Married and Poor

Are poor marriages more likely to divorce?

Basic Characteristics of Economically Disadvantaged Couples in the U.S.

David J. Fein, Ph.D.
Abt Associates

David Fein ( is Senior Associate at Abt Associates Inc. He gratefully acknowledges the contribution of unpublished analyses from Paul Amato.

The Supporting Healthy Marriage Evaluation is being conducted by MDRC and its research partners, Abt Associates, Child Trends, and Optimal Solutions Group, under contract to the Office of Planning, Research and Evaluation, Administration for Children and Families, U.S. Department of Health and Human Services. For further information please contact: Virginia Knox (, Project Director, MDRC.

The prisms social scientists have used to study marriage mostly have not been focused on the lower end of the economic spectrum. There has been considerable attention to racial and ethnic minorities and, more recently, to relationships among unwed parents. Although these populations are disproportionately poor, their distinctive attitudes and behaviors could reflect many influences other than economic status. Many analyses of marriage outcomes in the general population have included economic indicators as covariates. Very few, however, have examined carefully the effects of economic or other causal variables among the most disadvantaged sample members (Fein, 2003; Fein et al., 2003).

Emerging federal initiatives seeking to support marriage have increased the need for improved information on low-income married couples. These needs begin with basic descriptive statistics. Research on fragile families has demonstrated that simple facts can be very useful in stimulating thinking about interventions for couples. For example, the finding that a substantial majority of unwed couples are involved romantically around the time of birth but most of these relationships do not survive long after birth has stimulated interest in transition to parenthood programs (Dion et al., 2003). A similar body of descriptive evidence on low-income married couples is needed to support thinking about the broad population of interest, subgroups that might be particularly important to target, and the kinds of services and policy changes that may be most helpful.

One key need is to document the degree to which marriage outcomes vary across different forms and levels of economic disadvantage. Next, we must ascertain how different individual, family, and environmental characteristics of disadvantaged couples are associated with marriage outcomes. Beyond simple measures like marital satisfaction, it will be useful to assess how more specific aspects of marital interaction and related psychological processes — the proximate targets of relationship skills programs — vary across groups. Needed are analyses both of variation in outcomes at a point in time, as well as of changes in outcomes for a population over time.

This paper starts the enterprise by assembling and assessing recent descriptive statistics on the formation and stability, characteristics, and quality of marriages in the low-income population of the U.S. In addition to culling findings from published reports, it also provides new findings from several recent surveys.

Research Questions and Data Sources

Although the questions this paper explores are basic, their answers are neither well nor widely known: Are economically disadvantaged persons less likely to marry and to stay married? How different are poor married couples’ social and economic characteristics compared with those of more affluent couples? Are the numbers of married couples using government assistance programs very large? To what degree does marital quality tend to differ for less and more affluent couples? Do the answers to questions like these depend on how we measure economic disadvantage?

The concept of economic disadvantage has multiple dimensions, encompassing temporary and enduring aspects of current material circumstances (e.g., low family income, financial stress, economic dependency, crime-ridden neighborhood) as well as factors likely to affect future prospects (e.g., low education, disability, lack of health insurance, poor local job market). Some aspects of disadvantage may operate through individual- or couple-level characteristics (e.g., growing up in a family headed by a single parent with low education, low current family income), while others operate through neighborhood-level characteristics (e.g., few married couples in the community to serve as role models). Within any single dimension, where to draw the line between “disadvantaged” and “non-disadvantaged” may not be clear.

In this paper, we explore the implications of several different measures of disadvantage. We emphasize also the need to distinguish clearly the concept of economic disadvantage from that of membership in a racial/ethnic minority group. There are many reasons why racial and ethnic differences in marriage are important to understand, but such differences should not be attributed automatically to economic disadvantage. [1]

As mentioned, this paper assembles findings from published reports, as well as from new tabulations of data from the CPS and two surveys of married couples. A large, high-quality survey, the CPS offers rich detail on demographic and economic characteristics. The March 2003 CPS sample included 70 thousand married adults aged 18-59, representing 89 million married adults nationwide. The March CPS contains little information about these couples’ marriages, however.

