Welfare reform has traditionally focused on helping single mothers get jobs, perhaps with the help of government-funded job training and child care. But such efforts are expensive and have shown only modest results. Recently, long-overdue efforts to reduce out-of-wedlock births have been made. But, again, results have been largely unimpressive. What about marriage as a route off welfare? Despite the hopes of feminists that women will become economically independent of men, marriage remains a major source of economic security for women with children, especially if they have few job skills. In fact, many divorced and unwed mothers escape welfare (and poverty) only through marriage. Hence, encouraging such unions ought to be as important an issue in welfare reform as job training and work.

Current welfare law, however, discourages marriage by imposing steep financial penalties on couples that wed. The poorest welfare mothers, disproportionately African American, suffer the stiffest "marriage penalties." The evidence suggests that these penalties, by poisoning attitudes about marriage and reducing the marriageability of many black men, are implicated in the low rates of marriage -- and high rates of illegitimacy -- across the African-American community.

At any given time, 62 percent of welfare recipients are in the midst of a series of welfare spells that will eventually last a total of nine or more years, according to estimates by the Urban Institute's Ladonna Pavetti. Many mothers, especially older ones with better job skills or work experience, find jobs and leave welfare relatively quickly. But others, especially younger mothers who dropped out of high school or have little work experience, find it much more difficult to work their way off welfare.

It is easy to see why. Nationally, the total benefit package for a mother on welfare with two children is worth upwards of $8,500, with another $4,500 worth of Medicaid benefits, and, for the 30 percent lucky enough to receive them, about $5,000 in housing benefits. Most long-term welfare recipients simply cannot earn this much in their first jobs, but, even for those who can, work often "does not pay." They end up working 40 hours a week for the same money that the welfare system provides them for staying at home. And they have precious little time and energy left for their children. Job-training programs simply don't equip women with the skills needed to earn enough to escape this trap.

These realities are what make marriage such an important antipoverty mechanism. The economics is simple enough. The addition of a husband adds another wage earner to the household, reducing the need for paid child care or making it easier for the mother to work. Although income statistics by family structure also reflect underlying differences in who gets married, they make the point: In 1994, the median income for female-headed families with children was $14,902, compared to $47,244 for a married couple with children. About one-half of all single-mother households have incomes below the poverty line, compared to only 8 percent of married couples. Married couples also have sharply lower rates of dependence on government programs: In 1991 and 1992, about one-third of all single mothers participated in a major means-tested government-assistance program (such as welfare or Food Stamps) for the entire two years, compared to only 4 percent of married couples.

The role of marriage as a route off welfare is, however, easily underrated. For example, Kathleen Mullan Harris of the University of North Carolina at Chapel Hill found that about two-thirds of the monthly exits from welfare were because of a new job or a pay increase at an existing job. Only about 5 percent were due to marriage. But, as she notes, most of the mothers who leave welfare for work soon return to the rolls.

Indeed, the story is quite different for more lasting welfare exits. Mary Jo Bane and David Ellwood, then both at Harvard University, studied exits that lasted for at least one year. For these women, exits due to work and marriage were roughly equal. Marriage also helps many mothers escape poverty altogether. Greg Duncan of the University of Michigan studied the economic impact of both marriage and work on female-headed families. He found that, in seven years, the incomes of the poor mothers who married rose 50 percent faster than those of the single mothers who started working without marrying. It is true that Bane and Ellwood and Duncan relied heavily on data from the 1970s, before the importance of work increased for women (as more joined the labor force and their jobs paid better). Nevertheless, the continued centrality of marriage to the economic well-being of many mothers (and children) is evidenced by the data on income and program participation cited above, as well as in more recent data on welfare exits analyzed by Duncan.

There are basically two groups of single mothers on welfare: divorced mothers and never-married ones. In 1994, two-thirds of the children on Aid to Families with Dependent Children (AFDC) had mothers who were never-married, compared to only 30 percent who were the children of divorced or separated mothers. The remainder were widows, the temporarily disabled, and "unknown."

In years long past, relatively few divorced mothers remarried, and few unwed mothers who kept their children married. But, as both divorce and illegitimacy increased in prevalence, and as the stigma associated with both declined, the likelihood of subsequent marriage has also increased. Larry Bumpass, James Sweet, and Teresa Castro Martin of the University of Wisconsin at Madison estimate that 72 percent of women who divorced between 1980 and 1984 will ultimately remarry, one-half within seven years.

