Benefit and Loan Programs: Improved Data Sharing Could Enhance Program
Integrity (Letter Report, 09/13/2000, GAO/HEHS-00-119).

Pursuant to a congressional request, GAO provided information on data
sharing among federally funded benefit and loan programs, focusing on:
(1) the data that selected benefit and loan programs use to verify
self-reported, eligibility-related information; (2) whether additional
data sharing with federal, state, and local governments, or private
entities, could avoid some improper payments; and (3) the legal or other
barriers to enhanced data sharing.

GAO noted that: (1) federally funded benefit and loan programs require
similar types of information about individuals to correctly determine
their eligibility for assistance; (2) such information includes their
identity, earned income and unearned income, assets, citizenship status,
and household composition; (3) while numerous factors can affect an
individual's eligibility for benefits, income is generally one of the
most important and most prone to error or inaccurate reporting; (4) some
programs obtain information from independent, third-party sources such
as federal and state agencies, or private companies such as credit
bureaus to verify the accuracy of self-reported data; (5) the three
programs GAO reviewed could use enhanced data sharing to make more
timely and accurate eligibility determinations; (6) other legal
restrictions, as well as management, administrative, and technological
challenges, limit the ability of federally funded benefit and loan
programs to effectively share information with one another; (7) a number
of laws have been enacted over the past 25 years that limit access to
sensitive data sources in an effort to protect individual privacy and
the confidentiality of sensitive information or to address concerns
about taxpayer compliance with tax laws; (8) these statutes include: (a)
section 6103 of the Internal Revenue Code (IRC), which governs the
disclosure of taxpayer information; (b) provisions in the Social
Security Act that restrict access to the Office of Child Support
Enforcement's (OCSE) National Directory of New Hires; and the (c)
Privacy Act, which balances the government's need to collect and
maintain sensitive information about individuals against their right to
privacy; (9) providing more federally funded programs access to
restricted data sources and the ability to share this information with
state and local agencies that administer various benefit and loan
programs often requires amending federal laws governing the use of the
data; and (10) however, increasing the access to sensitive data for
benefit and loan programs can be balanced with the need for personal
privacy, confidentiality, and tax compliance.

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

 REPORTNUM:  HEHS-00-119
     TITLE:  Benefit and Loan Programs: Improved Data Sharing Could
	     Enhance Program Integrity
      DATE:  09/13/2000
   SUBJECT:  Federal aid programs
	     Social security benefits
	     Interagency relations
	     Eligibility determinations
	     Public assistance programs
	     Student financial aid
	     Computer matching
	     Confidential communication
	     Internal controls
	     Right of privacy
IDENTIFIER:  HHS Temporary Assistance for Needy Families Program
	     Food Stamp Program
	     HUD Section 8 Housing Assistance Program
	     Pell Grant
	     Federal Family Education Loan Program
	     William D. Ford Federal Direct Loan Program
	     OCSE National Directory of New Hires

******************************************************************
** This file contains an ASCII representation of the text of a  **
** GAO Testimony.                                               **
**                                                              **
** No attempt has been made to display graphic images, although **
** figure captions are reproduced.  Tables are included, but    **
** may not resemble those in the printed version.               **
**                                                              **
** Please see the PDF (Portable Document Format) file, when     **
** available, for a complete electronic file of the printed     **
** document's contents.                                         **
**                                                              **
******************************************************************
GAO/HEHS-00-119

Appendix I: Comments From the Administration for Children and Families

50

Appendix II: Comments From the Department of Education

57

Appendix III: Comments From the Department of Housing and
Urban Development

58

Appendix IV: Comments From the Internal Revenue Service

60

Appendix V: Comments From the Office of Management and
Budget

64

Appendix VI: Comments From the Department of the Treasury

66

Appendix VII: Major Contributors to This Report

69

Table 1: Potential Sources of Data for the Three Programs in Our
Review 14

Table 2: Some Major Federal, State, and Private Data Sources 16

ACF Administration for Children and Families

AFDC Aid to Families with Dependent Children

BSRT Benefit Systems Review Team

CMPPA Computer Matching and Privacy Protection Act

DMDC Defense Manpower Data Center

DMV Department of Motor Vehicles

FAFSA Free Application for Federal Student Aid

HEA Higher Education Amendments of 1998

HHS Department of Health and Human Services

HUD Department of Housing and Urban Development

IEVS Income and Eligibility Verification System

IRC Internal Revenue Code

IRS Internal Revenue Service

OASDI Old Age, Survivors, and Disability Insurance

OCSE Office of Child Support Enforcement

OIG Office of Inspector General

OMB Office of Management and Budget

PARIS Public Assistance Reporting Information System

PHA public housing authority

PMO priority management objective

PRWORA Personal Responsibility and Work Opportunity Reconciliation Act of
1996

QHWRA Quality Housing and Work Responsibility Act of 1998

REAC Real Estate Assessment Center

SAVE Systematic Alien Verification for Entitlement

SESA state employment security agency

SSA Social Security Administration

SSI Supplemental Security Income

TANF Temporary Assistance for Needy Families

Health, Education, and
Human Services Division

B-283206

September 13, 2000

The Honorable Fred Thompson
Chairman, Committee on Governmental Affairs
United States Senate

The Honorable Joseph Lieberman
Ranking Minority Member, Committee on Governmental Affairs
United States Senate

Many federally funded benefit and loan programs rely on applicants and
current recipients to accurately report their own important information,
such as the amount of income they earn, that affects their eligibility for
assistance. To the extent that such information is underreported or not
reported at all, the federal government overpays benefits or provides loans
to individuals who are ineligible. In recent years, reviews that we have
conducted and that offices of inspectors general (OIG) and others have
conducted demonstrate that federally funded benefit and loan programs such
as housing and higher education assistance have made hundreds of millions of
dollars in improper payments.1 Some of these payments were made improperly
because the federal, state, and local entities that administer the programs
sometimes lacked adequate, timely data needed to determine applicants' and
current recipients' eligibility for assistance. Our previous work has
demonstrated that improper payments can be avoided or detected more quickly
by using data from other programs, or maintained for other purposes, to
verify self-reported information. For example, we estimated that direct
on-line connections between the Social Security Administration's (SSA)
computers and databases maintained by state agencies containing information
on wages, welfare benefits, unemployment insurance, and workers'
compensation benefits could have prevented or more quickly detected $131
million in Supplemental Security

Income (SSI) overpayments in one 12-month period.2 We also estimated that
about $648 million in SSI overpayments that occurred in one year could have
been avoided or more quickly detected if SSA had had access to the Office of
Child Support Enforcement's (OCSE) National Directory of New Hires, as well
as financial institution data.3

Data sharing, as we use the term in this report, means obtaining and
disclosing information on individuals from independent, third-party sources
such as federal and state government agencies or private organizations to
determine their eligibility for federally funded benefit and loan programs.
Commonly used data sources include tax return information from the Internal
Revenue Service (IRS), earnings information maintained by SSA, and
unemployment insurance data maintained by state employment security
agencies. You asked us to study whether improved data sharing among
federally funded benefit and loan programs could help them make more
accurate initial and continuing eligibility determinations. Specifically,
you asked us to determine (1) the data that selected benefit and loan
programs use to verify self-reported, eligibility-related information; (2)
whether additional data sharing with federal, state, and local governments,
or private entities, could avoid some improper payments; and (3) the legal
or other barriers to enhanced data sharing.

To respond to your request, we examined three federally funded benefit and
loan programs:4 (1) the Department of Housing and Urban Development's (HUD)
Public Housing and Tenant-based Section 8 programs, (2) the Department of
Health and Human Services' (HHS) Temporary Assistance for Needy Families
(TANF) program, and (3) the Department of Education's Student Financial
Assistance programs. These programs represent three different modes of
program administration: direct federal administration (Student Financial
Assistance), block grants with state and local administration (TANF), and
local agency administration with some federal involvement (HUD housing
assistance). We interviewed federal officials from each of these programs as
well as from other federal agencies.5 We also interviewed state and local
officials who administer the three programs in eight locations: Sacramento
and San Francisco, California; Miami and Tallahassee, Florida; Baltimore,
Maryland; Newark, New Jersey; Albany and New York City, New York. In
addition, we collected available data from federal, state, and local
governments on improper payments and cost savings related to the use of data
sharing. We also cosponsored a symposium on data sharing with the Senate
Committee on Governmental Affairs in June 2000. This conference examined a
number of different issues related to data sharing, including its benefits
for program integrity, the role of technology, and challenges to enhanced
data sharing, as well as privacy and information security issues. Numerous
representatives attended the conference from federal agencies, as well as
state and local governments and private sector organizations.6 We performed
our work between June 1999 and June 2000 in accordance with generally
accepted government auditing standards.

Federally funded benefit and loan programs require similar types of
information about individuals to correctly determine their eligibility for
assistance. Such information includes their identity (name, Social Security
number, date of birth), earned income (wages) and unearned income (food
stamps), assets (bank accounts, automobiles), citizenship status, and
household composition. While numerous factors can affect an individual's
eligibility for benefits, income is generally one of the most important and
most prone to error or inaccurate reporting. Some programs obtain
information from independent, third-party sources such as federal and state
agencies, including IRS or state unemployment insurance programs, or private
companies such as credit bureaus to verify the accuracy of self-reported
data. For example, some of the public housing agencies we visited obtained
information from national credit bureaus such as Equifax to verify
applicants' and recipients' financial resources. Matching automated computer
files and accessing online databases are two methods by which federally
funded benefit and loan programs obtain and share eligibility information.
In our earlier reviews of the SSI and Food Stamp programs, we noted the
efficacy of using federal and state data sources to verify self-reported
information and control improper program payments.

The three programs we reviewed could use enhanced data sharing to make more
timely and accurate eligibility determinations. While each of the programs
uses varying degrees of computer matching and other methods to verify the
information that applicants and current recipients provide, the programs
could benefit from access to additional data sources. For example,

ï¿½ HUD estimates that the lack of adequate information on applicants' and
tenants' income contributed to $935 million of excess rental subsidies in
1998.7 While federal HUD officials have access to data from various computer
matches (such as those conducted with IRS taxpayer data), under current law
they are not permitted to share federal tax information with public housing
agencies and private owners who administer HUD's rental assistance
programs.8

ï¿½ The Department of Education's OIG estimates that underreported income
contributed to roughly $109 million in excess Pell Grant awards in 1995-96.
Access to IRS taxpayer information could have helped Education prevent some
of these overpayments.

ï¿½ The Administration for Children and Families (ACF) within HHS estimates
that program savings of about $100 million could be realized in the TANF
program if all states participated in the Public Assistance Reporting
Information System (PARIS) match.9

An example of an information source that many program administrators cite as
being beneficial is OCSE's National Directory of New Hires. Access to this
source would provide these programs with the most timely and comprehensive
source of state wage, unemployment insurance, and new hire data available.
However, current law requires OCSE to establish and implement safeguards
designed to restrict access to confidential information to authorized
persons for authorized purposes and does not permit the three programs we
examined to obtain information from the directory for purposes of verifying
applicants' eligibility for benefits or loans.10

Other legal restrictions, as well as management, administrative, and
technological challenges, limit the ability of federally funded benefit and
loan programs to effectively share information with one another. A number of
laws have been enacted over the past 25 years that limit access to sensitive
data sources (or restrict how such data may be used) in an effort to protect
individual privacy and the confidentiality of sensitive information or to
address concerns about taxpayer compliance with tax laws. These statutes
include section 6103 of the Internal Revenue Code (IRC), which governs the
disclosure of taxpayer information; provisions in the Social Security Act
that restrict access to OCSE's National Directory of New Hires; and the
Privacy Act (including the Computer Matching and Privacy Protection Act
amendments of 1988, or CMPPA), which balances the government's need to
collect and maintain sensitive information about individuals against their
right to privacy. Providing more federally funded programs access to
restricted data sources (such as the National Directory of New Hires) and
the ability to share this information with state and local agencies that
administer various benefit and loan programs often requires amending federal
laws governing the use of the data.11 However, increasing the access to
sensitive data for benefit and loan programs can be balanced with the need
for personal privacy, confidentiality, and tax compliance. For example,
providing access to restricted data sources in this context does not mean
that personal information on citizens will be available in the public domain
or that the confidentiality of sensitive data will not be protected from
unauthorized disclosure. Rather, only agencies or their representatives with
the need to view the information for purposes of determining eligibility for
individuals who apply for benefits and loans should have access to such data
and would be responsible for protecting them from unauthorized disclosure.
In addition to legal restrictions, other issues complicate the ability of
benefit and loan programs to share information. For example, the lack of a
system through which the states can share data to detect individuals who
have obtained TANF benefits in more than one state makes it difficult for
states to ensure that only eligible individuals and families receive
benefits. This type of challenge will likely require federal facilitation to
address the management, administrative, and technological issues that are
involved. We are suggesting to the Congress and the heads of the Office of
Management and Budget (OMB), HHS, and HUD actions that will help address
some of these issues.

