PAUL R. AMATO,DAVID R. JOHNSON,ALAN BOOTH, AND STACY J. ROGERS The Pennsylvania State University

Continuity and Change in Marital Quality Between 1980 and 2000

We use data from two national surveys of married rising age at ®rst marriage, the continuing high individualsÐone from 1980 and the other from rate, and the declining remarriage rate 2000Ðto understand how three dimensions of demonstrate that marriage is a less permanent part marital quality changed during this period. Mar- of adult life now than in the recent past. Accord- ital happiness and divorce proneness changed lit- ing to a marital decline perspective, people are tle between 1980 and 2000, but marital interac- turning away from marriage because it has be- tion declined signi®cantly. A decomposition come increasingly dif®cult to maintain happy and analysis suggested that offsetting trends affected stable unions. marital quality. Increases in marital heterogamy, Other observers view recent changes in mar- premarital , ' extended hours of riage as being benign or even bene®cial (Coontz, employment, and wives' job demands were asso- 1992; Scanzoni, 2001; Skolnick, 1991; Stacey, ciated with declines in multiple dimensions of 1996). According to this view, although it is easier marital quality. In contrast, increases in economic for people to leave unhappy marriages these days, resources, decision-making equality, nontradition- the proportion of unhappy marriages in the pop- al attitudes toward gender, and support for the ulation has not necessarily increased. Indeed, be- norm of lifelong marriage were associated with cause divorce removes unhappy couples from the improvements in multiple dimensions of marital married population, existing marriages may be of quality. Increases in husbands' share of house- higher quality now than in the past. Moreover, in- work appeared to depress marital quality among creases in married women's education, employ- husbands but to improve marital quality among ment, and income have raised women's status and wives. given wives greater decision-making power, thus increasing the potential for less patriarchal and The belief that the institution of marriage is in more egalitarian marital relationships. decline is widespread (Bellah, Madsen, Sullivan, With respect to marital instability, the divorce Swidler, & Glenn, 1996; Popenoe, 1993; Tipton, rate increased dramatically during the 1970s, 1985; Zill & Nord, 1994). The growing popularity peaked around 1980, then declined slightly (U.S. of nonmarital cohabitation, the increasing per- Bureau of the Census, 2000, Table 77). Therefore, centage of children born outside of marriage, the although divorce is common these days, relatively little change in marital instability occurred during Department of Sociology, The Pennsylvania State Univer- the last two decades. Evidence about the quality sity, University Park, PA 16802-6207 ([email protected]). of existing marriages, however, is sparse. Using Key Words: decomposition analysis, change, family merged data from the General Social Survey, demography, marital quality, marriage. Glenn (1991) found that the percentage of people

Journal of Marriage and Family 65 (February 2003): 1±22 1 2 Journal of Marriage and Family reporting that their marriages were very happy de- marital happiness tends to wane continuously with clined between 1973 and 1988. Similarly, Rogers duration of marriage (VanLaningham, Johnson, & and Amato (2000) compared a sample of young Amato, 2000). Therefore, if there are fewer mar- adults married between 1981 and 1997 (and sur- riages of long duration these days, then mean lev- veyed in 1997) with a sample of young adults els of marital happiness in the population may married between 1964 and 1980 (and surveyed in have increased. 1980) and found that the more recent marriage cohort reported less interaction and more marital Remarriage. The percentage of marriages involv- con¯ict. Although these studies suggest a decline ing previously married increased substan- in the aggregate level of marital quality in the tially during the 1970s and stabilized during the population, the lack of research on this topic 1980s. Currently, about half of all marriages in- makes it dif®cult to reach conclusions with cer- volve a second or higher-order marriage for one tainty. The present study uses two national, rep- or both spouses (U.S. Bureau of the Census, 2000, resentative samples of married individuals in the Table 145). Spouses in second (or higher-order) United StatesÐone interviewed in 1980 and the marriages, compared with spouses in ®rst mar- other interviewed in 2000Ðto provide new evi- riages, tend to report lower marital quality (Booth dence on recent changes in the nature and quality & Edwards, 1992) and are more likely to see their of marriage. marriage end in divorce (White, 1990). Rates of remarriage following divorce have been declining in recent years, but if the current population of SOCIAL CHANGE 1980±2000 married individuals contains a higher proportion During the latter part of the 20th century, a variety of people in second or higher-order marriages, of demographic, economic, social, and cultural then the aggregate level of marital quality may changes in the U.S. population had the potential have declined. to affect people's marriages either positively or negatively. Children. The proportion of without chil- dren under the age of 18 increased slightly, from 48% in 1980 to 52% in 1999 (U.S. Bureau of the Demographic Characteristics Census, 2000, Table 65). Studies typically show Age at marriage. Age at ®rst marriage began to that couples with children report less marital hap- rise in the early 1960s and continued to rise piness (Glenn & McLanahan, 1982). Observers through the 1990s. For men, the median age at disagree about whether this association exists be- ®rst marriage increased from 24.7 in 1980 to 26.7 cause having children lowers marital quality in 1998. The corresponding ®gures for women (Cowan & Cowan, 1992) or because children re- were 22.0 and 25.0 (U.S. Bureau of the Census, duce the likelihood that unhappily married people 1998). People who marry at young ages, com- will divorce (White, Booth, & Edwards, 1986). pared with those who marry at older ages, spend Regardless of the explanation, the decline in the less time searching for suitable partners, have proportion of couples with children may have fewer ®nancial resources, and are less mature psy- raised the mean level of marital quality in the pop- chologically. For these reasons, people who marry ulation. at young agesÐespecially during their teen yearsÐare at increased risk of being unhappily Race and ethnicity. Because African Americans married and seeing their marriages end in divorce are more likely than Whites to divorce (White, (Booth & Edwards, 1985; Bumpass, Martin, & 1990), it is likely that African Americans also ex- Sweet, 1991). The increase in age at marriage, perience more marital discord, although there is therefore, may have improved the quality of re- little research that directly compares these two cent marriages. groups. Nevertheless, the proportion of all married individuals who are African American changed Duration of marriage. Because people are mar- little between 1980 and 1999 (U.S. Bureau of the rying at older ages, and because an increasing pro- Census, 2000, Table 53), suggesting few impli- portion of unions are second (or higher-order) cations for aggregate levels of marital quality. The marriages for one or both spouses, the average percentage of Hispanic married couples, in con- duration of marriage in the population may have trast, nearly doubled during this period (U.S. Bu- declined. Recent longitudinal data indicate that reau of the Census, 2000, Table 53). Rates of di- Marital Quality: 1980±2000 3 vorce vary considerably between Hispanic groups, may undermine later marital stability. It is possi- and little research is available on marital quality ble, therefore, that the increase in premarital co- in this population. Consequently, the implications habitation has had negative consequences for mar- of the increase in the married Hispanic population riage (Seltzer, 2000). for marital quality in the United States are unclear. Education, Income, and Employment Heterogamy. Although people tend to choose spouses much like themselves with respect to age, Both women and men experienced increases in education, religion, and race, marital heterogamy education between 1980 and 2000. For example, has increased in recent years. For example, the the proportion of women 25 years of age and old- proportion of marriages that are interracial in- er who had completed 4 or more years of college creased from 1.3% in 1980 to 2.4% in 1999 (U.S. rose from 13% in 1980 to 23% in 1999; the cor- Bureau of the Census, 2000, Table 54). The in- responding percentages for men were 20% and crease in interracial and interethnic marriage holds 27%, respectively (U.S. Bureau of the Census, for almost all groups, including ethnic Whites 2000, Table 250). A high level of educational at- (such as Italian Americans and Polish Americans), tainment is associated with individual and family Blacks, Latinos, Asians, and Native Americans characteristics that promote marital happiness and (Rosenfeld, 2002). Age-heterogamous marriages, stability. Well-educated individuals, compared especially those in which wives are older than with poorly educated individuals, earn more in- their husbands, also increased during this time come, possess better communication skills, are at (U.S. Bureau of the Census, 2000, Table 56). The lower risk of depression, and experience a stron- increase in heterogamy re¯ects a relaxation of ger sense of personal control (Amato & Booth, prohibitive social attitudes toward intermarriage. 1997; Farley, 1996; Ross & Wu, 1995). For these Despite a more supportive social climate, how- reasons, increases in education may have contrib- ever, individuals in heterogamous marriages uted to general improvements in marital quality. (based on age, race, religion, and education) re- Between 1980 and 2000, many married cou- port less marital happiness and divorce more often ples experienced improvements in economic well- than do individuals in homogamous marriages being. Correcting for in¯ation, the median family (Booth & Edwards, 1992; Heaton, 2002). The in- income of married-couple families rose from crease in marital heterogamy during the last few about $46,000 in 1980 to about $55,000 in 1998 decades, therefore, may have lowered the mean (U.S. Bureau of the Census, 2000, Table 748). level of marital quality in the population. During this same period, levels of unemployment and family poverty declined for all races. Eco- Premarital cohabitation. Cohabitation among un- nomic advancement, however, occurred primarily married couples has increased dramatically since among married-couple families with two-wage 1980. The number of unrelated adults of the op- earners. Among husband-earner married-couple posite sex living together rose from 1.6 million in families, median family income was relatively 1980 to 4.5 million in 1997 (U.S. Bureau of the stagnant during this time (White & Rogers, 2000). Census, 2000, Table 57). Studies consistently Despite the fact that economic growth during the show that couples who cohabited prior to mar- last two decades was not equally shared by all riage, compared with those who did not, report families, changes in the U.S. economy probably lower marital happiness, interact less frequently, raised the overall level of marital quality in the have more arguments, and are more likely to see population. their marriages end in divorce (Booth & Johnson, During the last two decades of the 20th cen- 1988; Bumpass et al., 1991). Many observers be- tury, the long-term convergence in husbands' and lieve that selection into cohabitation is responsible wives' economic roles continued. Between 1980 for these trendsÐthat is, the same traits that lead and 1999, the labor force participation rate de- people to cohabit (such as holding unconventional clined from 80% to 77% for married men and attitudes toward marriage or having a low level of increased from 50% to 61% for married women income) also increase the risk for later marital (U.S. Bureau of the Census, 2000, Table 651). problems (Booth & Johnson). Longitudinal stud- Married women now demonstrate a pattern of ies, however, indicate that cohabitation tends to consistent attachment to the labor force, with most weaken people's commitment to the norm of life- wives no longer leaving the labor force during the long marriage (Axinn & Thornton, 1992), which prime years of childbearing (Spain & Bianchi, 4 Journal of Marriage and Family

