Earned Income Tax Credit: Implementation of Three Tests Proceeded
Smoothly, But Tests and Evaluation Plans Were Not Fully 	 
Documented (30-DEC-04, GAO-05-92).				 
                                                                 
Research has shown that the Earned Income Tax Credit (EITC) has  
helped lift millions of individuals out of poverty. In recent	 
years, the Internal Revenue Service (IRS) has paid approximately 
$30 billion annually to about 20 million EITC recipients.	 
However, the program also has experienced a high rate of	 
noncompliance. IRS estimated that EITC overclaim rates for tax	 
year 1999, the most recent data available, were between 27 and 32
percent of dollars claimed or $8.5 billion and $9.9 billion,	 
respectively. We were asked to describe the three tests IRS has  
begun to reduce overclaims and how the funds appropriated for	 
them were spent; assess how well IRS implemented the tests and	 
describe planned refinements for the 2005 tests; and assess	 
whether IRS's evaluation plans had sufficient documented detail  
to facilitate managerial review and stakeholder oversight and	 
describe the status of the 2005 evaluation plans.		 
-------------------------Indexing Terms------------------------- 
REPORTNUM:   GAO-05-92						        
    ACCNO:   A15036						        
  TITLE:     Earned Income Tax Credit: Implementation of Three Tests  
Proceeded Smoothly, But Tests and Evaluation Plans Were Not Fully
Documented							 
     DATE:   12/30/2004 
  SUBJECT:   Appropriated funds 				 
	     Eligibility determinations 			 
	     Evaluation methods 				 
	     Income taxes					 
	     Internal controls					 
	     Noncompliance					 
	     Overpayments					 
	     Program evaluation 				 
	     Tax administration 				 
	     Tax credit 					 
	     Tax returns					 
	     Taxpayers						 
	     Earned Income Tax Credit				 

******************************************************************
** This file contains an ASCII representation of the text of a  **
** GAO 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.                                         **
**                                                              **
******************************************************************
GAO-05-92

                 United States Government Accountability Office

GAO Report to Congressional Requesters

December 2004

EARNED INCOME TAX CREDIT

 Implementation of Three New Tests Proceeded Smoothly, But Tests and Evaluation
                        Plans Were Not Fully Documented

                                       a

GAO-05-92

[IMG]

December 2004

EARNED INCOME TAX CREDIT

Implementation of Three New Tests Proceeded Smoothly, But Tests and Evaluation
Plans Were Not Fully Documented

                                 What GAO Found

IRS implemented three tests in 2004 to address leading sources of EITC
errors: a qualifying child test, where selected taxpayers were asked to
document that their child lived with them for more than half the year in
2003; a filing status test, where selected taxpayers were asked to provide
documentation to prove the accuracy of their 2003 filing status, and an
income misreporting test, where a new screening process was used to select
EITC returns that identify taxpayers likely to have the most significant
changes in their assessments due to underreporting of income on their tax
return.

Leading Sources of EITC Errors Contributing to Overclaims in Tax Year 1999

IRS's implementation of the tests proceeded smoothly and largely as
planned. However, some information, such as a key change in the filing
status test, was not well documented and the level and quality of some
services provided to test participants were not measured. This lack of
documentation hindered monitoring, oversight, and did not foster a common
understanding of the tests. For the 2005 tests, IRS made key changes to
the qualifying child test to encourage taxpayers to certify in advance of
filing their return and to attempt to simulate what might happen with
nationwide implementation. IRS also changed the sample selection criteria
for the filing status test to better target noncompliant taxpayers.

IRS's plans for evaluating the 2004 tests generally lacked documentation
and detail for many key issues, which undermined their value to managers
and stakeholders. For example, IRS did not specify how it planned to
analyze some qualifying child survey data. In essence, an evaluation plan
is the management plan or roadmap for the evaluation endeavor and
welldeveloped plans facilitate test management and oversight. Despite the
importance of having evaluation plans prior to implementation, IRS had not
completed its plans for the 2005 tests before two of the tests had begun.

                 United States Government Accountability Office

Contents

  Letter

Results in Brief
Background
IRS Implemented Three Tests on Leading Sources of EITC

Noncompliance and Reported Spending Most of the Funding Received on the
Tests Tests Implemented Smoothly, and Refinements for the Fiscal Year 2005
Tests Made

IRS's 2004 Evaluation Plans Lacked Sufficient Documented Detail to Allow
for Oversight; Evaluation Plans for 2005 Tests Were Not Completed Before
Two of the Tests Had Begun

Conclusions
Recommendations for Executive Action
Agency Comments and Our Evaluation

1 4 6

10

24

35 39 40 41

Appendixes

Appendix I: Appendix II:

Appendix III: Appendix IV:

Scope and Methodology

Updated Results from the Income Misreporting Test as of
September 30, 2004

Comments from the Internal Revenue Service

GAO Contacts and Staff Acknowledgments

GAO Contacts
Staff Acknowledgments

43

45

47

51 51 51

Tables           Table 1: EITC Requirements for Tax Year 2003            7 
           Table 2: Description of the Qualifying Child Certification Test 
                                                                       and 
                                      Subtests                             13 
              Table 3: Criteria for Assessing IRS's Test Implementation    44 
            Table 4: Status of Income Misreporting Test, June 26, 2004 and 
                             September 30, 2004 Results                    46 
Figures    Figure 1: Leading Sources of EITC Errors Contributing to        
                             Overclaims in Tax Year 1999                    9
           Figure 2: Characteristics of the Test Sample for the Qualifying 
                                     Child Test                            12 
              Figure 3: The Qualifying Child Certification Test Process    15 
                      Figure 4: The Filing Status Test Process             18 

Contents

Figure 5: Characteristics of the Taxpayers in the Income

Misreporting Test by Filing Status 21 Figure 6: The Income Misreporting
Test 22

This is a work of the U.S. government and is not subject to copyright
protection in the United States. It may be reproduced and distributed in
its entirety without further permission from GAO. However, because this
work may contain copyrighted images or other material, permission from the
copyright holder may be necessary if you wish to reproduce this material
separately.

A

United States Government Accountability Office Washington, D.C. 20548

December 30, 2004

The Honorable Amo Houghton Chairman, Subcommittee on Oversight Committee
on Ways and Means House of Representatives

The Honorable Earl Pomeroy Ranking Minority Member, Subcommittee on
Oversight Committee on Ways and Means House of Representatives

Researchers generally consider the Earned Income Tax Credit (EITC), a
federal program providing tax relief to low-income workers, to be a
successful antipoverty program. The Council of Economic Advisers
reported that because of the EITC, an estimated 4.3 million individuals-
including 2.2 million children-were lifted out of poverty in 1997.1 In
recent years, IRS has paid about $30 billion annually to about 20 million
EITC recipients. However, the EITC program has long experienced high
rates of noncompliance. IRS's most recent EITC compliance study
estimated that between $8.5 billion and $9.9 billion of the EITC claims
filed
for tax year 1999 should not have been paid. This amount represents an
estimated rate of EITC overclaims-total erroneous claims less any
amount that IRS recovered or expects to recover-of between 27 and 32
percent of EITC dollars claimed.

In February 2002, when the most recent compliance study was released, a
task force of IRS and Treasury officials was convened to find ways of
reducing EITC overclaims. The IRS/Treasury task force found that the
three leading errors were responsible for about $7 billion of overclaims
each year. These errors resulted from taxpayers (1) claiming children who
were not a qualifying child, meaning they do not meet certain
requirements,
primarily that the child did not live with them for more than half the tax
year ($3 billion in overclaims); (2) using an incorrect filing status of
either
single or head of household, when the correct status was married filing
jointly or married filing separately ($2 billion in overclaims); and

1Council of Economic Advisers, Good News for Low Income Families:
Expansions in the Earned Income Tax Credit and the Minimum Wage
(Washington, D.C.: December 1998). The Council of Economic Advisers gives
advice directly to the President and other senior members of the
administration on fiscal policy.

(3) misreporting, primarily underreporting, their income ($2 billion in
overclaims). In all three cases, these errors resulted in taxpayers
receiving a larger credit than they should have received.

IRS received about $52 million in fiscal year 2004 for a new EITC
five-point initiative to improve service, fairness, and compliance with
the EITC program. Included in that amount were funds for three EITC tests
to evaluate new methods for reducing the overclaim rate in the three
leading problem areas identified by the IRS/Treasury task force-qualifying
child, filing status, and income misreporting. After designing the tests
and evaluation plans to guide the agency in its assessment of the tests,
IRS implemented them beginning in fiscal year 2004. Although the tests and
evaluation of them continued into fiscal year 2005, IRS issued a status
report to Congress on the tests in August 20042 to comply with a
legislative mandate.3 IRS also decided to begin another round of EITC
tests in fiscal year 2005 in the same three areas, but with some
refinements. The qualifying child and income misreporting tests started in
the first quarter of fiscal year 2005 and the filing status test will
begin in the second quarter of 2005.

Because IRS's plans surrounding the tests have garnered much attention,4
you asked us to review each of the fiscal year 2004 tests and (1) describe
the tests and how funds appropriated for them were spent, (2) assess how
well IRS implemented the tests and describe IRS's planned refinements for
the fiscal year 2005 tests, and (3) assess whether IRS's plan for
evaluating the tests contained sufficient documented detail to facilitate
managerial review and stakeholder oversight and describe the status of
IRS's evaluation plan for the fiscal year 2005 tests.

To describe the tests and determine how the funds appropriated for them
were spent, we analyzed IRS documents and interviewed IRS officials. Our
assessment of IRS's implementation of the tests and documentation of its
evaluation plans for the tests was based on IRS's stated goals for them,

2 Internal Revenue Service, IRS Earned Income Tax Credit (EITC)
Initiative: Status Report to Congress (Washington, D.C.: August, 2004).

3 Pub. L. No. 108-199, 118 Stat. 3 (2004).

4 We reported on this topic in 2003, focusing on the test relating to
qualifying child errors. See GAO, Earned Income Credit: Qualifying Child
Certification Test Appears Justified, but Evaluation Plan Is Incomplete,
GAO-03-794 (Washington, D.C.: Sept. 30, 2003).

which were that the tests reduce overclaims and, for the qualifying child
test, also maintain EITC participation for eligible participants and
minimize taxpayer and IRS administrative burden.

To assess how well IRS implemented the tests and describe IRS's planned
refinements for the 2005 tests, we developed criteria to evaluate IRS's
implementation (i.e., the execution and day-to-day management) of the
tests. Those criteria included whether IRS sent correspondence to the
taxpayers' correct address, where taxpayers could go to receive assistance
about the tests and the quality of that assistance, IRS's hiring and
training of staff, and other aspects of administering the tests. To assess
the implementation based on these criteria, we analyzed IRS's status
reports; reviewed policies and procedures; observed test operations in
Kansas City, Missouri; Atlanta, Georgia; and Fresno, California and
selected those locations based on a variety of reasons, including the
location of key managers and work; interviewed IRS officials; and reviewed
a judgmentally selected sample of cases for each test at those locations
in order to review a variety of case types.

