Tax Administration: IRS' Use of Random Selection in Choosing Tax Returns
for Audit (Letter Report, 02/06/98, GAO/GGD-98-40).

Pursuant to a congressional request, GAO reviewed the Internal Revenue
Service's (IRS) use of random audits during fiscal years 1994 through
1996, focusing on the: (1) number of audits selected overall and at
random for tax returns filed by all taxpayers and by IRS employees
across the nation and in Georgia; (2) profile of the taxpayers subjected
to random audits by state, type of taxpayer return, taxpayer income
level, and taxpayer occupation; (3) results of the random audits in
terms of the number of audits for which additional taxes were
recommended as well as the amount of these additional taxes and the
number of referrals to IRS' Criminal Investigation Division; (4) known
burdens imposed on taxpayers subjected to random audits; and (5)
alternatives that IRS might have used other than random selection to
meet its objectives.

GAO noted that: (1) between the fiscal years 1994 and 1996, the number
of audits nationwide increased from 1.4 million to 2.1 million; (2) the
increases were due to audits of taxpayers claiming the earned income
credit (EIC); (3) during fiscal years 1994 through 1996, the IRS did not
randomly select returns for audit from either the population of all
taxpayers or all returns; (4) IRS has about 40 audit sources, which are
programs and techniques used to select potentially noncompliant returns
for audit; (5) IRS audit sources do not rely on random selection from
the population of all returns but rather IRS selects returns having
characteristics indicative of potential noncompliance; (6) IRS did
identify six projects involving subpopulations of taxpayers with
indications of noncompliance from which taxpayers were randomly selected
for audit; (7) IRS chose six subpopulations for the six projects
nonrandomly on the basis of known or suspected high noncompliance rates
and other criteria, including geographic location or business size; (8)
the number of audits generated by random selection for these six
projects was small compared with the million or more audits done each
year; (9) IRS does not randomly audit its 100,000-plus employees; (10)
IRS treats its employees the same as other taxpayers for the purposes of
audit selection, with one exception: IRS has a special program for
auditing returns filed by specific types of employees; (11) this special
program has not used random selection; (12) although the IRS data show
the projects covered taxpayers in almost all states, 16 states had fewer
than 10 random audits, and 10 states had more than 100 such audits;
these 10 states, generally, had a higher number of audits because an IRS
field office for those states ran 1 of the 6 projects; (13) most audited
individuals in the six projects reported positive income below $25,000;
(14) audit results for the two projects with more than 200 audited
returns showed that the percentage of audits recommending additional
taxes was 46 percent for the EIC project and 80 percent for the eating
and drinking establishment project; (15) according to IRS, any audit
imposes some level of costs and burden on taxpayers; (16) IRS had no
alternative data sources that would accomplish the objectives of the six
projects other than random audits; and (17) IRS officials said they have
little incentive to randomly select taxpayers for audits because IRS'
regular audit programs generally find more noncompliance at lower costs.

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

 REPORTNUM:  GGD-98-40
     TITLE:  Tax Administration: IRS' Use of Random Selection in 
             Choosing Tax Returns for Audit
      DATE:  02/06/98
   SUBJECT:  Tax administration systems
             Tax return audits
             Tax nonpayment
             Taxpayers
             Statistical data
             Data collection
             Income taxes
IDENTIFIER:  Earned Income Tax Credit
             EIC
             
******************************************************************
** This file contains an ASCII representation of the text of a  **
** GAO report.  Delineations within the text indicating chapter **
** titles, headings, and bullets are preserved.  Major          **
** divisions and subdivisions of the text, such as Chapters,    **
** Sections, and Appendixes, are identified by double and       **
** single lines.  The numbers on the right end of these lines   **
** indicate the position of each of the subsections in the      **
** document outline.  These numbers do NOT correspond with the  **
** page numbers of the printed product.                         **
**                                                              **
** No attempt has been made to display graphic images, although **
** figure captions are reproduced.  Tables are included, but    **
** may not resemble those in the printed version.               **
**                                                              **
** Please see the PDF (Portable Document Format) file, when     **
** available, for a complete electronic file of the printed     **
** document's contents.                                         **
**                                                              **
** A printed copy of this report may be obtained from the GAO   **
** Document Distribution Center.  For further details, please   **
** send an e-mail message to:                                   **
**                                                              **
**                                            **
**                                                              **
** with the message 'info' in the body.                         **
******************************************************************


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


Report to the Honorable
Paul Coverdell, U.S.  Senate

February 1998

TAX ADMINISTRATION - IRS' USE OF
RANDOM SELECTION IN CHOOSING TAX
RETURNS FOR AUDIT

GAO/GGD-98-40

Random Audits

(268798)


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

  AIMS - Audit Information Management System
  CID - Criminal Investigations Division
  DIF - discriminant function
  DORA - District Office of Research and Analysis
  EIC - earned income credit
  IGP - Information Gathering Project
  IRS - Internal Revenue Service
  SOI - statistics of income
  SSN - Social Security number
  TIN - taxpayer identification number
  TCMP - Taxpayer Compliance Measurement Program

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


B-277015

February 5, 1998

The Honorable Paul Coverdell
United States Senate

Dear Senator Coverdell: 

In a letter dated March 25, 1997, you requested that we answer a
series of questions about the Internal Revenue Service's (IRS) use of
random selection in choosing tax returns for audit (also referred to
as "random audits").  On April 14, 1997, we briefed your office on
the preliminary results of our work, and in this report, we provide
more complete data. 

In responding to your questions on IRS' use of random audits during
fiscal years 1994 through 1996, our objectives were to provide
information on (1) the number of audits selected overall and at
random for tax returns filed by all taxpayers and by IRS employees
across the nation and in Georgia; (2) the profile of the taxpayers
subjected to random audits by state, type of taxpayer return,
taxpayer income level, and taxpayer occupation; (3) the results of
the random audits in terms of the number of audits for which
additional taxes were recommended as well as the amount of these
additional taxes and the number of referrals to IRS' Criminal
Investigations Division (CID); (4) the known burdens imposed on
taxpayers subjected to random audits; and (5) the alternatives that
IRS might have used other than random selection to meet its
objectives.\1


--------------------
\1 During our work, you also requested that we answer questions about
a form of IRS audit project known as Information Gathering Projects
(IGP).  For fiscal years 1992 through 1994, you asked us to determine
the number of IGPs, the IRS controls over their use, and the types
and results of IGPs in IRS' Georgia District.  We are preparing a
separate report, IRS Use of Information Gathering Projects,
(GAO/GGD-98-39, Feb.  5, 1998), to answer these questions. 


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

Between fiscal years 1994 and 1996, the number of audits nationwide
increased from 1.4 million to 2.1 million.  Audits done through
correspondence with taxpayers on single issues accounted for most of
the increase as the number of more complex, face-to-face audits with
taxpayers dropped slightly.  During this same time, the number of
audits done in Georgia also increased (from 45,451 to 55,446), and
these were also largely correspondence audits done at the service
center.  Overall, the increases were due to audits of taxpayers
claiming the earned income credit (EIC). 

During fiscal years 1994 through 1996, IRS did not randomly select
returns for audit from either the population of all taxpayers or all
returns.  IRS has about 40 audit sources, which are programs and
techniques used to select potentially noncompliant returns for audit. 
IRS' audit sources do not rely on random selection from the
population of all returns but rather IRS selects returns having
characteristics indicative of potential noncompliance.  IRS officials
did identify six projects involving subpopulations of taxpayers with
indications of noncompliance from which taxpayers were randomly
selected for audit.  Available IRS data show that no taxpayers
outside of these six subpopulations were selected at random for
audit. 