For information on marital quality, we draw on special analyses prepared by Paul Amato based on pooled data from two surveys of married adults nationwide — the 1980 Marital Instability Over the Life Course Survey and 2000 Survey of Marriage and Family Life. These two surveys covered a wide range of topics bearing on marriage and family life (Amato et al., 2003). Their samples — 2,034 and 2,100 persons, respectively — are representative of the population of married adults aged 55 and under in the 48 contiguous states.


The following sections provide a basic profile of the population of economically disadvantaged married couples. We look first at rates of entry into, and exit from, first marriage. Next, we examine the timing of transitions to parenthood in relation to marriage. A series of analyses then document how basic socio-economic characteristics of married couples vary by income level and, among low-income couples, by race and ethnicity. Finally, we assess evidence on whether and how marital quality varies across couple income levels.

People with economic disadvantages are just as likely to marry as other people, but their marriages are substantially more unstable.
Characterizing the situation as one of “not as much marriage” among disadvantaged people misses an important distinction. Tying the knot does not seem to be an issue: rather, the problem appears to be maintaining the union thereafter. Statistics from the 1995 National Survey of Family Growth (NSFG), as reported by Bramlett and Mosher (2002), demonstrate this distinction for a variety of individual and community-level indicators of disadvantage.

Through their early 30s, economically disadvantaged adults actually are more likely to marry than advantaged adults. The proportions ever married by age are shown in Exhibit 1 by education (top panel) and neighborhood income level (bottom panel). Fractions ever married are much higher among women with no more than a high school degree in the young adult years, but begin to narrow by age 30. By age 35, other statistics show that the fractions ever married are virtually the same across education groups (Ellwood and Jencks 2001, Tables A11-13). A similar story appears in comparisons by neighborhood income level (Exhibit 1, bottom panel). Through their 30s, women from the most affluent neighborhoods (e.g., in the upper 25 percent of median family incomes) are less likely than those from less affluent neighborhoods to have married. The differences here are somewhat narrower than for education while women are in their 20s, likely because education provides a direct indication of marriage postponement for the sake of college and career.

Looking at Exhibit 2, we see that at every level of education, blacks are substantially less likely to marry than whites or Hispanics. This finding reinforces the warning that differences across race — ethnicity groups may not be informative about differences based on economic status.

In contrast to getting married, the difficulty of staying married increases substantially with levels of economic disadvantage. The probability of splitting up in each year after first marriage is consistently higher for women with less, than for those with more, education (Exhibit 3, top panel) and for those from less, compared with more, affluent neighborhoods (bottom panel). The effect of neighborhood income level is especially large. For example, the probability of breaking up within 10 years of marriage is nearly twice as high for women from the bottom quarter (44 percent break-up) as for those from the top quarter (23 percent break-up) of neighborhoods ranked by median family income.

At every level of education, black women have the highest risks of marital disruption over the 15 years after they first marry (Exhibit 4). In contrast, disruption rates for married Hispanic women are relatively low, particularly at low levels of education. Community-level indicators of economic disadvantage show generally similar patterns.

The fragility of marriages among disadvantaged couples is one strong rationale for efforts to strengthen marriage in low-income populations. The statistics indicate also the possible benefits of targeting black married couples and couples who are relatively young, since both are at higher risk of breaking up. Although substantial, it is worth noting that the evidence on risks of early marriage is based on analyses of general population samples (Bumpass et al., 1991; Heaton, 2003), and we do not know how much greater the risks are for disadvantaged couples who marry young.