To assess whether IRS's plans for evaluating the tests contained
sufficient documented detail to facilitate managerial review and
stakeholder oversight, we used GAO guidance and the social science
evaluation literature to identify key attributes of an evaluation. We
shared these attributes with IRS officials and they generally agreed to
their relevance. These attributes included the research design, outcome
measures,5 target and sample populations, data collection activities,
analyses, and dissemination of results. We obtained all available
documentation on IRS's evaluation plans for each of the tests and reviewed
that documentation to determine whether we could understand from the
documentation alone how IRS planned to address the key attributes. Where
we could not, we interviewed IRS officials to further understand whether
and how the officials planned to address those key attributes. Written
documentation should be complete, facilitate tracing of events, and be
readily available for examination to foster a common understanding of the
program and

5 Outcome measures are used to assess the tests' impact. Well-defined
outcome measures clearly name and define a measure as well as the
methodology used in its calculation.

facilitate oversight.6 To describe the status of IRS's evaluation plan for
the fiscal year 2005 tests, we primarily relied on interviews with IRS
officials. Appendix I provides more detail on the scope and methodology,
including a more detailed description of the implementation criteria used
in conducting our work.

We reviewed documentation and interviewed IRS officials to ensure that the
data that we received from IRS were reliable for the purposes used in this
report and determined they were. As we conducted our work, IRS continued
to collect and analyze significant amounts of data from each test. The
tests continued into fiscal year 2005, primarily to finish cases where
taxpayers had either not responded to IRS's request for substantiation or
the substantiation provided was deemed insufficient by IRS. Thus, the
analyses presented in this report are based on work that we performed
while the tests were ongoing. Final conclusions about the impact of the
tests on overclaim rates, burden, and participation should not be made
until all data are collected and analyzed and IRS publishes its final
report, which is due to Congress by June 30, 2005. We performed our work
from October 2003 to December 2004 in accordance with generally accepted
government auditing standards.

Results in Brief 	IRS implemented three tests-qualifying child, filing
status, and income misreporting-to evaluate the potential for reducing the
EITC overclaim rate beginning in fiscal year 2004. For the qualifying
child test, selected taxpayers were asked to document, when filing their
2003 tax return, that their qualifying child lived with them for more than
half the tax year. The filing status test had selected taxpayers provide
documentation to prove the filing status claimed on their 2003 tax return.
For the income misreporting test, IRS employed a new screening process to
select EITC tax returns from an existing program to identify taxpayers
likely to have the most significant underreporting of income on their tax
return and, therefore, the highest potential EITC overclaim amount. IRS
reported spending about $17.5 million on the three EITC tests-about $3.2
million less than planned because expected workloads did not materialize.
Because IRS spent less than planned on the tests themselves, it was able
to

6 See GAO, Internal Control Management and Evaluation Tool, GAO-01-1008G
(Washington, D.C.: August 2001); GAO, Standards for Internal Controls in
the Federal Government, GAO/AIMD-00-21.3.1 (Washington, D.C.: November
1999); and GAO, Designing Evaluations, GAO/PEMD-10.1.4 (Washington, D.C.:
May 1, 1991).

fund some activities under the five-point initiative that otherwise would
have gone unfunded.

Implementation of the tests generally proceeded smoothly. IRS addressed
the major issues that arose during implementation and based planned
refinements for the 2005 tests primarily on lessons learned from the 2004
tests. Some of the strengths of the tests' implementation include that
they were conducted largely as planned and IRS hired and provided staff
with training. An example of an issue that arose and IRS subsequently
addressed was when some examiners mistakenly required taxpayers who were
part of the filing status test to complete documentation required for the
qualifying child test. Once this problem was identified, according to IRS
officials, internal quality reviews helped ensure it was no longer
occurring. Although implementation generally proceeded smoothly, some
important information about the tests, including the basis for a key
policy change, was not well documented and the level and quality of some
services offered to taxpayers were not measured. As a result, it was
difficult for management or staff to gain a common understanding of the
program and rationale for changes and to fully monitor the implementation.
Further, developing and documenting such information would enable others
to review the methodology. IRS made refinements to all three tests for the
subsequent round of fiscal year 2005 tests. Two of the most significant
refinements were to the qualifying child test, including that (1)
taxpayers would be encouraged to certify in advance of filing their return
that their child met the EITC residency requirement and (2) a portion of
the taxpayers would be drawn from a single community.7 According to IRS
officials, earlier certification could help them process cases more
quickly and get eligible taxpayers their refund faster. Targeting one
community would simulate what might happen if a certification requirement
were imposed nationwide. Another key refinement was related to the filing
status test, whereby IRS would change the sample selection criteria to
better target noncompliant taxpayers.

IRS's plans for evaluating the three 2004 tests had critically important
strengths such as identifying test goals and linking evaluation objectives
and outcome measure to these goals. However, IRS's evaluation plans were
generally not in sufficient detail to facilitate managerial review and

7 Although IRS mailed letters to sampled taxpayers in this community, some
uncertainty exists about implementation because of a lawsuit that was
filed to block the test as we were finalizing our report.

stakeholder oversight and help ensure that the evaluations' results would
be communicated in such a manner that others could understand and judge
their strengths and limitations. In essence, an evaluation plan is the
management plan or roadmap for the evaluation endeavor and would
facilitate oversight. As such, the more completely a plan is developed,
the more likely it will be useful to managers in ensuring that the
evaluation is well-executed. IRS's evaluation plans lacked detail, for
example, about key planned analyses that IRS intended to conduct. For
example, IRS did not specify how it planned to analyze some qualifying
child survey data. Despite the importance of having detailed plans prior
to implementation, IRS has not completed its evaluation plans for the
fiscal year 2005 tests even though two of those tests have begun.

We are recommending that the Commissioner of Internal Revenue ensure that
the rationale for key policy decisions and other significant events be
documented; obtain some information on the quality and level of service
for all types of taxpayer assistance provided as part of the EITC tests;
when disseminating the test results ensure they clearly state test and
evaluation limitations and complete the development of comprehensive and
adequately detailed evaluation plans for the 2005 tests.

We requested comments on a draft of this report from the Commissioner of
Internal Revenue. We received written comments, which are reprinted in
appendix III. In his comments, the Commissioner said that IRS agreed with
the recommendations. We further discuss the Commissioner's comments in the
"Agency Comments and Our Evaluation" section of the report and in other
sections where appropriate.

Background	The EITC, enacted in 1975,8 was originally intended to offset
the burden of Social Security taxes and provide a work incentive for
low-income taxpayers. The credit has been modified several times since its
introduction, and three laws9 have been enacted in recent years aimed at
resolving some concerns with EITC rules. Despite modifications, the
original goal of the credit remains intact and the EITC continues to
provide a substantial benefit to millions of American families.

8 26 U.S.C. Sec. 32.
9 Pub. L. No. 107-16, 115 Stat. 38 (2001), Pub. L. No. 106-170, 113 Stat.
1860 (1999), and Pub.

L. No. 105-34, 111 Stat. 788 (1997).

The EITC is a refundable tax credit, meaning that qualifying working
taxpayers may receive a refund greater than the amount of income tax they
paid for the year. EITC payments have a (1) phase-in range in which higher
incomes yield higher EITC amounts, (2) plateau phase in which EITC amounts
remain the same even as income rises, and (3) phase-out range in which
higher incomes yield lower EITC amounts. The amount of credit a taxpayer
receives is based on several other factors, such as the presence and
number of qualifying children. In general, taxpayers with one or more
qualifying children receive a higher credit than taxpayers without
qualifying children. For tax year 2003, the amount of EITC that could be
claimed with two qualifying children ranged from $0 to $4,204 per tax
return filed, depending on income and filing status.

EITC requirements for tax year 2003 include rules for all taxpayers
claiming the credit and additional rules that differ depending on whether
or not a taxpayer has qualifying children (see table 1).

                  Table 1: EITC Requirements for Tax Year 2003

Additional rules for Rules for all taxpayers Additional rules for
taxpayers taxpayers without a claiming the EITC with a qualifying child
qualifying child

Must have a valid Social Income less than: If one child: Income less than:
$11,230

Security number	$29,666 (or $30,666 if married (or $12,230 if married
filing jointly). If more than one filing jointly) child: $33,692 (or
$34,692 if married filing jointly).

           Cannot use married filing Child must meet age, relationship, Must  
           be at least 25 years                                               
                    separately status and residency tests a old, but under 65 
                        Must be a U.S. citizen or Child can be claimed by one 
                                        resident alien all year taxpayer only 

Cannot be the dependent of another person

Cannot file form 2555 or 2555-EZ

Cannot be a qualifying child of another taxpayer

Cannot be a qualifying child of another taxpayer

Must have investment Must have lived in U.S. income of $2,600 or less more than
                                 half the year

Must have earned income

Source: IRS.

a In general, a qualifying child must be under 19 years old at the end of
the tax year and have lived with the taxpayer for more than half the year.
To meet the relationship test, the qualifying child had to be a son,
daughter, adopted child, stepchild of the taxpayer, or a descendent of any
such individual. Sisters, brothers, stepsisters, stepbrothers, and
descendents of any such individuals also qualify if the taxpayer cares for
the individual as they would their own child. In addition, a foster child
can qualify for the relationship test if certain conditions are met.

IRS has periodically measured EITC compliance.10 For tax year 1999, (the
most current data available), IRS estimated the EITC overclaim rates at 27
to 32 percent of EITC dollars claimed, or $8.5 billion to $9.9 billion.11
IRS has limited data on underclaims, which for tax year 1999 were
estimated to be $710 million to $765 million. Because of the persistently
high rates of noncompliance, we also have identified the EITC program as a
high-risk area for IRS since 1995.12

In February 2002, the compliance study was released and the Assistant
Secretary of the Treasury, Tax Policy, and IRS Commissioner convened a
joint IRS/Treasury task force to identify ways of reducing EITC
overclaims, while maintaining participation among eligible claimants and
minimizing taxpayer and IRS's administrative burden. The task force found
that the leading causes of errors resulting in EITC overclaims were due to
taxpayers (1) claiming children who were not a qualifying child, (2) using
an incorrect filing status, and (3) misreporting their income. With this
information, the task force designed what ultimately became initial
versions of the three tests, as show in figure 1. As envisioned by the
task force, even if fully implemented, IRS does not plan to apply the test
requirements to the entire EITC population because IRS can use available
data to verify the eligibility of certain taxpayers.

10 IRS measured EITC overclaim rates in the past using differing
methodologies. In 1997, IRS estimated the overclaim rate at 23.8 to 25.6
percent of EITC dollars claimed.

11 Department of the Treasury, Internal Revenue Service, Compliance
Estimates for Earned Income Tax Credit Claimed on 1999 Returns
(Washington, D.C.: Feb. 28, 2002).

12 Prior to 2001, EITC was part of a broader IRS tax filing fraud
high-risk area that we defined. Beginning in 2001, the focus of that
designation was narrowed to EITC. GAO, High-Risk Series: An Update,
GAO-01-263 (Washington, D.C.: January 2001).

