Many U.S. households do not have enough savings to help manage temporary losses of income or increased expenditures from unexpected events. Increased savings might particularly help low- and moderate-income families avoid resorting to high-cost (sometimes “payday”) loans or failing to meet monthly rent bills and minimum credit card payments. To support the buildup of savings, some experts have proposed encouraging low- and moderate-income individuals to save part of their annual tax refunds, capitalizing on these large, one-time influxes of cash. Some past research suggests that this approach might be promising; other research indicates that many low- and moderate-income individuals need their refunds to pay bills or reduce debt.

The SaveUSA evaluation, a randomized controlled trial launched by MDRC in 2011, contributed strong evidence relating to several aspects of this debate. SaveUSA (studied and operated as part of the Mayor’s Fund and Center for Economic Opportunity Social Innovation Fund Project) was a voluntary tax-time savings program. Through a 50 percent matching incentive, SaveUSA aimed to make the accumulation of emergency savings more attractive to low- to moderate-income families. When filing their taxes, these households were presented with the opportunity to directly deposit some or all of their tax refunds into special savings accounts at participating financial institutions. If they kept the amount that they pledged to save in the savings account for approximately one year, then they would receive a 50 percent match.

SaveUSA targeted tax-time savings because tax refunds are generally the largest one-time cash infusion for low- to moderate-income individuals. The SaveUSA model also employed lessons from behavioral economics research, such as simplified options, a separate account for savings, electronic deposit into the account, incentives to maintain savings, and disincentives to remove even small amounts of savings. SaveUSA also differed from many other savings programs in that the use of both the initial deposit amount and any savings match were unrestricted, that is, they could have been used for any purpose.

Few programs exist that help low- and moderate-income individuals build up unrestricted savings using tax refunds, and rigorous studies of the effects of such programs are even rarer. A report describing SaveUSA’s implementation and early effects (after 18 months) was released in early April 2014. A final report, which examined SaveUSA’s effects over a 42-month follow-up period, was released in December 2015.


Agenda, Scope, and Goals

Operating in New York City, Tulsa, OK; San Antonio, TX, and Newark, NJ, SaveUSA enrolled approximately 2,500 participants in the study’s first year. The program’s effects  were measured in New York City and Tulsa, where enrollees were randomly assigned to either the SaveUSA group (eligible to open a SaveUSA account) or the regular tax filers group (not eligible to do so). This randomized controlled trial method allowed MDRC to follow both groups over time and compare their outcomes; any differences in outcomes represent the impacts of the SaveUSA program.

The Newark and San Antonio locations did not participate in the randomized controlled trial; however, MDRC used data from those locations along with data from the other two to evaluate the implementation of the program.

The SaveUSA study addressed several groundbreaking research questions:

  1. Can a program that provides incentives to save, such as SaveUSA, be implemented in different settings? To whom does it appeal and to what extent?
  2. Can such a program actually increase savings beyond what low- and moderate-income tax filers would have accumulated on their own? Can it help people develop a habit of saving?
  3. If SaveUSA succeeds in increasing savings beyond what would have happened in the absence of the program, can the increases improve financial stability or reduce hardship in any identifiable way?

The findings from the SaveUSA study provide strong evidence indicating that small increases in amounts of unrestricted-use savings can improve the financial well-being of low- and moderate-income individuals, and also offer guidance on the design and implementation of savings initiatives.

Design, Sites, and Data Sources

SaveUSA began in 2011. As part of the evaluation, deposits into SaveUSA accounts continued through the 2013 tax-preparation season and savings matches continued through early 2014. In New York City and San Antonio, the program continued beyond the evaluation period.  A cumulative total of 17 Volunteer Income Tax Assistance (VITA) sites offered SaveUSA across the four participating cities. To be eligible for SaveUSA, tax filers had to be at least 18 years old and meet certain income requirements (earn $50,000 or less annually for filers with dependents, and generally earn $25,000 or less for filers without dependents).

On their tax returns, SaveUSA participants instructed the Internal Revenue Service or state taxing agency to deposit at least $200 from their tax refunds directly into the SaveUSA accounts. Each participant also pledged to keep a certain amount of the initial deposit, from $200 to $1,000, in the account until February 1 of the following year. A participant who fulfilled this pledge received a 50 percent match, up to $500, on the pledged savings amount.

Five banks and one credit union voluntarily participated in SaveUSA and set up SaveUSA accounts. Staff members from these partnering financial institutions facilitated the on-site opening and continuing maintenance of the SaveUSA accounts, but the institutions were not required to determine match eligibility or match amounts. MDRC performed these functions by analyzing periodic data files shared by the financial institutions with MDRC.

MDRC used the following data sources to evaluate the SaveUSA program:

  • Baseline surveys
  • 18-month and 42-month follow-up surveys
  • Tax return data
  • SaveUSA account activity and balance data from the participating financial institutions
  • Interviews with VITA staff members
  • Observations of program operations