Means-Tested Programs: Information on Program Access Can Be an Important Management Tool

Means-Tested Programs: Information on Program Access Can Be an
Important Management Tool (11-MAR-05, GAO-05-221).

Federal agencies that administer means-tested programs are
responsible for both ensuring that people have appropriate access
to assistance and ensuring the integrity of the programs they
oversee. To balance these two priorities appropriately, it is
important for agencies to have information on program integrity
and program access. Knowing the proportion of the population that
qualifies for these programs relative to the numbers who actually
participate can help ensure that agencies can monitor and
communicate key information on program access. To better
understand participation in low-income programs, this report
provides information on: (1) the proportion of those eligible who
are participating in 12 selected low-income programs; (2) factors
that influence participation in those programs; and (3)
strategies used by federal, state, and local administrators to
improve both access and integrity, and whether agencies monitor
access by measuring participation rates.
-------------------------Indexing Terms-------------------------
REPORTNUM: GAO-05-221
ACCNO: A19137
TITLE: Means-Tested Programs: Information on Program Access Can
Be an Important Management Tool
DATE: 03/11/2005
SUBJECT: Child care programs
Disadvantaged persons
Entitlement programs
Federal aid programs
Federal aid to states
Federal social security programs
Food programs for children
Food relief programs
Housing programs
Income maintenance programs
Medicaid
Program management
Public assistance programs
Statistical data
Strategic planning
Family support programs
Vouchers
Earned Income Tax Credit
Federal Pell Grant Program
Food Stamp Program
Head Start Program
HHS Child Care and Development Fund
HUD Housing Choice Voucher Program
HUD Public Housing Program
Special Supplemental Nutrition Program
for Women, Infants and Children

SSA Supplemental Security Income
Program

State Children's Health Insurance
Program

Temporary Assistance for Needy Families
Program

******************************************************************
** This file contains an ASCII representation of the text of a **
** GAO Product. **
** **
** No attempt has been made to display graphic images, although **
** figure captions are reproduced. Tables are included, but **
** may not resemble those in the printed version. **
** **
** Please see the PDF (Portable Document Format) file, when **
** available, for a complete electronic file of the printed **
** document's contents. **
** **
******************************************************************
GAO-05-221

United States Government Accountability Office

GAO Report to the Ranking Minority Member, Committee on the Budget, House
of Representatives

March 2005

MEANS-TESTED PROGRAMS

Information on Program Access Can Be an Important Management Tool

GAO-05-221

[IMG]

March 2005

MEANS-TESTED PROGRAMS

Information on Program Access Can Be an Important Management Tool

What GAO Found

For 12 federal programs supporting low-income people, we found that the
proportion of those eligible who are enrolled varies substantially both
between and within programs. Among entitlement programs-those programs
that provide benefits to all applicants that meet program eligibility
criteria-these rates range from about 50 to more than 70 percent. Among
non-entitlement programs-those with limited funding-these rates ranged
from less than 10 percent to more than 50 percent. While it may be neither
feasible nor desirable for programs to serve 100 percent of those eligible
for benefits, information on the share of those eligible who are enrolled
in means-tested programs and on particular recipient groups such as the
elderly or families with children, can help program managers more
effectively address issues related to program access. However,
participation rate estimates must be interpreted carefully because of
limitations in the data sources and estimation methodologies used to
calculate the estimates.

Many factors influence access to low-income programs-including the type of
benefits, ease of access, misperceptions about program requirements, and
application and eligibility verification procedures. These factors can
impact not only the share of eligible people who participate in low-income
programs, but other aspects of program access as well, including the
composition of the program caseload and how programs work together to
serve low-income individuals and families.

Federal, state, and local administrators have implemented many strategies
to achieve the goals of access and integrity, but federal agencies
generally put more emphasis on tracking information and outcomes related
to program integrity than program access. To better ensure that program
administrators achieve program integrity goals, agencies have begun to
develop measures to track and report on program integrity. Federal
agencies have developed participation rate estimates for several
low-income programs, but only four-CCDF, food stamps, WIC, and EITC-either
currently collect and report information on the extent to which they are
reaching their target populations or plan to do so. Such information can
guide administrators in setting priorities and targeting scarce resources,
even among programs that were not intended to serve everyone eligible for
program benefits.