Figure 1: Leading Sources of EITC Errors Contributing to Overclaims in Tax
Year 1999

Source: GAO.

Because a new analysis of EITC compliance using 2001 tax return
information is not expected to be complete until spring 2005, IRS did not
know whether compliance has significantly changed since 1999 when
developing the EITC tests. However, IRS officials do not think EITC
compliance has improved substantially since then. In October 2004,
Congress passed a new law to make the definition of a qualifying child
uniform in various IRS provisions, but those changes are not effective
until tax years after December 31, 2004. In general, the revised
definition appears to mainly affect other tax situations, such as claiming
dependents, more than just affecting the EITC.13 IRS is studying whether
the change would affect any testing that may be done in 2006.

13 Pub. L. No.108-311, 118 Stat. 1166 (2004).

Having a Complete Evaluation Plan Before Implementation Helps to Ensure
Success

IRS completed its initial evaluation plans for the three EITC tests in
December 2003. In September 2003, we recommended that IRS accelerate the
development of its qualifying child evaluation plan to help ensure the
success of the test. 14

An evaluation plan ideally should be completed and disseminated for review
and feedback before beginning the research activity (or in this case,
test). As we reported, although an evaluation plan need not precisely
identify all issues and how they will be evaluated before implementation,
the more complete a plan is, the more likely the evaluation will be
sufficient and support future decisions. IRS's Internal Revenue Manual15
also recognizes the desirability of having an evaluation plan in place
before a project is implemented; for example, it requires such plans
before reorganizations.

IRS Implemented Three Tests on Leading Sources of EITC Noncompliance and
Reported Spending Most of the Funding Received on the Tests

In an effort to implement the joint IRS/Treasury task force
recommendations, IRS implemented three new tests-qualifying child
certification, filing status, and income misreporting-in 2004. IRS
reported spending about $17.5 million on the three EITC tests-about $3.2
million less than planned. Because IRS spent less than planned on the
tests, it was able to fund some activities under the five-point initiative
that otherwise would have gone unfunded.

Qualifying Child The purpose of the qualifying child certification test
was to evaluate the Certification Test Required impact on the test goals
of asking taxpayers to substantiate-when filing Substantiation of Child
their tax return-that their qualifying child lived with them for more than

half the tax year, as required by the EITC (see table 1). Under current
rules,Residency taxpayers are only required to substantiate that their
child satisfied this residency requirement if they are being audited by
IRS on this issue.

14 GAO-03-794. 15 See Internal Revenue Manual 1.1.4.

This test involved two random samples of 25,000 taxpayers who claimed one
or more qualifying children for tax year 2002: a test sample, whose
members were asked to substantiate their qualifying child's residency, and
a control sample, whose members had similar characteristics to the test
sample, but were not asked for any substantiation. Both samples were
designed to include taxpayers (1) most likely to make errors and (2) whose
qualifying child eligibility could not be verified from information
available to IRS. IRS used prior research results to determine which
taxpayers would be most likely to incorrectly claim a qualifying child.

The research showed that with those taxpayers most likely to make errors,
the errors often correlated with the taxpayer's relationship to the child
and the taxpayer's gender and filing status. Taxpayers most frequently
making qualifying child errors included both fathers and males and females
who were not the child's parents and who filed as single or head of
household.

IRS also used available data to obtain evidence about taxpayers and
whether their qualifying children met residency and relationship
requirements. For example, a child's residency could be established with
some certainty by using the Department of Health and Human Services'
Federal Case Registry, and a child's relationship to the taxpayer could be
established with some certainty using the Social Security Administration's
KIDLINK.16 When this available evidence supported the taxpayers' EITC
claim that they had a qualifying child, those taxpayers were excluded from
the qualifying child test.

Prior research showed that taxpayers who comply with the residency
requirement also comply in most cases with the relationship requirement.
Thus, if a taxpayer's child met the residency requirement, there was a
high probability that the relationship requirement would be met as well.
Given this analysis and difficulties IRS encountered in identifying
documents that taxpayers could readily obtain to prove their relationship
to the child, any taxpayer whose EITC eligibility was not verified from
available data became eligible to be selected for the qualifying child
test in which they would be asked to substantiate that the child lived
with the taxpayer for

16 The Federal Case Registry compiles court and other records that
indicate who is the custodian for a child. IRS assumes that children live
with the custodian of record. KIDLINK ties parents' and children's Social
Security numbers for children born after 1998 in U.S. hospitals.

more than 6 months during the year. Our September 2003 report contains a
more detailed explanation on how the sample was designed.17

As figure 2 shows, males filing as single or head of household comprised
the majority of the test sample. The control group had characteristics
similar to the test group.

Figure 2: Characteristics of the Test Sample for the Qualifying Child Test

Married filing jointly

                      Single or head of household females

                       Single of head of household males

The qualifying child test had three components-a general test and two
subtests. Under the general test, taxpayers received test documentation in
English only18 and could have provided substantiation in one or any
combination of three ways-records, letters, or a Schedule A, also known as
the general affidavit.19 Records that a taxpayer could provide included
school, medical, landlord, or child-care provider documentation. Letters
were statements from certain individuals, such as a member of the clergy
or a community based organization official, on official letterhead.

17 GAO-03-794.

18 Taxpayers were told on the forms that they could request that documents
be provided in Spanish.

19 Schedule A, referred to as the general affidavit, can be signed by
parties including an attorney, child care provider, clergy, court
official, employer, health care provider, Indian tribal official,
landlord, school official, or social service agency or other government
official.

Affidavits were legal documents in which an individual attests that the
taxpayer's qualifying child resided with the taxpayer for a certain period
of time. To be accepted, the document(s) had to contain various data, such
as the names of the qualifying children and the dates the child lived with
the taxpayer.

In response to concerns that taxpayers may have difficulty obtaining
certification through the official sources cited on the Schedule A, such
as through an attorney or landlord, and that English-only documents might
weaken participation among taxpayers with limited English proficiency, IRS
also implemented two subtests. The Schedule B, also known as the friends
and neighbors affidavit subtest, for 1,000 of the 25,000 taxpayers,
broadens the definition of the individuals allowed to certify the child's
residency to include those who have personal knowledge of a taxpayer's
circumstances, such as certain family members.20 The purpose of this
subtest was to determine whether such individuals could facilitate an
increase in residency certification for eligible taxpayers. The taxpayers
in the Spanish subtest, 1,000 of the remaining 24,000, received documents
in both English and Spanish. The purpose of this subtest was to determine
whether Spanish language documents would increase the number of taxpayers
attempting to certify their child's residency. Table 2 describes the test
and subtests.

Table 2: Description of the Qualifying Child Certification Test and
Subtests

                      Test or Subtest Description of Test

General Test  o  Documentation sent in English only (23,000 taxpayers)  o 
Only General Affidavit (Schedule A) provided

Schedule B Subtest  o  Documentation sent in English only
(1,000 taxpayers)  o  Friends and Neighbors Affidavit (Schedule B)
provided

Spanish Subtest  o  Documentation sent in English and Spanish (1,000
taxpayers)  o  Only General Affidavit (Schedule A) provided

Source: GAO analysis of IRS data.

20 The Schedule B affidavit, referred to as the friends and neighbors
affidavit, differs from the Schedule A affidavit in that it broadens the
list of individuals who can sign the affidavit to anyone who has records
or personal knowledge, other than the taxpayer's spouse, dependent,
qualifying child, or parent of the qualifying child.

IRS sent the selected taxpayers21 five documents in December 2003
informing them about the test, including:

1.	Notice 84-A, a letter informing the taxpayer about the new
certification requirements;

2.	Form 8836, Qualifying Children Residency Statement, to be completed by
the taxpayer and returned to IRS;

3. Schedule A or B (an affidavit) that could be used for certification;

4.	Publication 3211M, Earned Income Tax Credit Questions and Answers; and

5.	Publication 4134, Free/Nominal Cost Assistance Available for Low Income
Taxpayers.

Under the test and subtests, once taxpayers received the documents from
IRS, they were supposed to obtain supporting documents to prove their
qualifying child's residency and send that documentation back to IRS in
the same envelope as their 2003 tax return. IRS would withhold, or
"freeze," the EITC portion of the taxpayers' refund until acceptable
documentation proving a child's residency was received. Once IRS received
the documentation, IRS examiners in Kansas City, Missouri, would review it
and send a letter to the taxpayer accepting the claim, asking for
additional information, or rejecting the claim. If the taxpayers provided
acceptable documents, IRS would release the taxpayer's EITC portion of
their refund.

If acceptable documentation was not provided or if no response was
provided following a second notification letter, the taxpayer's EITC claim
would be denied and the taxpayer would be informed of his or her right to
appeal to the U.S. Tax Court.22 This process is depicted in figure 3.

21 Some taxpayers were removed from the 25,000 sample and not replaced
because IRS learned that they had mitigating situations, such as serving
in combat or living in a disaster zone. The true sample size (i.e., the
number of taxpayers IRS required documentation from) was actually 24,711.

22 IRS allowed for internal appeal hearings only if the taxpayer
maintained contact with IRS and requested such a hearing. In the original
contact letter, IRS told taxpayers they could request an internal appeal
hearing.

Figure 3: The Qualifying Child Certification Test Process

                                  Source: GAO.

Filing Status Test Required Substantiation That the Correct Filing Status
Was Used

Another cause of EITC errors is when taxpayers claim an incorrect filing
status. EITC filing status errors occur when married taxpayers incorrectly
use single or head of household. Married taxpayers who incorrectly file
individually as single or head of household could qualify for a larger
EITC than they would otherwise be entitled to if they claimed the correct
filing status. This is because, pursuant to statute,23 IRS considers the
combined income of married taxpayers who file jointly for purposes of
determining the amount of EITC for which the taxpayer(s) qualifies.24
Using combined

23 I.R.C. Section 32(b)(2)(B).
24 Taxpayers who file as married filing separate are not eligible for the
EITC.

income may result in taxpayers exceeding the EITC income ceiling,
therefore receiving no credit at all, or qualifying for a lesser credit
amount. For example, in tax year 2003, married taxpayers filing jointly
with $17,500 of income each, or a combined earned income of $35,000, and
four qualifying children would not be eligible for the EITC. However, if
each taxpayer incorrectly filed as head of household, claimed two
qualifying children, and their $17,500 income, they would each receive a
credit of $3,405 or a combined total of $6,810. IRS's databases offer
limited ability to independently or systematically identify taxpayers who
may be claiming an incorrect filing status.

The primary purpose of the filing status test was to evaluate the impact
on overclaims of requiring taxpayers whose filing status has changed from
married to single or head of household any time between 1999 through 2002
to substantiate the filing status they claimed on their 2003 tax return.
To select the population for the filing status test, IRS started with a
computer file of approximately 1.6 million taxpayer returns, or a 10
percent sample of all taxpayers who claimed the EITC with one or more
qualifying children on their 2002 return. IRS eliminated the qualifying
child and income misreporting test populations and, applied other
exclusions, such as taxpayers subject to an audit examination, or
taxpayers with more than one potential EITC-related issue. From that
population, IRS selected taxpayers whose returns showed a filing status of
married at least once in the previous 3 years. This resulted in a sample
of 69,000 taxpayers, which IRS sorted by gender, zip code and filing
status. Using a random sampling method, IRS selected 36,000 of these
taxpayers for this sample who filed as single or head of household on
their 2003 tax return. Females filing as single or head of household
comprised 96.9 percent of the test sample.

