Welfare Reform: Four Years Later

by Douglas J. Besharov & Peter Germanis

This article appeared in the Summer 2000 issue of the Public Interest.

        Four years ago this August, a Republican Congress pushed a reluctant President Clinton to sign a bill that ended welfare as we had known it. But since the 1996 welfare-reform act expires on September 30, 2002, its eventual fate is not yet clear. Much will depend on how the law's impact is viewed. So far, it certainly seems to be a success. By June 1999, welfare rolls had fallen an amazing 49 percent from their historic high of five million families in March 1994. That's nearly seven and a half million fewer parents and children on welfare.

        It has suited the purposes of both the Clinton administration and the Republican Congress to claim that "welfare reform" has caused this dramatic decline--and that over two million former recipients are now working because of the new law. But that's not quite true. The strong economy--and massively increased aid to the working poor--almost certainly have had more impact than welfare reform per se. Moreover, as many as 40 percent of the mothers who left welfare are not working regularly but are instead relying on support from boyfriends, family members or friends, and other government and private programs.

        Both liberals and conservatives have found it convenient to ignore this reality--conservatives because it gives the "Clinton economy" and the president's success in expanding aid to the working poor too much credit and Republican welfare reform too little, and liberals because it suggests that many welfare recipients didn't "really need" government benefits. But the failure to be clear about why the rolls have declined so much prevents an accurate accounting of the law's impact--and what needs to be done next.

Welfare's rise and fall

        For nearly 60 years, it seemed that welfare rolls could only grow. With the exception of a few short-lived declines, the rolls grew from 147,000 families in 1936 to about 5 million in 1994, from less than 1 percent of all American families with children to about 15 percent.

        Between 1963 and 1973, there was a striking 230 percent increase--not because of a bad economy (unemployment was actually quite low during most of this period) nor simply because of an increase in family breakdown (both divorce and illegitimacy were rising, though not nearly as fast as the welfare caseload). Rather, the increase was largely the result of programmatic changes that made it easier for income-eligible families to get benefits, as well as the destigmatization of being on welfare. Where once welfare agencies discouraged applicants (by pressing them to seek other means of support or by imposing a grueling eligibility process), the obstacles to enrollment were now lowered. New York City's rolls almost tripled in only five years (between 1965 and 1970) under liberal mayor John Lindsay. The same liberalization was taking place across the nation, as welfare came to be seen more as a "right" than as a temporary safety net. Some of the drive behind this national movement was undoubtedly the long-overdue repeal of Jim Crow-like rules in the South that kept African-American mothers off welfare.

        After this liberalization, caseloads stayed roughly steady for almost 15 years. They rose again, by 34 percent, between 1989 and 1994, largely because of the weak economy. But there were other important causes: a spike in out-of-wedlock births among some groups; an increase in immigrants applying for means-tested benefits, either for themselves or their American-born children; half-a-decade's outreach efforts to get single mothers to sign up for Medicaid (and thence welfare benefits); and an increase in child-only cases, perhaps as a result of the spread of crack addiction among mothers and an increase in cases of parental disability.

        Regardless of what caused rolls to rise in the past, they rarely fell back very far. Thus no one predicted the recent halving of welfare since 1994. Fifteen states have had declines of over 60 percent; three report declines of 85 percent or more. Indeed, almost everywhere, welfare rolls are way down and work is way up. For example, never-married mothers, the group most prone to long-term welfare dependency, were 40 percent more likely to be working in 1999 than in 1994. What's responsible for the decline in welfare and the increase in work?

The end of welfare as we know it

        In 1992, Barbara Sabol, then New York City's welfare commissioner, visited two of her own welfare offices dressed in a "sweatshirt, jeans, and scarf or wig." She told the welfare workers she needed a job in order to care for her children. But try as she would, she could not get the workers to help her find a job.

