Through recent decades, America's social welfare policies have oscillated between two contradictory impulses. The 1960s were marked by a campaign against poverty; in the 1980s, welfare policy was increasingly concerned with fighting dependency. By the early 1990s, when welfare rolls hit an all-time high, the fear of unintended consequences--that welfare was discouraging work and marriage, and encouraging out-of-wedlock childbearing--led to reforms that limited how long someone could receive welfare benefits, the incorporation of strict work requirements, and, thus, the end of welfare as we knew it.
At issue is an age-old conundrum: The jobless poor need cash benefits to escape destitution and to have any hope of moving up. Yet cash benefits can reduce the incentive to work (and, some believe, to marry)--especially if policies make the welfare package more attractive than a low-wage job (or a low-wage husband). Under the old welfare system, many families ended up no better off financially for having gone to work. Beyond losing cash benefits, families that suddenly moved from welfare to work incurred new expenses such as transportation and child care, and also risked losing government health care benefits.
Would it be possible to design a system of cash support that encouraged work, reduced poverty, and improved the well-being of families and children without deepening welfare dependency? A system in which recipients could retain both necessary in-kind services such as child care and cash subsidies for an extended period as they became experienced workers?
The Manpower Demonstration Research Corporation (MDRC), a nonprofit group that conducts demonstration programs and evaluations, has worked with federal, state, and local policy makers to launch three large-scale experiments aimed at determining what difference work incentives packaged together with other changes (including elimination of various disincentives to work and marry) would make in rewarding work, minimizing dependency, and lifting workers and their families out of poverty. The experiments began prior to the 1996 welfare reform legislation mandating time limits on welfare, but the just-released three-year follow-up findings from two of the experiments shine new light on the triple challenge of increasing work and reducing both dependency and poverty in this era of time-limited welfare.
Acting independently of one another, policy makers in Minnesota, Wisconsin, and Canada set out to design and test alternatives to the traditional safety net system: Rather than paying people when they did not work, these alternatives aimed to support people when they did. The evidence from ongoing rigorous evaluations of these programs conducted by MDRC is encouraging: When cash supplements were paid on top of earnings, particularly to long-term welfare recipients, these programs increased employment and income, lowered the fraction of families with below-poverty-level incomes, reduced the incidence of divorce, and generally improved family and child well-being.
The results have important implications. More than 40 states now attempt to bolster low-wage workers by disregarding some earnings when calculating welfare benefits. Pennsylvania disregards 50 percent of all earnings, while California disregards the first $225 plus 50 percent of remaining earnings. In effect, this liberalization of "earned-income disregards" allows welfare recipients to work and receive cash payments that supplement their earnings from entry-level jobs. These little-noticed "make work pay" provisions were part of the time-limited welfare changes states made in reaction to the 1996 federal welfare reforms--suggesting that states were as concerned about poverty reduction as they were about dependency reduction. States also made changes designed to keep families together, namely elimination of the rule requiring two-parent families to lose welfare eligibility as soon as one parent works more than 100 hours a month, regardless of earnings.
While the founding impetus was different for each of the make-work-pay programs MDRC studied--the Minnesota approach was launched by state government; in Milwaukee, Wisconsin, a coalition of community groups set up a targeted program for specific areas of the city; and in Canada, two provinces worked in partnership with the national government--these three shared an underlying approach: They attempted to make low-wage work pay by providing work incentives in the form of monthly cash payments. These payments were made only when people worked, and the amount of each month's cash payment depended on that month's earnings.
These were not work-for-your-benefits "workfare" models. Rather, the idea was to create real, monetary incentives--to increase the payoff of low-wage work through work-conditioned earnings supplements. All three programs encouraged welfare recipients to work full time (at least 30 hours a week), either by tying incentive payments to full-time work or by requiring those who were not working full time to participate in job preparation services. Aggressive outreach and communication about the value of the incentives were hallmarks of these programs. In each place, of course, the welfare safety net system continued to operate. In Canada and Milwaukee, where the programs operated outside of the welfare system, people who lost their jobs (and thus their work incentive payments) could return to welfare by reapplying. In Minnesota those who lost their jobs were automatically converted back to the traditional welfare benefit structure, without applying.
