oversight

Tax Administration: IRS Needs More Reliable Information on Enforcement Revenues

Published by the Government Accountability Office on 1990-06-20.

Below is a raw (and likely hideous) rendition of the original report. (PDF)

                  United   States   General   Accounting   Office
                  Report to the Ranking Minority
(iA0              Member, Committee on the Budget,                  ’.
                  U.S. Senate


June 1990
                  TAX
                  ADMINISTRATION
                  IRS Needs More
                  Reliable Information
                  on Enforcement
                  Revenues




                RESTRICTED--      Not to be released outside the
                General Acconnting Ofl’lce unless specifically
                approved by the Office of Congressional
                Relationa


GAO/GGD-90-85
-&!&II gig2iiLm3~
            t **
      General   Government   Division

      B-239247

      June 20,199O

      The Honorable Pete V. Domenici
      Ranking Minority Member
      Committee on the Budget
      United States Senate

      Dear Senator Domenici:

      This report responds to your request for our assessment of IRS’estimates of revenues
      generated by its various enforcement programs. More specifically, the report provides
      information on IRS’(1) overall enforcement revenue estimates, (2) budget estimates for
      collection and document matching programs, (3) Enforcement Resource Allocation Model,
      and (4) efforts to develop more reliable data on the results of IRS’ enforcement actions.

      We are sending copies of this report to the Secretary of the Treasury, the Commissioner of
      Internal Revenue, and the Director, Office of Management and Budget. As arranged with
      your office, unless you publicly announce its contents earlier, we plan no further distribution
      of this report until 30 days from the date of issuance. At that time, we will send copies to
      congressional committees interested in the matters discussed and to other interested parties.

      The major contributors to this report are listed in appendix III. Please contact me on 275-
      6407 if you or your staff have any questions.

      Sincerely yours,




      Jennie S. Stathis
      Director, Tax Policy and
        Administration Issues
__   .~

Executive Summary


             With continuing high federal budget deficits, can Congress and the
Purpose      administration look to the Internal Revenue Service (IRS) for greater rev-
             enues from its enforcement programs? If so, by how much?

             GAO  previously questioned IRS’ estimates of revenues produced by one
             enforcement program, the examination of tax returns. Senator
             Domenici, the Ranking Minority Member of the Senate Committee on the
             Budget, asked GAO to evaluate (1) US’ revenue estimates for two other
             enforcement programs- the collection of unpaid taxes and the matching
             of information on tax returns with data provided by third parties,
             (2) IRS’ estimates of total enforcement revenues, and (3) the use and use-
             fulness of a model IRS developed to allocate enforcement resources.


                 makes two types of revenue estimates-projections     for the year
Background   IRS
             ahead and estimates of actual results at the end of each year. It esti-
             mates overall enforcement revenues and the revenues it expects from
             proposed staffing increases. The President’s budget for fiscal year 1991,
             for example, proposes adding 3,600 staff to examine more tax returns,
             collect more unpaid taxes, and do other tax enforcement jobs. IRS esti-
             mated that with this additional staff, it would generate about $500 mil-
             lion of additional revenue in 1991 and about $6.5 billion by the end of
             fiscal year 1995.

             IRS’enforcement revenue estimates are used in several ways. The
             Department of the Treasury adds IRS’ revenue estimates for proposed
             staffing increases to its own revenue projections to arrive at total esti-
             mated federal tax receipts. The Office of Management and Budget and
             congressional budget and appropriations committees use IRS’ estimates
             in deliberating on proposed staffing increases and in determining the
             extent to which IRS’ enforcement programs can be relied on to help meet
             deficit reduction targets.

             IRS develops its estimates from information supplied by each enforce-
             ment program. Each program, such as collection of unpaid taxes, esti-
             mates revenues from its own activities using different data sources,
             different methodologies, and different assumptions. IRS must estimate
             the revenues actually generated by its enforcement programs because,
             as GAO has reported on several occasions since 1981, IRS does not have
             the information systems needed to develop adequate data on the results
             of its enforcement efforts.




             Page 2                                  GAO/GGWOJ35 IRS’ JXnforcement Revenues
                   Executive Summary




                   There is strong evidence to suggest that IRS’ after-the-fact estimates of
Results in Brief   the amounts collected from its total enforcement effort have been far
                   too high-on the order of $20 billion or about 40 percent off. There
                   have also been some large differences between the revenues IRS esti-
                   mated it would derive from IRS’ collection and document matching pro-
                   grams and the revenues it said it eventually realized.

                   Considering the inherent difficulties in estimating results many months
                   in advance and all the unanticipated events that can occur after an esti-
                   mate is developed, it is not unreasonable for estimates to deviate from
                   actual results. What troubles GAO is the absence of sufficient documenta-
                   tion to assess the reasons for the variances.

                   Even more troubling to GAO is IRS’ continuing inability to report what
                   revenues actually result from its enforcement programs. IRS recognized
                   the need for such information as early as 1976, and GAO has reported
                   that need in various reports and congressional testimonies over the
                   years. Nonetheless, IRS has had little success in developing the requisite
                   information systems. IRS did develop a model to help allocate additional
                   enforcement staff in a way that would maximize revenues. That model
                   is not being used as intended because of problems with the model’s input
                   data.

                   To ensure that IRS’ current efforts succeed, IRS and congressional com-
                   mittees need to exert strong oversight. In response to concerns raised by
                   GAO and others, IRS has several efforts planned or underway, including
                   development of an enforcement management information system, that
                   are directed at improving information on the results of IRS’ enforcement
                   efforts and developing better estimates of future enforcement revenues.

                   In the meantime, Congress is still faced with the need to make prudent
                   funding and staffing decisions for IRS’ enforcement programs. IRS now
                   audits only about 1 percent of all tax returns, and the inventory of
                   unpaid taxes IRS needs to collect exceeds $60 billion. Thus, it seems rea-
                   sonable to assume that adding IRS enforcement staff will generate addi-
                   tional revenues, especially over the longer term.

                   Until IRS develops more reliable information on the results of its enforce-
                   ment efforts, however, Congress should use the most conservative
                   assumption in its deliberations about additional revenues that can be
                   expected from additional enforcement staff. Congress also needs to
                   ensure, as it considers staffing increases, that IRS is being funded at
                   levels sufficient for it to adequately maintain current operations. As GAO


                    Page 3                                  GAOp3XMO-86   IRS’ Enforcement Revenues
                            testified in March, funding shortages in fiscal year 1990 caused IRS to
                            defer enforcement staffing increases that Congress had authorized, thus
                            negating expected revenue increases.



Principal Findings

Historical Perspective on   GAO reported in 1981 and subsequent years that IRS needed better infor-
                            mation on the costs and revenues associated with its enforcement pro-
Problems With IRS’          grams. As GAO noted in 1982, IRS recognized the need for an agencywide
Enforcement Revenue         management information system as early as 1976 but had made little
Data                        progress in developing such a system.

                            The situation today is essentially the same. IRS still does not know how
                            much revenue its enforcement programs actually generate. As the gov-
                            ernment’s tax collector, IRS has unparalleled responsibility to maintain
                            financial management systems that are second to none. In addition to
                            providing the foundation for improved revenue estimates, actual data
                            on case costs and revenues generated will also allow IRS to make more
                            informed decisions on allocating resources among enforcement pro-
                            grams. Further, such cost information is a necessary component of a
                            comprehensive cost accounting system-another         long-term goal of the
                            agency. (See pp. 42 and 43.)


Magnitude of IRS’ Total     Policymakers have generally thought the annual revenue impact of IRS’
                            enforcement programs was about $50 billion. IRS cited numbers of that
Enforcement Revenues Is     magnitude in testifying before Congress, and the Office of Management
Uncertain                   and Budget used those numbers in its budget analyses. In fact, however,
                            IRS does not know how much revenue its enforcement programs actually
                            generate. In recent years, IRS and the Department of the Treasury’s
                            Office of Tax Analysis explored better ways to calculate total enforce-
                            ment revenues. The results for fiscal year 1989 showed revenues of $30
                            billion. With such a wide variation, Congress and the administration can
                            have little confidence as to the overall revenue impact of IRS’ enforce-
                            ment programs. Various projects IRS has underway, especially the devel-
                            opment of an enforcement management information system, may
                            eventually rectify this situation. (See pp. 18 to 21.)




                            Page 4                                  GAO/GGD90-85 iRS Enforcement Revenues
                              Executive Summiuy




Insufficient Information to   Two of IRS’ major enforcement programs are the collection of unpaid
                              taxes and the matching of information on tax returns with information
Assess Revenue Estimates      on documents filed by third parties. For these two programs, GAO com-
for Two Enforcement           pared statistics on revenues generated with estimates of revenues
Programs                      expected when the year began. Those variances, some of which were
                              quite large, might be due to factors, like reduced staffing, that could not
                              be anticipated when the estimates were developed. Attempts to under-
                              stand reasons for the variances, however, are hampered by the uncer-
                              tainty about actual revenues and the absence of sufficient information
                              on how the revenue projections were made, including assumptions and
                              data sources. IRS needs to better document its estimating process and
                              compare estimates with results, identify reasons for differences, and
                              adjust its estimating methodology accordingly. (See pp. 22 to 35.)


Enforcement Resource          IRS developed the Enforcement Resource Allocation Model in 1986. As
Allocation Model Not Used     the name implies, the model was to help allocate additional staff among
                              enforcement programs in a way that would maximize revenues; it was
as Intended                   also to project the total costs and resulting revenues from a given
                              increase in staffing. The model is not used as intended because its prin-
                              cipal users (IRS’ Finance Division, the Treasury Department, and the
                              Office of Management and Budget) lack confidence in it.

                              One concern is that the model uses adjusted average, rather than mar-
                              ginal, revenue and cost data for the collection and document matching
                              programs. Another problem is its outdated assumptions about amounts
                              and timing of tax collections. The model assumes, for example, that 95
                              percent of tax assessments made by Examination’s auditors are actually
                              collected-an assumption based on a study of 1972 audit results that IRS
                              acknowledges is outdated and invalid. (See pp. 36 to 41.)


Intense Management            In response to these data problems and to improve its revenue estimates,
                              IRS plans an integrated enforcement management information system.
Attention Needed to           Several other related projects are also underway or planned. So that IRS’
Improve Enforcement           current effort succeeds where others have failed, IRS needs to develop a
Revenue Data                  strategy that shows how individual projects will contribute to reliable
                              enforcement program information and how they will be integrated to
                              provide improved revenue data. Together, the projects need to track the
                              costs and revenues associated with all kinds of enforcement cases from
                              start to finish but avoid unnecessary duplication or overlap. IRS needs to
                              closely monitor implementation of the projects. Given the significance of



                              Page 6                                  GAO/GGDso-85 IRS Enforcement Revenues
                    this issue, and IRS’ continued inability to develop the requisite informa-
                    tion, congressional oversight is also essential. (See pp. 42 to 47.)


                    The Budget and Appropriations Committees should monitor IRS’ pro-
Recommendationsto   gress in improving its enforcement revenue data information system.
Congressional       The Committees may want IRS to provide plans and milestones and peri-
committees          odic status reports. Until more reliable information is available, the
                    Committees should use the most conservative revenue assumptions in
                    their deliberations about additional revenues that can be realized by
                    providing additional enforcement staff for IRS. (See pp. 48 to 50.)


                    The Secretary of the Treasury should direct IRS to provide Congress
Recommendationsto   with information on the actual revenues generated by IRS’ enforcement
the Agency          programs as soon as it becomes available. In the interim, the Secretary
                    should direct IRS and the Office of Tax Analysis to identify the most
                    reliable methodology for estimating total actual enforcement revenues.
                    The Secretary should also direct IRS to explore ways to link the
                    improved revenue data with proposed staffing increases to provide Con-
                    gress with more reliable estimates of the revenue to be generated from
                    these increases. (See p. 21.)

                    GAO   is also making several recommendations to the Commissioner of
                    Internal Revenue directed at developing actual revenue and cost data
                    for IRS’ enforcement programs and using that data to improve the relia-
                    bility of its processes for estimating enforcement revenues. (See pp. 31,
                    35, and 47.)


                        agreed with many of GAO’S recommendations and said that it had
Agency Comments     IRS
                    already made changes to its revenue-estimating methodologies. Among
                    other things, IRS said that its fiscal year 1991 budget estimates are based
                    on more conservative and better documented methodologies than prior
                    years’ estimates and that the true issue in IRS’ budget is not whether IRS’
                    revenue-estimating procedures are good but whether IRS gets the overall
                    resources it needs to carry out its basic tax administration responsibili-
                    ties. IRS noted in that regard that significant budgetary shortfalls over
                    the past 2 years had kept it from maintaining the level of enforcement
                    effort it had originally intended. Chapter 8 contains a detailed discus-
                    sion of IRS’ comments, which incorporate the views of Treasury’s Office
                    of Tax Analysis, and GAO’S evaluation of those comments. (See pp. 51
                    and 52.)


                    Page 6                                   GAO/GGD90-86 IRS’ Enforcement F&wenues
Page 7   GAO/GGlS!3O-M IRS’ Enforcement Revenues
Contents


Executive Summary                                                                                          2

Chapter 1                                                                                                12
Introduction                   IRS’ Generation of Enforcement Revenues Involves
                                    Several Programs
                                                                                                         12

                               IRS’ Revenue-Estimating Process Involves Many Players                     15
                               Objectives, Scope, and Methodology                                        16

Chapter 2                                                                                                18
Magnitude of IRS’              Computations of Actual Enforcement Revenues Vary
                               IRS’ Information Indicates That Past Projections of Total
                                                                                                         18
                                                                                                         19
Total Enforcement                   Enforcement Revenues Have Far Exceeded Amounts
RevenuesIs Uncertain                Realized
                               Accuracy of Information on Total Enforcement Revenues                     20
                                    Affected by Uncertain Reliability of Component Data
                               Conclusions                                                               21
                               Recommendations to the Secretary of the Treasury                          21

Chapter 3                                                                                                22
Reliability of Revenue Insufficient  Information Available to Explain Variances
                            Between Estimated and “Actual” Collection
                                                                                                         22
Estimates From              Revenues
Collection Activities  Improved     Data Could Enhance Estimating Process                                28
                                                                                                         30
Needs to Be Improved Conclusions
                       Recommendations to the Commissioner of Internal                                   31
                                     Revenue

Chapter 4                                                                                                32
IRS Needs to Fully              Overview of the Process for Estimating Underreporter
                                    Revenues
                                                                                                         32
Document and Monitor            Inadequate Documentation to Assess Reasons for                           33
Its New Methodology                 Variances Between Estimated and “Actual”
                                    Underreporter Revenues
for Estimating IRP              Conclusions                                                              35
Underreporter                   Recommendations to the Commissioner of Internal                          35
Revenues                             Revenue




                                Page 8                                GAO/GGD90-85 IRS’ Enforcement Revenues
                       Contents




Chapter 5                                                                                      36
Data Limitations       ERAM Designed to Improve Revenue Estimating and                         36
                           Enhance Budgeting
Prevent IRS From       ERAM Not Being Used as Intended                                         37
Evaluating Relative    Conclusions                                                             40
Merits of Proposed
Increases in
Enforcement Staffing
Chapter 6                                                                                      42
IRS Needs to Ensure    Not Yet Clear Whether Planned Enforcement
                           Management Information System Will Meet IRS’
                                                                                               43
That Various Efforts       Needs
to Improve Revenue     IRS Needs to Ensure That Planned Improvements to the                    45
Data Are                   Revenue Estimation Process Are Effectively
                           Implemented
Comprehensive and      Conclusions                                                             46
Appropriately          Recommendations to the Commissioner of Internal                         47
                           Revenue
Coordinated
Chapter 7                                                                                      48
Reliability of         Recommendations to the Senate and House Budget and                      50
                           Appropriations Committees
Enforcement Revenue
Data Has Broad
Implications
Chapter 8                                                                                      51
Agency Comments and
Our Evaluation
Appendixes             Appendix I: IRS Projects Relating to Enforcement                        54
                           Revenue Estimates
                       Appendix II: Comments From the Internal Revenue                         57
                           Service
                       Appendix III: Major Contributors to This Report                         75




                       Page 9                              GAO/GGD-90-86 IRS’ Enforcement Revenues
Tables   Table 3.1: Comparison of Estimated and “Actual”                        23
             Accounts Receivable Revenues
         Table 3.2: Comparison of Estimated and “Actual”                        24
             Revenues Generated by Notices and TDAs
         Table 3.3: Comparison of Estimated and “Actual”                        24
             Revenues Generated by First Notices
         Table 3.4: Comparison of Estimated and “Actual”                        25
             Revenues Generated by Delinquent Returns
         Table 3.5: Comparison of Estimated and “Actual”                        26
             Revenues Generated by the IRP Nonfiler Program
         Table 3.6: Percent Differences Between Estimated and                   26
              “Actual” Results for Components of IRP Nonfiler
             Revenue Estimates
         Table 4.1: Comparison of Estimated and “Actual” IRP                    34
             Underreporter Revenues and Workload

Figure   Figure 1.1: Interrelationships Among IRS’ Major Revenue-               13
              Generating Enforcement Programs




         Abbreviations

         EMIS      Enforcement Management Information System
         ERAM      Enforcement Resource Allocation Model
         IRP       Information Returns Program
         IRS       Internal Revenue Service
         OMB       Office of Management and Budget
         OTA       Office of Tax Analysis
         TCMP      Taxpayer Compliance Measurement Program
         TDA       Taxpayer Delinquent Account


         Page 10                              GAO/GGD9&85 IRS’ Enforcement Revenues
Page 11   GAO/GGD-90-M IRS’ Enforcement Revenues
Chtq$~v- 1

Introduction


                       The Internal Revenue Service (IRS) generates revenues for the federal
                       government when it enforces existing tax laws. For example, it might
                       identify more taxes owed upon the examination of tax returns. Delibera-
                       tions on ways to reduce the federal budget deficit often raise the issue of
                       whether IRScould generate additional revenues by intensifying its
                       efforts. An increased interest in the revenue-generating capabilities of
                       IRS’ enforcement activities has brought with it an increased scrutiny of
                       IRS’ estimates of the revenues that have been and can be generated by
                       those activities.

                       In an August 1988 report to the Senate Budget Committee,l we identified
                       several flaws in IRS’ process for estimating the revenues from its Exami-
                       nation program. As a follow-on to that report, the Ranking Minority
                       Member of the Senate Budget Committee asked us to expand our inquiry
                       to IRS’ two other major revenue-generating activities-the    collection of
                       unpaid taxes and the matching of information on tax returns with infor-
                       mation submitted by third parties, such as employers and banks. The
                       Ranking Minority Member also asked us to inquire into IRS’ estimates of
                       total enforcement revenues and a model IRS developed for allocating
                       enforcement resources.


