The EITC compliance problem
Jeffrey B. Liebman*

ince the 1960s, economists have advocated the use of the tax system as a means of transferring income to low-income families. Milton Friedman and James Tobin argued that replacing the welfare system with a negative income tax would provide greater incentives for work and would reduce administrative costs. Studying the Earned Income Tax Credit (EITC) offers an opportunity to learn how well the tax system functions in performing tasks traditionally handled by the welfare system.

Recent research on the EITC mostly confirms Friedman's and Tobin's conjectures that the tax system can be used to transfer income with fewer distortions to labor supply and with lower administrative costs than are achieved in the current U.S. welfare system. Papers by Eissa and Liebman and Meyer and Rosenbaum find that the EITC increases labor force participation among single women, and Eissa and Liebman fail to observe the predicted adverse effect of the EITC phase-out on hours of work for those already in the workforce. In contrast, most evidence indicates that AFDC discourages work. The administrative cost of the EITC is difficult to separate out from other IRS costs, but it is likely to be less than 1 percent of dollars transferred. The administrative costs of AFDC in 1995 were 16 percent of benefits paid, in large part because of the need to pay for case workers. By relying on taxpayers to self-report their eligibility on their tax returns, the EITC  avoids this cost. In addition, administering benefits through the tax system avoids imposing on transfer recipients the stigma and direct costs of participating in the welfare system. These low costs may explain why the take-up rates Karl Scholz has estimated for the EITC are somewhat higher than those that have been estimated for AFDC.

Experience with the EITC suggests, however, that there is a major disadvantage to using the tax system to transfer income to the poor: high rates of noncompliance. Tabulations from the Internal Revenue Service's Taxpayer Compliance Measurement Program (TCMP) indicate that one-third of 1985 and 1988 recipients were ineligible for the credit, primarily because they did not have children entitling them to claim the credit. Since 1988, there have been changes to EITC eligibility rules designed to reduce non-compliance, and the IRS has adopted a series of measures to reduce erroneous EITC payments. However, the maximum EITC has more than doubled in real terms, increasing the return to fraud. In addition, the technology of tax filing has changed, with the percentage of EITC taxpayers filing electronically increasing from less than 1 percent in 1988 to 26 percent in 1994. A 1997 IRS study found that in tax year 1994, 26 percent of EITC dollars were overclaimed, down from 35 percent in 1988. The study also estimated that if the additional EITC compliance efforts implemented since 1994 had been in place, the overpayment rate would have been 21 percent.

These basic results raise three important questions. First, who are the ineligible EITC recipients? If the ineligible taxpayers are low-income taxpayers with children, and therefore similar to eligible taxpayers, then society may be more willing to tolerate high noncompliance rates, especially given the EITC's low administrative costs. Second, are the ineligible taxpayers making inadvertent errors or are they deliberately committing fraud? Even if the ineligible taxpayers are mostly low income families with children, society might want to assign a low or even negative social welfare weight to dollars transferred to noncompliant taxpayers who are committing fraud. Third, how much of an effect have IRS efforts to reduce EITC noncompliance had on noncompliance rates and what is the prospect for further reducing EITC overpayments?

Who are the ineligible EITC recipients?
In 1988, 86 percent of ineligible EITC recipients had incomes low enough to qualify for the EITC if they had been otherwise eligible. Most were ineligible because they did not have a child entitling them to claim the credit.

Under pre-1991 rules, a child qualified a taxpayer for the EITC if the child lived with the taxpayer for more than half of the year and if the taxpayer provided at least half the cost of supporting the child. The support test implied, for example, that a taxpayer who received $4,000 in AFDC benefits and $3,000 in earnings would not have been eligible to receive the EITC Largely in response to complaints that the support test was difficult to enforce, OBRA 1990 eliminated the support test as an eligibility requirement for the EITC, leaving the child residency test as the main eligibility requirement.