If divorced mothers frequently marry, what about unwed mothers, especially those who had their first babies as teenagers? We could find only one study on the subject, so we decided to explore this question ourselves using the National Longitudinal Survey of Youth (NLSY). The NLSY began in 1979 as a national sample of 12,686 males and females, ages 14 to 22. The NLSY collects longitudinal data on both the fertility and marriage of young women and tracks enough unwed teenage mothers to allow for statistically reliable analyses of various social, economic, and demographic factors. Importantly, it contains a relatively recent cohort of unwed mothers.

We created a subsample of young women whose childbearing and marital histories we could follow: 2,783 women who were still participating in the survey in 1993 and who, at the time of the first interview in 1979, were under 20, had never been married, and had never given birth. Starting with 1979, we followed each of these women through subsequent surveys until they were 28 years old, between 1988 and 1993.

Almost 7 percent of the young women in our subsample had babies as unwed teens. Compared with the others in our subsample, the unwed teen mothers were much more likely to be African American, to have grown up in poverty, and to have scored lower on the available measures of human capital. Surprisingly, 54 percent of the unwed teen mothers in our subsample had married by age 28. (We capped our analysis at this age because it was the oldest point that all the women in our subsample reached.) Most of these marriages took place when the women were between 20 and 24 years old. The most frequently observed pattern was that of a young woman who was 15 at the first survey (1979), gave birth at age 19 (1983), and married at age 24 (1988).

This 54 percent marriage rate for the unwed teen mothers, however, is about one-fourth lower than that of the women who did not have babies as unwed teens -- almost 70 percent of whom had married by our cut-off age of 28. And this disparity will probably grow over time as the more middle-class and better-educated women in our subsample, who tend to marry later in life, do so in greater numbers.

Marriage, of course, is only part of the story. It does an unwed teen mother little good to marry and then quickly divorce. Unfortunately, the marriages of unwed teen mothers appear substantially less stable than those of other women: By age 28, about 47 percent of the unwed teen mothers who had married were subsequently divorced, compared to 31 percent for the married women who were not unwed mothers.

Despite the higher divorce rate for unwed teen mothers, marriage remains a valuable institution for many. Robert Lerman of the Urban Institute has translated the NLSY's cold data into actual case histories. Here is one:

Diana managed against the odds. As a black high school dropout with very poor academic skills, she became an unwed mother at age 18 in 1981. For the first year of motherhood, she received AFDC, Food Stamps, and child support, but was poor with an income of only about $4,000. She married the following year and her family income rose to about $14,000. In 1983, when she started working extensively in the labor market, family income doubled to $28,000, ranged from the high $20s to mid $30s through the late 1980s, and reached $41,000 in 1989.

Aggregate statistics about marriages among divorced and unwed mothers, however, mask important racial differences. According to Bumpass and his colleagues, three-quarters of white divorcees will eventually remarry, one-half within six years. But one-third fewer black women will remarry, and it will be eight years before even one-quarter do so. We found a similar racial gap in marriages among women who had their first babies as unwed teens.

In our subsample, the eventual marriage rates of white and Hispanic unwed teen mothers were sharply different from African-American ones: 76 percent of white unwed teen mothers had married by age 28, compared to 65 percent of Hispanics (not a statistically significant difference) and only 46 percent of African Americans. Thus the 23 percent difference in marriage rates between the unwed teen mothers and the other women in our subsample is an artifact of the composition of the two groups: The disproportionate number of unwed teenage mothers who are African American brings down the overall marriage rate of unwed teen mothers.

Within each racial and ethnic group, cumulative marriage rates were roughly equal by age 28. The non-Hispanic whites and the African Americans showed a steady rate of accumulating marriages, but the Hispanics seem to delay marriage until age 23 or 24 when there is a spurt of marriages. This would be consistent with ethnographic studies reporting that Hispanic unwed teen mothers tend to wait a few years after their children's birth before marrying and that, unlike mothers in the other two groups, they tend to marry the fathers of their children.

The only other major research on the subject is consistent with our findings. Neil Bennett of Yale University, David Bloom of Harvard University, and Cynthia Miller of Princeton University estimated the marriage rates of women using the 1987 version of the same NLSY data set that we used, as well as three additional data sets. Parallel to our findings, they documented substantially lower marriage rates for black unwed mothers than for white ones.

Their research also suggests that the gap between the marriage rates of unwed mothers and other women, although large, may be smaller than in the past. When Bennett and his colleagues used the NLSY data set that we used, they, too, found that unwed teen mothers have marriage rates roughly equivalent to those of other women of the same race. They found substantially lower marriage rates for unwed mothers compared to other women of the same race only in the other data sets -- involving older women, many of whose marriages would have occurred a decade or more earlier. This change over time makes sense. If one of the reasons for the increase in out-of-wedlock births is a decrease in the stigma associated with them, then it stands to reason that unwed teen mothers are at less of a disadvantage in the marriage market. And, if unwed parenthood is more common, there may be less adverse selection to it, thereby making the overall pool of unwed mothers more marriageable.