IRS chose the subpopulations for the six projects nonrandomly on the
basis of known or suspected high noncompliance rates and other
criteria, including geographic location or business size.  Three
projects studied the compliance of taxpayers (1) claiming the EIC
across the nation,\2 (2) claiming dependent exemptions duplicated on
other returns in Florida as well as other locations, and (3)
operating certain types of eating and drinking establishments in
Ohio.  The remaining three projects studied the compliance of
self-employed individuals who appeared to be (1) filing questionable
Schedule Cs in Illinois, (2) claiming false business losses to be
eligible to claim the EIC in Georgia, or (3) not paying
self-employment tax in Missouri.\3 Three of the six projects included
taxpayers from Georgia.  The number of taxpayers in these
subpopulations ranged from 2,348 to 15.1 million.  By comparison, the
population of all taxpayers was more than 100 million.  Appendix I
summarizes these projects. 

The number of audits generated by random selection for these six
projects was small compared with the million or more audits done each
year.  The six projects randomly selected a total of 7,421 taxpayers
from the six subpopulations.  As of July 1997, IRS had completed
audits of 2,961 of the returns filed by 2,629 of the 7,421
taxpayers.\4

IRS does not randomly audit its 100,000-plus employees; that is, IRS
does not randomly select employees for audit from the population of
IRS employees.  A few IRS employees were audited as part of the six
projects that used random selection because they were members of the
project subpopulation.  Of the 2,961 returns audited after random
selection from the project subpopulations, 4 involved IRS employees. 
According to IRS officials, IRS employees were unlikely to be part of
the subpopulations being studied, such as the self-employed.  IRS
officials also said IRS treats its employees the same as other
taxpayers for the purposes of audit selection, with one exception: 
IRS has a special program for auditing returns filed by specific
types of employees, such as new hires, executives, and employees in
sensitive positions.  This special program has not used random
selection. 

The profile of taxpayers selected randomly for audit in the six
projects reflected the location and the nature of the subpopulations
under study.  Although IRS data show the projects covered taxpayers
in almost all states, 16 states had fewer than 10 random audits, and
10 states had more than 100 such audits; these 10 states, generally,
had a higher number of audits because an IRS field office for those
states ran 1 of the 6 projects. 

Most audited individuals in the six projects reported positive income
below $25,000.  This is because 2,472, or 84 percent, of the audited
returns dealt with the EIC, which is designed to assist lower income
individuals.  The project involving the EIC studies was a response to
a congressional mandate to monitor and improve compliance with the
EIC, which IRS' regular audit programs were unlikely to address on a
large scale because of the lower incomes reported by EIC recipients. 

Audit results for the two projects with more than 200 audited returns
showed that the percentage of audits recommending additional taxes
was 46 percent for the EIC project and 80 percent for the eating and
drinking establishment project.  The average amount of additional tax
recommended per audit was $1,653 for the EIC project and $12,711 for
the eating and drinking establishment project. 

According to IRS, any audit imposes some level of costs and burdens
on taxpayers.  IRS has not measured these costs and burdens for any
type of audit.  As discussed in several of our testimonies,
accurately measuring taxpayer costs and burdens is difficult.\5 Even
so, IRS has recognized the importance of these measures and has been
making efforts to develop them. 

According to officials, IRS had no alternative data sources that
would accomplish the objectives of the six projects other than random
audits.  Further, they said compliance data that addressed these
objectives did not exist outside IRS.  We did not evaluate whether
the designs of the six projects were adequate to meet the objectives
or identify alternative data sources that might have addressed the
objectives.  However, beyond these six projects, our previous work as
well as research by others has pointed to a lack of tax compliance
data that IRS could use to guide its compliance efforts.  IRS
officials said they needed statistically valid data that can be used
to test ways to correct tax noncompliance and improve their audit
selection methodologies to better target noncompliant taxpayers for
audit.  Outside these research purposes, IRS officials said they have
little incentive to randomly select taxpayers for audit because IRS'
regular audit programs generally find more noncompliance at lower
costs. 


--------------------
\2 IRS' CID, together with the Examination Division, reviewed
taxpayers' apparently erroneous claims for the EIC, which could lead
to erroneous tax refunds.  At the conclusion of the CID review,
Examination closed the reviews as audits through IRS' service
centers. 

\3 The Form 1040 Schedule C is used to report profit or loss from a
business.  IRS describes a "questionable Schedule C" as one in which
the taxpayer claims unusual or excessive deductions that appear to be
incorrect. 

\4 The number of returns audited is greater than the number of
taxpayers because some audits led IRS to audit other returns filed by
the same taxpayer. 

\5 Taxpayer Rights and Burdens During Audits of Their Tax Returns
(GAO/T-GGD-97-186, Sept.  26, 1997); Tax System:  Issues in Tax
Compliance Burden (GAO/T-GGD-96-100, Apr.  3, 1996); and Tax System
Burden:  Tax Compliance Burden Faced by Business Taxpayers,
(GAO/T-GGD-95-42, Dec.  9, 1994). 


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


      IRS AUDITS OF TAX RETURNS
---------------------------------------------------------- Letter :2.1

IRS audits tax returns to check compliance in reporting income or
deductions and in other tax issues as well as in paying the correct
tax liability.  IRS auditors check taxpayers' documents in support of
data reported on tax returns.  Through IRS' 33 district offices,
auditors either visit the taxpayers to review the documentation or
ask taxpayers to bring it to the IRS office.  These district-based
audits often focus on two or more tax issues. 

Tax examiners in IRS' 10 service centers are to check taxpayers'
documentation through correspondence audits.  These audits usually
involve one tax issue, such as the number of tax exemptions claimed
on a return. 

Regardless of the type of audit, IRS auditors decide whether to
recommend that additional taxes be assessed.  If auditors recommend
additional tax assessments, the taxpayer can agree with or appeal the
change.  If the taxpayer wins the appeal, the additional taxes
recommended in the audit would not be assessed and collected. 

IRS has about 40 audit sources, which are programs and techniques
used to select potentially noncompliant returns for audit.  The major
source is the discriminant function (DIF) formula, a
computer-generated score designed to predict returns that, if
audited, would be most likely to result in additional taxes owed. 
The other sources prompting audits include (1) referrals from outside
or inside IRS, (2) information provided by a third party, (3)
indications of fraud or noncompliance from another audit, (4) actions
of tax return preparers, and (5) returns filed by IRS employees who
hold sensitive positions.  Service center processes also identify
potential noncompliance, such as apparently improper claims for the
EIC on tax returns. 

IRS has established procedures to better ensure that auditors promote
a fairer tax system by focusing on potentially noncompliant returns. 
IRS also has established nine audit standards to guide auditors'
behavior in areas such as probing for unreported income and
developing evidence of noncompliance.  After an audit, selected IRS
staff across the country are to review a small sample of audits
closed by district offices to measure auditors' adherence to these
standards. 


      WHY RETURNS MAY BE RANDOMLY
      SELECTED FOR AUDIT
---------------------------------------------------------- Letter :2.2

In its simplest form, random selection is a process by which all
members of a specific population or subpopulation have an equal
chance of being selected for study.\6 This process eliminates
personal biases and subjectivity from the selection process and
allows study results to be generalized to a larger group.  On this
basis, IRS could study compliance for a specific taxpayer
subpopulation by randomly selecting a sample of taxpayer returns from
that subpopulation for audit.  The results of these audits could be
projected to that entire subpopulation.  Indications of noncompliance
would not be considered in the random selection of returns for audit. 

