Earned Income Credit: IRS' Tax Year 1994 Compliance Study and Recent
Efforts to Reduce Noncompliance (Letter Report, 07/28/98,
GAO/GGD-98-150).

Pursuant to a congressional request, GAO reviewed the Internal Revenue
Service's (IRS) 1994 Earned Income Credit (EIC) compliance study,
focusing on: (1) evaluating IRS' study methodology to determine if the
reported results were reasonably accurate; (2) identifying the primary
sources of EIC noncompliance found in the study; and (3) determining
whether recent IRS compliance efforts are designed to address the
primary sources of noncompliance.

GAO noted that: (1) IRS' estimate of $4.4 billion in EIC overclaims has
a 95-percent confidence interval of $4.0 billion to $4.9 billion; (2)
GAO's evaluation of the study methodology showed that the estimate is
reasonably accurate and representative of EIC claimants filing between
January 15 and April 21, 1995; (3) some aspects of the study methodology
affected the precision of the results; but, given the scale of the
findings, these limitations do not affect the study's message or its
usefulness in designing compliance approaches; (4) although it is a
reasonable estimate of EIC overclaims, the entire $4.4 billion should
not be viewed as a potential savings to the government had IRS somehow
been able to prevent or correct all of these errors; (5) for returns
filed with an EIC claim, the tax year 1994 study was designed to
evaluate taxpayers' compliance with each EIC eligibility filing
requirement, to produce an overall estimate of EIC amounts claimed in
error, and to identify the sources of these errors; (6) the study was
not designed to detect or quantify EIC claims that taxpayers could have
made; (7) the largest source of taxpayer error identified by the tax
year 1994 study relates to EIC requirements that are difficult for IRS
to verify--those related to qualifying children; (8) unlike income
transfer programs, the EIC was designed to be administered through the
tax system; (9) this choice generally should result in lower
administrative costs and higher participation rates and emphasizes that
the credit is for working taxpayers; (10) EIC eligibility is difficult
for IRS to verify through its traditional enforcement procedures; (11)
thoroughly verifying qualifying child eligibility requires IRS to do an
audit of the type done in the EIC compliance studies; (12) with new
enforcement tools provided by Congress and an increase in funding
designated for EIC-related activities, IRS began implementing in fiscal
year 1998 a plan that, over a period of 5 years, calls for attacking EIC
noncompliance; (13) most of the efforts that make up the EIC compliance
initiative had not progressed far enough at the time GAO completed its
audit for it to make any judgment about their effectiveness; (14) IRS
plans to measure the overall impact of the compliance initiative on the
overclaim rate through annual studies of EIC compliance starting with a
baseline study of tax year 1997 returns; and (15) IRS plans to measure
the results of the individual initiative components implemented in 1998.

--------------------------- Indexing Terms -----------------------------

 REPORTNUM:  GGD-98-150
     TITLE:  Earned Income Credit: IRS' Tax Year 1994 Compliance Study 
             and Recent Efforts to Reduce Noncompliance
      DATE:  07/28/98
   SUBJECT:  Noncompliance
             Income taxes
             Tax credit
             Tax administration
             Overpayments
             Tax return audits
             Tax refunds
             Tax violations
IDENTIFIER:  Earned Income Tax Credit
             IRS Earned Income Credit Compliance Initiative
             EIC
             
******************************************************************
** This file contains an ASCII representation of the text of a  **
** GAO report.  Delineations within the text indicating chapter **
** titles, headings, and bullets are preserved.  Major          **
** divisions and subdivisions of the text, such as Chapters,    **
** Sections, and Appendixes, are identified by double and       **
** single lines.  The numbers on the right end of these lines   **
** indicate the position of each of the subsections in the      **
** document outline.  These numbers do NOT correspond with the  **
** page numbers of the printed product.                         **
**                                                              **
** No attempt has been made to display graphic images, although **
** figure captions are reproduced.  Tables are included, but    **
** may not resemble those in the printed version.               **
**                                                              **
** Please see the PDF (Portable Document Format) file, when     **
** available, for a complete electronic file of the printed     **
** document's contents.                                         **
**                                                              **
** A printed copy of this report may be obtained from the GAO   **
** Document Distribution Center.  For further details, please   **
** send an e-mail message to:                                   **
**                                                              **
**                                            **
**                                                              **
** with the message 'info' in the body.                         **
******************************************************************


Cover
================================================================ COVER


Report to the Chairman, Committee on Ways and Means, House of
Representatives, and to the Honorable
Larry E.  Craig, U.S.  Senate

July 1998

EARNED INCOME CREDIT - IRS' TAX
YEAR 1994 COMPLIANCE STUDY AND
RECENT EFFORTS TO REDUCE
NONCOMPLIANCE

GAO/GGD-98-150

IRS' Tax Year 1994 EIC Compliance Study

(268815)


Abbreviations
=============================================================== ABBREV

  AGI - adjusted gross income
  EIC - Earned Income Credit
  HHS - Department of Health and Human Services
  IRS - Internal Revenue Service
  OTA - Office of Tax Analysis
  QRP - Questionable Refund Program
  SOI - Statistics of Income
  SSA - Social Security Administration
  SSN - Social Security number
  TRA97 - Taxpayer Relief Act of 1997

Letter
=============================================================== LETTER


B-278069

July 28, 1998

The Honorable Bill Archer
Chairman, Committee on Ways and Means
House of Representatives

The Honorable Larry E.  Craig
United States Senate

In April 1997, the Internal Revenue Service (IRS) published the
results of its tax year 1994 Earned Income Credit (EIC) compliance
study.  The study showed that of $17.2 billion in EIC claimed during
the study period, taxpayers overclaimed about $4.4 billion, or about
26 percent.  In response to your request, we (1) evaluated IRS' study
methodology to determine if the reported results were reasonably
accurate, (2) identified the primary sources of EIC noncompliance\1
found in the study, and (3) determined whether recent IRS compliance
efforts are designed to address the primary sources of noncompliance. 


--------------------
\1 "Noncompliance," as used in this report, includes errors caused by
taxpayer mistakes, negligence, or fraud.  Determining whether an EIC
claim is fraudulent requires knowing the taxpayer's intent, which is
difficult to prove. 


   RESULTS IN BRIEF
------------------------------------------------------------ Letter :1

IRS' estimate of $4.4 billion in EIC overclaims has a 95 percent
confidence interval of $4.0 billion to $4.9 billion.  Our evaluation
of the study methodology showed that the estimate is reasonably
accurate and representative of EIC claimants filing between January
15 and April 21, 1995.  Some aspects of the study methodology
affected the precision of the results; but, given the scale of the
findings, these limitations do not affect the study's message or its
usefulness in designing compliance approaches. 

Although it is a reasonable estimate of EIC overclaims, the entire
$4.4 billion should not be viewed as potential savings to the
government had IRS somehow been able to prevent or correct all of
these EIC errors.  For returns filed with an EIC claim, the tax year
1994 study was designed to evaluate taxpayers' compliance with each
EIC eligibility filing requirement, to produce an overall estimate of
EIC amounts claimed in error, and to identify the sources of these
errors.  The study was not designed to detect or quantify EIC claims
that taxpayers could have made but did not.  For example, the $4.4
billion overclaim estimate includes about $780 million in overclaims
associated with errors in applying the "adjusted gross income (AGI)
tiebreaker" rule.  That rule provides that if a child meets the
conditions to be a qualifying child of more than one person, only the
person who had the highest AGI may treat that child as a qualifying
child.  As the 1994 study was designed, if IRS determined under the
AGI tiebreaker rules that a person claiming the EIC was not entitled
to it because there was another person in the household with a higher
income, IRS would disallow the claim and include it as an overclaim
in computing the study results.  However, because these overclaims
are not offset by any claim that could have been made by the other
person involved in the tiebreaker, the ultimate savings to the
government could be less than $780 million. 

The largest source of taxpayer error identified by the tax year 1994
study relates to EIC requirements that are difficult for IRS to
verify--those related to eligibility of qualifying children. 
Taxpayer returns with qualifying child errors accounted for about
two-thirds of the $4.4 billion in overclaims.  Failure to meet the
child residency test (living with the taxpayer for more than 6
months, or 1 year if a foster child) was the most common of the
qualifying child errors, followed by errors in applying the AGI
tiebreaker rules. 

Unlike income transfer programs such as Temporary Assistance for
Needy Families and Food Stamps, the EIC was designed to be
administered through the tax system.\2

This choice generally should result in lower administrative costs and
higher participation rates and emphasizes that the credit is for
working taxpayers.  The trade-off, however, is higher noncompliance. 
EIC eligibility, particularly related to qualifying children, is
difficult for IRS to verify through its traditional enforcement
procedures, such as matching return data to third-party information
reports.  Correctly applying the residency test and AGI tiebreaker
rules, for example, often involves understanding complex living
arrangements and child custody issues.  Organizations that administer
programs like Food Stamps are set up to investigate and verify this
type of eligibility before payment is made; IRS is not.  Thoroughly
verifying qualifying child eligibility basically requires IRS to do
an audit of the type done in the EIC compliance studies--a costly,
time-consuming, and intrusive proposition.  IRS has designed some
compliance efforts to reduce qualifying child noncompliance but
cannot fully address a significant root cause--design of the EIC
itself. 

With new enforcement tools provided by Congress and an increase in
funding specifically designated for EIC-related activities, IRS began
implementing in fiscal year 1998 a plan that, over a period of 5
years, calls for attacking EIC noncompliance through expanded
customer service and public outreach, strengthened enforcement, and
enhanced research.  Together, these activities make up the "EIC
compliance initiative." Many parts of that initiative are targeted at
the primary sources of EIC noncompliance identified in the tax year
1994 compliance study.  Most of the efforts that make up the EIC
compliance initiative had not progressed far enough at the time we
completed our audit for us to make any judgment about their
effectiveness.  However, in reviewing IRS' efforts for tax year 1997,
we identified several implementation issues that could diminish the
initiative's impact.  For example, EIC claimants often file quite
early in the filing season because they receive sizable refunds;
however, IRS offered several of its EIC-related customer service
programs during the 1998 filing season after many EIC claimants had
already filed their returns. 

IRS plans to measure the overall impact of the compliance initiative
on the overclaim rate\3 through annual studies of EIC compliance
starting with a baseline study of tax year 1997 returns.  However,
the 5-year initiative could be into its fourth year before IRS has
tax year 1997 and 1998 study data to compare in assessing the
initiative's results.  That would be too late for IRS to identify and
implement meaningful adjustments to the initiative.  IRS also plans
to measure the results of individual initiative components
implemented in 1998, but some of these results will not be available
for initial planning of fiscal year 1999 activities. 


--------------------
\2 As a refundable tax credit, EIC amounts in excess of tax liability
paid through the income tax system are transferred to taxpayers
through an income tax refund. 

\3 The overclaim rate refers to the amount of EIC overclaimed divided
by the amount of EIC claimed on taxpayer returns, corrected for math
errors. 


   BACKGROUND
------------------------------------------------------------ Letter :2

The EIC is a refundable tax credit available to low-income, working
taxpayers.  Congress created the credit in 1975 to offset the impact
of Social Security taxes on low-income families and encourage
low-income workers to seek employment rather than welfare. 