Federally funded benefit and loan programs provide cash or in-kind
assistance to individuals who meet specified eligibility criteria. TANF,
SSI, Food Stamps, housing assistance, and student loans are representative
of such programs. Some are administered centrally by federal agencies (such
as SSI), while states and localities are involved in the administration of
others (such as TANF). Many of these programs rely on applicants and current
recipients to self-report important information that can affect their
eligibility for benefits, such as the amount of income they earn. Because
many benefit and loan programs require the same information, it is more
efficient for them to share the necessary data with one another rather than
requiring each program to independently verify similar data. Sharing
important eligibility-related information across programs can also reduce
the burden on individual applicants and recipients, businesses, and other
entities for repeatedly supplying it. These programs may verify
self-reported information by comparing their records with independent,
third-party data sources from other federal or state agencies as well as
private organizations.

Benefit and loan programs can compare large amounts of information on
applicants and recipients by using computers to match automated records.
Electronic transmission of data and online access to agencies' databases are
additional tools program administrators can use to share important
information on applicants and recipients in a timely, efficient manner. If
used consistently, they can help program administrators check the accuracy
of individuals' self-reported statements, as well as identify information
relevant to eligibility that the applicants and recipients themselves have
not provided.

We discuss two types of accuracy checks related to data sharing in this
report: positive verification, or confirming the accuracy of information
that an applicant or recipient has voluntarily reported, and negative
verification, or checking independent, third-party data sources to detect
information that may not have been reported. The following example
illustrates the difference between the two concepts. To confirm the amount
of earned income that an applicant reports at an eligibility interview, a
public assistance program administrator sends a letter to an employer the
applicant cited and requests the employer to certify the applicant's wages.
This is a form of "positive verification" because the administrator is able
to confirm only employment and wage information that is voluntarily
provided. "Negative verification" in this example could consist of the
program administrator's accessing the central state wage database and
checking all the wages reported to the state by all employers on behalf of
the applicant. This type of verification is superior because it allows
program administrators to identify more relevant earned income information
for an individual, including information that is not volunteered on an
application.12 Moreover, sharing important data across benefit and loan
programs with similar information needs may result in more efficient program
administration, including more timely and accurate eligibility
determinations.

HUD funds and regulates the Public Housing and Tenant-based Section 8
programs, both of which local public housing agencies administer under
contract to HUD.13 The Public Housing Program provides subsidies to nearly
3,200 public housing authorities (PHA) to help them pay a large portion of
the cost of operating and maintaining public housing units. In fiscal year
1998, this program spent about $8 billion to subsidize 1.3 million public
housing units.14 The Section 8 Rental Assistance Program subsidizes
low-income families' choice to live in privately owned housing by providing
rental assistance payments to landlords on behalf of families. In fiscal
year 1998, HUD spent about $8.3 billion to help about 1.6 million families
obtain affordable, private housing. In general, tenants are required to pay
30 percent of their anticipated income toward rent, with HUD paying the
balance of the rental amount. The PHAs are responsible for verifying the
information applicants and current tenants provide that may affect their
eligibility for assistance. In contrast to entitlement program benefits,
which generally go to all who qualify, the benefits of HUD's housing
assistance programs are limited by budgetary constraints to about one-fourth
of the households that are eligible.

TANF was created by the Personal Responsibility and Work Opportunity
Reconciliation Act of 1996 (PRWORA) and replaced the federal entitlement
awarded to eligible families under the Aid to Families with Dependent
Children (AFDC) program with temporary benefits contingent on successful
completion of work and education requirements. TANF has been implemented in
the form of a block grant to states, designed to help low-income families
with children reduce their reliance on welfare and move toward economic
independence. The annual block grant to states is $16.5 billion. In fiscal
year 1999, states spent a total of $11.2 billion of federal funds and $11.3
billion of their own funds on this program.15 While HHS oversees TANF at the
federal level, states have much greater flexibility than under AFDC to
design and implement programs to meet state and local needs and to establish
eligibility requirements. However, among its other provisions, PRWORA
requires that states impose work requirements for adults, meet specified
requirements for the percentage of adults who participate in work
activities, and enforce a 5-year lifetime limit on receiving TANF benefits.
States and localities are also responsible for verifying the accuracy of
information that applicants and TANF recipients provide.

The Department of Education's Office of Student Financial Assistance for
higher education provides federal student aid in the form of grants and
loans to eligible applicants based on key financial and student status
information. The grant and loan programs include Pell Grants, the Federal
Family Education Loan Program, and the Federal Direct Loan Program.
Applicants report pertinent information on the Free Application for Federal
Student Aid (FAFSA) form. In general, Education relies on institutions (such
as universities) to verify the accuracy of this information. It also
conducts various internal and external computer edits to detect errors on
applications. Education currently oversees a total of about $51 billion in
federal student aid grants and loans to nearly 8.4 million students and
parents.

A number of different laws and regulations govern the ability of benefit and
loan programs to obtain access to and share sensitive information such as
the Social Security numbers of and income data on applicants and recipients.
Among the most important of these are the Privacy Act, including the
Computer Matching and Privacy Protection Act amendments of 1988 (CMPPA) and
1990, and IRC section 6103.

The Privacy Act (5 U.S.C. section 552a) is intended to balance the
government's need to collect and maintain information about individuals
against their right to be protected against unwarranted invasions of their
privacy resulting from federal agencies' collections and use of information
about them. The act addresses four basic policy objectives: (1) restricting
the disclosure of records that contain personal identifiers maintained by
agencies; (2) providing individuals the right of access to agency records
about them; (3) granting individuals the right to seek amendment of agency
records on them if they can show that such records are not accurate,
relevant, timely, or complete; and (4) establishing a code of fair
information practices that requires agencies to comply with statutory norms
for collecting, maintaining, using, and disseminating records. During the
June 2000 data sharing symposium, some observers stated that the act was
primarily designed to address privacy concerns in a paper environment. As
such, it is not well suited to addressing some of the issues that arise from
the widespread use of computerized records and electronic dissemination of
information and may need to be updated.

CMPPA permits federal agencies to conduct matches with one another, subject
to various provisions. In general, all matching agreements must include the
purpose and legal authority for a match, anticipated results, a cost-benefit
analysis, procedures to notify applicants and recipients who are identified
in a match, and procedures for verifying the information to be produced. In
addition, CMPPA (see 5 U.S.C. sect.sect. 552a(o), (r), and (u)) requires agencies to
publish matching agreements in the Federal Register, report matching
programs to the Congress and OMB, and establish internal data integrity
boards to review and approve matching programs. An initial matching
agreement may be in force for 18 months, which may be followed by a 12-month
renewal period. To continue a match beyond 30 months, a new agreement must
be negotiated between the agencies, even if the provisions of the matching
agreement remain the same.

Under IRC section 6103, federal, state, and local agencies may receive IRS
taxpayer information only for certain specified purposes, including
administering state tax programs and verifying eligibility for various
welfare and public assistance programs (such as public housing).16 According
to the Department of the Treasury, this section of IRC was enacted to limit
the instances in which other agencies used tax data for nontax purposes to
cases in which the need for the information outweighs the related concerns
of taxpayer privacy and continued compliance with tax laws. Section
6103(l)(7) specifies that taxpayer information may be disclosed to federal,
state, or local agencies administering any of nine program categories, such
as unemployment compensation or housing assistance.17 Agencies are also
required to protect the confidentiality of the information they receive and
to implement safeguards that are designed to prevent unauthorized access,
disclosure, and use. Unauthorized disclosure or access may result in
criminal penalties (section 7213) or civil damages (section 7431).
Individual taxpayers may also authorize IRS to disclose their tax return
information to specified entities by providing a written consent (IRC
section 6103(c)). However, such disclosures are not subject to the same
safeguards or penalties that exist for disclosures to specified agencies
under IRC section 6103.

Many federally funded benefit and loan programs (including the three
programs in our review) require common information on applicants and
recipients to accurately determine their eligibility for benefits. These
data include a person's identity (name, date of birth, Social Security
number), address, income (both earned and unearned), assets (such as bank
accounts), household composition (including number of dependents),
citizenship status, incarceration status, and whether the person has died.18
Applicants' and recipients' income are the most vital pieces of information
in making an accurate decision, because income is often a main requirement
for an individual's (or family's) eligibility. In addition, assets such as
bank accounts can affect applicants' and current recipients' eligibility for
assistance. In many instances, program administrators rely on applicants and
recipients to self-report important information, such as income, which can
be subject to error and abuse. Administrators of some benefit and loan
programs may gain access to records maintained by other federal, state, or
local agencies by means of computer matching or on-line access to "live"
databases to verify the accuracy of self-reported statements. However, they
do not consistently use such sources to make timely and accurate eligibility
determinations. Table 1 shows potential sources for some commonly required
types of data, including federal, state, local, and private entities.

 Type of data
 needed         Federal         State         Local           Private
 Identity
 (name, Social
 Security       SSA
 number, birth
 date)
                --Immigration
                and
 Citizenship    Naturalization
 status         Service

                --SSA
                --IRS           --Department
                                of motor
                --OCSE,         vehicles
 Address        National        (DMV)                         Credit
                Directory of                                  bureaus
                New Hires       --State
                                directory of
                --SSA           new hires
 Household
 compositiona
                                --State
                                employment
                --IRS           security
                                agency
                --OCSE,         (SESA)
 Earned income  National
 (wages)        Directory of    --Department
                New Hires       of labor

                --SSA           --State
                                directory of
                                new hiresb
                                --Department
                                of human
                                services or
 Unearned                       the           Various
 income (public                 equivalent    agencies (such
 assistance     Various                       as housing      Financial
 benefits,      departments and --Department  authorities)    institutions
 pensions,      agenciesc       of labor or   administering
 interest)                      the           federal and
                                equivalent    state programs

                                --SESA
                                                              --Credit
                                              Various         bureaus
 Assets (bank                   --DMV         agencies
 accounts,      IRS                           maintaining     --Data
 automobiles)                   --Lottery     property        vendors
                                agencies
                                              records
                                                              --Financial
                                                              institutions
 Incarceration  --Department of Prisons and
 status and     Justice         criminal
 criminal                       justice       Jails           Data vendors
 history        --SSA           agencies

 Death                          Bureau of
 information    SSA             vital                         Data vendors
                                statistics

Note: The data and sources are not exhaustive but illustrate some of the
more common types of data needed, as well as some of their potential
sources.

aAlthough no single source of data is generally available to verify
household composition, program administrators may use a combination of
sources that may provide limited information, including state bureaus of
vital statistics, local marriage license data, prison match data from SSA,
and local school attendance forms.

bPRWORA required states to set up databases of new hires to help locate
individuals involved in child support cases and to verify eligibility for
certain programs. See 42 U.S.C. 653(h).

cFederal departments and agencies that maintain information on unearned
income, such as public assistance benefits, include HUD, IRS, SSA, and the
Veterans Administration.

