Title: Cute butts and housework: A gynocentric theory of assortative mating
Abstract: In the 1992 screenplay for Sleepless in Seattle Tom Hanks's character, Sam, a recently widowed single father, is facing the dating pool after 14 years. “I want to know what it's like over there,” he asks his friend Jay. The answer comes back, “… Pecs and a cute butt. … You can't even turn on the news without hearing about how some babe thought some guy's butt was cute. … The good news is, split the check” (Arch, Ephron, & Ephron, 1992). The message is clear: It is a whole new world out there in the 1990s because women's requirements for potential mates are very different from what they used to be. On one hand, men have to measure up to compete. On the other hand, there are economic benefits for men now that women are working. Megan Sweeney and Maria Cancian's (2004) new approach to understanding the marriage market focuses solely on “splitting the check” and does not theorize women's evaluation of men—that is, women's new demand for cute butts and sharing the household labor. Put another way, the authors' economistic and androcentric approach neglects crucial sociological insights about changing gender and family relations—as well as rising social inequality—in the post-1965 period. Sweeney and Cancian's (2004) interesting and provocative article, “The Changing Importance of White Women's Economic Prospects for Assortative Mating,” asks a very important question for the understanding of family formation after the massive social change of women's increased labor force participation in the 20th century: How does women's new employed status alter their value to potential mates? Their careful, erudite analysis adds to the literature by starting a dialogue—using methods that have not been used before—about how best to specify a model of assortative mating in an era of wage-earning women. The article investigates the effect of women's increased economic activity on their value to men as potential wives. Using data from the 1970 U.S. Census, the National Longitudinal Surveys of Young Women, and the National Longitudinal Surveys of Youth, they set out to test the hypothesis that “women's earning capacity has become a more important factor over time” (p. 1017). The authors test this hypothesis by estimating husbands' expected earnings and occupational status as a function of wives' premarital wage rate for two cohorts: those marrying predominantly in the 1970s and those marrying in the late 1980s and 1990s. Unlike previous studies, their data are longitudinal and individual level, allowing them to look at assortative mating without the complexities of divorce, for example, or the endogeneity of spouses' earnings during marriage. On one hand, their model assumes that the “quality” of a husband is his economic value only, and that husbands' earnings are stable over time. A wife's “quality,” on the other hand, includes her economic and home production value, the latter of which, it turns out, cannot be measured. Results show that, as expected, women's premarriage wage is positively and significantly associated with men's expected earnings and occupational status in the late cohort when compared with the earlier cohort. Sweeney and Cancian (2004) are correct that the raw correlations they present in Table 2 of the article are insufficient to demonstrate their thesis that men's weights on women's economic value have changed. Many factors could be affecting the wages of men and women, and the raw correlation could be masking the movement of these factors within a single bundled measure. A simple example of this phenomenon is the role of education as a type of capital for women. Suppose that men have always placed the same high value on an educated wife as they do now, but historically there were fewer educated women. Over time, women have increasingly gained access to education, so now, more men can achieve their desire for a wife who is as educated as they are. At the same time, the earnings returns to education have increased. The correlation between husbands' and wives' earnings would increase over time, but the weight that men place on women's earnings could have remained constant. Therefore, the correlation is not a sufficient statistic to reveal what has happened to men's weight on women's earnings. The authors correctly use regression analysis to parse out the effect of wives' premarital wage. We may begin to assess the article's main regression result by assuming that the estimated regression coefficients in Table 3 are correct. What are the possible interpretations of the finding? The first row in the table shows the regression coefficient for wife's premarriage wage. Results reveal a coefficient of 0.025 for the late cohort in the model that predicts earnings, a large increase from 0.009 in the earlier cohort. These are the α'1 coefficients from Equation 4. They suggest that the large increase in the regression coefficient over time implies that men value women's earnings capability more than they used to. This hypothesis makes intuitive sense; before women had their current level of economic opportunity, there was much less variation in women's economic potential. Unless poised to inherit wealth, wives were to be judged on the basis of their family, social class, race, religion, education, beauty, wit, obedience, cooking, and so forth. That is, they were judged on physical, social, cultural, and psychological factors more than economic ones. But I suggest that the same regression results shown in Table 3 could have an alternative explanation. Women's increased