Here's a closer look:
The Minnesota Family Investment Program (MFIP) used the welfare system (which combined food stamps and welfare into one cash grant) to reward work by changing the way earned income was treated: It increased both the "earned-income disregard" and the basic benefit level for workers. These changes had the effect of increasing a worker's income by as much as $250 a month above what the basic welfare system would have paid someone taking a job. In addition, for long-term welfare recipients who were not working at least 30 hours a week, MFIP required participation in employment-focused services designed to help them find jobs. Failure to participate could result in a 10 percent grant reduction. For two-parent families, the program also eliminated the 100-hour rule. MFIP first operated as a pilot program; in modified form, it became Minnesota's statewide welfare program in 1998.
A family assigned to MFIP would meet with an eligibility specialist to understand how benefits would be affected when work began. Caseworkers enthusiastically conveyed the "work pays" message; for the first time, they could honestly tell recipients they would clearly be better off if they went to work. Long-term recipients--for two years or more--who were working less than 30 hours a week were then referred to an MFIP-contracted service provider and a case manager who would regularly reinforce the "work pays" message and help develop an employment plan that could include education and training. For two-parent families, the participation requirement kicked in after six months. When a parent got a job, MFIP would help arrange and pay for child care, continue her Medicaid benefits, and establish a new benefit structure to supplement her earnings.
The Canadian Self-Sufficiency Project (SSP) tested a work-based alternative to welfare that paid a substantial monthly earnings supplement for up to three years--as much as $400 a month above the amount they would have received if they had worked and remained on welfare--to long-term, single-parent welfare recipients who left welfare and worked full time (at least 30 hours a week). SSP operated outside the welfare system; it was administered by private agencies in Vancouver, British Columbia, and parts of New Brunswick. It ran from 1992 through 1999. Participation was voluntary.
Single parents who had been on welfare for a year or more were contacted by a private social services agency and invited to a three-hour workshop where worksheets and exercises were used to explain how SSP would affect their income if they found full-time jobs and left welfare. The typical recipient who received about $10,000 a year on welfare was shown that if she took a job for 30 hours a week or more, her earnings plus her SSP supplement would total about $20,000, assuming she worked full time for the entire year. Once working, each month she would complete a voucher, attach her pay stub, and submit it to the SSP office, where staff would verify that total hours met the 30-hour-a-week eligibility threshold, and then calculate her supplement payment and deposit it in her bank account.
Milwaukee's New Hope Project was a community antipoverty initiative designed to test a comprehensive set of financial and other supports for low-income workers. It was open to all low-income people living in two target areas and offered a package of incentives consisting of earnings supplements plus child and health care subsidies and--for people who could not find jobs--access to temporary community service jobs. Participants had to work at least 30 hours a week to receive New Hope's incentives package, and they could be eligible for up to three years. The program operated from 1994 through 1998.
New Hope staff flooded the neighborhood with flyers and went door to door, reaching out to working and nonworking residents. Individuals who enrolled and were employed full time were asked for documentation of hours worked and wages, and interviewed about their child care and health insurance needs, after which staff calculated benefit payments, signed each recipient up with one of the program's health providers, helped arrange child care (if the person had children), and made payments to the chosen provider. People who were not working full time were helped to find a job. Those who could not find one were offered a temporary community service job. No one who worked less than 30 hours a week was able to receive benefits, but because of the community jobs, no one was denied this opportunity.
The studies were unusually rigorous. A random, lottery-like process was used to assign welfare recipients in MFIP and SSP, or in New Hope's case, a broad cross section of low-income people, to either a group that was eligible for the program's work incentive payments (e.g., the MFIP group) or to a group that was not (a control group eligible for the regular welfare program). Differences in employment and other outcomes between the two groups is the effect or "impact" of the program.