Traditionally, the only IRS program using widespread random selection
has been the Taxpayer Compliance Measurement Program (TCMP).  Under
TCMP, IRS auditors did line-by-line audits of randomly selected tax
returns from large taxpayer populations such as individuals,
partnerships, or small corporations.  TCMP was IRS' program for
gathering comprehensive and reliable tax compliance data.  IRS used
the data for measuring compliance levels, estimating the tax gap,
identifying compliance issues, developing DIF formulas for
objectively selecting returns for audit, and allocating audit
resources.  In addition, Congress used TCMP data for policy analysis,
revenue estimating, and research.  IRS did the last TCMP for
individuals who filed returns for tax year 1988; these audits were
generally done during 1990, 1991, and 1992.  IRS had planned a TCMP
for tax year 1994 but postponed it indefinitely because of concerns
about the costs to IRS and the burdens on taxpayers. 

Outside TCMP, IRS' Examination Division initiates audit projects at
IRS district offices and service centers to improve the selection of
the audit workload.  Also IRS' Compliance Research function runs
research projects at IRS districts and service centers through
District Research Offices.  IRS occasionally does research projects,
such as the ongoing series of EIC studies, through other functions.\7
All of IRS' projects focus on measuring the extent and nature of
noncompliance for specific tax issues or specific groups of taxpayers
in which compliance problems have occurred.  These projects also tend
to focus on finding solutions to the compliance problem that use
nonenforcement means, such as taxpayer assistance, rather than
enforcement means, such as audits.  IRS officials said they may
choose to randomly select samples of tax returns in these projects
for research purposes. 


--------------------
\6 In more complex types of random selection, every member of a
specific subpopulation has a known chance of being selected, but
these chances may not be equal since selection depends on
statistically derived formulas and weights. 

\7 As of September 1997, IRS had finished an EIC study for filing
years 1994 and 1995 and was still completing the EIC study for filing
year 1996. 


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

You asked 12 questions about IRS audits and IRS' use of random
selection for audits closed during fiscal years 1994 through 1996,
which were

1.  The total number of IRS audits closed nationwide and in Georgia;

2.  The number of audits across the nation and in Georgia that were
randomly selected (i.e., random audits);

3.  The number of IRS closed audits of returns filed by IRS
employees;

4.  The number of audits of IRS employees that were randomly
selected;

5.  IRS' plans or efforts to measure taxpayers' burdens and costs
from being subjected to random audits;

6.  The number of nationwide random audits, by location of the
taxpayer;

7.  The number of nationwide random audits, by type of taxpayer;

8.  The number of nationwide random audits, by taxpayer income
levels;

9.  The number of nationwide random audits, by type of taxpayer
business;

10.  The number of nationwide random audits that resulted in
additional taxes recommended and the amounts of those additional
taxes;

11.  The number of nationwide random audits that resulted in
referrals for criminal investigation; and

12.  The alternatives other than random audits IRS might have used to
meet its objectives. 

To answer your questions, we defined random audit as any audit of a
taxpayer's return that was randomly selected, including returns
randomly selected from subpopulations that were nonrandomly targeted
because of suspected or known noncompliance.  We used the returns as
the basis of our analysis in order to make some comparisons to the
overall audit universe.  However, when IRS selects one taxpayer's
return for audit, through whatever means, it may also audit other
returns filed by the taxpayer in that or another tax year.  Although
the first selection may be random for the subpopulation, the
selection of additional returns may not.  In reporting the overall
number of random audits, we distinguished between the number of
taxpayers and number of returns. 

Within this context, we responded to your 12 questions through 5
objectives on IRS' use of random audits during fiscal years 1994
through 1996.  Our objectives were to provide information on (1) the
number of audits selected overall and at random for tax returns filed
by all taxpayers and by IRS employees across the nation and in
Georgia; (2) the profile of the taxpayers subjected to the random
audits by state, type of taxpayer return, taxpayer income level, and
taxpayer occupation; (3) the results of the random audits in terms of
number of audits for which additional taxes were recommended as well
as the amount of these additional taxes and the number of referrals
to IRS' CID; (4) the known burdens imposed on taxpayers subjected to
random audits; and (5) the alternatives other than random selection
that IRS might have used to meet its objectives. 

To identify the number of audits done nationwide and in Georgia as
well as of returns filed by IRS employees for fiscal years 1994 to
1996, we collected related data from IRS officials in the National
Office and Georgia District.  We also used data from IRS' database on
closed audits--the Audit Information Management System (AIMS)--for
each of these fiscal years.  We interviewed Examination Division
officials in the National and Georgia District offices to understand
these data as well as IRS' procedures for selecting and doing audits. 
We collected data on IRS procedures, including those on audits of IRS
employees. 

To identify the number of audits that used random selection, we first
reviewed IRS' audit selection procedures and interviewed Examination
Division and Research Division officials in the National Office to
discuss the types of audits that may use random selection.  These
types included audits done for IGPs in the Examination Division,
research projects in the District Office Research and Analysis (DORA)
unit, and EIC project audits.  IRS does not track the use of random
selection, so we asked officials to query the district offices and
DORAs about projects that may have used it.\8

We reviewed all the information IRS had available on the projects in
the National and Georgia District offices to see whether we could
identify other uses of random selection.  Some information was
unavailable, particularly for projects that started in earlier years. 
Examination officials explained that IRS is not required to maintain
these records.  They also said some records were discarded or lost in
consolidating from 63 to 33 district offices and in shifting
responsibilities for record maintenance during recent reorganization
efforts.  As a result, we are not sure whether all projects that used
random selection have been identified.  IRS officials said any
omission would be minor because IRS lacks the resources to use random
selection extensively. 

For one of the six projects that IRS identified as using random
selection, the EIC project, our analyses included the first two EIC
studies for filing years 1994 and 1995 but not the third, because IRS
was still checking the EIC and other claims for filing year 1996. 
IRS officials would not share the approximately 2,000 taxpayer
identification numbers (TIN) associated with the tax year 1995
returns that were randomly selected for the 1996 EIC study, because
IRS was still analyzing the audit results and wanted to prevent early
release of data on EIC noncompliance.  Our having the TINs and AIMS
data, however, would not have allowed us to accurately compute the
level of noncompliance.  IRS provided some summary data on 789
taxpayers whose returns were in the tax year 1995 sample and had been
audited, but we did not incorporate those data into our responses to
the questions because we could not analyze and verify the data to the
extent we could have done with the AIMS data.  Appendix II summarizes
the data provided by IRS. 

For any project that we or IRS identified as randomly selecting
returns for audit, we asked for the TIN on the audited returns.  We
used these TINs in two ways.  First, we matched them to data on IRS
employees in 1994, 1995, and 1996 to identify IRS employees subjected
to random audits.  Second, we matched the TINs to AIMS data to answer
the questions about the profile of the taxpayers (e.g., income level
and state location) and audit results (e.g., additional taxes
recommended and referrals to Criminal Investigation).\9

For the taxpayer profile and audit results, we reported the available
AIMS data on each random audit.  We did not test the reliability of
the AIMS data.  We did not report estimates that may be developed
from the data because we did not evaluate the statistical validity of
IRS' random selection; further, most of the random audits were not
finished.  For these reasons, in combination with the small number of
random audits compared with all audits, we did not attempt to draw
conclusions about the profile and results.  Also, we did not report
results of audits involving fewer than three taxpayers because of IRS
disclosure rules that protect taxpayer privacy. 

To identify IRS' efforts to measure taxpayers' costs and burdens, we
talked to Examination officials and collected data on IRS' plans and
surveys to define and measure taxpayer burden.  We reviewed IRS'
ongoing efforts for measuring taxpayer satisfaction with the audit
process through a survey.  We reviewed IRS' 1992 and 1996 survey
results for measuring taxpayer burden and attitudes of large
corporations.  We did not attempt to evaluate these surveys or
plans.\10

To identify whether IRS had any alternatives to random audits, we
first interviewed IRS officials and collected IRS data about the
objectives of each project using random selection.  Next, we reviewed
our prior reports as well as a Price Waterhouse report on compliance
data and the trade-offs with alternative sampling strategies.\11 We
interviewed IRS officials and researched the literature to see
whether we could find any other data sources that would meet IRS
objectives.  We did not attempt to evaluate the worthiness of IRS
objectives. 