A relatively small number of federal and state agencies as well as private
companies maintain substantial information that is required by many
federally funded benefit and loan programs. For example, IRS is a main
source for taxpayer information on earned and unearned income, including
wages, interest from bank accounts, and investment income.19 OCSE's National
Directory of New Hires also contains important information for all 50
states. The directory includes centralized sources of state wage data,
unemployment insurance, and new hires data that could help benefit and loan
programs make more timely, accurate eligibility determinations. During the
data sharing symposium we cosponsored in June 2000, several representatives
from federal, state, and local government agencies inquired about how they
could obtain access to information contained in the directory. However,
provisions in the Social Security Act specify that in addition to OCSE, only
three agencies (IRS, SSA, and the Department of Education) may use
information maintained in the directory, and they may use it only for
specific purposes. The data in the directory are not legally accessible to
most benefit and loan programs.

At the state level, one of the main sources of information for data sharing
is the Income and Eligibility Verification System (IEVS).20 IEVS draws on
numerous federal and state data sources, including IRS Form 1099 data and
Social Security benefits and earnings information, as well as state wage and
unemployment insurance data. PRWORA also required states to continue using
IEVS in administering their TANF programs.

Another source of valuable information (particularly for verifying
applicants' or recipients' assets) is private financial institutions such as
banks. However, only two federal agencies that administer benefit or loan
programs (OCSE and SSA) currently have legislative authority to request data
from financial institutions.21 For example, the Foster Care Independence Act
of 1999 authorizes SSA to require applicants and current recipients of SSI
benefits to provide SSA with the authority to check the records of all
financial institutions as part of its eligibility determination process.
However, most federally funded benefit and loan programs do not have this
type of authority. Table 2 describes some of the types of data that may be
available from some selected federal, state, and private sources.

 Source                                       Description
                                              Federal taxpayer information,
                                              including individual income
                                              tax, Form W-2 (wage
                 IRS Individual Master File   statements received from
                 and related databases        SSA), Form 1099 (unearned
                                              income), taxpayer name,
                                              mailing address, and marital
                                              and tax filing status.
                                              Quarterly state wage data,
 Federal agenciesOCSE, National Directory of  new hires data, and
 and systems     New Hires                    unemployment information from
                                              all 50 states and the
                                              District of Columbia.
                                              Form W-2 (wage statements);
                                              Social Security number
                                              verification; Old Age,
                 SSA databases                Survivors, and Disability
                                              Insurance data; SSI data;
                                              death information; and
                                              prisoner data from states and
                                              localities.
                                              States use this mechanism to
                                              compare data that applicants
                 IEVS                         and recipients of welfare
                                              programs (TANF, Food Stamps,
                                              and Medicaid) supply with
                                              various federal data sources.

 State agencies  State directory of new       New hires data from states.
 and systems     hires
                 Employment security          Quarterly wage data,
                 agencies                     unemployment information.

                 Bureaus of vital statistics  Births, deaths, marriages,
                                              and divorces.

                 DMV                          Address, asset information
                                              (such as automobiles).

 Private         Private data vendors,        Credit history, address
 companies       financial institutions, and  information, assets.
                 credit bureaus

Note: Access to some of these data sources, including but not limited to IRS
taxpayer data and information contained in the National Directory of New
Hires, is restricted to agencies specified in current law.

While each of the three programs we examined (HUD assisted housing,
Education's Student Financial Assistance loans and grants, and TANF) use
some data sharing to control their payments, weaknesses still exist. HUD's
housing assistance programs rely heavily on applicants and current tenants
to self-report important information, such as income that affects their
eligibility for assistance. This is particularly true during the initial
application process, yet little is done to check independent data sources
for information that applicants may omit or overlook. Inaccurate reporting
in these programs contributed to several hundred million dollars in excess
rental subsidies in 1998. Similarly, the Department of Education relies on
applicants for student financial aid to accurately report their income and
other important information. Excess awards of more than $100 million have
been documented in the Pell Grant program as a result of applicants'
underreporting their income. States we visited that are administering the
TANF program use numerous data sources to verify applicants' and recipients'
eligibility for benefits. However, state and local officials told us that
they currently lack a comprehensive, reliable source of data to track
duplicate receipt by recipients, program disqualifications, and time limits
imposed by welfare reform across state lines.

Accuracy of Eligibility Determinations

Despite HUD's improvements in the scope of its computer matching activities
to verify important information such as income, the Public Housing and
Tenant-based Section 8 programs still lack adequate information for making
accurate and timely eligibility determinations for applicants and
recipients. For example, in its fiscal year 1999 financial statements, HUD
reported $935 million in subsidy overpayments for calendar year 1998. 22
These improper payments are largely attributed to tenants' unreported or
underreported income. Given that the population that uses housing assistance
is a low-income population, it is unlikely that more than a small fraction
of these improper payments will be recovered. Moreover, subsidy overpayments
to some tenants may result in other eligible low-income families not being
served because of the limited availability of public housing units and
rental vouchers. Although HUD uses negative verification to verify Social
Security and SSI information for its rental assistance programs, it could
make more accurate and timely eligibility determinations by using the wage,
unemployment insurance, and new hire data maintained in OCSE's National
Directory of New Hires. However, current law does not permit HUD to access
this information source.

PHAs that administer the Public Housing and Tenant-based Section 8 programs
at the local level often do not have access to the third-party data sources
they need to make accurate eligibility decisions. For example, restrictions
in IRC section 6103 do not permit HUD to provide applicants' or tenants' tax
information to PHAs.23 Instead, the PHAs we visited rely extensively on
applicants and current tenants to self-report information vital to their
eligibility determination, such as income, assets, and household
composition.24 Most of the PHAs perform some form of positive verification
by following up with third parties to verify data reported by the applicant
or tenant. In general, applicants and tenants are required to have employers
or program administrators from other public assistance programs (such as
TANF) sign verification forms confirming the wages or benefits they reported
to the PHA. For example, the San Francisco PHA obtains signatures from
employers, financial institutions, and SSA representatives, confirming
income that applicants or tenants reported. Housing eligibility workers may
also telephone such entities if they suspect that the information provided
is incorrect. Officials in this and other PHAs told us that the current
process is helpful for verifying the accuracy of information that is
volunteered but cannot detect wages, bank accounts, or other vital pieces of
information that are intentionally concealed or overlooked by applicants and
tenants. For example, PHAs do not have independent access to potentially
helpful sources of asset information, such as financial institutions. In
addition, staff at most of the PHAs we visited said that they do not have a
reliable source of data to verify an applicant's previous address or rent
history. This information can be important for determining a family's
eligibility for housing assistance, because public housing and Section 8
tenants sometimes vacate their apartments once they are notified of an
income discrepancy during an

annual recertification rather than repay rent retroactively. 25 If such a
family applies for housing assistance at a PHA in another jurisdiction and
does not disclose its previous address or landlord, that PHA may not be able
to accurately determine the family's eligibility for assistance.

Some of the PHAs we visited have limited access to independent, third-party
databases for identifying unreported income or assets. For example, the
Baltimore, Hialeah, Newark, and San Francisco PHAs use national credit
bureaus such as Equifax to obtain information on applicants' and tenants'
assets and outstanding loans (which can provide additional information about
the financial resources the applicants have access to for their own
support).26 The Newark PHA also recently began implementing a new system
that provides access to state wage data over the Internet. A PHA official
demonstrated how this on-line system, once fully implemented, would provide
eligibility workers with real-time access to individual applicants' wage
histories and employers for several preceding quarters. In addition, the
Newark PHA has already used the new system to identify tenants who have
failed to report or who have underreported their income. For example, in a
recent sample of 161 families that claimed to have no income, the PHA found
76 (47 percent) with unreported income.27 The total amount of unreported
wages for these 76 cases was about $1.2 million during calendar years 1998
and 1999 and resulted in about $350,000 in back rent owed to the PHA. PHA
officials told us that they intend to use the system to verify earned income
for all applicants and current tenants.

Overall, PHA and HUD officials we interviewed emphasized the importance of
using timely, third-party data sources to verify applicants' and current
tenants' eligibility. In an effort to improve its ability to verify
unreported and underreported income, HUD has taken various steps to improve
its data sharing capabilities in recent years.28 For example, in 1999 HUD's
Real Estate Assessment Center (REAC) started to conduct its first
large-scale computer match using Form 1099 data provided by IRS and W-2 data
provided by SSA. 29 Of the 2.1 million households included in the match,
REAC has identified about 280,000 with potential income discrepancies. HUD
also conducts regular matches with SSA to identify tenants who receive
unearned income from old age, disability, or SSI benefits. Discrepancy
reports may be sent to PHAs to alert them to inconsistencies found as a
result of the computer matches.

Although HUD's computer matches are useful for verifying the eligibility of
current tenants, HUD officials concede that HUD lacks the ability to
adequately verify important information, such as income, for applicants
during their initial certification for assistance. Moreover, several HUD and
PHA officials we interviewed acknowledged that given the difficulty of
recovering improper payments, accurately verifying individuals' eligibility
during initial application becomes even more important for controlling
program payments. In addition, some HUD and PHA officials told us that
certain QHWRA provisions will likely make accurate initial eligibility
determinations even more important. For example, section 523 of the act
specifies that PHAs may provide qualified public housing tenants with flat
rents (as opposed to standard rents indexed to the family's income) that
remain fixed for up to 3 years. Further, the act only requires the
recertification of families who select the flat rent option once every 3
years (as opposed to annually). Thus, with longer periods between
eligibility checks for some tenants, the need to make accurate decisions
during the initial application phase is even more important.

The current lack of adequate data sharing and payment controls may be
attributed to a number of factors, including a lack of adequate resources,
time constraints, legal restrictions, and technological limitations. In
particular, staff at the PHAs we visited cited inadequate staffing levels,
insufficient computer resources, and the inability to access third-party
data sources as their most significant challenges. For example, officials at
the New York City PHA told us that they rely heavily on applicants to
self-report eligibility-related information, because they do not have
adequate staff or time to verify the data that are provided. They also
emphasized that the housing authority has few computers that can access the
Internet or that are linked with independent sources of data such as New
York's Welfare Management System.30 In addition, HUD and the PHAs do not
have the authority to independently match applicants and recipients of
housing assistance against data from all financial institutions, such as the
authority recently granted to SSA.31 Therefore, PHAs must rely on
self-reported information. Staff at one PHA also said that they were unaware
that sources of data such as state wage information existed and therefore
had not taken steps to obtain them. In other instances, some PHA staff
identified the need for a system that links the PHAs together to better
share information on applicants' address and rent histories. However, a
comprehensive, national system to track such information does not currently
exist.

Information

To obtain student financial assistance, applicants must provide various
pieces of information on the FAFSA, including their Social Security number,
adjusted gross income, assets, federal taxes paid, and citizenship status.32
Applications are processed by Education's Central Processing System, in
which a number of consistency tests are conducted to determine whether the
information on the applications is accurate. These tests include computer
matches to verify information (such as citizenship) that can affect the
applicants' eligibility for student aid.33 Following the matches,
eligibility reports are sent to the individual institutions, such as
colleges and universities, where financial aid packages are determined.
However, neither Education nor the institutions have access to third-party
data sources to independently verify most applicants' or parents' income
before disbursing loan and grant payments. Insufficient data are available
to precisely document the extent to which student aid applicants do not
report or underreport their income and related information. However, some
studies suggest that the problem could be substantial. For example, a 1997
Education OIG report found that at least 102,000 students were awarded
approximately $109 million in excess Pell Grant funds because they either
failed to report or underreported their income.34 More than 300 of these
grant recipients underreported their adjusted gross income by more than
$100,000.