Whereas the vast bulk of first transitions to parenthood among upper middle class couples follow first marriages, first births among disadvantaged newlyweds are far more likely to precede marriage.
There is substantial interest in targeting marriage interventions around the point that couples have their first child, a time of both optimism and strain when couples may be especially receptive to marriage education services (Cowan and Cowan 1995, Dion et al., 2003). In this regard, a useful piece of information is that a substantial fraction of disadvantaged first-time newlyweds already have children.

Ellwood and Jencks analyzed CPS data on the timing of women’s first births relative to first marriages in successive marriage cohorts after dividing each cohort into thirds based on its level of education. Among couples who married in 1990, one-third of those in the bottom education category had their first child before marriage, compared with one-tenth of those in the top education category (Exhibit 5). Among African Americans, the fraction is even higher: over half (51 percent) married after their first transition to parenthood. [2]

Thus, programs targeting only marital first transitions to parenthood will exclude a substantial number of newlyweds. One response is to target such couples before marriage, and there currently is a great deal of interest in programs that do just that. Alternatively, it might be helpful to target couples who are already parents at marriage on the basis of subsequent (marital) births, as well as on other developmental milestones (e.g., children reaching school age or puberty).

Many premarital transitions to parenthood occur more than three years before marriage: 55 percent of premarital first births in the bottom educational third, and 82 percent of premarital first births to African Americans. These earlier births are more likely to have been fathered by a partner prior to the current spouse. Statistics from a recent Florida survey confirm that low-income married couples are more likely to be living with children from prior relationships. Nearly half (49 percent) of low-income (under 200 percent of poverty) married couples with children were living with at least one child from a prior relationship, compared with 31 percent of couples with incomes of at least 400 percent of poverty (Karney et al. , 2003, Table 5).

The fact that at least one spouse often has children from prior (unmarried) partners suggests that relationships with prior partners around parenting are likely to be more salient for low-income married couples. Thus issues concerning relationships with prior partners and stresses associated with step parenting are important subjects for marriage programs for disadvantaged newlyweds. Research on unwed parents, among whom rates of multiple partner fertility are even higher (Carlson and Furstenberg, 2003, Tables 4-6), has inspired similar recommendations.

The pace of fertility after marriage also is more rapid among less well-educated women. Over half (53 percent) of women in the lowest education category who married before having their first child had a birth within the next three years, compared with 41 percent in the top category (calculated from percentages in Exhibit 5). It is likely that unplanned and premarital conceptions partly account for somewhat greater pace of childbearing. One of the better established findings in the academic literature is that premarital conceptions and births substantially increase the risk of subsequent marital disruption (Upchurch et al., 2001).

Notwithstanding their faster pace of fertility, a substantial fraction of women in the bottom education category (43 percent of women childless at marriage) did not have their first birth until at least three years after marriage (Exhibit 5). An implication is that programs targeting newlyweds will include many couples who do not have children for a substantial period after they receive services.

Disadvantaged married couples are comparatively young and disproportionately Latino. African Americans constitute a relatively small share of this population.
CPS tabulations reveal a number of important differences in the characteristics of married couples at different ends of the economic spectrum. We look first at basic demographic characteristics of couples who were poor, or near-poor, compared with more affluent families.

Of all U.S. married women in March 2003, 2.5 million (5 percent) were poor, and eight million (17 percent) had family incomes below 200 percent of poverty. [3] Low-income married women are substantially younger than their middle- and higher-income counterparts. Among married women, 47 percent of those who were poor were still in their prime childbearing ages (under 35), compared with only 18 percent of those with incomes at least six times above poverty (see Exhibit 6). The fractions with children under age six in these two income groups were 42 and 14 percent, respectively.

There are several reasons why low-income married adults tend to be younger than more affluent couples. One reason is that, as shown in Exhibits 1-4, poor people marry at younger ages and have shorter marriages. A second reason is that the age distribution of poor populations tends to be younger due to higher fertility rates compared with more affluent groups. A third reason is that young couples are at earlier stages in their careers and tend to have lower average earnings than older couples.