                       As shown in figure 1.1, IRS’ generation of enforcement revenues involves
IRS’ Generation of     several interrelated programs including Examination, Collection, the
Enforcement Revenues   Information Returns Program (IRP), and Appeals/Tax Litigation.’
Involves Several
Programs




                       ‘Tax Administration: Difficulties in Accurately Estimating Tax Examination Yield (GAO/
                            - - 119, Aug. 8, 1988).

                       ‘IRS has other enforcement programs, such as Criminal Investigation, which also generate revenues.
                       Those programs are not included in our discussion because, according to IRS, to the extent that their
                       activities produce enforcement revenues they would be included in the yield of other functions, such
                       as Examination.



                       Page 12                                               GAO/GGD90-85 IRS’ Enforcement Revenues
__-.
                                                Chapter 1
                                                introduction




Figure 1.l : Interrelationships   Among IRS’ Major Revenue-Generating         Enforcement Programs

                                    Filers of Tax Returns                                                                Nonfilers




                                                                                                information
                   Examination                                                               Returns Program




                                         Accounts Receivable:                                                   Delinquent Returns:
                                                                                                                     (Collection)


                                           Enforcement      Receipts (tax, penalties and interest)




                                                               Voluntary Receipts




                                                 Source Evaluation of the IRS System of Projectrng Enforcement Revenue (Department   of the Treasury,
                                                 Internal Revenue Serwce Research Dwislon, Publication 1501 (Nov. 1989).


                                                 Under the Examination program, IRS audits income, estate, gift, employ-
                                                 ment, and certain excise tax returns to determine whether taxpayers
                                                 correctly calculated their tax liabilities. In most cases, an examination




                                                 Page 13                                              GAO/GGMJO43S IRS’ Enforcement Revenues
Chapter 1
Introduction




results in the auditor recommending the assessment of additional tax.3
Taxpayers might do one of several things after learning of the auditor’s
recommendations. They might agree with the auditor and immediately
pay the additional tax, penalties, and interest due; they might agree but
not immediately pay, in which case the amount due would become an
account receivable to be handled by the Collection function; or they
might appeal the auditor’s findings through administrative channels in
IRS or through litigation, in which case the Appeals/Tax Litigation func-
tion would become involved.

The Collection program includes two major components-accounts
receivable and delinquent returns. Under the accounts receivable com-
ponent, IRS collects past due taxes and associated penalties and interest.
The amounts collected include dollars assessed by other enforcement
programs, such as Examination, as well as those due from taxpayers
who failed to pay the full tax due when they filed their returns. Under
the delinquent returns component, IRS takes enforcement action against
individuals and businesses that failed to file required tax returns. Those
nonfilers are generally identified through special compliance projects
operated by Collection or through documents matched under IRP.

Through IRP, a computerized document matching program, IRS matches
information documents, such as bank interest statements, with related
income tax returns to detect individuals who either (1) did not report all
their tax liability on filed returns (underreporters) or (2) did not file
returns at all (nonfilers). Within the underreporter component, certain
larger and more complex cases are forwarded to Examination for fur-
ther investigation. Under the nonfiler component, cases are forwarded
to Collection for assessment and collection of the amount owed.

Through the Appeals/Tax Litigation programs, IRS’ Chief Counsel’s
Office (1) provides an administrative appeals process for taxpayers who
dispute IRS’ audit findings or other enforcement actions and (2) repre-
sents IRS in cases litigated before the Tax Court. The amount assessed
after an appeal or litigation may range from nothing to the full amount

%3S uses different terms to measure revenue-recommended, assessed.and collected amounts. “Rec-
ommended” is the amount of additional tax and penalties IRS initially believes a taxpayer owes after
auditing the taxpayer’s return. “Assessed” is the amount of taxes and penalties IRS actually decides
is due and payable from the taxpayer. That amount often differs from the recommended amount
because of reductions resulting from taxpayers’ appeals of audit findings. ‘Collected” is the amount
IRS receives in payment of the assessedamount, including any penalties and interest. Because IRS is
not always able to collect owed taxes, the amount of additional taxes and penalties eventually col-
lected as a result of a particular audit may be less than the amount recommended. On the other hand,
the imposition of interest causes the amount eventually collected to increase.



Page 14                                              GAO/GGD90-85 IRS’ Enforcement Revenues
                        Chapter 1
                        Introduction




                        in dispute plus any interest and penalties. If the taxpayer does not pay
                        the assessed amount, the unresolved case is forwarded to Collection as
                        an account receivable.


                             process for estimating enforcement revenues starts with the indi-
IRS’ Revenue-           IRS’
                        vidual enforcement functions (Examination, Collection, etc.). Each func-
Estimating Process      tion uses different data sources, assumptions, and methodologies to
Involves Many Players   estimate the revenues to be derived from its activities. In our August
                        1988 report, we described the process then being used by IRSto develop
                        its estimates of examination yield.4 In chapters 3 and 4 of this report, we
                        discuss the processes used in developing revenue estimates for the
                        accounts receivable and delinquent returns components of Collection
                        and the underreporter and nonfiler components of IRP.

                        IRS’Finance Division consolidates each function’s estimates into an
                        overall estimate of enforcement revenues. The person responsible for
                        directly overseeing the preparation of enforcement revenue estimates is
                        the Budget Revenue Coordinator. IRSestablished that position in the
                        Finance Division in July 1988 to be more responsive to the Service’s rev-
                        enue estimation needs and those of Treasury, the Office of Management
                        and Budget (OMB), and Congress. Among other duties, the Coordinator is
                        to review estimates submitted by the individual functions and adjust
                        them to correct for such things as double counting.

                        The Budget Revenue Coordinator is also responsible for maintaining a
                        computer model, called the Enforcement Resource Allocation Model
                        (ERAM).IRS developed ERM to (1) project the costs and revenues associ-
                        ated with a given increase in enforcement staff and (2) determine how a
                        staffing increase would best be allocated among enforcement programs
                        to maximize revenue. We discuss ERAM’S use and usefulness in chapter 5.

                        As interest in IRS as a potential generator of additional revenues has
                        increased, the many users of enforcement revenue estimates have
                        become more sensitive to their reliability. One of the key users is Trea-
                        sury’s Office of Tax Analysis (OTA), which is responsible for projecting
                        total federal tax receipts and for estimating the revenue effects of tax
                        law changes and budgetary initiatives that are intended to raise reve-
                        nues through increased enforcement. In chapter 2, we discuss CITA’S

                        “To correct problems identified in our report, Examination has revised its estimating methodology.
                        We recently began assessing that methodology in response to a request from the Senate Committee on
                        the Budget.



                        Page 15                                             GAO/GGD90-85 IRS’ Enforcement Revenues
-
                          Chapter 1
                          introduction




                          recent involvement in trying to ensure the reliability of   IRS’   estimates of
                          total enforcement revenues.

                          Other key users of enforcement revenue estimates are OMB; the Congres-
                          sional Budget Office; and Congress, especially those committees directly
                          involved in the budget/appropriations   processes. The reliability of IRS’
                          enforcement revenue estimates is especially critical to these users
                          because they rely on the estimates to make decisions about the level of
                          IRSfunding and the revenues that can be expected from these programs.
                          They also rely on IRS’estimates to evaluate the relative merits of various
                          enforcement programs in generating additional revenues with increased
                          staffing.


                          As requested by the Ranking Minority Member, our objectives were to
Objectives, Scope,and     (1) assess the reliability of IRS’estimates of total enforcement revenues
Methodology               and the estimates of Collection and IRP revenues included in IRS’budget
                          submissions to Congress and (2) evaluate the use and usefulness of
                          ERAM.

                          To achieve these objectives, we

                          compared the various methodologies and data sources used by IRS and
                          u-r.4to calculate total enforcement revenues;
                          monitored IRSefforts to develop more reliable estimates of overall
                          enforcement revenues;
                          interviewed IRS National Office officials to obtain information on the
                          methodologies used to estimate Collection and IRP revenues, including
                          information on relevant models and databases;
                        . compared information in IRS’ budget submissions to Congress on (1) the
                          revenues IRS expected to realize and (2) the revenues it subsequently
                          determined it had realized;
                        . reviewed IRS documentation on how the revenue data in its budget sub-
                          missions to Congress were calculated, to the extent that such informa-
                          tion was available, and talked to appropriate IRSofficials about those
                          calculations;
                          identified IRS’ efforts to improve its enforcement revenue databases and
                          estimating methodologies and assessed its oversight of those efforts;
                          talked to officials from Treasury, OMB, and the Congressional Budget
                          Office about their reliance on IRS’ revenue estimates and estimation
                          models in preparing budget submissions;
                        . reviewed ERAM’S assumptions and data input; and




                          Page 16                                 GAO/GGD!M435 IRS’ Enforcement    Revenues
-
      Chapter 1
      Introduction




    . assessed ERAM'S use and usefulness by reviewing IRSrecords, internal
      studies, memoranda, and other supporting documentation and by inter-
      viewing officials in IRS,Treasury, and OMB about their use of ERAM.

      In evaluating IRS' estimates of Collection and IRP revenues, we compared
      the amount of revenue IRSsaid it actually received in a particular fiscal
      year with the amount it estimated it would receive as reported in its
      budget request for that year. The budget estimates we used in our com-
      parisons were the ones available to Congress when it deliberated on IRS'
      appropriations. Although IRSlater updated those estimates to reflect
      more recent workload information, we did not use those updates in our
      analyses because they were not available until after Congress had com-
      pleted its deliberations. To gain an understanding about trends in IRS’
      estimates, we focused our review on the period between fiscal years
      1983 and 1989.

      We did our audit work between October 1988 and February 1990 using
      generally accepted government auditing standards. The Commissioner
      of Internal Revenue provided written comments, which included input
      from CTA, on a draft of this report. The Commissioner’s major points are
      summarized and evaluated in chapter 8. The written comments and our
      additional analyses are included in appendix II.




      Page 17                                GAO/GGD-90436 IRS’ Enforcement Revenues
Magnitude of IRS’ Total Enforcement Revenues
Is Uncertain

                     The process of estimating future revenues begins with data on past per-
                     formance. IRS does not have the information systems it needs to tally
                     actual revenues generated through its enforcement programs. Absent
                     those systems, IRShas to estimate the actual results of IRS’enforcement
                     efforts. As discussed in chapter 6, IRShas several efforts underway,
                     including development of an enforcement management information
                     system, that are directed at improving enforcement revenue data.

                     There is much uncertainty about the level of revenues actually gener-
                     ated by these programs. But IRS’ recent recalculation suggests that
                     actual enforcement revenues were about $30 billion in 1989. This varies
                     significantly from the $50 billion range estimated for budget purposes in
                     recent years.


                     IRS and WA have used three different data sources in the past few years
Computations of      to calculate actual revenues from IRSenforcement programs. Their
Actual Enforcement   varying results create uncertainty as to the overall revenues generated
RevenuesVary         by those programs.

                     For fiscal year 1987, IRS estimated that it had collected a total of $45.9
                     billion in enforcement revenues. IRScalculated its total by consolidating
                     data supplied by the component enforcement programs-its          usual
                     method.

                     OTA   analyzed other data for fiscal year 1987 that led it to conclude that
                     total enforcement revenues that year might be closer to $24 billion. CVA
                     used an IRSreport (known as the “S-2 Report”) which shows allocations
                     of corporate and individual income tax revenues distributed by tax lia-
                     bility year. OTA arrived at its figure by assuming that the only portion of
                     total tax revenues shown on the S-2 Report that could be attributed to
                     IRS’enforcement activities were those stemming from tax liability years
                     at least 2 years earlier. (This was a reasonable assumption because of
                     the lag in beginning enforcement actions as well as the time it normally
                     takes to conclude an enforcement action.) IRS and OTA were unable to
                     reconcile the discrepancy between their respective numbers because of
                     the different data sources and assumptions used in compiling them.

                     In June 1988,~because of the differences in their numbers, IRS and CIIA
                     initiated a joint project to develop information on annual enforcement




                     Page 18                                  GAO/GGD-90-85 IlW Enforcement Revenuea
                       Chapter 2
                       Magnitude of IRS’ Total Enforcement
                       Revenues Ia Uncertain




                       revenues from IFS’ master files.’ To do this, IRSand OTA had to make cer-
                       tain assumptions as to which amounts in the master files reflected
                       actual enforcement revenues. In effect, they decided to define actual
                       enforcement revenues as payments received after an enforcement
                       action, such as an audit or nonfiler investigation. Using that assumption,
                       the master file data showed that IRS’enforcement programs generated
                       $29.6 billion in fiscal year 1989-about $5 billion more than the $24.6
                       billion derived for that year using data in the S-2 Report,

                       IRS considers the fiscal year 1989 master file analysis the best available
                       information on total annual enforcement revenues; but IRSis still
                       assessing the validity of the estimate. One validity check being consid-
                       ered is a further analysis of master file data to determine if fiscal year
                       1989 was an anomaly or if the data for other years also show revenues
                       significantly lower than the figures derived from IRS’enforcement
                       functions.


                           has, in past years, provided congressional and administrative deci-
IRS’ Information       IRS
                       sionmakers with information to indicate that it collects about $50 billion
Indicates That Past    a year as a result of its enforcement activities. IRS’recent recalculation
Projections of Total   of actual enforcement revenues for 1989 of $30 billion suggests that the
                       estimates have been far too high.
Enforcement Revenues
Have Far Exceeded      IRS’ estimates of total enforcement revenues have been provided to offi-
                       cials in Congress, OMB, and Treasury. In March 1987 testimony before a
Amounts Realized       House Appropriations Subcommittee, for example, the Commissioner of
                       Internal Revenue said that IRS’ proposed fiscal year 1988 budget would
                       generate $49 billion in enforcement revenues. An OMB official who over-
                       sees Treasury’s budget also told us that his analyses of IRS’ budget are
                       based on the assumption that IRS’enforcement programs generate about
                       $50 billion in revenues each year.

                       To support its budget submissions, IRS estimates projected revenues for
                       future years, broken down by type of enforcement. These estimates of
                       future enforcement revenues, like those of actual enforcement revenues,
                       are developed from data submitted by the component enforcement pro-
                       grams Each component program uses its own definitions, assumptions,
                       and methodologies for estimating revenues, IRS’ and OTA’S recent efforts


                       ’ IRS’ Individual and Business Master Files are databases that record information about taxpayers’
                       filing histories and their tax transactions.



                       Page 19                                               GAO/GGD99-86 lR!3’ Enforcement Revenues
                        Chapter 2
                        Magnitude of IRS’ Total Enforcement
                        Revenues Is Uncertain




                        to recompute actual enforcement revenues raise questions about the
                        validity of the projections.

                        In a March 1988 memo to the Secretary of the Treasury, OMB expressed
                        its interest in validating IRS’ enforcement revenue estimates. The memo
                        listed several significant policy decisions relating to IRS. One decision
                        was to review the methodology IRSused to estimate total enforcement
                        revenues and determine “whether the current enforcement receipt man-
                        agement goals of $50 billion for 1988 and $53 billion for 1989 are to be
                        revised.” In an April 1988 memo to IRS,Treasury’s Acting Assistant Sec-
                        retary for Management said that “(i)mproved estimation and quantifica-
                        tion of enforcement receipts is a high priority item for the Department
                        . . . and supports the Administration’s efforts to quantify the relation-
                        ship between funding levels and . . . revenue yields.”

                        IRSwas unable to determine why the calculation of actual enforcement
                        revenues derived from the master files ($29.6 billion) differed so much
                        from the projected estimate for fiscal year 1989 ($52.6 billion). Some of
                        the variance may be due to problems with the data submitted by the
                        various enforcement programs.


                        IRS’estimates of total enforcement revenues actually received and its
Accuracy of             projections of future revenues have been based on estimates developed
Information on Total    by the component enforcement programs. Each program uses different
Enforcement Revenues    data sources, assumptions, and methodologies. In combining these esti-
                        mates, the Finance Division has attempted to eliminate double counting
Affected by Uncertain   of revenues by the programs and has converted the revenue figures sub-
Reliability of          mitted by each program to a common measure-dollars      collected. Data
                        submitted on tax assessments have been translated into dollars collected
Component Data          using assumptions about the rate at which assessments will eventually
                        be collected and the timing of those collections.

                        Problems with the data supplied by the enforcement programs have
                        affected their reliability. In our August 1988 report, for example, we
                        questioned several key assumptions underlying Examination’s revenue
                        estimates. We found, among other things, that (1) IRS’ assumptions about
                        the number of audits completed per examination staff year varied sig-
                        nificantly from actual results, and (2) assumptions about the percent-
                        ages of recommended taxes that are actually assessed and the
                        percentages of assessed amounts eventually collected were based on an
                        outdated study that tracked 1972 audits.



                        Page 20                                GAO/GGDSO-M IRS’ Enforcement Revenues
-
                       Chapter 2
                       Magnitude of IRS’ Total Enforcement
                       Revenues   Is Uncertain




                       As discussed in more detail in chapters 3 and 4, data for Collection’s
                       accounts receivable and delinquent returns components and for the
                       under-reporter and nonfiler components of IRP have unexplained vari-
                       ances between budget estimates and estimates of revenues actually gen-
                       erated. Development of their estimates are not well documented; and
                       some estimates, such as those for accounts receivable, include amounts
                       that are not true enforcement revenues.


                       Congress and the administration need reliable information on the
Conclusions            revenue-generating capabilities of IRS’ enforcement programs. This
                       information has become even more critical in recent years as deci-
                       sionmakers have deliberated the prospects of increasing tax revenues by
                       expanding the size of IRS’ enforcement staff.

                       IRScannot provide decisionmakers with the actual amount of total reve-
                       nues generated by its enforcement programs because it does not have
                       the necessary information systems. As we will discuss in chapter 6, IRS
                       has several efforts under way, including development of an Enforce-
                       ment Management Information System, that may eventually enable IRS
                       to rectify that situation. Until then, IRS must estimate its total enforce-
                       ment revenues.

                       IRSmust also estimate for budget purposes the revenues to be expected
                       from a given level of staffing. Data available on the way the estimates
                       are developed and the wide variance between the estimates and IRS’
                       recent calculation of actual revenues give little assurance that Congress
                       can rely on those estimates.

