Gender Division

Cohen / GENDER DIVISION GENDER & SOCIETY / April 2004 10.1177/0891243203262037 OF LABOR

THE GENDER DIVISION OF LABOR “Keeping House” and Occupational Segregation in the United States
PHILIP N. COHEN University of California, Irvine
This article explores the effect of women’s movement into the labor market on the gender segregation of work, using the Current Population Survey from 1972 to 1993. The author includes as working those respondents who were “keeping house” and codes keeping house as an occupation. The results show higher estimates of gender segregation, and slightly steeper declines over time, than were seen in previous studies. Analysis of one-year longitudinal changes reveals less movement out of female-dominated occupations when keeping house is included as an occupation. Finally, a decomposition of the segregation trend shows that the movement of women away from keeping house contributed as much to the overall decline in gender segregation as did the desegregation of paid occupations. The author concludes that the movement of women’s work from the household to the labor market has been a driving force in the changing nature of gender inequality.

Keywords:

occupational segregation; housework; gender inequality

Gender segregation in the labor market is high, fueled by gendered and discrimi-

natory practices and assumptions (Baunach 2002; Nelson and Bridges 1999; Reskin 1993). But for reproducing an institutionalized gender division of labor, and devaluing women’s work, the labor market is still no match for the “gender factory” of the married-couple family (Berk 1985). This article explores the effect of women’s labor moving into the paid market on the overall gender segregation of work and therefore on the changing nature of gender inequality. The gender division of labor is a central feature of gender inequality, both in its economic aspects and in the social construction of gender identities (Huber 1991; Lorber 1994). As Chafetz wrote, “undergirding all systems of gender stratification is a gender-based division of labor, by which women are chiefly responsible for different tasks than are men” (1991, 77). However, the empirical literature on the gender division of labor is uncomfortably divided between those who examine the

AUTHOR’S NOTE: I would like to thank Christine Bose, Nancy Folbre, Matt Huffman, and the Gender & Society reviewers for their comments and suggestions and Liana Sayer for her contributions. REPRINT REQUESTS: Philip N. Cohen, Department of Sociology, University of California, Irvine, CA 92697-5100.
GENDER & SOCIETY, Vol. 18 No. 2, April 2004 239-252 DOI: 10.1177/0891243203262037 © 2004 Sociologists for Women in Society

239

240

GENDER & SOCIETY / April 2004

division of household labor and those who study gender segregation in the paid labor market. The overall gender division of labor has not been considered in empirical studies of the United States (Miller and Garrison 1982). The problem with separate studies of housework versus occupational segregation is that they cannot show the dynamic relation between the two. In this article, I bring together the division of household and labor market work in one, partial attempt to fashion a unified measure of the division of labor for one period of recent history. The results underscore the importance of the movement of women’s labor from the household to the labor market in reducing gender inequality. HOUSEWORK AND OCCUPATIONS Research consistently has shown that women do the lion’s share of unpaid labor within households (Coltrane 2000). Although this inequality has decreased in recent decades, the household division of labor remains highly gendered (Bianchi et al. 2000). At the same time, gender segregation in the labor market remains high, although after a half century of apparent stability, there were declines in the 1970s and 1980s (Blau, Simpson, and Anderson 1998; Cotter et al. 1995; Reskin 1993; Wells 1999). Of course, change in these two arenas is linked, but that connection is rarely the subject of direct examination. The entry of greater numbers of women into the labor force occurred as household services, products, and technology reduced women’s housework obligations and increased the demand for female labor in the market (Cohen 1998; Cohen and Bianchi 1999; Cotter, Hermsen, and Vanneman 2001; Cowan 1983; Presser 1999; Strasser 1982; Uttal 2002). The movement of care work from within the family to the market represents a fundamental shift (Folbre and Nelson 2000). We have not reached the point at which we might “eliminate the home as a place of work and housewives as a functional group of the population” (Durand 1946, 222), but we have unquestionably moved in that direction (Stacey 1993). With women more likely to be employed, the segregation of paid work has increased in importance as a component of gender inequality. Chang (2000, 1658) argued that “the long-standing presumption has been that occupations are the backbone of the class stratification system, but as women enter into the formal economy in ever-increasing numbers, the occupational structure becomes the main locus of gender stratification as well.” This echoes an earlier body of research on the shift from home to market, which stressed the continuity of gender segregation in the new context of the labor market:
The sexual division of labor reappears in the labor market, where women work at women’s jobs, often the very jobs they used to do only at home. . . . As these jobs are low-status and low-paying, patriarchal relations remain intact, though their material

Cohen / GENDER DIVISION OF LABOR

241

base shifts somewhat from the family to the wage differential, from family-based to industrially-based patriarchy. (Hartmann 1981, 25)

