The Changing Importance of Economic Prospects for Assortative Mating

In light of recent changes in the labor force participation and socioeconomic standing of women, we ask whether a woman's position in the labor market has become more important over time as a determinant of her position in the marriage market. To test this hypothesis, we examine change over time in the association of wives' wages and husbands' socioeconomic standing, using data on first marriages among members of two cohorts from the National Longitudinal Survey of Labor Market Experience. Unlike much prior research on assortative mating, we take an individual-level approach to the analysis and rely on improved measures of labor market position, such as measuring wives' wages before marriage and considering various indicators of husbands' socioeconomic standing. Our findings do suggest some increase over time in the importance of economic prospects for assortative mating, with stronger evidence of change observed when husband's longer-term position in the labor market is considered. The Changing Importance of Economic Prospects for Assortative Mating Social scientists consider the question of who marries whom to be a central organizing feature of social life. A large body of literature has established that people tend to seek spouses with particular characteristics, and prefer to marry within their own social group. The tendency toward homogamous mating has been demonstrated along a number of different dimensions, including race and ethnicity (e.g. Alba and Golden 1986; Pagnini and Morgan 1990; Qian 1997; Schoen and Wooldredge 1989), religion (e.g. Bumpass 1970; Johnson 1980; Kalmijn 1991a), social background (e.g. Blackwell 1998; Kalmijn 1991b), age (e.g. Qian 1998; Sweet and Bumpass 1987), educational attainment (e.g. Jacobs and Furstenburg 1986; Kalmijn 1991a; Lichter, Anderson, and Hayward 1995; Mare 1991; Rockwell 1976; Schoen and Wooldredge 1989; Spanier and Glick 1980), and occupational characteristics (e.g. Jacobs and Furstenberg 1986; Hout 1982; Kalmijn 1991a, 1994). Recent changes in factors related to marriage in contemporary societies -particularly the improved labor market position of women and increases in their labor force participation -have contributed to social scientists’ growing interest in investigating shifts over time in patterns of assortative mating. To the extent that current cohorts of women spend more of their lives in the labor market than did previous cohorts, we might expect an increased emphasis on women's socioeconomic characteristics for marriage. Indeed, a model of marriage in which both men and women seek spouses with good labor market prospects underlies much recent research on marriage timing and assortative mating. While some empirical evidence does point to an increasing ECONOMIC PROSPECTS AND ASSORTATIVE MATING 2 association between the educational attainment and occupational characteristics of spouses (e.g. Kalmijn 1991a, 1991b, 1994; Mare 1991), most prior research has examined cross-sectional samples of marriages existing at a particular point in time and has considered a limited array of measures of socioeconomic characteristics. Both factors limit the potential of this research for testing hypotheses about changing marriage behavior. The current research addresses the fundamental question of whether and how the association between the socioeconomic characteristics of husbands and wives may be changing over time. We offer several important extensions on previous research. First, we examine a variety of indicators of husband's socioeconomic status, including observed earnings, expected future earnings, a composite index of occupational status, as well as separate measures of occupational education and occupational earnings. Our approach allows us to consider change in the association of wives' labor market position with both economic and cultural aspects of husbands' occupational standing, and to investigate the sensitivity of this association to the time-horizon over which husband's standing is considered. Second, we use longitudinal data and take an individual-level approach to the analysis, which has several notable advantages over most previous studies. For example, our analysis is less affected by attrition through divorce than are studies examining the characteristics of cross-sectional samples of married couples. Because we have 1 When couple characteristics are reported after any extended period of marriage, selective attrition through differential patterns of divorce becomes problematic. Given some evidence that divorce is related to levels of marital homogamy and other characteristics of spouses (Bumpass, Martin, and Sweet 1991; ECONOMIC PROSPECTS AND ASSORTATIVE MATING 3 measures of wives' wages before marriage, our results should also be less influenced by the impact spouses have on one another after marriage. Finally, our approach permits an examination of the association of spouses' socioeconomic characteristics net of basic controls for factors such as age at marriage and region of residence. THEORY AND PREVIOUS RESEARCH Both sociologists and economists have written extensively on the underlying processes determining partner choice. Economists typically analyze marriage as a voluntary union of rational individuals, with the aim of maximizing some concept of joint consumption. As explained in the classic discussion by Becker (1973, 1974), the division of labor facilitated by marriage will be of greatest advantage for couples with the greatest difference in wage rates. Within this model, negative assortative mating on wages produces greater gains the larger the wage advantage of the spouse who works for wages relative to the spouse who concentrates her (his) efforts at home. Recent additions to this literature in economics explore other views of potential gains from marriage, with implications for patterns of assortative mating. For example, marriage brings benefits as individuals can achieve economies of scale (i.e., it takes very little extra effort to cook a meal for two people rather than one). Marriage also provides a context for investments in household public goods--goods that are collectively consumed by the household in which one spouse's consumption does not reduce the other's (for