This does not mean, however, that young women pay no price for having a baby out of wedlock. Our data is what statisticians call "right censored." Because our last observations involve women who were only 28 years old, we are comparing the marriage rates of relatively young women. By that age, only three-fourths of the women who are expected to marry have already done so. Thus, as mentioned above, the marriage rate for the other women in our subsample should continue to climb, as more middle-class and better-educated women marry in greater numbers, while the rate for unwed mothers may have plateaued. For example, although many of the women in our subsample were older than 28 at the last interview in 1993 (with some as old as 34), few of the unwed teen mothers married after age 28, while many of the others did.

The persisting cost of out-of-wedlock parenthood is also revealed by racially disaggregated divorce rates. The white unwed teen mothers in our subsample had a divorce rate almost twice as high as that of the other white women (50 percent versus 28 percent), while the black divorce rates were 52 percent and 39 percent, and Hispanic ones about 30 percent and 34 percent (a statistically insignificant difference).

We found one factor that powerfully explained marriage patterns among the women of our NLSY subsample: AFDC recipiency. We calculated year-to-year probabilities of marriage based on the woman's age, years of education, race, urban or rural status, number of children, AFQT score (a measure of skill level and intelligence), income, poverty status as a teenager, marital expectations as a teenager, teen motherhood, and AFDC recipiency. Controlling for all of these factors, AFDC recipiency stood out as the predominant factor (other than race) associated with non-marriage. A woman receiving AFDC was about 30 percent less likely to marry in the ensuing year than a woman who was not.

This finding points to the penalties for marriage that permeate American welfare laws. Marriage automatically disqualifies a single mother from many programs, and, where it does not, even relatively low earnings by a new husband rapidly reduce benefits in a host of programs, including Food Stamps, school meals, and child care. The basic AFDC program, for example, is limited to single mothers and their children. Married couples are not eligible, and, if a welfare mother marries, she is removed from the program. (Married couples can only receive benefits under AFDC-UP, the "UP" standing for "Unemployed Parent." As we will describe later, however, the UP program does not erase the marriage penalty.)

The Urban Institute's Eugene Steuerle calculated the marriage penalty faced by a couple consisting of a man who is working full time at the minimum wage and a mother on welfare with two children. He concluded that, if they marry, the new family's combined earnings plus benefits would be $16,194, or $3,862 (almost 20 percent) less than the couple's premarriage income.

Steuerle's calculation was a theoretical illustration of how large the marriage penalty might be. Recent work by Stacy Dickert-Conlin of the University of Kentucky and Scott Houser of California State University at Fresno begins to explicate the actual scope and distribution of the penalty. They identified a systematic relationship between a wife's characteristics and her husband's earnings, documenting, in effect, patterns of assortive mating. For example, wives with higher than average levels of education tended to marry men with higher than average earnings. Similar relationships were found between the husband's earnings and the wife's age, race, place of residence, and other characteristics. Using this information, they developed a model that predicts the earnings of the men who would likely marry the women in their sample, if the latter were to marry.

Once they had the likely incomes of potential couples, Dickert-Conlin and Houser calculated the effect of marriage on the new family's income, taking into account changes in state and federal income taxes (including the Earned Income Tax Credit), AFDC, Food Stamps, and Supplemental Security Income. They concluded that most couples would lose money by marrying, with the median marriage penalty for a single mother under age 45 being $1,701 -- about 6 percent of the potential couple's total premarriage income.

But this is the penalty for all couples and essentially reflects the relatively modest impact of the income tax, more than compensated for by other economic and emotional benefits of marriage. Marriage penalties were much stiffer for poor women and women on AFDC, whose median penalties were about $3,000 -- or about 15 percent of the couples' premarriage income. Moreover, those were the medians; 25 percent of the couples would have suffered penalties of more than $4,800 -- or 23 percent. And even this estimate does not include the loss of other benefits, such as child-care subsidies, housing assistance, school lunches, and so forth. Where cohabitation rather than marriage is an option, penalties of this magnitude certainly discourage marriage.