Health and Human Services Agriculture Department Temporary Assistance for
Needy Families . Food Stamps Child Care and Development Fund . Women,
Infants, and Children Nutrition Head Start Internal Revenue Service
Medicaid . Earned Income Tax Credit

State Children's Health Insurance Program Social Security Administration
Housing and Urban Development . Supplemental Security Income

Housing Choice Vouchers Education DepartmentPublic Housing . Earned Income
Tax Credit

Source: GAO.

United States Government Accountability Office

Contents

Letter

Results in Brief
Background
Use of Federal Low-Income Assistance Varies Greatly by Program

and by Subgroup
Many Factors Influence Participation but Their Impact on
Entitlement and Non-Entitlement Programs Differs

Program Administrators Have Strategies That Improve Both
Access and Integrity, but Federal Agencies Have Generally
Focused More on Measuring Program Integrity Outcomes

Conclusion
Recommendations to Executive Agencies
Agency Comments

3 5

35 47 48 50

Appendix I Objectives, Scope, and Methodology

Appendix II Limitations Affecting Use of Participation and Coverage Rates

Appendix III Agency Estimates of Improper Payments for Programs Reviewed

Appendix IV Comments from the Department of Education

Appendix V Comments from the Department of Health and
Human Services 68

Appendix VI Comments from the Department of the Treasury 78

Appendix VII Comments from the Social Security Administration 80

Appendix VIII GAO Contacts and Staff Acknowledgments 83

GAO Contacts 83 Staff Acknowledgments 83

Related GAO Products

Tables

Table 1: Fiscal Year 2003 Annual Expenditures and Agencies Responsible for
Administering 12 Selected Low-Income Programs

Table 2: Descriptions of 12 Low-Income Programs Table 3. Level of
Government Responsible for Funding and Design of 12 Low-Income Programs
Table 4: Estimated Participation Rates for Entitlement Programs for the
Most Recent Year Data Were Available

Table 5: Available Subgroup Participation Rate Estimates for Entitlement
Programs for the Most Recent Year Data Were Available

Table 6: Estimated Coverage Rates for Non-Entitlement Programs for the
Most Recent Year Data Were Available

Table 7: Coverage Rate Estimates for TANF Cash Assistance and WIC
Subgroups for the Most Recent Year Data Were Available

Table 8: Potential Cost of Providing Benefits to Eligible Nonparticipants
for the Most Recent Year Data Were Available Table 9: Status of Agencies
Identification of and Reporting on the Amounts of Improper Payments by
Program

Table 10: Status of Agencies Efforts to Use Information on Participation
or Coverage Rates in Managing Their Programs

Table 11: Summary of Participation and Coverage Rate Methodologies Table
12: Improper Payment Estimates Reported in Agency Fiscal Year 2003
Performance and Accountability Reports

6 7

14 19

22 26 43

45 54 65

Figures

Figure 1: Food Stamp Participation Rate Estimates for Fiscal Years 1999 to
2002 17 Figure 2: Head Start Coverage Rate Estimates for 1997 to 2003 24

Abbreviations

ACF Administration for Children and Families
CCDF Child Care and Development Fund
CMS Centers for Medicare & Medicaid Services
CPS Annual Social and Economic Supplement to the

Current Population Survey ED Department of Education EITC Earned Income
Tax Credit FNS Food and Nutrition Service FRED Fraud Early Detection FY
fiscal year HCV Housing Choice Vouchers HHS Department of Health and Human
Services HIPAA Health Insurance Portability and Accountability Act HUD
Department of Housing and Urban Development IPIA Improper Payments
Information Act IRS Internal Revenue Service NRC National Research Council
OMB Office of Management and Budget PRWORA Personal Responsibility and
Work Opportunity