The taxpayers in the 36,000 sample who filed a 2003 tax return claiming
the EITC received a letter from IRS about 2 weeks after filing their
return informing them that the EITC portion of their refund would be
delayed until IRS reviewed their return. Within 30 days, IRS sent a second
letter asking taxpayers to verify their filing status, using the enclosed
Form 886-FS, Filing Status Information Request and send it back to the
IRS. This form requires taxpayers to provide documentation as to why they
did not file as married for tax year 2003. Taxpayers were asked to provide
IRS with documentation that they were divorced or legally separated as of
December 31, 2003, they did not live with their spouse for the last 6
months of the year, the spouse was deceased, or some other reason existed
to

warrant a change of filing status.25 IRS examiners reviewed the form and
accompanying documentation and sought clarification or additional proof,
if needed. If IRS examiners accepted the documentation, they released the
EITC portion of the taxpayer's refund and closed the case.

If a taxpayer did not respond or IRS found the taxpayer's documentation
unacceptable, then IRS sent the taxpayer a notice stating that IRS (1)
changed the taxpayer's filing status from single or head of household to
the married filing separate status, (2) disallowed the EITC, and (3)
changed the taxpayer's standard deduction to the appropriate amount. In
addition, IRS forwarded the taxpayers a letter informing them of their
right to appeal the changes to U.S. Tax Court.26 This process is depicted
in figure 4.

25 This is the same documentation IRS would request in an audit if filing
status were an issue.

26 Similar to the qualifying child test, IRS allowed for internal appeal
hearings only if the taxpayer maintained contact with IRS and requested
such a hearing.

Figure 4: The Filing Status Test Process

Source: GAO.

The filing status test also included a subtest to gather additional
information on EITC claimants who used the head of household filing
status. The IRS/Treasury task force found that taxpayers using the head of
household filing status were more likely to misstate their filing status
than taxpayers using a different one. IRS selected 500 taxpayers who filed
as head of household on their 2003 tax return. The sample of 500 taxpayers
showed 99 percent females and 1 percent males with the head of household
filing status.

Unlike the test for the 36,000 sample, IRS did not ask taxpayers in the
subtest of 500 who filed a 2003 return to provide supplemental
documentation to support their filing status until after they had received
their EITC refunds. And, unlike taxpayers in the 36,000 sample, where IRS
had some information they had filed as married at least once from 1999
2002, IRS did not have such information on the taxpayers in the 500
sample.

In fact, IRS could not determine whether these taxpayers were ever
married. As a result, IRS asked these taxpayers to confirm their
eligibility for the head of household filing status, which they claimed on
their 2003 tax return, by either (1) calling IRS on a special toll-free
number and stating that they used the correct filing status or (2)
completing a stub that was attached to the letter they received, checking
yes or no, and mailing or faxing it to IRS. IRS did not ask these
taxpayers to provide substantiation to support the filing status they
claimed. This was, in part, because IRS had not identified any
documentation that would be available to support a taxpayer's claim that
he or she had never been married. If taxpayers indicated they were not
eligible to use the head of household filing status, they could correct
their filing status by sending in an amended tax return either by mail or
fax.27 IRS asked taxpayers to provide the information within 45 days from
the date on the letter. All taxpayers who did not respond would be subject
to an examination before their 2004 EITC refund would be released.

In another aspect of the filing status test, IRS planned to determine
whether a third-party service that attempted to locate the address of
taxpayers could be as reliable as the filing status test in identifying
taxpayers who had used an incorrect filing status. The locator service
used information from credit bureaus to determine whether taxpayers were
living together and possibly married. The information from the locator
service had no impact on taxpayers for this year's filing status test.

Income Misreporting Test Used New Screening Process to Find Cases Likely
to Yield the Highest Assessments

Although some taxpayers could receive a larger EITC by over-reporting
their income, misreporting of income for EITC is generally an
understatement, according to IRS, resulting in the taxpayer receiving a
higher credit amount than entitled. The purpose of the income misreporting
test was to evaluate the impact on the test goal of a new screening
process to select EITC tax returns 28 that identify taxpayers likely to
have the most significant changes in their tax assessments due to
underreporting of income on their tax return.

27 IRS did not assess any penalties, interest, or additional tax due if
the taxpayer sent an amended return. This decision was for the 2004 test
only.

28 Unlike the other two tests, the income misreporting test was based on
2002 tax returns. This was because IRS does not match information returns
from third parties until several months after April 15th.

Income misreporting is a component of an existing program known as
Automated Underreporter (AUR). 29 Under that program, IRS attempts to
match income information as reported by the taxpayer on the tax return to
information reported by third-party sources, such as a taxpayer's employer
or bank. In instances where this matching process identifies
discrepancies, IRS may assess additional taxes on the taxpayer. The annual
AUR matching program identifies far more cases than IRS has staff to work.
In determining which cases to work, IRS selects not only cases that it
believes will generate the highest probable assessment, but also cases
involving taxpayers who underreport different types of income (e.g.,
wages, interest). In the past, some of those cases-roughly 300,000 per
year- involved the EITC. However, EITC was not one of the different types
of categories from which IRS historically had chosen cases.

For the income misreporting test, IRS attempted to select---from all the
EITC cases for which AUR found an income mismatch on 2002 tax
returns-300,000 EITC cases expected to provide the highest EITC
assessments. IRS employed a computer selection tool that used variables
such as the taxpayer's filing history, filing status, and number of
children to rank the cases in terms of the highest probable EITC
assessments.

Additionally, IRS designed the test to determine whether certain
characteristics of the selected cases made them more likely to yield
higher assessments. Thus, IRS placed each of the selected cases in one of
four groups: (1) "repeater egregious," cases in the same income category
for the third year in a row and were assessed an additional tax for the
previous 2 years; (2) "repeater worked," cases worked at least once during
the last 3 years; (3) "repeater not worked," cases in the income
misreporting inventory at least once in the last 3 years, but not worked;
and (4) "other criteria," cases randomly selected from the other three
categories and other criteria, such as first-time underreporters.

As figure 5 shows, the majority, 62 percent, of the taxpayers selected for
the income misreporting test filed their return using the head of
household filing status, while 30 percent claimed married filing jointly
and 8 percent claimed a single filing status.

29During fiscal year 2003, the AUR program assessed taxpayers about $2.9
billion.

Figure 5: Characteristics of the Taxpayers in the Income Misreporting Test
by Filing Statusa

Single and others

Married filing jointly

Head of household females

aIRS was unable to provide characteristics by gender.

IRS added the income misreporting test cases back into the general AUR
inventory, and examiners in Atlanta, Georgia; Austin, Texas; and Fresno,
California worked the test cases using the same processes as for all other
AUR cases. Examiners manually screened all cases for simple math errors or
errors that could not be picked up by a computer (e.g., placing an amount
on the wrong line). If such an error was found and resolved, the tax
return was accepted, and the case was closed. If the examiner could not
resolve the discrepancy, the examiner sent a notice30 to the taxpayer
explaining that IRS found a discrepancy on his or her return. The taxpayer
was given 30 days to respond to the notice. If no response was received,
IRS sent another notice informing the taxpayer that the IRS had determined
there was a deficiency in the return and the taxpayer must pay an
assessment based on the deficiency or file a petition with the U.S. Tax
Court within 90 days. If IRS received a response that took issue with
IRS's assessment, the examiner would then determine whether the response
was sufficient to support the taxpayer's original tax return.31 If the
response was sufficient, the examiner would close the case with no
additional tax assessed. If the response was not sufficient or a response
was not

30 In most of these cases, a CP-2000 notice, "We Are Proposing Changes to
Your Tax Return", is sent to the taxpayer; however, IRS sends a different
notice in rare instances, such as when there was an income discrepancy of
at least $100,000.

received, the IRS examiner would assess the taxpayer the additional tax.
This process is depicted in figure 6.

                     Figure 6: The Income Misreporting Test

                     Source: GAO and Art Explosion (image).

31 IRS officials said the Internal Revenue Manual guides this
determination and that examiners receive months of training in how to make
this determination.

IRS Reported Spending Less on Tests Than Anticipated

IRS reported spending about $17.5 million on the three EITC tests-about
$3.2 million less than planned.32 This funding was part of the
Consolidated Appropriations Act of 2004,33 which provided IRS with $52
million in fiscal year 2004 for a five-point initiative to improve
service, fairness, and compliance with the EITC program. IRS announced the
new initiative in June 2003. The initiative addresses:

o 	reducing the backlog of pending EITC examinations to ensure that
eligible taxpayers whose returns are being examined receive their refunds
quickly;

o 	minimizing burden and enhancing the quality of communications with
taxpayers by improving the existing audit process;

o 	encouraging eligible taxpayers to claim the EITC by increasing outreach
efforts and making the requirements for claiming the credit easier to
understand;

o 	ensuring fairness by refocusing compliance efforts on taxpayers who
claimed the credit, but were ineligible because their income was too high
(or filing status was incorrect); and

o 	piloting a certification effort to substantiate qualifying child
residency eligibility for claimants whose returns are associated with a
high risk of error.

Of the $52 million budgeted, IRS reported spending or obligating $51.8
million in fiscal year 2004. Of that, IRS officials said they spent about
$17.5 million on the tests---$7.4 million was spent on the income
misreporting test, $5.6 million on the filing status test, and $4.5
million on the qualifying child test. IRS officials noted that, in some
cases, the amounts they reported spending differed from what they
budgeted. For example, IRS originally budgeted $7.2 million on the filing
status test, but reported spending $5.6 million on direct costs for that
test.

32 We could not verify these figures because IRS does not have a cost
accounting system. See GAO, Financial Audit: IRS's Fiscal Years 2004 and
2003 Financial Statements, GAO05-103 (Washington, D.C.: Nov. 10, 2004).

33 Pub. L. No. 108-199, 118 Stat. 3 (2004). The Act did not earmark a
specific amount for the tests.

According to IRS officials, they spent about $3.2 million less than
anticipated on the tests primarily because some planned work did not
materialize. For example, for the filing status test, IRS originally
planned to work more cases but about 10,000 taxpayers who were originally
selected for the filing status test were not included for various reasons,
such as they did not claim the EITC. IRS officials said that, as a result,
they redirected funding to improvement projects within the five-point
initiative that would otherwise have gone unfunded.

Tests Implemented IRS's implementation of the tests generally proceeded
smoothly because of

IRS actions including use of a detailed project plan and
managementSmoothly, and involvement. IRS addressed most of the major
issues that arose duringRefinements for the implementation and released a
status report to Congress in August 2004. Fiscal Year 2005 Tests IRS's
plans for most refinements for the 2005 tests are based on the lessons

that it learned from the 2004 tests.