                       IRSand OTA need to resolve the inconsistency between IRS’two estimating
                       techniques. If IRS considers the fiscal year 1989 master file analysis to be
                       the best available information on annual enforcement revenues, it needs
                       to develop a way to use that data for budget estimates.


                       We recommend that the Secretary of the Treasury direct IHSto provide
Recommendationsto      Congress with information on the actual revenues generated by IRS’
the Secretary of the   enforcement programs as soon as it becomes available. In the interim,
Treasury               the Secretary should direct IRS and CTA to identify the most reliable
                       methodology for estimating actual enforcement revenues. The Secretary
                       should also direct IRS to explore ways to link improved revenue data
                       with proposed staffing increases to ultimately provide Congress with
                       more reliable estimates of the revenue expected from those increases.


                       Page 2 1                                GAO/GGD9O-85 IRS’ Enforcement Revenues
Reliability of Revenue Estimates From
Collection Activities Needs to Be Improved

                        In recent years, IRS’ estimates of revenues to be generated by its Collec-
                        tion activities have often varied significantly from estimates of actual
                        revenues.’ These variances, in and of themselves, do not mean that IRS’
                        estimates are unreliable. There are several reasons why estimates may
                        differ from actuals. Some of those reasons (such as unanticipated hiring
                        freezes) are outside its control, while others (such as misused trend
                        information) are within its control and could warrant adjustments to its
                        estimating methodology.

                        We were unable to determine what reasons applied to the variances in
                        IRS’ Collection numbers and thus were unable to assess their reliability
                        because IRSdid not have sufficient documentation supporting its esti-
                        mates. Our evaluation of the variances in IRS’numbers was also ham-
                        pered by the fact that IRSestimates “actual” revenues, making it unclear
                        how much of the variance is due to erroneous estimates of future reve-
                        nues versus inaccurate estimates of actual revenues.

                        To enhance its process for estimating Collection revenues, IRS needs to
                        (1) document the process and the results therefrom, (2) correct the defi-
                        nition used in computing “actual” revenues to exclude amounts that are
                        not true enforcement revenues, and (3) develop data linking the cost of
                        processing Collection cases to the revenue generated-information    that
                        IRS needs to forecast the impact of additional staff in generating
                        revenues.


                            develops estimates of the revenue to be generated by Collection’s two
Insufficient            IRS
                        major components (accounts receivable and delinquent returns). In all
Information Available   btu one year since fiscal year 1983, the “actual” revenues from the
to Explain Variances    delinquent returns component varied from the estimate by at least 25
                        percent. Variances for the accounts receivable component were much
Between Estimated       lower, ranging from 3 to 30 percent. We were told of certain aspects of
and “Actual”            IRS’ revenue-estimating process that could help explain those differ-
                        ences. We were unable to determine any causal relationships, however,
Collection Revenues     because IRS did not adequately document its estimating process.


Accounts Receivable     The process for collecting accounts receivable involves three stages: cor-
                        respondence, telephone contact, and personal contact. The process

                        ‘&cause IFS does not have the kind of information necessary to identify the amount of revenue
                        actually generated by its Collection activities, it estimates those results. For ease of presentation, we
                        refer to those estimated actuals as “actual.”



                        Page 22                                                 GAO/GGD90-86 IRS’ Enforcement Revenues
                                         chapter3
                                         Reliability of Revenue Estimates From
                                         Collection Activities Needs to Be Improved




                                         begins with a series of up to five computer-generated notices to individ-
                                         uals owing past due taxes. If the notices prove unsuccessful and the
                                         amount due is above a certain level, IRS establishes a Taxpayer Delin-
                                         quent Account (TDA). Generally, IRS first attempts to resolve these TDAS
                                         through telephone contact with the taxpayer. If that fails, the TDA is
                                         sent to a district office where the case is scored on the basis of its poten-
                                         tial yield. Cases below a certain score remain in a holding file. The
                                         remaining cases are assigned to revenue officers, who can visit the tax-
                                         payer and take various actions, such as seizing physical assets, in an
                                         attempt to satisfy the taxpayer’s liability.

                                         As shown in table 3.1, “actual” revenues from this process varied from
                                         the estimate by more than 10 percent in 3 of the 5 fiscal years between
                                         1985 and 1989.2

Table 3.1: Comparison of Estimated and
“Actual” Accounts Receivable Revenues    In millions of dollars collected
                                                                                                                       Difference
                                                                                                                               Percent of
                                         Fiscal vear                          Estimate        “Actual”          Amount           estimate
                                         1985                                       17,738       18,252              514                        3
                                         1986                                       17,301       19,594            2,293                       13
                                         1987                                       17,492       22,766            5,274                       30
                                         1988                                       23,869       23,205              WV                          (3)
                                         1989                                       26,287       23,491            (2,796)                     (II)
                                         Source: GAO analysis of IRS budget data.


                                         Even in 1985, when the overall variance was small, comparisons of the
                                         components making up the overall estimate showed large variances. In
                                         that year, “actual” collections for the notice component varied from the
                                         estimate by 22 percent ($2.4 billion) while those for the TDAcomponent
                                         varied by 29 percent ($1.9 billion). As shown in table 3.2, large vari-
                                         ances in those components have not been uncommon over the past few
                                         years.

                                         Even in fiscal year 1988, when “actual” TDA collections were within 1
                                         percent of the estimate, the detailed elements of that estimate varied
                                         greatly from “actual” results. Collection’s records indicated that the
                                         number of TDAS closed in fiscal year 1988 was 20 percent lower than
                                         estimated, and the average yield per case closed was 25 percent higher.

                                         ‘Our analysis excluded fiscal years 1983 and 1984 because estimated and “actual”    budget data for
                                         those years were not comparable.




                                         Page 23                                              GA0/GGD90-36 IRS’ Enforcement Revenues
                                                          Chapter 3
                                                          Reliability of Revenue Estimates Prom
                                                          Cdkction Activities Needs to Be Improved




Table 3.2: Comparison of Estimated and “Actual”                     Revenues Generated by Notices and TDAs
In mllllons   of dollars collected
                                                         Notices’                                                               TDAs
                                                                           Difference                                                         Difference
                                                                                        Percent of                                                         Percent of
Fiscal year                 Estimate       “Actual”        Amount                        estimate      Estimate   “Actual”      Amount                      estimate
I 983                         $             $               fi                                           $5,528      $5,760        $232                             4
1984                                                                                                      5,934       5,159         (775)                         (13)
i 985                             11,174        13,580           2,406                          22        6,564       4,672       (1,892)                         (29)
1986                              11.692        14.524           2.832                          24        5,609       5,070         (539)                         (10)
1987                              12,765        16,151           3,386                          27        4,727       6,615        1,888                           40
1988                              16,849        16,220             (629)                        (4)       7,020       6,985            (35)                        (1)
1989                              17,733        16,443           (1,290)                        (7)       8,554       7,048       ( 1,506)                        (18)
                                                          aNotlce Information IS not shown for fiscal years 1963 and 1964 because the estimated and “actual’
                                                          figures for those years were not comparable.
                                                          Source: GAO analysis of IRS budget data.


                                                          According to IRS’ data, about 60 percent of the revenue from notices in
                                                          fiscal years 1985 through 1989 was generated by first notices. As shown
                                                          in table 3.3, although IRS’ estimates of first notice revenues varied from
                                                          “actual” results by at least 10 percent for all but one of those years, the
                                                          size of the variance has generally been decreasing.

Table 3.3: Comparison of Estimated and
“Actual” Revenues Generated by First                      In millions of dollars collected
Notices                                                                                                                                     Difference
                                                                                                                                                    Percent of
                                                          Fiscal year                                 Estimate      “Actual”         Amount           estimate
                                                          1985                                          $7,104         $9,190           $2,086                     29
                                                          1986                                           7,789          8,998              1,209                   16
                                                          I 987                                          8,603          9,832              1,229                   14
                                                          1988                                          10,444         10,001              (443)                    (4)
                                                          1989                                          10.795          9.662            (1.133)                  (10)

                                                          Source GAO analysis of IRS budget data




Delinquent Returns                                        The second major component of IRS’ Collection program involves delin-
                                                          quent returns. IRS’ first step after identifying an apparent nonfiler is to
                                                          send a series of notices. If the case is not resolved through the notices,
                                                          either by securing the delinquent return or obtaining a satisfactory
                                                          explanation as to why a return did not have to be filed, IRS creates a
                                                          delinquency case that is handled in the same manner as a TDA.



                                                          Page 24                                                   GAO/GGD9@85 IRS Enforcement Revenues
                                             Chapter 3
                                             Reliability of Revenue Estimates From
                                             Collection Activities Needs to Be Improved




                                             As shown in table 3.4, revenue estimates for delinquent returns varied
                                             from “actual” results by between 25 and 148 percent in all but 1 year
                                             between fiscal years 1983 and 1989. Excluding that year, the dollar
                                             magnitude of the discrepancies ranged from $1.2 billion to $3.0 billion.

Table 3.4: Comparison of Estimated and
“Actual” Revenues Generated by               In millions of dollars assessed
Delinquent Returns                                                                                                  Difference
                                                                                                                            Percent of
                                             Fiscal year                        Estimate       “Actual”      Amount           estimate
                                             1983                                     $1,856      $4,532      $2,676               144
                                             1984                                      2.059       5,105       3.046               148
                                             1985                                      4,512       5,726       1,214                27
                                             1986                                      4,466       7,308       2,842                64
                                             1987                                      5,326       8,033       2,707                51
                                             1988                                      8.206       7,809        (397)               (5)
                                             1989                                      8,993      11,232       2,239                25
                                             Source GAO analysts of IRS budget data


                                             The supporting documentation that was available on Collection’s esti-
                                             mates provides some insight into factors that contributed to the differ-
                                             ences between estimated and “actual” delinquent returns assessments.
                                             In fiscal year 1987, for example, when IRS’ estimate of assessments
                                             varied from “actual” results by 51 percent,

                                         l   the number of cases closed was 33 percent lower than projected;
                                         l   the number of delinquent returns secured per case was 75 percent
                                             higher; and
                                         l   the average yield per return secured was 27 percent higher.

                                             Collection’s documents also show that even in fiscal year 1988, when
                                             “actual” delinquent returns assessments were within 5 percent of the
                                             estimated amount, the detailed elements of the estimate varied signifi-
                                             cantly from “actual” results. In that year, 29 percent fewer cases were
                                             closed than estimated, 19 percent more staff years were used, and
                                             average yield per return secured was 14 percent higher.

                                             The delinquent returns revenues cited in table 3.4 include revenues gen-
                                             erated by the nonfiler component of IRP. Under that component, IRS iden-
                                             tifies individuals who have not filed required tax returns by comparing
                                             computerized data from information documents (such as wage and tax
                                             statements) submitted to IRSby employers and other third parties



                                             Page 25                                           GAO/GGD9045   IRS Enforcement Revenues
                                         Chapter 3
                                         Reliability of Revenue Estimates From
                                         Collection Activities Needs to Be Improved




                                         against the individuals’ tax return filing histories recorded in the master
                                         file.

                                         The methodology IRS uses to estimate the yield from delinquent returns
                                         is the same one it uses to estimate the yield from the nonfiler component
                                         of IRP. As shown in table 3.5, use of that methodology to estimate IRP
                                         nonfiler yield produced (1) large variances from “actual” results in
                                         fiscal years 1986, when IRS started budgeting separately for IRP, and
                                         1987; (2) a small variance in fiscal year 1988; and (3) almost no variance
                                         in fiscal year 1989.

Table 3.5: Comparison of Estimated and
“Actual” Revenues Generated by the       In millions of dollars assessed
IRP Nonfiler Program                                                                                                 Difference
                                                                                                                                  Percent of
                                         Fiscal year                         Estimate         “Actual”           Amount            estimate
                                         1986                                       $542         $1,091            $549                   101
                                         1987                                        997          1,953             956                    96
                                         1988                                      1,876          1,947              71                     4
                                         1989                                      2,112          2,111               (1)                       a

                                         aLess than l-percent difference.
                                         Source. GAO analysis of IRS budget data


                                         Using available supporting documentation, we analyzed factors that
                                         contributed to differences between estimated IRP nonfiler assessments
                                         and “actual” results. As shown in table 3.6, there were generally large
                                         differences between estimated and “actual” amounts for the various
                                         components of the revenue estimates, even in fiscal year 1989.

Table 3.6: Percent Differences Between
Estimated and “Actual” Results for                                                            Percent difference between estimated
Components of IRP Nonfiler Revenue                                                              and “actual” amount in fiscal year
Estimatesa                               Component of estimate                                       1987          1988          1989
                                         Number of cases disoosed                                         (34)            (52)            24
                                         Number of returns secured                                         31             (31)            (13)
                                         Average number of returns secured per
                                           dlspositlon                         ~-                         98                44            (30)
                                         Average yield per return secured                                 58                41             30
                                         Average number of cases disposed per staff
                                           year                                                           (42)            (54)             14
                                         Total staff years                                                 14               5               9
                                         aPercent figures In parentheses represent cases where the estimated amount was higher than the
                                         “actual” results
                                         Source GAO analysis of IRS data




                                         Page 26                                              GAO/GGD96435 IRS’ Enforcement Revenues
                              Chapter 3
                              ReLtabili~ of Revenue JMimatea From
                              CMlection Activities Needs to Be Improved




                              As discussed below, we were unable to further assess the differences
                              noted in tables 3.1 through 3.6 because IFSdid not adequately document
                              its estimating methodology, including any related trend analyses and
                              adjustments. Without such documentation, neither we nor IRS can deter-
                              mine why IRS’ estimates differ from “actual” results.


IRS’ Process for Estimating   There are various reasons why an estimate may differ from actual
Collection Revenues Not       results. Some of those reasons, such as hiring freezes that limit expected
                              staffing increases, cannot be anticipated and would not, if they occur,
Adequately Documented         invalidate IRS’estimating methodology. There are other reasons, how-
                              ever, that could reflect on the methodology. Variances between IRS’ esti-
                              mated and “actual” Collection revenues, for example, could indicate that
                              changes are needed in the way IRS analyzes and adjusts trend
                              information.

                              There are many factors (such as first notices issued, cases disposed,
                              returns secured, dollars assessed or collected, and staffing) for which IRS
                              develops historical trends to project future Collection workload and
                              results. According to the official responsible for developing the esti-
                              mates, the number of years of data reviewed for historical trends has
                              varied from year to year, and IRShas sometimes assigned different
                              weights to the various years included in the trend analysis. As part of
                              its methodology, IRSjudgmentally adjusts the results of its trend
                              analyses to account for such things as anticipated productivity enhance-
                              ments, program revisions, and tax law changes.

                              To adequately assess IRS’ methodology for estimating Collection reve-
                              nues, we would need to determine the extent to which the inconsisten-
                              cies and judgmental adjustments described above contributed to the
                              variances between estimates and “actuals” noted in tables 3.1 through
                              3.6. To do that, we would need to evaluate information on IRS’trend
                              analyses and its assumptions about staffing levels, staffing produc-
                              tivity, workload accomplished, and yield associated with the workload,
                              and include a comparison of those analyses and assumptions to “actual”
                              results. However, we were unable to make those kinds of evaluations
                              because IRS lacked sufficient supporting documentation for its estimates.
                              The official who develops the Collection estimates told us that IRShas
                              never documented the trend analyses.

                              Federal internal control standards require that agencies document
                              important transactions and processes, such as the basis for revenue esti-
                              mates. Without such documentation, management cannot adequately


                              Page 27                                     GAOPXXMO-85 IRS’ Enforcement Revenues
                      Chapter 3
                      Reliability of Revenue Estimates From
                      Collection Activities Needs to Be Improved




                      monitor its estimates over time, identify significant deviations from
                      actual, or assess the need for any methodological adjustments to correct
                      for those deviations. In its November 1989 report on enforcement rev-
                      enue estimation, IRS recognized the need for documentation:

                      “A universal complaint voiced by people outside IRS who use enforcement revenue
                      estimates is that documentation has been lacking-concerning     not only the methods
                      and the models in principle, but also concerning the assumptions that underlie spe-
                      cific revenue estimates.“:’

                      The only documentation we found in support of IRS’estimates of Collec-
                      tion revenues were some of the spreadsheets on which IRS recorded the
                      results of its adjusted trend analyses. These spreadsheets-which     are
                      developed for both accounts receivable and delinquent returns-pro-
                      vide details on the elements of the revenue estimates. IRS had discarded
                      the spreadsheets for all of the estimates that we reviewed except the
                       1988 and 1989 estimates for accounts receivable and delinquent returns
                      and the 1987 estimate for delinquent returns.? The official who is
                      responsible for developing the estimates said that he cannot reconstruct
                      the estimates for other years because the information is not available.


                      Although insufficient information was available to assess IRS’process
Improved Data Could   for estimating the revenues generated by its Collection activities, we
Enhance Estimating    identified two limitations with IRS’data that, if corrected, could enhance
Process               the results of that process. Those limitations involve (1) apparently
                      inaccurate data on “actual” revenues and (2) the absence of data linking
                      staffing and results.


Data on “Actual”      Accurate data on revenues realized in past years can facilitate the esti-
                      mation of future years’ revenues. The data in IRS’ budget submissions on
Revenues May Not Be   “actual” accounts receivable revenues come from an internal report-
Accurate              the Collection Yield Report. This report, which is derived from IRS’
                      master files, extracts payments and credits posted to a delinquent tax-
                      payer’s record.




                      %valuation of the IRS System of Projecting Enforcement Revenue, Department of the Treasury.
                      Internal Kevenue Service Kesearch Uwislon, Pubbcation lw,.       1989).

                      ‘As reported in this chapter, we used the information from the available spreadsheets to analyze
                      elements of the yield estimates for those years



                      Page 28                                              GAO/GGD-90-86 IRS’ Enforcement Revenues
                            Chapter 3
                            Reliability of Revenue Estimates F’rom
                            Cdkction Activities Needs to Be Improved




                            According to Research Division officials, the definition of revenue for
                            purposes of the Collection Yield Report includes amounts that do not
                            represent true enforcement revenues. For example, the Chief of the
                            Research Division’s Compliance Analysis Group said that the method-
                            ology used in producing the Yield Report (1) counts some overpayments
                            made by taxpayers and (2) includes payments that taxpayers made on
                            time but that IRSdid not correctly post until after their accounts became
                            delinquent. He also said that some IRS function other than Collection
                            may have actually secured the payment/credit that was counted as Col-
                            lection revenue because the master files do not contain sufficiently
                            detailed data to make that determination.