Dickert-Conlin applied this model to the actual behavior of couples to see whether the marriage penalty affected matrimonial decisions. Although she was unable to find an association between the penalty and the decision to marry (perhaps because of her small sample), she did find that the marriage penalty encouraged marital separations, especially among poor couples. She found that halving the penalty for couples in their early thirties who face the largest reductions in income would reduce the likelihood of annual separation by 36 percent (from 4.2 percent to 2.7 percent).

Significantly, this effect on separations echoes findings by Bane and Ellwood from two decades ago. In a widely cited study of the impact of welfare on family structure, they used data from the 1976 Survey of Income and Education to conclude that benefit levels have no impact on out-of-wedlock births. Less noted was their finding of a correlation between benefit levels and marital dissolution. Bane and Ellwood found that high benefit levels had only a slight effect on the decision of older mothers to divorce or separate but a large impact on younger mothers. They estimated, for example, that, for both white and non-white mothers under 24, a 50 percent increase in monthly AFDC benefits could "increase the number of divorced and separated mothers by as much as 50%." Speculating about why this might be, the authors suggest that young women's marriages may be less stable than those of older women and, therefore, more susceptible to the impact of a marriage penalty. As they put it: "Welfare may offer an alternative to an unhappy early marriage."

In a separate analysis, Bane and Ellwood also found the effect to be greater for non-whites. For a white mother who married before age 20, that same 50 percent increase in monthly benefits raised the probability of divorce or separation by age 20 from about 8 percent to 12 percent. For a non-white mother, the probability rose from 15 percent to 25 percent.

This brings us back to the question of race. One reason for the racial differences in marriage rates could be that, although similarly situated black and white AFDC recipients suffer equal marriage penalties, the marriage penalty has a far greater impact on the black community. Because black single mothers are one-third more likely than whites to be on AFDC (38 percent versus 28 percent), they face much steeper marriage penalties. According to Dickert-Conlin and Houser, the median single black mother will suffer a marriage penalty of nearly $1,900, compared to the white median of $1,575. Significantly, this represents almost 9 percent of the black couple's earnings, compared to only 4.6 percent of the whites. Again, these are only medians and do not reveal the full force of the penalty on poor black single mothers. More tellingly, black single mothers are 50 percent more likely than white ones to face a marriage penalty that exceeds 10 percent of their income (46 percent versus 31 percent).

These racial differences are magnified among AFDC recipients. The median recipient, of either race, faces a marriage penalty that is about 15 percent of total income, with more than one-fourth of all recipients, of both races, facing marriage penalties that exceed 23 percent of their income. Black women (ages 15 to 44), however, are 5.5 times more likely to be receiving AFDC than white women (20 percent versus 3.5 percent). This means that, across the board, the proportion of all black women facing a substantial marriage penalty is quintuple that of white women.

This finding sheds light on why the overall African-American marriage rate is one-half that of whites, and why, although rates in both groups have been declining in recent years, the black rate has been declining faster. The most popular explanations are the lack of marriageable black males and different attitudes about marriage and responsibility in the African-American community. Our findings lend support to both propositions.

According to the lack-of-marriageable-males argument, a variety of forces -- including low earnings, unemployment, incarceration, and death -- conspire to reduce the ratio of eligible black men to black women. A man's marriageability embraces his ability to contribute to an economically advantageous union. As we have seen, marriage penalties have their greatest impact on black women, which, in turn, would make black men disproportionately less marriageable (compared to white men) than they might otherwise be. (Although this would be expected to result in lower marriage rates for all black women, its greatest effect would be on never-married teen mothers, since their potential mates are likely to be low earners or unemployed.)

Our findings also suggest that the marriage penalties could be having an effect on attitudes toward marriage across a large swath of the black community. Steuerle describes how government policies can have "signaling" effects beyond the individual financial incentives that they create. He points out that, despite the fact that American workers might be expected to want to retire at different ages, they tend to retire at about the same age -- partly because of the signals they receive from the Social Security program. Similarly, if marriage is penalized by the welfare system, and a large enough portion of the community feels the bite, then rearing children outside of wedlock could more easily become the norm. Our analysis reveals just such a signal from AFDC against marriage -- heard five times louder in the black community. That cohabitation is more accepted in many black communities only accentuates this effect.

Marriage is, of course, no panacea: Too many disadvantaged women are already exploited and abused by the men in their lives. But, given the importance of marriage to the economic well-being of so many disadvantaged mothers, one would think that welfare policy would at least not discourage it. But, during the last decade of welfare reform, it has hardly been mentioned. Perhaps it is just too politically incorrect to suggest that mothers should trade dependence on welfare for dependence on a husband.

In any event, the marriage penalty can be reduced in two basic ways: (1) allow married couples to receive welfare or welfare-like benefits, or (2) make welfare less attractive to mothers. Any combination of the two will lower the costs of leaving welfare for marriage.