Reconciliation Act SCHIP State Children's Health Insurance Program SIPP
Survey of Income and Program Participation SSA Social Security
Administration SSBG Social Security Block Grant SSI Supplemental Security
Income program TANF Temporary Assistance for Needy Families TRIM3 Transfer
Income Microsimulation Model, version 3 USDA Department of Agriculture WIC
Special Supplemental Nutrition Program for Women

Infants, and Children

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

United States Government Accountability Office Washington, DC 20548

March 11, 2005

The Honorable John M. Spratt, Jr.
Ranking Minority Member
Committee on the Budget
House of Representatives

Dear Mr. Spratt:

Each year through more than 80 means-tested programs, the federal
government provides benefits and services to individuals and families with
low incomes-just 12 of which account for as much as $330 billion in
annual federal expenditures. As stewards of these funds, federal agencies
have many roles, among them: to ensure that people have appropriate
access to this assistance and to ensure the integrity of these programs by
guarding against improper payments and unqualified participation. These
responsibilities can be complementary, but they can also be practiced in
ways that work at cross purposes. Outreach without appropriate
screening, for example, can result in service to the wrong recipients, but
cumbersome enrollment procedures can discourage those who qualify
from applying. To meet requirements related to the Improper Payments
Information Act of 2002, federal agencies administering these programs
will be required to institute new steps to assure integrity by annually
measuring and reporting improper payments. As they step-up efforts to
ensure integrity, they will be challenged, as well, to ensure appropriate
access. Without a sense of the proportion of the population that qualifies
for these programs relative to the numbers who actually participate,
however, it may be difficult for most agencies to know whether they have
struck an appropriate balance in the weight of their strategies and
whether
they need to pursue methods that better serve both access and integrity.

To better understand participation in low-income programs, we have
agreed to provide you information on: (1) the proportion of those eligible
who are participating in 12 selected low-income programs; (2) the factors
that influence participation in those programs; and (3) strategies used by
federal, state, and local administrators to improve both access and
integrity, and whether agencies monitor access by measuring participation
rates. The 12 programs included in this review are among the largest and
were selected to cover a range of benefits and services aimed at
supporting needy families and individuals. They include: the Child Care
and Development Fund (CCDF), the Earned Income Tax Credit (EITC),

the Special Supplemental Nutrition Program for Women, Infants, and
Children (WIC), the Food Stamp Program, the Supplemental Security Income
program (SSI), the Head Start program, the Pell Grant program, the
Medicaid program, the State Children's Health Insurance Program (SCHIP),
the Temporary Assistance for Needy Families (TANF) program, the Housing
Choice Voucher (HCV) program, and the public housing program.

To address our research objectives, we compiled estimates of the
proportion of those eligible who participate in each of the programs for
the most recent year for which data were available1 using different
methodologies. We refer to this proportion as a participation rate when
discussing entitlement programs and other programs that provide program
benefits to all eligible applicants: EITC, food stamps, Medicaid, Pell
Grants, and SSI. We refer to it as a coverage rate when discussing
nonentitlement programs that do not necessarily provide benefits to all
eligible individuals who apply for the program: CCDF, Head Start, Housing
Choice Vouchers, Public Housing, SCHIP, TANF, and WIC. For seven programs,
we contracted with the Urban Institute to provide us with information
about their microsimulation model estimates of participation rates.2 These
model estimates are also used to calculate estimates of the potential cost
of serving eligible nonparticipants in a subset of programs. For the
remaining programs, we provide participation or coverage rate estimates
based on administrative and national survey data. We held discussions with
federal agency officials to discuss the reliability of program data used
in our estimates, reviewed related documentation, and ensured that the
agencies conducted tests of the data for omissions and errors. In
addition, we reviewed literature on factors that influence participation
in the 12 programs covered by the review and strategies to improve program
access and program integrity. We surveyed and interviewed the federal
agencies that administer the 12 programs on factors they have identified
that influence program participation and on

1The most recent participation rate estimates are available for the years
1999 to 2003 depending on the program.