labor market participation and improved economic independence have led them to place less weight on men's economic value when choosing a husband. I will call this interpretation the gynocentric model. Viewed as a mathematical result, Sweeney and Cancian's (2004) increased coefficient is consistent both with men placing more importance on women's economic value and/or women placing less importance on men's economic value (a copy of the statistical proof is available from the author). Within the framework of mate matching, Sweeney and Cancian's estimation procedure finds changes in the relative weight that men and women place on each other's economic value, but they do not interpret the change as a relative one; they see it as an absolute shift upward in the weight that men give to women's economic value as future wives. How do we know which interpretation of the result is correct—their androcentric model or the alternative gynocentric model? We cannot distinguish between the two hypotheses using the information given for two reasons. First, the regression coefficients on wives' premarital wage actually combine the underlying coefficients of interest (men's view of women and women's view of men) as discussed above. Second, to reach their conclusion, Sweeney and Cancian (2004) would need to show both an increase in the regression coefficient from the early cohort model to the late cohort model, and an increase in the explained variation (as measured by the R2) from the early to the late model. However, we cannot use the change in R2 as a measure because the estimates of men's earnings for the later cohort are faulty. To explain this second point in more detail, if Sweeney and Cancian's (2004) interpretation of the results is correct, the amount of variation explained should increase in the later cohort's model. If men are placing more weight on women's premarital wages, these wages should explain more variation in husbands' future earnings. Alternatively, if women now place less weight on men's future earnings, women's premarital wages should explain less of the variation in husbands' future earnings. In Sweeney and Cancian's (2004) Table 3, we see that indeed the percent of variation explained does go up over time, which would be consistent with the authors' interpretation of the results if their data for husbands' earnings were accurate. Their estimates of husbands' expected earnings and occupational status are problematic, however. The estimate for the later cohort is constructed using census data from the 1960s, together with the assumption that men's earnings have been constant over time. Therefore, the explained variation for this model is also related to men's earnings from the late 1960s and is not related to men's earnings from the new marriage market and the social context of the late 1980s and early 1990s. The method for computing men's earnings is consistent with the authors' assumption, but their assumption is incorrect: Men's wages have not been stable over time, even though it appears in Table 1 as if they have been stable. Sweeney and Cancian's (2004) Table 1 indicates that, in the present data, husbands' future earnings have virtually no deviation from a mean of $10 across time. But in reality, between 1963 and 1989, the average weekly wage of working men increased by 20%. Further, this increase masks increasing wage inequality. Wages for the least skilled fell by about 5%, whereas wages for the most skilled increased by about 40% (Juhn, Murphy, & Pierce, 1993; also see Bluestone & Harrison, 1988, and Levy, 1989). This enormous increase in wage inequality in White male wages is not captured by the authors' estimates of future earnings. If men's wages are more dispersed and some men are worse off in real earnings while others are better off, the actual increased dispersion of male wages does not meet their maintained assumption that men's earnings have been stable over time. Thus, although the improvement in explained variation (R2) seems to support their androcentric hypothesis, we really cannot evaluate it because there are no data for men's actual wages in the later cohort. We therefore cannot know whether men altered their evaluation of women, or whether the converse is true. Moreover, that men's wages have become dispersed since the 1970s is a finding about increased social inequality among White men, another important factor that is undertheorized in Sweeney and Cancian's (2004) model. It is likely that both explanations of the findings—the authors' androcentric model and the gynocentric model—are true, but it also could be that the authors' hypothesis better explains lower class mating, and the gynocentric model better explains higher class mating. Statistically, this would mean that the same regression coefficient may have a very different interpretation depending on social class. For working-class men, the coefficient would indicate more relative weight on women's economic value. For the higher class sample, the coefficient would indicate less relative weight on men's economic value from women's point of view. Why would social class have such an effect on assortative mating? Juhn et al. (1993) show increasing social inequality among White men. As lower skilled men experience a diminishing real value of wages, highly skilled men experience increasing real wages. Therefore, low-skilled men are now economically less independent than they used to be. They must focus on finding a wife who can help support their family and maintain their standard of living because of the loss of good jobs that allow men to earn a “family