By conditioning incentive payments on full-time work, the programs increased both work and income among single parents at risk of long-term welfare dependency, without incurring many of the unintended negative consequences on employment among the working poor that have plagued past policies. The employment and earnings gains among long-term welfare recipients were among the largest found in any previously evaluated welfare-to-work programs; the income gains and accompanying poverty reductions were unprecedented. These economic gains were the catalyst for a chain of mostly positive and, in some cases, extraordinary effects on families and children. More employable groups--welfare applicants, two-parent families, and groups that were working full time when they first learned about the programs--also benefited, but not as consistently. Overall costs rose, however, as the amount paid out in earnings supplements exceeded the welfare savings that resulted when more people went to work.
When work incentive programs were linked to participation mandates or were conditioned on full-time work, they substantially increased the employment, earnings, and total income of long-term welfare recipients. By the second and third year of follow-up, the Minnesota and Canada programs increased annual employment rates relative to a control group that was not eligible for work incentive payments by a fourth to a third, earnings by one- to two-fifths, and before-tax income by a sixth to a fourth. For example, over nearly a three-year period, MFIP increased quarterly employment rates by 35 percent (50 percent for MFIP versus 37 percent for the control group), while quarterly earnings were 23 percent higher ($955 for MFIP versus $779 for the control group).
Not surprisingly, given the rules governing each program, most of the three programs' effects were on full-time employment, with cascading effects on job retention and advancement. At its peak, the SSP program nearly doubled the full-time employment rate (29 percent for MFIP versus 16 percent for the control group). Job retention rates were high--two of every three people who got a job because of MFIP or SSP stayed employed for at least a year. There was evidence of wage growth in both programs as well; in SSP about half of those employed experienced wage growth of more than 10 percent over two years. Full-time employment was key to both job retention and wage advancement.
Moreover, the increased income that long-term welfare recipients obtained typically led to a substantial decline in both poverty and the poverty gap--that is, roughly an 11 percent drop in the number of people with incomes below the poverty line as well as a sizeable decrease in the size of the average gap between a recipient's income and the poverty line.
The evidence also showed that participants spent as much as half of their increased income on basic necessities--food, clothing, housing, and child care--a sign of previous impoverishment. Moreover, people made an effort to increase their assets by opening savings accounts and taking other actions; in SSP the fraction of people with $500 or more in savings rose by nearly 30 percent.
Like the Earned Income Tax Credit (EITC) and other policies that redistribute income, these work incentive programs typically increased receipt of transfer payments, resulting in modest increases in transfer payment amounts (6 to 14 percent higher than the welfare payments made to a control group). At the same time, the share of people relying solely on welfare declined substantially as more people combined work and earnings supplements.
These extra costs bought remarkable changes in family life and child outcomes. Incidents of domestic violence decreased 18 percent in MFIP (49 percent for MFIP versus 60 percent for the control group), while marriage rates increased (11 percent of the MFIP group was married three years later versus 7 percent of the control group). In SSP, marriage increased by a like amount in New Brunswick--a heavily Catholic province--but decreased in British Columbia, suggesting that cultural norms also influence whether work and income gains will encourage "independence" or marriage.
The data suggested improvements in several child outcomes as well, particularly for children ages two to nine when the programs began, including better performance in school, higher test scores, and fewer behavioral problems. For adolescents there was some suggestion of negative effects. With less parental supervision, adolescents may have been more likely to engage in minor delinquency behavior and experiment with alcohol, smoking, and possibly drugs.
For welfare applicants, the incentives-only program model in Minnesota, without participation requirements or full-time work conditions, had more limited effects, modestly increasing employment and income but not earnings. The New Hope program did increase employment, earnings, and income for people who were not working full time when they learned about that program, although the effects were smaller than with MFIP or SSP.