There clearly is a connection between the work women do at home and the occupations that are female dominated in the labor market. However, it is misleading to collapse the two entirely because there are very few paid occupations that are as female dominated as “women’s work” in the home. Many people who study the segregation of paid occupations are concerned theoretically with the overall gender division of labor. For example, in his international comparison, Jacobs (1989) reviewed data on the gender division of labor, including paid and unpaid work. But in the data analysis, he examined paid work only. Because women tend to move between occupations that are more or less female dominated during the course of their careers, through “revolving doors,” he suggested that gender segregation is continually reproduced through processes of social control that define male- and female-dominated work, rather than women’s “taste” for certain jobs or their human capital assets. The question he examined— how gender segregation is reproduced—is important for all kinds of work, but the labor force data he used restrict his analysis to paid employment. Similarly, in Valerie Oppenheimer’s (1970) landmark study, she acknowledged that “census statistics best reflect . . . paid employment outside the home, rather than the trends in all kinds of productive work carried out by women.” But she added, “This limitation should not, however, be a serious drawback in the analysis of the labor force of an industrial society” (p. 10). From the perspective of these labor force studies, then, a defining characteristic of industrial society—of modernity, in fact—is that the paid labor market replaces the home as the central site of gender inequality. Thus, many researchers concerned with gender inequality have moved to focus primarily on occupational inequality, even as most feminists stress the continuity in the division of labor between home and market (Cohen and Huffman 2003; Cotter et al. 1997). But the transition from unpaid labor at home to paid labor in the market is itself a source of change in the gender division of labor. Direct comparisons are difficult to find, but consider the examples of cooking and cleaning. In 1995, women did 74 percent of all unpaid cooking at home, but in the market, only 45 percent of all cooks were women (this category excludes those working in private households, a tiny fraction of the total). Similarly, women did 80 percent of unpaid housecleaning at home, but only 35 percent of janitors and cleaners were women.1 Insofar as the division of labor is a cornerstone of gender inequality, then, women leaving home and going to work may itself reduce gender inequality. In fact, the market’s ability to pull women from the household (Cotter, Hermsen, and Vanneman 2001) has been a leading factor in the partial redivision of housework in the past few decades. This may be seen in the many studies that show less gender inequality in couples’ housework when women are employed (Batalova and Cohen 2002; Bianchi et al. 2000; Coltrane 2000).

242

GENDER & SOCIETY / April 2004

Existing studies of trends in occupational segregation implicitly treat women who enter the labor market as if they are just beginning to work: The level of gender segregation is assessed only in their paid work capacity.2 In this analysis, I treat women who leave home for the labor market as if they were exchanging one job for another. This is possible because the Current Population Survey (CPS) until 1993 identified respondents who were “keeping house.” Although this method raises definite problems, as discussed below, it also offers a unique opportunity to see the role of women’s increasing labor force participation in the overall gender division of labor. KEEPING HOUSE: COUNTING HOUSEWORK AS WORK Feminists argue that women’s work is devalued by the failure of mainstream economics to account for unpaid work, both internationally (Waring 1999) and in the United States. As Folbre and Abel (1989, 547) reported, in the nineteenth century, “the census institutionalized a definition of ‘work’ as ‘market work’ that literally devalued women’s unpaid work”—and it still does. Women’s participation in the paid labor force did not reach 50 percent until the late 1970s, even among Black women, whose participation rate was historically higher than white women’s (Goldin 1990, 17). Thus, neither national economic accounts nor labor force statistics take into account the work that for much of American history was the focus of most women. Similarly, contemporary analyses of occupational segregation do not include the disproportionate share of unpaid labor women perform. Reskin and Hartmann (1986, 7) acknowledged this problem in their study of sex segregation, noting as an aside that “the occupation of most women not in the labor force, homemaker, is one of the most segregated occupations.” That selectivity is appropriate for studying some labor market dynamics, but it precludes us from evaluating the overall division of labor. The U.S. Census Bureau introduced the term “keeping house” in 1870. In the instructions to assistant marshals for the 1870 census, under “occupation,” the Bureau wrote the following:
The term “housekeeper” will be reserved for such persons as receive distinct wages or salary for the service. Women keeping house for their own families or for themselves, without any other gainful occupation, will be entered as “keeping house.” Grown daughters assisting them will be reported without occupation.3

Those women coded as “keeping house” were not included in tabulations of the labor force. British census takers, on the other hand, counted keeping house as an occupation from 1851 to 1881, as did Massachusetts from 1875 to 1905 (Folbre and Abel 1989). In 1910, the U.S. Census Bureau instructed enumerators to count women doing unpaid labor on family farms as “farm laborers,” resulting in an