One of the more striking findings in this analysis is the high proportion of Latinos among low-income couples. Over one-third (35 percent) of poor married couples are Hispanic, less than half (47 percent) are non-Hispanic white, and only one-tenth are non-Hispanic black. Latinos’ disproportionate representation among poor married couples is due to the fact that—as shown earlier—they are relatively likely both to marry and stay married, as well as to be poor in the first place. By weight of numbers, Latino couples appear to deserve substantial attention in marriage-strengthening initiatives and related research.

African Americans are far less prominent among low-income married couples than among new unwed parents. In large cities covered in the 1999 Fragile Families baseline survey, blacks accounted for 44 percent of births to unwed parents, but only 13 percent of births to married parents (McLanahan, 2003).

Priority target groups for marriage services must be established on the basis of need and ability to benefit from services, however, as well as numbers. In this regard, married African Americans deserve a prominent spot on the marriage agenda, due to their high risk of marital disruption.

Low-income married couples have substantially lower levels of education and employment than higher-income couples.
With good-paying jobs increasingly requiring advanced education, it is little surprise to find that education is highly correlated with income among married couples. Among poor husbands, 41 percent have no high school diploma, and only 27 percent have some education beyond high school (Exhibit 6). The corresponding statistics for the most affluent husbands are 2 percent and 82 percent, respectively. Education distributions for wives show very similar income differences.

Among poor couples, only 51 percent of husbands and 14 percent of wives hold full-time jobs, compared with 92 and 66 percent of husbands and wives, respectively, in the top income category. Together, poor husbands and wives averaged only 33.5 hours of work per week, compared with 72.6 hours for couples in the most affluent group.

Low education and employment levels are important to consider in designing marriage programs. Some adaptation is likely to be needed in marriage skills curricula developed for well-educated, middle-class couples in order to be appropriate for populations with low levels of literacy and experience in formal classroom settings. It may be helpful to add material on how couples can recognize and respond constructively to economic stresses that may affect their relationships. Finally, employment services and related supports could prove valuable in reducing economic worries so that couples can focus on improving their relationships.

Millions of married couples use a wide variety of government assistance programs.
Notwithstanding the popular view of public assistance recipients as single-parent families, a great many married couples also receive government aid. Among all recipients, the share married ranges from 13 percent for rent subsidies to 66 percent for workers compensation benefits (see Exhibit 7, top panel). [4] Married women accounted for about one-fifth of all women who received TANF and food stamps and nearly half (46 percent) of all women with a family member registered for Medicaid. Of the fourteen programs shown in Exhibit 7, seven served over a million married couples.

In addition to providing valuable supports to married couples and their families, these programs represent potential avenues for providing marriage and family strengthening services to married couples. The varying rules and services of different programs suggest possibilities for reaching families with varying socioeconomic characteristics and needs, as well as for integrating marriage education services with an array of other services.

The characteristics of major racial and ethnic groups within the low-income married population vary by age, age of youngest child, education, region of the U.S., and urban-rural residence.
Racial and ethnic differences in characteristics may have implications for targeting, recruitment, and service strategies. Exhibit 8 compares the characteristics of white, black, and Hispanic married couples with annual family incomes under 200 percent of poverty. The statistics correspond to estimated total populations of 4.1, .8, and 2.5 million non-Hispanic white, non-Hispanic black, and Hispanic married couples, respectively.

Latinos are especially likely to be recent parents. Half (51 percent) of all low-income married Latino couples have young children, compared with 37 and 34 percent of whites and blacks, respectively. Latinos constitute 40 percent of all low-income married parents with children under age six (not shown in exhibit).

Disadvantaged Latino couples also have especially low levels of formal education. Three in five husbands lack a high school degree, compared with only one in five white and black husbands. Only 14 percent of Latino husbands have some education beyond the high school level, compared with 37 and 34 percent of low-income white and black husbands, respectively. Statistics for wives’ education are very similar.

