Reforming Welfare and Rewarding Work: A Summary of the Final Report on the Minnesota Family Investment Program

The Minnesota Family Investment Program (MFIP) represents a new vision of welfare as a system that can simultaneously encourage work, reduce dependence on public assistance, and reduce poverty. It attempts to break loose from the historical tradeoffs among these goals by implementing two complementary policies: financial incentives to reward work and reduce poverty and, for long-term welfare recipients, mandatory participation in employment-focused services to encourage and require work and reduce dependence.

MFIP was initially implemented as a pilot program in the three urban counties of Hennepin (Minneapolis), Anoka, and Dakota, and the four rural counties of Mille Lacs, Morrison, Sherburne, and Todd. The pilot program operated from April 1994 to June 1998 and was evaluated by the Manpower Demonstration Research Corporation (MDRC) under contract to the Minnesota Department of Human Services (DHS). The evaluation was also supported by the agencies and foundations listed at the front of this summary. A modified version of MFIP is now Minnesota’s statewide welfare program.

This document summarizes the results presented in the evaluation’s final report. Volume 1 of that report examines MFIP’s effects on employment, earnings, welfare receipt, income, marriage, and other outcomes for adults in single- and two-parent families for up to three years after they entered the study. Volume 2 presents the results of a special study of MFIP’s effects on children and other aspects of family well-being for single mothers who had at least one child aged 2 to 9 when they entered the study.

MFIP’s results are particularly important because more than 40 states have incorporated a “make work pay” approach in conjunction with work requirements as part of their new, time-limited welfare reforms, which followed enactment of the 1996 federal Personal Responsibility and Work Opportunities Reconciliation Act (PRWORA). Most commonly — as in MFIP — states have aimed to make work pay by increasing their “earned income disregard”: More of a family’s earnings are disregarded (not counted) when their welfare grant is calculated. This policy allows more people to combine work and welfare. As discussed later, MFIP also included other financial incentives to work. The MFIP pilot program did not include time limits on welfare receipt, but the newer, statewide version does.

The evaluation results speak directly to three goals that have emerged as high priorities under PRWORA: ensuring that long-term welfare recipients make substantial strides toward self-sufficiency before reaching their time limits on welfare receipt, supporting the efforts of low-income workers to advance in their jobs and provide adequately for their families, and assuring that social policies do not discourage marriage.

To assess MFIP’s success in achieving its ambitious goals, the evaluation used a rigorous, random assignment research design. Between April 1994 and March 1996, more than 14,000 families in seven Minnesota counties were assigned, using a lottery-like process, to either the MFIP program (the “MFIP group”) or the traditional Aid to Families with Dependent Children (AFDC) program (the “AFDC group”). MFIP’s effects were estimated by following the two groups over time and comparing their employment, welfare receipt, and other outcomes. The difference in outcomes between the two groups is the effect, or impact, of the MFIP program. For example, an “increase” in employment means that the MFIP group achieved a higher employment rate than the AFDC group.

MFIP’s designers set out to increase employment, reduce poverty, and reduce dependence — a set of goals rarely achieved by any previous program. It succeeded in meeting the first two of these objectives — and, by some measures, the third — for the group many policymakers consider key: single-parent long-term recipients, who are the majority of the caseload at any given time and are least likely to enter employment on their own. [1] MFIP’s results also provide the first hard evidence that changes in welfare policies can lead to increases in marriage and marital stability among program participants — another central goal of many welfare reformers. MFIP had less consistent effects for recent applicants to welfare than for families who had already been receiving welfare when they entered the program. Figures 1 and 2 illustrate some of the main results.

As they developed a new vision of welfare during the late 1980s, Minnesota public officials tried to address a problem that had plagued the welfare system for decades: For many families receiving AFDC, welfare paid better than work. Parents who went to work typically lost most of their assistance because benefits were reduced nearly one dollar for every dollar of earnings, and most of these workers did not earn wages high enough to offset this benefit reduction and make their families better off. In fact, after deducting child care and other expenses, they often found that they were worse off. Many families adjusted to three decades of stagnant or declining wages for low-wage workers by sending both parents into the workplace, a societal development with a range of potential implications for families and children. Single parents, however, did not have this option. Although they usually report that they prefer work to welfare, they have been faced with a tradeoff between being poor and dependent on welfare, with the stigma such a choice brings, and being poor and working, with all the stress associated with managing a job and a family alone.

Thus, the challenge for policymakers was to design a program that would encourage work and reduce dependence on welfare but at the same time reduce poverty. Most policies that focus on one of these goals do not achieve the others. On the one hand, approaches aimed primarily at increasing employment (policies characterizing most earlier welfare-to-work programs) usually increase work and reduce welfare caseloads. They do not typically make families better off, however, because earnings are not high enough to compensate for lost benefits. On the other hand, policies designed to increase incomes by increasing welfare benefits can reduce poverty, but at the cost of increasing dependence on welfare and possibly reducing work effort.