        The same year, candidate Bill Clinton showed that he was a New Democrat by ambiguously promising to "end welfare as we know it." After the election, his administration granted many state waivers that, among other things, toughened work requirements and imposed partial time limits on benefits--ultimately culminating in the Republican-inspired 1996 welfare-reform law.

        T he Republican bill bore a superficial resemblance to what Clinton proposed, so both sides were able to claim credit for reforming welfare. But the changes in welfare were largely based on the Republican plan. While both bills imposed time limits on benefits, the Clinton proposal included an entitlement to a public job afterward. The Republican bill had no such entitlement, and also transformed the program into a capped block grant, which gave states an incentive to cut caseloads because they would get to keep any unexpended funds.

        Today, Sabol would find that welfare workers are eager to find jobs for their clients. Across the nation, the culture of welfare offices has changed--from places where mothers are signed up for benefits (with almost no questions asked) to places where they are helped, cajoled, and, yes, pressured to get a job or rely on others for support. The U.S. General Accounting Office described the change this way:

        Under states' welfare reform programs, participation requirements are being imposed sooner than under JOBS [the old welfare regime], with many states requiring participation in job search activities immediately upon application for assistance. Before reform, recipients could wait months--or even years--before being required to participate, and many never were required to participate because of the lack of sufficient services and staff.

        Many welfare offices are now "job centers," where workers help applicants and recipients find employment. Depending on the office, they teach how to write r�sum�s and handle job interviews; provide access to word processors, fax machines, telephones, and even clothes; offer career counseling and financial-planning services; and refer applicants to specific employers with job openings. In a survey of former welfare recipients in Texas who left the rolls in December 1996, over 60 percent said the welfare agency "gave me the kind of help I needed."

        Some of this is boosterism, plain and simple, with welfare workers giving young mothers the moral support they so often need. As one caseworker said, "Some of these women never thought that they could get a job. We give them the confidence to try." But the assistance also can be quite concrete. Besides large expansions in Medicaid and child care, many states provide cash assistance to families on welfare to help them leave or stay off. These payments range from a few hundred dollars to over $2,000. For example, Texas provides stipends to help such families pay for employment-related expenses, such as transportation, education, and training. Virginia gives transportation allowances for up to a year after leaving welfare. And about a dozen states have created or expanded EITC-like tax credits for low-income families which can be used for any purpose.

        In a real break from the past, however, few welfare agencies seem to put any stock in job training. Administrative data from the states indicate that less than 2 percent of adult welfare recipients are in some sort of formal job-training program. Instead, agencies emphasize immediate job placement and on-the-job work experience. This gives mothers much-needed work experience, but it also adds to the pressure they feel to leave welfare or not apply for it in the first place. For there is also a sharper edge to welfare reform.

        In most places, the welfare application process has a new element: "diversion." Diversion is a straightforward effort to keep families off welfare. It is encapsulated in two simple questions now asked of welfare applicants: Have you looked for a job? Can someone else support you? Many welfare agencies maintain a bank of phones that applicants must use to call as many as 20 potential employers before they can even apply for benefits. When told of these requirements, many applicants simply turn around and walk out.

        In New York City's "Job Centers," for example, all applicants are encouraged to look for work (and offered immediate cash support for child care) or to seek support from relatives or other sources. Those who still decide to apply for welfare are required to go through a 30-day assessment period during which they complete the application process and undergo a rigorous job-readiness and job-search regimen involving many sessions at the Job Center and other offices. At the end of this period, eligible able-bodied adults who choose to receive assistance are required to participate in the city's workfare program. As a result, New York City officials estimate that the percentage of mothers entering these Job Centers who are eventually enrolled has fallen by about 40 percent, from about half to about a third of applicants.

The hassle factor

        Being on welfare has also changed but not as much as many people think. When welfare reform was being considered by the Congress, most analysts expected the states to establish large, mandatory work programs in order to satisfy the bill's "participation" requirements. However, because those requirements were set in relation to 1995 caseloads, the sharp decline in the rolls since then has obviated the need for such programs--and few places beyond Wisconsin and New York City have established them.