2The Transfer Income Microsimulation Model (TRIM3), developed by the Urban
Institute and funded by HHS, uses national survey data from the Annual
Social and Economic Supplement to the Current Population Survey (CPS), the
Survey of Income and Program Participation (SIPP), the National Survey of
America's Families, and information on eligibility rules to estimate the
population eligible for means-tested programs, including CCDF, food
stamps, Medicaid, SCHIP, SSI, TANF, and WIC. The model estimates the
amount each individual or household would receive based on their
characteristics.

Results in Brief

strategies they have implemented to improve program access and program
integrity. In addition, we interviewed federal, state, and local program
administrators of various programs in California, Connecticut, Georgia,
Maryland, and Minnesota, on factors that affect participation and
strategies they have employed to address the goals of program access and
program integrity. These states were chosen on the basis of innovative
programs and initiatives related to program integrity and program access
and demographic and geographic diversity. We performed this work between
November 2003 and January 2005 in accordance with generally accepted
government auditing standards. See appendix I for additional information
on scope and methodology.

For 12 federal programs supporting low-income people, we found that the
proportion of those eligible who are actually enrolled varies
substantially both between programs and among subgroups enrolled in a
single program. Given the differences in the goals, design,
administration, and funding of the 12 programs, it is not surprising that
we found substantial variation in these proportions. For entitlement
programs-those designed to support all those who qualify-the estimated
proportion of eligible people who were enrolled ranged from about 50
percent to more than 70 percent. For example, approximately three-quarters
of those who were eligible took advantage of the Earned Income Tax Credit
program in 1999, the most recent year for which data were available. For
nonentitlement programs-those with limited funding and not necessarily
intended to cover all eligible persons-the estimated proportion of the
eligible who were enrolled ranged from less than 10 percent to more than
50 percent. In the Head Start program, for example, funded enrollment was
sufficient to cover half of all potentially eligible children in 2003.
Within some programs, we found that some subpopulations were enrolled in
different proportions than other groups. For example, in the Food Stamp
Program, the participation rate for families with children was higher than
the participation rate for other groups. Some evidence shows that
enrollees tended to be those among the eligible who had greater needs-as
indicated by the levels of assistance for which they qualified. While
information on participation rates can help program managers more
effectively address issues related to program access, several factors must
be taken into consideration when interpreting this information. For
example, because of limitations of the data and methodologies used to
measure rates of participation, we could not make statistically reliable
comparisons across programs nor could we describe fluctuations over time
for most programs.

The size and type of program benefits, ease of access, misperceptions
about program requirements and eligibility verification requirements put
in place to ensure program integrity can affect program access in both
entitlement and non-entitlement programs. For example, agency officials
told us that participation in the WIC program is highest among infants in
part because infants are eligible for the highest benefits, but that
factors such as extended office hours can also impact participation. The
factors we identified can affect several aspects of program access,
including not only the number of program participants, but also the extent
to which programs reach the total population eligible for benefits and
services, how well agencies are allocating resources among targeted
subpopulations, and, from an agencywide or even governmentwide
perspective, how well specific means-tested programs complement and
interact with programs that serve similar populations or provide for
similar needs. While concerns about the allocation of scarce resources and
interactions with other programs apply to all low-income programs, they
may be of particular importance to non-entitlement programs whose
participation is limited by funding constraints. Participation or coverage
rate information can guide program administrators in setting priorities
and targeting scarce resources, even among programs that were not intended
to serve everyone eligible for program benefits.

Federal, state, and local administrators have implemented a variety of
strategies to achieve both program access and program integrity, but
federal agencies generally put more emphasis on tracking information and
outcomes related to program integrity than program access. Through federal
agency surveys and site visits, we identified several strategies currently
being used that have the potential to achieve two of the fundamental goals
common to all means-tested programs-program access and program integrity.
These strategies ranged from innovative use of information technology to
special outreach programs; some were in use nationwide and others were
state or local efforts. The initiation and control of these strategies is
often beyond the direct control of the federal agencies managing the
programs. However, federal managers do play a role in that they encourage
and facilitate such strategies, even in the most decentralized program, by
emphasizing the importance of program access and program integrity-even
absent specific related laws or regulations related to these issues. The
Improper Payments Information Act of 2002 has built upon existing
govermentwide efforts to emphasize to federal agencies the importance of
having an internal control framework to address program integrity issues.
At the same time, internal controls are important management tools to
ensure that agencies meet their basic goals of reaching eligible families.
The federal agencies that oversee these