MFIP	AFDC
· Recipients were eligible for welfare until their income reached 140 percent of the poverty line	· Recipients faced a sharp reduction in benefits as earnings increased.
· Child care subsidies were paid directly to the provider if the recipient worked while receiving welfare	· Child care was reimbursed through the recipient’s AFDC grant

Under the traditional AFDC system, recipients had little incentive to work, given the sharp reduction in benefits as their earnings increased (an implicit tax on earnings of nearly 100 percent). For example, under AFDC, a single parent who went to work part time and earned $520 per month would have her AFDC and Food Stamp benefits reduced by $407, leaving her only marginally better off in terms of total income. MFIP increased the incentive to work for both single- and two-parent families. When a parent went to work, her basic grant was increased by 20 percent to offset work-related expenses, and then 38 percent of her earnings were disregarded (not counted as income) in calculating the family’s grant level. Using the above example, because more of the parent’s earnings would be “disregarded,” the single parent working part time would have her benefits reduced by only $170 under MFIP. Thus, MFIP increased the reward for working by $237 compared with AFDC. MFIP’s incentives raised the reward for working more for part-time than full-time work.

MFIP also provided an additional incentive to work through its child care payment policy. For parents working or participating in employment-related activities, MFIP paid child care costs directly to providers. Under AFDC, in contrast, parents paid child care costs themselves and were reimbursed later, a practice that may have discouraged them from going to work but may have also hindered their ability to stay employed.

If incentives are to affect parents’ decisions about work, then recipients need to be given a clear explanation of how the new incentives work. MFIP financial workers provided the initial explanation of the new rules during the eligibility interview that took place on the day each family was assigned to MFIP. Their role in explaining and promoting the financial incentives led MFIP financial workers to view their jobs differently than they had under the AFDC program. In particular, they felt empowered to discuss work and work-related topics with the parents on their caseloads, rather than focusing solely on eligibility determination.

MFIP	AFDC
· Mandatory participation in employment and training activities for single parents who had received assistance for at least 24 of the prior 36 months and were not working full time	· Voluntary, education-focused STRIDE welfare-to-work program for single parents
· For two-parent families, mandatory participation required after the family had received assistance for at least 6 of the prior 12 months	· Mandatory job search/Community Work Experience Program for most two-parent families

Under the AFDC system, single-parent recipients were eligible to volunteer for Minnesota’s welfare-to-work program, known as STRIDE. The STRIDE program provided education, training, and other services and, throughout the evaluation period, placed most enrollees in longer-term education and training activities. Under MFIP, all recipients who received welfare for at least two of the prior three years, had no children under age 1, and were not currently working at least 30 hours per week were required to participate in MFIP’s employment and training services. In contrast to STRIDE, MFIP emphasized quick entry into the workforce through the use of career workshops, job search classes, and other employment-focused activities. The participation mandates were targeted to long-term recipients to minimize costs and focus the services on individuals most likely to need them.

For two-parent families, the participation requirements also differed under MFIP compared with AFDC, although the differences were less dramatic, since two-parent families who received AFDC — usually through the AFDC-UP (Unemployed Parent) program — were also subject to mandates. Under AFDC, the family could continue receiving benefits only if the primary wage earner worked, searched for a job, or worked in exchange for benefits through the Community Work Experience Program. Under MFIP, at least one of the parents was required to work 30 hours per week or to participate in employment-focused activities after the family had received welfare for six months.

Because of these participation requirements, single-parent long-term recipients in MFIP were more likely to participate in employment and training activities than families assigned to the AFDC program. Participation increased most in job search activities, although a significant proportion also participated in education or training. MFIP’s mandates made less difference for single-parent recent applicants, a large proportion of whom left welfare before the mandates applied to them, and for two-parent families, whose AFDC counterparts had similarly strict participation requirements.

Interestingly, the existence of the financial incentives affected the way the employment and training services were implemented. A significant part of the orientation to employment and training services was devoted to explaining the financial incentives in detail, so that participants would understand the benefits of going to work relatively quickly. MFIP employment and training staff followed up by reinforcing the “work pays” message in their monthly contact with recipients. Finally, because MFIP’s incentives ensured that work would make families better off, more MFIP than STRIDE staff reported that they encouraged parents to enter employment quickly or to mix education and training with part-time work.

Simplification of rules and procedures

MFIP	AFDC
· Consolidation of AFDC, Food Stamps, and Family General Assistance; Food Stamps were “cashed out” (that is, their value was included in the MFIP grant)	· Separate programs with different rules
· Elimination of the work history requirement and 100-hour rule for two-parent families	· Work history requirement and 100-hour rule for two-parent families

Under AFDC, parents often faced a confusing array of programs, each with its own eligibility criteria and benefit calculations. Partly for this reason, staff focused primarily on complex eligibility issues rather than on helping families move toward self-sufficiency. MFIP combined AFDC, Food Stamps, and Family General Assistance (a state-funded cash assistance program enrolling a small proportion of the caseload) into a single program with one set of rules and procedures and one monthly payment. Recipients also received their Food Stamp benefits as part of their cash grant, rather than separately as coupons. In addition, the eligibility criteria for two-parent families were made comparable to those for single-parent families. Under AFDC-UP, two-parent families were ineligible to receive benefits if the primary earner in the family worked more than 100 hours per month or if this earner did not have a recent work history. Both of these restrictions were eliminated under MFIP.

Determining what difference MFIP made requires knowing how the parents in MFIP would have behaved if they had not been in MFIP. For example, tracking employment rates over time for families in the MFIP program does not, by itself, indicate how many of those families went to work because of MFIP and how many would have worked anyway. The most reliable way to determine the number of people who would have worked anyway is by using a random assignment research design. Between April 1994 and March 1996, more than 14,000 recipients of and applicants for public assistance were randomly assigned to either the MFIP program (the MFIP group) or the AFDC program (the AFDC group). Because people were assigned at random to the two groups, there were no systematic differences between the groups at the beginning of the study. They were similar in their demographic characteristics as well as in their history of employment and welfare receipt. For this reason, any differences that emerged between them after they entered the study can reliably be attributed to the MFIP program. MFIP’s effects were estimated by following the two groups over time and comparing their employment, welfare receipt, income, and other family outcomes. The difference in outcomes between the two groups is the “impact” (effect) of MFIP. [2] The accompanying box presents several key definitions used in the analysis.