                            The IRS official responsible for estimating Collection revenue told us that
                            the Collection Yield Report counts as revenue overpayments or late-
                            posted payments because these transactions represent part of Collec-
                            tion’s workload, and Collection needs to quantify its workload for man-
                            agement purposes. He said that IFSmay need to develop a separate
                            report on revenues for preparing the estimates included in IRS’ budget
                            submissions while retaining the Collection Yield Report to quantify
                            workload.

                            IRS does not know the dollar impact of the definitional problems
                            described above. To identify that impact, the Research Division is devel-
                            oping an alternate version of the Collection Yield Report intended to cor-
                            rect those problems. Research officials expected the results to be
                            available in June 1990.


IRS Needs Data Linking      To accurately project future enforcement revenues, IRS needs to under-
                            stand to what extent the past efforts of its staff have resulted in pro-
co11.ection Case Staffing   ductive case outcomes. IRS is limited in its ability to do that, however,
and Results                 because it does not have a database that (1) reports the amount of staff
                            time spent on Collection cases and (2) links this to the results in terms of
                            revenue. As noted above, IRS uses master file data on payments and
                            credits that are posted to delinquent taxpayers’ accounts. The master
                            files do not contain information on the amount and type of staff
                            resources applied to those delinquent cases. Therefore, IRS cannot mea-
                            sure the impact that additional staff would have on case results. Also,
                            the master file does not identify whether Collection or some other
                            enforcement function, such as Examination, was responsible for
                            securing the payment or credit. Consequently, IRS must make assump-
                            tions about which master file cases generate revenues that should be



                            Page 29                                    GAO/GGD90435 IBS’ Enforcement Revenues
              Chapter 3
              Reliability   of Revenue Estimates Fkum
              Cdection      Activities Needs to Be Improved




              attributed to Collection-which                  reduces the certainty that those figures
              are accurate.

              IRS recognizes the need to develop a database on Collection’s case
              results. It believes that its proposed Enforcement Management Informa-
              tion System will eventually meet that need. IRS plans to incorporate Col-
              lection cases into the system by July 1991. Even then, however, as
              discussed in chapter 6, IRSmay still not have data on the staffing costs
              associated with working individual Collection cases.


              Although revenue estimation is not an exact science, federal internal
Conclusions   control standards and prudent management require that IRS be able to
              support its estimates. Revenue estimates should be based on a consistent
              methodology for evaluating the historical trends and anticipated future
              changes that will influence the estimates. Changes to a revenue-
              estimating methodology or its underlying assumptions should be based
              on an understanding of the causes of problems with the methodology or
              assumptions.

              We were unable to adequately assess IRS’ estimating process or deter-
              mine the reasons for differences between estimates and “actual”
              because IRS kept little documentation of the methodology used in pre-
              paring the estimates or the assumptions behind that methodology.
              Accordingly, Congress and the administration have little basis for
              relying on IRS’ estimating process to produce the information they need
              to assess the revenue-generating capabilities of IRS’Collection activities.

              IRSneeds to document its estimating process and track estimates against
              results to identify differences and determine whether the differences
              call for adjustments to the estimating methodology. To facilitate that
              tracking, IRSneeds to generate data on actual results. Actual data is crit-
              ical to the revenue-estimating process because it provides (1) a basis to
              assess the reasonableness of past estimates and (2) a foundation for
              making future estimates. Two key data elements for which IRS needs
              actual data are (1) the actual amount of revenue generated by its Collec-
              tion activities and (2) the amount of staff time spent on Collection cases.




               Page 30                                              GAO/GGDBCKl5 IRS’ Enforcement Revenues
                          Chapter 3
                          Reliability of Revenue Estimates Fkom
                          Collection Activities Needs to Be Improved




                          We recommend that IRS develop actual data on Collection revenues by
Recommendationsto         (1) defining revenues to be attributed to collection activities and
the cCmImk%iOner of       (2) developing a database that records for each Collection case the reve-
Internal Revenue          nues resulting from that case and the staffing applied to it.

                          To improve estimates of future revenues, we recommend that IFS

                      . develop and document a consistent methodology for reviewing historical
                        trends, incorporating anticipated program changes, and estimating reve-
                        nues in the Collection function;
                      . fully document the data used in applying the methodology, including the
                        trend analyses performed and the assumptions underlying those esti-
                        mates, and the results therefrom; and
                      l monitor estimates against actual results over time, using the most reli-
                        able data on actual results available, to assess the causes of any signifi-
                        cant discrepancies and to identify any adjustments needed in
                        assumptions or methodology.




                          Page 31                                      GAO/GGJHO435 IRS’ Enforcement Revenuea
IRS Needs to F’ully Document and Monitor Its
New Methodology for Estimating IRP
Underreporter Revenues
                             Since IRS started budgeting separately for the Information Returns Pro-
                             gram, estimates of the revenue generated by the underreporter part of
                             that program have varied from “actual” results by amounts as high as
                             41 percent of the estimate.’ We were unable to analyze possible causes
                             for the differences because IRS did not retain adequate documentation.

                             IRS is now using a new methodology that appears to provide a more
                             refined basis for developing underreporter estimates. At the time of our
                             review, however, IRS did not have documentation available on the
                             assumptions underlying its new methodology and was not monitoring
                             estimates against “actual” results to identify the possible need for
                             changes in methodology or assumptions.


                             The underreporter component of IRP identifies individuals who may
Overview of the              have misreported income by matching data on information documents
PrOCeSS for Estimating       (such as bank interest statements) to figures on individual tax returns.
Underreporter                IRS contacts taxpayers to verify facts and amounts in question before
                             assessing additional taxes due or refunding overpayments.
Revenues
                             To prepare its underreporter revenue estimate for fiscal year 1989, IRS
                             began using a computer model. To estimate revenues, the model extra-
                             polated past case results as follows:

                         l Cases were first sorted into 155 categories by combining two factors:
                           (1) type of income, such as wages, interest, and dividends; and
                           (2) potential amount of tax adjustment.
                         . Cases were further sorted into 1 of 15 processing activities. These activ-
                           ities represent a series of related technical and clerical steps through
                           which underreporter cases may flow as they are processed. One
                           activity, for example, includes the following steps: (1) case is screened
                           by tax examiner, (2) a notice is sent to taxpayer, (3) taxpayer responds
                           to the notice, and (4) case is closed out agreed.

                             In estimating fiscal year 1989 underreporter revenues, IRS estimated the
                             caseload for that year by reviewing historical trends and adjusting for
                             program changes. The model sorted this estimated caseload into the 155
                             categories on the basis of the percentage distribution of the most recent

                             ‘Because IRS does not have the kmd of information necessary to identify the amount of revenue
                             actually generated by its underreporter activities in a given fiscal year, it estimates those results For
                             ease of presentation. we refer to those estimated act&s as “actuals.” IRS has recently implemented
                             an IRP Management Information System. As described to us, that system is designed to track cases to
                             their conclusion, thus measuring actual dollars collected by fiscal year.



                             Page 32                                                 GAO/GGDSOSS IRS’ Enforcement Revenues
                    w-4
                    IRSNabtOFUIlY      Document and Mollit4Dr Ita
                    NeulU~for~lRP
                    underreporterBevenuea




                    actual case results. After establishing the estimated case volume in each
                    category, the model then calculated the flow of cases through each
                    activity by category. The volume of cases in each activity was then mul-
                    tiplied by the processing rat@ for that activity to calculate costs for the
                    category. The volume of cases in each activity was also multiplied by
                    the average assessment for each category to calculate yield (net of
                    refunds). Finally, the model calculated a yield/cost ratio for each case
                    category and ranked the categories in descending order on the basis of
                    their ratios.

                    Compared to the process used before fiscal year 1989, the computer
                    model provides a more refined basis for estimating underreporter rev-
                    enue. As noted above, in preparing the fiscal year 1989 estimate, the
                    model sorted cases into 166 categories on the basis of the type of income
                    and potential amount of tax @ustment. By contrast, the process used to
                    compute the fiscal year 1988 estimate analyzed case results into seven
                    categories on the basis of the potential amount of tax adjustment, and
                    the process used to compute estimates in earlier years analyzed cases in
                    even less detail.


                    As shown in table 4.1, the estimated underreporter revenues and
Inadequate          workload included in IRS’budget submissions to Congress for fiscal years
Documentation to    1986 through 1988 varied significantly from “actual” amounts subse-
AssessReasonsfor    quently reported by IRSfor those years. The cause of these discrepancies
                    is not apparent from the workload data in the budgets. Although the
Variances Between   budget data show that the “actual” number of underreporter notices
Estimated and       issued varied greatly from the estimated number, the amounts and, in 1
                    year, the direction of the variances do not appear consistent with the
‘ ‘Actual”          variances in revenues. For example, although “actual” under-reporter
Underreporter       assessments exceeded the estimated amount by 29 percent in fiscal year
Revenues             1986, the “actual” number of notices issued was lower than the esti-
                    mated amount by 23 percent.




                     “IRS defines the   processing
                                                rate as the number of staff   hours required to handle 1,000 cases.



                     Page 33                                                  GAO/GGD9O-85 IRS’ Enforcement Revenues
                                         chapter 4
                                         IRS Needs to Fblly Document and Monitor Its
                                         New Methodology for Estimating IRP
                                         Underreporter Revenues




Table 4.1: Comparison of Estimated and
“Actual” IRP Underreporter Revenues                                                                             Difference
and Workload                                                                                                            Percent of
                                         Fiscal year                     Estimate          “Actual”      Amount           estimate
                                         Assessments (in millions of dollars)
                                         1986                               n ,406            $1,808        $402                29
                                         1987                                 2,027            1,201        (826)              (41)
                                         1988                                 2,220            1,944        (276)              (12)

                                         Notices issued (in thousands)
                                         1986                                      4,161       3,200         (961)             (23)
                                         1987                                      5,500       2,200       (3,300)             w
                                         1988                                      5,000       3,800       (1,200)             (24)
                                         Source: GAO analysis of IRS budget data


                                         There are various reasons why estimated and “actual” revenues may
                                         vary, some of which are outside of IRS’control and cannot be anticipated
                                         when the estimate is being developed. According to IRS, for example, the
                                         41-percent variance in assessments for fiscal year 1987 was due, in part,
                                         to unanticipated computer problems that adversely affected the quality
                                         of its data. Because of inadequate documentation, however, we were
                                         unable to assess the extent to which the discrepancies between esti-
                                         mated and “actual” results for fiscal years 1986 through 1988 were due
                                         to factors that were within or outside IRS’ control.

                                         IRS did not retain any documentation  of the underreporter estimates sub-
                                         mitted to Congress for 1986 or 1987. Although documentation for the
                                         1988 estimates was retained, it does not support the congressional
                                         budget estimates, and IRS was unable to explain why. The official
                                         responsible for these estimates told us that he does not monitor “actual”
                                         results against budget estimates because the methodologies are continu-
                                         ally changing.

                                         As noted earlier, IRS used a computer model to develop its underreporter
                                         estimates for fiscal year 1989. The “actual” underreporter assessments
                                         for that year (about $2.1 billion) were within about 5 percent of the
                                         estimate (about $2.0 billion)-a    smaller variance than IRS had experi-
                                         enced using its previous methodology. IRS, however, was still unable to
                                         provide the kind of documentation needed to assess the reliability of
                                         estimates generated by the model and to evaluate the cause of any
                                         problems that may arise with future estimates. IRS did not document the
                                         analyses of historical trends or the judgmental adjustments that form
                                         the basis for the estimated caseload input to the model. Further, it did



                                         Page 34                                           GAO/GGD90-35 IRS’ Enforcement Revenues
                       Chapter 4
                       IRS Needs to Fully Document and Monitor Its
                       New Methodology for Estimating IRP
                       Underreporter Revenues




                        not retain any computer records showing the model’s analyses of past
                        case results, which constitute the major assumptions underlying the
                        future years’ revenue estimates. These analyses would show the per-
                        centage distributions for cases by category and activity along with the
                        data on average assessments, case processing times, and case staffing
                        costs.


Conclusions             mates provides a more refined basis for developing those estimates than
                        did the methodology it replaced. Because the model was first used to
                        produce fiscal year 1989 estimates, it is too soon to evaluate its useful-
                        ness. Any attempt to make such an evaluation in the future would be
                        hampered by the absence of adequate documentation. As discussed in
                        chapter 3, it is important that IRS adequately document and monitor its
                        revenue estimates. Without doing so, it will not have an appropriate
                        basis for identifying needed changes.

                        An evaluation would also be hampered by the absence of data on actual
                        IRP underreporter revenues. IRS has implemented an IRP Management
                        Information System, which, as described to us, will generate actual rev-
                        enue data on IRP cases. We did not independently assess the adequacy of
                        that system.


                        We recommend that
Recommendationsto                                IRS

the Commissioner of   . fully document its IRP underreporter revenue estimates, including docu-
Internal Revenue        mentation of any analyses used to estimate future revenues and all
                        other assumptions underlying the estimates;
                      . monitor “actual” underreporter results against estimated amounts to
                        identify the causes of discrepancies and any needed changes in the
                        methodology or assumptions used to estimate revenues; and
                      . use actual information, once it is available, to monitor the reliability of
                        IRP underreporter estimates.




                         Page 36                                     GAO/GGD90-36 IRS’ Enforcement Revenues
Data Limitations Prevent IRS From Evaluating
Relative Merits of Proposed Increases in
Enforcement Staffing
                    Deliberations on proposed IRS staffing increases would be enhanced if IRS
                    had reliable information on the relative yields of its enforcement pro-
                    grams in comparison to their costs. IRS developed the Enforcement
                    Resource Allocation Model (EFaM) in an attempt to (1) determine how
                    additional enforcement staff could be allocated among enforcement pro-
                    grams in order to maximize revenue and (2) estimate the total costs and
                    total revenue flows for a given increase in enforcement staffing. Several
                    limitations with ERAM’Sinput data, however, have affected the model’s
                    usefulness. As a result, ERAM is not being used as intended.


                    In the past, IRS’ estimates of the revenue yield to be derived from
ERAM Designed to    staffing increases had several problems. First, projected revenues from
Improve Revenue     each enforcement function were not additive because (1) each enforce-
Estimating and      ment function expressed revenue in different terms (that is, dollars rec-
                    ommended, assessed, or collected); and (2) some of the accounts
Enhance Budgeting   receivable revenue was double-counted when both Collection and Exam-
                    ination took credit for some of the same cases. Second, the estimates
                    were incomplete because they omitted costs incurred by other func-
                    tions-such as Appeals/Tax Litigation -in support of, or as a follow-on
                    to, a staffing increase. Finally, the multiyear effect of a staffing
                    increase, in terms of both revenues and costs, was not addressed; only
                    the budget year was estimated.

                    In 1982, we discussed IRS’ need for more complete cost and revenue data
                    on its enforcement programs1 We recommended that IRS develop addi-
                    tional revenue data so that resources could be better allocated among all
                    enforcement programs. Additionally, we recommended that to improve
                    IRS’resource allocations and overall management of IRS’ enforcement
                    resources, the Commissioner implement a system to provide cost and
                    revenue data from a total program and agencywide perspective.

                    Treasury and OMES  had expressed similar concerns about IRS’ revenue
                    estimates. In response to those concerns, IRS developed the Program-
                    matic Budgeting Model in 1984. That model was designed to express all
                    enforcement revenues in a single term (dollars collected), eliminate
                    double-counting of accounts receivable revenues, and include support
                    and follow-on costs in the cost estimates. IRS developed ERAM in 1986 to
                    further improve its estimates of the costs, workloads, and revenues
                    associated with proposed changes in enforcement staff. In addition to

                    ‘Ruther Research Into Noncompliance Is Needed to Reduce Growiqj Tax bsaes (GAO/GGD-J32-34,
                    Julr a19w.
                               Chapter 5
                               Data Limitations Prevent IRS From
                               Evaluating Relative Merib of Proposed
                               Increases in Enforcement Staffing




                               incorporating all of the Programmatic Budgeting Model’s features, ERAM
                               was designed to compute a staffing increase’s revenue effect over 5
                               years and to allow the user to allocate resources across enforcement
                               programs to maximize revenue.’

                               IRSdocuments portray ERAM as a useful tool for providing information to
                               administration and congressional officials on IRS’budget decisionmaking
                               process and the revenue-generating capabilities of its enforcement pro-
                               grams For example, a May 1988 memorandum from IRS’Deputy Com-
                               missioner for Planning and Resources to Treasury’s Acting Assistant
                               Secretary for Management states that “the Service currently uses the
                               Enforcement Resource Allocation Model (ERAM) for estimating revenue
                               expected from enforcement programs, and for making calculations of
                               total program costs.” Also, in its November 1989 report on enforcement
                               revenue estimation, IRS describes ERAM as a model that can estimate the
                               total yield and cost consequences of a given staffing increase (or
                               decrease) and optimize resources across enforcement programs on a
                               marginal yield-to-cost basis.”


                               Although IRS designed EPA4 to estimate the revenue associated with a
ERAM Not Being Used            staffing increase and to optimally allocate additional staff to maximize
as Intended                    revenue, it does not use the model for those purposes. Also, according to
                               OMB and Treasury Department officials, neither of those agencies-the
                               two primary customers for whom the model was designed-uses ERAM.
                               In each case, ERAM’S nonuse was attributed to concerns about the model’s
                               reliability-concerns  that can be traced primarily to limitations with the
                               model’s input data.


Limitations With ERAM’s        ERAM’S input data (such as adjusted average yield and revenue flow
                               rates) are developed by each of the enforcement functions using varying
Input Data                     information sources, methodologies, and assumptions. Those data have
                               several limitations that affect the model’s accuracy and usefulness.

ERAM’s Yield Curves Based on   ERAM  is designed to estimate the revenues associated with a staffing
Fewer Data Points Than         increase by constructing yield curves derived from data generated by
Functions’ Own Yield Curves    the functions. Yield curves are graphic depictions of the relationship



                               ‘FXAhf’s revenue estimates are discounted to their present dollar value.
                               “Marginal yield is the change in total yield arisiig from an additional enforcement case worked.



                               Page 37                                               GAO/GGfMO-S5 IRS’ Eufomzment Revenuea
                                    Chapter 5
                                    Data Umltations Prevent IRS From
                                    J3valuatIng ReIatIve Merits of Projto6ed
                                    IIICNWW in Enforcement Staffing




                                    between the amount of revenue and a given level of staffing. ERAM con-
                                    structs these yield curves from four staffing points supplied by the
                                    functions. Those four points show the revenue associated with (1) the
                                    maximum additional staff that the function believes it can absorb
                                    during 1 year, (2) a staffing increase less than the function’s maximum
                                    absorption figure, (3) the function’s budgeted staffing figure, and (4) no
                                    staffing increase for the function.