1996). Although wives' employment improved Cherlin, 2000). This difference may occur because the economic well-being of families, as noted ear- wives who adopt less traditional views often en- lier, it also increased the potential for work-family counter resistance from their husbands, whereas con¯ict (Perry-Jenkins, Repetti, & Crouter, 2000). husbands who adopt less traditional views often For this reason, we assume that with income con- receive support from their wives. The conse- trolled, the increase in dual-earner arrangements quences of this shift in attitudes for marital qual- has raised the level of discord in marriage, espe- ity, therefore, may vary with sex. cially when wives work long hours or have highly A rise in individualistic values, rooted in the demanding jobs. desire to maximize self-ful®llment, may have un- dermined people's commitment to the norm of lifelong marriage (Bellah et al., 1985; Glenn, Gender Relations 1996). Individuals with a strong belief in marital Married women's increased commitment to paid permanence are likely to invest a good deal of work has been accompanied by a decline in time and effort in trying to resolve marital dis- household work among wives and a modest in- agreements and problems. In contrast, individuals crease in household work among husbands (Col- who are tolerant of divorce may prefer to jettison trane, 2000). Moreover, economic resources are a an existing marriage (rather than invest more ef- basis for marital power, and wives often expect fort in it) to ®nd greater happiness with a new (and get) more decision-making power when they partner. Amato and Rogers (1999) found that in- make ®nancial contributions to the marriage dividuals who adopted more favorable attitudes (Blumberg & Coleman, 1989). How might these toward divorce tended to report declines in marital changes have affected marital quality? On the one happiness and increases in marital con¯ict. If peo- hand, contention over the household division of ple are less likely to endorse the norm of marital labor and the balance of marital power has be- permanence now than in the recent past, then a come prevalent since 1980 (Hochschild, 1989). shift in attitudes may have lowered the aggregate But on the other hand, shared involvement in fam- level of marital quality in the population. ily work and paid work has the potential to create It is not clear whether people in the United egalitarian, and perhaps more intimate, marital re- States have become more or less religious during lationships (Hood, 1983; Scanzoni, 1978). Obser- the last few decades. People with a strong reli- vational research indicates that asymmetric pow- gious orientation tend to support the norm of life- erÐwith husbands having more power than long marriage, have relatively low rates of divor- wivesÐis more common among dissatis®ed cou- ce, and report high levels of marital happiness ples than satis®ed couples (Gottman, 1994, p. 57). (Heaton & Pratt, 1990). For these reasons, in- These contrasting views make it dif®cult to pre- creases in religiosity may have improved the qual- dict whether the growing convergence in work ity and stability of existing marriages. and family roles (increases in wives' share of household income, husbands' share of housework, GOALS OF THE CURRENT STUDY and decision-making equality) has resulted, over- all, in greater tension or greater cohesion between Our review suggests that social change during the spouses. last two decades of the 20th century had complex and contradictory effects on marital quality. Some changes, such as the increase in age at marriage, Attitudes and Values a possible decline in marital duration, the rise in As the economic roles of men and women in mar- education, the decline in marital fertility, the riage have changed, men as well as women have growth of wives' earnings and family income, the adopted less traditional views about marriage. For increase in decision-making equality, and a pos- example, the proportion of people who believe sible rise in religiosity, may have improved the that husbands should be breadwinners and wives quality of existing marriages. Other changes, such should be homemakers has declined substantially as the growing popularity of premarital cohabita- since the 1950s (Thornton & Young-DeMarco, tion, the increase in remarriage, the rise in heter- 2001). Several studies indicate that adopting less ogamy, the increase in employment demands for traditional gender attitudes is associated with low- wives (and corresponding work-family con¯ict), er marital quality among wives but higher marital and a decline in support for the norm of lifelong quality among husbands (Amato & Booth, 1995; marriage, may have been detrimental to marital Marital Quality: 1980±2000 5 quality. And the implications of other changes, (Booth, Johnson, White, & Edwards, 1981). The such as shifts in the racial and ethnic composition target population consisted of all married individ- of the married population, husbands' growing in- uals in households in the contiguous United States volvement in family work, and changes in atti- in 1980 with a telephone, both spouses present, tudes about gender roles in marriage, are dif®cult and both spouses 55 years of age or less. Tele- to predict. Consequently, it is not clear whether phone interviewers used random-digit dialing to the aggregate level of marital quality in the pop- select a sample of households and a second ran- ulation improved or declined in recent decades. dom procedure to determine whether to interview Of course, it is possible that these positive and the husband or . Seventeen percent of targeted negative trends offset one another, resulting in lit- individuals could not be reached after 20 tele- tle net change in marital quality. phone calls. Of those individuals contacted, 78% We take advantage of two national surveys, one gave complete interviews, for an overall response conducted in 1980 and the other conducted in rate of 65%. The ®nal sample consisted of 2,034 2000, to explore recent changes in marriage and married persons (not couples). marital quality. Our research has four goals. First, The second sample, the Survey of Marriage we use these surveys to describe changes in rele- and Family Life, was collected in 2000 (Booth, vant demographic, economic, social, and cultural Amato, Johnson, & Rogers, 2002). As in the 1980 variables during this 20-year period. Second, we study, telephone interviewers contacted a random document changes in three dimensions of marital sample of households and randomly selected the quality during the same period. Although some re- husband or wife to serve as the respondent. Be- searchers restrict the term marital quality to spous- cause response rates to telephone surveys have de- es' reports of satisfaction (e.g., Fincham & Brad- clined during the last 20 years, we sent selected bury, 1987), other researchers view marital quality households a letter (along with a $2 bill) prior to as multidimensional (e.g., Amato & Booth, 1997). calling. Pretesting revealed that this procedure We include spouses' reports of happiness with mar- substantially increased levels of cooperation. The riage, the frequency of shared activities (which we overall response rate for the 2000 survey (includ- refer to as marital interaction), and thoughts or ac- ing refusals and failures to locate targeted respon- tions that may lead to divorce (which we refer to dents) was 63%. The total sample consisted of as divorce proneness). Prior research indicates that 2,100 individuals who met the same sampling cri- these three marital dimensions are conceptually and teria used in 1980, that is, respondents were mar- empirically distinct (Johnson, White, Edwards, & ried, respondents' spouses lived in the household, Booth, 1986). Furthermore, using three marital out- and both spouses were 55 years of age or less. comes (rather than one) allows us to cast a wide Inspection of the 1980 and 2000 Current Pop- net in determining how people's experiences within ulation Surveys revealed that our 1980 and 2000 marriage have changed. Third, we conduct a de- samples were representative of the U.S. married composition analysis to determine the extent to population (age 55 or under) in many respects. which changes in demographic, economic, social, Both samples contained the usual survey biases, and cultural variables (our explanatory variables) however, with men, young respondents, minorities, may have contributed to changes in aggregate lev- poorly educated individuals, and people residing in els of marital quality in the population. Finally, in metropolitan areas being underrepresented. For this keeping with a long research tradition based on reason, we weighted the 1980 sample to represent ``his and her'' marriages (Bernard, 1972), we con- the 1980 U.S. population of married individuals sider whether these trends differed for husbands (age 55 or under) with respect to sex, age, race and wives. (non-Hispanic White, non-Hispanic Black, Hispan- ic, and other), years of education, metropolitan sta- tus, and household size. Correspondingly, we METHOD weighted the 2000 sample to represent the 2000 U.S. population of married individuals (age 55 or Sample under) on the same characteristics. Our analysis involves two national probability samples of married individuals. The ®rst sample, Variables collected in 1980, was the ®rst wave of a six-wave The wording of all survey items was identical in longitudinal study that has come to be known as 1980 and 2000. Descriptive information on all the Marital Instability Over the Life Course study variables is available in Table 1. 6 Journal of Marriage and Family

TABLE 1. DESCRIPTIVE DATA FOR ALL VARIABLES Miss- ing Data

Variables Range (%) xÅ1980 S1980 xÅ2000 S2000 xÅ2000 Ϫ xÅ1980 Demographic variables Wife respondent 0±1 0.00 0.50 0.50 0.50 0.50 0.00 Age at current marriage 13±54 0.07 22.90 5.55 25.86 6.71 2.96*** Years married 0±38 0.07 13.27 9.65 13.27 9.12 0.00 Either remarried 0±1 0.00 0.20 0.40 0.29 0.45 0.09*** Any preschool children 0±4 0.00 0.32 0.47 0.30 0.46 Ϫ0.02 Any school-age children 0±6 0.00 0.52 0.50 0.50 0.50 Ϫ0.02 Black respondent 0±1 0.63 0.07 0.26 0.09 0.28 0.02 Latino respondent 0±1 0.63 0.06 0.23 0.11 0.32 0.06*** Other non-White respondent 0±1 0.63 0.02 0.14 0.06 0.24 0.04*** Heterogamy index 0±5 0.00 0.76 0.87 0.89 0.99 0.13*** Premarital cohabitation 0±1 0.46 0.16 0.37 0.41 0.49 0.24*** Employment and income Education respondent 0±28 0.17 12.93 2.81 13.89 2.85 0.96*** Husband employed 0±1 1.45 0.95 0.23 0.92 0.26 Ϫ0.03* Wife employed part-time 0±1 2.20 0.15 0.36 0.21 0.41 0.06*** Wife employed full-time 0±1 2.20 0.32 0.47 0.34 0.47 0.02 Wife extended hours 0±1 2.20 0.11 0.31 0.21 0.41 0.10*** Husband job demands 0±4 0.10 1.35 1.11 1.37 1.19 0.02 Wife job demands 0±4 0.10 0.46 0.84 0.68 0.98 0.22*** Family income (in 1,000s of constant 2000 dollars) 2.50±271.78 8.75 55.60 26.93 70.00 34.27 14.40*** Public assistance 0±1 3.87 0.11 0.31 0.07 0.26 Ϫ0.04*** Finances getting better 0±1 0.44 0.60 0.49 0.63 0.48 0.03 Finances getting worse 0±1 0.44 0.14 0.34 0.07 0.26 Ϫ0.07*** Gender relations Wife proportion income 0±1 8.79 0.21 0.22 0.32 0.21 0.11*** Husband share housework 0±1 2.18 0.27 0.18 0.35 0.19 0.08*** Equal decision making 0±1 0.59 0.49 0.50 0.64 0.48 0.15*** Attitudes and values Traditional gender attitudes 1±4 2.00 2.42 0.46 2.26 0.43 Ϫ0.16*** Support lifelong marriage 1±4 3.17 2.62 0.41 2.73 0.43 0.11*** Religiosity 1±5 0.78 3.64 1.23 3.81 1.29 0.17*** Marital quality Marital happiness Ϫ4.03±1.26 2.83 0.00 1.00 0.00 1.00 0.00 Marital interaction Ϫ3.41±1.55 2.10 0.16 0.96 Ϫ0.16 1.01 Ϫ0.32*** Divorce proneness Ϫ0.99±2.41 0.15 0.01 0.97 Ϫ0.01 1.02 Ϫ0.01 Note: Sample size is 2,034 in 1980 and 2,100 in 2000. Signi®cance tests (two-tailed) are based on ␹2 for dichotomous variables and t tests for ordered variables. *p Ͻ .05. **p Ͻ .01. ***p Ͻ .001.