Key Definitions in the MFIP Evaluation

Welfare. Although the term welfare is typically associated with cash assistance, welfare is defined more broadly in the MFIP evaluation reports. For families in the AFDC group, welfare is defined as income from AFDC payments, Food Stamp benefits, and General Assistance payments. For families in the MFIP group, welfare is defined as MFIP payments, part of which is Food Stamp benefits given in the form of cash.

Dependence (or self-sufficiency). There are various ways to define dependence on welfare, reflecting families’ varying degrees of reliance on welfare relative to other sources of income. A parent who is relying solely on welfare is more dependent, for example, than one who is working and supplementing her earnings with a small welfare grant. Several measures of welfare receipt are presented here in order to reflect this continuum of dependence.

Poverty. For purposes of this evaluation, a family is considered to be in poverty if the parents’ earnings plus welfare benefits are below the official poverty line. (Because of resource constraints, data were not collected on all sources of family income for the whole follow-up period.) Because this measure of family income includes Food Stamps and does not include cash income from other sources, the poverty figures presented here are not comparable to the official poverty rate.

To enable the evaluation to disentangle the effects of the financial incentives from the effects of adding participation mandates to the incentives, some single parents were randomly assigned to a third group, which received the financial incentives but were never subject to a participation mandate. When the remainder of this discussion attributes a particular effect of the program to either the financial incentives or the addition of the participation mandates, that conclusion was based on evidence from this aspect of the research design (although, for brevity’s sake, the results for this third group are not presented in this summary document).

Data on sample members’ earnings and welfare receipt were obtained from state Unemployment Insurance records and public assistance benefits records. Data on other aspects of family well-being, plus additional information on employment, were obtained from a survey administered to a subset of the full evaluation sample three years after they entered the study. The survey asked families a battery of questions about family circumstances, such as health care coverage and material hardship. Families who had at least one child between the ages of 2 and 9 at entry into the study were given an expanded survey, which included the original questions plus a series of questions covering the well-being of their children and additional family circumstances.

The sections that follow present the program’s effects for four different types of families: single-parent long-term recipients, two-parent recipient families, single-parent recent applicants (a category that includes new applicants and people who had received welfare for less than two years), and two-parent applicant families. Each of these family types has very different welfare and employment patterns (even in the absence of MFIP) and faced different rules under MFIP and AFDC.

Single-parent long-term recipients are heads of families who had received welfare for two years or more in the prior three when they entered the evaluation. Except for those who were already working at least 30 hours a week (and had no children under age 1), people assigned to the MFIP group were required to participate in MFIP’s employment and training services immediately upon entering the program because they already met the criteria for mandatory participation. Because they were the only group whose mandate applied immediately and whose AFDC counterparts did not have any participation requirements, the effects for this group most fully capture the combined effect of MFIP’s incentives and mandates.

Economic Effects

· MFIP increased employment by 35 percent and increased earnings by 23 percent, on average.

One of MFIP’s key goals was to increase work, and the results show that it did so quite substantially among long-term recipients. In fact, MFIP’s employment gains were on a par with those of two of the most successful welfare-to-work programs previously evaluated — the Riverside (California) GAIN program and the Portland (Oregon) JOBS program. Table 1 presents MFIP’s effects for recipients in urban and rural counties during the first 9 quarters after they entered the study. In an average quarter, 49.9 percent of MFIP parents worked, compared with 37 percent of AFDC parents — a 35 percent increase in employment rates. Their earnings were also 23 percent higher, on average.

The results for urban and rural counties are combined in this summary, but throughout much of the full report they are presented separately. MFIP’s effects on employment and earnings were considerably larger in urban than in rural counties. Aside from that distinction, MFIP had fairly consistent impacts across most types of families, including across racial and ethnic groups. One exception is that it increased employment and earnings relatively less among single parents who had been previously married when they entered the study. It may be that other (unmeasured) characteristics of previously married mothers actually caused their smaller impacts; for example, they may have tended to leave welfare more quickly than never-married mothers because they were more likely to marry or for other reasons. Nevertheless, MFIP’s smaller impacts for previously married single parents do help to explain MFIP’s smaller effects in rural counties, since the majority of rural recipients had been previously married.

A common concern raised about welfare-to-work programs that (like MFIP) emphasize entering employment quickly is that they may push recipients into unstable, poor-quality jobs. The results for MFIP, however, show that many of the MFIP parents who got jobs stayed employed fairly continuously and worked in jobs that offered some benefits. For example, more parents in the MFIP group than in the AFDC group went to work during the first year after they entered the study and stayed employed for at least 12 months (not shown in Table 1). Also, as of the three-year mark, more parents in the MFIP group reported working or having worked in full-time jobs (at least 30 hours per week), and more reported that their jobs offered health benefits.

· MFIP increased the number of families receiving welfare because it allowed more of them to work and still receive some benefits. However, MFIP families were also 21 percent less likely to be solely dependent on welfare.

