Our ability to study how gender structures family life improved vastly when, a few decades ago, surveys began to ask about time spent in household work. Using such data, the sociologist Julie Brines (1994) published an innovative, rigorous article on housework that has come to be known for one main idea—when men are unemployed or earn less than their wives, they respond by reducing their housework. If money talks in relationships (and is a reason women generally have less power), then we would have expected men to up their domestic contribution when they earn little or nothing. The finding that they do the opposite was interpreted to mean that, when husbands are unable to display socially approved versions of masculinity in one domain (having a job and outearning their wives), they seek another way to display it—through not doing much housework. The idea was not merely that most men eschew housework to look like “real” men, though that may be true, too. Rather, the focus was on which men felt the most pressure to display gender by avoiding housework—those who are gender deviant in not outearning their wives. About a decade later, other scholars added findings purporting to show an effect for women, too; they claimed that when women outearned their husbands, the wives increased their housework (Bittman, England, Sayer, Folbre, & Matheson, 2003; Evertsson & Nermo, 2004). Here the interpretation was that women responded to the gender deviance of outearning their husbands by doing more housework to show that they were, nonetheless, “real” women or to relieve their husbands from the additional threat to masculinity of having to do more housework. This basic idea that gender deviance in who earns more is responded to by redoubled traditionalism in who does the housework spread like wildfire among scholars, and Brines (1994) became a canonical citation. The perspective was variously called compensatory gender display and gender-deviance neutralization.

In this issue, Oriel Sullivan (2011) makes a convincing case that the received wisdom on this matter is largely wrong—or at best, the effects supporting gender-deviance neutralization are so small, or pertain to such a small subgroup, that they are unworthy of the amount of attention they have received. Her incisive review of quantitative and qualitative studies shows that the gender-deviance neutralization effect for men was, where present, largely limited to small subset of men earning less than their wives—men unemployed for a long time or those with the very lowest relative earnings. Moreover, she points out, the analogous finding for women—that is, in the range in which women earn more than their husbands, additional earnings increase their housework—was a mistaken inference resulting from incorrectly specified equations. Specifically, as Gupta (2007) has shown for the United States, it is women's low absolute earnings, not their earnings relative to their husbands, that are associated with more housework after appropriate controls.1

My main reaction to Sullivan's review in this issue is an emphatic “Amen!” As a participant in the research community studying gender and household work over the past 25 years, I want to reflect on how and why we collectively got things so wrong on this topic and on what we should have focused on instead. This is self-critique as much as a critique of others, as I contributed to what I now regard as the undue attention to the phenomenon of compensatory gender display. My diagnosis and prognosis below are based on reading research articles over the years and on how I have heard scholars talk about this area in colloquia, on conference panels, and in other informal gatherings. Most of my diagnoses of the causes of the problems are not at all unique to this area of research but occur in many areas of study.

First, articles inevitably get reduced to sound bites in academic collective memory. Sometimes this entails little distortion of the central idea, but, here, with a complicated curvilinear relationship, it led to Brines's (1994) paper being misinterpreted from the beginning. She showed that, in the range in which women earn more than men, each incremental addition of the proportion of income earned by the woman reduced men's housework, the opposite of what dependency or exchange and bargaining theory says. But people forgot that this generalization applied only to the minority of couples in which women earned more than men. Across the range of relative earnings from women earning 0–50% of the couple's money, which is where the vast majority of couples fell, Brines's findings fit the standard view of “money talks,” variously known as the dependency, relative resources, exchange, or bargaining view; the more you earn relative to your partner, the less housework you do and the more your partner does. The research experts understood that this curvilinear relationship was what Brines showed, but as it moved into literature reviews, textbooks, and informal discussions, all was often lost but the idea that men do less housework as their wives earn more than they do. Brines's article did not show that most women are unable to translate earnings into bargaining power in marriage, but given the way the findings were often summarized, you would think that she had.

Second, as in many areas of research, we lost sight of what we teach our students—statistical significance does not mean substantive significance. The effect for men found in the original article was tiny, as best as one can tell from the graphs (Brines, 1994, Figures 7–8), and was small or nonexistent in later studies (Bittman et al., 2003). This was not easily discernible from the regression tables, as the key relationship featured a nonlinear relationship captured with relative earnings and its square; the magnitude of decrease across any range could not merely be read off coefficients but required computing predicted values from the regression.