In order to learn more about the characteristics of ineligible EITC recipients, I became a special sworn Census Bureau employee and used a data set which matched the March 1991 Current Population Survey to the tax returns of CPS respondents. I estimated the percentage of 1990 EITC recipients who told the CPS interviewer that there were no children living in the household. I found that between 10 and 21 percent of all EITC recipients lacked children and therefore were unlikely to be eligible for the credit. Male taxpayers filing as household heads were particularly likely to be ineligible - between 25 percent and 53 percent lacked children, depending on the exact measure of eligibility used. Many of these male taxpayers may be noncustodial parents. These results suggest that most EITC dollars are indeed reaching low-income households with children - 87 percent of EITC recipients had children living in their households at the time that they received the credit.

Taxpayer error or taxpayer fraud?
In a recent paper, I estimate the share of EITC noncompliance that is due to error and the share that is due to tax evasion. The basic insight motivating the estimation is that the amount of inadvertent error should not respond to the size of the tax credit available to a taxpayer who claims the EITC. In contrast, tax evasion should increase when the return to such behavior grows. My data for this study were 1985 and 1988 cross-sections from the IRS's Taxpayer Compliance Measurement Program. In these data, I can observe the number of children that each taxpayer claimed on his or her tax return and the number of children that the auditor determined the taxpayer was entitled to claim. I test whether the probability that a childless taxpayer claims a child on his or her tax return depends on the reduction in tax liability from making such a claim. Because the Tax Reform Act of 1996 increased the return to claiming children by different amounts at different levels of income, I am able to separate the effect of the EITC from any underlying relationship between income and compliance behavior that happens to be correlated with the EITC in the cross section. I find that the EITC is positively correlated with the probability of noncompliance. My results imply that of every additional dollar spent on the EITC, 24 cents go to ineligible taxpayers - 11 cents to taxpayers who are responding to the EITC incentive to wrongly claim children and 13 cents to taxpayers who would wrongly claim children even in the absence of the EITC (and who could be making inadvertent errors).

Is noncompliance declining?
During the 1990s, the IRS had taken a number of important steps to reduce EITC noncompliance. For example, before sending out tax refunds, the IRS now verifies the Social Security numbers of children claimed on tax refunds. While overpayment rates have declined from 35 percent in 1988 to around 21 percent today, a decline of roughly this magnitude would have been expected without additional IRS efforts because of the OBRA 1990 elimination of the support test which made many previously ineligible taxpayers eligible for the EITC. Therefore, it is worth asking whether recent compliance efforts have had any effect.

While it is impossible to be certain, it seems likely that the reforms have had a substantial impact. The results of my study of the noncompliance response to the 1987 expansion of the EITC indicate that a 45 percent increase in the maximum EITC increased the noncompliance rate by 14 percent. It is highly speculative to extrapolate from these results to more recent EITC expansions. Nonetheless, between 1990 and 1994 the value of the EITC more than doubled, so we might have predicted an increase in noncompliance of 33 percent from the 21 percent post support test level to 28 percent. Thus, if recent reforms have reduced the rate of noncompliance to 21 percent, then they have eliminated one-quarter of EITC noncompliance. Last year, at the Treasury Department's request, Congress passed six new reforms that are likely to reduce EITC noncompliance rates further - possibly to a level that is below the 17 percent overall individual income tax noncompliance rate.

Implications
Using the tax system to transfer income to low-income families, it is possible to reduce administrative costs and increase program participation rates by eliminating welfare case workers. However, without case workers it becomes difficult to verify recipient eligibility. In the case of the EITC, we have seen that very high rates of noncompliance can occur when recipients are permitted to self-determine eligibility status. The tradeoffs between administrative costs and participation rates on the one hand and compliance rates on the other occur within a given transfer system as well as across different systems. In the past few years, the IRS has devoted considerable additional resources to verifying taxpayer eligibility for the EITC and to recognizing emerging patterns of fraud. While noncompliance rates have fallen, it is clear that the IRS actions have raised administrative expenditures and have discouraged some eligible taxpayers from receiving the credit.

Back
*Jeff Liebman is assistant professor of Public Policy at the John F. Kennedy School of Government, Harvard University.

TOP