By some measures, MFIP reduced dependence, but by others it did not. Because MFIP allowed families to keep more of their benefits when they worked, one could argue that it modestly increased dependence: 85.3 percent of MFIP recipients received welfare in an average quarter, compared with 80.6 percent of AFDC recipients. However, families rely on welfare to varying degrees, and a family combining work and welfare would generally be considered less dependent than one relying solely on welfare, despite the fact that both receive some benefits. Using another definition, then, MFIP substantially reduced dependence: In each quarter after program entry, an average of 54.5 percent of recipients in the AFDC group relied only on welfare compared with 42.4 percent of recipients in the MFIP group.

One feature of MFIP that distinguished it from many other welfare-to-work programs was its antipoverty focus, and its impacts on income and poverty bear this out. Although the Riverside and Portland programs, for example, moved a substantial number of people into jobs, most of them were not made better off in terms of total income, since their benefits were reduced almost dollar for dollar as their earnings increased. In contrast, MFIP made single-parent long-term recipients better off. Because MFIP families had higher earnings and were able to keep more of their welfare benefits, their incomes (defined here as the sum of earnings and welfare benefits) were 15 percent higher on average than incomes of AFDC families, and their incomes were more likely to rise above the poverty line. In addition, the income gain of 15 percent is a conservative estimate, given the substantial additional benefits available to working families through the federal Earned Income Credit (EIC) and Minnesota’s Working Family Credit (WFC). Families’ EIC and WFC benefits are estimated in the full report but not included in the income measures presented here.

· Compared with recipients in the AFDC group, recipients in the MFIP group were 13 percent more likely to have had continuous health care coverage over the three years of follow-up.

Health insurance is an important issue for low-income families, since many lose coverage when they leave welfare and have jobs that offer either no insurance or insurance they cannot afford. MFIP increased the continuity of recipients’ health coverage: 68.9 percent of MFIP recipients reported that they had uninterrupted coverage over the three-year period, compared with 61.2 percent of AFDC recipients. This effect is most likely a byproduct of the increase in the percentage of families receiving welfare when they work, since such families are automatically eligible for Medicaid.

· Taking all monetary effects of the program into account, MFIP added about $2,000 per year to government costs per family. Single-parent long-term recipients in the MFIP group were, on average, about $2,000 better off per year than their AFDC counterparts and also experienced a number of important nonfinancial improvements in their lives.

Just as MFIP’s financial incentives enabled the program to raise family incomes more than in a typical welfare-to-work program, they also made MFIP more costly than the typical “work first” program, many of which save the government money rather than adding to government costs. For families, the two largest sources of financial gains were increased welfare benefits and increased earnings and associated fringe benefits. Also important were increased Medicaid payments for working families (because members of the MFIP group remained on welfare longer), increased EIC and Minnesota WFC payments, and increased child care payments. The program was particularly efficient at producing gains for this particular group — single-parent long-term recipients. Each dollar spent by taxpayers resulted in an equivalent financial gain to families, as well as a set of nonfinancial gains that these dollar values do not capture, such as the dramatic improvements in family life and child well-being discussed below.

Effects on the Well-Being of Families and Children

Because MFIP, as its name expresses, was designed as an investment in families, the evaluation included a study of its effects on families and children to assess whether the program’s potential economic effects brought measurable improvements in children’s environments and well-being. As described below, the program’s effects on families’ financial status did produce a chain of striking effects in other areas, including marriage, domestic abuse, and children’s well-being.

· At the three-year follow-up point, MFIP recipients were more likely to be married than were AFDC recipients.

As shown in Table 1, 10.6 percent of MFIP parents were married at the end of the third year, compared with 7.0 percent of AFDC parents. There are a variety of ways in which MFIP might have affected marriage rates — for example, one might expect that moving single parents into the workforce would increase their chances of marriage by expanding their social networks. However, results shown in the full report indicate that MFIP’s incentives and benefit rules (rather than the participation mandates) were largely responsible for the increase in marriage rates.

First, the enhanced incentives provided parents with more income than they would have had otherwise, and Food Stamp benefits were provided in the form of cash — both changes that give parents more discretion over their spending. Parents with a greater sense of financial security may be more likely to marry.

Second, single parents in the MFIP group may have been aware that two-parent families faced fewer eligibility restrictions under MFIP than under AFDC; MFIP eliminated the work history requirement and the work hours restriction (the “100-hour” rule). Many observers believe such rules to be disincentives to marry — for example, a single parent who married a full-time worker could jeopardize her welfare eligibility. Thus, removing these rules might have encouraged higher rates of marriage among single parents in the MFIP group.

· Over the three years of follow-up, MFIP reduced the reported incidence of domestic abuse by 18 percent for a group of single-parent recipients with young children.

Nationally, and in Minnesota, a startlingly high number of welfare recipients experience domestic abuse. Such abuse can have a wide range of negative effects on parents, in addition to the physical and psychological pain of the abuse, serving as a barrier to finding and keeping a job and affecting how they interact with their children. The evaluation’s three-year survey measured various types of abuse against the mothers. More than half of the survey respondents in the AFDC group reported having been abused during the three-year follow-up period, with more reporting physical abuse (for example, sexual abuse or hitting) than nonphysical abuse (for example, threats or yelling).

MFIP substantially reduced the incidence of abuse: Among single, urban mothers with young school-aged children,[3] those in the MFIP group were 18 percent less likely to have experienced abuse during the three-year period.[4] (These reductions were in abuse by intimate partners, such as husbands or boyfriends, as well as family members or other individuals.) MFIP’s effects on domestic abuse are particularly notable, since abuse is such a difficult dynamic to change, even by programs that are designed to target the problem much more directly. Moreover, this effect has potentially widespread implications for the well-being of both parents and children.