      EIC ELIGIBILITY
---------------------------------------------------------- Letter :2.1

The amount of a taxpayer's credit depends on the number of qualifying
children who meet age, relationship, and residency tests and on the
nature and amount of qualifying income. 

Taxpayers with children can claim the EIC if they (1) have at least
one EIC qualifying child,\4 (2) meet income tests, (3) file with any
filing status except "married filing separately," and (4) were not a
nonresident alien for any part of the year.  To claim the EIC without
a qualifying child, taxpayers have to meet requirements 2, 3 and 4,
be at least 25 but less than 65 at the end of the year, have lived in
the United States for more than half the year, and must not be
claimed as a dependent on another return. 

The credit amount gradually increases with increasing income,
plateaus at a maximum amount, and then gradually decreases (in a
"phase-out range") until it reaches zero when the taxpayer's earned
income or AGI exceeds the allowable maximum.  Taxpayers with AGI
falling in the credit's phase-out range are to receive the lesser
amount resulting from using their earned income or AGI in calculating
the credit. 

Recently, Congress made changes to EIC eligibility rules in the
Personal Responsibility and Work Opportunity Reconciliation Act of
1996 (P.L.  104-193) and the Taxpayer Relief Act of 1997 (P.L. 
105-34).  These changes, affecting returns filed for tax year 1996
and after,

  -- denied the EIC to any taxpayer with investment income over a
     certain threshold ($2,250 for tax year 1997);\5

  -- defined a "modified AGI" to be used in calculating the credit
     that excludes certain losses from investments and businesses;\6

  -- denied the credit to taxpayers without valid Social Security
     numbers (SSN);\7 and

  -- excluded certain workfare payments\8 from wages for EIC
     purposes. 

Table 1 compares the maximum EIC amounts and income limits for tax
years 1994 and 1997. 



                                Table 1
                
                 Maximum EIC Amounts and Income Limits
                      for Tax Years 1994 and 1997

                            Tax year 1994           Tax year 1997
                        ----------------------  ----------------------
                                       Maximum                 Maximum
Number of EIC              Maximum   allowable     Maximum   allowable
qualifying children     EIC amount      income  EIC amount      income
----------------------  ----------  ----------  ----------  ----------
None                          $306      $8,999        $332      $9,769
One                          2,038      23,754       2,210      25,759
Two                          2,528      25,295       3,656      29,289
----------------------------------------------------------------------
Source:  IRS EIC tables. 


--------------------
\4 Complex living arrangements, such as when two unrelated families
share a home or when divorced parents have joint custody of a child,
often make it difficult to determine who can claim a child for the
EIC. 

\5 The investment income exclusion is basically an indirect wealth
test intended to eliminate certain taxpayers from the EIC program. 
Investment income includes taxable and nontaxable interest, taxable
dividends, net rent and royalty income derived from sources outside
the taxpayer's ordinary course of trade or business, capital gain net
income, and passive activity net income.  This provision was
effective beginning in tax year 1996 with a threshold of $2,200.  The
threshold is to be indexed for inflation and was increased
accordingly to $2,250 for tax year 1997. 

\6 Effective beginning in tax year 1996, modified AGI for the
purposes of the EIC meant AGI determined without regard to (1) net
capital losses, (2) net losses from trusts and estates, (3) net
losses from rents and royalties derived outside the taxpayer's normal
course of trade or business, and (4) 50 percent of net losses from
trades or businesses.  The losses subject to the 50-percent exclusion
are to be computed separately with respect to sole proprietorships
other than farming, farming sole proprietorships, and other trades or
businesses.  Effective for tax year 1998, taxpayers are required to
add to modified AGI (1) tax-exempt interest; and (2) nontaxable
distributions from pensions, annuities, and individual retirement
arrangements if not rolled over into similar vehicles during the
applicable rollover period.  Also beginning with tax year 1998, the
disregarded amount of net losses from trades or businesses increased
from 50 percent to 75 percent with the same computation rules. 

\7 Taxpayers were already required to provide valid SSNs for
qualifying children. 

\8 Generally, under the Personal Responsibility and Work Opportunity
Reconciliation Act of 1996, the receipt of certain government
assistance payments is denied unless the recipient meets certain work
requirements.  Wages earned through state-subsidized work experience
and community service programs are referred to as "workfare payments"
and are excluded from earned income for the purposes of the EIC. 


      HOW IRS DETECTS AND PURSUES
      NONCOMPLIANT RETURNS
---------------------------------------------------------- Letter :2.2

IRS checks individual returns, with and without the EIC, for
compliance while the return is initially being processed and in the
months after filing.  Some noncompliance involves mathematical errors
and other obvious mistakes made by taxpayers or their representatives
in preparing the returns.  Other noncompliance involves mistakes that
can be detected only through an audit of the return. 

The easiest EIC mistakes to identify and correct are those that IRS
classifies as math errors.  These errors, identified as the return is
processed, include EIC computation errors and certain qualifying
errors (e.g., missing SSNs for taxpayers and their children).\9 For
returns filed on paper, staff in IRS' service centers are to enter
tax return and Schedule EIC data into computers that check for math
errors.  If a math error that affects EIC eligibility or the size of
the EIC claim is found, IRS is to reduce or deny the EIC accordingly. 
IRS is to then send a notice to the taxpayer explaining the change to
his or her tax liability and refund.  Taxpayers have 60 days to
protest IRS' actions, either in writing or by telephone, and to
provide additional data supporting their original claims.  If
taxpayers do not respond to IRS' notice, they are to get no further
correspondence from IRS about that matter unless they fail to pay any
additional tax that was assessed as a result of IRS' change. 

Returns that taxpayers attempt to submit electronically are subject
to a series of computerized "filters" that screen the submission for
accuracy and completeness.  Submissions with computational mistakes
or missing or invalid data are to be rejected.  A taxpayer whose
electronic submission has been rejected can either correct the
mistake(s) and resubmit the electronic return or file the return on
paper (with or without the corrections).  If filed on paper, the
return would be subjected to the math error procedures described in
the preceding paragraph. 

The most serious form of noncompliance involves deliberate attempts
to defraud the government through, for example, phony refund claims. 
IRS' primary effort to identify fraudulent refund claims, including
those involving the EIC, is the Questionable Refund Program (QRP),
established in the 1970s and run by IRS' Criminal Investigation
Division.  Using a scoring system based on known noncompliance
patterns, an IRS computer program analyzes all incoming returns to
identify those that are potentially fraudulent.  Then, questionable
refund detection teams in the 10 service centers are to perform more
in-depth reviews and, if a return is considered fraudulent, stop any
refund before it is issued. 

IRS' examination units in service centers and district offices review
other potentially erroneous EIC claims that do not meet the criteria
for inclusion in the math error or questionable refund programs. 
Service center staff review cases that do not require face-to-face
contact with the taxpayer.  Cases requiring face-to-face contact are
done by district offices.  Questionable refund detection teams are to
refer cases with EIC errors that are not considered fraudulent to the
examination units.  Examination staff may also review cases included
in special enforcement or compliance research projects.  When
examination staff determine that an EIC claim is erroneous, they are
to notify the taxpayer of that finding and advise the taxpayer of his
or her appeal rights.  If the taxpayer agrees with IRS' finding or
disagrees with the finding but fails to overturn it on appeal, the
claimed EIC is to be disallowed or adjusted in accordance with the
examiner's findings. 


--------------------
\9 Until the 1997 filing season, SSN errors were not considered math
errors and had to be corrected using the examination procedures
discussed at the end of this section.  In 1997, as a result of a
provision in the Personal Responsibility and Work Opportunity
Reconciliation Act of 1996, IRS began treating missing or incorrect
SSNs as math errors, similar to the way it had historically handled
computational mistakes. 


      IRS' EIC COMPLIANCE STUDIES
---------------------------------------------------------- Letter :2.3

IRS has undertaken a series of EIC compliance studies in recent
years.  In the first study, IRS sampled returns with EIC claims that
had been filed electronically during a 2-week period in January 1994. 
The results, which could be generalized only to electronic returns
filed during that 2-week period, showed that 39 percent of the
returns involved overstated EIC claims that represented 26 percent of
the dollars claimed.  To learn more about EIC compliance, IRS
conducted a broader study of tax year 1994 returns filed both
electronically and on paper.  The results of that study, released in
April 1997, are the subject of this report.  In 1996, IRS began a
third study involving tax year 1995 returns.  As of June 1998, IRS
had not completed its analysis of the data from that study.  All
three of these EIC compliance studies predated the SSN-related math
error procedures that were first implemented in 1997.  However, as
noted later, IRS adjusted the findings of its tax year 1994 study to
show what the noncompliance rate would have been if those procedures
had been in place then. 

As part of a 5 year EIC compliance initiative begun in fiscal year
1998 and discussed later in this report, IRS plans to measure its
progress in reducing the EIC overclaim rate through annual studies of
returns filed with an EIC claim.  According to IRS, the first study
of about 2,500 tax year 1997 EIC returns filed from January through
May 1998 is designed to provide a baseline measure of the validity of
EIC claims and types of EIC errors.  IRS' time line for the study
shows that it expects to have a final report prepared by December 31,
1999.\10 The results of subsequent studies are to be compared with
that baseline to identify changes in EIC compliance. 


--------------------
\10 In commenting on a draft of this report, IRS officials told us
they plan to issue an interim report earlier in 1999 based on
preliminary data they expect to be available by the end of 1998. 


   OBJECTIVES, SCOPE, AND
   METHODOLOGY
------------------------------------------------------------ Letter :3

Our objectives were to (1) evaluate IRS' tax year 1994 EIC compliance
study methodology to determine if the reported results were
reasonably accurate, (2) identify the primary sources of EIC
noncompliance found in that study, and (3) determine whether recent
IRS compliance efforts are designed to address the primary sources of
noncompliance. 

To evaluate IRS' study methodology and the accuracy of IRS'
compliance study results, we reviewed written documentation on the
study's methodology, reviewed 122 case files,\11 interviewed IRS and
Treasury officials involved in the study, reviewed computer programs
written by IRS and Treasury's Office of Tax Analysis (OTA) that were
used to create and edit the final dataset, and calculated confidence
intervals for the data presented in IRS' April 1997 report.  To
assess IRS' methodology, we determined whether IRS used generally
accepted social science standards, which include the use of (1)
unbiased sample selection procedures, (2) data collection controls,
(3) procedures to ensure quality of data used, and (4) appropriate
statistical procedures to generalize the data gathered and analyzed. 
In doing so, we considered the following questions: 

  -- Does the study population appear to represent the population of
     all EIC filers during the period from January 15 through April
     21, 1995? 

  -- Was the sample drawn in accordance with probability selection
     principles? 

  -- Were sufficient data verifying compliance with all EIC
     eligibility requirements collected from the EIC claimant and
     other sources? 

  -- Were IRS staff collecting the data knowledgeable of how to apply
     EIC eligibility rules? 

  -- Did the data collection procedures include controls to help
     ensure consistency in the evaluation of cases? 

  -- Was data entry into the final database verified? 

  -- Was the database checked for internal consistency, outliers, and
     invalid codes? 

  -- How precise were the reported overclaim estimates? 