Sweet and Bumpass 1987; Tzeng and Mare 1995), some bias from this source is expected. Looking at first marriages among recently married couples minimizes, but does not remedy, this problem of selection. ECONOMIC PROSPECTS AND ASSORTATIVE MATING 4 example, a picture on the wall or the well-being of children). This aspect of marriage depends on the spouses having similar tastes, implying positive assortative mating on factors associated with taste for public goods (Lam 1988). To the extent that these factors (for example, education) are also related to wages and other measures of socioeconomic standing, this approach implies positive assortative mating on these characteristics. Economic models of marriage have heavily influenced the thinking of sociologists. Social exchange theory (e.g. Edwards 1969, Schoen and Wooldredge 1989), for example, makes use of the economic metaphor of social relationships as extended markets in which individuals attempt to maximize gains and minimize costs through marriage. The focus of exchange theory is on the resources individuals trade in order to maximize rewards. It is assumed that men and women will most often marry spouses with similar levels of resources, largely because of the tendency to reject those with fewer resources than themselves. The resources that are exchanged, however, need not be identical in nature so long as they are considered equivalent. Departures from homogamy are expected to occur most often when these equally valued, but not identical, resources are exchanged. Because of the tendency for men and women to fill different roles in society, social exchange theory suggests that such departures from homogamy will likely involve the exchange of men’s socioeconomic resources (such as income and status) for women’s non-economic resources (such as social and domestic services). Oppenheimer (1988), drawing on ideas from social exchange theory and job search theory in economics, has suggested that the basis for assortative mating has ECONOMIC PROSPECTS AND ASSORTATIVE MATING 5 changed over time. As the tendency for both women and men to remain attached to a work career increases, Oppenheimer argues that the characteristics considered important in a spouse also change. In particular, women are increasingly evaluated as potential spouses on the basis of their own achieved socioeconomic status, rather than more traditional characteristics such as religion, family background, and physical attractiveness. Thus Oppenheimer's theory implies an increasing association over time between wives' wages and the socioeconomic standing of their husbands. Oppenheimer further emphasizes that the long-run benefits of marriage, and thus also assortative mating outcomes, depend heavily on the expected future characteristics of spouses. Indeed, she argues that the difficulty of assessing important future characteristics of spouses based on the incomplete information available at any given time seriously complicates the process of partner choice. The central assumption underlying Oppenheimer's theory is that a woman's achieved socioeconomic standing, rather than her productivity in the home or social background, has increased in importance over time as a determinant of her position in the marriage market. If Oppenheimer's theory were correct, we would expect that women's productivity in the labor market would display a strengthening association over time with the socioeconomic standing of the men who they marry. Several recent investigations of change in patterns of assortative mating have directly or indirectly tested this hypothesis, mostly relying on log-linear analyses of contingency tables of husbands' and wives' postmarriage characteristics. For example, several studies report an increase in the association of spouses' educational attainments between 1940 and the 1980s (Blackwell 1998; ECONOMIC PROSPECTS AND ASSORTATIVE MATING 6 Kalmijn 1991a; Mare 1991). Evidence of some increase in educational homogamy persists when trends are adjusted for changes in the timing of marriage and school leaving during this period, particularly at the upper end of the schooling distribution (Mare 1991). In other research, Kalmijn (1994) considers change in the association of newlyweds' (i.e. couples married no more than two years) occupational standing between 1970 and 1980. One important strength of this study is that Kalmijn considers multiple dimensions of occupational status, considering both spouses' economic status (occupational earnings) and their cultural status (occupational education). While Kalmijn find evidence of some increase in the importance of occupational earnings for partner choice during this period, he reports some reduction in the importance of occupational education. In both periods, however, he finds stronger homogamy with respect to occupational education than with respect to occupational earnings. Although these prior investigations have provided important insights into trends in patterns of assortative mating, they are limited in their ability to provide a solid test of Oppenheimer's theory. For example, greater educational attainment may contribute to productivity both in the labor market and in the home (Cancian 1995). Another empirical complication faced by previous analyses is the potential endogeneity of labor market activities of spouses-in particular, the potential for wives’ labor force participation to respond to their husband’s earnings, commonly referred to by economists as the "income effect." Suppose, for example, we observe that high-wage men are now more likely to be married to women also earning high wages. It may be that high wage men are increasingly marrying women with the potential to earn high wages. Alternatively, ECONOMIC PROSPECTS AND ASSORTATIVE MATING 7 marriage patterns may have remained stable, and labor force participation patterns of married women may have changed. In particular, it may be that women married to highearning men generally did not pursue high-wage jobs after marriage in the earlier period. These women may have had high wage potential, but relatively low observed wages after marriage. In order to measure changes in assortative mating we must distinguish between changes in marriage formation and changes in the impact of husbands' socioeconomic