While Education's verification procedures (such as computer matching edits
to identify error-prone applications) are reasonable for detecting and
correcting mistakes on applications, they cannot identify students who
intentionally underreport their income. Federal regulations require
institutions to verify certain information for at least 30 percent of
theapplications they receive annually.35 However, this process is generally
limited to requiring applicants to produce additional self-reported
information, such as copies of tax returns, in some instances Form W-2
(wages reported to SSA), or other tax-related forms. The process provides
the institutions with documentation from which verification can be
conducted, but there is no way for institutions or Education to verify that
the applicants provide accurate copies of the forms they send to IRS.

The weakness in this process was documented by Education's OIG, which traced
a sample of adjusted gross income data from applications that institutions
had verified. Although the results from the institutions' verification
procedures indicated that 99.5 percent of the cases were accurate or within
Education-specified tolerances, IRS records revealed significantly higher
income for all the cases they examined. To address this type of problem,
institutions can request applicants selected for verification to sign an IRS
Form 4506, consenting to have copies of their tax returns released to the
institutions as a means of independently verifying income information
provided on applications.36 However, this process can be time-consuming and
cumbersome and may be impractical for large-scale use with millions of
applications.

While a match between Education application data and IRS taxpayer
information could be a more efficient method of verifying applicants'
adjusted gross income, such a match may not be timely enough for initial
eligibility verification purposes. Education and university officials told
us that the student aid packages are processed before IRS has complete
taxpayer data for the prior year available for matching. However, because
student loan and grant payments are typically not made until late August or
September, it may be possible to conduct a match with IRS using fairly
complete taxpayer data, before making such payments. This would allow
Education to verify the information on more students' financial aid
applications than under current procedures.

Although section 483(e) of the Higher Education Amendments of 1998 (HEA) was
apparently aimed at giving Education access to specific IRS taxpayer
information, IRS had not granted Education permission to receive taxpayer
information for use in verifying data provided on FAFSAs as of

June 2000.37 IRS officials told us that the language of the 1998 amendments
was not sufficient to grant Education access to taxpayer data because the
HEA provision did not expressly override the general prohibition on
disclosure of those data found in IRC section 6103(a). Because section
6103(a) permits exceptions to section 6103's general rule of confidentiality
only as provided in title 26, an amendment to section 6103 is necessary to
effectuate IRS's disclosure authority for the purposes contemplated in HEA.
Education believes, however, that it needs not only authority to obtain the
taxpayer data from IRS but also the authority to redisclose the data to
contractors and schools in order to effectively determine applicants'
eligibility for assistance. One alternative would be for Education to
request that every student aid applicant sign an IRS Form 4506. But, as
noted above, both Education and IRS have suggested that this consent-based
approach could be administratively cumbersome, given the large number of
applications that are received annually for student aid.38 Moreover,
Treasury and IRS officials told us that a large-scale consent-based approach
would release a substantial amount of taxpayer data from the protections of
IRC 6103, an approach that Treasury and IRS generally do not support.

Education and IRS are conducting two statistical pilot matches that would
involve summary information, but not individual-level taxpayer data, to
determine how effective a full-scale match would be. The first pilot
involves matching the Social Security numbers of 50,000 student aid
applicants against IRS taxpayer data. The second pilot expands on the first
by including the Social Security numbers of parents for applicants who are
dependents. According to Education officials, this step is important because
it will allow Education to more effectively gauge a family's true ability to
pay for higher education, independent of self-reported information on the
FAFSA. At the time of our work, Education did not yet have the results from
these pilot matches.

PRWORA significantly changed the federal role in ensuring that states have
access to accurate and timely data on the eligibility of their clients for
TANF-funded benefits and services. PRWORA contains an explicit limitation on
federal regulation and enforcement and provides specific, limited
responsibilities for the federal government. The TANF block grants are
fixed, and the states receive the funds regardless of the number of clients
who are eligible for benefits. Block grants present unique challenges to
providing adequate accountability for federal funds. Block grants give
states flexibility to adapt funded activities to fit state and local needs,
and they devolve responsibilities to the states themselves to oversee their
TANF programs. For example, under PRWORA the states establish their own
program eligibility requirements. In addition, any savings achieved as a
result of cost avoidance and eligibility disqualification accrue only to the
state. The federal government does, however, have an interest in ensuring
that certain eligibility requirements are met that require the states to
share client data. For example, PRWORA prohibits families that include an
adult from receiving federal TANF-funded assistance for more than 5 years.
Without the ability to verify interstate historical data on an individual's
receipt of TANF-funded benefits and services, states cannot ensure
compliance with the time limits.

The agencies that administer TANF in the states we visited (California,
Florida, Maryland, and New York) make fairly extensive use of data sharing
to determine applicants' and current recipients' eligibility for benefits
and verify self-reported information. Most of their data sharing activities
are focused on obtaining information on individuals' income, assets,
household composition, and other factors that frequently have an effect on
eligibility, such as citizenship, incarceration, and recipients' death.
However, less is done to identify information across states that can affect
eligibility, such as duplicate receipt of benefits, TANF disqualifications,
and time limits. State TANF officials estimated that improper payments
related to unreported or underreported income may account for between 40
percent and 70 percent of all payment errors. Household composition and
unreported assets are also sources of payment error, according to officials,
although to a somewhat lesser degree than unreported and underreported
income.

Quarterly state wage data are generally viewed as the most valuable source
for verifying earned income in each of the states we visited. They are also
a relatively timely source of information; state officials said that the
data are generally between 4 and 6 months old by the time they are available
to TANF eligibility workers. Employers are required to provide wage
information to the state in order to facilitate the administration of the
unemployment insurance programs. Access to these data may also result in
significant savings or cost avoidance for a program. For example, New York
estimated that its use of state wage data resulted in about $72 million in
cost avoidance between January and September 1999.39 Some but not all state
TANF programs can also draw on the recently created state directories of new
hires to detect unreported or underreported household income.40 While these
databases do not contain wage information, they can show eligibility workers
that applicants or recipients (or other household members) were recently
employed as indicated by a new hire record. New hire data are generally
about a month old when they are available to TANF eligibility workers.41 In
this regard, such data sources are useful in detecting unreported or
underreported income during the initial application process. Maryland
officials reported that using the new hire data has dramatically reduced the
opportunity for TANF applicants and recipients to misreport or fail to
disclose income during the application process. However, officials at ACF
confirmed that not all state TANF programs have access to their state's
directory of new hires data, because they have not completed an agreement to
obtain access to the database.

Unearned income, such as benefits from other public assistance programs, can
also affect individuals' eligibility for TANF benefits. Information that an
applicant or recipient does not report may be obtained from various federal
or state agencies. For example, New York City reported that between July
1993 and June 1999, its computer matching programs identified cost savings
of about $66 million from unreported unemployment insurance benefits, $28
million from Old Age, Survivors, and Disability Insurance (OASDI), and $69
million from SSI. Other relevant sources of unearned income such as
veterans' benefits and workers compensation benefits are not routinely
checked in the states we visited. Access to such data sources could result
in additional cost savings for the states. For example, Maryland reported
$500,000 in savings from a 1997 match with veterans' benefit data, and New
York reported about $600,000 in savings from a workers compensation match
between 1993 and 1999.42

Data sources for other information that affects eligibility for benefits
such as assets, household composition, and citizenship status are not always
timely and may not be available at all. For example, the primary data source
for assets is matches with IRS unearned income (Form 1099) data.43 Because
of the annual nature of tax reporting, IRS Form 1099 data may be 18 to 24
months old by the time match results are available to eligibility workers.
However, IRS is currently the only centralized source for Form 1099 data,
which leaves program administrators with few other options. Access to other,
more timely sources of asset information could be helpful. For example, New
York City supplements IRS Form 1099 data with a match that uses data from
large banks doing business in the city, such as Astoria Federal Savings and
Apple Bank.44 New York City attributed about $8.7 million in cost savings to
the identification of unreported assets through the bank match from 1993 to
1999.45 With regard to household composition, no sources of information
directly or adequately measure this variable. Citizenship status may be
confirmed with the Systematic Alien Verification for Entitlement (SAVE)
system that the Immigration and Naturalization Service administers under
section 1137 of the Social Security Act. However, the states we visited
reported that this system is not always reliable or timely, particularly for
making initial eligibility decisions.

In addition to the problems with data sharing operations just described,
officials in the states we visited consistently pointed to the need for more
timely sources of information suitable for timely and accurate initial
eligibility decisions. Access to more timely data is important because
overpayments are difficult to recover. States generally rely on computer
matches whose results may not be available until after an applicant has
begun to receive benefits. Local eligibility workers reported that fewer
overpayments would occur if more timely data were available during the
application process. For example, fraud units in Florida have access to more
online sources than do eligibility workers. One district's fraud unit
reported cost savings of about $200,000 during November 1999. Additionally,
Maryland, New York, and eight other states are participating in a pilot
project that permits online access to SSA benefit data to help them better
control TANF payments.46

State officials also emphasized that the TANF programs they administer lack
a comprehensive data source to track interstate receipt of benefits, TANF
disqualifications, and benefit time limits across all states.47
Traditionally, states have tried to address the problem of interstate
receipt of benefits by negotiating matches with neighboring states. Florida,
Maryland, and New York reported conducting such matches. However, state
officials explained that in their experience, the surrounding states do not
always agree to participate in a match, thereby limiting its coverage and
effectiveness. In 1997, HHS submitted a report to the Congress that outlined
the status of automated data processing systems operated by the states and
noted that, among other findings, virtually no state agency could identify
duplicate TANF cases or implement the TANF time limit across state
boundaries.48 The report defined five alternative systems for a centralized
national client database that would address this type of weakness in the
states' systems. In 1999, the Department of Agriculture produced a
supplemental report that assessed the current state agency efforts to
identify interstate, duplicate receipt of benefits, as well as state agency
capabilities to participate in a national client database. This report built
on the findings in the HHS report and concluded that a national database
system would be feasible and cost-effective if it were configured in a
specified manner and encompassed both the TANF and Food Stamp programs.49

In a recent report on welfare reform, we also found that five of six states
we reviewed reported that they do not collect data on recipients' prior
receipt of TANF in other states or that they rely on recipients to disclose
this information.50 ACF initiated a project in 1993 that may be useful in
addressing this problem. Its Office of State Systems created the Public
Assistance Reporting Information System (PARIS) with the computer operations
support of the Defense Manpower Data Center (DMDC) facility in Monterey,
California. This system uses the facility's computer matching capabilities
to compare various state and federal program databases.51 Although the
number of states participating in PARIS has been limited, the project has
identified significant potential cost savings.52 For example, Pennsylvania
has estimated its annual cost savings at $2.8 million from the PARIS
interstate match. In addition, ACF estimates that the August 1999 PARIS
match with the DMDC databases identified potential cost savings of about
$28.6 million. Increased savings and cost avoidance estimated at more than
$100 million could be realized if the match were conducted nationally with
all states participating. In addition, ACF estimates that cost savings of
$15.3 million could have been achieved in 1999 from matching Veterans
Administration data with 16 states and the District of Columbia. As
discussed previously, the savings resulting from PARIS benefit the states;
the federal government does not realize any direct savings from this
activity. In addition, while the PARIS project demonstrates the efficacy of
cross-state data sharing, the computer matches it uses are primarily
targeted at posteligibility verification (once benefits have already been
dispensed to recipients). A more efficient approach would be to help states
obtain eligibility information on new applicants, before benefits are paid,
and to track time limits and disqualifications across state lines. However,
a nationwide system with these capabilities does not currently exist.