As with the effects on marriage, results shown in the full report indicate that MFIP’s effects on abuse were due to its changed incentives and benefits. It is difficult to pinpoint how the changed welfare rules reduced domestic abuse, but the explanation is probably more complex than simply “increase income, decrease domestic abuse.” MFIP probably also helped women feel more control over their lives and finances, perhaps changing the dynamic in some of their close relationships. Not only did the mothers have more money, but the money was linked to working; Food Stamps were now provided as cash, giving the women more options in their spending; and MFIP staff encouraged them to take advantage of their new opportunities and decrease their reliance on welfare. Further, as discussed in the full report, MFIP’s incentives also reduced mothers’ risk of depression, and depression is correlated both with feelings of helplessness and with abuse.

The bottom panel of Table 1 summarizes MFIP’s effects on young school-aged children, a group for whom extensive information was collected about child care, parenting, and child outcomes. Compared with mothers in the AFDC group, mothers in MFIP were less likely to report that their children exhibited problem behaviors such as cheating or being cruel, disobedient, or moody. These and the other aspects of problem behavior about which sample members were asked make up the Behavior Problems Index (BPI); the BPI is used in a number of national studies and is highly predictive of a child’s future well-being such as performance in high school. Mothers in MFIP also reported that their children did better in school on average and were more engaged in school, as measured, for example, by completing homework and getting along with teachers. MFIP’s effects on children were more pronounced for particular groups: girls, children aged 6 to 9 at the time they entered the study, and African-American children. For no group of children of long-term recipients was MFIP found to have negative effects.

How did MFIP — a program that did not include parenting classes, school attendance requirements, or other services aimed directly at parenting or child development — produce these positive effects for children? Analyses presented in the full report indicate that MFIP’s effects on child well-being are primarily attributable to the program’s financial incentives — the additional money available to working families — rather than its participation mandates. The incentives (and other changes such as the Food Stamp cash-out) affected many aspects of family life that, in turn, could have influenced children’s outcomes. As discussed above, MFIP not only increased parents’ employment and reduced poverty, but it also increased the likelihood that parents married,[5] and it substantially reduced domestic abuse. Another change that has direct implications for children’s well-being is that the children spent more time in formal child care, and their care arrangements were more stable than for those in the AFDC group. (Unlike other changes for children, the increase in formal day care is attributable to the participation mandate and therefore results from a combination of substantially increased employment, higher incomes, and MFIP’s direct child care payment system). Given all of these positive developments in families’ lives, it is perhaps not surprising that the children in those families were better off.

MFIP’s effects for two-parent recipient families are presented in Table 2. Two-parent families are defined as families in which the recipient reported living with a partner or spouse at the time they entered the study. These families had been receiving benefits for at least one month when they entered the program, but most had received welfare for at least six months, making them immediately subject to the mandate for two-parent families. It is important to remember, however, that for two-parent families MFIP is primarily being compared with the AFDC-UP program, which had fairly strict work requirements already. Thus, the key ways in which MFIP differed from AFDC-UP were not in its work requirements, but rather in its eligibility criteria and enhanced financial incentives. In the discussion that follows, many of the outcomes on family and child well-being presented for single-parent families are not presented because these data were not collected for two-parent families.

· Two-parent recipient families in MFIP were as likely as those in the AFDC group to have at least one parent work, but less likely to have both parents work, leading to lower combined earnings.

With two parents potentially available as workers, it is not surprising that two-parent families receiving welfare are more likely than single-parent recipients to have at least one parent in the workforce. In fact, the majority of two-parent recipient families in both the MFIP and AFDC groups had at least one employed parent. On average, 60.2 percent of MFIP families worked in each quarter of the evaluation’s follow-up period, compared with 62.5 percent of AFDC families, a difference that is not statistically significant. Instead, MFIP’s primary effect was to cause one parent in some of the families to cut back on work, either by reducing his or her hours worked or leaving work entirely. For this reason, spouses’ combined earnings were somewhat lower for MFIP families than for AFDC families.

· By allowing more families to combine welfare and work, MFIP increased welfare receipt by 16 percent.

As for single-parent recipients, MFIP increased welfare receipt among two-parent families: 76.4 percent of MFIP families received benefits each quarter, on average, compared with 66 percent of AFDC families. This effect is the result of MFIP’s enhanced incentives and removal of the 100-hour rule, which allowed more working families to remain eligible for benefits. In other words, the additional families who received benefits were all working families. This can be seen from the fact that more MFIP families than AFDC families combined welfare and work.

· MFIP families were 40 percent more likely to stay married than their counterparts in the AFDC group.[6]

Two-parent recipient families often face unstable family lives. By the end of the third year, for example, nearly 30 percent of couples in the AFDC group who were married at program entry had separated or divorced. MFIP’s most dramatic effect for two-parent families was to reduce this aspect of family instability. Table 2 shows that 67.3 percent of MFIP recipient families reported that they were married at the end of year 3, compared with only 48.3 percent of AFDC families, for a nearly 20 percentage point, or 40 percent, increase. Because of the magnitude and importance of this result, it was confirmed using an additional source — public divorce records from the State of Minnesota. These data showed that parents in the MFIP group were less likely than those in the AFDC group to divorce, and that this difference persisted for five years after entry into the study. MFIP’s effects on marital stability were similar, and of similar magnitude, for a wide range of families — defined, for example, by race/ethnicity, number of children, and prior work experience.