                                    According to IRS’ Budget Revenue Coordinator, revenue estimates based
                                    on the four-point yield curves are not as accurate as revenue estimates
                                    developed for specific staffing increases by the functions during the
                                    budget process. He said that the functions’ estimates are generated from
                                    a far greater number of yield points.

Unsubstantiated   Collection Flow   ERAM  estimates revenues using assumptions developed by the individual
Rates                               functions about the extent to which assessments will actually be col-
                                    lected and the timing of those collections over the years. These flow rate
                                    assumptions greatly influence the total amount and 5-year distribution
                                    of revenues projected by ERAM.

                                    On the basis of our current and prior work, the flow rate assumptions
                                    developed for Examination, Collection, and IRP Underreporter revenues
                                    are not reliable. For example, IRSbased its Examination flow rates on
                                    the results of a study tracking the results of 1972 audits-a study that
                                    IRSofficials acknowledge is outdated and invalid. The Collection official
                                    responsible for developing revenue estimates said that Collection’s flow
                                    rates for the delinquent returns and IRP nonfiler programs were devel-
                                    oped some time ago on the basis of a study of a sample of cases worked
                                    in Collection. IRS had no documentation of the study’s methodology or of
                                    its results. The official who develops IRP underreporter revenue esti-
                                    mates said that the flow rates for under-reporter cases worked by Exam-
                                    ination are based on the 1972 case results discussed above, and the flow
                                    rates for the remaining under-reporter cases are judgmental and were
                                    developed years ago from input provided by several IRS managers.

Marginal Yield and Cost Data        IRS has stated that ERAM can be used to estimate the optimal revenue-
Generally Not Available             maximizing allocation of additional resources across its enforcement
                                    programs on the basis of the relative marginal yield-to-cost ratios of the
                                    various programs. However, IRS does not have marginal yield and mar-
                                    ginal cost data for Collection and IRP. For those programs, therefore, IRS
                                    substitutes adjusted average yield and cost figures for marginal data in
                                    ERAM.



                                    Page 38                                    GAO/GGM   IBS’ Enfonxment   lkvennes
Chapter 6
Data Limitations Prevent IRS F’rom
Evaluating I&dative MerIta of Proposed
Increaw in Jhforcement Staffing




In estimating the amount of revenue that can be generated from
increased resources, marginal yields should be compared. Marginal yield
is the change in total yield arising from an additional enforcement case
worked. If IRS’ enforcement programs generally pursue cases with the
greatest revenue potential first, as IRS claims, the yield generated by
additional enforcement staff will generally be less than the average
yield. To allocate additional enforcement resources in a way that maxi-
mizes yield, IRS needs to evaluate how the marginal yield-to-cost ratios
change for various levels of staffing. By substituting adjusted average
yield and cost data for marginal data, IRS could be misstating its revenue
projections. We do not know to what extent that might be happening
because IRS had no documentation to show how it computed adjusted
average yield and cost.

The Collection function-which     estimates accounts receivable and delin-
quent returns revenues-submits      judgmentally adjusted average, rather
than marginal, yield and cost figures to ERAM because it lacks the neces-
sary data to report on the effect of its activities at the margin. To obtain
marginal yield and cost information, Collection would have to track the
revenue resulting from individual cases and the amount of time and
staff spent on closing those cases. With this information, Collection
could then generate yield curves that project future revenues expected
at different staffing levels.

As noted in chapter 3, Collection does not have a database linking case
results and the resources applied to those cases. IRS officials said that
they estimate Collection’s marginal yield for ERAM by analyzing master
file data for different categories of cases, broken down by potential
yield. As discussed in chapter 3, however, the master file does not con-
tain information on the amount and type of staff resources used for
these cases. IRS, therefore, cannot measure the impact of additional
staffing necessary for marginal yield calculations.

The Examination function also provides ERAM with adjusted average
yield and cost data for the IRP under-reporter component because it lacks
marginal data. The recently developed IRP Management Information
System mentioned in chapter 4 may eventually provide the data needed
to calculate marginal yield and cost.




Page 39                                  GAOpXXMU35   IIES’ Enforcement Revem-
                           Chapter 6
                           Data Limitations Prevent IRS From
                           EvahatIng ReIatlve Merits of Propo6ed
                           Increases in Enforcement Staffhg




IRS Uses ERAM Only to      Because of data limitations, IRS does not use ERAM to estimate revenues
                           or to allocate additional enforcement resources. IRS only uses the model
Estimate Support and       to estimate the staffing costs that occur in support of, or as a follow-on
Follow-On Costs            to, an enforcement staffing increase. For example, a staffing increase in
Associated With Staffing   Examination would be expected to eventually increase the number of
Increases                  returns audited. That increased workload would necessitate the hiring
                           of (1) additional support staff, such as clerks, technicians, and
                           paraprofessionals; and (2) additional professional staff in other func-
                           tions, like Collection and Appeals/Tax Litigation, to handle the
                           increased workload that those functions can expect as a follow-on to the
                           increased audits. The data ERAMuses to estimate those support and
                           follow-on costs come from the individual functions.


Treasury and OMB Do Not    IRS designed ERAh! to project alternate scenarios of an enforcement initia-
                           tive’s revenues and costs for OMB and Treasury to use in considering pro-
Use ERAM                   posed staffing increases. Both OMB and Treasury officials said, however,
                           that they do not use ERAM’Sestimates or modelling capabilities. These
                           officials said that they thought ERM’S concept was valuable, but they
                           believed the model’s output to be unreliable.

                           As an example of ERM’S unreliability, the OMB official who oversees IRS’
                           budget cited the fact that W’S collection flow rates indicate that IRS
                           realizes positive revenues in the first year of an Examination staffing
                           increase-a result that we questioned in our 1988 report on IRS’ process
                           for estimating the revenue from its Examination program and in our
                           1990 report on the costs associated with training new Examination
                           staff.4


                               needs the capability, like that envisioned for ERAh4, to compare
Conclusions                IRS
                           enforcement programs with each other to determine the optimal alloca-
                           tion of enforcement resources and to inform the administration and Con-
                           gress about the relative merits of proposed staffing increases. Because
                           of limitations with the input data provided by the individual func-
                           tions-limited   yield data, unsubstantiated collection flow rates, and
                           adjusted average rather than marginal yield data-ERAM has not filled
                           that need and is not being used as intended.


                           4Tax Administration: Difficulties in Accurately Estimating Tax Examination Yield (GAO/
                           aD-8&119,    Aug. 8,19f!8); Tax Admix&ration: Potential Audit Revenues hst While Training New
                           Revenue Agents (GAO/GGDW)-77, Apr. 6,199O).



                           Page 40                                           GAO/GGBS946     IPB’ Enforcement Rem-
Chapter 5
Data Limitations Prevent IRS From
Evaluating Relative Merits of Proposed
Increases in Enforcement Staffing




As discussed in chapter 6, IRS has various efforts underway or planned
to improve its revenue estimates and other enforcement data. We
believe that IRS needs to determine, among other things, how the results
of those efforts will affect the use of or need for ERM. Accordingly, we
are making no specific recommendation regarding ERAM.




P4e41
IRS Needs to Ensure That Various Efforts to
Improve Revenue Data Are Comprehensive and
Appropriately Coordinated
              As far back as 1981, we reported on the need for IRSto develop better
              information on the results of its enforcement programs. In 1981, for
              example, we testified that:
              I‘
                   the Service needs better data on the cost and revenue yield of its compliance
                   .   .


              programs. Generally, IRS has data showing only ‘average’ dollar yields. IRS’ manage-
              ment information systems do allow measurements of yield ‘at the margin’ for the
              examination program and do permit a rough estimate for the information returns
              program, but they do not provide marginal yield data for other compliance
              programs.“’

              In a 1982 report, we noted that IRS’ information systems generally do not
              provide adequate data on the cost and revenue impact of its compliance
              programs and that, except for the Examination program, IRS’ informa-
              tion systems generally do not provide data necessary for estimating
              marginal yield.2 We continued to make that point in subsequent years’
              testimonies on IRS’ budget requests3 In our 1982 report, we described
              IRS’ long-standing need for agencywide management information and
              noted that:

              “IRS’ management information systems do not provide adequate data to optimally
              allocate resources within and among its various compliance programs. The systems
              generally do not accumulate and report complete cost and revenue data for a compli-
              ance program, even though the data components exist at various places within the
              agency. Generally, each division or function within IRS has its own information sys-
              tems, designed to serve its own needs, and the systems usually do not track cases or
              projects across divisional lines. As a result, cost/revenue data for a particular pro-
              gram are often incomplete.

              “IRS recognized the need for an agencywide management information system as
              early as 1976    IRS’ progress in developing a system to meet those needs has been
              slow, however.”

              The situation we described then is essentially unchanged today. Recent
              concerns about the reliability of IRS’ revenue estimates, however, have
              again highlighted the need for a comprehensive, inter-functional
              database on the results of IRS’ enforcement programs. In response to that

              ‘Statementof William J. Anderson,Director, General Govemment Division, before the Subcommittee
              on Oversight,HouseCommitteeon Waysand Means, on the adequacy of IRS’ compliance resources
              for fiscal year 1982 (May 11, 1981).

              %uther ResearchInto NoncomplianceIs Neededto ReduceGrowing Tax Losses(GAO/GGD-82-34;
              July 23, 1982).
              3See,for example, statementsof Johnny C. Finch, SeniorAssociate Director, General Government
              Division, before the Subcommittee on Oversight, House Committeeson Ways and Means, on the admin-
              istration’s budget requestsfor IRS for fiscal years 1986 and 1987 (Apr. 29,198s and May 12,1996).



              Page 42                                            GAO/GGD99-35 lRS’ Jhforcement       Revenues
                        Chapter 6
                        IRSNeedstoEnsureThatVariousJZfortato
                        Improve Revenue Data Are Comprehensive
                        and Appropriately Coordinated




                        need, IRSnow plans to establish an integrated enforcement management
                        information system -the first phase of which it expects to be opera-
                        tional in August 1990. It also has several other efforts underway or
                        planned that are intended to develop more reliable data on the results of
                        its enforcement actions and that could provide a basis for more reliable
                        revenue estimates.

                        Because IRS’ plans for future phases of the integrated information
                        system have not been finalized, it is unclear whether the system, when
                        fully operational, will provide the comprehensive data needed to meet
                        IRS’ objective. Considering IRS’past record in developing such a system,
                        however, we think there is reason for some concern. To better ensure
                        that IRS’ current efforts succeed where others have failed, IRS needs to
                        (1) develop a strategy that shows how each of its efforts will contribute
                        to the ultimate objective of more reliable enforcement data and revenue
                        estimates and (2) provide executive oversight to ensure effective imple-
                        mentation of that strategy.


                        Recognizing that IRSneeds a comprehensive, interfunctional database on
Not Yet Clear Whether   the results of its enforcement programs, IRS’ Research Division began
Planned Enforcement     planning the Enforcement Management Information System (EMIS) in
Management              July 1989. IRS envisions EMIS as an integrated system that will track the
                        workflow and measure accomplishments for each of IRS’ major enforce-
Information System      ment functions (Examination, Collection, Criminal Investigation, Chief
Will Meet IRS’ Needs    Counsel, and IRP). The key feature of IRS’ plan for EMIS is that it will
                        track case activities across all enforcement programs through comple-
                        tion. In doing this, IRS hopes that EMIS will allow IRS executives to better
                        manage the enforcement area and will ultimately provide more accurate
                        measures of enforcement revenues and costs for Treasury, OMB, and
                        Congress. IRS had planned to implement this system by June 30, 1990-a
                        milestone identified as one of IRS’ critical success factors in its strategic
                        planning process.

                        IRShas described EMIS as a system that must be able to integrate case
                        data from numerous functions, so that each function uses the same data
                        element definitions, the data sources do not overlap, and cases are not
                        double-counted. Although EMIS’ development is incomplete, IRS’current
                        plans for implementing the system raise concerns about whether it will
                        achieve that objective.

                        IRS’current plans for EMIS call for incorporating and tracking Examina-
                        tion and certain other cases by the June 30, 1990, milestone. According


                        Page 43                                  GAO/GGD90-85 IRS’ Enforcement Revenues
Chapter 6
Jl?SNee&toEnmuwThatVariouaEfXortato
Improve Revenue Data he Ccbmprehenslve
ad Appipprktely  ~rdhated




to those plans, Collection cases will be added to the system in 1991. The
plans also note that IRP cases that are not referred to Examination, Col-
lection, or Appeals will not be incorporated in EMISbut that a future
enhancement to EMISmay provide for a direct link to the IRP Manage-
ment Information System.

In developing plans to implement EMIS,IRS decided to primarily focus its
initial efforts on the Examination program because Examination aIready
had a case tracking system in place. That system, known as the Audit
Information Management System, tracks Examination cases through the
assessment of taxes. Under EMIS’design, IRSwill track these cases until
collection. Consequently, EMISis expected ultimately to contain informa-
tion on the actual dollars collected from all audits.

According to Research Division officials, expanding EMISto include the
large volume of IRP underreporter and nonfiler cases would be very
costly.4 These officials believe that the IRP Management Information
System may provide adequate information on IRP case results either on
its own or through integration of that system’s data into EMIS.They are
concerned, however, about the potential for doublecounting IRP cases
because some of the cases included in the IRP Management Information
System are also reported in the Audit Information Management System,
which is being used to develop the Examination case data in EMIS.

The Research Division plans to include Collection cases in EMIS’design
by July 1991. Case data on these programs would have to be developed
from the master files because Collection has no database tracking its
case outcomes. According to Research Division officials, the master files
are not readily accessible and, unlike the Audit Information Manage-
ment System, do not provide all the needed data to track a given Collec-
tion case. In particular, the master files do not contain any data on the
staffing costs associated with working individual cases. Without such
data, IRS will be unable to develop marginal yield-to-cost ratios for the
Collection function. These ratios are needed to allocate Collection
resources in an optimal manner.




4For tax year 1986. for example, nearly 10 million cases were worked under IRP’s underreporter
program.



Page 44                                              GAO/GGDSO-65 IRS’ Enforcement Revenues
                               Chapter 6
                               IRS Needs to Ensure That Various Efforts to
                               Improve Revenue Data Are Comprehensive
                               and Appropriately Coordinated




                                   recognizes the need for more reliable enforcement revenue estimates.
IRS Needs to Ensure            IRS
                               In addition to EMIS, IRS has 12 other ongoing or planned projects that
That Planned                   may contribute to that end. Some of these projects are specifically aimed
Improvements to the            at improving IRS’ enforcement revenue estimates, while others are
                               designed to develop better enforcement program information. These
Revenue Estimation             projects, which are being developed in different organizations within IRS,
ProcessAre                     are described in appendix I.
Effectively                    Because it is critical that IRS develop more reliable information on
Implemented                    enforcement revenues and because attainment of that goal requires the
                               involvement and cooperation of several functions, IRS needs to
                               (1) develop a strategy that spells out its ultimate objective, identifies the
                               various efforts that will contribute to the objective, and describes how
                               the results of those efforts will be integrated; and (2) provide top man-
                               agement oversight over the strategy’s implementation.


IRS Needs a Strategy for       Considering the number of efforts IRS has ongoing or planned that would
                               contribute to improved enforcement revenue data and more reliable rev-
Coordinating and               enue estimates, IRS needs a strategy for ensuring that the efforts are
Integrating Its Numerous       appropriately coordinated and that their results are coherently linked.
Projects                       In doing so, IRS needs to develop specific plans that include both a long-
                               term strategy for developing the needed data and an interim strategy
                               for improving the estimates in the short term. Key questions to be
                               addressed in such a strategy include the following:

                           .   How will IRS’ numerous projects relate to its objective of improving
                               enforcement revenue estimates?
                           .   How will the various management information systems, databases, and
                               estimating models resulting from these projects be integrated?
                           .   Will the projects’ results complement each other?
                           .   How, if at all, will the results of these projects affect the use of or need
                               for ERAM?
                           .   Are the various projects properly timed so that their results can be
                               integrated?
                           .   Can the number of projects (13 or greater) be reduced to a number that
                               can produce the same results at a lower cost?
                           .   Are there any redundancies among these projects, and, if so, can they be
                               eliminated or minimized?
                           .   How will EMIS address the requirements for improving each enforcement
                               function’s revenue data, such as (1) updating collection flow rates,
                               (2) linking marginal yield and marginal cost data on cases, and



                               Page 45                                       GAO/GGD9@85 IRS’ Enforcement Revenues
                       Chapter 6
                       IRS Needs to Ensure That Various Efforts to
                       Improve Revenue Data Are Comprehensive
                       and Appropriately Coordinated




                       (3) attributing    revenue collected to particular enforcement functions
                       and actions?


Success of IRS’ Data   As discussed earlier, IRS has long recognized the need for improved data
                       on enforcement results. Despite that recognition, IRS has made little pro-
Improvement Efforts    gress in meeting its need. We believe that the success of IRS’ current
Depends on Effective   attempt to develop more reliable management information will depend
Oversight              on top management’s active involvement in overseeing the effort.

                       Oversight responsibility now appears to rest with an Enforcement Rev-
                       enue Estimation Steering Committee that IRS established in July 1989.
                       That Committee is comprised of assistant commissioners from each
                       revenue-producing function and is chaired by the Assistant Commis-
                       sioner for Planning and Research. We do not believe that a committee so
                       comprised is in the best position to provide effective oversight of this
                       effort. In our opinion, oversight needs to be at a level that can transcend
                       the parochial interests of the individual functions and make decisions on
                       an n&-wide perspective.

                       The steering committee’s ability to effectively oversee and coordinate
                       IRS’ revenue estimation efforts might also be impaired by the fact that it
                       had only identified for its oversight 9 of the 13 projects described in
                       appendix I. The four missing projects are the accounts receivable man-
                       agement information enhancements, the Collection Budget Formulation
                       Model, the IRP Management Information System, and the Taxpayer Com-
                       pliance Measurement Program for individual nonfiler cases. Two senior
                       IRS Research officials said they could not recall who compiled the list of
                       nine projects for the steering group. Thus, we were unable to determine
                       why the four projects, which are aimed at improving enforcement rev-
                       enue data and would likely affect the quality of future revenue esti-
                       mates, were excluded from the list.

                       We also found no evidence in minutes of the committee’s meetings that it
                       had defined or was pursuing an integration strategy. Most of the meet-
                       ings focused on the first phase of EMIS’ implementation, which, as noted
                       earlier, primarily involves the tracking of Examination cases.


