¹ Keynes, John Maynard, Essays in Biography, Collected Works of John Maynard Keynes, vol. 10 (London, 1972), p. 258Google Scholar.

² The monopsony argument is made by Madden, Janice Fanning, The Economics of Sex Discrimination (Lexington, MA), 1973Google Scholar.

³ U.S. Congress, Report of the Immigration Commission, Senate Doc. (Washington, D.C., 1911)Google Scholar, and U.S. Department of Commerce, Bureau of the Census. 1970 Public Use Sample, 15 percent one-in-a-thousand county group version (Washington, D.C., 1970)Google Scholar.

⁴ McGouldrick, Paul and Tannen, Michael, “Did American Manufacturers Discriminate Against Immigrants Before 1914?” this Journal, 3 (Sept. 1977), 723–46Google Scholar.

⁵ The Immigration Commission Report contains information on some native-born workers, but as will become apparent later, there are serious problems in explaining their wages through the use of the variables we employ.

⁶ The data do not include hours worked per week, but this should not pose a problem since hours worked for full-time workers were very similar by sex and also by industry. The 1970 data source does include part-time workers (and so a separate hours variable must be entered there), but the 1909 survey excludes them (respondents were asked to report normal work weeks and normal earnings).

⁷ The reason for not including these variables is that unpublished regressions incorporating these showed no statistically significant effects on wages. The ethnicity contrasts of these results with both the column B all-manufacturing ones and with our earlier study of men alone is striking, leading us to suspect that what was true for manufacturing as a whole was not true for the textile-clothing subgroup. For the latter had been using immigrant labor from the southern, and eastern parts of Europe for years and was also highly competitive.

⁸ See Michael Tannen, “Earnings, Skill, and Nativity: A Study of Single Women Prior to World War II,” unpublished manuscript, 1980.

⁹ Our primary interest is not in the degree of overall explanatory power shown by R², which will be affected by the weighting procedure, but rather in the statistical reliability of the individual regression coefficients. We also conducted a Chow test to determine whether the Table 1, Panel A regressions had coefficients for men and women identical except for sampling variation. This test indicated that we should reject the hypothesis that the coefficients were identical. Therefore, we used sex interaction variables, as explained in the text, to test systematically for differences in the coefficients.

¹⁰ Michael Tannen, “Earnings, Skill and Nativity,” found a noticeable superiority of earnings of single, second-generation native-born women over earnings of all other single women in retail sales during 1907 and 1908, after standardizing to remove effects of length of work year, experience, schooling, father's earnings, home ownership, and other family characteristics. This is consistent with the hypothesis that foreign-born women (or women born here of foreign-born parents) were willing to accept lower, apprentice-type wages in retailing in exchange for the opportunity to learn English on the job. Although the inclusion in the “all other” group of single women with one but not both parents born abroad clouds this result, it should not falsify it.

¹¹ The 1907–1908 Labor Department study, Woman and Child Wage Earners, published in 1909 had grouped data for both sexes by occupation, but lacked crucial skill proxies such as age and marital status. An 1895 study by the U.S. Commissioner of Labor, Eleventh Annual Report (Washington, D.C., 1895)Google ScholarPubMed, attempted to standardize men and women by skills and efficiency using a ranking system; but in an unpublished paper we reject the methods employed in that study as ill-defined.

¹² According to human capital theory, persons who intend to remain in the labor force longer (for example, men as compared to women) have a greater incentive to invest in human capital over a prolonged period and in more depth. Investment differences of this type should produce different ageearnings profiles for men and women. However, within a particular subgroup of low-to-medium skill industries, investment differences by men and women observed in grouped data are likely not to be very pronounced. A critical additional point is that the two data sets used in our regressions differ from those used in regressions in other studies in that they exclude many or most highly skilled and supervisory workers. In 1909 they exclude the native-born of both sexes, whereas in 1969 they exclude all workers in occupations where a minimum number of women were not employed. The latter exclusion had to be made to align the data set more closely with that for 1909. Thus, the true experience-earnings profiles for men and women could have been different in 1969 because men who had invested more in training and schooling were promoted from jobs in the scope to jobs not in the scope of the data sets. And in 1909, factors peculiar to immigration could have forced an equality of ageearnings profiles among immigrants that was not true of the native-born. Finally, it is entirely possible that age-earnings profiles by sex could vary significantly in a broader industry sample because of interindustry mobility and greater variation of skills.

¹³ For 1909 we have presented evidence of a tight covariation between age and experience for women, but the 1969 situation may well have been different. For 1969, however, we standardized for many other variables (including schooling, employment status five years earlier, marital status, and occupation) that should be correlated with experience, and the growth of the pay gap we observe is still so large that we do not believe that the missing experience data, if made available, would entirely eliminate that growth.

¹⁴ The R²s naturally are lower than those of Table 1 because we are now using individual rather than grouped observations.

¹⁵ This is a technical point that arises from the fact that age variables alone combine the effect of schooling and experience, whereas age variables in addition to a schooling variable allow separate coefficients to be applied to schooling and experience, subject to measurement error on the experience terms. This point, and the nature of the biases present in the age coefficients, is explained more fully in Mincer, Jacob and Chiswick, Barry, “Time Series Changes in Personal Income Inequality in the United States from 1939, with Projections to 1985,” Journal of Political Economy, 80 (May/June 1972, pt. 2), S34–66Google Scholar. Of course, this reason applies to the direction of change and not to the magnitude, and so a diminished level of investment in on-the-job training in these industries cannot be ruled out as a supplemental cause.

¹⁶ Following what has become an accepted line of reasoning, we can relate the labor force commitment of women to the lack of human or non-human capital held by their spouses or families, ceteris paribus. This would imply that women from southern and eastern Europe in 1909 had a greater commitment to paid work than did women from northern and western Europe; since men and families from the former regions had less non-human capital and lower skills. Hence, women from southern and eastern Europe should have earned more than those from northern and western Europe, on a standardized basis. When we tested this hypothesis by adding both ethnic and ethnic-sex dummy variables to the regressions, we found no statistically significant difference of the type postulated. Of course, if discrimination against workers from southern and eastern Europe had been present it might have offset the favorable effects on wages of superior work intensity.

¹⁷ A closing note seems in order. The proportion of all employed women who were working in these industries declined substantially between 1909 and 1969, mainly because clerical and service employment expanded so rapidly. Thus, the 1969 situation of women in clothing and textiles was undoubtedly much less representative of the situation for employed women in general than it had been earlier. On the other hand, this point does not weaken the implications of our findings for men and women who were at the bottom of the occupational ladder over the sixty-year period.