Third, most of this literature gave us limited purchase on causality because studies seldom used panel data with statistical techniques, such as fixed effects, to remove omitted-variable bias. Fortunately, recent articles by Gershuny, Bittman, and Brice (2005); Killewald and Gough (in press); and Baxter and Hewitt (2009) have corrected this, using longitudinal data.

Fourth, I suspect that many of us were willing to ignore the generally small magnitude of the effects because the interpretation regarding gender-deviance neutralization was new and theoretically interesting.2 It resonated with the emerging view among gender scholars that the gender system is a powerful force at micro, interactional, and macro levels. The idea suggested that the gendering of household labor not just was a linear reflection of earnings in the market (as important as that would be) but also affected the very way money could be converted to interpersonal power. It resonated with the emerging emphasis on intersectionality—men at different class locations seemed to respond differently to who brought home more of the bacon. These resonant ideas provided the obligatory theoretical hook for research articles. As a former journal editor (American Sociological Review from 1994 to 1996), I believe that our collective research enterprise is hurt by our rigid adherence to the standard that, to be published, every empirical article should test theory and lead to new theoretical conclusions. We would better grasp the big picture if reviewers and editors selected rigorous empirical research identifying sizable effects on interesting questions, if empirical articles did not stretch for theoretical implications beyond what the analysis reveals, and if theoretical conclusions were derived more from reviews of accumulated studies, with most attention given to studies featuring the best designs for discerning causality and finding sizable effects. At least in sociology, we authors have an incentive to reach for juicy theoretical implications of even tiny effects to “sell” our articles to journals, and it is not surprising that we respond to those incentives.

Although we as a field were making much ado about almost nothing, some of us also gave too little attention to large patterns and trends. Over time, the really big picture is the long-term increase in women's employment, the sizable reduction in women's housework time, and the lesser increase in men's housework time, which scholars such as Bianchi, Robinson, and Milkie (2006) have chronicled. But, unfortunately, another canonical citation in the literature on gender and household work, Hochschild's (1989) The Second Shift, fostered a basic misconception about the result of these trends. Hochschild claimed that the lack of adjustment of women's and men's housework as women entered employment led to a situation in which women do two shifts of work (paid and unpaid), whereas men do one (paid). Near the beginning of her book, Hochschild (1989) stated that she averaged estimates of paid work plus household work from major studies of the 1960s and 1970s and concluded that women do about 15 hours a week more total work (paid plus unpaid) than men, leaving them that much less time for sleep or leisure. Although her book was excellent in its qualitative analysis of how couples talk about and negotiate housework, the quantitative claim seriously misled the field. She must have simply made a mistake, because neither the time-use studies she cited nor more recent work has supported anything remotely close to a 15-hour gender gap in total work.3 Recent U.S. data show that, on average, husbands average slightly more (3 hours per week) total (paid plus unpaid) work, although in some types of families, such as those in which wives are employed full-time and have a preschool child, women average about 5 hours more work per week (Sayer, England, Bittman, & Bianchi, 2009). Most time-use experts have long known that Hochschild's (1989) claim was incorrect, but the book probably kept young scholars from seeing that the real gender issue is not inequality in the total work time men versus women are burdened with but, instead, what determines and changes how spouses divide paid and unpaid work. And given that women still do less paid work, and thus make less money,4 the issue that animated Brines's (1994) initial inquiry—whether economic dependence reduces individuals' power to get their way in a relationship—is still relevant, despite the reduction in the magnitude of women's economic dependence. We need to revisit the question of how much money talks in relationships, using measures of getting what one wants other than freedom from housework.5

Another large pattern ignored until recently in scholarship on gender and household work is how time spent in child care varies by gender, by socioeconomic status, and over time. Most of the studies reviewed above excluded child care from their measures of household work. Sullivan (2010) in a recent article chronicled an especially large increase in the time that highly educated parents are putting into child care, thus reflecting a new norm of intensive parenting and creating large class differences in time spent in parenting. One recent analysis showed that college-educated married mothers spend 5–6 more hours per week in child care than those who finished less than high school; the analogous educational differential for married fathers is 3–4 hours per week (England & Saraff, 2010). These differences are striking because the well educated have fewer children and average more time in market work (because more of them are employed). The extra time they nonetheless put into child care appears to come from leisure, sleep, and housework. The causes and consequences of these class differences and the large remaining gender differences in child-care time are an important area for future research.