                       As we reported on various occasions during the 198Os, IRS needs better
Conclusions            information on the results of its enforcement programs. Although IRS
                       has long recognized that need, it has been unsuccessful in developing
                       systems to generate the information. There is now a renewed emphasis


                       Page 46                                       GAO/GGDSW35 IRS’ Enforcement Revenues
                        Chapter 6
                        IRS Needs to Ensure That Various Efforts to
                        Improve Revenue Data Are Comprehensive
                        and Appropriately Coordinated




                        within IRSto address the problem. To ensure that the emphasis is prop-
                        erly directed, IRSneeds to develop a strategy that shows how the results
                        of its various projects will be integrated to achieve its objective of more
                        reliable enforcement data and revenue estimates.

                        To ensure that the strategy is effectively implemented and that the cur-
                        rent emphasis is sustained over time, IRS’top management needs to
                        closely oversee the effort. Because IRS’effort to improve its enforcement
                        revenue estimates is so critical and cuts across the entire organization,
                        we believe that oversight needs to be provided at the Deputy Commis-
                        sioner level, where decisions can be made from an IRS-wide perspective.


                        We recommend that IRS:
Recommendationsto
the Commissioner of .   Develop a comprehensive strategy for identifying and coordinating the
Internal Revenue        various enforcement revenue projects and integrating their results. As
                        part of the strategy, IRS should ensure that revenue and cost data on all
                        enforcement cases, including Collection and Information Returns Pro-
                        gram cases, are either incorporated into the EMISdatabase or are linked
                        with EMISdata in a way that avoids double-counting.
                        Assign responsibility at the Deputy Commissioner level for overseeing
                        development and implementation of the strategy.
                        Use the results of its various projects to improve the enforcement
                        revenue-estimating process.




                         Page 47                                      GAO/GGD90-85 IRS’ Enforcement Revenues
Reliability of Enforcement Revenue Data Has
Broad Implications

               As concerns about the growing deficit mount, Congress and the adminis-
               tration look to IRS for increased enforcement revenues. The President’s
               budget for fiscal year 1991, for example, includes a request for about
               3,600 additional enforcement staff for IRS to use in doing more audits;
               collecting more unpaid taxes; and verifying certain deductions, credits,
               and exemptions--a staffing increase that IRS has estimated will generate
               about $500 million in additional revenue in 1991 and about $6.5 billion
               by the end of fiscal year 1995. A proposal introduced by Senator Conrad
               calls for a much larger staffing increase, including 3,000 additional
               examiners and 2,500 additional collection personnel, with the expecta-
               tion of generating $1.7 billion in fiscal year 1991 and $32.9 billion by the
               end of fiscal year 1995.

               To adequately assess such proposals, Congress and the administration
               need reliable data on how much revenue they can reasonably expect
               each of IRS’ enforcement programs to generate at various levels of
               staffing and how soon they can expect that revenue to come into the
               Treasury. Reliable enforcement data are also essential if IRS is to effec-
               tively manage its programs. Without such data, IRS cannot make
               informed decisions as to how to use its resources in a way that maxi-
               mizes revenues.

               IRS provides Congress and the administration with two types of enforce-
               ment revenue data-information       on the amount of revenue it expects to
               generate with increased staff and information on the amount of revenue
               it estimates it actually realized. As discussed in previous chapters, that
               information shows some big variances between expectations and final
               results.

               Considering the inherent difficulties in trying to estimate results many
               months before the fact and all of the unanticipated events that can
               occur after an estimate is developed, it is not unreasonable for estimates
               to deviate from actuals. Our concern is not that variances exist but that
               IRS does not have the documentation and does not do the kind of moni-
               toring necessary to evaluate those variances. Through such an evalua-
               tion, IRS could identify and correct problems with its methodology, such
               as inappropriate assumptions, and thus enhance future estimates.

               What we find most intolerable is IRS’ continuing inability to say what it
               actually receives in revenues as a result of its various enforcement pro-
               grams. We think it is vital that Congress and the administration be given
               accurate information that they can use to judge the actual effectiveness



               Page 48                                 GAO/GGD9O-85 IRS’ Enforcement Revenues
Chapter 7
Reliability of Enforcement Revenue Data Has
Broad Implications




of IRS’various enforcement programs in generating revenues-informa-
tion that they can use in deliberating on future changes to those pro-
grams. As we reported in 1982, IRS recognized the need for improved
enforcement revenue information as early as 1976.’ Despite that recog-
nition, IRS has made little, if any, progress in filling that need and must
still estimate its actual enforcement revenues.

Now, with growing pressure from Congress and the administration, IRS
has expressed its intent to improve the reliability of its enforcement rev-
enue data and has various projects underway or planned toward that
end. Considering the fact that IRShas been unable to reach that goal
since 1976, however, and recognizing that its success will require sev-
eral years of sustained effort, we believe that continuing congressional
oversight is essential. Congressional oversight would be enhanced if IRS
provided Congress with detailed plans and milestones for improving its
enforcement revenue data and periodic reports on the status of its
efforts and if Congress held IRS accountable for meeting those plans and
milestones.

Until I= starts generating more reliable revenue information, however,
Congress will be faced with the need to make prudent funding and
staffing decisions involving IRS’ enforcement activities. Faced with an
audit coverage of about 1 percent and an accounts receivable inventory
of over $60 billion, it seems clear that there are benefits to be gained by
increasing IRS’enforcement staff. In the absence of more reliable rev-
enue data, however, Congress needs to ensure that its deliberations on
future staffing increases are based on the most conservative assump-
tions-remembering,      as we have reported previously, that there are
limits to the number of additional staff that IRS can absorb at any one
time and that the benefits from increased enforcement staff are more in
the long term rather than the short term.’ The Congressional Budget
Office used some conservative assumptions, for example, in assessing
the revenue estimate associated with the administration’s request for
additional IRS enforcement staff in fiscal year 1991. It reduced the
administration’s estimate of first year revenues from $500 million to
$300 million after deciding that revenues would not be generated as
quickly as the administration assumed.

’ Further Research Into Noncomphance Is Waded to Reduce Growing Tax Losses (GAO/GGD-82-34;
July 23, 1982)

“Tax Administratron: IRS’ Implementation of the 1987 Revenue Initiative (GAO/GGD-88-16. Dec. 2.
1987); Tax Administration: Difficulties in Accurately Estimating Tax Examination Yield (GAO/
GGD-88-119. .4ug. 8. 1988); Tax Administration: Potential Audit Revenues Lost While Training New
Revenue Agents (GAO/GGD-90-77, Apr. 6,199O).



Page 49                                            GAO/GGD90-86 IRS’ Enforcement Revenues
                       Chapter 7
                       I&liability of Enforcement Revenue Data Has
                       Braad Implications




                       As we testified in April 1989, Congress also needs to ensure, as it con-
                       siders increases in enforcement staff, that IRS is being funded at levels
                       sufficient to maintain current, needed operations.3 IRS’ recent exper-
                       iences provide a case in point. As we testified in March 1990, funding
                       shortages in fiscal year 1990 caused IRS to defer most of the enforcement
                       staffing increases that Congress had authorized-thus       negating
                       expected revenue increases4


                       The Senate and House Budget and Appropriations Committees should
Recommendationsto      closely oversee IRS’ progress in improving its enforcement revenue data.
the Senate and House   In doing so, we suggest that the Committees require IRS to provide
                       detailed plans and milestones and periodic status reports. A&o, until
Budget and             more reliable data are available, the Committees should use conserva-
Appropriations         tive revenue assumptions in deliberating on future enforcement staffing
Committees             increases, including those being proposed for fiscal year 1991.




                       ,lAdministration’s Fiscal Year 1990 Budget Proposals for IRS and the Tax Court (GAOEGGD-89-16,
                       Apr. 4, 1989).

                       “IRS’ Budget Request for Fiscal Year 1991 and Status of the 1990 Tax Return Filing Season (GAO/T-
                            - 0-26 , Mar. 22, 1990).



                       Page 50                                             GAO/GGD90-65 IRS’ Enforcement Revenues
Agency Comments and Our Evaluation


    .-
              In his June 11, 1990, comments on a draft of this report (see app. II), the
              Commissioner of Internal Revenue expressed agreement with many of
              our recommendations and said that IRSmust continue to improve its
              enforcement revenue estimating models. As discussed more fully below,
              the Commissioner said that IRS has already taken action to implement
              some of our recommendations, especially as they relate to better docu-
              menting its estimating processes.

              As a result of a comment by the Commissioner, we revised our first rec-
              ommendation in chapter 6 to make it clear that information on all
              enforcement cases could either be incorporated into the EMIS database or
              linked with EMIS data. The Commissioner was concerned that our recom-
              mendation, as previously worded, would have required IRS to include IRP
              data in the EMIS database itself when IRS felt that linking IRP data with
              EMIS data would serve the same purpose.

              The Commissioner did not fully agree with our second recommendation
              in chapter 6, in which we call for assigning responsibility at the Deputy
              Commissioner level for overseeing development and implementation of a
              comprehensive strategy for identifying and coordinating the various
              enforcement revenue projects. The Commissioner said that he felt over-
              sight responsibility properly rested with IRS’ Controller-a   position IRS
              recently created in response to a previous GAO recommendation.’ Consid-
              ering the Controller’s responsibility for establishing administrative and
              revenue accounting standards, tracking actual enforcement revenues
              against plans and coordinating estimates across IRS, and because the
              Controller will be reporting directly to IRS’Deputy Commissioner for
              Planning and Resources/Chief Financial Officer, we agree that making
              him responsible for IRS’strategy would be appropriate. However, given
              the importance of this work to IRS, we would still expect the Deputy
              Commissioner to actively oversee those efforts.

              The Commissioner also had several broad comments that he thought
              needed to be recognized in order to put our report in perspective. His
              first comment was that our report does not put the revenue estimating
              issue in context. He refers to the fact that IRS’ enforcement efforts, and
              thus enforcement revenues, have been seriously curtailed by operational
              budget shortfalls and that without additional resources IRS will be “ham-
              strung in its ability to enforce the tax laws” with a resultant loss in rev-
              enues and decline in voluntary compliance. Accordingly, while

              ‘Managing IRS: Actions Needed to Assure Quality Service in the Future (GAO/GGDS9-1. Oct. 14.
              1988).



              Page 51                                           GAO/GGD90-85 IRS Enforcement Revenues
Chapter 8
&ncy Comments and Our Evaluation




acknowledging the need to improve estimating procedures, the Commis-
sioner said that the “true issue” in IFS’ budget is whether IRS gets the
overall resources to carry out its basic tax administration responsibili-
ties We are on record as supporting IRS’needs for additional resources
for the same reasons cited by ~-an      increasing accounts receivable
inventory, a decreasing rate of audit coverage, and a sizeable tax gap.
IRS must recognize, however, that it will become more and more difficult
for IRS to defend that position and us to support it in the absence of
credible data on the value of such staffing increases-especially     data on
the revenues actually realized as a result of IRS’ enforcement efforts.

The Commissioner’s second point is that IRS is continuing to improve its
estimating process and that the revenue estimates in the administra-
tion’s fiscal year 1991 budget are based on more conservative and better
documented methodologies and thus are more defensible than the ones
we reviewed. We know from other work we have done and are doing
that IRS methodology for estimating the revenues to be generated by the
increased Examination staff requested for fiscal year 1991 is more con-
servative than its past methodology. We could not assess the methodolo-
gies behind the Collection and IRP revenue estimates in the fiscal year
 1991 budget or the adequacy of IRS’ documentation, however, because
data relating to those estimates were not available until our audit work
was virtually complete. Even if data had been available in time, it would
have been difficult to compare current and past methodologies because
of insufficient documentation behind the past methods. We acknowledge
that IRS has taken steps toward improving its enforcement revenue data
and estimating methodologies and have tried to recognize those efforts
throughout the report, especially in chapter 6.

The Commissioner’s final point is that significant budgetary shortfalls
over the past 2 years kept IRS from maintaining the level of enforcement
effort it had originally intended and caused it to revise its revenue pro-
jections. We do not disagree with IRS and, as noted in chapter 7, are on
 record in support of the need for Congress to ensure that IRS has enough
 funds to maintain necessary operations. We would only point out that
 improved information on IRS’ enforcement efforts, including the amount
 of revenue actually realized, might go a long way in helping convince
 Congress and the administration as to the adverse effects of inadequate
 funding.




Page 52                                  GAO/GGD-9045 IRS’ Enforcement Revenues
Page 53   GAO/GGD9O435 IRS’ Enforcement Revenues
IRS Projects Relating to Enforcement
Revenue Estimates

                           During our work we identified 13 ongoing and planned projects that
                           may contribute to improving IRS’ enforcement revenue estimates. One of
                           those projects, EMIS, is discussed in chapter 6. Descriptions of the other
                           12 projects follow.


Enhancements to Accounts   IRS  has underway a number of efforts to improve its accounts receivable
                           inventory data. Among these efforts are the development of two reports
Receivable Management      that together will track cases in IRS’ accounts receivable inventory by
Information                 fiscal year, age of receivable, and status of the receivable. Each case will
                           be tracked through closure or until the 6-year statute of limitation
                           expires. The purpose of these two reports is to allow IRS to take snap-
                           shots of the flow of collections from inventory cases at regular intervals.
                            IRS is developing this information primarily to provide a basis for pro-
                           jecting the portion of the accounts receivable inventory that will likely
                            be collected. Information on the timing and flow of collections would
                            also be useful, however, in projecting future revenues from cases in the
                            inventory.


Collection Budget          Collection is developing this model, which is also known as the Collec-
                           tion Resource and Database Tracking System, to (1) formulate revenue
Formulation Model          and workload data for its budget submissions and (2) execute its budget
                           by allocating staff and generating a work plan for field offices. The
                           model will summarize revenue data by district office and yield potential
                           of the cases but will not track individual case costs or revenues. The
                           model will use the same definition of yield as the Collection Yield Report
                           discussed in chapter 3.


IRP Management             This system, developed under the Assistant Commissioner for Human
                           Resources Management and Support, is designed to track cases to their
Information System         conclusion, thus measuring dollars collected. As described to us, once
                           the case tracking is completed, IRS will have historical case data that can
                           be used to project future IRP revenues.


Taxpayer Compliance        TCMP   is designed to measure the extent to which taxpayers comply in
                           filing an accurately completed, timely, and fully paid return. IRS’
Measurement Program        Research Division is planning a new KMP to develop comprehensive esti-
(TCMP)
_- ^.. For
         - Individual      mates of individual nonfilers in tax year 1988. IRS will randomly select
Nonfiler Gases             and work a statistical sample of cases from the population of potential
                           nonfiler cases generated through the IRP document matching program,


                           Page 54                                  GAO/GGD9W35 IRS’ Enforcement Revenues
                           Appendix I
                           IRS Projects Relating to Enforcement
                           Revenue Estimates




                           IRSinvestigative leads, and other sources. Among other things, the
                           survey results will help IRS to estimate the total number of individual
                           nonfiler cases and the value of the associated tax yield.


TCMP Estimated             Under this study, IRS will update TCMP revenue data on certain cases,
Checksheet and Unagreed    including those where the taxpayer disagreed with the audit results.
Case Study

Examination Revenue        Under this project, the Examination function is evaluating ways to
                           improve its revenue estimation and resource allocation model, along
Estimation Project         with the data sources used as input to that model. In the meantime, IRS
                           has introduced an interim method of estimating the revenues associated
                           with an increase in examination staff that is intended to update the
                           assumptions on which the estimates are based and provide short-term
                           improvement to Examination’s revenue-estimating process.


Multi-Year Input-Output    IRS’ Research Division plans to develop a new system for estimating
                           enforcement revenue called the Multi-Year Input-Output Model. This
Model                      system will reflect the flow of workload from one enforcement function
                           to the next, with each function receiving estimated inputs from other
                           functions, as appropriate, and estimating its own outputs. In the case of
                           Appeals/Tax Litigation, for example, the model would include as input
                           information on the audited tax returns that Examination expects will be
                           appealed and would include as output information on the cases that
                           Appeals/Tax Litigation would expect to transfer to Accounts Receivable
                           for collection. IRS has not yet developed an action plan for this model,
                           and the responsible Research Division officials told us that implementa-
                           tion will take about 5 years.


Direct Enforcement         IRS’ Finance Division and Treasury’s Office of Tax Analysis jointly

Revenue Estimation Study   undertook this project to develop improved estimates of total enforce-
                           ment revenues from the master files. As discussed in chapter 2, this pro-
                           ject was initiated in response to Treasury and OMB concerns about the
                            reliability of IRS past estimates, which were compiled from data sub-
                            mitted by the enforcement functions.




                           Page 55                                 GAO/GGD-9035 IRS’ Enforcement Revenues
                           Appendix I
                           IRS Projects Relating to Enforcement
                           Revenue Estimates




Collection Yield Study     IRS’ Research Division is implementing this project, which is aimed at
                           revising the current definition of collection yield. Chapter 3 describes
Group                      this project in more detail.


Chief Counsel Revenue      IRS’Chief Counsel’s Office is developing a statistical model for projecting
Projection Model           the additional revenue it can expect to generate through an increase in
                           Appeals/Tax Litigation staff.


Chief Counsel Resource     IRS’Chief Counsel’s Office is developing this model for use in (1) allo-
                           eating Appeals/Tax Litigation resources, (2) planning workload, and
Allocation Model           (3) projecting revenues.


Sustention Rate/Recovery   IRS’Research Division is conducting this study to determine the rate at
                           which adjustments proposed by Examination are eventually upheld in
Rate Study                 the Appeals/Tax Litigation function.