Our work was done at IRS' National Office in Washington, D.C., the
Georgia District Office, Atlanta Service Center, and Southeast
Regional Office between March and November 1997 in accordance with
generally accepted government auditing standards.  We requested
comments from the IRS Commissioner on a draft of this report and
these comments are discussed at the end of this report. 


--------------------
\8 IRS officials explained that they do not track the use of random
selection because it is used rarely outside TCMP. 

\9 For audits done at service centers, AIMS data did not identify the
state in which the taxpayer resided.  As a result, we used zip code
information from AIMS to determine state location.  We merged this
state information with the state information from AIMS on random
audits at IRS districts. 

\10 Examination Customer Satisfaction Survey (June 22, 1997) and
Measuring Taxpayer Burden and Attitudes for Large Corporations:  1996
and 1992 Survey Results (Mar.  5, 1997). 

\11 Tax Administration:  Alternative Strategies to Obtain Compliance
Data (GAO/GGD-96-89, Apr.  26, 1996); Tax Research:  IRS Has Made
Progress but Major Challenges Remain (GAO/GGD-96-109, June 5, 1996);
and Price Waterhouse, Alternatives to the Taxpayer Compliance
Measurement Program (Feb.  28, 1997). 


   OVERALL AND RANDOM AUDITS DONE
   NATIONWIDE AND IN THE GEORGIA
   DISTRICT
------------------------------------------------------------ Letter :4

Between fiscal years 1994 and 1996, according to IRS data, audits
increased nationwide as well as in Georgia.  During this period, IRS
did not randomly select any taxpayers from the population of all
taxpayers for audit.  IRS did identify six subpopulations with known
or suspected noncompliance from which it randomly selected taxpayers
for audit.  Compared with the overall number of audits, IRS did very
few random audits during this period-- both across the nation and in
Georgia.  Similarly, very few of the audits of IRS employees involved
random selection. 


      TOTAL AUDITS DONE NATIONWIDE
      AND IN GEORGIA
---------------------------------------------------------- Letter :4.1

Table 1 shows that the total number of nationwide audits increased
from 1994 to 1996.  During that time, the number of audits done in
Georgia also increased (from 45,451 to 55,446).  Both of these
increases resulted from a change in emphasis in the types of audits. 
IRS increased the number of service center correspondence audits and
decreased the number of district office audits; correspondence audits
can be done more quickly than audits at a district office. 



                                Table 1
                
                Total Returns Audited Nationwide and in
                    Georgia, Fiscal Years 1994-1996

                                                Returns audited
                                          ----------------------------
Locality                                      1994      1995      1996
----------------------------------------  --------  --------  --------
Nationwide                                1,426,57  2,100,14  2,136,81
                                                 3         4         9
District Office                           1,009,16   969,365   948,425
                                                 3
Service Center                             417,410  1,130,77  1,188,39
                                                           9         4
Georgia                                     45,451    57,617    55,446
Georgia District\a                          37,866    38,220    28,430
Atlanta Service Center\b                     7,585    19,397    27,016
----------------------------------------------------------------------
\a Georgia District Office encompasses the state of Georgia. 

\b Only includes the taxpayers within the Atlanta Service Center who
are located in Georgia. 

Source:  IRS data for fiscal years 1994-1996. 


      RANDOM AUDITS CONDUCTED
      NATIONWIDE AND IN THE
      GEORGIA DISTRICT
---------------------------------------------------------- Letter :4.2

During the 3 years, IRS did not randomly select any taxpayers from
the population of all taxpayers for audit.  According to IRS data,
only those belonging to one of the six project subpopulations were
eligible for being randomly selected.  IRS chose these six
subpopulations nonrandomly on the basis of historically high
noncompliance rates or other evidence of suspected high noncompliance
rates.  These subpopulations represent small segments of the
population of taxpayers and include taxpayers in a specific
occupation, industry, geographic area, or economic activity; or with
specific characteristics, such as being EIC recipients. 

Table 2 shows the number of random audits across the six project
subpopulations for the nation and Georgia during fiscal years 1994
through 1996.  Compared with all its audits, IRS rarely did random
audits.  For the 3 years, IRS audited 2,961 returns across the
nation--including 157 in Georgia--in these six projects.  Most of
these audits took place at IRS service centers--2,524 nationwide and
133 in Georgia.  These IRS data reflect the number of returns audited
and not the number of taxpayers audited. 



                                Table 2
                
                   Returns Audited After Using Random
                 Selection From Project Subpopulations,
                Nationwide and in Georgia, Fiscal Years
                               1994-1996

Project                                      Nationwide        Georgia
----------------------------------------  -------------  -------------
EIC studies\a                                     2,472            130
Eating and drinking establishments                  247              0
Duplicate dependent SSNs                            162             14
Questionable Schedule Cs                              5              0
EIC Schedule-C loss                                  13             13
Self-employment tax                                  62              0
======================================================================
Total                                             2,961            157
----------------------------------------------------------------------
\a IRS had not completed its work on its filing year 1996 EIC study
and would not provide the tax year 1995 sample. 

Source:  IRS data for fiscal years 1994-1996. 

Of the six projects shown in table 2, three included taxpayers from
Georgia.  The EIC Schedule C project is managed by the Georgia
district; the duplicate dependent Social Security number (SSN)
project is managed by a neighboring district; and the EIC studies
project is national in scope.  The EIC studies accounted for the bulk
of the random audits both nationwide and in Georgia.  The other two
projects did not involve more than seven random audits of Georgia
taxpayers in any of the 3 years.  Additional Georgia taxpayers may be
subjected to random audits because the three projects were ongoing in
1997. 

As table 3 shows, the 6 projects included 7,421 taxpayers.  For
fiscal years 1994 through 1996, as of July 1997, IRS had audited only
2,629 taxpayers--largely in the filing year 1994 and 1995 EIC
studies.  During the audits of these 2,629 taxpayers, IRS audited
2,961 returns because some audits led IRS to audit additional returns
from the same taxpayer.  The other 4,792 taxpayers had not yet had
their audits completed because the projects had recently begun;
specifically, 3,702 of these taxpayers were selected for the
duplicate dependent SSN project, which started during 1996. 



                                     Table 3
                     
                      Taxpayers Randomly Selected for Audit,
                     Those Audited, and the Number of Returns
                      Audited by Project, Fiscal Years 1994-
                                       1996

                                   Taxpayers    Taxpayer
                    Subpopulatio    randomly  audits not   Taxpayers     Returns
Project                        n    selected  yet closed     audited     audited
------------------  ------------  ----------  ----------  ----------  ----------
EIC studies\a
--------------------------------------------------------------------------------
Filing year 1994       4,966,000       1,060         125         935       1,070
 study
Filing year 1995      15,100,000       2,048         674       1,374       1,402
 study
Eating and                24,000         118           7         111         247
 drinking
 establishments
Duplicate              3,200,000       3,835       3,702         133         162
 dependent SSNs
Questionable               2,348          12           7           5           5
 Schedule Cs
EIC Schedule-C             9,343         175         165          10          13
 loss
Self-employment           25,469         173         112          61          62
 tax
================================================================================
Total                         \b       7,421       4,792       2,629       2,961
--------------------------------------------------------------------------------
\a IRS had not completed its work on its filing year 1996 EIC study
and would not provide the tax year 1995 sample. 

\b Not applicable, because some taxpayers fit in more than one of the
subpopulations. 

Source:  IRS data for fiscal years 1994-1996. 

