Revised July 23, 2003

by Robert Greenstein

An Updated Version of this report is available here:
Earned Income Tax Credit Overpayment and Error Issues


PDF of this report

View Related Analyses

If you cannot access the files through the links, right-click on the underlined text, click "Save Link As," download to your directory, and open the document in Adobe Acrobat Reader.

The magnitude of current EITC overpayments is not known.  The last study of EITC overpayments, an IRS study of overpayments in tax year 1999, has significant methodological shortcomings that likely result in an overstatement of the overpayment rate.  In addition, substantial changes have been made in the EITC since tax year 1999 to reduce overpayments, including two significant measures enacted in 1997 but only partially in effect by 1999, and four further changes enacted in 2001.  A recent paper by the Treasury’s two leading EITC experts estimates that the 2001 changes could reduce EITC overpayments by approximately $2 billion a year.[1]  Some recent statements made by Treasury and IRS officials in defending a controversial new IRS initiative to reduce EITC overpayments have overstated the findings of the 1999 study and overlooked the findings of the Treasury’s experts on the impacts of the 2001 legislation.  While EITC overpayments clearly remain substantial, their dimensions are not known.

What the IRS Study of Errors in 1999 Reported

The IRS study[2] estimated that the EITC overpayment rate in tax year 1999 was between 27% and 31.7%, and placed the overpayment amount at between $8.5 billion and $9.9 billion.  This range reflects the fact that a significant fraction of families selected for the study did not respond to the IRS when contacted.  This is a typical occurrence when studies of this nature are conducted in means-tested programs, and past evaluations in other means-tested programs have found that many people who do not respond in such circumstances actually are eligible for program benefits but are intimidated by the request, do not understand the request (due to literacy problems or lack of fluency in English), or simply wish to avoid the hassle.  (For example, two USDA studies of families whose children lost free or reduced price school lunches because the families failed to respond to a request for verification found that more than 80 percent of the non-responders were actually eligible for free or reduced-price lunches.)

The 27% EITC overpayment rate estimate — the low end of the range and in the IRS study — assumes that overpayments among the non-responders occurred at the same rate as among those who did respond.  The 31.7% overpayment rate assumes that all of the non-responders were ineligible for every dollar of EITC benefits they received.  The 31.7% figure, which represents the upper bound of the range, clearly overstates the true error rate.

Are The Study Results a Valid Measure of Overpayments in 1999?

Many analysts regard the 27% to 31.7% overpayment rate estimate as being of questionable validity.  An analysis issued by the Treasury Inspector General for Tax Administration found “significant weaknesses in its [the IRS’] methodology” and reported that some of the audits of tax filers whose EITC claims were examined as part of the IRS study “lacked the necessary information to support the IRS’ results.”[3]  In addition, at a Brookings Institution conference on tax administration in November 2002, both Eugene Steuerle, a Senior Fellow at the Urban Institute and a former Deputy Assistant Secretary of the Treasury for Tax Analysis in the Reagan Administration, and Nina Olson, the National Taxpayer Advocate at the IRS, raised methodological concerns about the study.

Olson observed that the experience of the National Taxpayer Advocate’s office and low-income taxpayer clinics shows that when the validity of EITC claims is being challenged by the IRS, an interactive process involving numerous contacts with low-income taxpayers whose claims are being questioned is often necessary to clarify that a claim that IRS officials initially believe to be erroneous is, in fact, valid.  The IRS study of EITC overpayments in 1999 did not provide for a sufficient process of this nature, she explained.  This may have resulted in premature judgments in a number of cases that EITC payments were made in error.

The National Taxpayer Advocate’s recent report to Congress issued in December 2002, explains that “Taxpayers often obtain completely different (favorable) results from the original IRS determination” when they receive assistance from the Taxpayer Advocate Service in EITC audits and other IRS actions challenging EITC claims.  The Taxpayer Advocate’s report to Congress notes that Taxpayer Advocate Service (TAS) assistance resulted in reversals of IRS rulings in 51 percent of the EITC audit cases in which the TAS was involved in fiscal year 2002.  The report further explains that “TAS intervention often involves helping the taxpayer understand what documentation is needed and then working with the taxpayer to find alternative ways of substantiating an EITC claim.”