We also reviewed available data on IRS' design of the tax year 1997
EIC compliance study to see how, if at all, that study addressed
problems we identified with the tax year 1994 study. 

To determine the primary sources of EIC noncompliance on tax year
1994 returns, we analyzed the tax year 1994 study dataset as provided
by IRS and modified through OTA editing programs.  All data are
estimates based on the study sample.  Accordingly, we calculated
confidence intervals at the 95 percent confidence level to indicate
the precision of the estimates.  Unless otherwise noted, the
confidence intervals for percentages are
 5 percentage points or less; for other statistics, the intervals
are
 10 percent or less of the reported value. 

To determine whether recent IRS compliance efforts addressed the
primary sources of noncompliance, we

  -- reviewed IRS documents to identify the scope of EIC-related
     activities and related implementation plans;

  -- interviewed officials responsible for designing and implementing
     EIC-related activities at IRS' National Office, its Brookhaven,
     Cincinnati, and Fresno Service Centers, and its Northern
     California District Office; and

  -- obtained available data on the results of EIC programs. 

We did our work from September 1997 through May 1998 in accordance
with generally accepted government auditing standards.  We requested
comments on a draft of this report from the Commissioner of Internal
Revenue and the Secretary of the Treasury, or their designees.  The
Commissioner of Internal Revenue and Treasury Deputy Assistant
Secretary (Tax Analysis) responded in letters dated July 2, 1998, and
June 29, 1998, respectively.  Their comments are summarized at the
end of this letter and are reprinted in appendixes I and II.  On July
1, 1998, we met with IRS officials, including the Deputy Chief of
Operations, the Acting Assistant Commissioner for Customer Service,
and the Assistant Commissioner for Research/Statistics of Income, to
discuss the Commissioner's comments.  In addition, IRS and OTA
provided technical comments on our draft.  We made changes to the
report in response to the comments where appropriate. 


--------------------
\11 Our objectives in reviewing study case files were to improve our
general understanding of how the study cases were audited, to
research case-specific data questions raised during our initial
analyses, and to verify corrections and additions to the data made by
Department of the Treasury staff.  Given these objectives, we
judgmentally selected 122 cases to review.  Because Treasury did
extensive internal consistency testing on the data, we did not verify
data entry on a statistical sample of returns; rather, we verified
data entry for key variables on these 122 returns. 


   IRS' TAX YEAR 1994 EIC STUDY
   METHODOLOGY FOLLOWED GENERALLY
   ACCEPTED SOCIAL SCIENCE
   STANDARDS
------------------------------------------------------------ Letter :4

IRS found that of $17.2 billion in EIC claimed during the January 15
to April 21, 1995, study period, taxpayers overclaimed $4.4 billion,
or 25.8 percent of total EIC claimed.\12

To determine whether this $4.4 billion overclaim estimate is
reasonably accurate, we evaluated IRS' study methodology.  Our
evaluation was based on the extent to which IRS used generally
accepted social science standards for a research project of this
kind.  These standards include use of (1) unbiased sample selection
procedures, (2) data collection controls, (3) procedures to ensure
quality of data used, and (4) appropriate statistical procedures to
generalize the data gathered and analyzed. 

Before discussing our analyses, however, it is important to put IRS'
study findings in perspective.  For returns filed with an EIC claim,
the tax year 1994 study was designed to evaluate taxpayers'
compliance with each EIC eligibility filing requirement, to produce
an overall estimate of EIC amounts claimed in error, and to identify
the sources of error.  The study was not designed to detect or
quantify EIC claims that taxpayers could have made, but did not.  For
example, the $4.4 billion overclaim estimate includes about $780
million in overclaims associated with errors in applying the AGI
tiebreaker rule.  That rule provides that if a child meets the
conditions to be a qualifying child of more than one person, only the
person who had the highest AGI may treat that child as a qualifying
child.  As the 1994 study was designed, if IRS determined under the
AGI tiebreaker rules that a person claiming the EIC was not entitled
to it because there was another person in the household with a higher
income, IRS would disallow the claim and include it as an overclaim
in computing the study results.  However, because these overclaims
are not offset by any claim that could have been made by the other
person involved in the tiebreaker, the ultimate savings to the
government could be less than $780 million. 

With this basic limitation in mind, we found that overall, IRS' study
was designed and conducted in such a way that it produced a
reasonably accurate estimate of noncompliance on returns filed with
an EIC claim.  Although some issues with the study design affected
the precision of the results, our analysis showed that these
limitations did not affect the study's major message or its
usefulness in designing compliance approaches. 


--------------------
\12 The study does not reflect IRS procedures initiated in the 1997
filing season that allowed IRS, under math error authority, to reduce
or deny EIC claims on returns filed with missing or invalid SSNs for
qualifying children.  According to IRS and OTA analysis, had this SSN
math error authority been available when taxpayers filed their 1994
returns, overclaims would have been reduced by about $814 million and
the overclaim rate would have been about 21 percent.  According to
OTA, this estimate does not include overclaims that could have been
detected if the taxpayer did not provide a valid SSN for a child
under the age of 1; such errors could not be detected using the tax
year 1994 study data because taxpayers were not required to provide
SSNs for infants.  Also according to OTA, the estimate does not
include overclaims associated with invalid SSNs for primary or
secondary taxpayers, because the data provided by IRS regarding
invalid SSNs did not include this information. 


      IRS' STUDY METHODOLOGY
      SUPPORTED THE REPORTED
      FINDINGS
---------------------------------------------------------- Letter :4.1

IRS' study is representative of taxpayers filing an EIC claim on a
tax year 1994 return filed between January 15 and April 21, 1995.\13
We found that IRS used an appropriate statistical sampling procedure
to select the 2,046 returns included in the study,\14 and the sample
appears to represent taxpayers who filed an EIC claim during that
period.  As an indicator of whether the EIC study sample was
representative of EIC returns filed during the study time frame, we
assumed that study returns should have characteristics similar to EIC
returns filed during the entire year, as measured in IRS' SOI sample
of tax year 1994 returns filed throughout 1995.  IRS' April 1997
compliance study report compared weighted data on the distribution of
claimants by paper or electronic filing, filing status, number of
qualifying children, AGI range, and source of income for the study
sample and the full tax year 1994 SOI sample.  With the exception of
the percentage of EIC claimants reporting self-employment income (6.3
percent among taxpayers in the compliance study compared to 15.3
percent of taxpayers in the SOI sample), weighted compliance study
data closely paralleled data based on the SOI sample.  The apparent
underrepresentation of claimants reporting self-employment income is
discussed in more detail later. 

In evaluating how IRS collected data on the accuracy of taxpayers'
EIC claims, we considered the following criteria:  (1) whether IRS
collected sufficient data to verify each aspect of EIC eligibility,
(2) whether field agents and case reviewers knew how EIC rules were
to be applied, and (3) whether the study procedures included controls
designed to ensure consistency among cases.  After reviewing study
documentation and selected case files, we concluded that overall,
IRS' study methodology met these criteria. 

IRS' data collection proceeded in two stages:  initial taxpayer
interviews and supplemental data collection in the field and a final
review of complete case files at the Cincinnati Service Center.  For
each taxpayer in the study, IRS built a case file including
transcripts of prior years' returns, the tax year 1994 return,
associated information reports (W-2s and 1099s), information on
duplicate use of or invalid qualifying child SSNs, and dates of birth
for filers and their children.  Field agents were given initial case
files containing data available at the time, written instructions on
the information required to verify the claim, and checksheets to
record findings.  These instructions and checksheets covered all
aspects of EIC eligibility.  Field agents were required to contact,
in person, the taxpayer; employers; the transmitter of the
electronically filed return, if any; and the paid preparer, if any. 
If additional information was needed to verify the claim, the agents
were to contact neighbors, schools, or state agencies as
appropriate.\15 This type of face-to-face contact with the taxpayer
was necessary to verify the claim because eligibility of qualifying
children is self-determined; and, other than SSNs, IRS does not have
third-party information that can be used to verify the children's
eligibility.  Data collection by the field agents was followed by a
"best and final" review at the Cincinnati Service Center.  These
reviewers had access to additional information, primarily third-party
income reports, that was not available at the time the field agents
contacted the taxpayers.  Using this additional information, the
reviewers made final decisions regarding disposition of the claim. 

Most of the taxpayer interviews and other field data collection were
done by IRS special agents from the Criminal Investigation Division
or revenue agents from the Examination Division.  The service center
reviewers were Criminal Investigation Division tax examiners.  On the
basis of our prior work, we consider these staff generally to be
adequately trained in audit techniques and how to apply EIC rules. 

IRS' study methodology included controls designed to ensure
consistency among cases.  IRS used standard data collection
checksheets and written instructions as one means to ensure
consistent data collection.\16 In addition, after completing an
investigation, the field agents were instructed to call a study
coordinator at IRS' Cincinnati Service Center to discuss the case. 
This study coordinator was to review the findings to ensure
completeness and consistency with other cases before the case was
sent to the Cincinnati Service Center for final review.  In spite of
these controls, we found one consistency-related issue, regarding the
corrected filing status for married taxpayers who erroneously filed
as head of household, that may have systematically affected the study
findings.  This issue is discussed in more detail later. 

In reviewing IRS' procedures for ensuring the quality of final study
data used, we considered whether IRS (1) verified data entry into the
final database; and (2) checked for internal consistency, outliers,
and invalid codes.  We found that OTA staff did a comprehensive
review of the database to find and correct internal consistency and
data transcription errors.  This review of the data was necessary
because IRS did not verify data entry or check for internal
consistency within case records as the database was created.\17 We
verified data entry of key variables for the 122 cases included in
our case file review and found no data errors.  We used the
OTA-corrected data for our analysis. 

IRS and OTA used the data to estimate the total amount of EIC
overclaimed by the population represented by the sample of taxpayers
filing EIC claims from January 15 to April 21, 1995.  We replicated
this analysis and arrived at the same totals.  Data from the study
are estimates based on the sample of EIC returns.  To indicate the
precision of these estimates, we calculated confidence intervals at
the 95 percent confidence level.  For example, as shown in table 2,
IRS determined that taxpayers overclaimed a total of $4,448 million
in EIC.  The 95 percent confidence interval for this estimate is 
$412 million ( 9.3 percent of $4,448 million).  This indicates that
we are 95-percent confident that the actual overclaim amount is
between $4,036 million and $4,860 million.  The confidence interval
for the 25.8 percent overclaim rate ranges from 23.4 percent to 28.2
percent. 