entitlement and non-entitlement programs can benefit from having up-todate
information on the extent to which their programs reach eligible
individuals. However, while all of the programs we reviewed have begun to
take steps to measure the extent of improper payments and all make efforts
to assess how well their programs are working, only four of the programs
we covered in this review-CCDF, food stamps, WIC, and EITC-either
currently collect and report information on the extent to which they are
reaching their target populations in key performance and program reports
or plan to do so.

To better ensure that agencies have information on program access, we
recommend to the Secretaries and Commissioners whose programs do not
currently use participation rate information to consider the use of
participation rate information in managing their programs. We also make
some technical recommendations on how the usefulness of these measures
could be improved. The agencies generally agreed with these
recommendations.

Background The federal government funds a wide array of programs intended
to provide benefits or services to low-income individuals, families, and
households. The 12 programs included in this review include the largest of
these programs, Medicaid, as well as some relatively small programs such
as WIC. The 12 programs are administered by seven different federal
agencies.3 Table 1 shows the agencies responsible for each program and
federal expenditures for fiscal year 2003, the most current year
available.

3Two of these agencies, the Centers for Medicare & Medicaid Services and
the Administration for Children and Families, are part of the Department
of Health and Human Services.

Table 1: Fiscal Year 2003 Annual Expenditures and Agencies Responsible for
Administering 12 Selected Low-Income Programs

FY 2003 federal
expenditures a
Program Agency (dollars in billions)
CCDF Department of Health and Human $7.3
Services
(HHS), Administration for
Children and
Families (ACF)
EITC Department of Treasury, Internal 37.9b
Revenue
Service (IRS)
Department of Agriculture
Food Stamp Program (USDA), Food
and Nutrition Service (FNS)
Head Start HHS/ACF
Housing Choice 13.4c
Voucher Program Department of Housing and Urban
Development (HUD), Office of
Public and
Indian Housing
Medicaid HHS, Centers for Medicare & 160.7
Medicaid
Services (CMS)
Federal Pell Grant Department of Education (ED),
Program Office of
Federal Student Aid
Public Housing HUD, Office of Public and Indian 7.6d
Program Housing
SCHIP HHS/CMS
WIC USDA/FNS
SSI Social Security Administration
(SSA)
TANF HHS/ACF $16.3

Source: Federal agency officials.

aThis includes the total amount of federal expenditures in fiscal year
2003, which is the most recent year for which these data were available.
It includes current and prior year federal funds expended in fiscal year
2003; it does not include state expenditures.

bThis includes both credits paid out in refunds and reduced tax
liabilities.

cThis includes funding for the housing choice voucher and moderate
rehabilitation programs.

dThis includes expenditures for the public housing capital fund, the
public housing operating fund, and revitalization of severely distressed
public housing (HOPE VI).

These programs provide different benefits and services to different target
populations. Some of the programs provide benefits that phase out
gradually, so some people may be eligible for a very low benefit. For
example, in fiscal year 2005, monthly food stamp benefits for a
threeperson household can be as low as $1 per month. Other programs, such
as Medicaid, generally offer comparable benefits to every resident of a
state who meets eligibility requirements. Table 2 provides a brief
description of each of the programs covered in this report and the types
of benefits offered by each program.

Table 2: Descriptions of 12 Low-Income Programs

Program Description

CCDF The Personal Responsibility and Work Opportunity Reconciliation Act
of 1996 (PRWORA) authorized CCDF to help provide child care to low-income
families and to give states more flexibility to design child care
policies. CCDF is funded through an annual block grant to states and
required state funds and may also include transfers from TANF. States are
permitted to transfer up to 30 percent of their TANF block grant to CCDF.
Under CCDF, child care services are available to eligible families through
certificates or contracts with providers, and parents may select any
legally operating child care provider, including home-based or
center-based providers.