                           Page 56                                 GAO/GGD90-65 IRS Enforcement Revenues
Appendix    II

Comments From the Internal Revenue Service


Note GAO comments
supplementing those in the
report text appear at the
end of this appendix.                                     DEPARTMENT           OF THE TREASURY
                                                            INTERNAL          REVENUE          SERVICE
                                                                WASHINGTON.             D.C.   20224




                             Mr. Richard L. Fogel
                             Assistant   Comptroller     General
                             United States General Accounting                       Office
                             Washington,   D.C.      20548
                             Dear Mr.     Fogel:


                                     We have reviewed    your draft   report   concerning    IRS'
                             enforcement    revenue estimates.       We agree that IRS must continue
                             to improve its direct       enforcement    revenue estimating       models and
                             agree with many of your recommendations.            However, it should be
See comment 1                pointed    out that GAO examined estimates        of total   direct    revenue.
                             Neither    the indirect    enforcement   effects   nor the accuracy       of the
                             FY 1991 incremental      revenues were specifically        examined.
                                    We have made substantial      changes to our revenue projection
                             techniques    over the past several       years.    We believe     our current
                             estimates   are reasonable,     defensible,      and conservative.
                             me President's          FY   1991 Bug
Seep   51                    Qmration
                                       The report    does not put the revenue estimating                     issue in
                             context.        While it properly        urges us to improve our techniques
                             for counting        the trees,       we can't    lose sight of the fact that the
                             forest      is burning.        IRS' enforcement        efforts,        and thus
                             enforcement       revenues,      have been seriously            curtailed       by
                             operational       budget shortfalls.            Without     added resources           as begun
                             in the President's            FY 1991 Budget,       the IRS will          be hamstrung      in
                             its ability       to enforce       the tax laws -- a continuing                 condition
                             that means significant             revenue loss to the Federal government                     and
                             a concomitant         decline    in voluntary       compliance.           Although
                             difficult       to measure, the indirect            effects      of enforcement          on
                             improved voluntary            compliance     are at least as important                as the
                             direct      revenue effects.
                                     For these reasons,           while we agree the estimating                procedures
                             in place when GAO did their                 review needed improvement,              the true
                             issue in IRS' budget is whether we get the overall                            resources      to
                             carry out our basic tax administration                     responsibilities.             At a
                             time when our accounts            receivable        inventory    has increased
                             significantly,      our audit rate is down to less than one percent,
                             and the tax gap stands at $80 to $100 billion,                          it would be penny-
                             wise and pound-foolish           to focus on a discussion                 over the
                             methodology     for projecting            incremental      revenue enhancements             and
                             lose sight     of the fact that IRS must receive                   additional         funding
                             in order to accomplish             its tax administration            responsibilities.
                             Any discussion      of enfOrCeHEnt             revenue estimating          must be put into




                                   Page57                                                          GAO/GGD9O-S5IRSEnforcementRevenues
                         Appendix II
                         Comments From the Internal Revenue Service




                this context     so Congress  will           have the benefit   of your advice
                when considering     the important            issues of raising   revenue and
                funding  this agency to collect               that revenue.
Seep   52   I   IRS Continues        to ImDrove     Its     Estimatins   Procese
                        The revenue estimating       procedures      in place at the time GAO
                did its review have been modified.               While we have and should
                continue     to improve our revenue estimating           methodologies,    we
                believe    the new estimates      are both conservative        and defensible.
                In fact,     we have already    incorporated       many of GAO's earlier
                recommendations      for improving      estimating     models and look forward
                to your review of our FY 1991 estimates.
                       We are continuing        to refine  our current   revenue estimating
                methodologies,       and are also committed to a comprehensive,             cross-
                functional     information      system which can accurately       track all
                enforcement      results    through to completion.      The first      stage of
                this Enforcement         Management Information    System will      be
                operational      by August 15, 1990.
                        We have also established        recently     a Controller      organization
                that is responsible        for establishing       both administrative          and
                revenue accounting       standards.       For the first     time, one
                organization     will  have Service-wide         accounting     responsibility        for
                both operational      purposes and management information               purposes,
                including    the tracking      of actual    enforcement     collections        against
                planned collections       and coordinating        estimates     across the
                Service.
Seep   52       Steadv     Budaets    Mean Steadv         Proaress
                         IRS has suffered    from significant      budgetary    shortfalls    over
                the past two years.        These shortfalls       and absorptions       --
                primarily     mandatory   cost absorptions      -- required     IRS to shift
                funds just to meet payroll.           In essence, we mortgaged tomorrow to
                pay for yesterday's       costs.    In doing so, we found ourselves           in a
                position     where we could not always do the job we had been
                originally     budgeted to do.      As a practical      matter,     we have not
                been able to maintain        our level    of revenue officers        or revenue
                agents for almost two years -- another year would be even more
                detrimental     to revenue collections.         As a result,     revenue
                projections     prepared   several    years before funds were appropriated
                had to be revised       to reflect    the approved funding        levels.
                         The FY 1991 budget request     begins to restore       IRS to a sound
                footing     budget-wise;     it does not solve totally    all of the
                shortcomings       from prior   years.  In fact,    as we testified      this
                year, IRS will        have to absorb several    hundred million     dollars       in
                unbudgeted      costs in the coming fiscal      year.   As GAO itself
                testified     in March of this year, "The initiatives          that call      for




                         Page58                                          GAO/GGIHO-86 IRS Enforcement Revenues
        Appendix II
        CommentsFrom the internal Revenue Service




                                              3
additional        enforcement  staff     will,    in effect,    only serve to
replace      staff    that IRS lost    in the     last year    and a half due to a
hiring     freeze."
        Detailed    comments on the recommendations,             as well as other
specific       comments on the report  are included            as enclosures  with
this    letter.     This includes  comments provided            by Treasury's
Office      of Tax Analysis.
       Thank you for      the   opportunity       to comment on this     report.
       Best   regards.




Enclosures




        Page59                                           GAO/GGD90-86 IRS'EnforcementRevenues
                          Appendixll
                          CommentsFromthelnternalBevenueService




                                                      IRS'    COMMENTS

                        LE OF CONTEN!I!S AND m
                        We suggest   that   the term "were not realized"        be used when
                 discussing    the results      of our previous   enforcement    revenue
See comment 2.   estimates.      The report     uses the term "overstated"      and ngrossly
                 overstated"     in characterizing      IRS' estimates    of enforcement
                 revenue in the past.


Now p 3                  Page 4 -- The report states that Congress needs to ensure
                 that   deliberations       on future staffing         increases     are based on the
See comment 3    most conservative         assumptions.       We suggest that this statement           be
                 revised     to reflect     that Congress should use the most conservative
                 revenue assumptions          in their   deliberations        about "additional
                 revenue"      rather   than about future       staffing      increases.       (See page
                 7 of the Executive         Summary.)     Deliberations         about IRS' staffing
                 levels    should be discussed         in terms of the overall           mission
                 Congress expects IRS to accomplish,                not in terms of revenue
                 assumptions.
Now p 3                  Page 4 -- We suggest that the addition                       of IRS enforcement
See comment 4
                 staff    will generate    "substantialn       additional               revenue.      Likewise,
                 we believe    that the report        should note that                there would be
                 significant     increased    indirect     revenues from                such initiatives.

                                        FLICTING     INFO-ON          ON MAGNITUDE OF IRS'                TOW

                           endauons     to the    Secretarv       of the   Tre?iSUN


                          We recommend that the Secretary               of the Treasury    direct    IRS
                          and OTA to provide           Congress with information        on the actual
                          revenues generated           by IRS' enforcement       programs as soon as
                          it becomes       available.       In the interim,      the Secretary    should
                          direct     IRS and OTA to identify          the most reliable      methodology
                          for    estimating      actual   enforcement     revenues.     The Secretary
                          should also direct           IRS to explore     ways to link improved
                          revenue data with proposed staffing                increases   to ultimately
                          provide     Congress with more reliable            estimates   of the revenue
                          expected      from those increases.
                 IRS Response:
                         We do not disagree    with this recommendation         with respect   to
                 the IRS; however, OTA provides          information    on total    federal
                 collections    but does not furnish        information    on enforcement
See comment 5    revenue to Congress.        However, we would simply point out that the
                 most reliable     methodology    requires     linking  the enforcement
                 functions'    data bases to data on the IRS master file.               The IRS is




                          Page60                                           GA0/GGB90-%IRS'EnforcementRevenues
                         Appendix II
                         Comments From the Internal Revenue Service




                                                               2
                implementing     that methodology           in the development     of EMIS.   IRS
                currently    cannot provide    data         on direct  enforcement    revenue
                 (other than IRP) by function.
                Other    Comments:
                         The report     states      that IRS is unsure of the actual             levels of
                enforcement        revenues.        The Direct    Enforcement        Revenue Study
See comment 6   alluded     to in the report           is, in fact,      the first      step towards
                resolving      this uncertainty.           IRS now estimates          that the amount of
                enforcement        revenue generated         in FY 1989 was at least $29.6
                billion.       Absent data on actual           collections        from enforcement
                activities,        this is the most accurate             estimate     the Department has
                on enforcement         collections.




                        We recommend that         IRS
                        --    Develop and document a consistent           methodology      for
                              reviewing  historical     trends,    incorporating      anticipated
                              program changes,      and estimating     revenues    in Collection;
                        --    Fully   document the data used in applying            the
                              methodology,      including      the trend analyses     performed            and
                              the assumptions        underlying    those estimates,      and the
                              results   therefrom;       and
                        mm    Monitor    estimates    against    actual   results    over time,
                              using the most reliable         data on actual      results
                              available,     to assess the causes of any significant
                              discrepancies      and to identify       any adjustment     needed           in
                              assumptions      or methodology.
                        While we agree with these recommendations                      and believe      that
                RMIS will     accomplish       these goals,       we must again point out that
See comment 7   the report     critiques       a revenue estimating            model that has already
                been improved significantly              -- and which can be documented.
                While we agree that more can be done to improve revenue
                estimating     in the Collection           function,       the report     does not
                acknowledge      efforts      Collection      has made over the past two years
                to improve     its estimating          such as:       using Collection          research
See comment 8   file    data to validate         assumptions;        designing     a model that will
                use more discrete          data in developing           the estimates;        hiring    an
                analyst     to specialize        and focus on developing             a statistically
                sound estimating         technique;      and exploring         the use of RWMSscores
                to improve estimates           of marginal       yield.




                         Page 61                                           GAO/GGD-90-85 IRS’ Enforcement Revenues
                                                               3
                        Assumptions      were also changed based on historical          data.
                  This resulted      in lower projected       revenue per additional      staff
                  year, specifically       the estimated      yield   per additional  staff     year
See comment 9     was reduced after       examining   results      of the FY 1988 Collection
                  program.    Collection      has documented its methodology.
                          The changes made to Collection's    methodology    were first                used
                  to prepare the FY 1991 revenue estimates.         To our knowledge,                  GAO
                  has not reviewed     this interim   model which we believe     provides                a
                  more accurate     and reasonable  estimate  until  EMIS includes
                  Collection    cases.
                  Other    Comments:
Now p 27                  Page 38 -- The report's            discussion      of the use of different
                  weights      and different      historical       patterns      glosses   over the use of
See comment 10    judgmental      factors.       The historical         data contains      significant
                  peaks and valleys          that would skew a straight-line               estimate.
                  Examples of this are the 1985 filing                   season, introduction            of
                  Collection's        Resources and Workload Management System (RWMS), and
                  the rescheduling         of delinquent        returns     delinquency      checks.
                  Adjustments       that Collection         made to its estimates          were an attempt
                  to account for known, but not precisely                    quantifiable,        variations.
Now p 29.                 Pages 42-43 -- The report      states     that Collection       does not
See comment 11.
                  have a data base with time expended on Collection                  cases and
                  results    in terms of revenue.      While it is true that we do not
                  have a single     data base, this    information       is available      on
                  different    data  bases.    There are plans to revise           time reports    to
                  collect    case information,     but this    information     will    not be put on
                  IRS@ master file.

                                                                           AND MONITOR ITS NEW
                          DOLOGY FOR ESTIMATING              UNDE-ER             REVENUES
                                            . .
                  RecommeDggtions to the Commssloner                of Intern al Revenue :
                          We recommend that       IRS
                          --    fully    document its IRP underreporter         revenue estimates,
                                including     documentation    of any historical      trend
                                analyses     used to estimate    future  revenues and all other
                                assumptions      underlying  the estimates.




                           Page62                                          GAO/GGD90-85IRS'EnforcementRevenues
                            Appendixll
                            Comments From theIntemalRevenueService




                                                                 4
                    IRS Response:
                           We have already      implemented     this recommendation.         For the FY
See comment 12      1991 budget submission,          IRS has fully    documented records        of Tax
                    Year case inventories         for 1987 and all assumptions         (flow    rates,
                    average assessments       or refunds     by subcategory,       and average      costs)
                    underlying    the revenue projection          model.    Beginning   with FY 1989,
                    IRS has documentation        of analyses      of past case results       which
See comment 13      shows the percent      of distribution        for cases by category        with
                    assessments     by category      or subcategory      of use and average costs.
                            It is important    to note that IRS has improved the
                    documentation     of its methodology       in the past two years.     While we
See comment 14      believe    adequate documentation       exists   for FY 1989, either   in the
                    form of model documentation        or source documents which the model
                    used, we have improved the model so that for FY 1991, it has all
                    documentation     immediately    available     in the program itself.
                            se       monitor    "actual"    underreporter    results     against
                                     estimated     amounts to identify     the causes of
                                     discrepancies       and any needed changes in the methodology
                                     or assumptions       used to estimate     revenues;
                    IRS Response:
                            We agree with        this recommendation     and are establishing
                    management reports           of actual   results   to be provided    on a more
                    current    basis than        in previous    years.
                            --       use   actual      information   once it is available,     to monitor
                                     the   reliability        of IRP underreporter  estimates.
                    IRS Response:
                           We agree and would note that actual               information      is now
See comment 15.     available       for tax year 1985 which has been used to compare
                    against     IRP underreporter         estimates.      Our preliminary       comparison
                    of gross revenues         indicate     that our estimated       assessments were
                    within     one half of one percent           of actual   assessments.        This
                    comparison        is not meant to validate         the fiscal     year revenue
                    estimate      itself,   rather     it permits    IRS to validate       the model's
                    construction.
                    Other        Comments:
Now pp 32 and 35.          Pages 46 and 51 -- The report          states    that IRS is not
See comment 16.     adequately    documenting    the assumptions        underlying      the new
                    methodology    for  IRP underreporter        estimates.       This is incorrect.
                    IRS now has costs,      flow rates,      assessed amounts by case type,
                    number of cases in each type, etc. based on historical                      data.
                    This information      is readily    available      for FY 1991.        This
                    information    permits    IRS to monitor      estimates      against    actual




                            Page63                                          GA0/GGD9085IRSEnforcementRevenues
                             Appendix U
                             CommentsFromtheIntemalRevenueService




                                                              5
                  results.
Now pp 33-34              Pages 49-50 -- The report           discusses      the lack of
                  documentation       in assessing      the reasons for variances            between
Seep   52         estimated      and *@actual" underreporter          revenues.        As stated,    IRS
                  now has adequate documentation              for its IRP revenue estimates.
                  However, as the report           notes, there are significant             external
                  factors     for which IRS cannot project            in advance when making its
                  estimates.       Specifically       new legislation        in the early to mid
                  1980s affecting        information      reporting      (sales of securities,         IRA
                  deductions,      mortgage interest         deductions)       greatly   affected
                  historical      experience      on which projections          were made.
Seep   34                 The importance     of recognizing       these "external        influencesI'      is
                  clearly     seen in FY 1987, the year with the greatest                  variance
                  between the estimate        and actual     revenue (41%).          The major
                  variance     between the FY 1987 estimate           and the actual         revenue
                  resulted     from a variety      of concrete      causes.   First,       tax   year    1984
                  cases, which were worked in 1987, were adversely                    affected      by SCRS
                  and its impact on information           return     quality.     Second, many
                  erroneous      IRP cases were created because many IRA Forms 5498 were
                  not filed      or processed    in 1985.      This contributed        to a high
                  screen-out      rate and many unproductive          cases.    Third,       in this
                  first    year of matching      security    sales,     the basis for the security
                  sales was not present.           After this information         was obtained         from
                  the taxpayer       through  correspondence,        most cases were eliminated.
                  None of these problems could have been or were anticipated                          when
                  projections       were made for 1987.
Now p 35                 Page 52 -- GAO states     that "IRS is implementing         an IRP
See comment 17.
                  Management Information      System."     IRP MIS m       implemented    in
                  January 1987.    While it continues        to undergo refinements      and
                  modifications,    Examination    considers    it to be operational.        When
                  Tax Year 1986 cases (the first        tax year on the system) have been
                  worked to their    completion,     IRP MIS will    be able to address the
                  revenue impact of working underreporter           cases.

                                  LACKS CAPAB;U+ITY TO EVALUATE REJATIVE MEPITS OF
                                  em


                  DOPOSED INCREASES IN ENFORCEMENTSTAFFING
Now p 37                 Page 57 -- The report          contends that "ERAM is not being used
                  by the IRS as intended."             The EXAM itself      was never intended      to
See comment 18    be used in every day budget development.                  In particular,     the
                  yield    curves were only intended            to be used for general      estimates
                  of yield,      for quick estimating,          and hence, were updated only
                  annually.        Requests for more precise         estimates    are therefore
                  directed     to the enforcement         functions   in order to use the latest
                  assumptions       and other current        input data.      The ERAM costing     and
                  yield    subroutines        (percent  collected,    revenue flow factors,        and
                  support     and follow-on        cost factors)    were adopted by the
                  enforcement       functions      and are used in normal budget




                             Page64                                       GAO/GGD96-65IRS'EnforcementRevenues
                        Appendix II
                        CommentsFromtheIntemalRevenueService




                                                               6
                 development,    but    the model     itself       is    not used in the normal
                 budget process.
Now p 38                Page 59 -- The report            states    that IRS does not have marginal
                 yield    and marginal     cost data for IRp.              This is not correct     for
See comment 19   IRP.     While IRS does have marginal               yield    data for underreporter
                 cases on an individual          basis,       it does have marginal       yield   data
                 within    the categories       of IRP issues.             This permits   IRS to rank
                 the almost 200 issue categories                 according     to yield and average
                 cost ratio       when deciding      which cases to work first.             In our
                 opinion,     this marginal      yield      data permits       IRS to develop
                 reasonable       and defensible       revenue estimates.

                 c            --
                 REVENUE DATA ARE COMPREHENSIVE AND APPROPRIATELY COORDINm

                 Recommendations       to the   Commissioner            of Internal      Revenue
                       We recommend that        IRS
                       --     Develop a comprehensive           strategy   for identifying   and
                              coordinating     the various       enforcement   revenue projects,
                              and integrating      their     results.    As part of the
                              strategy,     IRS should ensure that revenue and cost data
                              on all enforcement         cases, including     Collection   and
                              Information     Returns Program cases, are incorporated
                              into EMIS.
                        --    Assign responsibility       at the Deputy Commissioner   level
                              for overseeing      development   and implementation  of the
                              strategy.
                        --    Use the results    of its various               projects     to improve   the
                              enforcement   revenue estimating                process.
                        We agree that there should be a comprehensive       strategy   but
                 do not agree that revenue and cost data from IRP necessarily           must
See comment 20   be included     in EMIS.    IRP MIS captures the necessary    revenue
                 information     and appropriate   cost data can also be readily
                 obtained.     IRS will   ensure that the output   from these information
                 systems are included      with EMIS data, while avoiding    double
                 counting.
Seep   51               We believe     oversight        responsibility    properly     rests with the
                 Controller.       As you know this position           was created       at the
                 recommendation      of GAO to play a more prominent               role in IRS'
                 budget formulation,          including      revenue estimation       and will   provide
                 adequate executive         oversight       of this area.     The newly chartered
                 Controller     organizat'ion       will    be the focal point for establishing
                 both administrative          and revenue accounting         standards.       For the




                        Page65
                            AppendizII
                            Comments From the Internal Revenue Service




-.