Notwithstanding their low education levels, Latino husbands are substantially more likely to be working full-time (74 percent) than either white (62 percent) or black (55 percent) husbands. Among low-income husbands, fewer than one in five (17 percent) Latinos is not working, compared with 30 percent of whites and 35 percent of blacks. Among low-income wives, blacks hold the employment edge: over a third (35 percent) of black wives works full-time, compared with 23 percent of whites and 25 percent of Latinos.

The geographic distribution of low-income married couples varies substantially by race and ethnicity. Nearly three quarters of low-income Hispanic couples live in the southwestern U.S., whereas the majority of low-income black married couples lives in the southeastern states. Both groups are disproportionately concentrated in the central cities of large metropolitan areas. In contrast, a relatively large fraction (37 percent) of low-income married whites lives in rural areas.

These race-ethnicity differences have a number of possible implications for marriage programs. One is that marriage education programs for Hispanic couples must be appropriate for persons with very low levels of formal education, as well as with English language difficulties. Another implication is that employment services for husbands might be especially helpful for black couples. Finally, geographic statistics suggest that substantial resources be allocated to serve Hispanics in inner city neighborhoods in southwestern states, blacks from inner cities in southeastern states, and whites from areas with relatively large rural populations.

There is a marital quality gap between low-income and other couples, but it is not as large as might be expected based on differences in marital disruption rates.
There is some concern that prevention-oriented marriage education programs developed for upper middle-class couples may be inappropriate if poor couples’ relationships are much more distressed. The fear is that poverty brings a substantial array of stresses that spill over into marriages and create abundant marital distress. Poor couples’ substantially higher disruption rates appear to support this concern.

The weight of evidence suggests that marital quality is at most only slightly lower for poor couples than for more advantaged ones, however. Statistics from a 2003 general population survey in Florida show only a small difference in mean levels on an index of relationship satisfaction between low-income (under 200 percent of poverty) and other married couples (Karney et al., 2003). An analysis of pooled General Social Survey data for 1972-96 reveals no relationship between responses to a simple question on marital satisfaction and (logged) income for either men or women, and a weak positive association for education (only among women), after controlling for age, year, and other demographic variables (Waite, 2000). Taking public assistance receipt as the indicator of disadvantage yields somewhat larger differences (Amato et al, 2003; Johnson et al., 2002; Karney et al., 2003), suggesting that public assistance receipt may signal more acute vulnerabilities to conditions harmful to marriage.

Exhibit 9 provides a sense of differences in relationship quality by income for a nationwide population of married adults, based on the question “How happy are you with your marriage?” [5] Percentages answering “very happy” and “not too happy” are plotted by family income-to-needs (poverty) ratio (the third category, “somewhat happy,” is not shown). The results are based on pooled data from two cross-sectional surveys — the 1980 Marital Instability Over the Life Course survey and the 2000 Survey of Marriage and Family Life — for a total sample size of 4,026 married persons.

The fraction reporting that they are “very happy” increases steadily with income, ranging from 56 percent in the bottom income group (poor) to 68 percent in the top group (income more than six times above poverty). The increase in happiness from one group to the next is fairly small, however. The fraction reporting they are “not too happy” declines only a little over the entire income range.

Next, we examine income differences for three different dimensions of marital quality. Each of the dimensions—marital happiness, positive interaction, and divorce proneness—is measured using an index combining a different series of survey items. [6] The marital happiness index reflects global satisfaction with one’s marriage. The positive interaction scale summarizes the frequency with which couples do things together. Finally, the divorce proneness measure reflects the degree to which people have thought, or are thinking, about divorce. Exhibit 10 reports the means for each index, standardizing individual scores so that their mean is zero and their standard deviation is one. In the top panel (Model I), we control for several demographic characteristics (age, race-ethnicity, gender, marriage duration, and marriage order) that might affect marital quality and vary by income. In the bottom panel (Model II), we control for wife’s education and employment status in addition to the other variables in Model I. We combine families under 200 percent of poverty in a single category due to small sample sizes.