Table 3 also shows that IRS' use of random selection did not cover
the entire population of taxpayers.  The subpopulations in the 6
projects accounted for a small portion of the more than 100 million
tax returns filed annually.  The subpopulations ranged from 2,348 to
15.1 million in any 1 year.  IRS randomly selected the 7,421
taxpayers after nonrandomly selecting subpopulations of taxpayers
known or suspected to be noncompliant. 


      TOTAL AND RANDOM AUDITS OF
      IRS EMPLOYEES COMPARED WITH
      ALL TAXPAYERS
---------------------------------------------------------- Letter :4.3

Table 4 compares all audited taxpayers and IRS employees with returns
audited as a result of random selection.  It makes two basic points. 
First, the percentage of returns audited as a result of random
selection for both groups was very small compared with the overall
number of returns audited.  For the 3 fiscal years in total, the
percentage of audited returns subjected to random selection was
slightly higher for IRS employees than for all taxpayers (0.06
percent compared with 0.05 percent, respectively). 

Second, IRS audited four IRS employees' returns after random
selection from a subpopulation.  According to officials, IRS
employees were unlikely candidates for being selected randomly for
audit because they generally did not fall into the subpopulations. 
Four of the six projects involved businesses or self-employed
individuals.  As a general rule, few IRS employees would also be
self-employed.  Two of the projects involved the EIC, which is
designed to help the working poor.  Many full-time IRS employees
would be ineligible for the program. 



                                Table 4
                
                   All Returns Audited, IRS Employee
                   Returns Audited, and the Number of
                 Returns Audited as a Result of Random
                   Selection, Fiscal Years 1994-1996

Returns                             1994      1995      1996     Total
------------------------------  --------  --------  --------  ========
All taxpayers
Audited                         1,426,57  2,100,14  2,136,81  5,663,53
                                       3         4         9         6
Audited as a result of               714     1,101     1,146     2,961
 random selection
Randomly audited returns as a      0.05%     0.05%     0.05%     0.05%
 percentage of all audited
 returns

IRS employees
----------------------------------------------------------------------
Audited                            1,135     2,828     3,103     7,066
Audited as a result of                 4         0         0         4
 random selection
Randomly audited returns as a      0.35%        0%        0%     0.06%
 percentage of all IRS
 employees' audited returns
----------------------------------------------------------------------
Note:  The "Audited as a result of random selection" rows do not
include any returns from IRS' tax year 1995 sample for its filing
year 1996 EIC study. 

Source:  IRS data for fiscal years 1994-1996. 

To provide another perspective on these percentages, we analyzed the
overall audit rates for individual taxpayers and IRS employees for
fiscal years 1994 through 1996.  For all individual taxpayers, IRS
audited 1.08 percent, 1.67 percent, and 1.67 percent, respectively,
of the returns filed.  For IRS employees, IRS audited 1 percent, 2.6
percent, and 2.7 percent, respectively, of the roughly 110,000 IRS
employees during each of the 3 years.  Most of the IRS employees
audited were new hires, executives, candidates for executive
positions, and employees promoted into sensitive positions; the rest
were selected for audit just like any other taxpayer would be. 


   PROFILE OF TAXPAYERS SELECTED
   FOR THE SIX PROJECTS
------------------------------------------------------------ Letter :5

To profile the characteristics of taxpayers audited through the six
projects, we analyzed IRS data on the 2,961 audited returns.  The
characteristics included the taxpayer's state, type of return, income
level, and occupation or business.  Our results cannot be considered
definitive because our analysis only involved completed audits, and
most of the projects are ongoing with the audits of 4,405 taxpayers
still to be closed. 


      STATE LOCATION OF THE
      TAXPAYER
---------------------------------------------------------- Letter :5.1

For fiscal years 1994 to 1996, 16 states had fewer than 10 random
audits, and 10 states had more than 100 random audits.\12 Most of
these audits resulted from the EIC and Ohio's eating and drinking
establishment projects.  In future years, Florida, Georgia, and
Missouri will likely have a higher number of random audits because of
three ongoing projects:  (1) the duplicate dependent SSN project in
Florida; (2) the EIC Schedule-C loss project in Georgia; and (3) the
self-employment tax project in Missouri.  (See app.  III, table
III.1.)


--------------------
\12 The 10 states were Florida, Georgia, Illinois, Louisiana,
Mississippi, New York, North Carolina, California, Ohio, and Texas;
the last 3 states had between 230 and 301 random audits. 


      TYPE OF TAXPAYER RETURN
---------------------------------------------------------- Letter :5.2

According to IRS data, the 2,961 returns audited in the 6 projects
fell into 3 categories of tax returns--individual, corporate, and
employment tax returns--except in 1996, when fewer than 3 partnership
returns were audited.  Of these 2,961 returns, individual returns
accounted for 2,781 of the audits.  For example, in fiscal year 1994,
all 714 random audits involved individual returns.  (See app.  III,
table III.2.)


      TAXPAYER INCOME LEVELS
---------------------------------------------------------- Letter :5.3

Of the 2,961 returns audited in the 6 projects, 2,572 returns
reported positive income of less than $25,000.  The project on the
EIC, which is designed to help lower income individuals, accounted
for 2,417 of these returns.  This income level accounts for most
individual tax returns and most IRS audits overall (See app.  III,
table III.3.).  For fiscal year 1996, these lower income taxpayers
filed about 59 million of the 116 million returns.  Further, 1.1
million (most of them with EIC claims) of the 2 million returns
audited in fiscal year 1996 also involved taxpayers from this income
level. 

Few of the random audits involved higher income taxpayers, but this
is also true for audits overall.  However, IRS usually audits a
higher percentage of returns that report higher income.  For fiscal
year 1996, IRS audited almost 3 percent of the individual returns
reporting at least $100,000 in positive income but less than 2
percent of the returns reporting less than $25,000 in positive
income.  Before the recent influx of audits focusing on EIC claims,
the audit rates for such lower income returns were well below 1
percent. 

Random audits in the eating and drinking establishment project
included nonindividual returns, such as corporate, employment, and
partnership returns.  For this project, 40 percent of the audited
nonindividual returns were filed by corporations with total gross
receipts under $250,000.  For corporate returns alone, this category
represented 69 percent of the audits.  In 1994, IRS data showed it
did not audit any nonindividual returns as a result of random
selection.  (See app.  III, table III.4.)


      TAXPAYER OCCUPATION OR
      BUSINESS
---------------------------------------------------------- Letter :5.4

AIMS has limited information on the occupation or business of
individual taxpayers.  In fact, it does not record an individual's
occupation.  Instead, the database indicates whether a taxpayer filed
a Form 1040 Schedule C (income from an individual business) or Form
1040 Schedule F (income from farming) and may indicate the type of
Schedule C business (e.g., retail sales, services).  Of the cases we
reviewed, AIMS had very little data on the types of businesses; the
data that did exist varied greatly, suggesting no discernable
pattern.  However, one of the six projects identified the type of
business by the title--eating and drinking establishments. 


   AUDIT RESULTS FROM THE SIX
   PROJECTS
------------------------------------------------------------ Letter :6

We also analyzed data on the reported results of the 2,961 audits in
the projects.  These results included the percentage of returns with
recommended additions to reported taxes, the amount of additional
taxes recommended, and the number referred to CID.  These results
cannot be considered definitive because our analysis only involved
returns with completed audits, and most of the projects are ongoing
with over 4,000 taxpayer returns still to be completed. 


      PERCENTAGE OF RECOMMENDED
      ADDITIONAL TAXES
---------------------------------------------------------- Letter :6.1

For the 3 fiscal years, the percentage of completed audits that
recommended additional taxes to a return was 80 percent or higher for
the projects on eating and drinking establishments and
self-employment tax.  The percentage recommending additional taxes
for the project on EIC studies averaged 46 percent for the 3 years. 
Each year, the percentage for this project rose--from 12 percent in
1994, to 50 percent in 1995, and to 69 percent in 1996.  For the same
period, the nationwide average across all audited returns was 67
percent.  (See app.  III, table III.5.)