        Instead, almost all states require recipients to sign "self-sufficiency agreements" describing their plan for becoming self-sufficient within a specified time frame. Iowa, for example, requires all able-bodied recipients without infant children to develop and sign a Family Investment Agreement. Failure to sign or comply with this agreement can result in immediate and complete termination of cash assistance. About 10 percent of those who begin this process seem to have their benefits terminated for failure to sign or comply with the agreement.

        In addition, most states now impose various behavior-related rules. Parents are required to have their children immunized and to send them to school; in a few places, mothers and fathers must even attend family or parenting skills classes. Failure to comply with these requirements can result in the welfare grant being reduced--and, in about 37 states, even terminated. According to state administrative data, in 1998, 6.2 percent (or 180,000) of the 2.9 million families that left welfare did so after a sanction. In some states, the percentage was as high as 30 percent.

        These and other new requirements raise what economists would call the "cost" of being on welfare. By a rough calculation that assumes recipients value their time at the minimum wage, these kinds of requirements can reduce the advantage of being on welfare versus working by about 50 percent. In very low-benefit states, the advantage can fall to zero.

        This amounts to the reintroduction of a long-gone aspect of being on welfare, "hassle." And it clearly leads some welfare recipients to seek other ways of supporting themselves. When these new requirements are explained to applicants and recipients, they often say things like: "I guess I might as well get a real job" or "I might as well move back home." Or they just walk out of the office--or stop responding to warnings that they will lose their benefits if they do not participate in work-related activities.

        In the 1996 Texas survey of former recipients, about a quarter said that important factors for leaving were either "unfriendly caseworkers" or "new program requirements." And in a survey of those who left welfare in South Carolina between January and March of 1997, 60 percent said they felt "hassled," and 13 percent said that is why they left. About a third said that the state's welfare program "wants to get rid of people, not help them." A similar survey was conducted in Wisconsin for those who left welfare in 1998, and the results were about the same. (Of course, hassle may have led others to leave welfare, though they cited some other reason, such as finding a job.)

        These are dramatic changes in welfare, and it is natural to assume that they are responsible for recent caseload declines. But welfare reform has coincided with the strongest economy in at least three decades, coupled with an unprecedented increase in aid to the working poor. The increased returns for low-skilled work are probably as responsible for the decline in welfare, and perhaps more so.

Work pays

        The strong economy, most experts agree, has played a key role in the welfare declines. The rolls started falling in 1994, two years before the enactment of the 1996 welfare law, and before the welfare waivers that allowed some states to "get tough" on welfare recipients could have had much impact. The weak economic conditions that helped drive up welfare rolls during the Bush presidency ended a few months before George Bush left office (not soon enough, of course, to affect the election). Since January 1993, the economic news has been truly remarkable: Real per capita Gross Domestic Product up about 25 percent in real dollars, 20 million new jobs, the highest ever employment-to-population ratio (64.3 percent), and the lowest unemployment rate since 1970 (4.1 percent).

        Most relevant to the welfare decline has been the increase in average real earnings, especially among low-wage workers. For example, since the second quarter of 1996, weekly earnings for full-time workers have grown 5.3 percent. The gains for low-income, full-time workers have been even larger: 7.0 percent for those at the 25th percentile of earnings and 8.5 percent for those at the 10th percentile of earnings.

        Also helping to reduce caseloads has been the progress in fulfilling Clinton's promise "to make work pay." Both Democratic and Republican Congresses have supported (the latter sometimes reluctantly) Clinton's initiatives for massive increases in government aid to the working poor. So much so that this spending now far exceeds what was spent on the old AFDC program. Between 1993 and 1999 alone, total aid to the working poor increased by about $25 billion a year, from about $40 billion to about $65 billion (in 1999 dollars). At its height, combined federal and state spending on AFDC never exceeded $30 billion.