The associations of marital happiness and marital interaction with income are positive and statistically significant in both Models I and II. As for the simpler happiness measure (Exhibit 9), however, the income differences are not that large: the difference between top and bottom income categories is .31 standard deviations for marital happiness and about half that (.18 standard deviations) for marital interaction. Controlling for wife’s education and employment (Model II) strengthens the relationship between income and positive interaction, but has little effect on the relationship between income and overall happiness.

In contrast, divorce proneness is not associated with income when only demographic adjustments are employed (Model I). When we control for wife’s education and employment (Model II), however, the expected negative relationship appears — divorce proneness is .16 standard lower in the highest than in the lowest income group. [7]

This last result suggests that thoughts about divorce may be affected differently by the couple’s economic situation and by the wife’s economic prospects. At a given level of financial wherewithal, wives with more education and employment may be more likely to think about leaving an unsatisfying marriage because they are more likely to perceive that they can make it on their own. And, because income is highly correlated with wife’s education and employment, the greater stability of more affluent couples is not apparent until we control for wife’s economic status.

Returning to the larger question, the evidence suggests that marital quality is positively associated with family income, but also that the association is relatively weak compared with differences in cumulative rates of actual marital disruption. One explanation may be that weaker social supports and cultural norms, and more stressful external circumstances, lead to more negative outcomes for poor couples who do experience distress. The most important differences thus may not be in relationship quality at a given point in time, but rather in the rapidity with which things fall apart in the face of stressful events.

Weak associations between income and marital quality also may signal protective mechanisms that help couples adapt to the stresses of poverty. Research on the pathways between economic situations and relationship quality has identified a number of moderating factors that may insulate couples from a certain amount of stress spillover (Conger et al., 1999, 2002; Karney et al., 2003; Vinokur et al., 1996). Couples appear to be able to recognize and discount for stress effects, at least up to a point (Tessor and Beach 1998). Facing external challenges successfully also can help bring couples closer together.

It is conceivable that poverty has aspects that are advantageous for couples. In Florida, poor couples report spending significantly more time together than higher-income couples, and this time is highly correlated with relationship satisfaction (Karney et al., 2003). Earlier, we saw that poor couples averaged substantially fewer hours of work than affluent ones. Thus, although low employment surely creates financial stress, it also may allow couples more time to be together and avoid some of the other stresses involved in balancing two full-time work schedules.

This descriptive evidence on marital quality has several possible implications for marriage interventions. The modest income differences in marital quality should alleviate worries that vast numbers of disadvantaged couples are too distressed to benefit from prevention-oriented marriage skills programs. At the same time, if poor couples’ relationships stand a relatively high risk of rapid deterioration, then the timing of education services, and of ongoing monitoring and supports, could prove critical. If greater relationship fragility reflects especially a lack of external supports and constraints, services addressing external situations—and helping to create supportive conditions—may be helpful. Finally, marriage programs should identify and build on the unique strengths, as well as the challenges, facing poor couples.


In this paper, we have provided an initial descriptive appraisal of the population of disadvantaged married couples in the U.S. Comparing couples along several indicators of economic disadvantage, we have looked at marriage entries and exits, at the timing of first parenthood relative to marriage, at demographic and economic characteristics, and at marital quality.

One theme in the findings is that it can be dangerous to attribute to disadvantaged married couples results observed for particular race-ethnicity or demographic groups. For example, blacks are less likely than non-blacks at all economic levels to marry, but poor people are just as likely as other people to marry. In contrast to their prominence in the unwed parent population, blacks constitute only a small fraction of low-income married couples, who are mostly white or Hispanic. To the extent that cultural differences linked to race and ethnicity affect marriage behaviors, we should not assume that findings for fragile families will apply to low-income married couples.