Demographic variables. The interview collected other), and whether the couple had lived togeth- information on the respondent's sex, age at ®rst er prior to marriage. We also created a heter- marriage, age at current marriage, number of ogamy index based on whether the respondent years married to current spouse, whether the re- and the respondent's spouse had different races, spondent or the respondent's spouse had been religions, levels of education (de®ned as a dif- married previously, the number of children age ference of 4 years or more), and ages (de®ned 0±5 in the household (preschool children), the as a difference of 6 years or more). To reduce number of children age 6±18 in the household the number of explanatory variables, we created (school-age children), the respondent's race or a single index of heterogamy by adding the ethnicity (non-Hispanic White, non-Hispanic number of differences, yielding a variable that Black, Hispanic, Asian, Native American, or ranged from 0 to 4. Marital Quality: 1980±2000 7

Education, employment, and income. The inter- we asked, ``Overall, considering all the kinds of view schedule contained standard questions on decision you two make, does your spouse more years of education and husbands' and wives' usu- often have the ®nal word or do you?'' Responses al hours of employment. Because the great ma- that referred to compromising or sharing decisions jority of husbands worked full-time, we collapsed were coded 1, and responses re¯ecting dominance husbands' employment into two categories: em- by one spouse (husband or wife) were coded 0. ployed versus not employed. We divided wives' Although researchers disagree about how marital employment into four categories: not employed, power should be operationalized (Mizan, 1994), part-time employment (1±34 hours per week), ®nal-word measures continue to be the most com- full-time employment (35±45 hours per week), monly used method for assessing this construct. and extended hours of employment (46ϩ hours per week). We distinguished between wives with Attitudes and values. We measured traditional typical full-time schedules and wives with more gender attitudes with a seven-item scale. Sample extensive work schedules to determine the con- items included ``A woman's most important task sequences of wives' time demands on marriage. in life should be taking care of her children'' and We did not include data on husbands' hours of ``Even though a wife works outside the home, the employment for three reasons. First, there was husband should be the main breadwinner, and the considerably less variability in husbands' work wife should have responsibility for the home and hours than in wives' work hours. Second, prelim- children'' (1 ϭ disagree strongly;2ϭ disagree; inary analyses revealed little evidence that hus- 3 ϭ agree;4ϭ agree strongly). Items were bands' work hours affected marital quality. And scored in the direction of traditional attitudes, and third, wives' full-time employment is more likely the mean of the responses served as the scale than husbands' full-time employment to be a score (␣ϭ.65 in 1980 and .66 in 2000). We source of marital tension, at least among couples measured attitudes toward the norm of lifelong with traditional views of marriage. To measure marriage with six questions, including ``Couples other job demands, we asked whether the hus- are able to get divorced too easily today'' and bands' or the wives' jobs involved irregular hours, ``Marriage is for life, even if the couple is unhap- shift work, evening meetings, or overnight trips. py'' (1 ϭ strongly disagree;2ϭ disagree;3ϭ We calculated the number of job demands, which agree;4ϭ strongly agree). Items were scored in ranged from 0 to 4, separately for husbands and the direction of support for lifelong marriage, and wives. the mean served as the scale score (␣ϭ.58 in With respect to economic well-being, couple 1980 and .61 in 2000). A single item measured income was measured in thousands of dollars, and religiosity: ``In general, how much do your reli- the wife's contribution was represented as a pro- gious beliefs in¯uence your daily life?'' (1 ϭ portion of the total. We also inquired about wheth- none;5ϭ very much). er the couple had relied on any form of public assistance during the previous 3 years (0 ϭ no;1 Marital quality. We used a 10-item scale of mar- ϭ yes). In addition, respondents indicated whether ital happiness. Sample items included ``How hap- their ®nancial situation during the last 3 years had py are you with the amount of understanding you improved, become worse, or stayed the same. Two receive from your spouse? With the amount of dichotomies (improved and became worse) rep- love and affection you receive? With your sexual resented this variable in the analysis. relationship? With your marriage overall?'' Re- sponses were scored in the direction of greater Housework and decision making. To measure hus- happiness (1 ϭ not too happy;2ϭ pretty happy; bands' share of housework, we asked, ``In every 3 ϭ very happy), and the mean response served family there are a lot of routine tasks that have to as the scale score. Alpha coef®cients were .87 in be done, such as cleaning the house, doing the 1980 and .89 in 2000. To measure marital inter- laundry, cleaning up after meals, cooking dinner, action, respondents were asked how often they en- and so on. How much of this kind of work usually gaged in six activities with the spouse: eating din- is done by you?'' Response options were none (0), ner, shopping, visiting friends, working on less than half (.25), about half (.5), more than half projects around the house, and going out for rec- (.75), and all of it (1.0). Depending on the re- reation. Responses were scored in the direction of spondent's sex, we coded responses to represent frequent interaction (1 ϭ never;4ϭ almost al- the husband's share. To measure decision making, ways), and the mean response served as the scale 8 Journal of Marriage and Family score. Alpha coef®cients were .64 in 1980 and .69 The equation partitions the change in yÅ (a dimen- in 2000. The ®nal scale was a 27-item measure of sion of marital quality) into four components: divorce proneness. Items dealt with divorce-relat- change in the intercepts of the two regression ed thoughts, such as thinking about divorce, as equations (⌬a), changes in the estimated effects well as behaviors, such as discussing divorce with of the explanatory variables (xÅ1⌬b), changes in the the spouse. Items were summed, but because the levels (means) of the explanatory variables (b1⌬xÅ), resulting distribution was severely positively and joint changes in the levels and effects of the skewed, the log (base 10) of the sum served as explanatory variables (⌬xÅ⌬b). The ®rst compo- the scale score. Alpha coef®cients were .92 in nent can be thought of as the residual change over 1980 and .92 in 2000. After pooling the 1980 and time with all explanatory variables controlled. The 2000 samples, the three marital quality variables second component represents the estimated effects were standardized to have means of 0 and stan- of changes in the b coef®cients between the two dard deviations of 1. periods, assuming no changes in the means of the In 1980, marital happiness correlated at r ϭ explanatory variables. The third component re- .44 with marital interaction; marital happiness cor- ¯ects the estimated effects of changes in the related at r ϭϪ.53 with divorce proneness; and means of the explanatory variables between the marital interaction correlated at r ϭϪ.32 with two periods, assuming no changes in the b coef- divorce proneness. In 2000, the corresponding ®cients. The fourth component captures the inter- correlations were .45, Ϫ.51, and Ϫ.26, respec- action between changes in the means and changes tively. Although all correlations were statistically in the b coef®cients. Summing the second, third, signi®cant (p Ͻ .001), the correlations were mod- and fourth component for each explanatory vari- erate enough to justify examining each dimension able provides a measure of the overall contribu- of marital quality separately. (The maximum 2 tion of each explanatory variable to the aggregate shared variance was .53 ϭ .28.) change in marital quality. (Our later discussion fo- cuses on this sum.) Analysis All independent variables are grand mean cen- Steps in the analysis. To identify overall changes tered (centered on the mean of the pooled sam- in marriage between 1980 and 2000, we compared ples) in the regression models used in the analysis. the means of the explanatory variables (age at Centering is a method used in the multilevel mod- marriage, family income, husbands' share of eling literature (Bryk & Raudenbush, 1992) to housework, etc.), as well as the marital quality solve the problem of arbitrary intercepts found in variables (happiness, interaction, and divorce components models with interval-level variables proneness), by decade. Next, we regressed the that have no true zero value. (For a discussion of three measures of marital quality on the explana- this issue, see Firebaugh, 1997, pp. 36±39.) After tory variables separately for the 1980 and 2000 conducting decomposition analyses for the entire samples. We then conducted a decomposition sample, we consider whether these patterns differ analysis for each marital quality variable, follow- for husbands and wives by conducting separate ing procedures described by Firebaugh (1997) for decomposition analyses for each sex. Because the analyzing social change. Decomposition analysis data are weighted, we used the survey analysis is a useful tool for examining repeated cross-sec- module in Stata (StataCorp, 1999) to take sample tions to determine how changes in multiple ex- weights into account when calculating standard planatory variables are related to changes in a de- errors. Across all of the analyses described in the pendent variable over time. following section, design effects caused by In the following equation, xÅ represents a vector weighting ranged from 1.17 to 2.23, with a mean of means and b represents a vector of regression design effect of 1.34. coef®cients. Using subscripts to denote time (1 ϭ 1980, 2 ϭ 2000), change in the mean level of Missing data. We used two procedures for dealing marital quality can be expressed as: with missing data. We ®rst imputed values for all yÅ21Ϫ yÅ ϭ (a 2ϩ xÅ 221b ) Ϫ (a ϩ xÅ 11b ) variables using the expectation maximization (EM) algorithm (Allison, 2001). This procedure ϭ a Ϫ a ϩ xÅ (b Ϫ b ) 21121 has been shown to generate accurate maximum