Despite my critical comments about allocation of energy in the research community I have been a part of for decades, I do not want to end too negatively. Fresh voices like that of Oriel Sullivan are providing correctives. I hope that some of my diagnoses about the sources of our collective misallocations of energy will be heeded.


  1. Top of page
  2. References
  • Baxter, J., & Hewitt, B. (2009, July 1617). Economic independence or bargaining power: The relationship between women's earnings and housework time. Paper presented at the annual Household, Income, and Labour Dynamics in Australia Survey Research Conference, Melbourne, Australia.
  • Bianchi, S. M., Robinson, J. P., & Milkie, M. A. (2006). Changing rhythms of American family life. New York: Russell Sage Foundation.
  • Bittman, M., England, P., Sayer, L., Folbre, N., & Matheson, G. (2003). When does gender trump money? Bargaining and time in household work. American Journal of Sociology, 109, 186214.
  • Brines, J. (1994). Economic dependency, gender, and the division of labor at home. American Journal of Sociology, 100, 652688.
  • England, P., & Saraff, A. (2010). Educational differences in parents' time spent in child care: Culture, incentives, or income constraints? Paper presented at annual meeting of the Population Association of America, Dallas, TX.
  • Evertsson, M., & Nermo, M. (2004). Dependence within families and the division of labor: Comparing Sweden and the United States. Journal of Marriage and Family, 66, 12721286.
  • Gershuny, J., Bittman, M., & Brice, J. (2005). Exit, voice, and suffering: Do couples adapt to changing employment patterns? Journal of Marriage and Family, 67, 656665.
  • Gupta, S. (2007). Autonomy, dependence, or display? The relationship between married women's earnings and housework. Journal of Marriage and Family, 69, 399417.
  • Hochschild, A. (1989). The second shift. New York: Avon.
  • Killewald, A., & Gough, M. (in press). A longitudinal evaluation of gender display in spouses' housework hours. Social Science Research.
  • Sayer, L., England P., Bittman, M., & Bianchi, S. M. (2009). How long is the second (plus first) shift? Gender differences in paid, unpaid, and total work time in Australia and the United States. Journal of Comparative Family Studies, 40, 523544.
  • Sullivan, O. (2010). Changing differences by educational attainment in fathers' domestic labour and child care. Sociology, 44, 716733.
  • Sullivan, O. (2011). An end to gender display through the performance of housework? A review and reassessment of the quantitative literature using insights from the qualitative literature. Journal of Family Theory and Review, 3, 113.
  • 1

    Killewald and Gough (in press) presented additional evidence that women's absolute earnings are more important than relative earnings for U.S. women's housework. Use of longitudinal data and a fixed-effects statistical model enhanced their purchase on causality. However, Baxter and Hewitt (2009), also using longitudinal data, found that women's relative earnings are more important in predicting their housework in Australia, which reinforces what Bittman et al. (2003) found in a cross-sectional analysis of another data set. Baxter and Hewitt (2009) found nonlinearity in the effect of women's relative earnings on their housework, such that as the proportion of his and her combined earnings that she contribute goes up, women's housework goes down, but where women contribute about 75% of joint earnings, the effect flattens and turns slightly positive. The positive upturn—the part supportive of gender-deviance neutralization—is quite trivial, at about 1 hour a week.

  • 2

    I can say only in my own defense, and that of my coauthors (Bittman et al., 2003), that the effect we identified for Australian women was not trivial in magnitude. Net of market hours worked, housework went down about 5 hours per week, as wives' contributions to couple's joint earnings went from 0% to 50%, and up about 5 hours per week, as wives' contributions went from equality (50%) to 100%. The finding was robust to removing the top 3% of outliers on her relative earnings. However, as reviewed here, effects for men have usually been trivial.

  • 3

    Hochschild (1989) does not mention which studies she averaged, but I have searched all the time-use studies cited in the book and cannot find evidence for the 15-hours-per-week figure in either their tables or text.

  • 4

    Of course, any extant gender differences in hourly wages will also produce earnings differences, even if women's hours of paid work converge with men's.

  • 5

    This is, of course, easier said than done, because with housework, it is a reasonable assumption that most people would prefer to do less of it, whereas with other outcomes (e.g., what money is spent on, what activities spouses do together, who spends more time with children), it is harder to know what individuals want without asking them, so it is difficult to discern who is getting more of what they want.