This suggests that some portion of the EITC claims classified as being erroneous in the IRS study may have been valid claims for which the tax filer, operating without assistance from a knowledgeable party, could not provide documentation that satisfied the IRS within the timeframes the IRS allowed.  In her most recent report to Congress, issued on June 30, 2003, Olson states that for these reasons, she believes the 1999 study “overstates the overclaim rate.”[4]

Further problems with the study of EITC overpayments in 1999 that result in an overstatement of EITC errors are acknowledged by the IRS in the body of the study itself.  One of the leading causes of EITC overpayments in 1999 consisted of cases where a mother and her child lived with a grandmother or similar relative, and the relative had higher adjusted gross income than the mother.  Under a complicated rule known as the “AGI tiebreaker rule” that was in effect in 1999, if the grandmother had higher adjusted gross income than the mother, the mother and grandmother were supposed to know that the grandmother — rather than the mother — was the proper person to claim the child for the EITC.  (The mother would have been eligible to claim her child for the dependent exemptions and the child tax credit, but not for the EITC).

Suppose the mother earned $18,000 in 1999 and the grandmother earned $20,000.  The overpayment would have equaled the amount by which the EITC that the mother received exceeded the EITC that the grandmother should have received.  In its study, however, the IRS counted the entire EITC payment to the mother as being in error, rather than determining the actual erroneous amount.  (The IRS did so rather than attempt to determine the EITC amount the grandmother would have received had she filed for the credit.)  This practice resulted in an overstatement of EITC overpayments.  (Note:  Since 1999, this complex rule has been changed in a manner that should eliminate the vast majority of the errors in these cases.)

In addition to issues regarding the accuracy of the study findings, there are questions regarding whether the study provides an adequate basis for designing IRS’ new EITC error-reduction initiative, known as pre-certification.  Gene Steuerle has voiced concerns that the methodology for the study was not developed well enough to be able to distinguish different sources of error.  For example, Steuerle notes, the study does not tell us approximately what percentage of errors are due to two people filing for the same child, and how often such double filing involves filers living apart or together.  Steuerle observes that such information would have helped greatly in figuring out how to design any administrative procedure — pre-certification or other — aimed at reducing error rates.  He counsels that any new administrative experiment, such as pre-certification, should be engaged only after a clear-cut statistical design has been developed that would provide fairly precise information on both the sources of error and the ability of the new procedures to detect and prevent different types of error.[5]

Why the Study Results May No Longer Reflect Current Rates of Overpayment

Major statutory changes were enacted in both 1997 and 2001 to reduce EITC overpayments.  Most of these changes had not yet taken effect in 1999.  In addition, the IRS has instituted some improved administrative procedures since 1999, including efforts to check Social Security numbers of second spouses listed on tax returns.  The effects of these statutory changes and administrative efforts are not reflected in the IRS study of overpayments in tax year 1999.

The study acknowledges this fact.  The study’s executive summary states: “The estimates presented in this report do not reflect the impact of recent legislative and administrative changes that are likely to improve compliance rates.”[6]  (See box below.)

The 1997 Reforms

A key change enacted in 1997 gives the IRS access to a new National Case Registry of child support cases.  (Establishment of the National Case Registry, or NCR, was mandated by the 1996 welfare law.)

The 1997 legislation also required that when a child applies for a Social Security number (which usually happens now when a child is born), the Social Security numbers of the child’s parents be included on the application.  Previously, the parents SSNs were not linked to the child’s.