ϩ b12(xÅ Ϫ xÅ 1) ϩ (xÅ 2Ϫ xÅ 1)(b 2Ϫ b 1) likelihood estimates with up to 50% missing data. If the amount of missing data is large, however, ϭ⌬a ϩ xÅ11⌬b ϩ b ⌬xÅ ϩ⌬xÅ⌬b. EM may underestimate standard errors, because it Marital Quality: 1980±2000 9 does not take into account variability (error) in- working extended hours (from .11 to .21). Our troduced by the imputation procedure itself. The measure of job demands (irregular hours, shift amount of missing data in the present study, how- work, evenings, overnight trips) increased for both ever, was small. As shown in Table 1, most vari- sexes, although the increase was signi®cant only ables were missing less than 1% of observations, for wives. The means for job demands in Table 1, with the largest percentages occurring for total however, were based on all individuals, irrespec- family income and wives' proportion of family tive of employment status. When we restricted the income. Moreover, preliminary analysis using comparisons to employed individuals, husbands' multiple imputation of missing data yielded results job demands increased from 1.99 to 2.12 (p ϭ that were virtually identical to those based on the .006), and wives' job demands increased from .95 EM algorithm. Consequently, we relied on EM to 1.16 (p Ͻ .001). Family income increased be- methods for all analyses reported here. tween 1980 and 2000 by over $14,000 (in con- stant 2000 dollars). In addition, the proportion of respondents who had used public assistance with- RESULTS in the previous 3 years declined signi®cantly, as did the proportion of respondents who viewed Explanatory Variables by Decade their economic situation as getting worse. In gen- Table 1 shows the means for all variables by de- eral, these results indicate an increase in married cade (1980 vs. 2000). Consistent with national couples' economic resources. data, age at current marriage was nearly 3 years Respondents reported signi®cant increases in higher in 2000 than 1980. (Although not shown the wife's percentage of income and in the hus- in the table, age at ®rst marriage also increased band's share of housework between 1980 and signi®cantly.) The means for years married, how- 2000. Correspondingly, respondents were signi®- ever, were virtually identical across decades. The cantly more likely to report equal decision mak- proportion of remarriages (for either spouse) in- ing, and gender attitudes became signi®cantly less creased signi®cantly, from .20 to .29. The decline traditional, indicating a shift toward more egali- in the proportion of marriages with either pre- tarian marital relations. Contrary to our expecta- school children or school-age children was too tions, scores on the scale of support for lifelong small to be signi®cantÐa result that re¯ects the marriage increased signi®cantly, re¯ecting less ac- relatively small drop in marital fertility at the na- ceptance of divorce. Respondents also described tional level. The proportion of Blacks, Hispanics, themselves as slightly more religious in 2000 than and other groups (Asian, Native American, and in 1980. mixed-race individuals) increased, although the Although not shown in Table 1, we also ex- difference for Blacks fell short of signi®cance. amined the 1980 and 2000 means separately for These racial and ethnic trends are consistent with husbands and wives. In general, trends for hus- national trends (described earlier) in the racial bands and wives were similar, and a signi®cant composition of the married population. In addi- interaction between sex and decade emerged for tion, the heterogamy index (based on religion, only one variable. The reported increase in the age, race, and education) increased signi®cantly, proportion of housework done by husbands was re¯ecting an increase in heterogamous marriages greater among husbands (from .29 to .39) than over time. (A supplementary analysis in which we wives (from .24 to .30). The increase was signif- assessed each form of heterogamy separately re- icant, however, irrespective of the respondents' vealed that heterogamy based on age and race in- sex. creased signi®cantly over time, whereas heterog- amy based on religion and education did not Marital Quality by Decade change.) More dramatically, the proportion of couples who cohabited prior to marriage increased Table 1 also reports the means for the three marital signi®cantly, from .16 to .41. quality variables. Despite the large number of Also consistent with national trends, the data changes that occurred between 1980 and 2000, revealed increases in years of education (about 1 mean levels of marital happiness were nearly year) between 1980 and 2000. The proportion of identical in the two periods. Similarly, little husbands in the paid labor force declined slightly, change occurred in divorce proneness. Marital in- from .95 to .92, whereas the proportion of wives teraction, in contrast, declined signi®cantly, by in the paid labor force increased, especially wives about one third of a standard deviation. Although 10 Journal of Marriage and Family

TABLE 2. DECOMPOSITION ANALYSIS OF 1980 TO 2000 CHANGE IN MARITAL HAPPINESS WITH MEAN CENTERED EXPLANATORY VARIABLES

(1) (2) (3) (4) (5)

b1 1980 b2 2000 b2 Ϫ b1 xÅ1 1980 xÅ2 2000 Demographic variables Wife respondent Ϫ.258** Ϫ.114* .143* .000 .000 Age married Ϫ.009 .000 .009 Ϫ1.503 1.456 Years married Ϫ.040** Ϫ.019 .020 .002 Ϫ.002 (Years married)2 .001** .000 Ϫ.001 5.130 Ϫ4.969 Remarried .142 Ϫ.053 Ϫ.195 Ϫ.046 .044 Preschool children Ϫ.132* Ϫ.227* Ϫ.095 .010 Ϫ.010 School-age children Ϫ.068 Ϫ.119* Ϫ.050 .007 Ϫ.007 Black Ϫ.152 Ϫ.158 Ϫ.006 Ϫ.009 .009 Hispanic .003 Ϫ.148 Ϫ.151 Ϫ.028 .027 Other non-White Ϫ.134 .075 .209 Ϫ.021 .021 Heterogamy index Ϫ.129* Ϫ.077** .052 Ϫ.068 .066 Cohabitation Ϫ.015 Ϫ.122* Ϫ.108 Ϫ.124 .120 Employment and income Education Ϫ.018 Ϫ.006 .012 Ϫ.487 .472 Husband employed Ϫ.179 Ϫ.095 .084 .011 Ϫ.010 Wife employed part-time Ϫ.077 .056 .133 Ϫ.031 .030 Wife employed full-time Ϫ.072 Ϫ.200* Ϫ.128 Ϫ.010 .010 Wife extended hours Ϫ.209 Ϫ.109 .099 Ϫ.049 .047 Husband job demands .000 Ϫ.014 Ϫ.015 Ϫ.011 .011 Wife job demands Ϫ.027 Ϫ.055 Ϫ.028 Ϫ.112 .108 Family income .004** .001 Ϫ.003* Ϫ7.317 7.088 Public assistance Ϫ.183* Ϫ.104 .079 .020 Ϫ.019 Finances better .043 .218** .175* Ϫ.014 .014 Finances worse Ϫ.055 Ϫ.122 Ϫ.067 .034 Ϫ.033 Gender relations Wife proportion income Ϫ.073 .095 .169 Ϫ.056 .055 Husband housework .398** .248 Ϫ.150 Ϫ.042 .041 Equal decision making .352** .288** Ϫ.064 Ϫ.075 .073 Attitudes and values Traditional gender attitudes Ϫ.063 Ϫ.203** Ϫ.140 .083 Ϫ.081 Lifelong marriage .398** .202** Ϫ.196* Ϫ.053 .051 Religiosity .081** .047* Ϫ.034 Ϫ0.85 .082 Constant (a) .058 Ϫ.042 R2 .145** .101** Sum of components Decomposition: yÅ Ϫ yÅ ϭ⌬a ϩ xÅ ⌬b ϩ⌬xÅb ϩ⌬xÅ⌬b 21͸͸͸ 1 1 Ϫ.002 Ϫ.100** .015 .113** Ϫ.030

Note: Sample size is 2,034 in 1980 and 2,100 in 2000. *p Ͻ .05. **p Ͻ .01. (two-tailed). not shown in Table 1, we also examined these level of divorce proneness changed little. The changes separately for husbands and wives. mean level of marital happiness decreased slightly Wives, compared with husbands, reported signif- for husbands and increased slightly for wives, but icantly less happiness, less interaction, and more neither change was signi®cant. divorce proneness in both periods. Nevertheless, the 1980±2000 differences for marital interaction Explaining Continuity and Change and divorce proneness were similar for both sexes: For wives as well as husbands, the mean level of The results thus far indicate that between 1980 interaction declined signi®cantly, and the mean and 2000, U.S. marriages changed substantially Marital Quality: 1980±2000 11

TABLE 2. EXTENDED and 2000 into components re¯ecting (a) changes in the intercepts, (b) changes in the levels (means) Decomposition of the explanatory variables, (c) changes in the slopes (b coef®cients) of the explanatory vari- (6) (7) (8) (9) Slopes Levels Joint Sum of ables, and (d) joint changes in the levels and

xÅ1 ⌬b ⌬xÅb1 ⌬xÅ⌬b components slopes of the explanatory variables. Because it is dif®cult to calculate standard errors for these com- ponents with weighted data, we used the bootstrap .000 .000 .000 .000 Ϫ.013 Ϫ.025 .025 Ϫ.013 procedure in the Stata survey module (with 500 .000 .000 .000 .000 replications) to estimate standard errors for sig- Ϫ.003 Ϫ.010 .006 Ϫ.007 ni®cance testing. .009 .013 Ϫ.018 .004 Table 2 provides details of the decomposition Ϫ.001 .003 .002 .004 analysis for marital happiness. Column 1 shows .000 .001 .001 .001 .000 Ϫ.003 .000 Ϫ.003 the regression equation (unstandardized b coef®- .004 .000 Ϫ.008 Ϫ.004 cients) for marital happiness in 1980. Most of the Ϫ.004 Ϫ.006 .009 Ϫ.001 signi®cant associations are consistent with the Ϫ.004 Ϫ.017** .007 Ϫ.014** general literature on marital quality (Bradbury, .013 Ϫ.004 Ϫ.026 Ϫ.016 Fincham, & Beach, 2000). Wives reported less happiness than husbands. Years married revealed Ϫ.006 Ϫ.017 .012 Ϫ.011 a curvilinear trend, with happiness declining in the .001 .004 Ϫ.002 .003 Ϫ.004 Ϫ.005 .008 Ϫ.001 early years of marriage and reaching a plateau in .001 Ϫ.001 Ϫ.003 Ϫ.003 the later years. In addition, marital happiness was Ϫ.005 Ϫ.020 .010 Ϫ.015* greater when no preschool children lived in the .000 .000 .000 .000 household, unions were homogamous (rather than .003 Ϫ.006 Ϫ.006 Ϫ.009 heterogamous), family income was high, families .024* .052** Ϫ.046* .029** .002 .007 Ϫ.003 .006 did not rely on public assistance, husbands did Ϫ.002 .001 .005 .004 more housework, decision making was equal, re- Ϫ.002 .004 .004 .006 spondents supported the norm of lifelong mar- riage, and respondents reported a high level of Ϫ.010 Ϫ.008 .019 .001 religious in¯uence. .006 .033** Ϫ.012 .027** The regression equation for the 2000 sample, .005 .052** Ϫ.010 .048** shown in column 2, is similar to the equation for the 1980 sample, although some variables that at- Ϫ.012 .010 .023 .022** tained signi®cance in 1980 were not signi®cant in .010* .041** Ϫ.020* .031** 2000, and vice versa. Most of the differences in .003 .014* Ϫ.006 .011** slopes, however, were not statistically signi®cant. Column 3, which shows the difference in slopes

(b2 Ϫ b1), indicates four signi®cant differences. .015 .113** Ϫ.030 .098** First, wives were less happy than husbands in both surveys, but the gap between husbands and wives grew smaller between 1980 and 2000. The nar- rowing of the gender gap in happiness may be attributable to a shift toward more egalitarian mar- ital relations, as described earlier. In addition, the with respect to a variety of demographic, econom- positive association between family income and ic, social, and cultural characteristics. With the ex- happiness became smaller, whereas the positive ception of a decline in marital interaction, how- association between perceiving one's ®nancial sit- ever, marriages did not change in terms of marital uation to be improving and happiness became quality. Why did marital happiness and divorce larger. The explanation for this pattern is not clear, proneness remain essentially constant when so but given that both variables are indicators of eco- many other aspects of marriage changed? nomic resources, it does not appear that economic To answer this question, we turn to the decom- factors, in general, became more or less important position analyses. As described earlier, we decom- to marital happiness. Finally, the b coef®cient re- posed the change in marital quality between 1980 lating support for lifelong marriage to marital hap- 12 Journal of Marriage and Family

