupted.” From the age of 25, some women's wages continued to rise, but others who “are at an unmistakable disadvantage as compared with those aged 21 to 24” found their wages beginning to decline. Neill, Charles D., Report on the Condition of Woman and Child Wage Earners in the U.S., 61st Congress, 2d Session, Senate Document No. 645, (Washington, D.C., 1910), p. 25.Google Scholar

⁵ This article is an abridged version of a working paper of the same name available from the author on request. Many of the arguments in the introduction to this version are elaborated upon there.Google Scholar

⁶ Bureau of Labor Statistics of the State of California, Fifth Biennial Report (Sacramento, 1893).Google Scholar

⁷ Here I am using the term “occupation” broadly, to denote a range of related tasks undertaken in a specified line of work. Thus, we define garment work as an occupation and consider sewing, cutting and basting as different tasks undertaken in that occupation or line of employment. The term “current industry” might be used in place of “current occupation” if the service sector were not included.Google Scholar

⁸ California Bureau of Labor Statistics, Report, p. 10.Google Scholar

⁹ These are figures for California, from U.S. Commerce Department, Bureau of the Census, 11th Census of Population (Washington, D.C., 1897), Vol. 1, pt. 2, Table 84, pp. 372–75. Estimates were constructed by weighting cell means and assuming an average age of 70 for individuals at least 65 years old. If the Census figures are to be believed, this indicates undersampling of older workers. Yet there is no obvious reason to suspect that surveys done at the plant systematically omitted older workers. The other possibility is that the Census, when taken at the household, failed to elicit complete information on labor force participation of child and juvenile workers.Google Scholar

¹⁰ This figure for women closely coincides with those in other surveys from the Progressive Era. See for example Wright, C., Fourth Annual Report of the Commissioner of Labor: Working Women in Large Cities (Washington, D.C., 1888);Google ScholarNeill, Report.Google Scholar See also the discussion in Goldin, Claudia, “Female Earnings Functions and Occupations,” Explorations in Economic History, 21 (01 1984),1–27.CrossRefGoogle Scholar

¹¹ Unfortunately, no information is provided on discontinuities in labor force participation, occupational tenure, or service.Google Scholar

¹² Bureau of Labor Statistics of the State of California, Third Biennial Report (Sacramento, 1888).Google Scholar

¹³ Little depends on this assumption; if, for example, the investment ratio as opposed to investment itself is assumed to decline linearly over time, the estimating equation remains unchanged and its interpretation differs only slightly. For further details, see Mincer, Jacob, Schooling, Experience and Earnings (New York, 1974), pp. 86–87. and passim.Google Scholar

¹⁴ See Becker, Gary, Human Capital (Chicago, 1975), pp. 49–52 for further discussion of this influence on investment.Google Scholar See also Zellner, Harriet, “The Determinants of Occupational Segregation,” in Lloyd, Cynthia, ed., Sex Discrimination and the Division of Labor (New York, 1975), pp.125–45.Google Scholar

¹⁵ See, for example, Smuts, Women and Work, pp. 69–70. The probability of a quit should be understood to reflect not only voluntary movements between jobs or out of the labor force but also forced moves due to relocation of a girl's parents.Google Scholar

¹⁶ The worker's age is sometimes included in such regressions as a measure of maturity. However, it need not (and cannot) be included here because it equals EXI plus SCHOOLING plus six. It is tempting to interpret the conventional measure of schooling used here as a proxy for maturity, however, since maturity is perfectly collinear with the sum of EXI and SCHOOLING, whereas EXI is directly controlled for in the regression.Google Scholar

¹⁷ As noted above, an alternative interpretation of this result is that the returns to maturity differ insignificantly between men and women.Google Scholar

¹⁸ Results like this have been taken in the literature to indicate either the presence of discrimination or the influence of other variables, such as physical strength, for example, which affect productivity and are correlated with gender but are omitted from the equation. To the extent that the negative coefficient on FEMALE is an indication of discrimination, this could reflect either the practice of paying inexperienced women less for the same work or the channeling of them into occupations in which labor productivity was lower. See Aldrich, M. and Albelda, P., “Determinants of Working Women's Wages During the Progressive Era,” Explorations in Economic History, 17 (1980), 332–36. These same authors argue that requirements for physical strength may have reduced female productivity in jobs such as coarse weaving in the cotton textile industry and horizontal warping in the silk industry.CrossRefGoogle ScholarBlau, F. D., “Immigration and Labor Earnings in Early Twentieth Century America,” Research in Populalion Economics, 2 (1980), 39, emphasizes that physical strength is likely to be only part of the explanation for gender-specific differences in earnings, pointing out that many women's industrial occupations were physically arduous. A referee has suggested that FEMALE may be picking up the effects of omitted industry variables and that including industry of employment might reduce the coefficient on gender. Were this the case, it would help to identify where either occupational discrimination or physical strength was important, but it would not help to differentiate between the two interpretations.Google Scholar

¹⁹ There are two interpretations of this finding. The insignificant earnings differential between single and married women may reflect the absence of an incentive effect; presumably fewer married women anticipated permanent responsibility for providing sole support for their family. Alternatively, the positive coefficient on FSINGLE may reflect depreciation of job skills while out of the labor force. Upon marriage, a woman might withdraw from the labor force to keep house or have children. This would involve some deterioration of skills previously acquired; presumably single women experienced fewer spells out of the labor force and, consequently, had higher earnings. In contrast, marital status for males had no such correlation with spells of nonmarket activity.Google Scholar

²⁰ A variety of interpretations of job segregation are provided by, among many others, Mitchell, Juliet, Woman's Estate (New York, 1971);Google ScholarGordon, David M., Edwards, Richard, and Reich, Michael, Segmented Work, Divided Workers: The Historical Transformation of Labor in the United Stares (London, 1982);Google ScholarHartmann, Heidi, “Capitalism, Patriarchy and Job Segregation,” in Women and the Workplace: The Implications of Occupational Segregation, Blaxall, Martha and Reagan, Barbara, eds. (Chicago, 1976), pp. 137–69.Google Scholar

²¹ This is similar but not identical to the procedure in Blinder, Alan, “Wage Discrimination:Reduced Form and Structural Estimates,” Journal of Human Resources, 8 (1973), 436–55.CrossRefGoogle Scholar

²² Note that the coefficients on years with current employer have switched signs. Since neither of these coefficients is statistically distinguishable from zero at standard levels of confidence, however, no significance should be attached to this result.Google Scholar

²³ Note that the coefficients on EX2 ² differ insignificantly between females and young men.Google Scholar

²⁴ The returns to firm-specific experience are negative because the coefficients on EX3 and the squared term have switched signs. The effect is small, however, and, for reasons indicated above, no significance should be attached to this anomaly.Google Scholar