        The Earned Income Tax Credit (EITC), for example, provides a cash subsidy to low-income, working parents. Between 1993 and 1999, total expenditures on the EITC rose from $18 billion to $30 billion (all in 1999 dollars). The increases for particular groups were striking: For example, the income supplement for a single mother (with two children) working at the minimum wage more than doubled, rising from about $1,700 to about $3,900 per year.

        Child-care aid has also expanded, becoming all but an entitlement for those families leaving welfare. Total annual federal and related state child-care expenditures rose from $8 billion to over $12 billion in the same years, providing child-care slots for well over one million additional children. Gaps in coverage remain, and take-up rates may be lower than many advocates say are appropriate (although the latter is probably because so many parents already have access to government subsidized child care or have other family members who can care for their children). In any event, subsidized child care is obviously helping many low-skilled and low-earning mothers work.

        Medicaid eligibility, too, has been substantially expanded. While Medicaid was once limited primarily to families receiving welfare, sequential expansions for pregnant women and children (beginning in the mid 1980s) have taken eligibility to between 100 and 250 percent of the poverty line (depending on the child's age and the state program). The welfare-reform law gave states authority to expand coverage for adults, and some have done so. As a result, total Medicaid and related health-care costs for low-income families with children rose from $15 billion in 1993 to $24 billion in 1999--making millions more children (and sometimes families) eligible.

        The absence of health-care coverage is not an insuperable barrier to work for mothers who are healthy and who have healthy children. But for those mothers who have chronic illnesses, or whose children have them, the threat of losing coverage can be a substantial disincentive to leaving welfare.

        Clinton also managed to push through the Republican-controlled Congress a two-stage increase in the minimum wage--from $4.25 an hour to $4.75 an hour on October 1, 1996, and then to its current $5.15 an hour on September 1, 1997. Moreover, additional expansions in aid to the working poor are looming. The child tax credit and Child Health Insurance Program will grow as they are fully phased in. The Clinton administration has proposed further expansions in child care and the EITC, and has taken steps to expand participation of low-income working families in Medicaid and Food Stamps. It also continues to push for increases in the minimum wage.

Explaining the decline

        A number of respected researchers have used econometric models to estimate how much of the caseload decline was caused by welfare reform itself compared to the economy and increased aid to the working poor. The models they use, unfortunately, are extremely sensitive to the assumptions and variables incorporated, making their findings imprecise. Nevertheless, most of the studies lead to a similar conclusion: In the early years of the caseload decline (1994-96), around 40 or 50 percent of the decline was due to the economy and the stronger job prospects for low-skilled workers. Later in the economic expansion (1996-99), the same and other studies conclude that the economy accounted for only about 10 or perhaps 20 percent of the decline.

        Many studies also attempt to gauge the impact of the increased aid to the working poor. This decade-long effort to "make work pay" may be an even more important factor in the declining rolls, accounting for perhaps 40 to 50 percent of the initial declines and 30 to 40 percent of the later declines.

        As for welfare reform itself, these studies usually estimate its impact at from 15 to 20 percent for the early declines and about 30 to 40 percent for later ones. (Most studies also find that the failure to increase welfare benefits, a 20-year trend, reduced rolls another 5 to 10 percent.) Consolidating the estimated impacts on initial and later declines (and weighting them for the size of each), here is what the studies suggest: The robust economy explains 15 to 25 percent of the decline; aid to the working poor 30 to 45 percent; increases in the minimum wage 0 to 5 percent; and welfare reform 30 to 45 percent.

        Placing too much confidence in the results of such econometric models is always questionable. These studies have many weaknesses, including the failure to include all policy changes, such as heightened child-support enforcement. Most also fail to consider at least partially independent demographic factors, such as declines in out-of-wedlock births, drug abuse, crime, and immigration. We doubt, for example, that increased aid to the working poor has had as much impact as they suggest; we would attribute more of the decline to the strong economy and to welfare reform generally. Nevertheless, these studies were carefully conducted, and their results are roughly consistent. Thus it seems reasonable to conclude that they correctly reflect the approximate contribution of these four factors to the decline in caseloads.