TABLE 3. DECOMPOSITION ANALYSIS OF 1980 TO 2000 CHANGE IN MARITAL INTERACTION WITH MEAN CENTERED EXPLANATORY VARIABLES

(1) (2) (3) (4) (5)

b1 1980 b2 2000 b2 Ϫ b1 xÅ1 1980 xÅ2 2000 Demographic variables Wife respondent Ϫ.024 Ϫ.052 Ϫ.028 .00 .000 Age married .003 Ϫ.002 Ϫ.005 Ϫ1.503 1.456 Years married Ϫ.043** Ϫ.040** .003 .002 Ϫ.002 (Years married)2 .001* .001** .000 5.130 Ϫ4.969 Remarried .107 Ϫ.004 Ϫ.111 Ϫ.046 .044 Preschool children Ϫ.177* Ϫ.223** Ϫ.046 .010 Ϫ.010 School-age children Ϫ.033 Ϫ.148** Ϫ.115 .007 Ϫ.001 Black Ϫ.343** Ϫ.266** .076 Ϫ.009 .009 Hispanic .070 .190 .120 Ϫ.028 .028 Other non-White .291* .262** Ϫ.029 Ϫ.021 .021 Heterogamy index Ϫ.142** Ϫ.066* .076 Ϫ.068 .066 Cohabitation Ϫ.231** Ϫ.059* .173 Ϫ.124 .120 Employment and income Education Ϫ.025* Ϫ.011 .014 Ϫ.487 .472 Husband employed Ϫ.167 .048 .214 .011 Ϫ.011 Wife employed part-time Ϫ.072 Ϫ.171* Ϫ.099 Ϫ.031 .040 Wife employed full-time Ϫ.143* Ϫ.360** Ϫ.218* .010 .010 Wife extended hours Ϫ.173 Ϫ.388** Ϫ.215 Ϫ.049 .047 Husband job demands Ϫ.025 Ϫ.041 Ϫ.016 Ϫ.011 .011 Wife job demands Ϫ.082* Ϫ.041 .041 Ϫ.112 .108 Family income .002 .001 Ϫ.001 Ϫ7.318 7.088 Public assistance Ϫ.120 Ϫ.286** Ϫ.166 .020 Ϫ.019 Finances better Ϫ.009 .185** .194* Ϫ.014 .014 Finances worse Ϫ.084 Ϫ.138 Ϫ.054 .034 Ϫ.033 Gender relations Wife proportion income .123 .280 .158 Ϫ.057 .055 Husband housework .802** .751** Ϫ.051 Ϫ.042 .041 Equal decision making .310** .146** Ϫ.164* Ϫ.076 .073 Attitudes and values Traditional gender attitudes Ϫ.094 Ϫ.123 Ϫ.029 .084 Ϫ.081 Lifelong marriage .174** .053 Ϫ.120 Ϫ.053 .051 Religiosity .020 .022 .002 Ϫ.085 .083 Constant (a) .200 Ϫ.197 R2 .134** .105** Sum of components Decomposition: yÅ Ϫ yÅ ϭ⌬a ϩ xÅ ⌬b ϩ⌬xÅb ϩ⌬xÅ⌬b 21͸͸͸ 1 1 Ϫ.315** Ϫ.397** Ϫ.004 .079* .007

Note: Sample size is 2,034 in 1980 and 2,100 in 2000. *p Ͻ .05. **p Ͻ .01. (two-tailed). piness declined by about half. Although the ex- shows the extent to which changes in marital hap- planation for this decline is unclear, this variable piness can be attributed to changes in the slopes continued to be associated positively and signi®- (effects) of the independent variables, assuming cantly with marital happiness in 2000. no change in levels (means). Column 7 shows the Columns 6 through 9 in Table 2 show the re- extent to which change in marital happiness can sults of the analysis in which we decomposed the be attributed to changes in the levels (means) of mean difference in marital happiness between de- the independent variables, assuming no change in cades (which was virtually zero) into components slopes (effects). Column 8 takes into account the based on changes in means and slopes. Column 6 joint effect of changes in levels as well as slopes. Marital Quality: 1980±2000 13

TABLE 3. EXTENDED 1), the increase in heterogamy implied a decline in happiness, as shown in column 7. The effect of Decomposition heterogamy declined slightlyÐbut not signi®cant- lyÐduring this period (column 3), and this minor (6) (7) (8) (9) Slopes Levels Joint Sum of change is re¯ected in a weak slope component in

xÅ1 ⌬b ⌬xÅb1 ⌬xÅ⌬b components column 6. The joint effect of changing slopes and changing means also was small (column 8). The overall contribution of heterogamy (column 9) .000 .000 .000 .000 .007 .008 Ϫ.014 .001 was negative and signi®cant. In other words, .000 .000 .000 .000 changes in the levels and effects of heterogamy .000 Ϫ.010 .001 Ϫ.010 between 1980 and 2000 accounted for a decline .005 .010 Ϫ.010 .005 of 0.014 of a standard deviation in the aggregate .000 .004 .001 .004 level of marital happiness in the population. (Fol- Ϫ.001 .000 .002 .001 Ϫ.001 Ϫ.006 .001 Ϫ.006 lowing Cohen [1988], we express all changes in Ϫ.003 .004 .007 .007 marital quality as effect sizes, that is, as propor- .001 .012 Ϫ.001 .012** tions of a standard deviation.) Similarly, the Ϫ.005 Ϫ.019* .010 Ϫ.014** change in wives' extended hours of employment Ϫ.021* Ϫ.056** .042 Ϫ.036** was associated with a decline of 0.015 of a stan- dard deviation in the level of marital happiness in Ϫ.007 Ϫ.024* .014 Ϫ.017* the population between 1980 and 2000. .002 .004 Ϫ.005 .001 .003 Ϫ.004 Ϫ.006 Ϫ.007 Changes in other variables appear to have had .002 Ϫ.003 Ϫ.004 Ϫ.005 positive implications for marital happiness. Ex- .010 Ϫ.017 Ϫ.021 Ϫ.027** pressed in standard deviation units, the increase in .000 Ϫ.001 .000 Ϫ.001 family income was associated with a rise of 0.029, Ϫ.005 Ϫ.018* .009 Ϫ.014* the increase in husbands' share of housework was .008 .023 Ϫ.015 .015 Ϫ.003 .005 .006 .008* associated with a rise of 0.027, the increase in Ϫ.003 .000 .005 .002 decision-making equality was associated with a Ϫ.002 .006 .004 .007 rise of 0.048, the decline in traditional gender at- titudes was associated with a rise of 0.022, the Ϫ.009 .014 .018 .022 growth in support for lifelong marriage was as- .002 .066** Ϫ.004 .064** sociated with a rise of 0.031, and the increase in .012* .046** Ϫ.024* .034** religiosity was associated with a rise of 0.011. Al- though not shown in the table, changes in these Ϫ.002 .015 .005 .018* six variables (collectively) were associated with .006 .018* Ϫ.012 .012* an increase of 0.17 of a standard deviation in the .000 .003 .000 .003 mean level of marital happiness in the population. The summary decomposition equation is shown at the bottom of Table 2. The difference Ϫ.004 .079* .007 .082** between intercepts (Ϫ.10) was negative and sig- ni®cant. This ®gure can be interpreted as the un- standardized b coef®cient for decade (2000 vs. 1980) in a model that includes all of the explan- atory variables. In other words, if the explanatory variables had not changed, then marital happiness Finally, column 9 shows the sum of the compo- would have declined by one tenth of a standard nents from columns 6, 7, and 8. The values in this deviation. This decline indicates that one or more ®nal column can be interpreted as the overall con- unmeasured variables lowered marital happiness tribution of change in each explanatory variable during this period. The levels component also was to the change in marital happiness. signi®cant, indicating that changes in the means To facilitate interpretation of these results, con- of all of the explanatory variables collectively sider the role of heterogamy. The mean for this raised happiness by about 0.11 of a standard de- variable increased between 1980 and 2000 (col- viation. The slope and joint components were not umns 4 and 5), and because heterogamy was neg- signi®cant. These results indicate that the mean atively associated with happiness in 1980 (column level of marital happiness in the population would 14 Journal of Marriage and Family

TABLE 4. DECOMPOSITION ANALYSIS OF 1980 TO 2000 CHANGE IN DIVORCE PRONENESS WITH MEAN CENTERED EXPLANATORY VARIABLES

(1) (2) (3) (4) (5)

b1 1980 b2 2000 b2 Ϫ b1 xÅ1 1980 xÅ2 2000 Demographic variables Wife respondent .038 .114* .077 .000 .000 Age married Ϫ.019** Ϫ.026** Ϫ.006 Ϫ1.503 1.456 Years married .048** .020* Ϫ.028 .002 Ϫ.002 (Years married)2 Ϫ.002** Ϫ.001** .001 5.130 Ϫ4.969 Remarried .174 .261** .088 Ϫ.046 .044 Preschool children .156* .112 Ϫ.044 .010 Ϫ.010 School-age children Ϫ.071 .008 .079 .007 Ϫ.007 Black .042 .046 .003 Ϫ.009 .009 Hispanic .142 .035 Ϫ.107 Ϫ.028 .028 Other non-White Ϫ.282 Ϫ.222** .060 Ϫ.021 .021 Heterogamy index .075** .026 Ϫ.049 Ϫ.068 .066 Cohabitation .325** .257** Ϫ.068 Ϫ.124 .120 Employment and income Education .020 .024** .004 Ϫ.487 .472 Husband employed .009 .072 .063 .011 Ϫ.011 Wife employed part-time .136 Ϫ.007 Ϫ.143 Ϫ.031 .030 Wife employed full-time .171 .088 Ϫ.083 Ϫ.010 .010 Wife extended hours .228* .208* Ϫ.020 Ϫ.049 .047 Husband job demands .023 .035 .011 Ϫ.011 .011 Wife job demands .063 .079** .016 Ϫ.112 .108 Family income .001** Ϫ.001 Ϫ.001 Ϫ7.318 7.088 Public assistance .222* .383** .161 .020 Ϫ.019 Finances better .037 Ϫ.130* Ϫ.166* Ϫ.014 .014 Finances worse .183* .326** .142 .034 Ϫ.033 Gender relations Wife proportion income .036 Ϫ.208 Ϫ.244 Ϫ.057 .055 Husband housework Ϫ.400** Ϫ.049 .351 Ϫ.042 .041 Equal decision making Ϫ.259** Ϫ.254** .005 Ϫ.076 .073 Attitudes and values Traditional gender attitudes Ϫ.106 Ϫ.037 .069 .084 Ϫ.081 Lifelong marriage Ϫ.283** Ϫ.104 .179* Ϫ.053 .051 Religiosity Ϫ.019** .012 .031 Ϫ.085 .083 Constant (a) .018 .016 R2 .101** .117** Sum of components Decomposition: yÅ Ϫ yÅ ϭ⌬a ϩ xÅ ⌬b ϩ⌬xÅb ϩ⌬xÅ⌬b 21͸͸͸ 1 1 Ϫ.012 Ϫ.002 .035 .023 Ϫ.068