Because these results provide new evidence that changes in the welfare system can increase the likelihood that families stay together, it is critical to understand how these effects occurred. Because both MFIP and AFDC included participation mandates for most two-parent families, the effects on marriage for two-parent families are more likely to have been caused by MFIP’s changes in benefit structure than by its mandates. (This is consistent with the conclusion drawn about the increase in marriage among single parents in the MFIP group, based on analyses in the full report that disentangle the effects of the incentives and the mandates for single parents.)

There are two likely ways in which MFIP’s changes in benefit structure could have increased marital stability, the first of which is MFIP’s less restrictive eligibility criteria. In particular, MFIP eliminated the 100-hour rule of AFDC-UP, in which the family lost eligibility for benefits if the primary earner worked more than 100 hours per month. As mentioned earlier, the 100-hour rule is widely thought to discourage marriage and might have caused some families receiving AFDC-UP to separate or divorce so that the mother and children could continue receiving benefits as a single-parent family. Second, MFIP may have increased marital stability through the benefit changes that applied to both single- and two-parent families. Its cash-out of Food Stamps increased families’ discretion over how they allocate their money, and its enhanced benefits for working families allowed one parent in some of these families to reduce his or her work hours without reducing family income. Perhaps these reductions in financial strain and in work hours helped reduce sources of marital stress.

· Compared with AFDC families, MFIP families had higher incomes, and twice as many of them owned a home at the end of the third year.

MFIP’s effects on marital stability led to other positive effects on families, the most obvious being that families who stayed together had higher incomes than families who did not. Table 2 shows that, after accounting for separations and divorces that occurred throughout the follow-up period, MFIP families had somewhat higher incomes than AFDC families and that fewer were in poverty. Thus, MFIP’s increased financial support for families could have helped increase the likelihood of parents’ staying together; conversely, keeping both potential earners in the household gave families access to higher incomes.

MFIP’s effects on income and marital stability may also have affected the rates of home ownership. Thirty-seven percent of two-parent recipient families in MFIP reported owning their home at the three-year mark, compared with only 18 percent of AFDC families. It may be that more families in the MFIP group bought homes, because families with higher income (including two-parent families) have greater access to mortgages, or it may be that fewer families in the MFIP group lost their own homes owing to divorce or financial difficulties.

· For two-parent recipient families, MFIP led to a total financial gain of about $1,400 per year relative to the AFDC group, and it increased government costs by about $3,800 per year.

These financial gains came from increases in welfare and Medicaid payments for working families who remained on welfare because of MFIP’s financial incentives. Because increased transfer payments to families in the MFIP group were partly offset by reductions in earnings, the program was less efficient in producing financial gains for this group than for single-parent families; each dollar spent by the government resulted in a 35-cent gain for two-parent recipient families. The fact that MFIP was more efficient in producing financial gains for single-parent than for two-parent families parallels the differences found in some previous income transfer studies (such as the well-known Negative Income Tax experiments of the 1970s).

Results for single-parent recent applicants are presented in Table 3, and results for two-parent applicant families are presented in Table 4. Single-parent recent applicants are defined as families who were applying for welfare, or who had been receiving benefits for less than two years, when they entered the program. Two-parent applicant families include only families who were newly applying for welfare as they entered MFIP.

The results for these groups provide information on how MFIP affected families who were new, or relatively new, to the welfare system. For both these groups, however, the evaluation primarily measured the effect of MFIP’s financial incentives (and, for two-parent applicant families, its less restrictive eligibility rules) rather than its mandates. Single-parent recent applicants did not face a mandate to work or to participate in activities unless they continued to receive benefits for 24 months, which many families did not. Two-parent applicant families did face a mandate within six months of entering MFIP, but their AFDC counterparts also faced work requirements. In addition, two-parent applicant families are very likely to leave welfare quickly on their own. For all these reasons, the effects of MFIP’s mandates on single-parent recent applicants and two-parent applicant families were likely to be small.

· MFIP had fewer effects on single-parent recent applicants than on long-term recipients.

Overall, MFIP produced fewer and less dramatic effects for recent applicants, as shown in Table 3. During the follow-up period, an average of 55.3 percent of parents in the MFIP group worked each quarter, compared with 52.1 percent of parents in the AFDC group. Despite higher employment rates, however, parents in the MFIP group did not have higher earnings on average because MFIP caused some of them to move from full-time to part-time jobs or to take lower-paying jobs than they would have otherwise. The increase in part-time work occurred early in the follow-up period and was shown in the interim report. By the end of the third year, however, more parents in the MFIP group than in the AFDC group reported working or having worked full time. That MFIP encouraged some parents to reduce their hours worked or to take lower-paying jobs than they would have otherwise is consistent with economists’ predictions that when more benefits are provided to parents who work, some may be encouraged to take new jobs or to work more, while some who are already working may use the extra income to reduce their work intensity — by cutting back their hours worked, reducing their weeks worked per month, or changing to a lower-paying job. In the case of recent applicants in MFIP, these effects offset each other to produce no change in average earnings.

Because MFIP allowed working families to remain eligible for benefits, parents in MFIP had higher incomes than those in AFDC — an 8.5 percent increase per quarter — and more of them had incomes above the poverty line. In addition, MFIP families were more likely than AFDC families to have had uninterrupted health insurance during the period: 61.7 percent of MFIP parents reported having continuous coverage, compared with only 49.7 percent of AFDC families. MFIP had few effects on other aspects of family and child well-being.