        It is possible that welfare reform has played a more substantial role, interacting with the strong economy and more generous aid to the working poor to encourage more single mothers to try working. By this way of thinking, the strong economy and more generous aid are necessary but not sufficient conditions--with welfare reform providing the needed motivation for people to seek jobs or the support of others who have jobs. After all, we have had strong economies in the past without concomitant welfare declines; sometimes welfare rolls have even risen. In other words, the impact of each of the factors may be greater than would otherwise be the case if they had occurred alone.

        The results of randomized welfare experiments, however, seem to confirm econometric estimates of welfare reform's only partial role in reducing caseloads. Starting in earnest in 1992, states were granted waivers from the old AFDC rules, but only if they established rigorous, random-assignment experiments to measure the impact of their new policies. Many of these new policies bear a close resemblance to the program restrictions in new-style state welfare regimes, such as tougher work requirements, time-limited benefits, family caps, and linking benefits to immunization and school attendance. The experiments also reflect many of the expansions in benefits that characterize welfare reform, such as liberalized resource limits, transitional benefits, eligibility for two-parent families, and earnings disregards (which allow working recipients to keep more of their benefits). About 10 of these experiments have yielded findings that provide an indirect measure of welfare reform's impact on caseloads.

        Across all 10 of these waiver studies, and regardless of the varying combination of program components, the difference between the experimental and control groups is rarely more than a few percentage points. The biggest declines in welfare receipt due to welfare reform do not exceed 15 percent or so, often over two or three years. This does not mean that welfare reform's contribution to the decline is only 15 percent. We recognize that these randomized experiments are an imperfect measure of welfare reform's potential impact because they do not capture either its role in discouraging people from going on the rolls ("entry effects") nor its broader impact on personal and agency behavior (partly through a change in community values). The point is that no rigorously evaluated program of welfare reform has ever had an impact even remotely comparable to what has happened to national welfare caseloads.

        Indeed, sometimes the group receiving the "reformed" welfare services was less likely to leave the rolls. This is because most of the waiver experiments, like most state-implemented welfare reforms, include components that both decrease caseloads (like work mandates) and increase them (like the expansion of earnings disregards). Thus Minnesota's "welfare reform," which expanded the state's earnings disregard, asset limits, and two-parent eligibility for benefits while imposing modest work requirements, increased caseloads by almost 5 percent for long-term recipients after 21 months.

Leaving welfare without working

        In addition to the often unappreciated contribution of the economy and aid to the working poor, another significant aspect of the caseload decline is that so many mothers seem to be leaving welfare without taking jobs.

        The best source of data about the families that have left welfare are surveys of former welfare recipients ("leaver studies") that have been conducted by various states and by the Urban Institute. Although they all have some weaknesses, such as low response rates and insufficiently detailed information, the best studies tell about the same story: Between 60 to 70 percent of those who left welfare were employed at the time they were surveyed (and 60 to 85 percent had been employed at some point since leaving). Of those who were working, about 60 to 80 percent seem to work full-time, earning about $6 to $8 per hour (or about $800 to $1,000 per month). The remainder worked fewer hours and thus earned less money. (Many studies, however, exclude the 20 to 30 percent of leaver families that returned to welfare, which tends to minimize the difficulty some mothers have finding work.)

        Broader measures of employment are consistent with this high level of nonwork among leavers--and also suggest that many of the single mothers who did not go on welfare are also not working. For example, between March 1994 and March 1999, the number of employed single mothers with children under age 18 increased by 1.251 million (from 5.712 million to 6.963 million). During the same period, welfare caseloads (almost all including a single mother) fell by 2.430 million (from 5.098 million to 2.668 million). Even if the entire 1.251 million increase in the number of single mothers working in this period represented those who were previously on welfare (or would have gone on welfare during the period), they would still amount to only about half of the caseload decline.