Note: Sample size is 2,034 in 1980 and 2,100 in 2000. *p Ͻ .05. **p Ͻ .01. (two-tailed). have declined between 1980 and 2000 (for un- attained signi®cance are consistent with the re- known reasons), but changes in the means of the search literature on marital quality. Moreover, col- explanatory variables offset this decline, resulting umn 3 reveals that the estimated effects of most in no net change. variables did not differ signi®cantly across sam- The analysis for marital interaction appears in ples. The decomposition analysis (especially col- Table 3. Columns 1 and 2 show the regression umn 9) indicates that ®ve variables were associ- equations for 1980 and 2000, respectively. With ated with declines in the level of marital the exception of education (which was negatively interaction between 1980 and 2000: the increase related to interaction in 1980), the variables that in heterogamy, the increase in cohabitation prior Marital Quality: 1980±2000 15

TABLE 4. EXTENDED ples using public assistance, the increase in hus- bands' share of housework, the increase in deci- Decomposition sion-making equality, the decline in traditional gender attitudes, and the increase in support for the (6) (7) (8) (9) Slopes Levels Joint Sum of norm of lifelong marriage. Taken together, these six

xÅ1 ⌬b ⌬xÅb1 ⌬xÅ⌬b components variables were associated with an increase of 0.15 of a standard deviation in the mean level of marital interaction in the population. Overall, these results, .000 .000 .000 .000 .009 Ϫ.057* Ϫ.019 Ϫ.066** combined with the summary decomposition equa- .000 .000 .000 .000 tion at the bottom of Table 3, indicate that inter- .004 .015 Ϫ.008 .011 action declined for largely unexplained reasons, but Ϫ.004 .016* .008 .020** the decline would have been greater if other off- .000 Ϫ.003 .001 Ϫ.003 setting changes had not occurred. .001 .001 Ϫ.001 .000 .000 .001 .000 .001 Table 4 shows the analysis for divorce prone- .003 .008 Ϫ.006 .005 ness. With the exception of education (which was Ϫ.001 Ϫ.012 .002 Ϫ.011* positively associated with divorce proneness), the .003 .010* Ϫ.006 .007 results of the regression analyses for 1980 and .008 .079** Ϫ.016 .071** 2000 are generally consistent with the literature on marital quality. Moreover, the slopes relating Ϫ.002 .019 .003 .021** the explanatory variables to divorce proneness ex- .001 .000 Ϫ.001 Ϫ.001 .004 .008 Ϫ.009 .004 hibited little change between decades. The decom- .001 .003 Ϫ.002 .003 position analysis (especially column 9) revealed .001 .022* Ϫ.002 .021** that changes in ®ve explanatory variables were .000 .001 .000 .001 associated with signi®cant increases in divorce Ϫ.002 .014 .004 .016** proneness: the increase in the proportion of re- .006 .001 Ϫ.011 Ϫ.004 .003 Ϫ.009* Ϫ.006 Ϫ.012** married individuals, the increase in cohabitation .002 .001 Ϫ.005 Ϫ.001 prior to marriage, the increase in education, the .005 Ϫ.012* Ϫ.009 Ϫ.017** increase in wives' extended hours of employment, and the increase in wives' job demands. Taken .014 .004 Ϫ.027 Ϫ.009 together, these ®ve variables accounted for a rise Ϫ.015 Ϫ.033** .029 Ϫ.019* of 0.15 of a standard deviation in the mean level .000 Ϫ.039** .001 Ϫ.038** of divorce proneness in the population. The anal- ysis also revealed several explanatory variables .006 .017 Ϫ.011 .012 that offset this trend: the increase in age at mar- Ϫ.009 Ϫ.029* .019 Ϫ.020** riage, the increase in the proportion of other non- Ϫ.003 Ϫ.003 .005 Ϫ.001 White individuals, the decline in the use of public assistance, the decline in the perception that one's ®nancial situation is getting worse, the increase in .035 .023 Ϫ.068 Ϫ.010 husbands' share of housework, the increase in de- cision-making equality, and the increase in sup- port for lifelong marriage. Overall, changes in these variables were associated with a decrease of 0.18 of a standard deviation in the level of divorce proneness in the population. In summary, the data to marriage, the increase in education, the increase suggest that continuity in the mean level of di- in the proportion of wives working extended vorce proneness between 1980 and 2000 con- hours, and the increase in wives' job demands. cealed two opposing trends, with some social Taken together, changes in these ®ve variables forces increasing divorce proneness and other so- were associated with a decline of 0.11 of a stan- cial forces decreasing divorce proneness. dard deviation in the mean level of interaction in the population. Six variables were associated with Summary increases in interaction: the rise in the proportion Table 5 presents a summary of the decomposition of other non-White individuals in the married results. In most cases, the direction of in¯uence population, the decrease in the proportion of cou- was consistent with our expectations. For exam- 16 Journal of Marriage and Family

TABLE 5. SUMMARY OF SIGNIFICANT ASSOCIATIONS BETWEEN CHANGES IN EXPLANATORY VARIABLES AND CHANGES IN DIMENSIONS OF MARITAL QUALITY: 1980±2000 Association With Marital Quality Direction of Change Marital Marital Divorce Explanatory Variable 1980±2000 Happiness Interaction Proneness Demographic variables Age married Increase ns ns Ϫ Years married No change ns ns ns Remarried Increase ns ns ϩ Preschool children No change ns ns ns School-age children No change ns ns ns Black No change ns ns ns Hispanic Increase ns ns ns Other non-White Increase ns ϩϪ Heterogamy index Increase ϪϪns Cohabitation Increase ns Ϫϩ Employment and income Education Increase ns Ϫϩ Husband employed Decrease ns ns ns Wife employed part-time Increase ns ns ns Wife employed full-time No change ns ns ns Wife extended hours Increase ϪϪϩ Husband job demands No change ns ns ns Wife job demands Increase ns Ϫϩ Family income Increase ϩ ns ns Public assistance Decrease ns ϩϪ Finances better No change ns ns ns Finances worse Decrease ns ns Ϫ Gender arrangements Wife proportion income Increase ns ns ns Husband housework Increase ϩϩϪ Equal decision making Increase ϩϩϪ Attitudes and values Traditional gender attitudes Decrease ϩϩns Lifelong marriage Increase ϩϩϪ Religiosity Increase ϩ ns ns Note: Column 2 indicates whether the explanatory variable increased or decreased between 1980 and 2000. In columns 3, 4, and 5, ϩ indicates a signi®cance increase in marital quality, Ϫ indicates a signi®cant decrease in marital quality, and ns indicates no signi®cant change in marital quality, based on last column (sum of components) of Tables 2, 3, and 4.

ple, the mean age at marriage (which increased did not conform to our expectations. We assumed between 1980 and 2000) and decision-making that increases in education would improve marital equality (which increased between 1980 and quality, but changes in this variable were associ- 2000) were associated with a signi®cant decline ated with less marital interaction and greater di- in divorce proneness. Similarly, the increase in vorce proneness. Although speculative, it is pos- heterogamy was associated with lower marital sible that the increase in education raised people's happiness and lower marital interaction. (A more expectations for marriage, perhaps unrealistically, detailed analysis of the heterogamy variable, in with negative consequences for marital interaction which the four forms were disaggregated, re- and divorce proneness. The table also indicates vealed that all forms of heterogamy were related that changes in wives' extended hours of employ- to at least one dimension of marital quality. How- ment, husbands' share of housework, decision- ever, because religious heterogamy and education making equality, and support for the norm of life- heterogamy did not change over time, age heter- long marriage were the most consistent predictors ogamy and racial heterogamy were driving these of changes in marital quality, yielding signi®cant ®ndings.) Education was the main variable that associations with each of the three dimensions. Marital Quality: 1980±2000 17