The 2001 Statutory Changes

The 2001 changes were major.  A recent paper by Treasury experts Janet Holtzblatt and Janet McCubbin, “Complicated Lives:  Tax Administrative Issues Affecting Low-Income Filers” (to be published in a forthcoming Brookings volume on issues in tax administration), estimates that the EITC changes enacted in 2001 could reduce EITC overpayments by approximately $2 billion a year.  These changes include:

The measures enacted in 2001 are significant both because of the authority they accord the IRS to deny claims from filers showing up as non-custodial parents in the National Case Registry and because of the important simplifications they make.  Many experts believe the EITC’s inordinate complexity is one of the main causes of EITC errors.  The number of pages of IRS instructions for the EITC exceeds the number of pages of instructions for the complicated individual Alternative Minimum Tax.

 New IRS Data on Errors Are Promising

New IRS data on EITC claims made this year through May 4, 2003 provide some indication that progress is being made.  These IRS data show the degree to which EITC claims are found during processing to have any of 17 different types of errors that lead to a denial of the claim or any of five other types of errors that lead IRS to recompute the EITC amount.  The percentage of EITC claims found during processing to have any of these 22 types of errors fell from 5.92 percent of claims for tax year 2000 to 3.55 percent of claims for tax year 2002, a fairly impressive decline.[8]  The data also indicate that a significant portion of the decline in the presence of these types of errors is due to the simplification measures enacted in 2001.

IRS Report States It Does Not Reflect the Impact of Recent Statutory Changes

The Executive Summary of the IRS report concludes with the following paragraph:

“The estimates presented in this report do not reflect the impact of recent legislative and administrative changes that are likely to improve compliance rates.  The Economic Growth and Tax Relief Reconciliation Act of 2001 (EGTRRA) substantially changed the rules applicable to individuals who have the same qualifying child.  It also modified the definition of a foster child and two EITC income concepts.  These provisions became effective in 2002.  In addition, the Internal Revenue Service (IRS) recently implemented the Dependent Database (which includes data from the Federal Case Registry of Child Support Orders and a Social Security Administration file linking parent and child social security numbers) for identifying returns with potential EITC errors for pre-refund audits.  As a result of the 2001 tax law change, the IRS will be authorized to deny EITC claims if the federal case registry indicates that the taxpayer is the child’s non-custodial parent.”

Since these data refer to types of errors that often are caught in processing, with the result that no overpayment is made, they do not translate directly into a reduction in the overpayment rate.  Nevertheless, some of these types of errors represent errors that are not always caught in processing and thus can lead to overpayments in some cases.  To the extent that the new data indicate that the incidence of these types of errors has fallen, it is likely that overpayments associated with these types of errors have declined.

End Notes:

[1] Janet Holtzblatt and Janet McCubbin, “Complicated Lives:  Tax Administrative Issues Affecting Low-Income Filers,” January 2003.  The authors are analysts in the Treasury Department’s Office of Tax Policy.  The paper will be published in a forthcoming Brookings volume on issues in tax administration.

[2] Internal Revenue Service, “Compliance Estimates for Earned Income Tax Credit claimed on 1999 Returns,” February 28, 2002.

[3] Treasury Inspector General for Tax Administration, “There are significant Weaknesses in the Internal Revenue Service’s Efforts to Measure Earned Income Tax Credit Compliance,” December 2001.

[4] Taxpayer Advocate Service, Internal Revenue Service, “The National Taxpayer Advocate’s Report to Congress:  Fiscal Year 2004 Objectives,” June 30, 2003, pp. 20-21.

[5] Personal communication from Gene Steuerle, May 2, 2003.

[6] Internal Revenue Service, op.cit., p.4.

[7] Non-custodial parents making child support payments may be eligible to claim the child for the dependent exemption and the child tax credit, but not for the EITC.  This inconsistency in rules may cause inadvertent error.  The errors in this area appear to represent both willful errors and inadvertent errors.

[8] These data cover tax year 2002 claims processed through May 4, 2002, as well as tax year 2000 claims processed through April 25, 2001, which are the most comparable data.  The source of these data is IRS, “Monthly Operational Review of Earned Income Credit,” May 16, 2003 and “Monthly Operational Review of Earned Income Credit,” June 6, 2002.