Made

Tests Were Executed Largely as Planned, Thus Meeting the Original Intent

The implementation plans for all three tests generally followed the
recommendations of the IRS/Treasury task force, and IRS's only significant
departure from those recommendations was based on an informed decision.
The task force recommended that taxpayers claiming the EITC (1) provide
IRS with documentation to prove a qualifying child's residency prior to
payment of the credit (the qualifying child test), (2) submit additional
data to establish that they are claiming the correct filing status (the
filing status test), and (3) use a new screening process to select tax
returns from an existing program to identify taxpayers likely to have the
most significant underreporting of income on their tax return and,
therefore, the highest potential EITC overclaim amount (income
misreporting test). In all three tests, IRS gathered information needed to
determine whether the task force recommendations have potential for
reducing the EITC overclaim rate without undue adverse effects. It was
important that IRS followed the task force recommendations; otherwise, the
validity of those recommendations would remain unknown.

IRS made an informed choice in not implementing one recommendation. The
task force had also recommended that taxpayers certify the child's
relationship to the taxpayer. However, IRS determined that this was a
lesser compliance problem than residency and that it could be difficult
for taxpayers to provide some of the documentation that IRS planned to

request for certification of the relationship. In addition, since both
residency and relationship requirements had to be met to claim the EITC,
if taxpayers fail residency certification, which is more likely according
to the compliance study, there would be no need to test for relationship.

To implement each test, IRS prepared a detailed project plan with time
frames for numerous action items such as developing notices, creating
organization charts, hiring staff, developing training materials, working
on systems needs, and determining samples. We found that IRS officials
used these plans extensively. For example, initially, IRS managers checked
the plan daily to determine if the schedule was being followed and less
often as the tests progressed. For a task to be marked as completed,
certain information had to be provided to the person in-charge of
monitoring the plan, including validation from a senior manager that the
task had been completed. According to IRS officials, the extensive use of
the project plan helped them execute and effectively monitor the
implementation of the tests.

Through Hiring, Training, and Management Actions, IRS Facilitated a Smooth
Implementation

Implementation went smoothly, in part because IRS hired sufficient numbers
of staff and provided adequate training to them. IRS hired about 410
staff, primarily examiners who processed cases and answered telephones, to
implement the three EITC tests in total. About 260 of the staff were for
the qualifying child and filing status test, while about 150 were for the
income misreporting test. The majority of the qualifying child and filing
status test staff were new to IRS, were hired on a temporary appointment,
and worked in Kansas City, Missouri. The income misreporting staff worked
in Atlanta, Georgia; Austin, Texas; and Fresno, California. According to
IRS officials, these staffing levels were appropriate to manage the
workload, thus contributing to the overall smooth implementation of the
tests.

IRS provided specific training for the qualifying child and filing status
tests. Among other things, the training included a history leading up to
the tests, a description of the test processes, the roles and
responsibilities of staff, several examples of how to determine whether
taxpayer substantiation was acceptable, and information on how to use the
Earned Income Credit Proof of Concept (EICPC) database, the database IRS
used to manage the qualifying child and filing status tests. Examiners we
met with in Kansas City told us the training was sufficient. However,
there was a gap between the time examiners first received training and
when they actually started doing the work. According to IRS officials,
this gap caused the staff to lose

some knowledge before they were able to apply it. To remedy this problem,
IRS provided the staff refresher training and a staff-developed job aid.
Examiners we interviewed said, that as a result, they felt confident in
making decisions to accept or reject taxpayer substantiation.

IRS did not provide specific training for the income misreporting test,
and instead relied upon the AUR program training because the process for
working cases remained the same-only how IRS selected the cases changed.
In our visits to Atlanta and Fresno, we found consistency in the training
that staff received for the income misreporting test, including how the
procedures were used when screening and working cases.

Management took steps to foster staff members' understanding of the
importance of the tests. Once the current EITC program director was
installed in late 2003, management oversight became more apparent for the
tests. Senior IRS management responsible for EITC were involved in
managing many details of implementation of the tests. To help garner staff
support, when the tests first began, IRS managers held meetings with the
examiners and took action based on their expressed concerns, such as
making key revisions to the EICPC system. In addition, front-line managers
with whom we spoke in Kansas City said the EITC director's involvement
helped marshal staff support at that location. Managers said this was
critical for smooth implementation of the tests, since they were the ones
dealing directly with the taxpayers. The examiners we interviewed also
said team meetings with managers helped them understand and effectively
convey information about the tests to taxpayers.

IRS Monitored Undeliverable Mail and Attempted to Resend It to Corrected
Addresses to Help Ensure Taxpayers' Response to Tests

IRS tracked undeliverable mail, mail that was sent to taxpayers and
returned to IRS by the U.S. Postal Service, which was critical to the
success of the tests. If taxpayers did not receive IRS's
correspondence-letters, forms, and notices-they would not have known they
needed to respond. And, had there been large volumes of undeliverable
mail, the feasibility of the tests could have been undermined. Ensuring
that those selected for the tests received the correspondence could have
been particularly difficult because research has shown that some EITC
claimants are highly mobile.

Although IRS used the most current address for test populations-in most
cases the address taxpayers provided on their 2002 tax returns-IRS
officials anticipated some mail being returned as undeliverable because
the taxpayer no longer lived at that address. IRS first learned that it
had a taxpayer's incorrect address when it received the undeliverable mail
from

the U.S. Postal Service. As it typically does for undeliverable mail, IRS
employed a locater service to attempt to find the taxpayer's new address
by using other kinds of information, such as addresses associated with any
credit cards. When the locator service found a new address, IRS resent
correspondence to the affected taxpayer.

IRS Provided Several Means to Help Answer Taxpayers' Questions and Found
Strong Performance Where Data Were Available

IRS provided several means for taxpayers selected for any of the three
tests to contact the agency for assistance. For example, in the initial
contact letter for the qualifying child test, taxpayers were informed that
they could get help from a toll-free telephone number where examiners
could answer their questions, any local IRS office-commonly known as
walk-in sites, and any of the approximately 200 low-income taxpayer
clinics (LITCs) in the U.S. In addition, the National Taxpayer Advocate34
was prepared to assist taxpayers selected for all three tests as needed.

Determining whether taxpayers received the correct information is an
important aspect of implementation. IRS's performance in terms of the
percentage of callers getting through to the agency and the quality of the
answers given was strong and comparable to similar IRS operations. IRS
received about 100,000 telephone calls from taxpayers about the qualifying
child test and about 75,000 calls about the filing status test, as of
September 30, 2004.35 Common questions about the qualifying child
certification test included "What documentation is acceptable?" and "When
will my refund be released?" According to IRS's telephone data, about 86
percent of callers got through and received service. Based on historical
data, IRS officials considered this level acceptable. Based on our annual
reviews of IRS's telephone performance during the filing season,36 we have
reported comparable levels of service. For example, in 2003, 85 percent of
taxpayer's calling IRS's main toll-free telephone lines got through and
received service. IRS's internal quality reviews showed that, as of
September 30,

34The National Taxpayer Advocate is available to help taxpayers when a
hardship arises or when their problem has not been resolved in a
reasonable time period.

35 IRS received calls about the income misreporting test on the general
AUR telephone line. IRS did not think it was necessary to track the test
calls separately from the general AUR because taxpayers were treated the
same under both the test and the program.

36 GAO, Tax Administration: IRS's 2003 Filing Season Performance Showed
Improvements, GAO-04-84 (Washington, D.C.: Oct. 31, 2003) and GAO, IRS's
2002 Tax Filing Season: Returns and Refunds Processed Smoothly; Quality of
Assistance Improved, GAO-03-314 (Washington, D.C.: Dec. 20, 2002).

2004, test examiners provided accurate responses to taxpayers seeking
assistance for the EITC tests via the telephone about 96 percent of the
time, which was somewhat higher than the quality of IRS's responses on its
tollfree telephone lines.

Because outside stakeholders expressed much concern about the tests, the
National Taxpayer Advocate decided to assist any taxpayer selected for the
tests, even if the assistance did not meet its established criteria.37 The
Advocate expected to assist about 2,600 taxpayers based on a needs
assessment, which was rooted in historical data. However, as of September
30, 2004, the Taxpayer Advocate assisted a total of 837 EITC taxpayers
participating in these tests. Most of the assistance provided included
helping taxpayers receive an expedited refund due to a financial hardship.
Internal quality reviews showed that the Advocate met its quality
standards 100 percent of the time for the test cases selected as part of
those reviews.

Steps Taken to Ensure Procedures Used by Examiners Led to Consistent
Decisions

For each test, IRS took several steps that were designed to ensure
consistency among the examiners making decisions about whether to accept
taxpayers' substantiation. The qualifying child and filing status cases
were worked in one location-Kansas City, Missouri-to make it easier to
ensure consistency among examiners. The income misreporting test cases
were worked in three locations-Atlanta, Georgia; Fresno, California; and
Austin, Texas. IRS officials said they did not believe there would be a
consistency problem in having the income misreporting test conducted
across these locations because the test was not a significant departure
from the general AUR work that had been done in these locations for the
past several years. Other examples of steps IRS took to ensure consistent
decision-making by examiners included holding multiple team meetings with
staff, sending out notices to staff when errors were noted, having certain
groups work only certain kinds of cases, preparing a job aid for
examiners, and conducting extensive quality reviews. According to IRS
managers and examiners we spoke with, these steps helped examiners make
consistent decisions in the cases they were reviewing.

37 The Taxpayer Advocate uses seven criteria to categorize and determine
the severity and level of attention that taxpayers require to resolve
their tax issue. For example, some issues, such as eviction or bankruptcy,
are categorized as imminent hardships for taxpayers and receive the most
prompt attention.

Another step IRS took to ensure consistency was to have managers in Kansas
City review all those cases where taxpayers provided substantiation for
the qualifying child test prior to filing their tax return-a total of
about 800 reviews. This review helped IRS identify and correct problems
that arose early in the tests. IRS officials stated that the review helped
provide for a smooth implementation because it identified problems, which
IRS corrected, and enabled IRS to issue supplemental guidance to ensure
repeat errors did not occur. For example, for the qualifying child test,
taxpayer substantiation did not always clearly indicate the exact dates of
a child's residency with the taxpayer-for example, some may have shown
"July through December 2003." Some examiners interpreted that to mean July
1 through December 31, giving taxpayers the time needed to certify for the
qualifying child's residency. Other examiners interpreted this same
information to mean July 1 through December 1, therefore not giving
taxpayers the time needed to qualify their child. This review identified
the inconsistent interpretation of dates, and IRS developed a policy
decision and issued guidance on how to interpret the dates when the dates
provided were vague.

