Hostname: page-component-cd9895bd7-gxg78 Total loading time: 0 Render date: 2024-12-16T15:26:19.008Z Has data issue: false hasContentIssue false
  • English
  • Français

Male-Female Wage Discrimination in Nineteenth-Century France

Published online by Cambridge University Press:  03 March 2009

Donald Cox  and
John Vincent Nye
Donald Cox
Affiliation:
Associate Professor of Economics, Boston College, Chestnut Hill, MA 02167
John Vincent Nye
Affiliation:
Assistant Professor of Economics, Washington University, St. Louis, MO 63130.
Get access Rights & Permissions [Opens in a new window]

Abstract

Traditional male-female wage discrimination measures rely on residuals from earnings functions that standardize for observable characteristics. But many productivity determinants are unobservable, and existing proxies for them are often difficult to interpret. Instead of using the earnings-function approach, we estimate production functions, using data from the 1839–45 and 1860–65 French industry censuses for textiles. While most of our findings cast doubt on the idea of discrimination against women in pay, they do not rule out some other forms of discrimination, such as occupational segregation.

Type
Articles
Information
The Journal of Economic History , Volume 49 , Issue 4 , December 1989 , pp. 903 - 920
Copyright
Copyright © The Economic History Association 1989

Access options

Get access to the full version of this content by using one of the access options below. (Log in options will check for institutional or personal access. Content may require purchase if you do not have access.)

References

We wish to thank seminar participants at Washington University, Northwestern, Caltech, and the Cliometrics sessions at the 1987 American Economic Association meetings in Chicago for helpful comments. We also wish to thank Paul Hohenberg and two anonymous referees for detailed comments and suggestions. We are grateful to Shin Cho for computational and research assistance. Of course, the authors are solely responsible for the final work.Google Scholar

1 Goldin's, Claudia “The Gender Gap in Historical perspective,” in Kilby, Peter, ed., Quantity and Quiddity: Essays in U.S. Economic History (Middletown, 1987), pp. 135–70, is the best survey in the literature of the historical pattern over the last century of relative female-male earnings in the United States.Google Scholar

2 Goldin, Claudia and Sokoloff, Kenneth, “Women Children, and Industrialization in the Early Republic: Evidence from the Manufacturing Censuses,” this JOURNAL, 42 (12. 1982), pp. 741–74,Google Scholarand “The Relative Productivity Hypothesis of Industrialization: The American Case, 1820–1850,” Quarterly Journal of Economics, 99 (08 1984), pp. 461–87.CrossRefGoogle Scholar

3 Oaxaca, Ronald, “Male-Female Wage Differentials in Urban Labor Markets,” International Economic Review, 14 (10. 1973), pp. 693–709. To illustrate, suppose the earnings functions for men and women are wm = Xmβm+∈m and wf = Xfβf + εf where wi = log wages, i = m, f; Xi= endowments of human capital; βi = returns to the endowments, and εi = stochastic terms. We can express the average wage difference as where δx = xm – xf and δβ = βm – βf The first term, δxβm, is the portion of the wage differential due to differences in endowments. The second term, xfδβ, is part of the wage differential due to differential returns to endowments. This is the portion of the wage differential thatOaxaca refers to as the “discriminatory” component.CrossRefGoogle Scholar

4 Ronald Oaxaca, “Male-Female Wage Differentials,” Mincer, Jacob and Polachek, Solomon, “Family Investments in Human Capital: Earnings of Women,” Journal of Political Economy, 82 pt. II (03. 04 1974), pp. S78–S108,CrossRefGoogle Scholarand Corcoran, Mary and Duncan, Greg, “Work History, Labor Force Attachment, and Earnings Differences Between the Sexes,” Journal of Human Resources 14(Winter, 1979), pp. 1–20.Google Scholar

5 A third problem is that of index numbers. The results depend on whether the differential is standardized by male or female average characteristics.

6 Leonard, Jonathan, “Anti-Discrimination or Reverse Discrimination: The Impact of Changing Demographics, Title VII and Affirmative Action on Productivity,” Journal of Human Resources, 19 (Spring 1984), pp. 145–74.CrossRefGoogle Scholar

7 Voos, Paula, “Wage Discrimination: A New Approach Based on the Direct Measurement of Productivity.” Paper presented at the American Economic Association Meetings, New York, NY, 1985.Google Scholar

8 Goldin, Claudia, “The Gender Gap,” p. 163.Google Scholar

9 Becker, Gary, The Economics of Discrimination (Chicago, 1971).CrossRefGoogle Scholar

10 As we show below, however, finding that equation 5 holds with equality does not rule out all forms of discrimination.

11 Spence, Michael A., “Job Marketing Signalling,” Quarterly Journal of Economics, 87 (08. 1973), PP. 355–74;CrossRefGoogle ScholarAigner, Dennis and Cain, Glen G., “Statistical Theories of Discrimination in Labor Markets,” Industrial and Labor Relations Review, 30 (01. 1977), pp. 175–87.CrossRefGoogle Scholar

12 Cain, Glen G., “The Challenge of Segmented Labor Market Theories to Orthodox Theory: A Survey,” Journal of Economic Literature, 14 (12. 1976), pp. 1215–57.Google Scholar

13 Goldin, Claudia, “Monitoring Costs and Occupational Segregation by Sex: A Historical Analysis, ” Journal of Labor Economics, 4 (07 1986), pp. 1–27.CrossRefGoogle Scholar