Expanding AFDC-UP, the welfare program for couples, is frequently suggested as the best way to pursue the first option. About 4.5 million female-headed households are on "basic" welfare, compared to only 350,000 couples on the UP program. One reason is UP's restrictive eligibility rules: The couple's principal wage earner (the parent with the highest earnings) must work no more than 100 hours per month. He (or she) must also demonstrate an attachment to the work force by either (1) having earned at least $50 in each of six or more of the previous 13 quarters, or (2) being eligible for unemployment compensation. In addition, the couple must wait 30 days after the principal wage earner becomes unemployed before receiving benefits and, in many states, benefits are limited to six months a year.

It is widely believed that these rules encourage unemployed fathers to leave home (or claim to leave) so that their wives and children can receive AFDC benefits after their UP benefits run out or they become ineligible because they are working more than 100 hours a month. Recently, more than 30 states have requested, and most have received, federal welfare waivers to lift one or more of these eligibility rules. Such changes may reduce the pressure unemployed husbands feel to abandon their families, but they are unlikely to encourage marriages. The reason is that UP eligibility depends not on marriage but on parenthood. That is, a couple can receive benefits without marrying. Moreover, the UP marriage penalty does not derive so much from work-rule ineligibility as it does from the high marginal tax on the new spouse's earnings (in the form of reduced benefits from AFDC, Food Stamps, and so forth).

Recognizing this underlying disincentive, New Jersey obtained a federal welfare waiver in 1992 which allows it to exclude some of the income of new spouses from the computation of AFDC benefits. When a New Jersey welfare mother now marries someone other than the father of her children, she loses her benefits, but her children can continue receiving all of theirs -- so long as the new spouse's income does not exceed 150 percent of the poverty line for the number of persons in the resulting family. This means that the children's benefits do not completely phase out until family income exceeds $25,000 a year. And, as an incentive for the mother to work, only some of her earnings are counted against her children's benefits.

Unfortunately, New Jersey's innovative approach seems to have had little immediate effect on marriages among welfare recipients. Perhaps more time is necessary for recipients to appreciate these changes. More likely, the reason for so little apparent impact is that New Jersey has only gone half way toward removing the marriage penalty: The couple loses the mother's AFDC benefit, and the Food Stamp phaseout for additional family earnings is not changed.

As suggested by New Jersey's UP program, however, fixing these problems would require a very long phaseout of benefits, reaching beyond families earning $30,000. This would entail astronomical increases in spending and would likely reduce the work effort of all those receiving the benefit.

Another way to reduce the marriage penalty is to change the basis of eligibility for benefits from its present emphasis on family structure to either a children's allowance or a wage supplement. But each alternative would create a new entitlement, itself perhaps a fatal weakness, as well as posing its own particular problems. Basing benefits on the number of children might avoid a marriage penalty but would create an incentive to have children (regardless of marriage) which is even greater than that under current law. A wage supplement would be fraught with design and implementation problems. Not only would the program need to specify who is eligible, the amount of the subsidy, and method of payment, but it would also have to guard against potential windfalls to employers who might lower the wages they pay in response to the subsidy.

These dangers might be acceptable if the changes involved were more likely to increase marriage. But the plain fact is that, although we know that the present system discourages marriage, we do not know the degree to which other social or demographic factors have contributed to the process or would mitigate the impact of changing it. As many have observed, whatever originally caused the decline in marriage, it may have taken on a life of its own.

The second way to reduce the marriage penalty is to make welfare less attractive. A smaller benefit package gives a single mother less to lose by marrying. If the right man is available and willing, and has sufficient earnings, a mother who prefers to stay home with her children rather than work might trade her now lower-value benefit package for the role of housewife, caregiver to her children, and, perhaps, part-time worker.

The most obvious way to reduce the value of the mother's benefit package, of course, is simply to cut its dollar amount. But there are more subtle ways, including work requirements and time limits: Mandatory work decreases the mother's ability to stay home or to work without telling the welfare agency. Time limits take away her benefits in the future.

These and other measures are imposed in the recently enacted welfare reform law, on the premise that they will encourage mothers to leave welfare for work. But, if the marriage penalty is as real as the data presented above suggest, such measures should also increase the welfare exits due to marriage, perhaps substantially.

Until now, however, evaluations of welfare programs have focused on whether employment increased. Evaluations of the new welfare regime should also gauge its impact on marriage. If this reform of welfare not only increases employment but also leads to more marriage, that would be frosting on the (wedding) cake.