We recently reported that action at the federal level could better
facilitate states' efforts to improve their automated systems in four areas.
Among these were addressing the need for states to have access to
cross-state information on individuals' TANF receipt to enable the
enforcement of the 5-year TANF time limit. We also recommended that the
Secretary of HHS establish an interagency group to identify, and develop
implementation plans for, federal actions that would facilitate states'
efforts to improve their automated systems for federal programs that serve
low-income families.53

Several legal as well as management, administrative, and technical
challenges limit the ability of benefit and loan programs to use data
sharing to control their payments. Central to some of these challenges are
concerns about protecting individuals' privacy, the confidentiality of
sensitive data and taxpayer compliance with tax laws, and balancing such
concerns with the government's need for certain information to provide
financial stewardship for its programs. Legal issues closely tied to
privacy, confidentiality, and tax compliance concerns include (1) IRC
section 6103 governing the disclosure and use of taxpayer data, (2) some
provisions of the Social Security Act that limit access to important data
sources such as OCSE's National Directory of New Hires, and (3) provisions
of CMPPA that limit the amount of time during which initial matching
agreements and renewals may be in effect. Management and administrative
challenges include balancing the objective of providing flexibility for the
states and localities to administer programs such as TANF against the need
for accountability and program performance information and the need for a
more comprehensive management approach by OMB to coordinate and support data
sharing activities across federal agencies. Technical issues include the
need for more timely and accurate data, as well as the need for consistent
eligibility criteria and data formats across programs. Balancing the
competing objectives of privacy, data security, and tax compliance with the
need for program integrity and efficiency involves some trade-offs. For
example, individuals who seek public benefits or loans may have to be
willing to provide sensitive information to the government so that their
eligibility for benefits can be accurately verified. The experience of
agencies such as SSA demonstrates, however, that maintaining and sharing
sensitive data is possible while maintaining the protections afforded by the
Privacy Act and related statutes.

IRC section 6103 is one of the main statutes that limits more extensive data
sharing among federally funded benefit and loan programs. To protect the
confidentiality of taxpayer information and ensure compliance with tax laws,
IRS is prohibited from disclosing taxpayer information to other federal
departments and agencies without specific statutory authorization in IRC
section 6103. Moreover, even in cases in which a federal department is
authorized to obtain protected information, it may be prohibited from
sharing that information with state and local entities that are involved in
administering programs that the department or agency funds. For example, IRC
section 6103(l)(7)(D)(ix) authorizes SSA and IRS to disclose income
information to HUD in order to verify applicants' and recipients' income.
However, only officers or employees of HUD may use the information. This
restriction precludes HUD from redisclosing the data to housing authority
administrators at the local level who need the information to accurately
verify individuals' income.54 Similar issues have surfaced with regard to
other federal agencies' use of private contractors to administer their
programs, such as in the case of Education, discussed earlier.

IRS officials told us that limitations on the dissemination of taxpayer
information are intentionally restrictive because of the risk of
unauthorized access when use is widespread. As a general rule, they stated,
the further that taxpayer information is disclosed from centralized
protections and controls, the more risk there is of improper disclosures,
intentional or otherwise. These officials were also skeptical of agencies'
ability to police themselves in terms of adequately protecting taxpayer
data. Allowing state or local entities (such as private contractors) access
to sensitive taxpayer data would be even more problematic, in their view.
The Joint Committee on Taxation reiterated this viewpoint in its January
2000 report when it recommended that "present-law disclosure rules for using
contractors for nontax administration purposes should not be expanded."55

One solution to these restrictions, according to IRS officials, is for
federal agencies to verify data they receive from IRS through an independent
party, such as an employer or bank. Any corroborated data the agency obtains
are not subject to the restrictions in IRC section 6103 and therefore may be
disclosed to the state and local entities (although the agency would have to
then delete the remaining IRS information from its databases). Another
alternative is for agencies to use a consent-based approach, in which
applicants and recipients provide written consent for IRS to release their
information to a designated third party. However, IRS and Treasury officials
are concerned about the widespread use of consent-based disclosures because
of the potential administrative problems as well as their inability to
protect the confidentiality of the data.

Section 453(l) of the Social Security Act, as amended, restricts access to
the National Directory of New Hires to particular agencies for specified
purposes. For example, Treasury (including IRS) has access to its database
to administer federal tax laws, as well as to verify claims for the Earned
Income Tax Credit. SSA also has access to the National Directory of New
Hires data to help it administer the SSI and OASDI programs. More recently,
Education was granted access for purposes of obtaining the addresses of
individuals who have defaulted on student loans or who owe grant repayments
to Education.

According to OCSE officials, numerous federal agencies and departments have
requested access to the National Directory of New Hires to help them
administer their programs. In most cases, such requests have been denied
because there was no statutory authority in the Social Security Act to allow
release of the data. OCSE officials told us of a variety of concerns
regarding how the National Directory of New Hires data are used and what
entities have access to them. For example, they emphasized that the
directory was designed primarily to help OCSE administer its
responsibilities for enforcing child support. In their view, opening the
database to widespread data sharing could reduce support for the database
among important groups, such as employers. In addition, they expressed
concern about the security of the data in the directory if they were to
become more widely accessible. Moreover, these officials told us that wider
access would likely raise additional questions about the privacy and
confidentiality of the information contained in the database.

We interviewed officials from various agencies who expressed frustration
with some of the provisions of CMPPA and who viewed the current renewal
procedures as unnecessarily burdensome. In particular, several officials
argued that the current 18-month and 12-month periods in which,
respectively, initial matching agreements and renewals may be in effect are
too short. Because the processes for getting new agreements approved and
existing contracts renewed are often lengthy and cumbersome, agreements that
are currently in force may lapse before new ones can be approved and
implemented. This can be disruptive to both agency operations and
relationships with other federal and state entities that are parties to a
match. Various officials familiar with the process of instituting matching
agreements suggested lengthening these time periods. Some officials told us
that extending agreement periods would not adversely affect the intended
confidentiality or due process rights that CMPPA was created to protect.
This is particularly true for routine matching agreements that do not change
substantively from year to year.

During the symposium on data sharing among federally funded benefit and loan
programs that we and the Senate Governmental Affairs Committee sponsored in
June 2000, representatives of various federal agencies stated that the
current time periods for implementing new matches and renewing existing
matches are unnecessarily burdensome. They emphasized that they are
committed to the privacy and data confidentiality protections that CMPPA was
created to protect and believe that longer time periods could be instituted
while maintaining such protections. Overall, these officials were uniform in
their assessment that the time limits on new and existing matching
agreements in CMPPA need to be lengthened. One agency official stated that a
legislative proposal is being considered that would extend the current time
limits for new agreements to 5 years and 3 years for extensions.

In 1997, the Benefit Systems Review Team (BSRT) led by OMB also found that
the procedures for renegotiating matching agreements was burdensome to the
agencies that conducted computer matches and could be significantly
streamlined.56 BSRT concluded that agencies' administrative burden could be
reduced by amending the computer matching sections of the Privacy Act to
allow for interagency agreements that would be effective for 3 years
initially and for an unlimited number of 2-year extensions. In addition,
BSRT concluded that such changes could be implemented without degrading the
protections intended by the Privacy Act. Finally, BSRT also recommended that
OMB work with the agencies to obtain an amendment to the Privacy Act to
permit the extensions to the time periods through which initial agreements
and renewals could be in force.

Welfare reform gave states and local governments additional flexibility to
design and administer public assistance programs such as TANF in a manner
that fits their needs. As state welfare agencies implement their plans, they
are drawing on numerous federal and state programs--often administered by
separate agencies--to provide a wide array of services such as child care,
food stamps, and employment training. These changes have had implications
for the states' information needs. For example, in the new welfare
environment, information systems to support expanded functions are necessary
in three key areas: case management, including eligibility determination;
service planning; and program oversight, with a new emphasis on outcomes and
results. In addition, the increased devolution of responsibility for program
operation and performance to states and localities increases the need for
information systems that can respond to the multiple needs of users at all
levels of government. We have found that shortcomings in some of the
automated systems that state and local officials use to administer TANF have
resulted in insufficient data to ensure some aspects of program integrity,
such as enforcement of the 5-year time limit.57 In this regard, the lack of
adequate data to ensure program integrity in some states may affect their
ability to ensure that taxpayer dollars are spent only on families who
qualify for benefits. However, improved automated systems coupled with an
enhanced emphasis on data sharing activities may improve the ability of
local, state, and federal programs to effectively communicate with one
another. Such improvements could benefit the ability of states to track and
assess the performance of service delivery, as well as help provide greater
program accountability by ensuring that the intended goals and requirements
of programs such as TANF (for example, the 5-year time limit) are met. It is
likely that additional federal assistance may also be necessary to help
states develop systems capable of performing these varied tasks. While
PRWORA reduced the role of HHS in overseeing systems funded solely with TANF
funds, HHS and other agencies (such as Agriculture and Labor) still play a
key role in funding and overseeing states' information systems for social
programs. The federal government could further facilitate states' automated
systems initiatives.

to Data

OMB is responsible for providing overall leadership and coordination of
federal information resources management within the executive branch.58 OMB
also has the core responsibility for managing federal computer security and
information technology, including data sharing among federally funded
benefit and loan programs. In our earlier work, we found that mission
fragmentation and program overlap are widespread in the federal government
and that crosscutting program efforts are not well coordinated.59 Without
adequate coordination, scarce funds are wasted, program customers are
confused and frustrated, and the overall effectiveness of the federal effort
is limited. We have found that OMB needs to take more initiative in setting
the agenda for this type of governmentwide issue. In particular, we have
called for a more decisive and assertive OMB role in defining the problems,
developing appropriate management strategies, and overseeing progress.

OMB has pursued such an approach in the past, as evidenced by the BSRT
report in 1997.60 Initiated by OMB in February 1995, this interagency effort
suggested numerous actions for improving data sharing among federally funded
benefit and loan programs, including moving from a "pay and chase" mode to a
more proactive and efficient approach (such as using data sharing to prevent
improper payments before they are made).61 Various pilots and data sharing
initiatives have been started by a number of federal agencies since the
inception of BSRT.

However, during our review, several federal officials we spoke with said
that more recently, OMB has not provided adequate leadership and
coordination to facilitate the efficiency and effectiveness of data sharing
throughout the federal government. Representatives of various federal
agencies indicated that federal data sharing efforts lack the necessary
coordination and leadership to maximize the efficiency of data sharing
activities. OMB activities appear to be largely geared toward resolving
problems case by case, without a more systematic, governmentwide approach to
these issues. As a result, some federal officials told us that federal data
sharing operations are not as effective or efficient as they could be and
suggested that OMB should take a more proactive, strategic approach to
federal data sharing issues. In light of the concerns about data security,
privacy, and confidentiality that proposals for improved data sharing raise,
it is even more important for OMB to demonstrate strong leadership that
emphasizes coordination and communication among all federal agencies.

OMB has recently indicated that it is taking a number of steps to improve
coordination across federal agencies, as well as the priority it is placing
on data sharing activities. For example, during our data sharing symposium
in June 2000, an OMB official said that the agency is preparing to issue new
guidance to all federal agencies that will focus on the interagency sharing
of sensitive data within the existing constraints of the Privacy Act. In
addition, OMB officials point to one of the 24 priority management
objectives (PMO) contained in the U.S. budget for fiscal year 2001 as
evidence of their commitment to managing data sharing activities.62 PMO 10
emphasizes the administration's focus on ensuring that administrative and
program payments to recipients are made accurately up front.63

State and local officials we interviewed in the three programs we examined
told us that their inability to obtain timely data from independent sources
is a major problem. A common complaint was that some data derived from the
IEVS were not timely. For example, IRS Form 1099 data were usually too old
to be useful, particularly for making timely eligibility decisions at the
point of application. State officials generally agreed that the lack of
timely income data from IRS makes access to alternative sources of income
information, such as the National Directory of New Hires or state wage data,
even more important. In addition to concerns about the timeliness of various
data sources, some officials questioned the reliability of the data they
receive. Two commonly cited examples of data that were sometimes unreliable
included SSA prisoner data and citizenship status data from INS' SAVE
system.