standing on wives' labor market decisions within marriage. Our research addresses these concerns, and builds on previous trend studies of patterns of assortative mating. First, to assess the importance of wives' potential in the labor market, we rely on measures of wages, which are generally considered by economists to explicitly reflect productivity in the labor market. Second, we analyze longitudinal data on labor market and marital histories, and are thus able to measure wife’s productivity in the labor market before marriage, indexed by her pre-marriage wage. Wages observed before marriage should be largely independent of husband's earnings. Finally, we consider multiple measures of husband's socioeconomic standing, including measures that reflect shorter-term and longer-term socioeconomic standing as well as differently signal the cultural and the more explicitly economic status that accompanies labor market position. DATA We use data from the National Longitudinal Survey of Young Women (NLSYW) and the National Longitudinal Survey of Youth (NLSY) to model changes in assortative mating over two cohorts of young women. These data are particularly well suited for the ECONOMIC PROSPECTS AND ASSORTATIVE MATING 8 current research, given their large and nationally representative samples of young adults, extensive information on socioeconomic characteristics of respondents and their husbands, and multiple cohort design. Information for the early cohort (NLSYW) comes from 12 surveys fielded over 16 years, from 1968 to 1983. Responses to 16 annual surveys conducted between 1979 and 1994 are used for the late cohort (NLSY). Our analytic sample was limited to white women who were between the ages of 14 and 17 and never married at first interview, and who subsequently married by the 1982 (NLSYW) or 1993 (NLSY) interview. We further eliminate women who married before the age of 17, as these are considered non-normative transitions which are not of direct interest to the current analysis. To be included in a particular regression analysis, respondents must also have non-missing data on own socioeconomic status, the socioeconomic status of their husbands, age at marriage, and the regional and SMSA status of their residence in the year of marriage. Because patterns of partner choice may vary by marriage order (e.g. Jacobs and Furstenberg 1986), our study examines only patterns of assortative mating among women entering first marriages. These restrictions lead to a maximum sample size of 789 early cohort women and 794 late cohort women. 2 The sample of young women was interviewed annually from 1968 through 1973, as well as in 1975, 1977, 1978, 1980, 1982, and 1983. Interviews were also conducted after 1983, but these data are not used in the current analysis 3 Actual sample sizes for particular regression models are indicated in the table of results. These vary due to differences in levels of missing data among particular measures of husbands' and wives' socioeconomic characteristics. ECONOMIC PROSPECTS AND ASSORTATIVE MATING 9 Although these surveys also contain reasonably large samples of black women and their husbands, we have limited our analysis to whites for several reasons. First, rates of non-marriage have been higher (and increasing more rapidly) among black women than among white women. Indeed, in 1990, 85.6 percent of white women ages 30 to 34 had married, compared with only 61.1 percent of similar black women (Norton and Miller 1992). We feel that modeling non-marriage is essential for understanding change in patterns of assortative mating among blacks. Second, previous research points to differences in the spouse preferences of blacks and whites, suggesting that the underlying process of assortative mating may differ by race. For example, South (1991) reports that black men are less willing than white men to marry someone who is unlikely to hold a steady job, who earns much less than they do, or has less education than themselves. South finds black women, however, to be more willing than white women to marry someone who is unlikely to hold a steady job, but less willing to marry someone with either more or less education than themselves or whom they do not consider to be good looking. Although beyond the scope of the current analysis, more research in needed to better understanding these racial differences in partner preferences and in patterns of assortative mating. 4 South’s sample, however, includes only unmarried and non-cohabiting people under age 35. Given high rates of cohabitation among people in this age group (Bumpass and Sweet 1989), the extent to which South's can be reasonably generalized to the population of unmarried people is questionable. ECONOMIC PROSPECTS AND ASSORTATIVE MATING 10 VARIABLES Husbands' Earnings Two approaches are taken to measuring husbands' earnings. First, we compute the natural log of husband's observed earnings in the year following marriage. These measures are summed from questions about wage and salary, farm and business, and military earnings taken directly from the surveys. Economic theory, however, suggests that the earnings measure of interest should be permanent income, rather than current earnings. Given differences in age-earnings profiles and the ability to smooth income and consumption over time, a prospective spouse should consider the present discounted value of total earnings, rather than current earnings. In her theory of marriage timing, sociologist Valerie Oppenheimer (1988) also emphasizes the long-run labor market position of potential spouses. We therefore also construct a second measure of longer-run earnings, assessing earnings expected ten years in the future. While our second measure does not account in full for the expected age-earnings profile, we believe it may be superior to current earnings. We base our measure of expected future earnings on actual trajectories observed in the 5 percent public use micro-sample of the 1970 census. In this year, individuals reported their earnings as well as the occupation they held five years prior to 5 The first valid observation of husband's earnings after the year of marriage is selected for this variable. All measures of observed wages and earnings are transformed into 1990 dollars using the Consumer Price