Sex Differences years. In other words, as the husband's share of housework increased, husbands reported less hap- To assess sex differences, we conducted decom- piness, and wives reported more happiness. Col- position analyses similar to those described in Ta- umns 4 and 5 reveal that the means for this var- bles 2 through 4 separately for husbands and iable increased between 1980 and 2000, wives. We then conducted tests to see if the over- regardless of whether husbands or wives were the all contributions of the explanatory variables to respondents. Correspondingly, column 9 reveals marital quality (slopes ϩ levels ϩ joint effects) that the increase in the husband's share of house- differed signi®cantly for husbands and wives. One work was associated with a decline in marital hap- explanatory variable, decision-making equality, piness among husbands equivalent to 0.07 of a yielded signi®cant sex differences for marital hap- standard deviation. For wives, in contrast, the in- piness (p Ͻ .01), interaction (p Ͻ .05), and di- crease in shared housework was associated with vorce proneness (p Ͻ .001). A second explana- an increase of 0.055 of a standard deviation. The tory variable, the husband's share of housework, results for divorce proneness yielded a similar pat- yielded signi®cant sex differences for marital hap- tern. For men, the husband's share of housework piness (p Ͻ .001) and divorce proneness (p Ͻ was positively and signi®cantly associated with .001). Because these were the only two explana- divorce proneness, at least in 2000. For wives, tory variables for which sex differences appeared however, this variable was negatively and signif- across more than one dimension of marital quality, icantly associated with divorce proneness in both we focus on these results. periods. The ®nal column reveals that the increase Table 6 shows elements of the decomposition in the husband's share of housework was associ- analysis (not the entire analysis) separately for ated with a signi®cant rise in divorce proneness husbands and wives. With respect to decision- among men and a signi®cant decline in divorce making equality, the b coef®cients in columns 1 proneness among women. Of all explanatory var- and 2 reveal that equality was positively associ- iables, the increase in the husband's share of ated with marital happiness and interaction and housework was the only one that had substantially negatively associated with divorce proneness, ir- different implications for men and women. (Al- respective of the respondent's sex, although the though not shown in the table, the results for mar- coef®cients were larger for wives than for hus- ital interaction did not differ for husbands and bands. Columns 4 and 5 indicate that husbands as wives. Regardless of sex, the increase in the hus- well as wives reported more equality in 2000 than band's share of housework was associated with an 1980, although the rise was somewhat larger increase in marital interaction.) among wives than husbands. The ®nal column shows that the rise in decision-making equality was associated with increases in marital happi- DISCUSSION ness, increases in marital interaction, and decreas- Our study was organized around four goals. First, es in divorce proneness for both sexes, but the we used surveys conducted in 1980 and 2000 to associations were stronger for wives than for hus- document how the population of married individ- bands. For example, the increase in equality was uals in the United States changed during the last associated with a rise of about 0.02 of a standard two decades of the 20th century. Our data dem- deviation in marital happiness among husbands, onstrate a substantial degree of change. Compared whereas the corresponding rise was 0.08 of a stan- with married people in 1980, couples in 2000 had dard deviation among wives. These results sug- married at older ages, were more likely to be in gest, therefore, that increases in decision-making second or higher-order marriages, were more like- equality between 1980 and 2000 improved marital ly to be Hispanics or other non-Whites, were more quality for respondents of both sexes, but more so likely to be in heterogamous marriages, were for wives than for husbands. more likely to have cohabited prior to marriage, The bottom half of Table 6 shows sex differ- were better educated, and had more economic re- ences with respect to the husband's share of sources. In addition, wives were more likely to be housework. For husbands, the b coef®cients for employed, wives contributed a greater proportion marital happiness were negative in 1980 and of family income, husbands and wives faced 2000, although only the latter coef®cient was sig- greater job demands, husbands performed a great- ni®cant. In contrast, the corresponding coef®cients er share of housework, and family decision mak- for wives were positive and signi®cant in both ing was more equal. With respect to attitudes, in- 18 Journal of Marriage and Family ETWEEN .051* .039* .067** .055** .017** .063** B .017* .080** .021** .050** (9) Ϫ Ϫ Ϫ Ϫ Sum of components IFFERENCES :D b ⌬ Å .071 .001 .042 .015 .009 .014 .001 .018 .002 .056** x (8) Joint ⌬ Ϫ Ϫ Ϫ Ϫ Ϫ Ϫ UALITY Q ARITAL M Decomposition 1 Åb x .009 .039* .031 .056* .011 .069* .017 .088** .020 .075* (7) ⌬ Ϫ Ϫ Ϫ Ϫ Ϫ Levels HANGE IN b ⌬ 1980±2000 C .011 .001 .000 .010 .007 .016 .001 .031** .004 .008 (6) 1 Å x Ϫ Ϫ Ϫ Ϫ Slopes IVES OUSEWORK TO W 2000 .085 .004 .068 .078 .085 .004 .068 .078 .068 .078 H (5) 2 Ϫ Ϫ Å x HARE OF 'S 1980 .015 .068 .054 .097 .015 .068 .054 .097 .054 .097 USBANDS AND (4) 1 H Ϫ Ϫ Ϫ Ϫ Ϫ Ϫ Ϫ Ϫ Ϫ Ϫ Å x USBANDS H 1 b Ϫ .707* .013 .009 .100 .422 .240 .012 .320 .077 .081 (3) 2 Ϫ Ϫ Ϫ Ϫ Ϫ Ϫ b QUALITY AND E 2000 .726** .616** .617** .132 .400** .178* .108 .171* .313** (2) AKING 1.119** 2 Ϫ Ϫ Ϫ Ϫ b -M ECISION D 1980 .304 .879** .092 .604** .142* .500** .165** .427** .094 .394** (1) 1 Ϫ Ϫ Ϫ Ϫ Ϫ b .01. (two-tailed). ONTRIBUTION OF Ͻ C p HE 6. T .05. ** Ͻ Husbands Wives Husbands Wives Husbands Wives Husbands Wives Husbands Wives p ABLE Divorce proneness Marital happiness Marital happiness Marital interaction Divorce proneness Note: Sample size is* 2,034 in 1980 and 2,100 in 2000. T Husband's share housework Decision-making equality Marital Quality: 1980±2000 19 dividuals were less traditional in their views about not able to explain the overall decline in marital gender, were more supportive of the norm of life- interaction between 1980 and 2000 with the var- long marriage, and reported more religious in¯u- iables in our analysis. But irrespective of the ex- ence. Overall, the married population looked quite planation, it appears that the lives of husbands and different at the beginning of the new millennium wives are becoming more separate, with couples than it had looked two decades earlier. being less likely to share activities such as eating Our second goal was to show how three central meals, shopping, working on projects around the dimensions of marital qualityÐmarital happiness, house, visiting friends, and going out for leisure marital interaction, and divorce pronenessÐ activities. The decrease in marital interaction may changed during this period. Despite substantial be due to couples having nonoverlapping work shifts in the composition of the married popula- schedules, a growing service sector that makes it tion, the means for marital happiness and divorce possible for spouses to be more independent of proneness were nearly identical in 1980 and 2000. one another, and the increasing demands on par- Marital interaction, in contrast, declined signi®- ents' time posed by children's after-school activ- cantly. Large shifts in demographic, economic, so- ities. cial, and attitudinal characteristics of the married The fourth goal of our study was to show how population, coupled with relatively little change in these trends differed for husbands and wives. marital quality, represents a puzzle. Why was the Across a large number of explanatory variables, aggregate level of marital happiness and divorce only two had consequences that differed consis- proneness stable during this period? tently for men and women. The increase in deci- Answering this question was our third goal. We sion-making equality appears to have improved used decomposition analysis (Firebaugh, 1997) to marital quality more strongly among wives than understand how a variety of explanatory variables among husbands. Given that traditional power ar- may have contributed to changes in marital quality rangements favor husbands, it is not surprising between 1980 and 2000. The results of these anal- that wives have more to gain from a shift toward yses allow us to reject the possibility that marital greater equality. Nevertheless, decision-making quality was unresponsive to shifts occurring in the equality is associated with improvements in mar- larger society. Instead, it appears that the stability ital quality for husbands as well as wives, so the of marital happiness and divorce proneness during shift toward equality does not represent a zero- this time was attributable to a variety of positive sum game with respect to happiness and stability. and negative forces that largely offset one another. Only one trendÐhusbands' share of houseworkÐ For example, increases in heterogamy and wives' had implications that differed for husbands and long hours of employment appear to have lowered wives. The increase in the proportion of house- marital happiness, whereas increases in family in- work done by husbands appears to have improved come, decision-making equality, support for non- the marital quality of wives but to have eroded traditional gender relations, support for the norm the marital quality of husbands. This ®nding sup- of lifelong marriage, and religiosity appear to ports the view of many observers, especially those have increased marital happiness. Similarly, in- writing from a feminist perspective, that house- creases in the proportion of second and higher- work is a major source of contention between men order marriages, premarital cohabitation, wives' and women (Ferree, 1991; Thompson & Walker, long hours of employment, and wives' job de- 1991). mands appear to have raised divorce proneness, Readers should be aware of several limitations whereas increases in age at marriage, ®nancial sta- of our study. First, a decomposition analysis is not bility, decision-making equality, and support for the same as a causal analysis. Although a path the norm of lifelong marriage appear to have low- analysis estimates the direct and indirect effects of ered divorce proneness. the explanatory variables, a decomposition anal- Marital interaction, although showing an over- ysis estimates only the direct effects. This differ- all downward trend, also appears to have been af- ence has implications for how one interprets the fected by contradictory forces. Indeed, the decline role of some explanatory variables, such as re- in marital interaction would have been greater if marriage. Compared with ®rst marriages, remar- it had not been offset partially by changes in hus- riages are more heterogamous and more likely to bands' share of housework, decision-making be preceded by cohabitation. If one views remar- equality, nontraditional gender attitudes, and sup- riage as a cause of heterogamy and cohabitation, port for the norm of lifelong marriage. We were then the decomposition analysis omits the indirect 20 Journal of Marriage and Family effects of remarriage (through heterogamy and co- by differential selection out of marriage. Because habitation) on marital quality. In addition, our divorce rates reached a plateau in 1980, there was analysis was not able to consider reciprocal rela- more selection out of marriage between 1980 and tionships between the explanatory variables and 2000 than in the two decades preceding the ®rst marital quality. For example, although wives' long survey. With increased selection of people out of hours of employment may interfere with marital troubled marriages, one might expect remaining satisfaction, some wives in troubled marriages marriages to be of higher quality in 2000 than in may seek out longer hours of employment. Given 1980. If this is true, then the apparent stability in the large number of variables in the current anal- marital happiness and divorce proneness between ysis, testing speci®c causal models must await fu- surveys would represent a decline in the real level ture research reports. Nevertheless, decomposition of these variables over time. The 1980 and 2000 analysis is a useful approach for taking a ``®rst Current Population Surveys indicate that of all pass'' through the data to locate the most prom- ever-married adults, 13% were divorced or sepa- ising variables for further investigation (Fire- rated in 1980 compared with 19% in 2000. (These baugh, 1997). percentages exclude widows and widowers.) To Another limitation involves our use of single see if this increase of 6 percentage points in the items to measure some explanatory variables, divorced population affected our results, we such a religiosity and equality. Measurement error adopted two strategies. associated with single-item indicators generally The ®rst strategy involved limiting the analysis attenuates correlations and can lead researchers to to individuals in their ®rst 2 years of marriageÐ underestimate the strength of associations in the a time when selection out of marriage through di- population. Moreover, although our analysis used vorce is modest. In fact, only about 7% of mar- three dimensions of marital quality, it is possible riages end in divorce within the ®rst 2 years that different trends would have appeared if we (Bramlett & Mosher, 2001). For individuals mar- had relied on alternative dimensions of marital ried less than 2 years (n ϭ 200 in 1980, n ϭ 146 quality (such as marital con¯ict). The availability in 2000), the mean standardized differences be- of only two time points is another constraint. For tween 1980 and 2000 were .08 for happiness, example, it is possible that marital happiness rose Ϫ.31 for interaction, and Ϫ.01 for divorce prone- (or fell) after 1980 but returned to its earlier level ness. Only the mean difference for interaction was by 2000, resulting in the mistaken conclusion that signi®cant (p ϭ .005). These calculations indicate no change occurred. that changes in marital quality for individuals in We were not able to determine whether chang- marriages of short duration were similar to the es in marriage between 1980 and 2000 re¯ected changes observed in the full sample. differential selection into marriage, cohort re- Second, we estimated how marital quality placement, or period effects. For example, the would have changed if there had been no selection family income of married couples (in constant out of marriage through divorce in 1980 or 2000. dollars) increased signi®cantly during these years. The 1980 sample was the ®rst wave of a panel This increase might have occurred because low- study, with additional interviews occurring in income individuals were more reluctant to marry 1983, 1988, 1992, and 1997. In marriages that in the 1990s than in the 1970s (differential selec- ended in divorce during the study, it was possible tion into marriage), because more recent marriage to examine marital quality prior to dissolution. We cohorts are better educated than earlier marriage used data on happiness, interaction, and divorce cohorts (a cohort explanation), or because gains proneness (drawn from the interview prior to mar- in economic productivity during the 1990s in- ital disruption) from 305 individuals who divorced creased the income of all married couples, irre- between 1980 and 1997. Not surprisingly, married spective of cohort (a period explanation). Al- respondents who were about to divorce, compared though determining the ultimate cause of marital with those who remained continuously married, change was beyond the scope of the current arti- reported lower marital happiness, less interaction, cle, the major ®ndings of our analysis do not de- and greater divorce proneness. We added these pend on whether speci®c marital changes were predisruption means to the observed means in driven by differential selection into marriage, co- 1980 and weighted the predisruption means by .13 hort replacement, or period effects. and the obtained means by .87 (based on the per- Our conclusions would be threatened, however, centage divorced in the 1980 Current Population if changes in marital quality were driven primarily Survey). We followed the same procedure for Marital Quality: 1980±2000 21