Some officials also said that varying data formats and eligibility criteria
can make efficient data sharing among public assistance programs difficult.
One state official told us that differing definitions of an "active" TANF
case in different states has made it more difficult to determine the
accuracy of match hits and complicates the follow-up required. For example,
while some states consider only a family that is receiving cash benefits an
"active" case, other states also include families whose applications are
pending or who have been suspended. Another example is the complexity that
is introduced to data sharing operations from nonuniform definitions of
"income" in various public assistance programs. Although some programs such
as SSI use a definition of income that is fairly broad and includes in-kind
support such as food and shelter provided by family or friends, other
programs have more restrictive definitions. Some officials told us that such
variations can make it difficult to share data across programs. In addition,
one official noted that different data formats across numerous programs and
systems for a commonly required variable such as an individual's address
complicates data sharing activities.

The objective of individual privacy can be compatible with the objectives of
program efficiency and integrity but may involve some trade-offs. For
example, individuals who wish to obtain public assistance or loans may have
to be willing to give administrators of benefit and loan programs access to
sensitive data. However, it is possible to improve access to important data
sources in a manner that is consistent with protections in the Privacy Act,
the Computer Matching Act amendments, and other privacy statutes. One
potential solution might include providing OCSE with the authority to share
data from the National Directory of New Hires with benefit and loan programs
such as HUD housing under controlled conditions within the limits of the
Privacy Act. For instance, periodic computer matches between benefit
programs' applicants and current recipients and the directory could specify
the type and amount of data that are shared and which federal, state, or
local entities could have access to the data. Moreover, federal, state, or
local agencies that have access to sensitive information could be required
to demonstrate adequate data security environments before receiving match
results.

SSA is an example of an agency that has significant experience in balancing
privacy with program integrity and efficiency. In recent years, SSA has
increased its use of data sharing as a tool to protect the integrity of the
SSI program. For example, SSA is expanding its use of on-line state data to
obtain more real-time applicant and recipient information. SSA is conducting
a pilot that gives selected field offices (that are responsible for
determining applicants and recipients' eligibility for SSI benefits) on-line
access to wage data, new-hire data, and unemployment insurance data that SSA
obtains from OCSE's National Directory of New Hires.64 To protect the
security of sensitive data in the directory and ensure that individual
privacy is maintained to the maximum degree possible, SSA is working to
strengthen a number of safeguards to limit access to the database, as well
as maintaining an audit trail that identifies the employees who have used
the database.65

SSA is also working on obtaining access to other important data sources. The
Foster Care Independence Act of 1999 (Pub. L. No. 106-169) includes
provisions that give SSA the authority to require SSI applicants and current
recipients to authorize SSA to obtain financial records from any and all
financial institutions by conducting periodic data matches. This new
authority should allow SSA to more effectively verify the assets of SSI
applicants and recipients, such as bank accounts. The Congressional Budget
Office estimates that this provision will reduce spending on SSI benefits by
$70 million from 2000 through 2004.66

Other federal agencies could benefit from the experience of agencies such as
SSA in terms of balancing privacy interests with the need for more timely,
accurate sources of eligibility information. For example, allowing HUD to
match applicants and current recipients of housing assistance programs
against the state wage, unemployment, and new hire data in OCSE's National
Directory of New Hires would provide HUD with the most timely, comprehensive
information available for verifying wages and unemployment compensation.
Such an arrangement could permit routine computer matches that OCSE could
administer and would not necessarily entail direct access to the sensitive
directory databases for HUD. Any matching agreement between HUD and OCSE
could specify what pieces of information would be matched and that only
information on individuals identified in a match could be provided to state
or local program administrators, such as public housing agencies. In
addition, HUD and individual housing agencies would be responsible for
ensuring that the privacy of individuals identified in a match is adequately
protected. HUD and public housing agencies could also be required to
demonstrate that they have sufficient data security protections in place
before receiving any data from OCSE.67 However, providing access to this
data source would require making changes to existing laws, because HUD does
not have access to the National Directory of New Hires, and providing
taxpayer data to local public housing agencies is not permitted. Moreover,
it would require HUD and numerous housing agencies to institute adequate
data protections in consultation with IRS and OCSE.

In addition, while programs like TANF rely on the initiative of state and
local program administrators to obtain some of the information they need to
make eligibility decisions (such as the activities in New York City to
obtain bank account information from local financial institutions), it may
not be the most efficient or comprehensive arrangement for obtaining such
information. Access to information sources such as financial institutions
through a more consolidated source (perhaps with the facilitation of a
federal agency such as HHS) could provide a more efficient method for
obtaining asset information across multiple states. However, most federal
agencies do not have authority to independently obtain asset information
from financial institutions (similar to the authority granted to OCSE and
SSA).

As the steward of taxpayer dollars, the federal government has the
responsibility to ensure that public assistance benefits and loans go only
to the persons who are eligible to receive them. When ineligible individuals
or families receive benefits, those who are eligible may not receive
assistance, and the public's confidence in the government may be undermined.
To ensure privacy and the confidentiality of personal information, access to
sensitive data and systems must be controlled so that only authorized
individuals are able to view specific pieces of information. Audit trails
are also necessary to track which individuals have accessed sensitive
databases. Thus, with adequate protections for privacy and data security in
place, it may be reasonable for applicants and recipients of benefits and
loans to provide the government with access to sensitive, private
information (as well as the authority to share such information with other
benefit and loan programs) in exchange for assistance. However, an ongoing
dialogue involving the Congress, OMB, other federal agencies, the states,
and the public will be required to reach consensus on how best to balance
personal privacy and data security with the objectives of program integrity
and efficiency.

Data sharing among federal, state, and local governments is a powerful
internal control tool that helps benefit and loan programs make more
accurate initial and continuing eligibility decisions. The three federally
funded benefit and loan programs we examined all use data sharing to
substantial but varying degrees to verify information that applicants and
current benefit recipients provide. The weaknesses in these programs'
eligibility determination processes could be mitigated if additional data
sources were available for sharing. The continued reliance on self-reported
information from applicants and recipients leaves these programs at risk for
improper payments. In recent years, federal departments such as HUD and
Education have attempted to improve their access to important data sources.
For instance, Education has repeatedly sought access to IRS taxpayer
information to help it administer its student loan and grant programs.
However, legal and other restrictions have impeded its ability to obtain the
information it needs. The programs we reviewed could all use access to
additional data sources to control their payments. For example, allowing
TANF and HUD's housing assistance programs to require applicants and
recipients to authorize them to obtain data from all financial institutions
(similar to the authority recently granted to SSA in the Foster Care
Independence Act) would provide useful tools for verifying self-reported
information that affects individuals' and families' eligibility.

Federal leadership could also be helpful in addressing the lack of a
mechanism to share some eligibility information across states (such as
duplicate benefit receipt and time limits on TANF usage), which hampers the
ability of program administrators to obtain some information they require to
make accurate eligibility decisions. Federal action to facilitate improved
data sharing and program coordination--particularly across states--while
protecting the confidentiality of sensitive data may be needed. Federal
departments that have primary oversight responsibility for benefit and loan
programs administered at the state and local levels--such as HHS (TANF), HUD
(housing assistance), and Agriculture (Food Stamps)--are candidates for
facilitating improved data sharing.68 Moreover, additional leadership and
coordination from OMB, contributing to a governmentwide dialogue on data
sharing issues, could further facilitate greater efficiency and
effectiveness in data sharing activities among federal programs. OMB could
draw upon the expertise and resources of interagency groups such as the
Chief Information Officers Council to achieve this objective. Because of
OMB's ability to look across agencies and mediate competing demands or
concerns, it is the appropriate federal entity to provide the leadership and
coordination necessary for enhanced data sharing. Its PMO 10, which
emphasizes verifying the accuracy of benefit payments (as articulated in the
fiscal year 2001 budget) is a positive development, but additional actions
are needed.

The programs we reviewed are making progress in expanding the range of
information they use to control their payments but could also benefit from
access to more timely and comprehensive sources of information for making
eligibility determinations, such as OCSE's National Directory of New Hires.
Enhanced front-end data sharing would help avoid some overpayments and allow
programs to move away from a "pay and chase" mode toward an approach that
would prevent erroneous payments. For example, providing HUD with access to
the directory and permitting HUD to redisclose specific, limited information
to the public housing agencies that administer the Public Housing and
Tenant-based Section 8 programs could help them make more accurate initial
eligibility determinations and avoid some of the nearly $1 billion in
estimated annual subsidy overpayments that have been identified in recent
years. In addition, providing Education with the ability to confirm
information on applications with IRS taxpayer information would help schools
more accurately determine applicants' eligibility for loans and grants.
However, the various challenges we have outlined make such front-end data
sharing problematic. In particular, current laws (such as parts of IRC
section 6103 and the Social Security Act) prevent many benefit and loan
programs from obtaining access to important and timely data sources, such as
the National Directory of New Hires. Other challenges we highlighted, such
as the burden placed on federal agencies by the time limits on initial and
renewed matching agreements in CMPPA, could be addressed with relatively
minor modifications to existing law. The current time periods (18 months for
initial agreements and 12 months for renewed agreements) are too short to
allow efficient operation of the matches and do not permit sufficient time
to negotiate new or extended agreements. Many federal officials involved in
computer matching activities agree that the time limits on initial and
renewed matching agreements could be extended while maintaining existing
privacy protections.

When considering changes to existing privacy and confidentiality provisions
in current laws, the Congress must weigh the concerns about sharing
sensitive data against the government's need for timely and accurate
information to ensure accurate payments in its benefit and loan programs. In
an age in which technological advances such as the Internet have facilitated
an increasingly free flow of information, the federal government has the
duty to protect the security and confidentiality of the data with which it
is entrusted. However, providing access to restricted data sources does not
mean that sensitive data become available in the public domain or that
existing data security protections are eliminated. Rather, agencies or their
representatives with the need to view certain information (such as income)
for purposes of determining eligibility for individuals who apply for
benefits and loans would have access to such data under controlled
conditions and would be responsible for protecting them from unauthorized
disclosure.

The Congress should consider how to improve the ability of federally funded
benefit and loan programs to obtain and share the information they need to
make timely and accurate eligibility determinations while protecting
personal privacy and the confidentiality of sensitive information.
Specifically, the Congress should consider

ï¿½ Amending the Social Security Act to allow HUD's Public Housing and
Tenant-based Section 8 programs to match applicants and current recipients
against OCSE's National Directory of New Hires. To facilitate accurate
initial eligibility determinations, HUD would have to be authorized to share
the results of a match with the local agencies that administer programs
(such as public housing agencies). To maximize privacy and data security
protections, HUD could screen the results of such a match to ensure that
only data vital to determining eligibility for individuals identified as
having reported erroneous information would be made available to state and
local program administrators. The local agencies should also be required to
demonstrate that adequate privacy and data security measures are in place.

ï¿½ Amending section 6103(l) of the Internal Revenue Code to authorize IRS to
disclose certain taxpayer data to officers, employees, and contractors or
other agents (such as schools) of Education for purposes of verifying
information reported on applications for financial aid. Again, only the
minimum amount of information necessary to determine eligibility for cases
that are identified in a match would be provided to schools or other
entities, subject to agreed upon privacy, data security, and safeguard
measures.

To decrease the burden of initiating and renewing computer matching
agreements, the Congress should amend the Privacy Act's computer matching
requirements by increasing the time limits under which initial matching
agreements and renewed agreements may be in force. The appropriate time
period through which initial and renewed agreements should be in force is
subject to debate. However, existing recommendations from various entities
(such as BSRT) range from doubling the existing limits (from 18 months and
12 months for initial and renewed agreements, respectively, to 3 years and 2
years, respectively) to as long as 5 years for initial agreements and 3
years for renewed agreements.

The Secretaries of HUD and HHS should develop a strategy for obtaining
access to automated data from financial institutions, determine the most
cost-effective, efficient method for obtaining such data, and report to the
Congress on any legislation that would be needed to implement this strategy.
HUD and HHS should also coordinate with SSA and OCSE to determine what
strategies are the most effective for obtaining this information.