        Some mothers who left welfare, of course, may not be reporting their employment. A four-city study conducted by researchers Kathryn Edin and Laura Lein in the early 1990s found that about 30 percent of low-income working mothers and about 50 percent of welfare mothers had unreported work. But there is no reason why the percentage not reporting work should have grown in recent years. If anything, the expansions in earnings disregards and the EITC should have encouraged more low-income mothers to report their employment.

        Thus only about 50 to 60 percent of the mothers who have left welfare (and stayed off) seem to be working regularly. The surprisingly large number of mothers leaving welfare without then working has been all but ignored by most commentators, including severe critics of welfare reform. Yet this has profound implications for the economic and social condition of low-income families.

Other sources of support

        How could so many mothers have left welfare without working? Work requirements and heightened levels of "hassle," by lowering the value of welfare, would be expected to cause mothers to leave welfare for work, even relatively low-paid work. But why would mothers leave welfare without having a job? The burdens placed on them hardly seem sufficient reason for them to abandon their only support for themselves and their children.

        The leaver studies suggest the answer: These mothers have other sources of support besides welfare. In South Carolina, for example, nonworking leavers were almost twice as likely as working leavers to have other sources of support--including other forms of government assistance, such as Social Security (13 percent vs. 6 percent) or SSI (20 percent vs. 8 percent); free housing from a parent or relative (15 percent vs. 10 percent); another adult in the home to help with the bills (17 percent vs. 7 percent); and help from someone outside the home (22 percent vs. 8 percent). A study of former recipients in Milwaukee, conducted by the Hudson Institute and Mathematica Policy Research, found that over two-thirds of all the mothers who left welfare received help (e.g., transportation assistance, a place to stay, and food) from family or friends. Those leavers who were not working were about 15 percent more likely to be receiving such help (72 percent vs. 63 percent).

        Most leaver studies do not identify the sources of support for working and nonworking mothers separately, but they do reinforce the importance of other household members or income sources. In Iowa, after families were dropped from welfare, they were about 33 percent more likely to be relying on others for a place to stay (25 percent vs. 33 percent). Similarly, in Florida, where families have begun to lose welfare due to a time limit, one-third of those who hit the time limit either moved or had a different living arrangement, such as adding another household member to help with the expenses. Finally, in Connecticut, 43 percent of the families that left welfare due to the state's 21-month time limit reported living with at least one other adult six months after benefit termination. (There is no comparison data for the period before the time limit was reached.)

        When welfare reform was being debated, many experts predicted increases in such "coresidency" or doubling up arrangements as mothers were pushed off welfare. So far, however, there is little evidence of substantial increases in coresidency (or marriage, for that matter). According to Christopher Jencks, for example, the total number of single mothers residing with another adult has remained essentially stable since 1988, with no discernible change after welfare reform. It is possible that many mothers entered such living arrangements while their total number remained constant--with as many mothers having left such arrangements because of the improving economy as entered them because they were pushed off welfare. But without more data, it is impossible to know for sure.

        There is another way that mothers can leave welfare without working: They can fall back on preexisting coresidency arrangements (together with other sources of support). Based on a study by Rebecca London, which used data from the Survey of Income and Program Participation, we calculate that, in 1990, before the declines in welfare caseloads, at least 37 percent of welfare mothers lived with other adults--18 percent with their parents, 6 percent with a boyfriend, and 13 percent with others.

        These findings may seem surprising, but for many years now the welfare system has largely ignored the household income in such coresidency arrangements. Depending on the situation, the income of the grandparents, with whom the adult welfare mother was living, would not be considered (for example, if the mother was an adult herself); and the man-in-the-house rule (which denied benefits to households with a cohabiting male) was abandoned years ago. We would suggest that when faced with the new work and behavioral requirements, mothers who had other sources of support sufficient to permit them to forgo welfare (predominantly those living in households with adequate economic resources) simply left welfare without looking for work.