IRS Internal Reviews Showed Few Significant Implementation Problems

The several on-going internal quality reviews during the tests generally
found few significant implementation problems. IRS managers conducted
reviews at the test sites, which examined accuracy, timeliness, and staff
professionalism. For the qualifying child and filing status tests, these
reviews showed generally high performance-case file documentation was
correct 87 percent and 93 percent of the time, respectively, as of
September 30, 2004. IRS did not capture this data for the income
misreporting test; however, IRS data show that, as of September 30, 2004,
95 percent of all AUR cases, which included the income misreporting cases,
contained correct documentation. The EITC Program Office also conducted a
review that assessed whether policies and procedures for the qualifying
child and filing status cases were being timely, accurately, and
consistently followed. According to IRS, the review showed good results.
For example, for the filing status test, the time between an examiner's
decision to accept taxpayers' documentation and the issuance of the
taxpayers' refund averaged fewer than 30 days.

IRS Addressed All But One Significant Problem That Arose During
Implementation

Although several problems surfaced during implementation, particularly in
the qualifying child test, IRS officials said that because they were able
to take quick actions to address the problems, the problems did not
adversely affect the tests or taxpayers selected for them to any great
extent. It was not surprising that most of the problems involved the
qualifying child test because it was a greater departure from past
practice than were the other two tests. For example, although IRS had
previously asked taxpayers to provide documents substantiating their
qualifying child upon audit, IRS has not previously allowed taxpayers to
provide affidavits to prove their claim; therefore, examiners had never
reviewed such documents. In contrast, IRS considers the filing status and
income misreporting tests expansions of existing IRS programs.

Examples of problems and IRS's actions to address them include:

o 	Early in the implementation of the test, some examiners advised
taxpayers who had called about letters received from IRS to complete
documentation for the qualifying child test even though they were selected
for the filing status test. Examiners were instructed via an email alert
to use the EICPC database to determine the test for which the taxpayer was
selected.

o 	Some qualifying child and filing status case files (paper and
electronic) had documentation deficiencies, such as not getting a
taxpayer's phone number for the case file or not obtaining
complete/required information for cases where the taxpayer agreed with
IRS's proposed changes. Through an e-mail alert, IRS officials reminded
examiners of the procedures they must follow to properly document files.

o 	Some files were missing for the qualifying child and filing status
tests. IRS established a new procedure that when a file could not be
located within 2 weeks after the taxpayer had submitted correspondence, a
new file would be created and marked "Possible Duplicate Folder."

In all three cases, IRS officials stated that the on-going quality reviews
helped ensure that examiners followed the correct procedures.

Although IRS addressed problems that arose during the implementation of
the tests, one significant problem still lingers. Some important
information about all three tests, including a key policy decision
regarding the filing status test, were either not well documented or not
documented at all. Internal control standards state that significant
decisions and events

should be documented and readily available for examination.38 When
documentation is lacking, it is difficult for management or staff to gain
an understanding of the program, refine work processes, or fully monitor
the implementation. Further, developing and documenting such information
would help ensure that test results are accurately determined and would
enable others to review the methodology.

IRS developed various management documents to organize, direct, and
monitor the test operations. However, while some important decisions about
the tests were made after these documents were developed and after test
implementation began, IRS did not explain the decisions by making
additions to the documents. For example, IRS's initial plan required that
the filing status subtest involving taxpayers who had never filed as
married, but had filed as head of household on their 2003 return, include
5,000 taxpayers. Several months later, IRS reduced the sample to 500, but
did not document the rationale for this decision until much later and at
our request. Also, certain other key information, such as when and how
information from a third-party locator service for the filing status test
would be used, was not fully developed or sufficiently detailed. The
Treasury Inspector General for Tax Administration (TIGTA) found similar
deficiencies in IRS's documentation about the tests occurring during
implementation.39 As a result, this lack of documentation hindered not
only test monitoring and oversight, but also did not foster a common
understanding of the tests.

According to IRS officials, time or other priorities caused some
significant decisions about the test not to be documented at the time
those decisions were made. Further, they said that changes to tests are
common during implementation and that they focused attention on ensuring
the tests were carried out correctly, rather than on documenting the
reasons for changes and other decisions as the tests proceeded. However,
IRS officials acknowledged that documenting significant events was
important.

In some correspondence to taxpayers about the tests, IRS referred
taxpayers to LITCs or walk-in sites for assistance. However, IRS did not
gather information on or measure the level or quality of assistance

38 GAO-01-1008G and GAO/AIMD-00-21.3.1.

39 Treasury Inspector General for Tax Administration, The Statistical
Sampling Method Used in the Earned Income Tax Credit Proof of Concept Test
Appears Valid, 2004-40-100 (Washington, D.C.: May 2004).

provided to test participants at LITCs or walk-in sites. IRS officials
said they did not collect these data because they thought taxpayer use of
this assistance would be minimal and there was no practical or cost
effective way to gather the information. In his response to our draft
report, the Commissioner echoed this sentiment, noting that because
qualifying child test participants were randomly selected from around the
country, efforts to measure services would be extremely difficult.
Further, IRS officials did not view this as an implementation problem, but
instead viewed it as a limitation of the tests. Whether it is a problem
with implementation or test design, there are some important reasons why
it would be useful to know the level and quality of services provided. For
example, our prior work found that the quality of service IRS walk-in
sites provided taxpayers was unknown.40 Further, face-to-face assistance
is costly, especially when compared to telephone services, which were used
extensively in the 2004 tests. Recognizing that options for providing
taxpayer assistance and outreach efforts are important, if IRS had data on
the level and quality of service provided, it would be in a better
position to determine the cost and benefit of providing this assistance.
Officials recognize that information on use of these forms of assistance
would be useful and indicate that they will collect information in
conjunction with a planned 2005 simulation of a nationwide test. The
simulation is discussed later in this letter.

Lessons Learned from 2004 Tests Prompt Most Refinements for New Round of
Tests

IRS officials identified lessons learned from the 2004 tests that were
implemented to help improve the 2005 tests. For example, IRS officials
plan to use of its automated telephone responses, which is important
because most taxpayers contacted IRS by telephone to obtain information
about the tests.

Changes to forms and letters based on case reviews and examiner input. IRS
officials told us that their modifications to letters and forms for the
qualifying child and filing status tests to be used for the 2005 tests
were primarily based on case file reviews and discussions with IRS
examiners who interacted with the taxpayers selected for the 2004 tests.
In April 2004, for example, EITC program officials reviewed case files and
met with examiners to discuss common taxpayer errors and questions about
letters and forms for the qualifying child and filing status tests.
Examples of taxpayer questions were: "How do I prove I did not live with
my spouse?"

40 See GAO-04-84.

or "Who can fill out the affidavit?" Examples of taxpayer errors on forms
included no signature, incomplete dates to prove residency, and either no
Social Security number listed on the form or an incorrect number. As a
result of their review, IRS officials revised the forms containing such
information (Form 8836, Qualifying Children Residency Statement and the
accompanying affidavit). For example, IRS changed the layout of the
affidavit to help reduce taxpayer errors involving dates and the amount of
time a child resided with the taxpayer. IRS did not make changes to the
letters and forms for the income misreporting test because they were the
same ones used under the general AUR program.

Improvements to key database based on examiner suggestions.

Examiners in Kansas City, the site responsible for the qualifying child
and filing status cases, suggested about 30 updates or other improvements
to the EICPC system that they said would either reduce errors in the
database, help IRS better manage the cases and workload, or improve
subsequent data analyses. For example, examiners noted they were lacking
computer fields to record certain information such as the taxpayer's
telephone number, whether the case was worked by the Taxpayer Advocate's
Office, or if an amended return had been received. As a result, examiners
suggested ways to capture these data, which have been incorporated into
the EICPC. IRS is continuing to update and make improvements to the EICPC.

Use of automated telephone voice response expanded. For the fiscal year
2004 qualifying child and filing status tests, IRS did not use automated
responses to answer routine telephone calls and did not have a mechanism
in place to obtain taxpayers' views about telephone services provided.
Both options were available for the income misreporting test and are
available for users of IRS's other toll-free telephone numbers. Officials
recognized that commonly asked questions, such as "Where do I mail the
required documentation?" or "Who can sign an affidavit?" could be answered
via automated responses, and plan to provide this option for the fiscal
year 2005 tests, leaving only the more difficult questions to be answered
by an examiner. IRS also decided to implement a random feedback survey of
taxpayers on the quality of service they received for the qualifying child
and filing status tests when they called the toll-free number. The survey
is a modified version of the one that IRS uses for its regular tollfree
telephone operations.

Changes made to the qualifying child test encourage early certification
and simulate implementation across the country. There

are two major changes to the qualifying child test for 2005: (1) taxpayers
will be encouraged to certify in advance of filing their return that their
child met the EITC residency requirement; and (2) a portion of the
taxpayers will be drawn from a single community-Hartford, Connecticut,
while the rest will be drawn randomly from across the nation.

IRS officials contend that an early certification could help reduce delays
in releasing EITC refunds because examiners would be able to validate
cases before the start of the tax filing season when workloads reach their
peak. Eligible taxpayers who provide acceptable documentation before the
start of the tax filing season could get their EITC refund more
expeditiously, IRS officials say, because the documentation would already
be validated at the time taxpayers file their tax returns. IRS has some
evidence that taxpayers are willing to certify in advance of the filing
season because about 800 taxpayers did so as part of the 2004 qualifying
child test, even though they were only asked to do so when filing their
returns.

Regarding the targeting of the single Hartford, Connecticut community, IRS
officials told us that they intend to simulate what might happen if an
early certification requirement were imposed across the country. This
change was the result of a recommendation from a contractor's review of
the 2004 test's sampling methodology. As part of this test, IRS plans to
mount an outreach campaign to include partnering with local governmental
and community-based entities to provide taxpayers assistance.

Need for refinement prompts reduction in filing status sample size.

Based on its experience with the sample selected for the 2004 filing
status test, IRS decided to dramatically reduce the sample size for next
year's test, while simultaneously trying to improve the criteria for
selecting the sample. As this year's test was implemented, IRS officials
realized that the test was yielding a high number of taxpayers claiming
the correct filing status, suggesting that the criteria for selecting them
could be improved and the burden on taxpayers to prove their filing status
was high, relative to the benefits gained. As a result, IRS officials
reduced the sample size from 36,000 to 5,000 for the 2005 test to minimize
taxpayer burden as IRS works to improve the selection criteria.

IRS also is testing two refinements in the sample selection criteria for
the 2005 filing status tests to determine whether the selection criteria
can be improved. First, IRS plans to apply TIGTA's finding, which IRS
officials said that they had also identified, that IRS could better use
information it possesses to verify the filing status of some taxpayers,
such as those whose

spouses have died or those who have submitted an amended return. Any such
taxpayers whose filing status could be verified using such information
would not be included in the sample. Second, IRS also plans to refine the
sample selection to not include taxpayers whose filing status of single or
head of household can be corroborated by information from the third-party
locator service, which was tested in 2004.

Income misreporting changes designed to improve sample selection.

IRS has planned minimal changes for the 2005 income misreporting test
because it found few issues that needed to be addressed. Changes were made
to selection criteria to help identify cases with a potentially higher
assessment amount. For example, IRS will no longer select cases where the
taxpayer's adjusted gross income is over the maximum amount for claiming
the EITC and EITC is claimed anyway because IRS found those cases yielded
a lower assessment than other cases.