The Secretary of HHS should facilitate the development of a system for
tracking duplicate benefit receipt, program disqualifications, and recipient
time limits across states in the TANF program. The Secretary should also
coordinate with other federal departments that oversee major public
assistance programs, such as Food Stamps, to ensure the compatibility and
cost-effectiveness of such a system.

To facilitate more efficient data sharing and help federal agencies that
administer benefit and loan programs to control their payments, OMB should
lead an effort that involves key departments and agencies (such as
Agriculture, Education, HHS, IRS, and SSA) to develop an overall strategy
for improving data sharing operations across all benefit and loan programs.
Such a strategy should include actions for further defining the problems
agencies encounter in sharing data, developing appropriate management
strategies to address these problems, and overseeing progress. OMB could
draw on the resources and expertise of the Chief Information Officers
Council and other related organizations to coordinate the appropriate
officials and federal agencies to achieve this objective.

We provided the federal departments and agencies we reviewed--Agriculture,
Education, HHS (including ACF and OCSE), HUD, OMB, SSA, and Treasury
(including IRS)--with the opportunity to comment on this report. Education
supports our recommendations, and HHS, HUD, and OMB indicated that they
generally support the direction and intent of our recommendations, with some
qualifications. Treasury noted that our report provides a fair and balanced
assessment of the challenges facing federally funded benefit and loan
programs in terms of making timely and accurate eligibility determinations.
However, Treasury and IRS also emphasized the unique challenges to the tax
system presented by the use of taxpayer data to verify self-reported
information, particularly with regard to taxpayer compliance with tax laws.
Formal comments from the agencies appear in appendixes I-VI.

With regard to our first matter for congressional consideration, to allow
HUD's Public Housing and Tenant-based Section 8 programs to match applicants
and current recipients against the National Directory of New Hires, HUD is
generally supportive. However, it suggested that this recommendation be
broadened to include HUD's project-based programs. OMB also agrees with the
intent of this recommendation but noted that responsible government must
promote data sharing in a manner that is consistent with strong privacy and
data security safeguards. OMB told us that it is currently examining whether
the National Directory of New Hires is an appropriate source of wage data
but emphasized that in its view any broad government or private sector
access to the financial data maintained in the directory could adversely
affect individual privacy and confidentiality. The other agencies did not
provide specific comments.

With respect to HUD's response, we note that our review focused exclusively
on HUD's tenant-based programs that are administered by public housing
agencies; we did not examine HUD's project-based housing programs.
Therefore, its suggestion to broaden this recommendation to include
project-based housing programs administered by private owners is beyond the
scope of this report. With regard to OMB's comments, we recognize the
sensitivity of the data maintained in the directory and the need to
carefully consider how the data could be made available to departments such
as HUD while preserving privacy and data security to the maximum degree
possible. As we discussed in the section on options for enhanced data
sharing, having OCSE perform a routine computer match using HUD's data on
housing applicants and current tenants, and limiting the data that public
housing agencies receive from such a match, would likely provide a greater
degree of individual privacy and data security than giving HUD online or
direct access to the National Directory of New Hires would afford. Moreover,
requiring HUD and public housing agencies to implement privacy and data
security measures that are consistent with the principles in the Privacy Act
(and any additional guidelines that HHS or OCSE may stipulate) could help
ensure that only authorized individuals have access to sensitive
information. Although computer matches are often viewed as less timely than
online access for purposes of making initial eligibility determinations, in
our view computer matching could still be used effectively to reduce the
incidence of unreported or underreported income in HUD's tenant-based
housing programs, particularly if the matches were performed at frequent
intervals.

Education supports our second matter for congressional consideration, to
amend section 6103(l)(13) of IRC to authorize IRS to release certain
taxpayer data to Education's contractors and schools. Although Treasury did
not address this recommendation directly, it emphasized that IRC section
6103 was enacted to limit the instances in which agencies used tax data for
nontax purposes to cases in which the need for the information outweighs the
related concerns of taxpayer privacy and continued taxpayer compliance. It
noted that in the past, the Congress has provided narrowly tailored
exceptions to section 6103 when it has determined that the need for taxpayer
information outweighs the concerns about taxpayer privacy and compliance.
Treasury also argued that tying tax reporting to needs-based government
benefits may lead some individuals to misreport their income, thus
undermining the integrity of the data provided to IRS and diminishing the
utility of the data for the very purpose for which they were originally
collected. Treasury believes that the integrity of the tax system could be
compromised and noted that a decrease of just 1 percentage point in taxpayer
compliance could cost the federal government more than $10 billion annually.
Treasury further reiterated its opposition to using taxpayer consent for
large-scale income verification programs. However, Treasury indicated that
it is willing to allow the use of consent forms to permit the disclosures
contemplated in HEA on a limited scale, pending an amendment to section 6103
and the outcome of the ongoing statistical pilot matches between Education
and IRS.

We acknowledge Treasury's position that its primary responsibility is to
protect the fiscal integrity of the U.S. Treasury and ensure taxpayer
compliance with existing tax laws and regulations. However, there is little
evidence to suggest that allowing IRS to disclose limited taxpayer
information on specific individuals who apply for needs-based benefits or
loans would lead to a reduction in overall taxpayer compliance. Numerous
federal as well as state and local agencies already have (or have recently
had) access to IRS taxpayer information. For example, as we noted in our
recent report on taxpayer confidentiality, 37 federal and 215 state and
local agencies received, or maintained records containing, taxpayer
information under provisions of section 6103 during 1997 or 1998.69
Moreover, these agencies are required to protect the confidentiality of the
information they receive and to implement safeguards that are designed to
prevent unauthorized access, disclosure, and use. Although IRS safeguard
reviews have identified deficiencies in some agencies' safeguard procedures,
it has recommended various corrections and has followed up to ensure that
the agencies have implemented acceptable solutions to these deficiencies.
Additional disclosures of taxpayer information such as those envisioned in
HEA always involve some degree of risk of unauthorized disclosure. However,
it is the responsibility of departments such as Education to implement
appropriate safeguards that comply with existing regulations to ensure that
privacy and data security are maintained. Pending any amendment to section
6103, we encourage Treasury and IRS to continue working with Education to
study the feasibility of using consent on a limited scale. It will also be
helpful to examine the results of the ongoing pilot matches between IRS and
Education to obtain the most up-to-date information on how widespread
underreporting of income currently is among applicants for student financial
aid.

Both HHS and HUD generally agree with our recommendation that they develop a
strategy for obtaining access to automated data from financial institutions.
HHS agrees that a system to accomplish this objective would be beneficial
but noted that there are a number of confidentiality and data security
issues as well as technical issues that must be addressed. HHS said that it
will consider implementing the recommendation in the context of existing
operational, budget, and personnel constraints. HUD stated that the
recommendation should be broadened to include data from state wage agencies
and should involve Labor because Labor has responsibility for overseeing
state wage information.

We agree that various confidentiality and data security issues will have to
be addressed to implement this recommendation, and we reiterate the need for
HHS and HUD to coordinate with OCSE and SSA to address such issues in the
most timely and efficient manner. To the extent possible, HHS and HUD should
implement this recommendation within the framework of existing privacy laws.
HUD's suggestion to broaden the recommendation to include wage information
is partly addressed in the first matter for congressional consideration that
suggests providing HUD with access to data (including wage data) in OCSE's
National Directory of New Hires. We agree with HUD that state wage agencies
are also a valuable source of wage information and should not be overlooked
as potential sources of these data, but the recommendation is directed at
information available at financial institutions.

HHS generally agrees with our recommendation to develop a system for
tracking duplicate benefit receipt, program disqualifications, and recipient
time limits in the TANF program. It noted that the 1997 report to the
Congress on data processing (and the more recent Agriculture report that
supplemented it) identified a number of systems that could accomplish these
objectives.70 HHS also indicated, however, that additional program authority
and resources would be required to implement this objective.

Our recommendation envisions HHS taking the next step to facilitate the
development of such a system in cooperation with Agriculture and the states.
If additional legislative or program authority and resources are required,
we encourage the Secretary to seek congressional approval to facilitate the
development of such a system. Further, we continue to believe that it is
important for HHS to work closely with Agriculture to ensure that any system
that is developed integrates the Food Stamp program to promote maximum
efficiency and cost-effectiveness.

OMB supports the general thrust of our recommendation to develop an overall
strategy for improving data sharing across all benefit and loan programs. It
also emphasized the need to carefully consider the potential implications
that enhanced data sharing may have for individual privacy, as well as the
security of sensitive information. For example, OMB stated that it is
studying whether additional uses of the National Directory of New Hires are
appropriate as a source of wage data but noted that expanding access to its
data requires careful study of the potential effect on personal privacy.

We agree that expanding access to data sources such as the directory,
whether through more traditional computer matches or online access, requires
careful consideration of the potential effect on personal privacy and data
security. We believe that OMB can help promote the necessary balance between
program integrity issues and privacy concerns by implementing our
recommendation and promoting an open, ongoing dialogue on these issues with
the Congress, federal agencies, states, private organizations, and the
public. The results of the OMB-led Benefit Systems Review Team demonstrate
that OMB (with the assistance of the Chief Information Officers Council and
other organizations) can facilitate this discussion and can develop a
strategy for improving data sharing among federally funded benefit and loan
programs.

A number of HUD's comments requested that we add language to the report to
include private owners who administer HUD's project-based housing programs.
HUD noted that about one-third of the total number of assisted households
receive rental assistance from the Office of Housing's programs that are
generally administered by private owners. Our review focused exclusively on
HUD's tenant-based programs (Tenant-based Section 8 and Public Housing) that
are administered by public housing agencies. Data sharing with private
owners was outside the scope of our work. Moreover, providing private owners
with access to sensitive information raises additional privacy and data
security issues that require careful consideration before data sharing can
be implemented.

Education, HHS, HUD, OMB, Treasury, and IRS also provided us with technical
comments that we have incorporated into our final report where appropriate.

As we agreed with your offices, unless you publicly announce the contents of
this report earlier, we plan no further distribution of it until 30 days
from the date of this letter. We will then send copies to the federal
agencies and states we reviewed that have administrative or oversight
responsibility for the public assistance and loan programs we discussed, and
we will make copies available to others on request.

Please contact me at (202) 512-7215, or Sigurd Nilsen at (202) 512-7003, if
you have any questions concerning this report or need additional
information. Other contacts and contributors to this report are listed in
appendix VII.
Cynthia M. Fagnoni
Director, Education, Workforce,
and Income Security Issues

Comments From the Administration for Children and Families

Comments From the Department of Education

Comments From the Department of Housing and Urban Development

Comments From the Internal Revenue Service

Comments From the Office of Management and Budget

Comments From the Department of the Treasury

Major Contributors to This Report

Jeremy Cox, Evaluator-in-Charge, (202) 512-5717
Nancy Cosentino
James Lawson
Behn Miller
Roland Miller
Chris Morehouse
Sigurd Nilsen

(207071)

Table 1: Potential Sources of Data for the Three Programs in Our
Review 14

Table 2: Some Major Federal, State, and Private Data Sources 16
  

1. Financial Management: Increased Attention Needed to Prevent Billions in
Improper Payments (GAO/AIMD-00-10 , Oct. 1999). See also OIG, Department of
Housing and Urban Development, U.S. Department of Housing and Urban
Development: Attempt to Audit the Fiscal Year 1999 Financial Statements,
00-FO-177-0003 (Washington, D.C.: n.d.), and OIG, Department of Education,
Accuracy of Student Aid Awards Can Be Improved by Obtaining Income Data From
the Internal Revenue Service, ACN: 11-50001 (Washington, D.C.: Jan. 29,
1997).