EITC EITC is a refundable federal tax credit available to eligible workers
earning relatively low incomes. Because the credit is refundable, a person
does not need to owe taxes to receive benefits. EITC is based on earned
income, adjusted gross income, and the presence of qualifying children, if
any. Under current law, there are three categories of EITC recipients:
childless adults, adults with one child, and adults with two or more
children.

Food Stamps The Food Stamp Program is the primary source of nutrition
assistance for many low-income households. It enables eligible low-income
households to buy nutritious food with electronic benefit cards at
authorized retail food stores across the country. State and local welfare
offices operate the program, and the federal government oversees the state
operation of the program. Participants must meet income and resource
standards and be U.S. citizens or eligible non-citizens, and all
able-bodied individuals between 16 and 60 without dependents must register
for work, take part in an employment and training program, or accept or
continue suitable employment. Households with incomes are expected to
spend about 30 percent of their income, after certain deductions, on food.

Head Start Head Start provides comprehensive developmental services for
low-income, pre-school children ages 3 to 5, and social services for their
families. Services can be provided through either a half-day or a full-day
program. Head Start provides diverse services in four components:
education, health, parent involvement, and social services. Grants are
awarded to about 1,400 local public or private non-profit agencies, and
the community must contribute about 20 percent of the total cost of a Head
Start program.

Housing Choice Vouchers The Housing Choice Voucher program is the federal
government's major program for assisting very low-income families, the
elderly, and the disabled with their housing needs. HUD gives public
housing agencies the funds to administer the program. Participants use
vouchers to find their own housing, including single-family homes,
townhouses and apartments. The voucher recipient pays the difference
between the actual rent charged by the landlord and the amount subsidized
by the program. Generally, the subsidy allows the tenant to pay no more
than 30 percent of adjusted monthly income towards the rent and utilities.

Medicaid Medicaid (title XIX of the Social Security Act) is a
federal/state entitlement program that generally provides health insurance
coverage for low-income families and individuals who are aged or disabled.
Medicaid is the largest source of funding for medical and health-related
services for America's poorest people. Within broad federal guidelines,
each state can (1) establish its own eligibility standards; (2) determine
the type, amount, duration, and scope of services; (3) set the rate of
payment for services; and (4) administer its own program. Medicaid
policies for eligibility, services, and payment are complex and vary
considerably, even among states of similar size or geographic proximity.

Pell Grant The Pell Grant program provides grants (i.e., aid that does
not have to be repaid) to needy undergraduates. It is the largest source
of grant aid for postsecondary education attendance funded by the federal
government and provided an estimated $13.1 billion to students in fiscal
year 2004. Pell Grants are intended to be the foundation for all federal
aid awarded to undergraduates and constituted an estimated 19 percent of
all federally supported aid in fiscal year 2004 that benefited
postsecondary education students. For fiscal year 2004-2005, grants ranged
from $400 to $4,050.

Program Description

Public Housing Public housing was established to provide decent and safe
rental housing for eligible low-income families, the elderly, and persons
with disabilities. Public housing comes in all sizes and types, from
scattered single family houses to high-rise apartments for elderly
families. HUD contributes both capital and operating funds to local
housing agencies that manage the housing for low-income residents at rents
they can afford. HUD also furnishes technical and professional assistance
in planning, developing, and managing these developments.

SCHIP As part of the Balanced Budget Act of 1997, Congress created SCHIP
as a federal/state partnership, similar to Medicaid, with the goal of
expanding health insurance to low-income children-i.e., those children
whose families earn too much money to be eligible for Medicaid, but not
enough money to purchase private insurance. States can (1) use SCHIP funds
to expand Medicaid eligibility to children who previously did not qualify
for the program; (2) design a children's health insurance program entirely
separate from Medicaid; or (3) combine both the Medicaid and separate
program options.