                                     Table 2
                     
                     EIC Overclaims on Tax Year 1994 Returns
                     by Number of Qualifying Children Claimed

                     (Number of returns and dollar amounts in
                                    millions)

                        EIC amount as claimed on
                            taxpayer returns           EIC amount overclaimed
                      ----------------------------  ----------------------------
Number of qualifying      Number of                     Number of
children claimed            returns         Amount        returns         Amount
--------------------  -------------  -------------  -------------  -------------
One or two
 Point estimate                12.0        $16,722            4.2         $4,368
 95% confidence          11.5 -12.5      $15,908 -       3.9 -4.5       $3,959 -
 interval                                  $17,536                        $4,777
 Confidence interval          4.2%                         7.1%
 as a                                        4.9%                         9.4%
 percentage of point
 estimate\a
None
 Point estimate                 3.0           $513            0.5            $81
 95% confidence            2.3 -3.7     $376 -$650       0.2 -0.8      $24 -$138
 interval
 Confidence interval         23.3%         26.7%         60.0%         70.4%
 as a
 percentage of point
 estimate\a
Total
 Point estimate                15.0        $17,235            4.7       $4,448\b
 95% confidence          14.2 -15.8      $16,429 -       4.3 -5.1       $4,036 -
 interval                                  $18,041                        $4,860
 Confidence interval          5.3%                         8.5%
 as a                                        4.7%                         9.3%
 percentage of point
 estimate\a
--------------------------------------------------------------------------------
\a To calculate the confidence interval as a  percentage of the
point estimate, we divided the dollar range of the 95 percent
interval by 2 to express it as a  value, then divided that figure by
the point estimate.  For example, the 95 percent confidence interval
for EIC amounts claimed by taxpayers with one or two qualifying
children is $15,908 to $17,536 million--a range of $1,628 million or
 $814 million from the point estimate.  The $814 million is 4.9
percent of the point estimate. 

\b Total does not add due to rounding. 

Source:  GAO analysis of IRS' tax year 1994 EIC compliance study
data. 

As shown in table 2, taxpayers with no qualifying children accounted
for only a small portion of the EIC dollars claimed for tax year 1994
and $81 million of the overclaim total.  The 95 percent confidence
interval is $24 million to $138 million. 

Besides the overclaim estimate in table 2, IRS included in its report
an estimate of total underclaims by taxpayers filing a tax year 1994
return with an EIC claim.\18 That estimate was $293 million\19 and
has a confidence interval of $129 million to $457 million, or  56.0
percent of the point estimate.  The sample of taxpayers with
underclaimed EIC is too small to allow IRS to make reliable estimates
by number of qualifying children. 


--------------------
\13 Based on a comparison to IRS data on EIC returns filed for the
full year, the sample represents about 80 percent of EIC returns for
tax year 1994.  The weighted estimate from the tax year 1994 EIC
study showed 15 million EIC returns filed between January 15 and
April 21, 1995.  The weighted estimate from IRS' Statistics of Income
(SOI) sample of all returns filed in 1995 showed 19 million EIC
returns.  This indicates that about 20 percent of EIC filers for tax
year 1994 filed outside the study period.  The study's estimate of
$4.4 billion in overclaims does not include overclaims among those 20
percent.  If the excluded 20 percent are equally as likely to be
noncompliant as the taxpayers represented by the sample, then the
overclaim estimate for the entire tax year will be about 20 percent
more than reported.  If the excluded filers are more or less
compliant, then the impact is unknown.  IRS' study of tax year 1997
returns extended sampling through May 1998 and should represent an
even larger proportion of the EIC filing population for that year. 

\14 IRS used a two-stage poststratified systematic sample design.  In
the first stage, IRS selected .04 percent of electronic and .05
percent of paper EIC returns filed at each service center.  The
returns were selected after the math errors had been corrected.  In
the second stage, the selected returns were classified into one of
four subgroups, or strata.  The strata and sampling rates were (1)
100 percent of male head of household or single filers claiming
qualifying children, (2) 25 percent of female head of household or
single filers claiming qualifying children, (3) 10 percent of married
taxpayers filing jointly or as qualified widow(er)s and claiming
qualifying children, and (4) 4 percent of filers claiming the EIC
with no qualifying children.  The final sample consisted of 2,046
returns--1,250 paper and 796 electronic. 

\15 In reviewing case files, we found six instances in which the file
did not contain all data relevant to the taxpayer's EIC claim.  In
one case, for example, a taxpayer filed as head of household and
claimed a qualifying child.  The Cincinnati Service Center reviewer
noted in the case file that the agent did not get an SSN for the
qualifying child's father, did not find out if anyone else lived in
the household, and did not verify that the qualifying child actually
resided there.  Within our sample, these appeared to be isolated
instances, however, and not systematic errors. 

\16 Several parallel questions on the field agent and final review
checksheets were not always worded the same way, or with the same
choices, which led to some inconsistent coding.  For example, the
field agent checksheet item regarding qualifying child errors did not
include a choice for "child does not exist," and initial coding of
these situations was inconsistent.  In editing the data, IRS and OTA
searched for and corrected these types of inconsistencies. 

\17 IRS' tax year 1997 EIC compliance study documents show that IRS
plans to use OTA's data verification programs at the time data are
entered into the database. 

\18 This underclaim estimate applies only to returns filed with an
EIC claim.  It does not include EIC claims that taxpayers could have
made but did not. 

\19 The $4.4 billion overclaim estimate does not net out these
underclaims. 


      MINOR METHODOLOGY PROBLEMS
      AFFECTED THE PRECISION OF
      IRS' STUDY RESULTS
---------------------------------------------------------- Letter :4.2

Through our review of IRS' study methodology, we identified two
issues that affected the final study results.  These issues are (1)
apparent underrepresentation in the sample of claimants reporting
self-employment income on a Schedule C and (2) apparent
inconsistencies in correcting the filing status of married taxpayers
who erroneously filed as head of household.  Although we were not
able to precisely quantify their net impact, our analysis showed that
neither of these issues was large enough in scale to alter the major
study findings. 

Filers who report self-employment income on a Schedule C appear to be
underrepresented in the tax year 1994 EIC study sample.  SOI data for
all taxpayers who filed in 1995 show that 15.3 percent of EIC
claimants reported Schedule C income.  In contrast, 6.3 percent of
claimants in the EIC compliance study filed a Schedule C. 
Self-employment income is often not subject to third-party
information reporting; consequently, IRS has found that Schedule C
filers in general are more likely to misreport their income than are
taxpayers with wage income.  A change in income, however, will often
result in an incremental change in the EIC rather than a full denial. 
The impact of underrepresenting Schedule C filers in the tax year
1994 study is unknown and depends on how the filers left out of the
sample might differ from those included.  IRS' ongoing compliance
study of tax year 1997 EIC returns includes specific sampling of
Schedule C returns and will sample through May; that study should be
more representative of Schedule C EIC filers. 

Inconsistencies in determining the correct filing status for married
taxpayers who erroneously filed as head of household also may have
affected the final overclaim estimate.  Taxpayers who use a filing
status of married filing separately are ineligible for the EIC. 
Married taxpayers filing a joint return can claim the EIC if their
joint income is within the eligible income range and they meet other
qualifying criteria.  When a taxpayer erroneously claimed head of
household and was living with his or her spouse, the data collection
instructions for the tax year 1994 study specifically directed field
agents to use the filing status most advantageous to the taxpayer,
usually married filing jointly (with appropriate changes to income,
dependents, exemptions, etc.). 

We, and OTA staff who also reviewed case files, found instances in
which field agents did not follow these instructions.  The $4.4
billion overclaim estimate included $631 million\20

accounted for by taxpayers whose filing status was changed by IRS to
married filing separately in the absence of qualifying child errors
and whose EIC was denied completely.  It appears that some of these
taxpayers may have been eligible for the EIC had IRS prepared a joint
return; and, to the extent that their joint income would have allowed
an EIC claim, the $631 million may be overstated.\21 Although this
filing status issue reduces the precision of the study findings,
particularly in terms of identifying sources of noncompliance, we
believe its impact to be relatively minor given the size of the total
overclaim estimate.  IRS data collection instructions for the tax
year 1997 study state that field agents should attempt to obtain a
copy of the spouse's 1997 return to insert into the case file when
filing status is changed to married filing jointly or married filing
separately.  However, the instructions did not specifically state
that married filing jointly should be the presumptive filing status. 


--------------------
\20 Confidence interval is  29 percent of the point estimate. 

\21 We do know that of the 423,000 returns (confidence interval 
20.6 percent of the estimate) that IRS changed from head of household
to married filing jointly, 58 percent (confidence interval
 11 percentage points) still qualified for some amount of EIC. 


   LARGEST SOURCE OF EIC
   OVERCLAIMS FOR TAX YEAR 1994
   WAS NONQUALIFYING CHILDREN
------------------------------------------------------------ Letter :5

The largest source of noncompliance found in the tax year 1994 study
relates to the EIC requirements most difficult for IRS to
verify--those related to the eligibility of qualifying children.  As
shown in figure 1, taxpayer returns with qualifying child errors
accounted for at least 65 percent of the $4.4 billion in
overclaims--56 percent from returns with qualifying child errors only
and an additional 9 percent from returns with qualifying child errors
made in conjunction with a filing status change to married filing
separately.\22 Claiming a child who did not meet residency
requirements was the most common qualifying child error, and errors
claiming head of household status often occurred with claims for
nonqualifying children.  Misreported income accounted for another 16
percent of the overclaim total; taxpayers whose filing status was
changed to married filing separately, in the absence of qualifying
child errors, accounted for most of the remainder. 

   Figure 1:  Sources of EIC
   Overclaim Errors, Tax Year 1994

   (See figure in printed
   edition.)

\a The "other error" category includes some returns with both
qualifying child and income errors. 

Source:  GAO analysis of IRS' tax year 1994 EIC compliance study
data. 


--------------------
\22 The "other error" category in figure 1 includes some returns with
both qualifying child and income errors. 


      FAILURE TO MEET RESIDENCY
      REQUIREMENTS WAS THE
      PREDOMINANT QUALIFYING CHILD
      ERROR
---------------------------------------------------------- Letter :5.1

In order for a taxpayer to claim a qualifying child,\23 the following
rules applied for tax year 1994. 

1.  Relationship:  The child must have been the taxpayer's son,
daughter, adopted child, grandchild, stepchild, or eligible foster
child.  A foster child is defined as any child cared for as the
taxpayer's own.\24
2.  Age:  The child must have been under age 19, or under age 24 and
a full-time student, or any age and permanently and totally disabled.
3.  Residence:  The child must have lived in the United States with
the taxpayer for more than half of the year or the entire year for
foster children.\25
4.  AGI tiebreaker:  If a child meets the conditions to be a
qualifying child of more than one person, only the person who had the
highest AGI may treat that child as a qualifying child.  This rule
does not apply if the other person is the taxpayer's spouse and they
are filing a joint return.  For example, if a child meets conditions
to be a qualifying child for both a parent and grandparent who share
a household and the grandparent has a higher AGI, the grandparent
must claim the child.  If the grandparent's AGI exceeds the maximum
income threshold, neither the parent nor the grandparent may claim
the EIC for that child. 

As shown in table 3, qualifying child errors were involved in
overclaims totaling $3.1 billion.  About $1.7 billion\26 of that
amount involved qualifying children who did not meet the residency
test, either alone or in combination with a failure to meet the
relationship test.  Failure to apply the AGI tiebreaker rules added
an additional $782 million in overclaims.  Together, these two types
of qualifying child errors accounted for about half of the $4.4
billion overclaim total.  As noted earlier, however, IRS' study did
not offset overclaims by claims that could have been made by other
taxpayers.  For example, in AGI tiebreaker cases, it is possible that
the taxpayer with the higher AGI might have been able to claim an
EIC.  It is also possible in residency cases that a taxpayer in the
household where the child actually lives could make an EIC claim for
the child in question.  The extent to which AGI tiebreaker and
residency cases involved an EIC claim that could have been made by
another taxpayer, but was not, is unknown. 