Cohen / GENDER DIVISION OF LABOR

243

upward spike in the historical trend for women’s labor force participation. They subsequently produced estimates to eliminate this anomaly, and the offending instruction to enumerators was not repeated (Oppenheimer 1970, 2-6). Durand (1946, 221) defended the exclusion of farm wives from the labor force, as “strictly accurate figures would greatly overstate the relative degree of participation of farm women in the labor force because a large percentage of them do very little gainful work,” but he provided no evidence to support this conclusion. Some researchers of national income in the nineteenth and early twentieth centuries attempted to estimate the value of this work, before interest apparently waned. Folbre and Wagman (1993, 279) concluded that “what varied across states and over time was women’s participation in market work, probably not their productive work in general.” If that is the case, we should recognize that the exclusion of houseworkers from existing occupational segregation studies is an artifact of patriarchal assumptions in the conception and collection of labor force data. With the work of Waring (1988, 1999), analysts once again started to factor women’s unpaid work into national economic estimates, but this newfound interest still has not reached occupational segregation studies. METHOD The analysis comprises three sections. First, I recode labor force data for respondents in the CPS to include those respondents who were recorded as keeping house, and I code keeping house as an occupation. Then I calculate gender segregation levels with and without the keeping house occupation across the period from 1972 to 1993 (when the question was dropped). Second, using the one-year longitudinal property of the CPS, I broaden Jacobs’s (1989) “revolving door”—which reflects the movement of women into and out of female-dominated work—to include the housework occupation. This will enable us to see the gender composition of occupations that women enter when they shift between keeping house and paid work. Finally, I decompose the trend to show the relative contributions to desegregation of women entering paid work versus paid occupations becoming less segregated. The data are from the CPS Annual Demographic Files (March). The CPS is a large, monthly, nationally representative survey conducted by the Census Bureau to measure attributes of the labor force. During this period, the sample for the March survey consisted of approximately 50,000 households per year. My sample includes noninstitutionalized civilian adults ages 25 to 54, the ages most commonly used in labor force studies (except in the longitudinal analysis, where I include women ages 18 to 64 to increase the sample size). Although these samples are large, I pool several years of data at each point to decrease fluctuations due to random variation. I use the person weights provided by the CPS. Until 1993, the CPS asked the “major activity” question of each household member: “What was X doing most of last week?” The categories offered were “working,” “looking for work,” “keeping house,” “going to school,” “unable to

244

GENDER & SOCIETY / April 2004

work,” and “retired.” I code all those who answered “keeping house” as working and include “keeping house” in the new analysis of occupational segregation. Slightly more than 10 percent of people listed as keeping house in the CPS were also listed as “in the labor force” on the CPS employment status variable. Presumably, these people were keeping house as their major activity but also gave information that led them to be coded as in the labor force. I include all those who chose “keeping house” in this category, regardless of their employment status, as this represents their self-described major activity. To measure the division of labor, I estimate occupational segregation using the index of dissimilarity (Blau, Simpson, and Anderson 1998). The number reflects the percentage of either men or women who would have to change occupations to achieve an equal distribution of men and women across occupations. Jacobs (1989) and Baunach (2002) made a persuasive case for using more elaborate measures of segregation (indexes of concentration and isolation). However, my objective is less to arrive at the most precise measure of segregation than it is to show the difference once houseworkers are included, which is unlikely to be affected by the choice of index. Because the CPS has smaller samples than the decennial censuses used by others, and small occupations are more likely to be segregated by chance (Cotter et al. 1997; Wells 1999), I limit my analysis to the largest 100 paid occupations at each time point. However, the results were nearly identical when I used 200 occupations instead. To analyze how the movement of women from housework into paid occupations, and vice versa, affects the gender division of labor, I take advantage of the one-year longitudinal property of the CPS. Each household in the CPS is interviewed for four consecutive months, then misses eight months, and finally is interviewed again for four more months (U.S. Census Bureau 2000). As a result, households interviewed in March are supposed to be interviewed in the following March’s survey. The matching across the two years is not perfect, however. For example, household composition may change, and people may move or die between March surveys; there are also nonresponse and recording errors. Furthermore, it is impossible to know for sure that matches are perfect (for example, if a person is divorced and then marries another spouse of the same age in the intervening year). However, using fairly restrictive criteria, it is possible to achieve most of the eligible matches with a high degree of confidence (Madrian and Lefgren 1999). I conduct the longitudinal analysis using the 1991 to 1993 March CPS, combining three years of transitions to increase sample size. Individuals are matched by household number and individual line number and included if they are the same sex and race/ethnicity and have aged between