As discussed earlier, a key reason that MFIP’s effects on employment were smaller for single-parent recent applicants is that few parents in this group faced a mandate to work or participate in employment activities until the third year; some left welfare before accumulating 24 months of receipt and others did not reach 24 months of receipt until very late in the study’s follow-up period. It is not clear, however, that MFIP would have increased employment and earnings more for single-parent recent applicants if they had faced the same treatment as long-term recipients, since recent applicants and long-term recipients differ in many ways. In particular, more recent applicants would have left welfare fairly quickly and returned to work in the absence of MFIP. Employment programs typically have a more difficult time increasing employment rates for these types of parents, since there is less room for improvement.

· Two-parent applicant families in MFIP were as likely to have at least one parent work but less likely to have both parents work, leading to lower combined earnings. They also had higher rates of welfare receipt.

As for single-parent families, two-parent families who are new applicants to welfare differ in many ways from recipient families. For one thing, more of them worked in the period after entering the study. Nonetheless, applicant families responded to MFIP’s incentives and less restrictive eligibility criteria in a similar way as recipient families, namely, one spouse in some of these families left work or reduced his or her work hours, leading to lower combined earnings. Finally, because MFIP allowed working families to remain eligible for benefits, more families in the MFIP group than in the AFDC group received welfare during the period. Information about marital stability and other family outcomes was not available for these families.

· Single-parent recent applicants in the MFIP group gained about $1,600 per year per family, while two-parent applicant families neither gained nor lost financially from MFIP. The program resulted in additional government costs of about $1,900 per year for single-parent recent applicants and $2,500 for two-parent applicant families.

For single-parent recent applicants, the largest sources of financial gain were increased welfare benefits and Medicaid payments for working families who remained on welfare because of MFIP’s financial incentives. Families gained 81 cents for each additional dollar spent by the government. For two-parent applicant families, gains from additional welfare and Medicaid payments were fully offset by reductions in earnings, producing no net gain for families. (However, the reduced work hours for second earners in some of these families may have led to nonfinancial benefits that were not measured.)

· MFIP’s enhanced incentives were critical to meeting the program’s goals of rewarding work, increasing families’ incomes and making families and children better off.

MFIP would not have increased families’ incomes if, as was the case under AFDC, benefits had been decreased nearly one dollar for each dollar of earnings. Moreover, MFIP’s incentives were largely responsible for the program’s effects on family and child well-being for single-parent long-term recipients.[7] These effects include the increase in marriage, the reduction in domestic abuse, and the improvements in children’s behavior and school performance.

· The participation mandates or work requirements complemented the incentives by increasing employment and earnings and reducing program costs.

The enhanced incentives were responsible for increased program costs because they increased the likelihood that families would receive some welfare. Adding the participation mandates to the incentives, in contrast, increased full-time work and increased earnings, thereby reducing welfare receipt and program costs from what they would have been if incentives had been offered by themselves. Because the incentives caused some parents to go to work, but also caused some working parents to reduce their hours, these findings suggest that if the goal is to increase full-time employment, incentives should be combined with a work or participation mandate, as in MFIP, or with a minimum hours requirement.[8]

· The financial incentives provided some families with extra benefits because they went to work (single-parent long-term recipients). Others received more benefits when they were already working, allowing some to reduce their work hours (single-parent recent applicants and two-parent families).

In this sense, MFIP’s incentives operated much like the EIC, encouraging some families to enter work and rewarding others who are already working with a supplement. In programs of this type, it is unavoidable that some parents will receive benefits because they are already employed, not because they increase their work effort. This may be an acceptable outcome if increasing the income of working families is a primary goal of the program. At the same time, states can maximize the number of parents who do increase their work effort by coupling incentives with active marketing campaigns to educate participants that “work now pays better than welfare” (and, as discussed above, with work or participation mandates).

· States should be aware that financial incentives and time limits may work at cross purposes.

MFIP’s results indicate that under welfare time limits, states face a choice. Should they encourage recipients to leave welfare quickly, “banking” their lifetime allocation of months on welfare and reserving the welfare safety net as a last resort? Or should they provide enhanced earned income disregards (or other supports) while families are on welfare, which may increase the chance for families to gain firm footing before reaching their time limit, but also have the side effect of keeping families on the rolls and using up their lifetime limit sooner? Some solutions to this dilemma include providing disregards but “stopping the clock” for full-time workers or extending their assistance beyond 60 months, gradually phasing out incentives over time, or providing supports for low-income workers primarily outside the welfare system, where time limits do not apply.

· Increasing the incomes of low-income working families can set in motion a positive chain of effects on families and children.

It is striking that the group whose economic status MFIP improved most dramatically — single parents who were long-term welfare recipients — was also the group for whom it had its most wide-ranging effects: increasing marriage, reducing domestic abuse, and improving children’s well-being. While many studies have shown that children in higher-income families do better on a range of outcomes than those from lower-income families, in non-random-assignment studies it is often difficult to isolate how much of this effect is due to differences in income versus differences in other family characteristics. MFIP’s research design enabled researchers to isolate the effects of incentives alone versus incentives plus mandates. Thus, MFIP’s results provide direct evidence that supplementing the income of very disadvantaged families when they go to work can make a measurable difference in the outcomes for their children.

However, it is important to recognize that MFIP’s financial incentives did more than simply transfer money to working families. MFIP staff actively encouraged parents to take advantage of the new, work-related benefits, and parents in the MFIP group knew that they had been given an important opportunity not available to others in the welfare system. Even parents who did not work gained added discretion in their spending through the provision of Food Stamps as cash. While it is impossible to disentangle the effect of additional income per se from the effect of how this income was provided, both staff and families reported that MFIP felt like a new and different kind of welfare system.