                                     Table 3
                     
                      EIC Overclaims Involving a Qualifying
                      Child Error, by Qualifying Child Test
                              Failed, Tax Year 1994

                     (Number of returns in thousands, dollars
                                   in millions)

            Point estimates for overclaims      Confidence interval at the 95%
            with a qualifying child error             confidence level\a
          ----------------------------------  ----------------------------------
Qualifyi
ng child
test
failed             Returns            Amount           Returns            Amount
--------  ----------------  ----------------  ----------------  ----------------
Residenc             1,102            $1,470        937 -1,267    $1,238 -$1,702
 y only
AGI                    578               782          450 -706          585 -979
 tiebrea
 ker
 only
Relation               228               324          155 -301          207 -441
 ship
 only
Residenc               132               181           80 -184           98 -264
 y and
 relatio
 nship
Other\b                257               324          169 -345          204 -444
================================================================================
Total                2,297          $3,080\c      2,067 -2,527    $2,740 -$3,420
--------------------------------------------------------------------------------
Note:  The table includes data for all qualifying child errors,
including those made in conjunction with income or filing status
errors. 

\a Confidence intervals for overclaim return and amount totals,
expressed as percentages, are  10 percent and  11 percent of the
point estimates, respectively.  Others range from  15 percent to 
46 percent. 

\b The "other" category includes qualifying child age errors,
unspecified child errors, qualifying children found not to exist, and
combinations of errors other than relationship and residency. 

\c Column does not add to total due to rounding. 

Source:  GAO analysis of IRS' tax year 1994 EIC compliance study
data. 


--------------------
\23 The definition of a qualifying child for the EIC differs from the
definition of a dependent used to claim a tax exemption.  One major
difference is that to claim a dependent, the taxpayer generally must
have provided over half the child's support during the calendar year. 
In contrast, EIC qualifying child eligibility depends on residency
and does not include a financial support test. 

\24 If the child was married at the end of the year, that child is
generally a qualifying child only if the taxpayer can claim the child
as a dependent. 

\25 A child is considered to have lived with the taxpayer for the
full year if the child was born or died during the year and the child
lived with the taxpayer for the entire time the child was alive
during the year.  Beginning in tax year 1995, the EIC was extended to
taxpayers living outside the United States because of a military
assignment if they meet all other criteria. 

\26 Confidence interval is  12 percent of the point estimate. 


      ERRONEOUS FILINGS AS HEAD OF
      HOUSEHOLD OFTEN OCCURRED
      WITH AN EIC OVERCLAIM
---------------------------------------------------------- Letter :5.2

Filing status per se does not affect either EIC eligibility or credit
amounts, except for married taxpayers filing separate returns who are
ineligible for the EIC.  As shown in table 4, however, head of
household errors occurred on returns accounting for $3.4 billion in
overclaims, or about three-quarters of the $4.4 billion in overclaims
on all returns.  For taxpayers whose filing status was changed to
single, qualifying child errors accounted for most of the overclaims. 
For taxpayers whose filing status was changed to married filing
jointly, most of the overclaims were attributed to income errors. 
Among taxpayers whose EIC was denied because their filing status was
changed to married filing separately, about 40 percent of the
overclaim amounts were also associated with qualifying child errors. 



                                Table 4
                
                  EIC Amounts Overclaimed by Taxpayers
                Erroneously Filing as Head of Household,
                  by Corrected Filing Status, Tax Year
                                  1994

                         (Dollars in millions)

                                                            Confidence
                                   Point estimates     interval at the
Head of household filing         for amount of EIC      95% confidence
status changed to:                     overclaimed             level\a
------------------------------  ------------------  ------------------
Single                                      $1,964      $1,712 -$2,216
Married filing jointly                         390            277 -503
Married filing separately                    1,008          792 -1,224
Total                                       $3,362      $3,029 -$3,695
----------------------------------------------------------------------
\a Confidence intervals expressed as percentages range from  10
percent to  29 percent of point estimates. 

Source:  GAO analysis of IRS' tax year 1994 EIC compliance study
data. 

Among all taxpayers who filed as head of household for tax year 1994,
regardless of final filing status, male taxpayers had an overclaim
rate nearly twice that of female taxpayers.\27

Of $3.2 billion in EIC claims by male head of household filers, $1.7
billion, or about 51 percent,\28 was overclaimed.  In contrast,
female head of household filers overclaimed $2.0 billion,\29 or about
25 percent,\30 of $8.2 billion in EIC they claimed. 


--------------------
\27 As the sample returns were selected, IRS used taxpayers' names to
determine gender.  For those cases in which the taxpayer's gender was
not obvious, IRS referred to SOI files, which include gender data. 
Gender information was confirmed during the taxpayer interview. 

\28 Percentage calculated using the actual estimates rather than the
rounded numbers presented in this sentence. 

\29 Confidence interval is  16 percent of the point estimate. 

\30 Percentage calculated using the actual estimates rather than the
rounded numbers presented in this sentence. 


      MISREPORTED INCOME ACCOUNTED
      FOR A RELATIVELY SMALL
      PORTION OF EIC OVERCLAIMS
---------------------------------------------------------- Letter :5.3

Errors in reporting income, with no other eligibility errors,
accounted for $708 million\31 in EIC overclaims, or 16 percent of
total overclaims.  Included in this group are taxpayers who (1) used
the correct filing status but misreported their income or (2) were
married and erroneously filed as head of household or single and
whose filing status was changed to married filing jointly.  The
filing status error, per se, had no impact on the EIC; however, when
IRS changed the filing status to married filing jointly and modified
the taxpayers' returns to include the correct combined income for
both parties, the EIC was often reduced or denied completely.  These
adjustments accounted for about $309 million\32 of the income-related
overclaims. 


--------------------
\31 Confidence interval is  27 percent of the point estimate. 

\32 Confidence interval is  34 percent of the point estimate. 


      RETURNS DONE BY INFORMAL
      PREPARERS HAD A HIGHER
      OVERCLAIM RATE
---------------------------------------------------------- Letter :5.4

In general, about half of EIC claimants use a return preparer rather
than completing the return themselves.  Using codes developed by
OTA,\33 we grouped prepared returns into the following three
categories:  those prepared by (1) "formal preparers," which includes
attorneys, Certified Public Accountants, national tax preparation
companies, and enrolled agents;\34 (2) "IRS preparers," which
includes staff at IRS walk-in sites and at IRS-supported volunteer
organizations like Volunteer Income Tax Assistance and Tax Counseling
for the Elderly; and (3) "local or informal preparers," which
includes anyone not in the other two categories. 

The study data show that there was little difference in EIC
noncompliance between self-prepared returns and those done by
preparers.  Both groups had overclaim rates of about 26 percent.  A
more detailed analysis, however, shows that overclaim rates varied by
type of preparer.  As shown in figure 2, the rate on returns prepared
by local or informal preparers was 31 percent; the overclaim rate on
returns prepared by formal preparers was 20 percent.  The sample
included too few IRS-prepared returns to allow us to make a reliable
overclaim estimate for that group. 

   Figure 2:  EIC Overclaim Rates
   for Returns Done by Paid
   Preparers, by Type of Preparer,
   Tax Year 1994

   (See figure in printed
   edition.)

Note:  Formal preparers include attorneys, Certified Public
Accountants, enrolled agents, and national tax preparation firms. 

Source:  GAO analysis of IRS' tax year 1994 EIC compliance study
data. 


--------------------
\33 According to OTA, it used two approaches to classify a return as
being done by a preparer.  Under the first approach, a return was
classified as being done by a preparer only if a preparer was
indicated on the filer's return.  Under the second approach, a return
was classified as being done by a preparer if a preparer was
indicated on the filer's return or if IRS investigators found
evidence that a preparer was used.  There was little difference in
the analysis results between the two groups; for our analysis, we
used the second approach. 

\34 An enrolled agent is a preparer who has demonstrated special
competence in tax matters on a written examination administered by
IRS. 


      TAX YEAR 1994 STUDY PROVIDED
      LIMITED DATA ON TAXPAYER
      INTENT
---------------------------------------------------------- Letter :5.5

Knowing the extent to which EIC overclaims are due to honest mistakes
versus intentional misstatements is important in targeting compliance
approaches.  If, for example, errors are due to a misunderstanding of
EIC rules, taxpayer education and assistance efforts would be
warranted.  Taxpayers intentionally misclaiming the EIC require
different approaches.  As part of the tax year 1994 study, IRS made a
determination of taxpayer intent.  Both field agents and Cincinnati
Service Center case reviewers were to classify taxpayers' errors as
intentional (e.g., the taxpayer knew that a child did not meet EIC
qualifying child tests); or unintentional (e.g., the taxpayer did not
understand the eligibility rules or EIC instructions).  We found that
field agents had not made determinations of intent in about 40
percent of the final overclaim cases.  In almost all of these
instances, however, Cincinnati reviewers made a determination of
intent as part of their best and final review.  Based on best and
final case data, about one-half of the 4.7 million returns with an
EIC overclaim and two-thirds of the total amount overclaimed were
considered to be the result of intentional errors.  These assessments
are judgmental in nature and should not be considered precise
measures of intentional and unintentional taxpayer errors.  However,
the results do indicate that IRS' compliance efforts should include
activities aimed at taxpayers who intentionally misclaim the EIC. 

Examiners working tax year 1997 compliance study cases are to collect
data related to taxpayer intent.  The data collection checksheet for
that study includes a question asking examiners to decide if errors
were due to complexity of the tax form, difficulty understanding the
law, a computational error, a potential fraud scheme, or some other
reason.  This provides more specific choices, particularly for
unintentional error, but the determinations of intent will still be
judgmental. 


   RECENT COMPLIANCE EFFORTS ARE
   AIMED AT MAJOR SOURCES OF
   NONCOMPLIANCE, BUT IT IS TOO
   EARLY TO MEASURE THEIR EFFECT
------------------------------------------------------------ Letter :6

With new enforcement tools provided by Congress and an increase in
funding specifically designated for EIC-related activities, IRS began
implementing in fiscal year 1998 a plan that calls for attacking EIC
noncompliance through expanded customer service and public outreach,
strengthened enforcement, and enhanced research.  Together, these
activities constitute what we refer to as the "EIC compliance
initiative." Many parts of that initiative are targeted at the major
sources of EIC noncompliance discussed in the prior section. 
However, in reviewing IRS' efforts for tax year 1997, we identified
several implementation issues that could diminish the initiative's
impact. 

As we have previously testified before Congress,\35 IRS' ability to
reduce EIC noncompliance is limited by the design of the credit. 
Unlike income transfer programs such as Temporary Assistance for
Needy Families and Food Stamps, the EIC is designed to be
administered through the tax system rather than through other state
or federal agencies.  This choice generally should result in lower
administrative costs and higher participation rates and emphasizes
that the credit is for working taxpayers.  The trade-off, however, is
higher noncompliance.  EIC eligibility, particularly related to
qualifying children, is difficult for IRS to verify through
traditional enforcement procedures, such as matching return data to
third-party information reports.  Correctly applying the residency
test and AGI tiebreaker rules, for example, often involves
understanding complex living arrangements and child custody issues. 
Organizations that administer programs like Food Stamps are set up to
investigate and verify this type of eligibility before payment is
made; IRS is not.  Thoroughly verifying qualifying child eligibility
basically requires IRS to do an audit of the type done in the EIC
compliance studies--a costly, time-consuming, and intrusive
proposition.  IRS has designed some compliance efforts to reduce
qualifying child noncompliance but cannot fully address a significant
root cause--design of the EIC itself. 