WIC WIC provides supplemental foods to low-income women, infants, &
children up to age 5 who are at nutritional risk. The WIC target
population consists of low-income, nutritionally at risk: pregnant and
breastfeeding women; nonbreastfeeding postpartum women; infants (up to 1st
birthday), and children up to their 5th birthday. It provides supplemental
nutritious foods, including infant formula; nutrition education and
counseling at WIC clinics; and screening and referrals to other health,
welfare and social services.

SSI The SSI program, title XVI of the Social Security Act, was enacted in
1972 and implemented in 1974 to ensure a minimum cash income to all aged,
blind, or disabled persons. SSI is provided to eligible individuals or
couples who have limited income and resources (the countable resource
limit is $2,000 for an individual and $3,000 for a couple). Federal SSI
benefits are paid from federal general revenues, but many states also
supplement payments.

TANF The Personal Responsibility and Work Opportunity Reconciliation Act
of 1996 (PRWORA) replaced Aid to Families with Dependent Children (AFDC)
with TANF, which marked the end of federal entitlement to assistance. TANF
ended unlimited matching funding for family cash welfare and created
fixed-block grants to states. The block grant covers benefits,
administrative expenses, and services targeted to needy families and gives
states great flexibility to design their own TANF programs. PRWORA imposed
a 5-year limit on TANF cash assistance paid with federal funds and
required states to achieve minimum participation rates in federally
recognized work activities. In addition, they must spend a specified
amount of state funds on eligible low-income families-at least 75 percent
of the state funds they spent in fiscal year 1994, known as the
maintenance-of-effort requirement. States may use these funds to pay for
separate state programs.

Program Administration

Source: Federal agency documents.

The programs covered by this review represent both entitlement and non
entitlement programs; programs that are administered entirely by federal
agencies and those that are administered through federal partnerships
with state and/or local agencies; programs that allow substantial state
and
local variation and those that provide uniform benefits throughout the
country. Table 3 summarizes the level of government with whom
responsibility for funding and design resides for each of the
12 programs.

Table 3. Level of Government Responsible for Funding and Design of 12
Low-Income Programs

Program Fundinga Designb Entitlement
EITC Federal Federal Yes
Food Stamp Program Federal Federal Yes
Medicaid Federal/state Federal/state Yes
SSI Federalc Federal Yes
Pell Grants Federal Federal No
CCDF Federal/state Federal/state No
Head Start Federal Federal/local No
Housing Choice Voucher Federal Federal No
Public Housing Federal Federal No
SCHIP Federal/state Federal/state No
TANF cash assistance Federal/state Federal/state No
WIC Federal Federal No

Program Performance Management

Source: GAO.

aDefined as the level of government that supplies the primary source of
funding for the support. If substantial funding comes from more than one
source, we list both sources. Some additional funding may come from
sources not listed in the table.

bDefined as the level of government that is primarily responsible for
availability, eligibility, and benefit amount determination.

cSome states contribute to the SSI program, but state funding is not
required and is not provided in all states.

Although state and local agencies responsible for administering several of
the programs have some control over program design and implementation,
primary responsibility for setting priorities, guiding policy, and
measuring performance rests with the federal agencies that oversee the
programs. The programs covered by this review may have many different
goals and objectives of varying degrees of importance, but all of these
programs- regardless of size, target population, funding structure, or
types of benefits offered-were established to assist persons with limited
income, and as such, all were tasked with the overarching goal of reaching
and serving those eligible for program benefits or services. Likewise, all
means-tested programs share the common goal of ensuring that the funds
allocated for program benefits are provided only to those eligible.

A key factor in achieving desired program outcomes, including program
access and program integrity, is the implementation of appropriate
internal control. As discussed in our prior work on this topic, internal
control is an integral component of an organization's management that

provides reasonable assurance that agencies are achieving outcomes related
to the effectiveness and efficiency of operations and compliance with laws
and regulations, among other things.4

Federal agencies have several key management tools and program reports
that they use to provide information about the programs they administer,
develop policy, help with management decision-making, and help focus
program administrators at all levels of government on achieving the
federal agency's highest priorities. For example, through their strategic
and annual performance plans and accountability reports, federal agencies
set program goals, measure program performance against those goals, and
report publicly on their progress. Some programs also issue annual or
biennial reports to Congress on specific programs they administer; these
generally include important program information.

