This report was prepared for the U. S. Department of Health and Human Services' Office of the Assistant Secretary for Planning and Evaluation under HHS Grant Number 00ASPE359A. The views expressed are those of the authors and should not be attributed to the U.S. Department of Health and Human Services or to the Urban Institute, its trustees, or its funders. The author thanks our project officer, Kelleen Kaye, and Elaine Sorensen for providing useful comments, Carolina Krawiec and Stephanie Riegg for excellent research assistance, and Avner Ahituv for help in implementing econometric specifications.
The nonpartisan Urban Institute publishes studies, reports, and books on timely topics worthy of public consideration. The views expressed are those of the authors and should not be attributed to the Urban Institute, its trustees, or its funders.
Over the last four decades, the declining proportion of married adults in the United States has contributed to a significant worsening of the economic status of families with children. The rise in single parenthood, together with limited child support payments, has meant that more children must rely primarily on the income of only one of their parents, usually the mother. As a result, despite healthy growth in per capita income, child poverty rates in the U.S. have remained at their 1970s levels. Researchers have demonstrated that reduced marriage propensities have caused substantially higher child poverty rates, even after accounting for the fact that the men unmarried mothers might marry have lower incomes than current married fathers (Lerman, 1996; Sawhill and Thomas, 2001).
On one level, it should be no surprise that single-parent families, with fewer potential earners or caregivers, would have much lower incomes than married couple families. But, in fact, understanding the decline in marriage and its implications for economic well-being is a complex problem. Given the dramatic increases in cohabitation and the high levels of co-residence of single mothers with their parents or other adults, many single mothers live with a second potential earner/caregiver and thus do not have a built-in economic disadvantage relative to married couple families. The simple distinctions between married parents and single parents are no longer sufficient for analyzing economic differences. The specific household form as well as the timing of marriage, divorce, separation, and non-marital childbearing will all be relevant to the way marriage and other family structures affect economic hardship. The analysis must take account of trends and patterns of marriage rates at each age, of childbearing rates within and outside marriage, of the duration of marriages, of cohabitation rates, of separation and divorce rates, and of household living arrangements of single parents.
The low and unstable incomes of potential husbands are another reason why marriage might not improve economic welfare of many mothers and children. According to ethnographer Kathryn Edin (2000), when asked about marriage, low-income mothers say that, "...marriage usually entails more risks than potential rewards." Although some of the risks relate to non-economic issues, such as domestic violence, trust, and sharing control of the household, women often mentioned the risk that potential husbands lacked the ability to earn a steady, adequate income and that they consequently become an economic burden.
In a recent paper (Lerman, 2001), I examined the role of marriage in limiting the degree of material hardship faced by families with children. The paper's focus was on whether marriage limited the incidence of material hardship, even among poor or near-poor families. The results showed that married, biological parents experienced lower rates of hardship than other parents with similar characteristics, including those with similar family income-to-needs ratios.1
This paper analyzes the relationship between marriage and economic well-being in a dynamic context. Using data on women and mothers over time enhances our ability to distinguish a causal effect of marriage from a selection effect. Cross section estimates are subject to bias because individuals who marry may have unobserved advantages affecting their incomes over individuals who do not marry. With panel data, we can observe the income profiles over time of individuals who marry and those who do not.
Still, limiting the selection problem is inherently difficult in the absence of an exogenous variable that reduces or stimulates marriages in one environment but not in another. I turn to four strategies. The most straightforward compares the economic outcomes of mothers by marital and family status categories, while taking account of differences in race, academic and technical ability, and other observed characteristics. The second is to compare women in "shotgun" marriages (marriages induced by childbearing, i.e., women with premarital pregnancies in marriages that take place after the pregnancy but before the birth of a child) to women who have premarital pregnancies and who do not marry before or soon after the child's birth. A third comparison is between women in shotgun marriages and women in conventional marriages (women who marry before becoming pregnant).
The third approach uses propensity score matching. Specifically, I estimate the probabilities of marriage among women who have a non-marital pregnancy and then comparing married and unmarried mothers who have similar probabilities of marriage. Fourth, I estimate the impact of marriage on outcomes, while controlling for unobserved differences among individuals using fixed and random effect models.
The diverse approaches deal with different questions about potential economic gains from marriage. The tabulations, basic regressions, and propensity score matching approaches ask mostly about what happens to the long-run economic gains from marrying before or soon after the birth of their first child. This approach focuses on the impact of an initial marriage on subsequent outcomes, incorporating direct and indirect effects of early marriages. This approach does not capture benefits from marriages that occur some years after initial childbearing and thus may understate income gaps associated with marriage. The reason is that income gaps might narrow between the initially married and initially unmarried as marital status changes for both groups; some of the initially married women become unmarried for some years and vice versa. To deal with the potential ongoing benefits of marriage, while at the same time controlling for unobserved differences among mothers, we employ fixed and random effects models to examine the question of how changes in marriage in specific years affect changes in economic status. These models control for the possibility that unobserved individual differences might be responsible for any observed connections between marriage and economic well-being.
The major goal of the paper is to bring new evidence to bear on the question of how marriage affects economic status. No single answer is plausible because any impacts from marriage itself are likely to vary with the circumstances such as which men and women marry and when they marry (say, before or after pregnancy or childbearing). Of particular interest are marriages to women with relatively low earnings capacities and to women who have a premarital pregnancy. A second goal is to describe in detail the marriage and family status experiences of women after their first pregnancy.
The following section takes up the task of clarifying the mechanisms by which marriage could enhance the economic status of mothers and children. I then discuss the data and present initial tabulations of patterns of marriage and family status over time. Next, I present descriptive evidence about the connection between marriage and economic well-being. The fifth section describes the propensity score methodology and the results drawn from applying this approach. The sixth section discusses and uses fixed and random effects models. I conclude with some summary remarks and interpretations.
1. Arguably, holding income-to-need levels constant may bias any observed impact of marriage downward. If marriage induced income gains, then married mothers with the same incomes may well have weaker earnings capacities than unmarried mothers. Still, the findings of reduced hardship among married women were robust and revealed consistent gains not only compared to single parents living with no other adults, but also compared to unmarried mothers living with a cohabiting partner or with other adults.
This report is available in its entirety in the Portable Document Format (PDF), which many find convenient when printing.