We have seen also signs that using different indicators of economic disadvantage can produce somewhat different answers to questions about married couples. The negative association between economic status and entries to marriage (before age 30) is stronger for education than for neighborhood income level, whereas the positive association between economic status and marital stability (over the 15 years following first marriage) is stronger for neighborhood income level than for education. Looking at who is most likely to contemplate divorce, we find that income reduces divorce proneness, but wife’s education and employment—another indication of economic status—increases divorce proneness. These findings underscore the need to examine a variety of indicators of economic disadvantage in descriptive analyses.

We have noted a number of possible implications for marriage initiatives. Most broadly, high marital instability rates (and attendant risks for children) affirm that low-income couples indeed do face special challenges in their marriages. At the same time, relatively modest differences in marital quality suggest that prevention-oriented marriage education programs developed for upper middle class couples could be appropriate for low-income couples as well. Low levels of formal education—especially among Latinos—point to the possibility that substantial adaptations in teaching methods will be needed. High unemployment rates also remind us that it may be useful to provide employment and other services in addition to marriage skills programs. Regional and rural-urban distributions help us to assess where key target groups may be concentrated. Finally, married couples’ substantial participation in a wide variety of public programs suggests a variety of possible channels for recruitment.

There remain a great many questions deserving further research. One general need is to learn more about the various demographic subgroups relevant to marriage intervention targeting and service design—especially married couples with infants and adolescents and with children from previous partners. It would be useful to learn much more about the characteristics of economically disadvantaged couples that might have a bearing on relationship outcomes, including a wide range of adult strengths and vulnerabilities and contextual aspects such as social networks and community characteristics. Finally, there is a great need for richer study of relationship processes and outcomes—such as thought processes, emotional connections, and behavioral interactions—of economically disadvantaged couples, particularly in comparison with those of more affluent married couples. The best existing marriage education programs have been based on careful research on upper-middle class couples, and the extent to which their principles will apply to disadvantaged couples is uncertain.


Amato, P., Johnson, D., Booth, A., and Rogers, S. (2003). Continuity and change in marital quality between 1980 and 2000. Journal of Marriage and Family, 65(February 2003), 1-22.

Bramlett, M., and Mosher, W. (2002). Cohabitation, marriage, divorce, and remarriage in the United States. National Center for Health Statistics. Vital Health Statistics, 23(22).

Bumpass, L. L., Martin, T. C., and Sweet, J. A. (1991). The impact of family background and early marital factors on marital disruption. Journal of Family Issues, 12(1), 22-42.

Carlson, M., and Furstenberg, F. (2003, October). Complex families: Documenting the prevalence and correlates of multi-partnered fertility in the United States. (Center for Research on Child Wellbeing, Princeton University, Working Paper # 03-14-FF).

Conger, R., Ebert-Wallace, L. , Sun, Y., Simons, R., McLoyd, V., and Brody, G. (2002). Economic pressure in African-American families: A replication and extension of the Family Stress Model. Developmental Psychology, 38(2), 179-193.

Conger, R. D., Rueter, M. A., and Elder, G. H., Jr. (1999). Couple resilience to economic pressure. Journal of Personality and Social Psychology, 76(1), 54-71.

Cowan, C. P., and Cowan, P. A. (1995). Interventions to ease the transition to parenthood: Why are they needed and what can they do? Family Relations, 44, 412-423.

Dion, M., Devaney, B., McConnell, S., Ford, M., Hill, H., and Winston, P. (2003). Helping Unwed Parents Build Strong and Healthy Marriages: A Conceptual Framework for Interventions. Washington, D.C. : Mathematica Policy Research, Inc.

Ellwood, D., and Jencks, C. (2001). The Growing Differences in Family Structure: What Do We Know? Unpublished manuscript, Harvard University.