      AMOUNT OF ADDITIONAL TAXES
      RECOMMENDED
---------------------------------------------------------- Letter :6.2

The amount of reported additional taxes recommended from audits of
individual returns for all 3 years exceeded $200,000 in three of the
six projects.  The project on the EIC studies recommended about $1.9
million in additional tax; the eating and drinking establishment
project recommended about $712,000, and the duplicate dependent SSN
project recommended about $208,000.  The amounts recommended in the
other three projects fell below $20,000 during the same period. 

Of the 2 projects with more than 200 audited returns, the average
amount of additional taxes recommended per individual return audited
for the EIC project was $1,653; and for the eating and drinking
establishments project, the average amount of additional taxes
recommended was $12,711--double the national average of $6,251 for
fiscal years 1994 to 1996.  (See app.  III, table III.6.)


      REFERRALS FOR CRIMINAL
      INVESTIGATION
---------------------------------------------------------- Letter :6.3

None of the 2,961 audited returns during fiscal years 1994 to 1996
resulted in criminal referrals to IRS' CID or the Department of
Justice.  In fact, IRS auditors referred very few cases to CID for
criminal fraud.  In fiscal year 1996, auditors only referred 783 of
more than 2 million audits. 

CID relies on various sources of information for initiating its
investigations, including information from (1) within IRS, such as
from the Examination Division; (2) other government sources, such as
U.S.  Attorneys; (3) banks and other financial institutions; and (4)
the public.  In an effort to increase the quality of fraud referrals
from other IRS groups to CID, IRS established formal fraud-referral
procedures, effective for fiscal year 1996.  According to CID
officials, the objective of these procedures was to increase
coordination between CID and other IRS divisions, particularly
Examination. 


   BURDEN IMPOSED ON TAXPAYERS
   SELECTED FOR AUDIT IN THE SIX
   PROJECTS
------------------------------------------------------------ Letter :7

According to IRS, any audit--whether randomly selected or
otherwise--imposes some level of cost and burden on taxpayers. 
However, IRS has no system or data to measure the costs and burdens
associated with any of its audits.  IRS recognizes this situation
and, as a result, is trying to develop measures of taxpayer costs and
burden as well as data sources. 

In considering ways to define and measure the burdens and costs
imposed on taxpayers, IRS plans to include all contacts with
taxpayers--from telephone calls and correspondence to audits and
collection notices--in its measures.  With these measures, IRS plans
to capture savings from burden reduction initiatives, such as
increasing telephone assistor access at Customer Service sites.  IRS
plans to measure taxpayer burden by dollars, in order to compare the
savings and costs of burden reduction initiatives versus tax law
enforcement initiatives.  IRS also plans to develop alternative
methods for measuring taxpayer burden and satisfaction with all IRS
products and services. 

In the interim, IRS began its current survey for measuring taxpayer
satisfaction with the audit process in July 1997.  The survey is
based on a Price Waterhouse study done in 1991 and a related prior
survey conducted by Booz-Allen in 1989.  IRS decided to use the Price
Waterhouse study as a model for its current survey. 

The purpose of the survey is to provide IRS with information from a
small sample of taxpayers on (1) their level of satisfaction with
recent income tax audits, (2) their suggestions to increase the level
of satisfaction and improve the audit process, and (3) recurrent
problems and how IRS could correct them.  IRS plans to compare
taxpayers' perceptions of the quality and efficiency of the audit
process with IRS' assessments.  IRS plans to conduct the survey
through the mail for a 1-year span in four of its district offices. 
Results of the survey are to be available as early as late 1998;
afterward, IRS plans to make decisions about future surveys. 


   IRS ALTERNATIVES TO CONDUCTING
   RANDOM AUDITS
------------------------------------------------------------ Letter :8

For the six projects, IRS officials said that they could not have met
the research objectives through alternatives to random selection. 
Internally and externally, no statistically valid compliance data
addressed the objectives of these projects according to these
officials.  We did not independently evaluate the designs of the six
projects to determine if they would meet their objectives.  We have
reported on IRS' lack of statistically valid data outside the data
from TCMP or specialized research that not only measured taxpayer
compliance but also offered insights on the nature of and reasons for
tax noncompliance.\13

Price Waterhouse has made similar points about IRS' ongoing need for
statistical compliance data.\14

IRS officials said that without such compliance data, IRS has few
options to using random audits for compliance research purposes,
particularly for statistical precision, data quality, and data
collection cost considerations.  The officials said they need some
source of statistical compliance data to be able to project research
results to a larger subpopulation as a way to improve audit selection
methodologies to better target noncompliant taxpayers for audit.  To
the extent that the random selection is adequately designed and
properly done, it allows IRS to develop estimates of noncompliance
for an entire subpopulation without burdening each taxpayer within
that subpopulation. 

Outside these research purposes, IRS officials indicated that they
would have little incentive to randomly select returns for audit
because IRS wants to invest its limited audit resources productively. 
According to IRS officials, random audits usually generate less
additional recommended taxes per audit hour compared with audits
selected for ongoing programs.  IRS wants to target audit resources
on returns selected through the ongoing programs, which attempt to
focus on the most noncompliant taxpayers, rather than on returns
selected randomly. 


--------------------
\13 GAO/GGD-96-89 (Apr.  26, 1996) and GAO/GGD-96-109 (June 5, 1996). 

\14 Price Waterhouse (Feb.  28, 1997). 


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

In a letter dated December 23, 1997, IRS' Acting Chief Compliance
Officer commented on a draft of this report (see app.  IV).  He
expressed disagreement with our definition of random audit.  He said
random audit involves random selection for audit in which every
taxpayer in the filing population would have an equal chance of being
selected.  He also said the six projects discussed in our report are
not random audits.  He said they are projects where returns were
selected using a statistical random return selection technique from a
subpopulation of returns that were nonrandomly selected because of
suspected or known noncompliance. 

We believe our report clearly makes this same distinction between
random selection from the population of all taxpayers and random
selection from subpopulations picked because of suspected or known
noncompliance.  And as we noted in the draft report, for the period
reviewed, IRS did not randomly select for audit any taxpayer from the
population of all taxpayers.  As a result, we made no changes to the
report on the basis of IRS' comments. 


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

As we arranged with your office, unless you publicly announce its
contents earlier, we plan no further distribution of this report
until 30 days from its date of issue.  We will then send copies to
the Commissioner of Internal Revenue and Members of the Georgia
Congressional Delegation, and we will make copies available to others
upon request. 

Major contributors to this report are listed in appendix V.  Please
contact me on (202) 512-9110 if you or your staff have any questions
about this report. 

Sincerely yours,

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


DESCRIPTIONS OF PROJECTS INVOLVING
RANDOM SELECTION
=========================================================== Appendix I

Earned Income Credit (EIC) studies--A series of nationwide studies
conducted on 1993, 1994, and 1995 tax returns to provide broader
information on taxpayer understanding of and assess compliance with
EIC qualification requirements.  The audits done in these studies
were unusual in that both Criminal Investigation Division (CID) and
Examination were involved.  Because of this circumstance, some audits
of returns may have been recorded as closed on the Audit Information
Management System well after CID's initial contact with the
taxpayers. 

Eating and drinking establishments--An Information Gathering Project
begun in 1993 at Internal Revenue Service's (IRS) Cleveland District,
seeking to measure the accuracy of income reported by eating and
drinking establishments that had fewer than 25 locations and that
were licensed to sell alcoholic beverages in Ohio. 