Among two-parent recipient families, MFIP was responsible for a dramatic increase in marriage, primarily because of a reduction in divorce and separation: At the end of the third year of follow-up, 67 percent of MFIP two-parent families were still married, compared with 48.5 percent of Aid to Families with Dependent Children (AFDC) recipients--a 38.1 percent increase. The program did not affect the likelihood that at least one parent would work, although the second earner in twoparent families typically cut back, suggesting that MFIP's effort to make work pay relieved stress on two-parent families. Total income still rose because supplement payment amounts exceeded the earnings decline; this extra income, coupled with a higher likelihood of staying married, led to a substantial increase in home ownership among the MFIP group (37 percent versus 18 percent).
Among the other findings:
The programs were not a panacea. In Canada only about a third of those eligible for the program ever took advantage of its earnings supplements, although that number rose to 50 percent when incentives were combined with job search services. By comparison, in Minnesota where both full- and part-time work were supplemented, about three-quarters of those eligible ever took advantage of the program.
The findings pose two philosophical issues: What constitutes dependency, and how to choose between universal and targeted strategies for reducing poverty? On the dependency question, there is consistent evidence across the three studies that earnings supplements increase work and reduce poverty, but by one measure, dependency arguably rises, albeit modestly. For example, in MFIP 85 percent of the group received benefits (welfare when they did not work, earnings supplements when they did) while 81 percent of the control group did. Conversely, by another measure, dependency falls quite substantially--54 percent of the control group relied solely on welfare, while only 42 percent of the MFIP group did. The two effects are inevitable: Incentives programs reward some people who would have gone to work anyway while they provide incentives to people who would not have worked otherwise. In these programs, the latter effect is clearly larger than the former. And don't forget, increased benefits for the former group meant more income, less poverty, and improved child and family outcomes without a significant decrease in employment rates.
On the question of universal versus targeted programs, there is evidence to support either position in these findings. In favor of targeting--as one moves up the employability ladder to more employable welfare applicants, then to two-parent families, then to working poor families--effects on employment, earnings, and income begin to decline, in part because there is not much room to affect employment in populations that work a lot already. But there are offsetting gains in favor of more universal approaches to consider as well. As Frank Levy explains in The New Dollars and Dreams, low-income two-parent families adjusted to 23 years of stagnant real earnings (1973-1996) by sending both parents into the workplace, a strategy that created new stresses on families and children. Hours reductions by the second earner in MFIP two-parent families and among the full-time employed group in New Hope appear to have relieved some of this pressure, resulting in less family splitting (MFIP), less stress, and better outcomes for children (New Hope).
In practical terms, the choice between universal and targeted strategies for making work pay is a choice between expanding the universal EITC program, which uses the tax system to supplement the pay of low-wage workers, and making changes to federal and/or state welfare programs operating under the Temporary Assistance for Needy Families (TANF) block grant created by the 1996 federal welfare reform bill. Increasing the generosity of the EITC avoids welfare's stigma, inequity, and bureaucracy, but there are disadvantages: The cost would be enormous, there is a risk of additional work reductions among working-class households as the credit moves further up the income stream, and because the credit comes at the end of the year and is not well-understood, it might not have the same work-inducing effects among more disadvantaged populations as a targeted program.
Using the TANF welfare system to make work pay produces a mirror image of pro and con arguments: On the one hand, the gains per person will be larger, the costs lower, and marketing easier; yet on the other hand, the welfare system as currently structured is not well-suited to supporting the working poor, and most concerning of all, welfare case loads have fallen by half since 1994, so recent welfare leavers would not benefit. Perhaps the most convincing arguments in favor of using the TANF structure are these: (1) MFIP and New Hope's effects occurred on top of the EITC, (2) most state welfare laws already include make-work-pay provisions, (3) TANF is a funding source, not a program, and thus is flexible enough to address at least some of the mentioned shortcomings, including being used to support a supplemental state EITC program.