IRS's 2004 Evaluation Plans Lacked Sufficient Documented Detail to Allow
for Oversight; Evaluation Plans for 2005 Tests Were Not Completed Before
Two of the Tests Had Begun

IRS's plans for evaluating the three 2004 tests lacked sufficient
documented detail to facilitate managerial review and stakeholders'
oversight and thereby help ensure that the evaluation of the tests'
results would be as sound as possible and the results would be
communicated with full recognition of their strengths and limitations. For
many aspects of IRS's evaluation plans, we were able to discern IRS
intentions by piecing together information from multiple sources,
including interviews with IRS officials. In essence, an evaluation plan is
used to manage the evaluation endeavor. As such, the more completely a
plan is developed, the more likely it will be useful to managers in
ensuring that the evaluation is wellexecuted. Despite the importance of
having detailed plans prior to implementation, IRS had not completed its
evaluation plans for the 2005 tests before two of those tests had begun.

Evaluation Plans Had Strengths Including Linkage Among Test Goals,
Evaluation Objectives, and Outcome Measures

Considering the written evaluation plans themselves, interviews with IRS
officials, IRS's status report to Congress and other documents, we found
that IRS's plans for assessing the three tests had important strengths.
For instance, IRS's evaluation plans:

o 	had clear goals for each the three tests. The primary goal of all three
tests was to reduce overclaim rates. There were additional goals for the
qualifying child test-maintaining EITC participation for eligible
participants and minimizing taxpayer and IRS administrative burden.

o 	linked evaluation objectives and outcome measures-which determine the
extent to which the goals were met-to the test goals. For example, the
income misreporting test had outcome measures that included the percentage
of cases where an EITC claim was reduced or disallowed and the average
amount of the change. These measures were clearly linked to the test's
goal of reducing EITC overclaim rates.

o 	selected samples to provide information that could be generalized to
the EITC population being targeted. Both TIGTA and an outside consultant
reviewed the samples for the qualifying child test and found that the
23,000 sample for the general test was sufficient-a conclusion with which
we also agree. TIGTA also reviewed the samples for the income misreporting
test and found that it should provide reliable results.41

Lack of Detail and Documentation in Evaluation Plans Undermined Their
Value

IRS's evaluation plans for the 2004 tests lacked sufficient documented
detail for us to determine how IRS planned to conduct key aspects of the
evaluations. When we were able to determine how key aspects of the
evaluations would be conducted, we often did so based on interviews and
analyses of various documents. The general lack of detail and
documentation undermined the value of the plans by, for example, limiting
IRS's and stakeholders' ability to oversee the evaluations, identify and
address limitations in the evaluations, and ensure that limitations will
be clearly communicated when the results are disseminated.

IRS's written evaluation plans for the three tests were essentially
outlines that were not comprehensive, meaning that they did not fully
document all key aspects of the evaluation. For example, IRS's written
plans did not provide information on the sampling methodologies used in
all three tests. These were not articulated until IRS issued its August
2004 status report to Congress. In addition to the status report, which
also provided additional insights into the types of analyses IRS plans to
conduct, we relied on multiple other sources to gain a complete
understanding of IRS's planned evaluation activities. We interviewed IRS
officials and reviewed other information and documents they provided, such
as the contractor's report on the qualifying child test's sampling
methodology. According to IRS officials, the lack of comprehensive and
detailed written plans was due to

41 TIGTA 2004-40-100.

other priorities, such as undertaking the numerous steps needed to
implement the tests themselves.

While we recognize competing demands on the EITC program office, striking
a balance between documenting evaluation plans and implementing and
evaluating the tests helps ensure all parties understand the evaluations
and the managers and stakeholders are able to oversee implementation and
evaluations. Well-developed evaluation plans have a number of benefits,
perhaps most importantly, increasing the likelihood that evaluations will
yield methodically sound results, thereby supporting effective policy
decisions. Such plans help (1) ensure that the agency has addressed the
principal aspects of the evaluation, including the research design,
outcome measures, target and sample populations, data collection
activities, analyses, and dissemination of results, (2) officials monitor
changes to tests and assess the impact of those changes on the planned
evaluations, and (3) facilitate management and stakeholder review. Having
comprehensive and detailed evaluation plans helps ensure that all those
working directly on the evaluation have a common understanding of how data
will be collected, analyzed and impacts assessed. Concerns or weaknesses
can be identified and corrected, and plans can be updated to reflect any
changes during implementation and afterwards, as the evaluation plan could
be considered to be a "living document." Finally, a well-developed plan
helps ensure that evaluation results can be communicated with appropriate
recognition of the evaluation's strengths and limitations so stakeholders
can better understand how to use the results when making decisions.

The following are illustrations of the overall lack of detail and
documentation in IRS's evaluation plans for the 2004 tests.

o 	Evaluation objectives were not documented in one place. Although we
found that IRS's evaluation plans had objectives linked to the test goals,
the objectives were not identified in any single location for any of the
three tests nor specifically identified as objectives. Thus, we pieced
together the information from multiple sources, including interviews with
IRS officials. For example, we had difficulty identifying the evaluation
objectives pertaining to the use of the third-party locator service for
the filing status test.

o 	Key outcome measures lacked important detail. IRS's evaluation plans
lacked important information for all the key outcome measures, such as
their definition and purpose, formula/methodology, data source

and collection method. For example, IRS's evaluation plans for the
qualifying child test did not identify the specific data that would be
used to produce the outcome measure-the number of taxpayers who claim (or
do not claim) the EITC. IRS has provided this type of information about
its measures for other programs. For example, for its telephone and other
operations, IRS annually prepares a comprehensive document known as a data
dictionary, which includes items such as the definition and purpose of the
measure and its formula/methodology. IRS officials agreed that providing
such information in the evaluation plans could have been valuable in
managing the EITC tests. Without knowing details on outcome measures,
stakeholders do not have enough information about a measure to know
whether it is valid and reliable.

o 	Limited information was provided on planned analyses. The evaluation
plans also lacked specificity with regard to the key analyses planned and
what those analyses were intended to accomplish. For example, IRS
conducted a survey to obtain information about a taxpayer's experience
with the qualifying child test. IRS originally planned to survey these
taxpayers in April 2004. The survey was not conducted until September
2004, primarily due to delays in selecting a contractor and developing the
survey instrument. The 5-month delay may substantially reduce the number
of taxpayers who accurately remembered the actions they took and thus
affect the quality of the responses (i.e., recall bias). The accuracy of
individuals' survey responses declines the further away those responses
are from the date of the actual events. IRS and the contractor are aware
that such recall bias could exist and stated that they will consider it
when analyzing the survey results, but no detail was available on how they
would do so. This is critical because the potential utility of the survey
results could be in question.

The lack of detail in IRS's evaluation plans also increased the risk that
reports disseminating the results of the tests would not fully disclose
the evaluations' potential limitations. In its August status report to
Congress, IRS did not make clear that the qualifying child test results
could only be generalized to taxpayers IRS had reason to believe were most
likely to make an erroneous claim for the EITC when filing for the EITC in
2002. Absent such clarity, stakeholders might incorrectly assume that test
results apply to all taxpayers claiming qualifying children for the EITC.
Also, IRS did not describe potential limitations of the outcome measures,
specifically, how non-respondents would be accounted for in measure
calculations.

IRS officials recognize that their final 2005 report will need to include
information on the evaluation limitations, and expect to provide
sufficient detail and explanation of limitations in that report.

Evaluation Plans for 2005 Tests Not Completed Before Two of the Tests
Began

As of early December 2004, IRS had not completed evaluation plans for the
2005 testing, even though the qualifying child and income misreporting
tests began in November. According to IRS officials, they had not yet
completed an evaluation plan for the 2005 tests because final decisions
about the testing were still being deliberated in October. In their view,
it was less important to finish an evaluation plan for these tests by the
time testing began, because IRS could use the 2004 evaluation plans in the
interim. IRS officials acknowledge that evaluation plans are important and
have started to develop them for the 2005 tests.

IRS can build upon the 2004 evaluation plans for all three tests. However,
IRS made substantial changes for the qualifying child and filing status
tests, which would need to be taken into account in developing
comprehensive and detailed evaluation plans for the 2005 tests. Therefore,
while we recognize that there will be similarities with the 2004
evaluation plans, the importance of having evaluation plans in place as
testing begins or soon thereafter is heightened because of planned changes
to the test.

Conclusions	The EITC program lifts millions of low-income taxpayers and
their families out of poverty. However, its high rates of
noncompliance-overclaims for the credit-could potentially undermine the
credibility of the program because billions of dollars are annually paid
out that should not have been. IRS's three tests-qualifying child
certification, filing status, and income misreporting-are major
initiatives to reduce overclaims by addressing the leading errors
taxpayers make. Given the importance of the EITC to many low-income
households and concerns about high overclaims, these tests are being
closely watched by numerous stakeholders.

Although IRS has generally implemented each of the tests smoothly, it did
not fully document some key management decisions and other significant
events. Documentation supports a common understanding among staff about
the program they are administering-particularly one as complicated as the
EITC-and helps managers monitor whether a program is implemented as
planned. Having adequate documentation during the 2005 tests could help
foster a better understanding of the tests, ensure results

are accurately determined, and facilitate review and oversight. In
addition, while IRS told taxpayers selected for the qualifying child test
they could visit various physical locations for assistance, including
LITCs and IRS walk-in sites, IRS did not collect information from those
sites to determine the level and quality of services provided. Because
officials believe relatively few taxpayers used these sites, collecting
information from the sites may not have been practical. However, the
single city simulation of nationwide implementation may offer an
opportunity to gather some information on these services.

The evaluations that IRS is conducting of each test are likely to yield
some useful information and results that will help IRS officials and other
stakeholders judge whether and how to proceed with further implementation
of the new approaches to reducing EITC overclaims. Nevertheless, the lack
of detail and documentation in the evaluation plans impeded officials'
ability to manage the evaluations as well as external stakeholders'
ability to review and understand the evaluations' strengths and
limitations.

As of early December 2004, IRS had not completed its 2005 evaluation
plans, although testing was underway for the qualifying child and income
misreporting tests. A well-developed and timely plan would help IRS to
improve on the 2004 evaluation plans and take into account changes in the
tests themselves.

Recommendations for Executive Action

The Commissioner of Internal Revenue should

o 	adopt a policy of documenting the rationale for key policy decisions
and other significant events as the 2005 tests are implemented;

o 	develop a means of gathering information during the 2005 tests on the
use of such locations as LITCs and walk-in sites on the level and quality
of service provided by those sites, particularly in light of IRS's plans
to draw its sample from a single community for the qualifying child test;

o 	ensure that reports disseminating the results of the 2004 and 2005 test
evaluations clearly outline aspects of test design and evaluation
shortcomings that limit the interpretation and utility of the results; and

o 	complete the development of comprehensive and adequately detailed
evaluation plans for the 2005 tests.

These actions should be done as soon as possible, with any significant
changes to the evaluation plan appropriately documented as the evaluation
unfolds.