Most of the efforts that make up the EIC compliance initiative had
not progressed far enough at the time we completed our audit for us
to make any judgment about their effectiveness.  IRS plans to measure
the overall impact of its compliance initiative on the EIC overclaim
rate through annual studies of EIC compliance starting with a
baseline study of tax year 1997 returns.  However, the 5-year
initiative could be into its fourth year before IRS has tax year 1997
and 1998 study data to compare in assessing the initiative's results. 
That would be too late for IRS to identify and implement meaningful
adjustments to the initiative.  IRS plans to measure the results of
individual programs implemented in 1998, but some of these results
will not be available for planning fiscal year 1999 activities. 


--------------------
\35 Earned Income Credit:  Noncompliance and Potential Eligibility
Revisions (GAO/T-GGD-95-179, June 8, 1995); and Tax Administration: 
Earned Income Credit Noncompliance (GAO/T-GGD-97-105, May 8, 1997). 


      RECENT LEGISLATION WAS AIMED
      AT REDUCING EIC
      NONCOMPLIANCE
---------------------------------------------------------- Letter :6.1

Upon release of IRS' April 1997 report on the results of its tax year
1994 EIC compliance study, the Department of the Treasury announced
six legislative proposals directed at reducing EIC noncompliance. 
Congress included four of the six proposals in the Taxpayer Relief
Act of 1997 (TRA97).  Specifically, these provisions (1) require paid
preparers to fulfill certain due diligence standards when preparing
EIC claims for taxpayers; (2) provide that taxpayers who fraudulently
claim the EIC can be denied the credit for 10 years, and those who
recklessly or intentionally disregard the rules and regulations can
be denied the credit for 2 years; (3) provide that taxpayers who are
denied the EIC through IRS' deficiency procedures\36 are ineligible
to claim the EIC in subsequent years unless they provide evidence of
their eligibility through a recertification process; and (4) allow
IRS to levy up to 15 percent of unemployment and means-tested public
assistance and certain other specified payments.  In addition, TRA97
included provisions that (1) give IRS access to the Department of
Health and Human Service's (HHS) Federal Case Registry of Child
Support Orders, a federal database compiling state information on
child support payments that could help IRS identify erroneous EIC
claims by noncustodial parents; and (2) require the Social Security
Administration (SSA) to collect SSNs of birth parents and provide IRS
with information linking the parents' and child's SSNs. 

Besides the new enforcement tools provided in TRA97, Congress began
funding the EIC compliance initiative.  For fiscal year 1998, the
first year of what is to be a 5-year effort, Congress appropriated
$138 million.  For the second year (fiscal year 1999), IRS has
requested $143 million.  Funding over the full 5 years is expected to
total $716 million.  IRS is using the compliance initiative funds to
expand existing EIC-related activities and to initiate several new
efforts, including implementation of the TRA97 provisions. 


--------------------
\36 As defined by IRS, deficiency procedures include administrative
procedures, other than procedures related to math or clerical errors,
that result in an assessment of a deficiency in tax. 


      IRS' EFFORTS ARE TARGETED AT
      MAJOR SOURCES OF EIC ERRORS
---------------------------------------------------------- Letter :6.2

The various activities being funded as part of the EIC compliance
initiative in fiscal year 1998 fall into three broad categories:  (1)
customer service and public outreach, (2) enforcement, and (3)
compliance research.  Primary efforts in each of those categories are
listed in table 5. 



                                     Table 5
                     
                      IRS' EIC Compliance Actions in Fiscal
                                    Year 1998

--------------------------------------------------------------------------------
EIC-related compliance activities for fiscal year 1998

================================================================================
Customer service and public outreach

Expanded telephone access for EIC-related issues to 7 days a week, 24 hours a
day

Provided Saturday assistance at more than 150 walk-in sites from March 7 to
April 11, 1998

Promoted March 28 as EIC Awareness Day and April 4 and 11 as Problem Prevention
days

Took various steps to alert taxpayers and practitioners about changes relating
to the EIC

Enforcement

Issued notice 97-65 notifying practitioners of new due diligence requirements
and allocated 18 staff years to do face-to-face compliance reviews of return
preparers

Increased QRP staffing to screen 1.3 million more questionable returns

Increased staffing in correspondence examination program to handle a total of
50,000 EIC case referrals, primarily from the QRP

Increased Criminal Investigation staffing by 40 special agents and 10 aides to
investigate potential fraud cases, including those involving return preparers

Issued notices informing more than 383,000 EIC taxpayers that IRS had
identified an SSN problem on their tax year 1996 returns and reminding them to
correct the problem before filing their tax year 1997 returns

Expanded the EIC math error program to include invalid SSNs for primary
taxpayers\a

Increased staffing to process a total of 25,000 cases involving taxpayers who,
after receiving a math error notice denying their EIC claim, protest IRS' action
but provide no documentation to substantiate their protests

Selected 140,000 returns for audit because of duplicate use of SSNs for EIC
qualifying children

Increased staffing to audit about 300,000 tax returns with potential errors
claiming head of household filing status and qualifying children

Compliance research

Baseline compliance study of tax year 1997 EIC returns

EIC claimant profile study and longitudinal study of EIC claimant filing
patterns

Feasibility analysis for using two federal databases as authorized under TRA97:
the SSA database linking parent and child SSNs and the HHS Federal Case Registry
of Child Support Orders

Study of noncompliance among EIC claimants who report income from self-
employment
--------------------------------------------------------------------------------
\a The "primary" taxpayer is the taxpayer filing the return or, on a
joint return, the taxpayer listed first. 

Source:  IRS EIC compliance initiative documents. 

As indicated in table 5 and discussed in more detail below, several
components of the EIC compliance initiative are directed at issues
that were identified by the tax year 1994 EIC compliance study as
major sources of EIC errors. 


         CUSTOMER SERVICE AND
         OUTREACH
-------------------------------------------------------- Letter :6.2.1

To the extent that EIC errors, whether they involve qualifying
children requirements, filing status, or misreported income, are
unintentional and due to a misunderstanding of the rules, IRS'
customer service and outreach efforts may help improve compliance. 
IRS data show that many taxpayers took advantage of the expanded
customer service IRS offered in 1998.  For example, IRS expanded
telephone access for EIC-related issues to 7 days a week, 24 hours a
day.  According to IRS data, 95,000 taxpayers called the EIC
assistance lines during the times when IRS' other assistance lines
were not available.  In addition, IRS provided Saturday walk-in
assistance at between 152 and 173 sites from March 7 through April
11, 1998.  IRS data show that staff available on these 6 Saturdays
helped 2,949 taxpayers prepare their EIC returns and provided 1,032
others with different types of EIC-related assistance.  According to
IRS, this is in addition to 185,305 EIC taxpayers assisted on
weekdays during the filing season. 

Some choices IRS made in implementing its assistance and outreach
efforts in 1998, however, limited the number of persons who might
have benefited.  For example: 

  -- IRS did not offer Saturday walk-in assistance until March 7, by
     which time millions of EIC claims had already been filed.  IRS
     reported that it had received about 7.4 million EIC claims as of
     February 21, 1998--2 weeks before the first Saturday that
     walk-in help was available.  EIC Awareness and Problem
     Prevention days were held even later in the filing season.  IRS
     said that it did not offer Saturday service earlier in the year
     because "prior to receiving the [EIC] appropriation, we had
     anticipated having Saturday service for only the last six weeks
     of the filing season" when, according to IRS officials, demand
     among all filers is generally higher.  The date for the EIC
     Awareness Day was selected so that IRS would have adequate time
     to publicize and provide for quality service to the public.  IRS
     officials said, in retrospect, it could have been more effective
     if scheduled earlier. 

  -- IRS did not advertise the 24-hour availability of telephone
     assistance for EIC-related issues.  IRS informed taxpayers of
     this service only if they received a notice from IRS about a
     problem with the EIC claims on their tax returns.  IRS officials
     told us that they did not advertise this service because they
     thought that it would lead to many non-EIC calls during the
     hours when other assistance lines were closed.\37

As noted earlier, TRA97 included provisions that allow IRS to deny
future EIC claims.  These provisions are to be implemented in 1999,
based on returns filed in 1998.  For example, persons found to have
intentionally disregarded the rules and regulations in filing their
EIC claims in 1998 can be denied the credit for the following 2
years.  IRS attempted to warn taxpayers about the implication of
these provisions before they filed their returns in 1998.  Those
outreach efforts were intended to create a deterrent effect by
providing an incentive for intentionally noncompliant taxpayers to
file a correct return and for other taxpayers to be sure that they
understand the EIC rules before filing.  To the extent that result
was achieved, the number of EIC errors may have been reduced. 
Although we have no way of knowing how successful those warnings were
in encouraging better compliance, we believe that the chances for
success might have been enhanced if IRS had done a better job of
publicizing those warnings. 

In that regard, IRS' income tax return instructions did not alert
taxpayers as clearly as they could have about the TRA97 provisions
and their implications.  The tax year 1997 Form 1040 tax package
included the following statement in its general information on
"what's new for 1997":  "Caution:  If it is determined that you are
not entitled to the EIC you claim, you may not be allowed to take the
credit for certain future years.  See [Publication] 596 for details."
A reference to this caution was not included later in the package
either with the instructions for filling in the EIC line item on the
tax return, the EIC worksheets, or the Schedule EIC that taxpayers
must submit with their returns to substantiate their EIC claims. 
Thus, IRS was relying on taxpayers to read the general information in
the front of the tax package before preparing their returns and,
assuming they did, to order Publication 596 for details.  For tax
year 1996, about 19 million taxpayers claimed the EIC and IRS
distributed about 636,000 copies of Publication 596.  We believe that
potential EIC claimants would have been more likely to read the
relevant information from Publication 596 if it had been included in
the Form 1040 instructions, along with statements in the EIC-specific
parts of those instructions that clearly alerted taxpayers to the
existence of that warning and where to find it.\38


--------------------
\37 This should not be a problem in 1999 when IRS plans to offer
24-hour assistance on all of its telephone assistance lines.  In
commenting on a draft of this report, IRS said that this expanded
service will be advertised in all tax packages and elsewhere. 

\38 The tax year 1997 version of Publication 596, entitled Earned
Income Credit, included the following two paragraphs:

"Beginning in 1997, if you improperly claim the [EIC] due to reckless
or intentional disregard of IRS rules or regulations, you cannot
claim the credit for the next 2 years.  Also, if you fraudulently
claim the [EIC] you cannot claim the credit for the next 10 years. 
These sanctions are in addition to any other penalty imposed, such as
the accuracy-related penalty or the fraud penalty.