2000 but weighted the predisruption means by .19 tural factors, including the health of the economy, and the obtained means by .81 (based on the pro- continued progress toward within portion divorced in the 2000 Current Population marriage, and resolution of the work-family di- Survey). This procedure, of course, lowered the lemmas facing many dual-earner couples. mean levels of marital happiness and interaction and raised the mean level of divorce proneness in NOTE both survey years. The differences in marital qual- We thank Glenn Firebaugh for helpful comments on an ity between surveys, however, were not dramati- earlier draft of this manuscript and for advice with data cally different from those reported in Table 1: analysis. This research was supported by Grant 5 R01 Ϫ.05 for marital happiness, Ϫ.29 for interaction, AG04146 from the National Institute on Aging and the and .00 for divorce proneness. This procedure Pennsylvania State University Population Research In- stitute, with core support from the National Institute of suggests that movement out of the married pop- Child Health and Human Development Grant 1 ulation through divorce did not substantially affect HD282663. our estimates of changes in marital quality. In conclusion, we return to the two perspec- REFERENCES tives on marriage described at the beginning of Allison, P. D. (2001). Missing data. Thousand Oaks, this article. Consistent with a marital decline per- CA: Sage. spective, we see some evidence that large-scale Amato, P. R., & Booth, A. (1995). Changes in gender changes in the United States have made marriage role attitudes and perceived marital quality. American Sociological Review, 60, 58±66. a more dif®cult arrangement in the last two de- Amato, P. R., & Booth, A. (1997). A generation at risk: cades. For example, the rise in premarital cohab- Growing up in an era of family upheaval. Cambridge, itation appears to have increased the perceived in- MA: Harvard University Press. stability of marriages. Moreover, increases in Amato, P. R., & Rogers, S. J. (1999). Do attitudes to- divorce affect marital quality? Journal of Fam- wives' extended hours of employment (more than ily Issues, 20, 69±86. 45 hours per week) and job demands (irregular Axinn, W. G., & Thornton, A. (1992). The relationship shifts, travel, evening meetings) are associated between cohabitation and divorce: Selectivity or with greater tension within marriageЮndings causal in¯uence? Demography, 29, 357±374. Bellah, R. N., Madsen, R., Sullivan, W. N., Swidler, A., that are congruent with recent research on work- & Tipton, S. N. (1985). Habits of the heart: Individ- family stress (Perry-Jenkins et al., 2000). Other ualism and commitment in American life. Berkeley, evidence, however, provides a more optimistic in- CA: University of California Press. terpretation of recent social change. The increase Bernard, J. (1972). The future of marriage. New Haven, CT: Yale University Press. in economic resourcesÐas re¯ected in real in- Blumberg, R. L., & Coleman, M. T. (1989). A theoret- come growth, a decline in the use of public assis- ical look at the gender balance of power in the Amer- tance, and increased optimism about family ®- ican couple. Journal of Family Issues, 10, 225±250. nancesÐappears to have strengthened marriage. Booth, A., Amato, P. R., Johnson, D., & Rogers, S. J. (2002). Survey of marriage and family life: Method- Similarly, changes in gender relations, especially ology report. University Park, Department of Soci- greater decision-making equality between spous- ology, Pennsylvania State University. es, appear to have improved the quality of mar- Booth, A., & Edwards, J. N. (1985). Age at marriage riage. Increased support for the norm of lifelong and marital instability. Journal of Marriage and the Family, 47, 67±75. marriage may have had positive implications for Booth, A., & Edwards, J. N. (1992). Starting over: Why multiple dimensions of marital quality. remarriages are more unstable. Journal of Family Is- Our results indicate that it is possible to sup- sues, 13, 179±194. port either a negative or positive view of recent Booth, A., & Johnson, D. R. (1988). Premarital cohab- itation and marital success. Journal of Family Issues, marital change by focusing selectively on partic- 9, 255±272. ular factors. A broader perspective, however, re- Booth, A., Johnson, D. R., White, L. K., & Edwards, J. veals a mixture of detrimental and bene®cial N. (1981). Female labor force participation and mar- trends, and, with the exception of a decline in ital instability: Methodology report. Lincoln, NE: Bu- reau of Sociological Research, University of Nebras- marital interaction, little change in people's eval- ka. uations of their marriages. Although married cou- Bradbury, T. N., Fincham, F. D., & Beach, S. R. H. ples face numerous challenges in adjusting to a (2000). Research on the nature and determinants of rapidly changing social context, marriage appears marital satisfaction: A decade in review. Journal of Marriage and the Family, 62, 964±980. to be an adaptable institution. Whether marital Bramlett, M. D., & Mosher, W. D. (2001). First marriage quality will improve or decline in the next two dissolution, divorce, and remarriage: United States. decades depends on a variety of structural and cul- Advance Data from Vital and Health Statistics, no. 22 Journal of Marriage and Family

323. Hyattsville, MD: National Center for Health Sta- (2000). Work and family in the 1990s. Journal of tistics. Marriage and the Family, 62, 981±998. Bryk, A. S., & Raudenbush, S. W. (1992). Hierarchical Popenoe, D. (1993). American family decline: 1960± linear models: Applications and data analysis meth- 1990: A review and appraisal. Journal of Marriage ods. Newbury Park, CA: Sage. and the Family, 55, 527±556. Bumpass, L. L., Martin, T. C., & Sweet, J. A. (1991). Rogers, S. J., & Amato, P. R. (2000). Have changes in The impact of family background and early marital gender relations affected marital quality? Social factors on marital disruption. Journal of Family Is- Forces, 79, 731±753. sues, 12, 22±44. Rosenfeld, M. J. (2002). Measures of assimilation in the Cherlin, A. J. (2000). Toward a new home socioecon- marriage market: Mexican Americans 1970±1990. omics of union formation. In L. J. Waite (Ed.), The Journal of Marriage and the Family, 64, 152±162. ties that bind: Perspectives on marriage and cohab- Ross, C. E., & Wu, C. (1995). The links between edu- itation (pp. 126±146). New York: Aldine de Gruyter. cation and health. American Sociological Review, 60, Cohen, J. (1988). Statistical power analysis for the be- 719±745. havioral sciences (2nd ed.). Hillsdale, NJ: Erlbaum. Scanzoni, J. (1978). Sex roles, women's work and mar- Coltrane, S. (2000). Research on household labor: Mod- ital con¯ict: A study of family change. Lexington, eling and measuring the social embeddedness of rou- MA: Heath. tine family work. Journal of Marriage and the Fam- Scanzoni, J. (2001). From the normal family to alternate ily, 62, 1208±1233. families to the quest for diversity with interdepen- Coontz, S. (1992). The way we never were: American dence. Journal of Family Issues, 22, 688±710. families and the nostalgia trap. New York: Basic Seltzer, J. A. (2000). Families formed outside of mar- Books. riage. Journal of Marriage and the Family, 62, 1247± Cowan, C. P., & Cowan, P. A. (1992). When partners 1268. become parents. New York: Basic Books. Skolnick, A. (1991). Embattled paradise: The American Farley, R. (1996). The new American reality. New York: family in an age of uncertainty. New York: Basic Russell Sage Foundation. Books. Ferree, M. M. (1991). The gender division of labor in Spain, D., & Bianchi, S. (1996). Balancing act: Moth- two-earner marriages: Dimensions of variability and erhood, marriage, and employment among American change. Journal of Family Issues, 12, 158±180. women. New York: Russell Sage. Fincham, F. D., & Bradbury, T. N. (1987). The assess- Stacey, J. (1996). In the name of the family: Rethinking ment of marital quality: A reevaluation. Journal of family values in the postmodern age. Boston: Beacon Marriage and the Family, 49, 797±809. Press. Firebaugh, G. (1997). Analyzing repeated surveys. StataCorp. (1999). Stata statistical software: Release Thousand Oaks, CA: Sage. 6.0. College Station, TX: Author. Glenn, N. D. (1991). The recent trend in marital success Thompson, L., & Walker, A. J. (1991). Gender in fam- in the United States. Journal of Marriage and the ilies: Women and men in marriage, work, and par- Family, 53, 261±270. enthood. In A. Booth (Ed.), Contemporary families: Glenn, N. D. (1996). Values, attitudes, and the state of Looking forward, looking back (pp. 76±102). Min- American marriage. In D. Popenoe, J. B. Elshtain, & neapolis, MN: National Council on Family Relations. D. Blankenhorn (Eds.), Promises to keep: Decline Thornton, A., & Young-DeMarco, L. (2001). Four de- and renewal of marriage in America (pp. 15±33). cades of trends in attitudes toward family issues in Lanham, MD: Rowman and Little®eld. the United States: The 1960s through the 1990s. Glenn, N. D., & McLanahan, S. (1982). Children and Journal of Marriage and the Family, 63, 1009±1037. U.S. Bureau of the Census. (1998). Marital status and marital happiness: A further speci®cation of the re- living arrangements (Current Population Reports, Se- lationship. Journal of Marriage and the Family, 44, ries P20-514). Washington, DC: U.S. Government 63±72. Printing Of®ce. Gottman, J. M. (1994). What predicts divorce? Hills- U.S. Bureau of the Census. (2000). Statistical abstract dale, NJ: Erlbaum. of the United States: 2000. Washington, DC: U.S. Heaton, T. B. (2002). Factors contributing to increasing Government Printing Of®ce. marital stability in the United States. Journal of Fam- VanLaningham, J., Johnson, D. R., & Amato, P. R. ily Issues, 23, 392±409. (2000). Marital happiness, marital duration, and the Heaton, T. B., & Pratt, E. L. (1990). The effects of re- u-shaped curve: Evidence from a ®ve-wave panel ligious homogamy on marital satisfaction and stabil- study. Social Forces, 78, 1313±1341. ity. Journal of Family Issues, 11, 191±207. White, L. K. (1990). Determinants of divorce: A review Hochschild, A. (1989). The second shift. New York: of research in the eighties. Journal of Marriage and Avon. the Family, 52, 904±912. Hood, J. (1983). Becoming a two-job family. New York: White, L. K., Booth, A., & Edwards, J. N. (1986). Chil- Praeger. dren and marital happiness: Why the negative corre- Johnson, D. R., White, L. K., Edwards, J. N., & Booth, lation? Journal of Family Issues, 7, 131±147. A. (1986). Dimensions of marital quality: Toward White, L. K., & Rogers, S. J. (2000). Economic circum- methodological and conceptual re®nement. Journal of stances and family outcomes: A review of the 1990s. Family Issues, 7, 31±49. Journal of Marriage and the Family, 62, 1035±1051. Mizan, A. N. (1994). Family power studies: Some major Zill, N., & Nord, C. W. (1994). Running in place: How methodological issues. International Journal of So- American families are faring in a changing economy ciology of the Family, 24, 85±91. and an individualistic society. Washington, DC: Child Perry-Jenkins, M., Repetti, R. L., & Crouter, A. C. Trends.