· Changing the structure of welfare benefits and eligibility rules can increase the likelihood of marriage for single parents and improve the stability of marriages for two-parent families.

Previous research has provided mixed evidence about whether welfare programs or transfer programs such as the Negative Income Tax of the 1970s discourage marriage. MFIP’s results offer perhaps the first evidence that changes in the welfare system can increase the likelihood of parents marrying or staying together. MFIP dramatically increased marital stability among two-parent recipient families, and modestly increased rates of marriage for single-parent long-term recipients. For both types of families, MFIP’s financial incentives (and, for two-parent families, its loosened eligibility rules) rather than its participation mandates were primarily responsible for these effects on marriage. Moreover, in two-parent families, by allowing one parent to cut back on hours worked, MFIP may have helped reduce the stress of balancing child care and work, increasing family stability.

· MFIP’s evaluation results provide a starting point for predicting the effects of the statewide MFIP program.

In early 1998, MFIP ended in the seven evaluation counties, and Minnesota implemented a modified version of it statewide to replace its AFDC system. The new program differs from MFIP in several key ways:

· applicants and recipients are required to work 35 hours per week, or to participate in services, after only one to six months of welfare receipt;

· sanctions are increased to 30 percent of the grant level for a nonworking family after the first event of noncompliance;

· the employment and training services are generally more work-first-focused than those in the pilot MFIP program; and

· the program is operating in the context of PRWORA’s five-year time limit on the receipt of federal cash welfare benefits.

Many of these changes are aimed at reducing program costs and increasing the urgency of the work message, and they could affect MFIP’s impacts on families. Some revisions may increase the likelihood of full-time work and of gains in earnings, while others decrease benefits for families, making the net effect of the revisions on total family income difficult to predict. Ultimately, the five-year time limit will also bring effects not seen in the pilot program. Thus, while this evaluation offers a foundation for predicting the effects of statewide MFIP relative to the AFDC system, the revisions to the program should be taken into account when making such projections.

· Minnesota’s characteristics during the pilot phase could affect the generalizability of these results within Minnesota or to other states.

First, the ability of parents to respond to the program by finding jobs may depend critically on both their demographic characteristics and the state of the economy. The economy — nationally and notably in Minnesota — was very strong during the evaluation period, with unemployment rates as low as 3 percent in some counties. Moreover, members of the MFIP research sample had a different composition than welfare recipients nationally: For example, a much higher proportion of Minnesota recipients have high school diplomas and a higher proportion are white. Second, if a state with very low welfare grants increases its earned income disregard, the effect on families’ likelihood of remaining on welfare will be different than in Minnesota, where the AFDC program had higher-than-average benefits. Third, it is possible that changes in community norms or other “community effects” may come into play in a program that is designed to saturate each county — rather than being implemented for a subset of the county’s caseload as was the case in the pilot program. Finally, Minnesota’s welfare recipients had never been subject to mandatory participation requirements before the MFIP pilot program. Once the whole caseload has been subject to these mandates for a period of time, the families that remain may have more labor market difficulties. Similarly, with welfare caseloads falling throughout the state and country, the families who remain on welfare are likely to have more significant barriers to employment than the families who were on welfare during the mid 1990s. If the MFIP pilot program was replicated with changes in one or more of these conditions, the results might be either more positive or negative than those described here.

[1] MFIP defined “long-term” as receipt of welfare for at least two of the prior three years, but more than half of the long-term recipients in the evaluation sample had received welfare for at least five years in their adult lives.

[2] Only differences that are statistically significant at the 10 percent level are considered program impacts, or described as increases or decreases caused by the program. Statistical significance is a measure used to denote the level of certainty that the difference observed between the groups is due to a true program impact rather than to chance. A difference that is statistically significant at the 10 percent level, for example, implies that there is only a 10 percent probability that the observed difference is due to chance and is not the result of the program.

[3] Information on domestic abuse (and child well-being) was only collected for single mothers with school-aged children. For these outcomes, the full report focuses primarily on the results in urban counties.

[4] Information for part of the MFIP evaluation was collected using audio-computer-assisted technology offering privacy to respondents when answering sensitive questions such as the incidence of domestic abuse and the perpetrator of the abuse. This method of data collection was very successful in increasing the validity of information on domestic abuse.

[5] The question of whether policies that would explicitly encourage parents to marry (or discourage them from separating) would be beneficial for children is a controversial one. In the case of MFIP, the marriage effects are the result of voluntary decisions by parents, presumably as a result of either their improved economic status or the reduction of disincentives to marry arising from welfare rules for two-parent families. Because the empirical evidence suggests that, all else equal, a two-parent household provides important advantages for children, it seems likely that marriages resulting from improvements in families’ economic circumstances and voluntary parental decisions will often benefit children.

[6] The numbers in this bullet and the paragraph that follows it — and the corresponding numbers in Table 2 — differ slightly (and in a positive direction) from those in the prepublication version of this report.

[7] Adding the mandates generally neither increased nor decreased the positive effects of the incentives on family and child well-being.

[8] Canada’s Self-Sufficiency Project is an example of a voluntary program that offered a substantial increase in benefits only to parents who worked full time. That program increased full-time employment fairly substantially. A potential limitation of this kind of work requirement is that not all parents can move immediately into full-time work, and the number who can may depend critically on the state of the local economy.