¹ The male and female employment weights used to compute these pay differentials from Immigration Commission earnings data come from the U.S. Census employment statistics in the same year (1909). (See Census Monographs IX, Women in Gainful Occupations, 1870 to 1920, Washington, D.C., 1929.) This use of Census weights avoids the possible objection that the Immigration Commission samples of employment may have been biased as relative measures of male and female employment in these industries.Google Scholar

² We argued in our 1980 article that women and men coming from southern and eastern Europe, Ireland, and French Canada had less non-human capital and familiarity with the new country than did immigrants from northern and western European countries. Hence if such characteristics induced immigrant women in general to work harder and more steadily, and to acquire more job-related skills, they should also have induced immigrant women of the first type to work harder, and so on, than did immigrant women of the second type. Thus, women of the first type should have had higher standardized earnings relative to those of men than did women from northern and western Europe. We did not find this to be the case in our earlier results. But the validity of this test depends upon the absence of complicating factors, such as discrimination against women from southern and eastern Europe, so the test was scarcely conclusive.Google Scholar

³ Results for immigrants in 1975 using a “human capital” equation (schooling and experience as the primary independent variables) are almost identical for the male-female pay differential as those shown. One of the authors, Michael Tannen, is also in the process of completing a larger story comparing the economic situation of immigrants in 1909 with those in 1975 using these data. Findings there should shed more light on the issue.Google Scholar

⁴ Strictly speaking, there is no simple method of comparing some of the coefficients in panel A with those in panel B, Table 1, since the former are obtained by regressing dummy variables on logs of individual wages whereas the latter come from regressing cell means on logs of cell mean wages. The following approximation, however, can serve as a guide in making comparisons if we assume that there are no sources of bias present in our results. For individual earnings in 1975, the coefficient on English proficiency implies that males speaking English had a. 2378 higher log of earnings, ceteris paribus, than those who did not. Evaluated at the mean of the log of earnings, this translates into 26.8 percent higher (non-log) earnings. If we then assume that one additional male in 100 spoke English, average cell earnings would rise by. 268 percent if we grouped by cells in 1975. This compares with alog. 00431 increase in the average earnings of males in 1909 if there was an addition of one worker speaking English to the cell. This translates to a. 431 percent increase (non-log) actual earnings of the cell (also evaluated at the mean, as in 1975). Thus, English proficiency apparently raised earnings more in 1909 than in 1975. The ratio of. 431 percent to. 268 percent, which we have just derived, is very close to the ratio of. 430 to. 238 which we would obtain if we shifted the decimal place in panel B two places to the right (to adjust for the fact a cell in which everyone speaks English in 1909 is measured by 100, whereas a person speaking English in 1975 is measured by a dummy variable value of 1). Some readers, therefore, might want to compare panel A and B coefficients by shifting the decimal place two points to the right in the panel B coefficients for cell means. There is always error of a non-linear type in doing this, however, because of the different data characteristics noted at the beginning of the footnote.Google Scholar

⁵ These results are based upon the use of value of 1 to indicate the presence of a characteristic in the 1975 results, and a value of 100 percent to indicate the presence of a characteristic in the 1909 results.Google Scholar

⁶ One of the writers, Paul McGouldrick, has written a detailed analysis of how government at all levels created for the first time skill-acquisition and job-commitment disincentives for women relative to those for men in the 60 years between 1909 and 1969. See “Why Women Earn Less,” Policy Review, Fall 1981. Of course, this may not be the whole story, but it is likely to be a large part of the story.Google Scholar

⁷ For example, plain and automatic looms differed in job structure, with weavers doffing bobbins on plain looms but delegating this job to less-skilled assistants when assigned to automatic looms. But many plain loom weavers had higher skills than automatic loom weavers, since automatic looms were used for coarser yarn counts and simpler cloth structures. Then too, more skilled workers in spinning departments would be assigned more spindles and would sometimes supervise assistants. Spinning frames also differed in speed (and therefore end-breakage rates) by vintage and counts of yarn spun. Many further examples of heterogeneous jobs masked by the same job titles could be cited from the abundant technical literature for the textile and clothing industries.Google Scholar

⁸ United States Congress, Senate (1910–1913), Report on Condition of Women and Child Wage Earners in the United States, S. Doc. 645, 61st Congress, 2nd Session.Google Scholar

⁹ In an early version of our 1980 paper, we presented much the same evidence by occupation as Thornton and Hyclak, and gave some reasons why we were dissatisfied with it. A referee pointed out the extent of the piece rate system to us, which was all the additional evidence we required to drop completely these calculations from our 1980 paper.Google Scholar

¹⁰ A description of these data are contained in Wise, Lauress L., McLaughlin, Donald H., and Steel, Lauri, The Project TALENT Data Bank Handbook, Revised Edition, American Institutes for Research (Palo Alto, California, 03 1979). The sample we use was drawn from ninth grade students in 1960, so that our data base is 1974 (the 1960 students scheduled to spend three more years in high school and then 11 years forward).Google Scholar