wage.”Sweeney and Cancian's (2004) findings make the most sense for this class, even though they do not have data on male earnings after 1970. Women have become much more economically independent, however. The value of education has increased, and more women are gaining an education. But there is also growing inequality among women. Well-educated women can better afford to judge potential mates based on noneconomic characteristics. If this is true, we would expect the percent of explained variation from the earlier to the later cohort to fall rather than increase. So, whereas in the past, the economistic model would have explained much because of women's limited options, now we would expect women's choice of a husband to play a more important role in assortative mating. We could tell many stories to understand what has happened to assortative mating in the last generation. Women have gained access to education, men's earnings have bifurcated, and the returns to education have increased for both men and women. Yet, the big story of the last 50 years is women's growing independence and economic opportunity, not men's changing views of women as wives. Using data from a similar period, Arlie Hochschild (1989) famously called this the “Stalled Revolution.” Women are changing, whereas men are staying the same or only changing a little. On one hand, as women's labor force participation increased, their gender ideology has tended to move from the traditional male breadwinner/wife homemaker model toward an egalitarian vision of marriage and family. On the other hand, men tended either to remain traditional or assume an emergent transitional gender ideology. Transitional men either do not mind if their wives work, or actively encourage it because they like the extra income. Either way, they expect wives to maintain their old-fashioned duties at home. There are men and women who exhibit all three of these ideal typical gender ideologies. The problem in modern marriage, according to Hochschild, is that there are many egalitarian women and not enough egalitarian men for them to marry. If an egalitarian woman hopes to maintain a job and to have children, she must determine the likely household production of her potential mate. Unfortunately, Sweeney and Cancian (2004) take an economistic point of view of this process: Couples do not talk about their future lives; they only evaluate each other (i.e., only men evaluate women) on the basis of market signals. Their model does not capture how people sort on work-family balance and compatibility issues. The assumption is that people are homogeneous in how they value household production such as housework and child care. Just as we have good reason to expect different findings based on social class, we also have good reason to expect different findings based on gender ideology. Sweeney and Cancian (2004) argue that they do not need to worry about this issue because men's household labor contributions have not changed much over time, and because men still do not do much household labor. But change over time is not relevant here. Men's family work does matter in the later cohort, and there is variation in men's housework. Many men do contribute to household labor, and the housework literature is full of models that explain which types of men do more or less of it. Sweeney and Cancian's model does not allow us to separate the effects of wage value and household production value. This separation is crucial because for women who hope to share family work, finding men to provide it is an important part of their search for a mate. It is women's perception that they want and can find a “good man” that we need to account for in the model, not necessarily the reality of how many are available at the point of partner choice. If we had data on gender ideology for this analysis, we could measure the relative ideologies for different dyads. On one hand, the most common would likely be an egalitarian wife and a transitional husband. This dyad would incorporate both the androcentric and gynocentric hypotheses: both men's increased weight on women's economic value and women's decreased weight on husbands' economic value relative to their household production. A traditional woman who wants to be supported by her husband, on the other hand, should seek a man according to his earnings potential. This pairing supports Sweeney and Cancian's (2004) formulation. Traditional women may be missing from the analysis because the authors excluded every woman who lacked wage data at the premarital point. Women who were not working as they searched for a breadwinner husband could fit into this group. Their omission hurts the authors' argument, which would be bolstered by the inclusion of “wannabe” homemakers. Alternatively, if we had a traditional man with an egalitarian or transitional wife, only the gynocentric hypothesis would hold. All of these effects of gender ideology would be further complicated by the simultaneous analysis of social class or of other noneconomic factors not yet considered, such as race, fertility, health, culture, and so forth. The fundamental economistic and androcentric formulation of their model sheds doubt on the finding that women's position in the labor market has become a more important determinant of her position in the marriage market. Researchers should put women, who have changed the most during this period, at the center of the analysis. Do women want cute butts and housework? The answer awaits future analyses. Thanks to Dr. Simon Potter, New York Federal Reserve Bank, for help working through the algebra.