• Intervention: Earnings supplements for welfare recipients who find work.
  • Key Findings: Randomized controlled trial shows a sizable increase in employment and earnings, and reduction in poverty, for single-parent, long-term welfare recipients (but not for other welfare recipients).

Description of the Intervention

The Minnesota Family Investment Program (MFIP) was implemented as a pilot program in seven urban and rural counties from 1994 to 1998 and served as an alternative to the federal welfare program (Aid to Families with Dependent Children (AFDC)). Minnesota’s current MFIP program, adopted in 1998 as a successor to this MFIP pilot program, is based largely on the pilot, but has slightly different features (namely, stricter work requirements and less generous financial incentives).

Under the MFIP pilot program, when a welfare recipient or applicant went to work, he or she received a cash supplement amounting to a 20 percent increase in cash benefits to offset work-related expenses (e.g. transportation, work clothes etc.). The MFIP pilot program also provided these individuals with full welfare benefits until their income (i.e., job earnings plus welfare benefits) reached 140 percent of the poverty line. By contrast, AFDC provided no cash supplement, and reduced welfare benefits one dollar for every dollar increase in earnings.

In addition, the MFIP pilot program required all participants who had received welfare for 1-2 years to participate in state-provided employment/training activities – i.e. job search assistance, education, or job training programs (whereas AFDC only included a training requirement for two-parent families). The MFIP pilot program also simplified the benefits process for its participants by combining AFDC, food stamp benefits, and state-funded cash assistance into one program with a single set of rules and procedures, and a single monthly payment.

The net cost of the MFIP pilot program to the government was between $2,000 and $4,700 more per family per year than AFDC in 2006 dollars (depending on the family’s demographic characteristics). These costs primarily consist of the pilot program’s more generous benefit payments and the cost of families’ continued enrollment in Medicaid while receiving benefits.

Click here for a more detailed description of the MFIP pilot program. See especially pages 3-10.

Click here for more information on the current MFIP program.


This program was evaluated in one randomized controlled trial of 11,473 families in seven Minnesota counties whose welfare status ranged from long-term welfare recipients to new welfare applicants, and who were eligible to participate in the MFIP pilot program based on their age, net annual income and total assets. Families were randomly assigned to the MFIP pilot program or to a control group that participated in AFDC.

Effects on workforce outcomes for the full sample 2 ½ years after random assignment:

  • Effects of the MFIP pilot program on single-parent, long-termwelfare recipient families(versus the control group):
    • 35% higher employment rate (50% employment rate for MFIP pilot group vs. 37% for controls)
    • 23% higher average annual earnings ($3,820 vs. $3,116)
    • Significantly lower percentage of families with overall income (i.e. earnings + welfare receipts) below the poverty line (75% vs. 85%).
  • Effects of the MFIP pilot program on two-parent long-term welfare recipient families(versus the control group):
    • The MFIP pilot program had no significant effect on employment, and resulted in 18% lower earnings ($8,772 vs. $10,728)—possibly because one parent cut back his/her work hours in response to increased MFIP benefits.
  • The MFIP pilot program had only a marginal impact on the employment and earnings of recent welfare applicants (both single-parent and two-parent applicants).

Effects on family structure/stability 3 years after random assignment (based on a survey of a randomly-selected subsample of 4586 families):

  • MFIP pilot long-term welfare recipient families (both single-parent and two-parent) were 40-60 percent more likely to be married than their control group counterparts.
  • The MFIP pilot had no effect on the marriage rate of recent welfare applicant families.
  • MFIP pilot single-parent, long-term welfare recipient mothers with young children living in urban areas were 18% less likely to have reported domestic abuse against themselves or their children (e.g., by a boyfriend) during the 3-year period than their control group counterparts (49.1% vs. 59.6%). This finding should be treated with caution, as it applies to a fairly specific subgroup. Subgroup effects of this type can sometimes appear by chance.

Effects on workforce and family outcomes for the full sample 5-6 years after random assignment (i.e., 2-3 years after the MFIP pilot program ended and was replaced by the modified version of MFIP now operating statewide):

  • The earlier large effects for single-parent, long-term welfare recipients on employment, earnings, and overall income faded to near zero. (One plausible explanation for this result is that, when MFIP became a statewide program, members of the MFIP pilot and control groups who remained in the welfare system became subject to the same welfare rules and benefits.)
  • Importantly, however, the large effects were sustained for the most disadvantaged families among the single-parent, long-term welfare recipients– namely, those headed by a parent who (i) had not worked in the year prior to entering the study, (ii) did not have a high school diploma or GED, and (iii) had received welfare payments for at least 11 of the 12 months prior to entering the study. The effects on these most disadvantaged families were as follows:
    • 32% higher employment rate (52% for the MFIP pilot group vs. 39% for controls)
    • 42% higher annual earnings ($6,604 vs. $4,664)
    • 21% higher annual income (i.e., earnings + welfare receipts) — $10,552 vs. $8,694.
    • No effect on annual welfare benefits.

While the effects for this subgroup are large, they should be treated with caution because they apply to a specific subgroup that constitutes just 4% of the overall sample. Subgroup effects of this type can sometimes appear by chance.

  • For all other subgroups of welfare recipients, there were no consistent, meaningful effects on earnings, employment, welfare benefits, or overall income. These other subgroups include two-parent, long-term welfare recipient families and recent welfare applicants (both single-parent and two-parent).
  • For the entire sample, as well as each major subgroup, there were no consistent, meaningful effects on marriage or divorce rates, or on educational achievement of the children. (There was suggestive, but inconsistent, evidence of positive effects on children’s educational achievement for specific subgroups; however, such subgroup effects should be treated with caution as they can sometimes appear by chance.)

Discussion of Study Quality (click here for glossary of terms)

  • The study had low attrition and a long-term follow-up: At the 2½ and 6 year follow-ups, official outcome data (administrative records of earnings, employment, welfare receipt, etc.) were collected for 100% of the sample. The 3-year follow-up survey (measuring marital status, etc. for a randomly-selected subsample of study participants) obtained outcome data for 81% of the subsample.
  • The study measured outcomes using an intention-to-treat analysis.
  • Prior to the intervention, the MFIP pilot and control groups did not differ significantly in the preponderance of their characteristics.
  • In measuring most outcomes (employment, earnings, welfare benefits, marriage and divorce rates, children’s educational achievement), the study used official government and school records rather than self-reported survey data.
  • This was a large, multi-site study conducted in typical welfare agencies with former AFDC caseworkers carrying out the intervention, thus providing strong evidence of the MFIP pilot program’s effectiveness in real-world settings.


(Click on linked authors’ names for their contact information)