Duplicate dependent Social Security numbers (SSN)--A research effort,
managed by IRS' North Florida District, examining those returns filed
in tax year 1995 where more than one taxpayer claimed the same
dependent (a duplicate dependent SSN).  IRS is testing the
effectiveness of notices in modifying taxpayer behavior for 1996,
securing amended returns for 1995, and learning more about the
subpopulation. 

Questionable Schedule Cs--A research effort, managed by IRS' Illinois
District, to determine the compliance of "questionable" wholesale and
retail sole proprietorships.  IRS drew two samples from tax year 1993
of Schedule Cs who claimed (1) zero gross receipts, zero other
income, and zero cost of goods sold, and (2) gross receipts of $350
or under, zero cost of goods sold, and total expenses greater than
$1,050. 

EIC Schedule-C losses--A research effort, managed by IRS' Georgia
District, attempting to determine if sole proprietorships might be
using losses from Schedule C to offset other income; such offsets
allow a taxpayer that would have been otherwise ineligible to qualify
for EIC.  IRS selected its sample from tax year 1994 returns
including a Schedule C and claiming EIC. 

Self-employment tax--A research effort, managed by the IRS'
Kansas-Missouri District, testing the impact of an educational letter
as a means of improving compliance for taxpayers filing a Schedule C,
Schedule F, and/or Other Income but not a Schedule SE with their
return.  The returns selected for testing were from tax year 1993. 


IRS SUMMARY DATA FOR IRS' STUDY OF
EIC CLAIMS FOR FILING YEAR 1996
========================================================== Appendix II

Because it was still drafting a report on EIC compliance during the
1996 filing season, IRS did not wish to share the tax year 1995
taxpayer sample.  IRS did provide summary data from the sample, which
follows. 



                               Table II.1
                
                  AIMS Summary Data by Audit Class for
                  Returns in IRS' Filing Year 1996 EIC
                                 Study

Audit class\a                                        Number of returns
--------------------------------------------------  ------------------
1040A, TPI under $25,000                                           706
Non-1040A, TPI under $25,000                                        57
TPI $25,000 under $50,000                                            4
Schedule C, TGR under $25,000                                       16
Schedule C, TGR over $25,000 under $100,000                         \b
Schedule C, TGR $100,000 and over                                   \b
Schedule F, TGR under $100,000                                      \b
----------------------------------------------------------------------
Legend:

1040A--U.S.  individual income tax return
TPI--Total positive income
TGR--Total gross receipts
Schedule C--Business income schedule
Schedule F--Farm income schedule

\a Type and class of return examined. 

\b In accordance with statistics of income (SOI) criteria, the data
have been deleted to avoid disclosure for specific taxpayers. 

Source:  IRS summary of 1996 EIC study data. 



                               Table II.2
                
                  AIMS Summary Data by Return Type for
                 Individual Returns in IRS' Filing Year
                             1996 EIC Study

Return type                                          Number of returns
--------------------------------------------------  ------------------
Nonbusiness                                                        767
Business\a                                                          22
----------------------------------------------------------------------
\a Form 1040A Schedule C or Schedule F filers. 

Source:  IRS summary of 1996 EIC study data. 



                               Table II.3
                
                AIMS Summary Data by State of Residence
                for Returns in IRS' Filing Year 1996 EIC
                                 Study

State                                              Number of taxpayers
----------------------------------------  ----------------------------
Alabama                                                             12
Alaska                                                               0
Arizona                                                             13
Arkansas                                                            12
California                                                          46
Colorado                                                             5
Connecticut                                                         45
Delaware                                                            \a
Florida                                                             44
Georgia                                                             14
Hawaii                                                              25
Idaho                                                                0
Illinois                                                            21
Indiana                                                              5
Iowa                                                                 4
Kansas                                                              10
Kentucky                                                             4
Louisiana                                                           25
Maine                                                                4
Maryland                                                            15
Massachusetts                                                        0
Michigan                                                             6
Minnesota                                                           11
Mississippi                                                         11
Missouri                                                            17
Montana                                                             \a
Nebraska                                                            \a
Nevada                                                               4
New Hampshire                                                        0
New Jersey                                                          14
New Mexico                                                          29
New York                                                            30
North Carolina                                                      28
North Dakota                                                        \a
Ohio                                                                35
Oklahoma                                                            14
Oregon                                                               6
Pennsylvania                                                        17
Rhode Island                                                         0
South Carolina                                                       8
South Dakota                                                         0
Tennessee                                                           22
Texas                                                              202
Utah                                                                 4
Vermont                                                              0
Virginia                                                             6
Washington                                                           6
West Virginia                                                        0
Wisconsin                                                            7
Wyoming                                                             \a
======================================================================
Total                                                              789
----------------------------------------------------------------------
\a In accordance with SOI criteria, the data have been deleted to
avoid disclosure for specific taxpayers.  However, deleted data are
included in the total. 

Source:  IRS summary of 1996 EIC study data. 


CHARACTERISTICS AND RESULTS OF
RANDOM AUDITS
========================================================= Appendix III



                              Table III.1
                
                 All Returns Audited and Number Audited
                   as a Result of Random Selection by
                     State, Fiscal Years 1994-1996

                                                Returns audited
                                          ----------------------------
                                                                Random
State                                               All      selection
----------------------------------------  -------------  -------------
Alabama                                          67,216             86
Alaska                                           17,660              0
Arizona                                          94,266             41
Arkansas                                         41,764             59
California                                    1,479,229            234
Colorado                                         77,195             23
Connecticut                                      65,110             21
Delaware                                         16,948              9
District of Columbia                             21,024             10
Florida                                         272,774            140
Georgia                                         158,187            157
Hawaii                                           23,092             \b
Idaho                                            25,332              8
Illinois                                        239,084            142
Indiana                                          88,712              8
Iowa                                             37,927             21
Kansas                                           40,809             23
Kentucky                                         40,525              7
Louisiana                                        85,554            152
Maine                                            19,626              9
Maryland                                        105,704             64
Massachusetts                                    96,787             39
Michigan                                        129,373             22
Minnesota                                        84,430             40
Mississippi                                      57,078            105
Missouri                                         77,213             67
Montana                                          14,338             \b
Nebraska                                         26,897              9
Nevada                                           61,230             13
New Hampshire                                    19,716              7
New Jersey                                      153,003             78
New Mexico                                       31,712             26
New York                                        366,323            158
North Carolina                                   97,187            136
North Dakota                                     12,719              4
Ohio                                            124,159            263
Oklahoma                                         67,270             49
Oregon                                           62,337             20
Pennsylvania                                    173,918             73
Rhode Island                                     22,498             11
South Carolina                                   60,220             62
South Dakota                                     11,518              5
Tennessee                                        66,782             95
Texas                                           453,770            302
Utah                                             29,245              5
Vermont                                           9,187             \b
Virginia                                        119,737             86
Washington                                      107,975             21
West Virginia                                    19,501             \b
Wisconsin                                        59,006             38
Wyoming                                           8,986              5
======================================================================
Total\a                                       5,641,853          2,961
----------------------------------------------------------------------
Note:  The "Random selection" column does not include any returns
from IRS' tax year 1995 sample for its filing year 1996 EIC study. 

\a Does not include returns from U.S.  territories or where the state
was unknown. 

\b In accordance with SOI criteria, the data have been deleted to
avoid disclosure for specific taxpayers.  However, deleted data are
included in the total. 

Source:  IRS AIMS data for fiscal years 1994-1996. 