        It also helps that many of these mothers are still receiving other government benefits--primarily Food Stamps and housing--which are often much more valuable than the basic welfare payment. (The continued availability of Medicaid also encourages mothers to leave welfare without finding work, even if the family does not sign up for coverage until someone takes ill.) Nonworking mothers on their own could not subsist on only these benefits, but nonworking mothers living with others (or getting support from others) could get by.

        This is in particular the case in low-benefit states where it may simply no longer "pay" to be on welfare as opposed, say, to Food Stamps. In Alabama, for example, in 1999, the welfare benefit for a family of three was just $164 per month, compared to a food-stamp allotment of $329. (Moreover, the food-stamp benefit comes with virtually no strings attached, whereas cash assistance can be accompanied by work and other behavioral requirements that reduce its value still further.) So mothers in low-benefit states can leave welfare and not suffer anything like a complete loss of income, especially if there are other adults in the household that have income.

        This makes economic sense. If one assumes that these mothers value their time at the minimum wage or above, then there is little incentive for them to engage in work activities for 20 to 30 hours per week to avoid a sanction that can be as little as $10 to $50 per week. The added income from complying with these requirements translates into an effective wage of fifty cents to two dollars per hour. This does not compensate for the lost free time (what economists call "leisure") that mothers can use, for example, to care for their children or take a job with unreported income. Data on this behavior are difficult to obtain, but its possible magnitude is suggested by the following: Between 1994 and 1998, the number of single-parent families on Food Stamps that were both not on welfare and with no earnings grew by 10 percent, or 55,000 families. While this is an imperfect measure, it could easily understate the phenomenon.

        The economy and aid to the working poor could also play a role here, as more households would become economically comfortable enough for the mother to leave welfare without working. This would be consistent with earlier patterns. Greg Duncan of Northwestern University and his colleagues used data from the Panel Study of Income Dynamics (PSID) to determine why mothers left welfare between 1986 and 1991. (Leaving welfare was defined as receiving welfare in one year but not the next year.) They found that about one-half of welfare exits were for work (or a rise in earnings), about a quarter for changes in marital status or living arrangements, about 5 percent because there were no children under 18, and the remainder were due to a variety of reasons such as an increase in other transfer income or a change in state of residence. About one-third of the earnings-related exits involved an increase in the earnings of an adult already in the household other than the mother, demonstrating the importance of shared living arrangements.

        These dynamics also explain the behavior of those mothers Larry Mead of New York University calls the "happily sanctioned." Such mothers accept less in welfare benefits rather than work or meet other behavioral requirements. In about 14 states, which include about half the national welfare caseload, the sanction for noncompliance is only a partial reduction in benefits--that is, the family's grant is reduced by some percent, usually representing the mother's share of the grant (about one-third of the welfare check). These mothers may not actually be happy, but since this reduction typically amounts to only one-sixth of the mothers' total package of benefits, one can see why they willingly make the trade-off.

Assessing "welfare reform"

        When congressional Republicans were pushing for the enactment of their welfare-reform bill, opponents predicted widespread hardship--including sharply increased homelessness. Happily, there is no evidence that welfare reform has caused substantial increases in homelessness or other indicators of extreme hardship, such as foster-care placements or substantiated reports of child abuse and neglect. And despite extensive efforts, journalists have found few individual horror stories of the harmful effects of welfare reform. As one administrator said, "We underestimated their ability to get jobs that meet their basic needs--or to get support from other sources."

