Savings Incentives for Low- and Middle-Income Families

Savings Incentives for Low- and Middle-Income Families

Since the introduction of Social Security in 1935, policy makers have agreed on the principle of a "three-legged stool" to ensure financial security in retirement - Social Security, employer-provided pensions, and personal retirement savings. Yet many older households have little support from the latter two legs of the stool. According to a recent report by the Congressional Research Service, four in ten persons over age 65 receive 90 percent or more of their income from Social Security. Only one-third of older individuals receive any pension income and only one-half receive any asset income; even among those with asset income, the typical person receives less than $1,000 per year.

The primary current policy to encourage retirement savings is the tax deduction for contributions to Individual Retirement Accounts (IRAs) and 401(k) plans and the deferral of taxes on account earnings. However, this approach has not enticed low- and middle-income families to contribute much to such accounts, in part because the value of the tax incentives is fairly low for these families in light of the low marginal tax rates they face.

An alternative policy to encourage retirement savings by low- and middle-income families would be for the government to match their IRA contributions. Many 401(k) plans offer an employer match of employee contributions, and studies suggest that the existence of such a match increases 401(k) saving. Yet this evidence may be of limited use in predicting how a match of IRA contributions would affect savings by lower-income families, since many 401(k) contributors are relatively affluent and firms may be more likely to offer a match when their employees like to save, making it difficult to distinguish the actual effect of the match from the underlying saving propensity of these workers.

In Savings Incentives for Low- and Middle-Income Families: Evidence from a Field Experiment with H&R Block (NBER Working Paper 11680), researchers Esther Duflo, William Gale, Jeffrey Liebman, Peter Orszag, and Emmanuel Saez provide new evidence of the effect of matching IRA contributions on saving from a large-scale, randomized field experiment.

The authors worked closely with H&R Block to design and execute the experiment, which was run in 60 H&R Block tax preparation offices in the St. Louis Metropolitan area between March 5 and April 5, 2005. The experiment was built around H&R Block's Express IRA (X-IRA) product, which allows clients to make IRA contributions at the time of tax preparation and to use part or all of their federal income tax refund to fund the contributions. Each client preparing a tax return was randomly assigned one of three match rates for X-IRA contributions: no match (the control group), a 20% match, or a 50% match. Contributions up to $1,000 were eligible for matching (up to $1,000 for each spouse in the case of married filers). H&R Block paid the direct costs associated with the experiment, including the matching contributions and the cost of training their tax professionals.

The authors' principal finding is that matching can have large effects on IRA participation. Only 3% of individuals in the control group (no match) choose to contribute to an X-IRA, versus 8% and 14% of those in the 20% and 50% match groups, respectively. Matching also has a significant effect on contributions, conditional on participation: average contribution levels (excluding the match) were $765 in the control group, vs. $1,100 and $1,110 in the 20% and 50% match groups. Including the value of matching contributions and incorporating both those who did and did not contribute to an X-IRA, the average value of IRA deposits are 4.5 and 10 times larger in the 20% and 50% match groups than in the control group.

Next, the authors examine which individuals are most likely to take advantage of the match. They find that the effect of the match on participation is larger for individuals with an income tax refund over $500, persons who have other savings, married people, and people with larger incomes. However, they note that the effect of the match is significant even for individuals in the lowest income quartile, who were three times as likely to contribute to an X-IRA (7.5% vs. 2.5%) if assigned to the 50% match group rather than the control group.

Another finding of note is strong "tax preparer effects" - the effect of the match on participation is much larger when the tax preparer had more experience with X-IRAs prior to the experiment. This may indicate that the tax preparer plays an important role in providing clarifying information about the product or giving financial advice to clients.

The authors also check for "gaming" of the system, since individuals could make easy money by contributing to an X-IRA in order to earn the matching contribution then withdrawing the money soon after. They find no evidence of this, however, as people in the matching groups are no more likely to withdraw funds from their IRAs than people in the control group.

Finally, the authors conduct an analysis of the Saver's Credit program, a small existing program that provides a match of contributions to retirement savings accounts through the tax code. They estimate the effects of the Saver's Credit program on savings to be far smaller than those observed in the H&R Block experiment. The authors suggest that people's lack of knowledge about the Saver's Credit program may account for the difference, highlighting the important role of information and simplicity in programs designed to promote saving.

In the end, the results of this experiment are both encouraging and discouraging for those who might seek to use matching to encourage saving by low- and middle-income households. On the one hand, the match of IRA contributions led to very substantial increases in both participation and contributions to IRAs, yet even with the 50% match, only one in seven participants chose to contribute to an IRA. The authors conclude that "the combination of a significant and readily understandable match for saving, easily accessible savings vehicles, the opportunity to use part of an income tax refund to save, and professional assistance could generate a significant increase in retirement saving participation and contributions, even among middle- and low-income households."

The authors thank H&R Block for the collaboration and resources it has devoted to this experiment. They also gratefully acknowledge financial support from the Pew Charitable Trusts, the Sloan Foundation, and the National Science Foundation.

National Bureau of Economic Research, 1050 Massachusetts Ave., Cambridge, MA 02138; 617-868-3900; email:

Contact Us