Indeed, as welfare rolls have decreased, states have accumulated a large and growing surplus of federal dollars--$7 billion or higher by most estimates. If states do not spend these resources, the budget pressures created by the so-called "pay as you go" provisions now required in each year's budget bill may mean that Congress will reduce the federal contribution to the welfare block grant, effectively eliminating the surplus in future appropriations.
But first, a basic contradiction must be resolved. States have put time limits on the total number of months families can receive welfare benefits (typically 60 months, sometimes less), which strongly discourage continued receipt of welfare assistance. As a result, when welfare recipients take jobs, the welfare system's more generous disregard policies have the perverse effect of holding job-takers on the welfare rolls, where they receive reduced grants and continue using up their lifetime limit of federal (and, in some cases, state) welfare benefits. This, in effect, perpetuates welfare's age-old conflict between reducing poverty and reducing dependency.
Moreover, why use work incentives to make work pay, when time limits will have the same effect while also reducing costs? The answer depends on the importance policy makers place on reducing poverty among single-parent families with children. Limiting welfare will raise employment rates, but given the labor market's new insistence on skills, welfare recipients are most likely to get low-wage jobs with little prospect of earnings growth. Disposable income, even after receipt of the EITC, will likely remain well below the poverty line.
What are the implications for states that want to reduce poverty and dependency (defined as those relying solely on welfare) while encouraging work? First, states should consider separating time-limited welfare from work incentive initiatives by stopping the clock on recipients' welfare time limits when they work full time (30 hours a week or more). Then, when recipients do get jobs, they can leave time-limited welfare and join a newly created program for the working poor. This program could offer a range of supports from work incentives to child care, health care, and information on the EITC. Illinois stops the clock for anyone working more than 25 hours a week now. Both Canada's SSP and Milwaukee's New Hope made their incentive payments only when someone worked at least 30 hours a week, and both programs were run outside the welfare system. New Hope's hours rule limited the work reductions among full-time workers to cutbacks in overtime work, and SSP's rule encouraged people who would not have worked at all to work full time--an important element in any long-term effort to attain self-sufficiency.
Second, states might make the incentive for full-time work more generous. When researchers simulated the effect of running an SSP-type program instead of the more modest disregards states have used in the past, the results suggested that an SSP-type program would have produced large increases in the number of long-term welfare recipients working full time and large increases in income, at only a modest net increase in costs. The magnitude of the simulated gains indicates that they would likely exceed those achieved by the states' current versions of enhanced earnings disregards.
Third, a more generous incentive payment could be structured to encourage and reward asset accumulation. Incentive payments could be broken into two components--one to supplement earnings and another to supplement savings. For example, if participants managed to save $1,500 of the $5,000 or more, their incomes could increase annually; if incentive funds matched those savings, over the course of three to five years, many participants might be able to afford to buy a home.
Fourth, if the stigma and inequity that accompany a targeted program are judged too high a price to pay despite its efficiency, policy makers might consider establishing a supplemental state EITC program or increasing the generosity of the federal universal EITC program, but only for people who work at least 30 hours a week. This strategy would encourage full-time employment and limit the work-hour reductions among full-time workers while retaining the part-time work benefits of the existing EITC program for single parents, whose parenting responsibilities may preclude full-time work. Policy makers might also consider an even more generous EITC program for married parents, to promote work and marriage.
The mounting TANF surplus indicates that states can afford to help make work pay. Indeed, most state welfare reforms have intended to do just that. Presumably, if supporting people when they do not work does not comport with America's values, then supporting people when they do work would. While additional experimentation is warranted, particularly to confirm the marriage effects of incentive programs, the evidence amassed so far on the economic effects of making work pay is compelling. The benefits are substantial, the costs can be modest, and the risks appear to be minimal.
- Manpower Demonstration Research Corporation
- Minnesota Family Investment Program
- Milwaukee's New Hope Project
- See the full list of Making Work Pay links.
- Minnesota Family Investment Program