Fein, D. (2003). Determinants of marriage and cohabitation among disadvantaged Americans: Two themes. Presented at the Marriage and Family Formation among Low-Income Couples: What Do We Know from Research? Georgetown University: National Poverty Center.

Fein, D., Burstein, N., Fein, G., and Lindberg, L. (2003). The Determinants of Marriage and Cohabitation among Disadvantaged Americans: A Literature Review. Bethesda, MD: Abt Associates Inc.

Heaton, T. B. (2002). Factors contributing to increasing marital stability in the United States. Journal of Family Issues, 23(3), 392-409.

Johnson, C., Stanley, S., Glenn, N., Amato, P., Nock, S., Markman, H., et al. (2002). Marriage in Oklahoma: 2001 Baseline Statewide Survey on Marriage and Divorce. Oklahoma: Bureau for Social Research, Oklahoma State University.

Karney, B., Garvan, C., and Thomas, M. (2003). Family formation in Florida: 2003 baseline survey of attitudes, beliefs, and demographics relating to marriage and family formation. Draft. University of Florida.

Karney, B., Story, L., and Bradbury, T. (In press). Marriages in context: Interactions between chronic and acute stress among newlyweds. In T. Revenson, K. Kayser and G. Bodenmann (Eds. ), Emerging perspectives on couples’ coping with stress. Washington, D.C. : American Psychological Association.

McLanahan, S. (2003, December). Fragile families and the marriage agenda. (Center for Research on Child Wellbeing, Princeton University, Working Paper # 03-16-FF).

Tesser, A., and Beach, S. R. (1998). Life events, relationship quality, and depression: An investigation of judgment discontinuity in vivo. Journal of Personality and Social Psychology, 74(1), 36-52.

Upchurch, D. M., Lillard, L. A., and Panis, C. W. (2001). The impact of nonmarital childbearing on subsequent marital formation and dissolution. In L. L. Wu and B. Wolfe (Eds. ), Out of Wedlock: Causes and Consequences of Nonmarital Fertility (pp. 344-382). New York: Russell Sage Foundation.

Vinokur, A. D., Price, R. H., and Caplan, R. D. (1996). Hard times and hurtful partners: How financial strain affects depression and relationship satisfaction of unemployed persons and their spouses. Journal of Personality and Social Psychology, 71(1), 166-179.

Waite, L. (2000). Trends in men’s and women’s well-being in marriage. In L. Waite (Ed. ), The Ties that Bind: Perspectives on Marriage and Cohabitation (pp. 368-392). New York: Aldine de Gruyter.

[1] Tabulations from the March 2003 Current Population Survey reveal that the vast majority of black and Latino adults (about four fifths) are above poverty, and most are above 200 percent of poverty. Although minority groups are disproportionately poor, a substantial body of research finds that differences in key family behaviors persist after controlling for economic circumstances (Fein et al., 2003).

[2] Ellwood and Jencks do not provide statistics for African Americans in different education categories.

[3] Income distributions for married men in the CPS are nearly identical to those for married women, but there were somewhat fewer men in the sample (see Exhibit 6). In theory, the counts for husbands and wives should be the same, since we restricted the sample to civilian adults who were married to, and living with, a civilian spouse. The apparent under-coverage of husbands relative to wives in the CPS could be due either to misclassification of marital status or higher omission rates from the survey.

[4] In Exhibit 7, “receipt” for most programs is defined as any participation during in the year prior to the March 2003 CPS interview. For subsidized rents and public housing, receipt indicates currently receiving these supports.

[5] I am indebted to Paul Amato for providing the analyses for Exhibits 9 and 10.

[6] See Amato et al. (2003) for a description of these indices.

[7] Because the sample is restricted to currently married couples, the measured levels of divorce proneness will tend to understate levels among ever-married couples (including those who did divorce). Such truncation also might attenuate the relationship between economic characteristics and divorce proneness if couples remaining married are less likely to act on thoughts of divorce.