"Beginning in 1997, if you improperly claim the [EIC] and IRS denies
it as a result of deficiency procedures, you cannot claim the credit
again unless you provide information required by the IRS that shows
you are eligible to claim the credit.  The IRS will send you
information about how to become recertified.  If you claim the credit
without first being recertified by the IRS, your claim will
automatically be denied.  The recertification procedures can apply if
you are subject to the above described 2- or 10-year disallowance
period."


         ENFORCEMENT AND RESEARCH
-------------------------------------------------------- Letter :6.2.2

The customer service and outreach efforts discussed above are
generally broad based and not targeted to specific sources of EIC
noncompliance.  In contrast, IRS' compliance initiative includes
several enforcement and research activities that are specifically
targeted on issues relating to qualifying children, the head of
household filing status, noncompliant return preparers, and
misreported income. 

Qualifying child errors.  IRS' tax year 1994 EIC compliance study
showed that qualifying child errors associated with the residency
requirement and AGI tiebreaker rules accounted for about half of the
$4.4 billion EIC overclaim total and 1.8 million\39 of the 2.3
million returns with a qualifying child error.  These errors
undoubtedly included both unintentional mistakes and intentional
noncompliance and involved a variety of complex living situations. 

IRS is able to verify some EIC eligibility criteria using tax return
or Schedule EIC information and does so through its math error
program as returns are submitted.  IRS receives few indicators,
however, of other problematic eligibility requirements, such as
qualifying child residency or the presence of another taxpayer in the
household who should be claiming the child.  IRS has targeted its
enforcement efforts on those compliance problems that can be
identified from tax return information or profiles of noncompliant
returns and is able to resolve some eligibility issues through
correspondence audits.  However, the bulk of noncompliance, primarily
related to qualifying children, can best be identified through
face-to-face audits. 

One component of the compliance initiative that combines elements of
customer outreach and enforcement is targeted on cases where a
qualifying child's SSN is used on more than one tax return for the
same tax year.  Because a qualifying child can be claimed only once,
resolution of these duplicate SSN cases should eliminate EIC claims
by taxpayers with whom the child did not reside.  For the outreach
portion of this effort, IRS identified about 225,000 qualifying child
SSNs that had been used by more than one taxpayer on tax year 1996
returns.  In December 1997, IRS sent taxpayers using these SSNs
(about 383,000 taxpayers) a notice informing them of the problem and
reminding them to file a correct return for tax year 1997.  To
evaluate the effectiveness of these notices, IRS plans to check for
duplicate use of these qualifying child SSNs on tax year 1997
returns.  According to IRS, it plans to begin its evaluation in
September 1998 and report the results in February 1999. 

For the compliance portion of this effort, IRS allocated additional
staff to audit as many as 140,000 taxpayers who had used about 92,000
duplicate qualifying child SSNs in both tax years 1995 and 1996.\40
According to IRS officials, as of May 16, 1998, about 103,000 of the
140,000 taxpayers had filed tax year 1997 returns, and IRS had frozen
their refunds.  Also as of May 16, 1998, however, IRS had released
49,000 of the refunds for taxpayers who had corresponded with IRS but
whose conflicting claims for the child(ren) in question were not
resolved.\41 In discussing the release of these refunds, IRS
officials told us that it could not process the amount of
correspondence received because IRS (1) did not have enough time to
adequately prepare for the start of this project (e.g., get staff
assigned, procedures developed, and training done); and (2) had
underestimated the volume of taxpayer contacts it would receive. 
Although IRS is continuing to investigate these cases, its
effectiveness in protecting the revenue has been compromised because
it is more difficult (and more costly) to recoup an erroneous refund
once it has been released.  IRS officials told us that meaningful
data on the results of this effort would not be available until
September 1998. 

Another way that IRS attempts to deal with qualifying child errors is
to deny EIC claims when the taxpayer has failed to provide valid SSNs
for the listed children.\42 This effort, which is part of IRS' math
error program, began before the compliance initiative and has
continued as part of the initiative.  As of June 4, 1998, IRS had
sent about 535,000 EIC SSN-related math error notices to tax year
1997 filers; at the same point in 1997, IRS had sent about 774,000
such notices to tax year 1996 filers.  IRS data for all of tax year
1996 show that it stopped approximately $876 million in erroneous
refunds through the EIC SSN math error program.  As of March 1998,
IRS data show that it stopped about $414 million in tax year 1997
refunds.\43 IRS expected to issue fewer SSN math error notices in
1998 because IRS, before the 1998 filing season, had sent notices to
about 600,000 taxpayers with known SSN problems telling them what to
do to correct the situation before filing their tax year 1997
returns. 

TRA97 included provisions giving IRS access to an SSA data file
linking parent and child SSNs and a Federal Case Registry of Child
Support Orders to be administered by HHS.  The Federal Case Registry
is to be a compilation of state child support and custody data. 
Access to both data files is intended to augment IRS' ability to
detect EIC claims for nonqualifying children.  Both, however, are
still in development, and IRS plans to do a "feasibility analysis"
regarding their use.  However, it will be several years before IRS
will be able to use these data.  Access, in terms of the specific
data fields IRS can obtain, is still a major issue to be resolved
among the three agencies.  In addition to access issues, IRS'
feasibility analysis is to include an assessment of data accuracy,
currency, and completeness--factors that will be especially important
for the custodial data to be useful. 

Filing status errors.  IRS' tax year 1994 study showed that a large
proportion of qualifying child errors occurred in tandem with
erroneous claims of head of household status.  One of the components
of the EIC compliance initiative involves increased staffing to
expand a project aimed at a universe of about 345,000 head of
household EIC claimants whose returns contain other indicators of
potential qualifying child problems.  This project was initiated in
1997 with audits of about 53,000 returns and expanded in 1998 to
313,000 returns.  As of March 1998, about 50,700 of the 53,000 audits
begun in 1997 had been closed; and about 43,400 of those closures (86
percent) resulted in tax changes totaling about $107 million.  On the
basis of those results, IRS expects that about 85 to 90 percent of
the 1998 audits will result in a change to the EIC claim.  According
to IRS, results of these audits will not be available until late 1998
or early 1999.  IRS officials estimated that about 25 percent of the
313,000 audits will be completed by September 30, 1998. 

Errors involving misreported income.  Misreported income accounted
for about 16 percent of the total EIC overclaims identified in IRS'
tax year 1994 EIC compliance study.  Many of IRS' traditional
compliance activities are designed to identify returns with
misreported income.  For example, EIC returns are subject to IRS'
document matching program, which compares W-2 wage reports and other
income information reports (e.g., those filed on Form 1099) with
income reported on tax returns.  Because misreported income is of
particular concern within that segment of the population that reports
self-employment income on Schedule C, the EIC compliance initiative
includes a study of noncompliance among EIC claimants who report
self-employment income.  IRS selected a sample of tax year 1997
returns, held the refunds, and plans to complete the audits by
September 1998.  IRS plans to issue a report of its findings in
February 1999. 

Paid preparer noncompliance.  The tax year 1994 study data showed
that returns prepared by local or informal preparers had a higher
overclaim rate (31 percent) than the returns prepared by formal
preparers (20 percent).  To address preparer noncompliance, TRA97
imposed due diligence requirements on paid preparers who complete EIC
returns and fines for preparers who fail to comply with those
requirements.  In December 1997, IRS issued specific due diligence
requirements\44 and publicized these requirements in mailings to
practitioners.  IRS did not institute at a national level specific
procedures to monitor compliance with the due diligence requirements
during the 1998 filing season.  However, individual field offices may
have done some monitoring.  At the Northern California District
Office, for example, we were told that staff phoned about 560
preparers in the district to inform them of the due diligence
requirements and inquired into conformity with those requirements as
part of the district's normal monitoring visits to about 100
preparers.  IRS informed us that national-level plans for the 1999
filing season include due diligence monitoring visits to EIC return
preparers, but IRS has not decided on the procedures for these
visits, the number of visits, or the extent to which they will target
those preparers most likely to be noncompliant (i.e., local or
informal preparers). 

As part of the EIC compliance initiative, IRS also planned to
increase district office Criminal Investigation staffing in fiscal
year 1998 to investigate potential EIC fraud cases, including cases
involving return preparers.  The increased staffing was to include a
total of 40 special agents and 10 investigative aides.  For fiscal
year 1998 as of May 31, 1998, 31 paid preparer cases have been opened
compared to 44 for all of fiscal year 1997. 

Fraud detection.  IRS' QRP is aimed at identifying tax returns with
potentially fraudulent refund claims.  The scoring system used to
identify these returns is based on known characteristics of
potentially fraudulent returns.  As part of the compliance
initiative, IRS expanded QRP staffing to allow screening of 1.3
million more returns in fiscal year 1998 than in fiscal year 1997,
for a total of 4 million returns.  According to IRS, as of April 30,
1998, QRP teams had scanned about 2.3 million potentially fraudulent
EIC returns and had identified 6,476 returns with erroneous EIC
claims totaling $17.6 million. 


--------------------
\39 Confidence interval is  11 percent of the point estimate. 

\40 The 140,000 taxpayers selected for audit as part of this project
were not included in the group of 383,000 taxpayers who received
notices related to duplicate SSN use. 

\41 Also as of May 16, 1998, IRS had released refunds for 21,000
taxpayers who did not claim the disputed child for 1997 and another
4,500 taxpayers who, on the basis of audits of 1996 returns, were
entitled to claim the disputed child(ren). 

\42 A valid SSN is one in which the SSN and associated name match the
information in SSA's files. 

\43 If taxpayers, after receiving the math error notice, correct the
SSN problems that caused their refunds to be stopped, IRS is to
release the refunds.  Because that process can take several months,
the $414 million in stopped refunds as of March 1998 is likely to
decrease as the year progresses. 

\44 Under these requirements, a paid preparer must (1) complete an
EIC eligibility checklist or some substitute form that contains the
same information; (2) complete the EIC worksheet, or keep a paper or
electronic record of the EIC computation that includes the
computation method and information used; (3) not know or have reason
to know that any information used to determine EIC eligibility is
incorrect; and (4) retain, for 3 years, a copy of the completed
checklist, the worksheet, and a record of how and when the
information was obtained and who provided the information. 


      TIMELY DATA ON THE RESULTS
      OF EIC COMPLIANCE EFFORTS
      WOULD FACILITATE
      DECISIONMAKING
---------------------------------------------------------- Letter :6.3

As is evident from our discussion of the various elements of the EIC
compliance initiative, there was little information available at the
time we completed our audit work on the results of IRS' efforts and
thus little basis for us or IRS to assess their effectiveness.  Such
data and assessments are crucial as IRS decides on the compliance
initiative's future direction. 

An obvious question one would ask in assessing IRS' results is "how
much has the EIC overclaim rate changed since the start of the
initiative?" Although the results of the tax year 1994 EIC compliance
study were the catalyst behind congressional funding of the
compliance initiative, IRS does not plan to use those results as the
baseline for measuring the initiative's impact on the EIC overclaim
rate.  Instead, it plans to measure the initiative's impact against
the results of a tax year 1997 compliance study, which IRS has begun
as part of the initiative.\45 However, by the time IRS completes the
tax year 1997 study, which is to become the baseline, and a tax year
1998 study that can be compared with the baseline to measure change,
IRS will be in the fourth year of its 5-year initiative.  IRS' time
frame for the tax year 1997 baseline study shows that the analysis
will not be completed until fiscal year 2000.  If the tax year 1998
study follows the same schedule, its results will not be available
until fiscal year 2001--the fourth year of the initiative.  It will
be too late at that point to make substantive changes to the
initiative. 