2. SSI is the nation's largest cash assistance program for the poor. In
1999, it paid about 6.6 million aged, blind, and disabled recipients more
than $28 billion in benefits. See Supplemental Security Income:
Administrative and Program Savings Possible by Directly Accessing State Data
(GAO/HEHS-96-163, Aug. 29, 1996).

3. Supplemental Security Income: Opportunities Exist for Improving Payment
Accuracy (GAO/HEHS-98-75 , Mar. 27, 1998).

4. We use "programs" to refer to individual programs as well as groups of
closely related programs. For example, Student Financial Assistance is a
group of several loan and grant programs including Pell Grants, Stafford
Loans, Direct Loans, PLUS Loans, and others.

5. The Office of Management and Budget (OMB), SSA, OCSE, Food and Nutrition
Services, the Department of Labor, and the Department of Treasury, including
IRS.

6. We will issue a separate report on the symposium in fall 2000.

7. This figure includes improper payments (as reported in the fiscal year
1999 financial statements) for HUD's project-based and tenant-based housing
programs. We did not, however, include the project-based housing programs in
our review.

8. HUD is permitted, however, to disclose to public housing agencies and
private owners the identity of individuals who have potential income
discrepancies. See 42 U.S.C. 3554(c)(2) (A)(ii).

9. This estimate was derived from cost-benefit analysis on the PARIS project
performed by ACF.

10. See 42 U.S.C. 653(m).

11. Providing such access may also require amending state laws.

12. However, this type of verification cannot identify wages individuals
receive that employers do not report.

13. HUD funds both tenant-based and project-based rental assistance
programs. Our review focused exclusively on the tenant-based programs
administered by public housing agencies.

14. This figure includes expenditures for the operating fund, capital fund,
debt service, drug elimination fund, and HOPE VI.

15. The states may carry forward funds that are not spent to meet program
needs in future years.

16. Taxpayer information may consist of an individual's name, Social
Security number, address, wages, self-reported earnings, unearned income
from interest and dividends, tax returns, and miscellaneous income
statements as well as other things. For a full description of IRC section
6103 and related issues, see Taxpayer Confidentiality: Federal, State, and
Local Agencies Receiving Taxpayer Information (GAO/GGD-99-164, Aug. 30,
1999).

17. See Joint Committee on Taxation, Study of Present-Law Taxpayer
Confidentiality and Disclosure Provisions as Required by Section 3802 of the
Internal Revenue Service Restructuring and Reform Act of 1998 (Washington,
D.C.: Jan. 28, 2000), p.46.

18. TANF administrators require additional information to accurately
determine applicants' and recipients' eligibility for assistance, including
interstate receipt of benefits, program disqualifications, and benefit time
limits.

19. See Taxpayer Confidentiality: Federal, State, and Local Agencies
Receiving Taxpayer Information (GAO/GGD-99-164 , Aug. 30, 1999).

20. IEVS was originally created by the Deficit Reduction of Act of 1984 and
was built on earlier state data sharing activities in the 1970s that used
state wage data to control welfare payments.

21. See section 353 of PRWORA and section 213 of the Foster Care
Independence Act of 1999, Pub. L. No. 106-169.

22. The estimate includes HUD's project-based housing programs for calendar
year 1998 and is statistically projected at the 95 percent confidence level,
plus or minus $133 million.

23. HUD matches its Multifamily Tenant Characteristics System against IRS
taxpayer information. However, it cannot provide federal tax data to PHAs
because they are not considered officers or employees of HUD. Section 6103
generally restricts access to taxpayer information to officers or employees
of agencies permitted to receive the information.

24. San Francisco,California; Hialeah and Miami-Dade, Florida; Baltimore,
Maryland; Newark, New Jersey; New York, New York.

25. PHAs are generally required to check all tenants' continued eligibility
for assistance annually or more often if circumstances warrant. An income
discrepancy report may result from one of HUD's periodic computer matches
with IRS or SSA. Moreover, failure to pay retroactive rent could result in
future ineligibility for housing assistance benefits. However, the Quality
Housing and Work Responsibility Act of 1998 (QHWRA) allows PHAs to provide
"flat rents" to tenants who meet certain qualifications. Under QHWRA, PHAs
may have to perform eligibility recertifications on tenants with flat rents
only once every 3 years rather than annually.

26. The Hialeah PHA uses credit bureau reports only for current tenants at
recertification.

27. According to PHA officials, this sample was judgmentally selected to
focus only on families that reported having no source of income. Thus, the
sample was not random or statistically projectable.

28. HUD has also taken other steps to improve its ability to verify tenants'
and applicants' eligibility, such as encouraging the states to share
automated wage data with public housing agencies.

29. In 1999, HUD transferred the income verification function from the
Office of Public and Indian Housing to REAC, which is primarily responsible
for the development, maintenance, and operation of systems used in computer
matching. It is also responsible for the administrative and technical
compliance of those systems within legal and regulatory requirements
concerning individual privacy.

30. This system contains individual-level information from various programs
such as TANF and SSI, as well as state wage information.

31. See section 213 of the Foster Care Independence Act of 1999, Pub. L. No.
106-169.

32. Adjusted gross income is a figure reported on IRS forms 1040, 1040A, and
1040EZ.

33. One such match is a comparison between the Central Processing System
file of applicant records and SSA data to verify the accuracy of Social
Security numbers; another matches up data with the National Student Loan
Data System to determine whether an applicant is in default on a federal
loan.

34. OIG, Department of Education, Accuracy of Student Aid Awards Can Be
Improved by Obtaining Income Data From the Internal Revenue Service,
ACN:11-50001 (Washington, D.C.: Jan. 29, 1997).

35. 34 C.F.R. 668.54(a)(2)(i) requires that institutions participating in
the title IV student aid programs verify key eligibility information for at
least 30 percent of the federal student aid applicants in attendance.

36. IRS Form 4506 can be used to give an applicant's consent to sending a
transcript of a tax return to a third party that the applicant designates.
Each return sent costs the applicant $23.

37. HEA's section 484(q) authorizes the Secretary of Education, in
cooperation with the Secretary of the Treasury, to confirm with IRS the
adjusted gross income, federal income taxes paid, filing status, and
exemptions applicants and their parents report on their federal income tax
returns for the purpose of verifying the information they report on student
financial aid applications. Education, IRS, and OMB were negotiating how to
grant Education limited access to taxpayer data when we completed this
report. The 1998 amendments reference IRC section 6103(l)(13) as a source of
authority for the release of tax return information to "officers or
employees of the Department of Education." According to Education officials,
if they were to perform an IRS match using FAFSA data, tax return
information would have to be released to Education contractors and to
individual schools.

38. IRS and Education are studying the feasibility of developing a system
that would enable financial aid applicants, on a pilot basis, to
electronically transmit consent forms to IRS over the Internet using
electronic signatures. However, such a system may not be available until at
least 2001.

39. Savings from data sharing activities in state TANF programs do not
translate into direct savings for the federal government, because the funds
have already been block-granted to the states.

40. PRWORA required the states to establish state directories of new hires
to assist in locating noncustodial parents for child support enforcement
purposes.

41. Florida officials reported that new hire data there are less timely,
from 4 to 7 months old.

42. Information on veterans' benefits is generally available to states from
the Veterans Administration. However, workers compensation data may be less
complete, because some states do not have data for self-insured employers
and benefits paid by insurance companies.

43. Some states also make use of other data sources that can provide
information on assets, such as DMVs or credit bureaus.

44. Not all banks doing business in the city participate. Additionally, the
bank match is not statewide, although certain large districts (counties)
outside New York City conduct their own bank matches.

45. New York City also uses local sources to obtain a variety of other kinds
of asset data, such as estates and liens, stocks and bonds, real property,
and legal settlements and awards.

46. The initial pilot, known as the State Online Query, includes Illinois,
Kentucky, Montana, North Carolina, Tennessee, Utah, Washington, and
Wisconsin. The pilot also allows states to verify applicants' and
recipients' Social Security numbers. In return, SSA will access state data
useful for controlling payments in the SSI program.

47. Under PRWORA, states can disqualify recipients from receiving benefits,
temporarily or permanently, for certain violations of program rules, such as
welfare fraud. PRWORA also set a maximum 5-year time limit on the receipt of
TANF benefits and allowed states to set shorter time limits. However, the
law did not establish a nationwide database of time limits that would allow
states to determine the number of benefit months remaining for clients who
have recently moved in from another state.

48. ACF, Report to Congress on Data Processing and Case Tracking in the
Temporary Assistance for Needy Families (TANF): Report of Data Processing
(Washington, D.C.: Dec. 1997).

49. See Department of Agriculture, Options for a National Database to Track
Participation in Federal Means-Tested Public Assistance Programs: Report to
Congress (Washington, D.C .: Nov. 1999), p. xii.

50. See Welfare Reform: Improving State Automated Systems Requires
Coordinated Federal Effort (GAO/HEHS-00-48 , Apr. 27, 2000).

51. The four states we visited had participated in PARIS.

52. A total of 11 states and the District of Columbia participated in the
August 1999 PARIS match.

53. GAO/HEHS-00-48 .

54. The Public Housing Tenant Integrity Act of 1997, a bill introduced in
January 1997, would have modified section 6103(l)(7) so that HUD could
redisclose federal tax return data to public housing agencies, but the bill
was never passed.

55. See Joint Committee on Taxation, Study of Present-Law Taxpayer
Confidentiality and Disclosure Provisions as Required by Section 3802 of the
Internal Revenue Service Restructuring and Reform Act of 1998 (Washington,
D.C.: Jan. 28, 2000), p. 9.

56. BSRT included information systems and program experts from the major
benefit programs in numerous federal departments, including Agriculture, the
Department of Defense, the Department of Labor, the Department of the
Treasury (including IRS), HHS, the Office of Personnel Management, SSA, and
the Veterans Administration.

57. GAO/HEHS-00-48 .

58. See OMB Circular A-130, sec. 9(h)(1), which provides governmentwide
information resources management policies as required by the Paperwork
Reduction Act of 1995, as amended.

59. See Office of Management and Budget: Future Challenges to Management
(GAO/T-GGD/AIMD-00-141 , Apr. 7, 2000).

60. OMB Benefit Systems Review Team, Strategies for Efficiency: Improving
the Coordination of Government Information Resources (Washington, D.C.: Jan.
1997).

61. "Pay and chase" refers to the labor-intensive and time-consuming
practice of trying to recover overpayments once they have already been made
rather than preventing erroneous payments in the first place.

62. According to the fiscal year 2001 budget document, OMB coordinates PMOs
with assistance from the National Partnership for Reinventing Government and
interagency working groups. Managers in the agencies have the primary
responsibility for achieving the objectives by implementing detailed action
plans, periodic reporting, and corrective action during the year.

63. See Budget of the United States Government, Fiscal Year 2001 (Washington
D.C.: 2000), p. 298.

64. Supplemental Security Income: Action Needed on Long-Standing Problems
Affecting Program Integrity (GAO/HEHS-98-158 , Sept. 14, 1998).

65. An independent auditor has found serious computer security weaknesses in
SSA's systems. SSA has put together a corrective action plan and is working
to address these deficiencies. See Financial Management: Agencies Face Many
Challenges in Meeting the Goals of the Federal Financial Management
Improvement Act (GAO/T-AIMD-00-178 , June 6, 2000).

66. See Congressional Budget Office, Congressional Budget Office Cost
Estimate: H.R. 1802, Foster Care Independence Act of 1999 (Washington, D.C.:
June 9, 1999).

67. IRS sets the standards for ensuring the security and confidentiality of
taxpayer information in Publication 1075.

68. In a recent report, we recommended that the Secretary of HHS establish
an interagency group to identify, and develop implementation plans for,
federal actions that would facilitate states' efforts to improve their
automated systems for federal programs that serve low-income families
(GAO/HEHS-00-48). Such a system could also be used to enhance data sharing
for more accurate eligibility determinations.

69. GAO/GGD-99-164 .

70. See our section above on TANF.
*** End of document. ***