                                                                 7
                    first    time, an organization           will    have Service-wide    accounting
                    responsibility       for operational          purposes and for management
                    information      purposes.        Tracking     actual  enforcement    collections
                    will   be a responsibility           of this organization        as well.     The
                    Controller     reports     directly      to the Chief Financial       Officer     who is
                    also the Deputy Commissioner               (Planning   and Resources).
                           We believe   that these organizational     changes              will   permit
                    much more control      and coordination    of the revenue              estimating
                    throughout     the Service.
Seep    51                  Finally,     it should be noted in the report        that the
                    Commissioner        and the Senior Deputy Commissioner         have been
                    involved       in these issues on an ongoing basis.          As the Steering
                    Committee has dealt with issues that cannot or should not be
                    resolved       at their    level,     these issues are raised with the
                    Executive        Committee.       For these reasons we believe     that the
                    oversight        and executive      involvement   the report advocates    is
                    largely     in place today.
                    Other    Comments:
Nowp    43                  Page 68 -- The report's      statement     that *'IRS' current      plans
                    for EMIS call     for incorporating       and tracking     Examination    cases*'
See comment    21   by June 30 is incorrect.          The first   phase of EMIS will       include
                    data on Examination,      Employee Plans,       Exempt Organizations,
                    Collection's     Employment Tax Examination          cases, and Appeals cases,
                    i.e.,    all cases on the Audit Information           Management System
                     (AIMS).
 Nowp   44                 Page 68 -- The report             observes that 'la future        enhancement to
                    EMIS may provide          a direct     link"     to the IRP MIS.     In fact,      such
 See comment   22   enhancements a             provide     this link.        Statements  in this section
                    also question      whether IRP MIS will              **provide adequate
                    information.**       IRP MIS provides            the same detailed     information        on
                    IRP cases that EMIS will             provide       to non-IRP cases.      There is,
                    therefore,      no reason to combine the two systems but we will                      link
                    their    outputs   after      assuring       that there is no double counting.
                    Combining the systems would be costly                    and would provide      no
                    additional      benefits.

                    CHAPTER 7 --         UNRl&I.ABLE DATA ON ENFORCEMENTREVENUES HAS BROAD
                    IMPLICATIONS
                    Recommendations        to the    Senate    and House Budaet        and Aovrooriations
                    Committees:
                            The Senate and House Budget and Appropriations             Committees
                            should closely     oversee IRS' progress      in improving    its
                            enforcement    revenue data.     In doing so, we suggest that the
                            Committee require      IRS to provide   detailed   plans and




                            Page66                                           GAO/GGD90-85 IRS’ Enforcement Revenues
                   AppendixII
                   CommentsFrom the Internal Revenue Service




                                                        8
                  milestones      and periodic     status   reports.      Also, until   more
                  reliable     data is available,        the Committees should use
                  conservative       revenue assumptions        in deliberating    on future
                  enforcement      staffing    increases,     including     those being
                  proposed for fiscal        year 1991.
            IRS Response:
Seep   52         This recommendation        is a sound one.       However, as already
            noted, and as described         below in more detail,       it is important    to
            note that the IRS has made substantial            progress     over the past
            year in reviewing       its estimating    systems to assure that they are
            both conservative       and consistent.      We are very willing       to work
            with outside    reviewers     in further    sharpening     our techniques.
            However, any revisions        of these estimates       should always be based
            on analytical     techniques,      not on reactions     to IRS' own estimates.
                     The revenue estimates         for the FY 1991 IRS compliance
            initiatives          were based on the IRS' updated historical        data using
            estimating         methodologies    that incorporate  recent   internal     and
            external        findings    on ways to improve IRS' estimating.         The
            current       IRS methodologies,       in every case, are improved and differ
            significantly           from those used by the IRS in the time period FY
            1983-1988,         the period    of study in the GAO report.
                    The Examination        initiative        estimates    were based on known
            workload       in inventory      nationwide        and the most recent results          of
            audits      on a class-by-class           basis.     In most cases the interim
            method was used and, where appropriate,                     opportunity    costs were
            included.         The interim      method bases estimates           of recommended tax
            and penalties          on the number of new cases started               as a result     of
            adding additional           examination       personnel.      The former method based
            estimates        on the number of cases closed as a result                  of additional
            resources.         Given the time some audits              can take, this is a
            significant        difference.         The former method used revenue flows
            derived      from 1972 data that is now outdated.                   The interim     method
            uses the IRS' most recent experience                    of the time cases spend in
            the appeals process -- a factor                  that significantly       extends    the
            flow of collections.
                   A few of the Examination        initiatives       were unique and could
            not be estimated     through the interim           method.   The estimates     for
            these initiatives      were calculated         using known data for the unique
            aspects of each initiative        and, while they necessarily           involved
            the making of assumptions,        efforts       were made to assure the
            estimates     were conservative.
                   While there were no FY 1991 initiatives      in the Information
            Returns function,   the IRP methodology     has also been improved
            since the period   studied   in the GAO report.     Beginning   with FY
            1989, a model has been used to determine        the revenue likely     to
            result   from each year's  planned IRP activities.       The model now




                   Pagc67                                           GAO/GGD!XM5 IRS’ Enforcement Revenues
                             AppendixII
                             Comments From the Internal Revenue Service




                                                                                                                       1
                                                                     9

                      ranks almost 200 issues on a marginal       basis and takes into
                      account the amount of time that will      have to be spent on each
                      issue and the resources   available   to do the work.     The model's
                      outputs  are based on actual    data beginning    with Tax Year 1985.
                               The Collection       function         has also been improving          its
                      methodology.         A new model for Collection                 is now in the
                      contracting      stage that,        upon completion,            should greatly      improve
                      the Collection         estimating        procedures.        The estimates       for the FY
                      1991 Accounts        Receivable        initiative      were based on the most recent
                      experience     of nationwide           Collection      activity       in the TDA area.
                      The estimates        included     the use of Collection               Research File data
                      to validate      assumptions        and the use of RWWSscores to determine
                      the cases that would be worked.                     In addition,        the estimates    were
                      discounted     more heavily          for time spent in training              than had been
                      done previously.           As a result          the estimates       for the FY 1991
                      initiative     are more realistic               than estimates        made in prior     years.
                                We would also like to see the report               recognize       that the IRS
                      proposes funding            and staffing    increases     not only to increase
See comment 1         direct      enforcement       revenues,    but also to further          voluntary
                      compliance        in our self-assessment          tax system.        Although      more
                      difficult       to document, most revenue estimators                 believe     there is a
                      substantial         indirect     effect  of additional       enforcement        efforts  on
                      voluntary       compliance.         It would be helpful        if GAO could examine
                      and advise Congress on the importance                  of increases        in voluntary
                      compliance        that are a by-product          of increases      in enforcement.

                                          COMMENTSBY THE OFFICE OF TAX ANAT,YSIS
                      EXECUTIVE S-
                      Backsround
See comment 23.               Page 2 -- The second paragraph            states     that Treasury        Wses
                      the revenue estimates          for proposed staffing           increases      in
                      projecting     total   federal     tax receipts."         This is not accurate.
                      Together    with the IRS, the Office            of Tax Analysis          (OTA) develops
                      estimates    of incremental        enforcement      receipts       associated     with
                      proposed changes in IRS staffing.                These incremental          receipts   are
                      added to 0TAl.s baseline          estimate    of total     federal      tax receipts.
                      However, in preparing          the baseline      estimates       themselves,
                      estimates    of direct     enforcement      receipts      are not needed or used
                      by OTA.




Now pp 15-16               Page 16 -- The report              states that in Chapter 2 it will
                      document OTA's involvement              in determining total  enforcement




                  -



                              Page 68                                            GAO/GGD90-85 IRS’ Enforcement Revenues
                          Appendix II
                          CbnmentaFromtheInternalRevenueService




                                                                  10
See comment 24    revenues.     OTA has never attempted     to estimate       total  enforcement
                  receipts.     OTA does not require    or use this      information     to carry
                  out its responsibility     to estimate    total    budget receipts.        This
                  information     is neither necessary   nor used by OTA in evaluating
                  estimates   of incremental   receipts   attributable        to enforcement
                  program initiatives.
Now p. 16                Page 17 -- OTA does not estimate               total     or overall
                  enforcement   revenues.
See comment 25.


Nowp   18                 Page 20 -- The report       states,     "IRS does not have the
See comment 26    information      systems it needs to tally           actual   revenues generated
                  through     its enforcement    programs.        Absent those systems,           IRS and
                  OTA have to estimate        the actual      results     of IRS' enforcement
                  efforts".       This is misleading.         Moreover,      to the extent      IRS
                  enforcement      efforts  lead to greater         voluntary     compliance       (which
                  is generally       agreed to be the case, although            difficult     to
                  measure) , only an estimate         of total      enforcement       revenues can ever
                  be available.
Nowp   18                 Page 20 -- The report  states   that IRS and OTA used three
See comment 27.   different     data sources to calculate    actual   revenues from IRS'
                  enforcement     programs.  As discussed    above, OTA does not calculate
                  actual    revenues from IRS' enforcement      programs.
Nowp   18.               Page 21 -- The report           says 'IOTA analyzed          other data for
See comment 28    fiscal    year 1987 which led it to conclude                 that total     enforcement
                  revenues that year were about $24 billion."                       OTA does not
                  estimate    total     enforcement      receipts.        While understanding        there
                  were obvious reasons why the S-2 Report may not be a good
                  indicator     of total      enforcement      revenue,     OTA asked IRS to
                  reconcile     its estimate        of total     enforcement      receipts    with
                  information       found in the S-2 Report.              OTA did not produce its own
                  estimate,     nor did it view the $24 billion                shown in the S-2 Report
                  as a measure of the total            direct      enforcement      revenues.
Nowp   19               Page 22 -- The report  discusses   "the 24.6 billion    derived
See comment 29    for the year  using OTA’s methodology.lV      As noted above,  OTA does
                  not have a methodology   for estimating    total  enforcement  revenues.
Now p. 19                Page 22 -- The term "budget estimate"   is ambiguous.   It
See comment 30    should be made clear this $50 billion     is a number in IRS' budget
                  documents and not an estimate    of a component of federal   budget
                  receipts.
Now pp. 19-20             Page 23 -- As noted above,              OTA does not make estimates        of
See comment 28    total    enforcement receipts.
Now p 21                 Page 26 -- The report           says, "the     wide     variance     between
                  estimates   and IRS' recent           calculations      of    actual    revenues give




                          Page69                                          GAO/GGD9@85 JRS'EnforcementRevenues
                       Appendix II
                       GxumentaFkomtheIntemalRevenueService




                                                            11
                 little    assurance    that Congress can rely on [IRS' estimates                  of
                 revenues from a given level            of staffing.]"        While this     "wide
See comment 31   variance"     in estimates     of total     direct    enforcement      revenues
                 exists,    it should be clarified         that such evidence          does not
                 support corresponding         conclusions      regarding     estimates     of
                 incremental     revenues associated         with increases        in IRS staffing.
                 The methodology      used by OTA and IRS to calculate               the revenue
                 effects    of additional      IRS staffing       does not rely on estimates          of
                 total   enforcement      revenues.
Now p 21                Page 27 -- The report        recommends that IRS and CTA should be
                 directed    to provide    information     on actual  revenues generated by
See comment 5    IRS enforcement      programs.      As previously   mentioned, this
                 information    can only be provided        by IRS.




                        Page70                                          GAO/GGD9OS6IRS Enforcement Revenues
--
                  Appendix II
                  Comments From the Internal Revenue Service




                  1. We did not attempt to examine the indirect revenue effects of IRS’
GAO Comments on   enforcement programs (e.g., the potential revenue gains from increased
IRS’ Letter       voluntary compliance with tax laws) because both IRSand Treasury
                  have acknowledged that the indirect effects are difficult, if not impos-
                  sible, to measure.

                  2. We revised the wording on pp. 3 and 8 to delete references to
                  “overstated.”

                  3. We revised the wording on p. 3 along the lines recommended by                       IRS.

                  4. Based on the work that we have done, we have no basis for asserting
                  that IRSenforcement staffing initiatives will result in substantial direct
                  and indirect revenue increases.

                  5. We revised the wording of the recommendation on p. 21 to direct that
                  IRS, not IRS in conjunction with CTI’A, furnish Congress with information
                  on actual enforcement revenues.

                  6. The Direct Enforcement Revenue Study referred to by                  IRS   is the
                  master file analysis that is discussed on p. 19.

                  7. Although EMIS is expected to provide actual data on enforcement
                  cases, it will not by itself accomplish IRS’ goal of improving documenta-
                  tion and monitoring of revenue estimates.

                  8. Regarding IRS’comments on Collection’s efforts to improve its esti-
                  mating process: (1) Collection had no documentation showing that data
                  from the Collection Research File was used to validate its revenue esti-
                  mating assumptions, (2) the model mentioned by IRSis discussed on p.
                  54, and (3) Collection’s “exploration” of using data from RWMS (which
                  is IRS’system for scoring Collection cases on the basis of their potential
                  yield) to improve estimates was not sufficiently developed or docu-
                  mented to warrant inclusion in the report.

                  9. These comments, which relate to           IRS’   fiscal year 1991 budget esti-
                  mates, are discussed on p. 52.

                  10. We believe that we have appropriately acknowledged and treated
                  the discussion of IRS’ judgmental adjustments to its revenue estimates.
                  Since IRS did not document the results of its historical trend analyses
                  and judgmental adjustments, it could not support (and thus we could not
                  evaluate) the factors mentioned in this comment.


                  Page 71                                             GAO/GGD-W-86 IRS’ Enforcement Revenues
Appendix II
Comments From the Internal Revenue Service




11. Our report makes the point that IRSdoes not have a database that
links Collection staffing data on cases to the results of those cases in
terms of revenue. IRS' comments seem to acknowledge that point.

12. We could not assess the adequacy of IRS’ implementation of this rec-
ommendation because sufficient information was not available when we
did our audit.

13. Documentation available for the fiscal year 1989 estimates was
insufficient to assess the basis for those estimates because, among other
things, it omits data on assumptions about average yield per case for
each case category.

14. As noted in comment 13, the documentation available for fiscal year
1989 was incomplete. Also, the model documentation and source docu-
ments that were retained and used in constructing the fiscal year 1989
model were not adequate or useful for assessing the model’s assump-
tions and results. Neither the source documents nor the available model
documentation show what data IRS extracted from the source documents
or how the model manipulated that data.

15. As IRS says in its comment, the comparison it made using tax year
1985 data was not meant to validate the fiscal year revenue estimate
but rather to validate the model’s construction. Our report, however,
does not criticize the model’s construction; we say on p. 32 that the
model appears to provide a more refined basis for developing revenue
estimates. Our point is that IRS needs to obtain actual results on a fiscal
year basis and use that information to monitor the reliability of its fiscal
year estimates.

16. We revised the wording on pp. 32 and 35 to indicate that we are
talking about the situation at the time of our review. We discuss fiscal
year 1991 estimates on p. 52.

17. We revised the wording on pp. 32 and 35 to indicate that the system
was implemented.

18. Our conclusion that ERAM is not being used as intended is based on
information in various IRS documents and discussions with IRS officials
as to why ERAM was developed and how it is being used. Our report does




Page 72                                      GAO/GGD-90-85 IRS’ Enforcement Revenues
Appendix Lf
Comments From the Internal Revenue Service




not state that ERAM was intended to be used in “everyday budget devel-
opment.” It states that IRS developed ERAM to estimate the costs and rev-
enues associated with an enforcement staffing increase and to optimally
allocate additional enforcement staff to maximize revenue.

19. Marginal yield is defined as the change in total yield arising from
each additional case worked. IRS’ revenue data, even when broken down
by category, does not constitute marginal yield data.

20. We revised the wording of the recommendation on p. 47 to allow for
either incorporating IRP data into the EMIS database or linking the IRP and
EMS data.

21. We revised the wording on p. 43 to show that Examination cases are
not the only ones scheduled to be incorporated into EMIS by June 30,
1990.

22. Documents recently provided to us by IRS state that the IRP Manage-
ment Information System and EMIS may be linked. We revised the recom-
mendation on p. 47 to allow for such linkage.

23. We revised the wording on p. 2 to more correctly state how Treasury
uses the estimates.

24. We revised the wording on pp. 15-16 to more clearly characterize
or..‘s involvement.

25. Although OTA does not, as a routine matter, estimate total enforce-
ment receipts, it did use the S-2 report to calculate those receipts as a
check on IRS’ estimate. We revised the wording on p. 16 to avoid using
the word “estimate.”

26. This statement is not misleading in that it is a straightforward
description of information that IRS does not have. We have clarified OTA’S
role in calculating enforcement revenues. See also comments 25, 27, and
28.

27. OTA did calculate a figure for such revenues using the S-2 report.

28. We agree that OTA does not, as a routine matter, estimate total
enforcement receipts. It did, however, make the analysis discussed in
our report.



Page 73                                      GAO/GGDSO-BS IRS’ Enforcement Revenues
Appendix II
Comments From the Internal Revenue Service




29. We changed the wording on p. 19 so as not to refer to a
“methodology.”

30. We changed the wording on p. 19 to delete reference to “budget
estimate.”

31. We do not think further clarification is necessary. We do not imply
that our conclusion relates to increases in staffing. Our analysis, as
stated in the report, addresses total enforcement revenues from the
given staffing level-not additional revenues from incremental staffing
increases.




Page 74                                      GAO/GGD90-86 IR!3’ Enforcement Revenues
mndix      III

Major Contributors to This Report


                        David J. Attianese, Assistant Director, Tax Policy and
General Government        Administration Issues
Division, Washington,   Rita A. Grieco, Evaluator-in-Charge
D.C.                    Victoria E. Miller, Evaluator
                        Monica L. Pickens, Secretary




(268418)                Page 76                                GAO/GGD-SO-g5   IRS’ Enforcement   Revenues