14 France, Statistique de la France 1835–73, 1st and 2nd series, vols. 10–14: Industrie; vol. 19: Enquête industrielle de la France, 1861–65 (Paris, 1913).Google Scholar

15 Although agriculture employed many more women in France than textiles, textiles and garment manufacture employed most of the female labor force in industry in the 1860s. Tilly, Louise A. and Scott, Joan W., Women, Work, and Family (New York, 1978), p. 69.Google Scholar

16 Although the accuracy of figures reported in any census is subject to question, evidence from worksheets available in departmental archives, such as at Colmar, indicates French census takers took pains to verify the figures reported whenever possible. In some instances survey forms had to be redone several times to pass their scrutiny.Google Scholar

17 The 1860–65 data contain 150 wool-producing arrondissements and 64 cotton-producing arrondissements. We dropped those arrondissements with no female workers from the data set (18 for wool, 1 for cotton). The descriptive statistics for employment and wages for the full data set are similar to our sample. A more detailed accounting of descriptive statistics for our sample is given in the Appendix, Table 6.Google Scholar

18 The F-statistics associated with the discrimination tests in the physical-output estimates are as follows: cotton–0.44, wool–2.47. The F-statistics for the discrimination tests in the value-added estimates are cotton–0.01, wool–2.03. None of these is significant at the 10 percent level. (See Appendix, Table 7).Google Scholar

19 A fourth possible modification would be to alter the functional form. Since these sample sizes are small, estimation of translog production functions is not feasible. We defer consideration of the translog form to a later section of the article, where we use a larger data set.Google Scholar

20 We estimated a constrained version of the production function, forcing the marginal product of women to equal that of men. We could not reject the equality-of-marginal-product hypothesis with the 1860–65 data.Google Scholar

21 The 1839–45 census may not have been as comprehensive as the 1860–65 census. There was an explicit concern to enumerate only those firms having 10 employees or more, though several of the observations are firms that are actually smaller than this. A total of 4.5 percent of the firms in the 1839–45 data set had no female workers, and 0.5 percent of the firms had no male workers. We dropped these firms from the analysis. Women comprise 39.3 percent of the adult employment in the full data set. They comprise 42.6 percent of adult employment in our sample.Google Scholar

22 The comparable figures for the entire data set, including firms with either no men or no women, are 47 percent for spinning and 36 percent for weaving.Google Scholar

23 For convenience, we will refer to the coefficients as “output elasticities” though they are actually the elasticity of value added with respect to inputs.Google Scholar

24 The 1839–45 data set contains information for children employed in addition to adult workers.We estimated Cobb-Douglas value-added function estimates with children included. This modification does not affect our basic finding of no employer discrimination against women. The F-statistics are not significant in all categories. The size and precision of the estimated output elasticities for male and female workers are fairly close to those in Table 9. For the first category, cotton spinning, the pattern of marginal products for men, women, and children matches the patterns of relative wages. The remaining categories exhibit some anomalies, however. For example, the results for cotton weaving indicate that children are underpaid relative to adults. And for wool weaving, their estimated output elasticity is negative. These results are available on request from the authors.Google Scholar

25 Note that the marginal product ratios are uniformly lower than the productivity ratio implied from Goldin's calculation of the differential in piece-rate earnings of 0.77 for United States manufacturing in 1895. This figure includes information from industries other than textiles (printing and cigar manufacturing). See “The Gender Gap.”Google Scholar

26 The F-statistics for the equality-of-marginal-product hypothesis are as follows: cotton spinning, 6.112; wool spinning, 19.108; cotton weaving, 4.458; wool weaving, 23.879; cotton and wool weaving, 0.344.Google Scholar

27 The advantage of the translog form is its flexibility (it imposes no restrictions on elasticities of substitution, for example). But the flexibility comes at a cost of weaker global properties (see, for example, Guilkey, David K., Lovell, C. A. Knox and Sickles, Robin “A Comparison of the Performance of Three Functional Forms,” International Economic Review, 24 [10. 1983] pp. 591–616). Further, inclusion of quadratic and interactive terms can create collinearity problems. The value-added function we estimate is given by: , where L c denotes the number of child workers.CrossRefGoogle Scholar

28 The no-discrimination constraint in the translog case is . The test statistics are as follows: cotton spinning, 4.842; wool spinning, 1.529; cotton weaving, 29.220; wool weaving, 6.486; cotton and wool weaving, 14.190. The translog estimates are available on request from the authors.

29 See, for example, Abbott, Edith, Women in Industry (New York, 1916), p. 305,Google Scholaror the discussions in Gullickson, Gay, Spinners and Weavers of Auffay (Cambridge, 1986),CrossRefGoogle Scholaror Tilly and Scott, Women, Work, and Family.Google Scholar

30 Abbott, Women in Industry, pp. 107–8.Google Scholar

31 “From Lancashire and Alsace came reports of insufficient work for men, abundant opportunity for women and children. The early factories used the women and children in unskilled jobs, often as assistants to male spinners.” Tilly and Scott, Women, Work, and Family, p. 76.Google Scholar

32 Gullickson, Spinners and Weavers.Google Scholar

33 Ibid., pp. 106–7.

34 Tilly and Scott, Women, Work, and Family, p. 76, quote Andrew Ure, “The effect of substituting the self-acting mule for the common mule is to discharge the greater part of the men spinners, and to retain adolescents and children. The proprietor of a factory near Stockport states that, by such substitution, he would save £50 a week in wages, in consequence of dispensing with nearly forty male spinners, at about 25s of wages each.”Google Scholar