Given the time frames associated with the broad compliance studies,
it is important that IRS closely monitor the results of the
initiative's individual components so that it can make more timely
and better informed decisions about revising, deleting, or expanding
those various components.  For example, information on the results of
the notices IRS sent users of duplicate SSNs in December 1997 would
be useful in deciding whether to send similar notices in December
1998.  As noted earlier, however, IRS does not plan to begin such an
assessment until September 1998 and does not expect to have final
results until well after December 1998. 


--------------------
\45 By using tax year 1997 compliance as its baseline, IRS will not
be measuring the impact of EIC changes implemented in tax year 1997
or earlier years. 


   CONCLUSIONS
------------------------------------------------------------ Letter :7

Although minor methodological problems in IRS' tax year 1994 EIC
compliance study could have led to some over- or understatement of
total EIC overclaims, these issues do not affect the relevance of the
study's findings.  The study demonstrates that EIC noncompliance is a
significant issue and that verifying qualifying child eligibility
lies at the heart of EIC compliance problems.  Targeting compliance
efforts at qualifying child errors, however, presents IRS with a
major challenge.  IRS is not set up to systematically verify
qualifying child eligibility.  Doing so would basically require IRS
to establish a process to verify eligibility before issuing a refund,
similar to the processes used in EIC compliance audits. 

IRS' EIC compliance initiative includes a broad array of customer
service, enforcement, and research activities aimed at reducing
noncompliance.  Some parts, like special audits of head of household
claimants and preparer due diligence requirements, are targeted
specifically at areas of noncompliance identified in the tax year
1994 study.  Others, like expanded walk-in and telephone assistance,
are more broadly based efforts aimed at improving taxpayers'
understanding of EIC rules.  Although it is too early to judge the
initiative's effect on noncompliance, we did identify some
opportunities for IRS to improve future implementation efforts.  For
example, IRS did not offer Saturday walk-in assistance until late in
the filing season when millions of EIC claims had already been filed. 
Also, for the TRA97 provisions allowing IRS to deny future EIC claims
to act as a deterrent, taxpayers must be aware of the circumstances
under which these penalties will be applied.  IRS' income tax return
instructions, however, did not alert taxpayers as clearly as they
could have about these provisions and their implications. 

IRS plans to do annual studies to measure the impact of its EIC
compliance initiative on the overclaim rate.  Based on IRS' time
frame for these studies, however, useful data on impact will not be
available until fiscal year 2001--the fourth year of the initiative. 
It will be too late at that point to make substantive changes. 
Absent timely data on the overall impact of IRS' efforts and given
the need for IRS to ensure that available resources are used as
effectively and efficiently as possible, it is important that IRS
have immediate information on the results of individual components of
that initiative.  Evaluation plans that are not timed to provide data
when data are most needed, as appears to be the case with IRS'
planned evaluation of the notices on duplicate SSNs, are of limited
value. 


   RECOMMENDATIONS TO THE
   COMMISSIONER OF INTERNAL
   REVENUE
------------------------------------------------------------ Letter :8

We recommend that the Commissioner of Internal Revenue

  -- ensure that customer service efforts aimed at EIC claimants be
     available earlier in the filing season when most EIC claims are
     filed, and

  -- include prominent information regarding the 2-year and 10-year
     sanctions and the recertification process in the Form 1040 EIC
     instructions and Schedule EIC. 

In addition, to provide a basis for continually improving and
refocusing EIC compliance efforts, we recommend that the Commissioner
develop evaluation plans for each compliance initiative component
that will provide, in succeeding years of the initiative, timely data
for decisionmakers. 


   AGENCY COMMENTS AND OUR
   EVALUATION
------------------------------------------------------------ Letter :9

We requested comments on a draft of this report from the Commissioner
of Internal Revenue and the Secretary of the Treasury, or their
designees.  The Commissioner of Internal Revenue responded in a July
2, 1998, letter generally agreeing with our recommendations (see app. 
I).  On July 1, 1998, we met with IRS officials, including the Deputy
Chief of Operations, the Acting Assistant Commissioner for Customer
Service, and the Assistant Commissioner for Research/Statistics of
Income, to discuss the Commissioner's comments.  Treasury's Deputy
Assistant Secretary (Tax Analysis) responded in a June 29, 1998,
letter (see app.  II). 

In response to our recommendation that IRS provide customer service
efforts earlier in the filing season, IRS said that it plans to
publicize EIC awareness events early in the 1999 filing season and to
hold EIC awareness activities beginning in January 1999.  IRS
officials told us that (1) Saturday service at walk-in assistance
sites during the 1999 filing season will begin on January 16 and
continue through the filing season, and (2) the first 6 Saturdays
will be publicized as EIC help days.  These actions, if effectively
implemented, will be responsive to our recommendation. 

In response to our second recommendation about more prominently
displaying information on the 2-year and 10-year sanctions and the
recertification process, IRS said that it will include such
information in the tax year 1998 Schedule EIC instructions but will
not revise the schedule itself.  IRS said that it did not believe the
Schedule EIC should be revised to address these issues because the
issues do not affect the majority of filers and providing the
information on the schedule may confuse filers who have not had their
EIC claim disallowed.  According to IRS officials, taxpayers must go
to the worksheet in the instructions to complete the schedule, and
IRS' intent is to place the information so that persons using the
worksheet will easily see it.  Although inclusion of the information
in the Schedule EIC instructions is an improvement, we continue to
believe that something should also be added to the schedule.  Because
one of the purposes of this information is to alert potential EIC
claimants to possible repercussions if they make erroneous claims,
the information affects all filers.  Also, although it is true that
taxpayers who choose to compute their own EIC have to use the
worksheet in the instructions, taxpayers who choose to have their
returns prepared by someone else do not have to use the worksheet and
thus would see only the Schedule EIC.  We are not suggesting that all
of the information on sanctions and recertification be included on
the schedule.  What we are suggesting is that a brief, but prominent,
cautionary statement be added to the schedule alerting users to
important information in the instructions that they should read
before filing their returns. 

Regarding our final recommendation, IRS said that it understood our
concern regarding more timely delivery of research data for
decisionmaking.  According to IRS, it has developed an information
delivery strategy that includes developing information systems that
will allow more timely delivery of both interim and final tax return,
audit, and research data.  The strategy includes using interim
reports to disseminate preliminary findings from various EIC
projects.  For example, IRS officials said that they hope to have, in
October 1998, some preliminary findings from audits of taxpayers who
had used duplicate qualifying child SSNs.  IRS also noted that using
interim data of this sort has limitations; it may not be adequate to
measure revenue or provide a full understanding of taxpayer behavior. 
Although there are certain limitations associated with interim data,
we believe, as IRS recognized in its comments, that such data can be
of value to decisionmakers. 

Treasury's letter addressed two statements in the draft report--our
characterization of the EIC as an income transfer program and our
statement that IRS cannot address a significant root cause of
noncompliance.  In response to our statement comparing the EIC to
other income transfer programs, Treasury said that "unlike income
transfer programs, the [EIC] makes work pay by reducing tax
liabilities," and that about 80 percent of the EIC's total costs
offset individual income, Social Security, and other federal taxes. 
We clarified our reference to income transfer programs where
appropriate. 

In response to our statement in a draft of this report that IRS
cannot address a significant root cause of noncompliance (IRS'
difficulty verifying qualifying child eligibility), Treasury said
that issues of verifying family relationships and living arrangements
are not unique to the EIC but also affect taxpayers' eligibility for
dependency exemptions, filing status, the child credit, and the child
and dependent care tax credit.  Also, both IRS and Treasury said that
they were hopeful that access to new data (an HHS registry of child
support orders and SSA data linking parent and child SSNs) will allow
IRS to detect some qualifying child problems during return
processing.  We modified our statement and related discussions in the
report to acknowledge IRS' ability to identify some noncompliance
related to qualifying children.  Our report also recognizes the
provision in TRA97 giving access to the HHS and SSA databases. 
However, IRS told us that it will not be testing use of these
databases until late 1999 or 2000 and that the amount of information
that can be initially expected is small. 


---------------------------------------------------------- Letter :9.1

As agreed with your offices, unless you publicly release its contents
earlier, we plan no further distribution of this report until 30 days
from the date of this letter.  At that time, we will send copies to
the Ranking Minority Member, Committee on Ways and Means; the
Chairmen and Ranking Minority Members of other interested
congressional committees; the Secretary of the Treasury; the
Commissioner of Internal Revenue; and other interested parties. 

Major contributors to this report are listed in appendix III.  If you
or your staffs have any questions, please call me on (202) 512-9110. 

James R.  White
Associate Director, Tax Policy
 and Administration Issues




(See figure in printed edition.)Appendix I
COMMENTS FROM THE INTERNAL REVENUE
SERVICE
============================================================== Letter 



(See figure in printed edition.)



(See figure in printed edition.)




(See figure in printed edition.)Appendix II
COMMENTS FROM THE DEPARTMENT OF
THE TREASURY
============================================================== Letter 



(See figure in printed edition.)


MAJOR CONTRIBUTORS TO THIS REPORT
=========================================================== Appendix 0

GENERAL GOVERNMENT DIVISION

David Attianese, Assistant Director
Deborah Parker Junod, Evaluator-in-Charge
James Fields, Senior Social Science Analyst
John Lesser, Senior Evaluator
MacDonald R.  Phillips, Economist
Robert Dunsky, Economist

KANSAS CITY FIELD OFFICE

Doris Hynes, Senior Evaluator

SAN FRANCISCO FIELD OFFICE

Susan Mak, Evaluator

RELATED GAO PRODUCTS

Earned Income Credit:  Noncompliance Relative to Other Components of
the Income Tax Gap (GAO/GGD-97-120R, June 13, 1997). 

Earned Income Credit:  Claimants' Credit Participation and Income
Patterns, Tax Years 1990 Through 1994 (GAO/GGD-97-69, May 16, 1997). 

Tax Administration:  Earned Income Credit Noncompliance
(GAO/T-GGD-97-105, May 8, 1997). 

Earned Income Credit:  IRS' 1995 Controls Stopped Some Noncompliance,
But Not Without Problems (GAO/GGD-96-172, Sept.  18, 1996). 

Earned Income Credit:  Profile of Tax Year 1994 Credit Recipients
(GAO/GGD-96-122BR, June 13, 1996). 

Earned Income Credit:  Noncompliance and Potential Eligibility
Revisions (GAO/T-GGD-95-179, June 8, 1995). 

Earned Income Credit:  Targeting to the Working Poor
(GAO/GGD-95-122BR, Mar.  31, 1995). 

Earned Income Credit:  Targeting to the Working Poor
(GAO/T-GGD-95-136, Apr.  4, 1995). 

Tax Administration:  Earned Income Credit--Data on Noncompliance and
Illegal Alien Recipients (GAO/GGD-95-27, Oct.  25, 1994). 

Tax Policy:  Earned Income Tax Credit:  Design and Administration
Could Be Improved (GAO/GGD-93-145, Sept.  24, 1993). 

*** End of document. ***