Agency Comments and Our Evaluation

In his December 22, 2004 letter, the Commissioner of Internal Revenue
agreed with our recommendations. Regarding the issue of documenting
significant policy decisions, he noted competing demands that can often
affect the quality of documentation, which we acknowledge in our report,
and that IRS has implemented a process to meet this recommendation. The
Commissioner noted that providing taxpayers with assistance is a top IRS
priority. As such, the Commissioner reported that IRS has plans to
identify the level and quality of services provided to taxpayers at LITCs
and walk-in sites in the single test community. Regarding dissemination of
results, the Commissioner reported that IRS is committed to ensuring all
aspects of the test design and evaluation will be clearly described to
stakeholders. Finally, the Commissioner reported that IRS intends to
complete the 2005 evaluation plans, in part, based on GAO's
recommendations about what a plan should contain. He also noted that IRS
may need to assess whether any modifications to the 2004 qualifying child
test criteria are appropriate in light of public events and community
leadership reaction in the single test community.

We are sending copies of this report to the Chairmen and Ranking Minority
Members of the Senate Committee on Finance and the House Committee on Ways
and Means. We are also sending copies to the Secretary of the Treasury;
the Commissioner of Internal Revenue; the Director, Office of Management
and Budget; and other interested parties. We will make copies available to
others on request. In addition, the report will be available at no charge
on the GAO Web site at http://www.gao.gov.

This report was prepared under the direction of Joanna Stamatiades,
Assistant Director. Other major contributors are acknowledged in

appendix IV. If you have any questions about this report, contact Ms.
Stamatiades at (404) 679-1900 or me on (202) 512-9110.

Michael Brostek Director, Strategic Issues

Appendix I

Scope and Methodology

For all three objectives, we reviewed and analyzed documents including
Treasury's EITC compliance study of 1999 tax returns; a joint IRS/Treasury
task force report; monthly status reports for each of the tests; draft and
final letters, forms, and notices for each of the tests; implementation
and evaluation plans; our prior reports; status results of the tests
reported by IRS and its contractors; and reports and EITC literature by
external stakeholders. We also interviewed Department of the Treasury and
IRS officials involved in the EITC tests, including the National EITC
Director, National Taxpayer Advocate, Director of Research Analysis and
Statistics, and other IRS officials involved with implementing the tests.
Additionally, we interviewed external stakeholders such as individuals at
the TIGTA, Center for Budget and Policy Priorities, and Urban Institute,
and reviewed and analyzed their reports.

We took steps to ensure that the data we received from IRS were reliable
for the purposes of this report and determined that they were. Some of
those steps included interviewing IRS officials knowledgeable about the
computer systems where the data we obtained came from and reviewing
documentation, such as system manuals and flowcharts. We identified and
assessed potential data limitations and compared those results to our data
reliability standards, noting no significant weaknesses. 1

In addition, to describe the three tests and determine how IRS was
spending the money appropriated it for the tests, we interviewed managers
and budget officials in the EITC Program office and reviewed and analyzed
IRS's fiscal year 2004 budget request and compared its planned to actual
EITC spending plan. Because IRS does not have an adequate cost accounting
system, we could not verify the accuracy of the figures IRS provided to
describe how funds appropriated for the tests were spent.

We identified attributes of sound program implementation based on reviews
of the social science literature, our prior work, and interviews conducted
with IRS research and program management officials and external
stakeholders, such as the Urban Institute.2 We tailored these

1 GAO, Assessing the Reliability of Computer-Processed Data, GAO-03-273G
(Washington, D.C: October 2002).

2 The three key sources we used to develop our implementation and
evaluation plan criteria were: P.H. Rossi, and H.E. Freeman. Evaluation: A
Systematic Approach, 4th ed. Newbury Park, Calif.: Sage, 1989, P.H. Rossi,
H.E. Freeman, and M.W. Lipsey. Evaluation: A Systematic Approach, 6th ed.
Thousand Oaks, Calif.: Sage, 1999, and GAO/PEMD-10.1.4.

Appendix I Scope and Methodology

attributes to apply them specifically to IRS's tests as shown in table 3.
Finally, to assess how well IRS implemented the tests and determine IRS's
planned refinements for further testing in fiscal year 2005, we reviewed
policies, procedures, and training documents; observed procedures and
operations in Kansas City, Missouri; Atlanta, Georgia; and Fresno,
California; and interviewed front line IRS managers and examiners in these
locations. We reviewed several case files for each test. Additionally, we
analyzed relevant interim reports prepared by IRS and its contractors; and
identified key results, and discussed them with IRS officials.

Table 3: Criteria for Assessing IRS's Test Implementation

Did IRS implement the test largely as planned?

Did IRS provide the necessary staff and training to accomplish the
workload associated with each test?

            Did IRS monitor the delivery of taxpayer correspondence?

Were taxpayers provided with a way to contact IRS or another entity if
they had questions and, if they did, did IRS assess the quality of the
interactions and the accuracy of the information provided?

Did IRS implement quality control procedures to ensure the quality of
decisions IRS staff made?

Source: GAO.

To assess whether IRS's plan for evaluating the tests contained sufficient
documented detail to facilitate managerial review and stakeholder
oversight, we used GAO guidance and the social science evaluation
literature to identify key attributes of an evaluation. These attributes
included the research design, outcome measures, target and sample
populations, data collection activities, analyses, and dissemination of
results. We obtained all available documentation on IRS' s evaluation
plans for each of the tests and reviewed that documentation to determine
whether we could understand from the documentation alone how IRS planned
to address the key attributes. Where we could not, we interviewed IRS
officials to further understand whether and how the officials planned to
address those key attributes. Written documentation should be complete,
facilitate tracing of events, and be readily available for examination to
foster a common understanding of the program and facilitate oversight. To
describe the status of IRS's evaluation plan for the fiscal year 2005
tests, we primarily relied on interviews with IRS officials.

Appendix II

Updated Results from the Income Misreporting Test as of September 30, 2004

In August 2004, IRS issued a status report to Congress, which was mandated
by the Consolidated Appropriations Act of 2004.1 The report presents an
overview of each of the three EITC tests, along with the design, status,
and preliminary findings as of June 2004. According to the EITC National
Director, the report contained some of the types of information that will
be needed to support future decisions about the full implementation of the
tests. Additionally, the EITC National Director noted that IRS also used
the status report to provide information on such items as the sampling
strategy that have been lacking in other documents.

IRS had updated results for the income misreporting test as of September
30, 2004. Updated results were not available for the qualifying child or
filing status tests. As IRS stated in its status report, which showed data
as of June 26, 2004, it is important to note that because the results are
interim, no conclusions should be drawn from the information provided and
no analyses about the impact of the tests were included.

As table 4 shows, IRS has screened all 300,000 tax returns for the income
misreporting test, and slightly more than half have been closed with
taxpayer agreement.

1Pub. L. No. 108-199, 118 Stat. 3 (2004).

Appendix II
Updated Results from the Income
Misreporting Test as of September 30, 2004

Table 4: Status of Income Misreporting Test, June 26, 2004 and September
30, 2004 Results

June 26, 2004 September 30, 2004 Number Percent Number Percent

          Total number of cases in AUR EITC test 300,000  100%  300,000  100% 
                          Number of notices sent 261,169   87%  261,188   87% 
                  Total number of cases closed a 102,545   34%  196,600   66% 
            Number of cases closed with taxpayer                         
                                     agreement b  66,981   22%  155,446   52% 
                  Number of cases screened out c  38,831   13%   38,812   13% 
         Number of cases closed with no change d  35,564   12%   41,154   14% 

Source: GAO analysis of IRS data.

a Total number of cases closed includes the number of cases closed with
taxpayer agreement and number cases closed with no change.

b Number of cases closed with taxpayer agreement includes cases when an
examiner determines a discrepancy in the case, and after sending a notice
to the taxpayer, an additional tax assessment is made and the taxpayer
agrees with this assessment.

c Number of cases screened out includes cases when an examiner reviews the
case and determines the computer mismatch (i.e., a math or other error)
can be resolved and thus closes the case without a change to the tax.
According to IRS officials, the number of cases declined slightly during
the period shown because some of the cases were transferred to
examination.

d Number of cases closed with no change includes cases when the taxpayer
is able to provide substantiation and thus the change is closed without a
change.

                                  Appendix III

                   Comments from the Internal Revenue Service

Appendix III Comments from the Internal Revenue Service

Appendix III Comments from the Internal Revenue Service

Appendix III Comments from the Internal Revenue Service

Appendix IV

                     GAO Contacts and Staff Acknowledgments

GAO Contacts	Michael Brostek (202) 512-9110 Joanna Stamatiades (404)
679-1900

Staff 	In addition to those named above, Tom Beall, Evan Gilman, Veronica
Mayhand, Susan Mak, Donna Miller, Libby Mixon, Chris Moriarity, Ed

Acknowledgments	Nannenhorn, Cheryl Peterson, Michael Rose, and Robyn
Trotter made key contributions to this report.

GAO's Mission	The Government Accountability Office, the audit, evaluation
and investigative arm of Congress, exists to support Congress in meeting
its constitutional responsibilities and to help improve the performance
and accountability of the federal government for the American people. GAO
examines the use of public funds; evaluates federal programs and policies;
and provides analyses, recommendations, and other assistance to help
Congress make informed oversight, policy, and funding decisions. GAO's
commitment to good government is reflected in its core values of
accountability, integrity, and reliability.

Obtaining Copies of The fastest and easiest way to obtain copies of GAO
documents at no cost

is through GAO's Web site (www.gao.gov). Each weekday, GAO postsGAO
Reports and newly released reports, testimony, and correspondence on its
Web site. To Testimony have GAO e-mail you a list of newly posted products
every afternoon, go to

www.gao.gov and select "Subscribe to Updates."

Order by Mail or Phone	The first copy of each printed report is free.
Additional copies are $2 each. A check or money order should be made out
to the Superintendent of Documents. GAO also accepts VISA and Mastercard.
Orders for 100 or more copies mailed to a single address are discounted 25
percent. Orders should be sent to:

U.S. Government Accountability Office 441 G Street NW, Room LM Washington,
D.C. 20548

To order by Phone:	Voice: (202) 512-6000 TDD: (202) 512-2537 Fax: (202)
512-6061

To Report Fraud, Contact:
Waste, and Abuse in Web site: www.gao.gov/fraudnet/fraudnet.htm

E-mail: [email protected] Programs Automated answering system: (800)
424-5454 or (202) 512-7470

Congressional	Gloria Jarmon, Managing Director, [email protected] (202)
512-4400 U.S. Government Accountability Office, 441 G Street NW, Room 7125

Relations Washington, D.C. 20548

Public Affairs	Susan Becker, Acting Manager, [email protected] (202)
512-4800 U.S. Government Accountability Office, 441 G Street NW, Room 7149
Washington, D.C. 20548

                               Presorted Standard
                              Postage & Fees Paid
                                      GAO
                                Permit No. GI00

United States
Government Accountability Office
Washington, D.C. 20548-0001

Official Business
Penalty for Private Use $300

Address Service Requested
*** End of document. ***