                                   Table III.2
                     
                       All Returns and Number Audited as a
                     Result of Random Selection by Tax Return
                     Type, Fiscal Years 1994, 1995, and 1996

                                     Returns audited
          ----------------------------------------------------------------------
                   1994                    1995                    1996
          ----------------------  ----------------------  ----------------------
                          Random                  Random                  Random
Tax type         All   selection         All   selection         All   selection
--------  ----------  ----------  ----------  ----------  ----------  ----------
Individu   1,225,707         714   1,919,437       1,045   1,941,560       1,022
 al
Corporat      78,014           0      71,233          38      80,087          68
 e
Employme      62,189           0      53,978          18      56,181      {56}\b
 nt
Partners       8,077           0       7,072           0       7,636
 hip
Excise        33,493           0      29,521           0      31,579           0
Estate        11,077           0      11,419           0      11,794           0
Fiduciar       4,662           0       4,326           0       4,511           0
 y
Gift           1,853           0       1,893           0       1,934           0
Luxury         1,399           0       1,149           0       1,316           0
Other\a          102           0         116           0         221           0
================================================================================
Total      1,426,573         714   2,100,144       1,101   2,136,819       1,146
--------------------------------------------------------------------------------
Note:  The "Random selection" columns do not include any returns from
IRS' tax year 1995 sample for its filing year 1996 EIC study. 

\a Includes returns not falling into one of the nine major tax types
and any returns where the tax type is not specified in the AIMS
database. 

\b In accordance with SOI criteria, the data in adjoining cells have
been combined to avoid disclosure for specific taxpayers.  However,
data are included in appropriate totals. 

Source:  IRS AIMS data for fiscal years 1994-1996. 



                                   Table III.3
                     
                     All Individual Returns Audited and Those
                     Audited as a Result of Random Selection,
                     by Taxpayer Income Level and by Project,
                              Fiscal Years 1994-1996

                                   Individual returns
         -----------------------------------------------------------------------
                                                                    Schedule F
                 Form 1040/1040A          Schedule C (business)     (farming)
         -------------------------------  ----------------------  --------------
                     TPI     TPI                     TGR
                  $25,00  $50,00                  $25,00
                       0       0     TPI     TGR       0     TGR     TGR     TGR
             TPI   under   under  $100,0   under   under  $100,0   under  $100,0
           under  $50,00  $100,0  00 and  $25,00  $100,0  00 and  $100,0  00 and
Project  $25,000       0      00    over       0      00    over      00    over
-------  -------  ------  ------  ------  ------  ------  ------  ------  ------
Nationw  2,759,0  648,02  473,00  353,30  355,27  268,40  187,73  20,552  21,352
 ide          62       0       4       8       3       3       0
EIC        2,417      10       0       0  {27}\b      13       3   {4}\b       0
 studie
 s\a
Eating        17       7       8      13               9  {14}\b               0
 and
 drinki
 ng
 establ
 ishmen
 ts
Duplica      110      22  {13}\b       4      10   {5}\b               0       0
 te
 depend
 ent
 SSNs
Questio        0       4               0       0                       0       0
 nable
 Schedu
 le Cs
EIC            4       4  {13}\b       0       0                       0       0
 Schedu
 le-C
 loss
Self-         24      18               0       8       0       0       0       0
 employ
 ment
 tax
================================================================================
Total      2,572      65      34      17      45      27      17       4       0
--------------------------------------------------------------------------------
Legend:

TPI--Total Positive Income
TGR--Total Gross Receipts

\a IRS had not completed its work on its filing year 1996 EIC study
and would not provide the tax year 1995 sample. 

\b In accordance with SOI criteria, the data in adjoining cells have
been combined to avoid disclosure for specific taxpayers.  However,
data are included in the appropriate totals. 

Source:  IRS AIMS data for fiscal years 1994-1996. 



                                   Table III.4
                     
                       All Nonindividual Returns and Those
                     Audited as a Result of Random Selection
                      by Tax Return Type With a Breakdown by
                       Corporate Assets, by Project, Fiscal
                                 Years 1995-1996

                         Corporate returns
         --------------------------------------------------
         Form 1120S
         (nontaxabl
             e)                Form 1120
         ----------  -----------------------------
                                  $250                  All
         Unde  $200     No  Unde     k   $1M          other
            r     k  balan     r  unde  unde  $50M     1120  Partnersh  Employme
         $200   and     ce  $250     r     r   and  related         ip    nt tax
Project     k  over  sheet     k   $1M  $50M  over    forms    returns   returns
-------  ----  ----  -----  ----  ----  ----  ----  -------  ---------  --------
Nationw  16,6  20,6  6,835  28,3  20,7  40,7  14,1    3,226     14,708   110,159
 ide       03    49           57    35    30    85
Eating     21     7      8    52     9     9     0        0     {74}\a
 and
 drinki
 ng
 establ
 ishmen
 ts
--------------------------------------------------------------------------------
Note:  In fiscal year 1994, IRS did not audit any nonindividual
returns as a result of random selection. 

\a In accordance with SOI criteria, the data in adjoining cells have
been combined to avoid disclosure for specific taxpayers. 

Source:  IRS AIMS data for fiscal years 1995-1996



                              Table III.5
                
                Number and Percentage of Returns Audited
                   Overall and as a Result of Random
                  Selection That Had Additional Taxes
                  Recommended by Project, Fiscal Years
                               1994-1996

                                                 Audited
                                            returns with
                                   Audited   recommended
Projects                           returns         taxes    Percentage
----------------------------  ------------  ------------  ------------
Nationwide                       5,663,536     3,820,467           67%
EIC studies\a                        2,472         1,129           46%
Eating and drinking                    247           198           80%
 establishments
Duplicate dependent SSNs               162           100           62%
Questionable Schedule C                  5             0             0
EIC Schedule-C loss                     13             9           69%
Self-employment tax                     62            51           82%
======================================================================
Total                                2,961         1,487           50%
----------------------------------------------------------------------
\a IRS had not completed its work on its filing year 1996 EIC study
and would not provide the tax year 1995 sample. 

Source:  IRS AIMS data for fiscal years 1994-1996. 



                              Table III.6
                
                Additional Taxes Recommended per Audited
                  Individual Returns Overall, and per
                  Audited Individual Returns Resulting
                 From Random Selection, by Project for
                         Fiscal Years 1994-1996

                                              Amount of    Recommended
                                            recommended      taxes per
Project                                           taxes         return
----------------------------------------  -------------  -------------
Nationwide (individual returns)           $21,524,190,4         $6,251
                                                     86
EIC studies\a                                $1,866,252         $1,653
Eating and drinking establishments             $711,835        $12,711
Duplicate dependant SSNs                       $207,952         $2,080
Questionable Schedule Cs                            0\b            0\b
EIC Schedule-C loss                             $18,889         $2,099
Self-employment tax                             $17,495           $343
----------------------------------------------------------------------
\a IRS had not completed its work on its filing year 1996 EIC study
and would not provide the tax year 1995 sample. 

\b Of the five returns audited for this project, none resulted in any
recommended taxes. 

Note:  For nonindividual returns audited, such as corporate returns,
the recommended taxes per audited return may be skewed by several
large recommended assessments that exceeded the range for the
majority of the recommended assessments.  For example, the 1996
recommended taxes per return for nationwide nonindividual audited
returns, which included some high recommended assessments against
very large corporations, was $200,164 per return; for eating and
drinking establishments, it was $8,199. 

Source:  IRS AIMS data for fiscal years 1994-1996. 




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



(See figure in printed edition.)


MAJOR CONTRIBUTORS TO THIS REPORT
=========================================================== Appendix V

GENERAL GOVERNMENT DIVISION,
WASHINGTON D.C. 

Tom Short, Assistant Director
Elwood D.  White, Evaluator
Pat H.  McGuire, Senior Computer Specialist
Joanne M.  Parker, Computer Specialist
Susan F.  Baker, Computer Specialist
Elizabeth W.  Scullin, Communications Analyst

ATLANTA FIELD OFFICE

Michelle E.  Bowsky, Evaluator-in-Charge
H.  Dean Perkins, Evaluator

*** End of document. ***