        For a while, it appeared that incomes of the poorest single mothers might be edging down--a sign that welfare reform might be squeezing those at the bottom. A widely disseminated study by Wendell Primus of the Center on Budget and Policy Priorities estimated that, from 1995 to 1997, the bottom quintile had experienced an 8 percent drop in income. Even though many of these mothers were not welfare leavers (nor likely to have gone on assistance before welfare reform), advocates latched onto this income decline as a sign that welfare reform should be reconsidered. However, Primus's further analysis after another year of data has reduced the estimated income loss for this group to about 4 percent. In the same period, 1995 to 1998, all the other quintiles of income for single mothers rose, with the middle quintile up 7 percent, going from $20,617 to $22,063.

        But if welfare reform has not been the social catastrophe some predicted, neither has it lifted large numbers of female-headed families out of poverty. Thanks to the EITC and other aid, most of the mothers who left welfare and are working now have more income than when they were on welfare. But average earnings are only about $12,000 a year. And as we saw, many mothers simply left welfare--without working. Unless they lived with someone or moved in with someone earning a great deal more, they probably suffered at least a partial loss of income.

        Moreover, some of those who gained income through work may not be immediately better off. They also have more expenses. Even if their child-care costs are fully covered, they still face other work-related expenses, such as transportation and clothing. And by working, they lose the ability to earn additional money off the books. Thus their higher income comes at the price of having to work many hours a week while also raising their children, often on their own.

        There are many weaknesses in the data that underlie the foregoing conclusions. For example, it is very difficult to find and count the number of homeless families and individuals, much less get detailed information on their characteristics. Data on substantiated cases of child maltreatment are a function of the number of reports received, the ability of the system to investigate them, and the willingness of states to report them. Even the much-cited income data used to measure trends in financial well-being are plagued by numerous problems. Perhaps most significantly, reported welfare receipt in the Current Population Survey (CPS) is over one-third lower than indicated by administrative records, a problem that has been worsening in recent years. These surveys also miss much of the income that is earned working "off-the-books" or received from boyfriends or other household members. For example, recipients may want to conceal this income from those administering the survey, for fear that it could affect their eligibility for welfare.

        This mixed picture of life after welfare is captured in the "before and after" questions asked in six of the more reliable leaver studies (Mississippi, New Mexico, South Carolina, Virginia, Washington, and Wisconsin). Depending on the study, between 20 to 40 percent of those responding said that "life was better" while on welfare. Conversely, 60 to 80 percent of former recipients think that life is the same or better being off welfare. (Three states asked separately about being better off. In all three, about 55 percent said they were "better off"; about 25 percent said they were doing the "same"; and about 20 percent said they were "worse off.")

        What should we make of these patterns? First, reducing welfare rolls is a tremendous and unprecedented achievement--especially given the apparently small amount of additional hardship. If this result had been guaranteed when welfare reform was being debated in 1996, most opposition would surely have melted away. Indeed, even some past opponents of welfare reform have been quieted by its apparent early success. Nevertheless, welfare reform should not be given credit for the consequences of a stupendous economy and unprecedented increases in aid to the working poor. More of the mothers who gained ground after leaving welfare can probably thank the latter two factors for their improved situation, and more of those who lost ground probably left assistance because of welfare reform and the added hassle associated with it.

        What about those mothers who are now working but not making much more than their previous welfare benefits, or those who are now relying on the support of others rather than welfare? Robert Haveman, an economics professor at the University of Wisconsin-Madison, says that they are "treading water, but staying afloat." We hope they are not just treading water, but building their skills or living in households where the prime earner is doing so.

        For most Americans, welfare reform was not just about reducing the rolls, nor was it some silver bullet that would immediately eradicate poverty. Instead, it was about reducing the deep-seated social and personal dysfunction associated with long-term dependency, thereby ultimately reducing poverty. For welfare reform to be a success on this measure will depend on whether the low-paying jobs taken by many leavers lead to better jobs, whether the household arrangements (and other sources of support) that have allowed mothers to leave welfare without working prove supportive and nurturing, and whether the eventual result is less dysfunctional behavior among parents and better outcomes for children. We may need a generation to find out.