A number of governmentwide initiatives have resulted in increased emphasis
by program managers on program integrity issues, including OMB guidance,
the President's Management Agenda, GAO's High-Risk series and the Improper
Payments Information Act of 2002 (IPIA). The IPIA requires the head of
each federal agency to annually review all programs and activities that
the agency administers and to identify all such programs and activities
that may be susceptible to significant improper payments.5 For each
program and activity identified, the agency is required to estimate the
annual amount of improper payments and submit those estimates to Congress
before March 31 of the following applicable year. OMB guidance then
directs federal agencies to include a measure of improper payments in
their annual Performance and Accountability Reports. All 12 programs we
reviewed are subject to these requirements.

4See GAO, Standards for Internal Control in the Federal Government,
GAO/AIMD-00-21.3.1(Washington, D.C.: Nov. 1999).

5Improper payments can include both overpayments and underpayments.

Use of Federal Low-Income Assistance Varies Greatly by Program and by Subgroup

The proportion of those eligible who are actually enrolled in 12 selected
low-income programs varies substantially both between and within programs,
but several factors must be considered to understand the implications of
this information for program access. The estimated proportion of eligible
people who were enrolled in entitlement programs- those designed to
support all those who apply and qualify-ranged from about 50 percent to
more than 70 percent. In contrast, the estimated proportion of the
eligible who were enrolled in non-entitlement programs-those with limited
funding and not necessarily intended to cover all eligible persons-ranged
from less than 10 percent to about 50 percent. Within programs, we found
that subpopulations were enrolled in different proportions. Some evidence
also suggests that enrollees in some programs, such as the Food Stamp
Program, tend to be those who are eligible for larger benefit amounts.

For Entitlement Programs, the Proportion of Those Eligible Who Were
Enrolled Ranged from about 50 Percent to More Than 70 Percent, but
Participation Rates Are Higher among Some Groups

For four of the five entitlement programs we examined, the proportion of
eligible people who were enrolled varied from around 50 percent in the
Food Stamp Program to more than 70 percent in the EITC and SSI programs;
and within programs, different types of participants, such as children or
the elderly, were enrolled in varying proportions.6 Table 4 provides
estimated participation rates for each of these programs, expressed in
ranges to reflect potential errors in sample survey data. The actual
participation rates may be outside the range of estimates shown because
some types of uncertainty in the estimation of participation rates cannot
be quantified. Estimates for the Food Stamp, Medicaid and SSI programs
were developed by the Urban Institute under contract with HHS and Food
Stamp Program estimates were developed by Mathematica Policy Research,
Inc., under contract with USDA. HHS and USDA are the only two agencies to
regularly estimate participation rates in low-income programs. We
estimated the EITC participation rate for 1999. However, we were unable to
estimate an enrollment rate for the Pell Grant program because of concerns
about the reliability of some of the data needed to estimate the rate.

6While the Pell Grant program is not an entitlement program, we have
included it in this group because program administrators told us that they
manage it to allow all eligible people who apply to receive benefits.

Table 4: Estimated Participation Rates for Entitlement Programs for the
Most Recent Year Data Were Available

Participation rate
estimates b
Program (year)a Eligibility unit (in percent)
EITC (1999) Households 75c
Food Stamp Program (2001): dHouseholds 46-48
HHS/Urban Institute
(2002): USDA/Mathematica Policy Households 48e
Research, Inc.
Individuals 54f
Medicaid (2000) Individualsg 66-70
Pell Grants Not available Not availableh
Individuals and
SSI (2001) married
i couples 66-73

Sources: GAO's analysis of CPS and IRS survey data and data from the Urban
Institute's TRIM3 model and Mathematica Policy Research, Inc.

aAll estimates are for the calendar year indicated in the table, except
the food stamp estimates created by Mathematica, which are for fiscal year
2002.

bThis range represents a 95 percent confidence interval based on estimated
sampling error in the national survey data. This interval does not take
into account errors introduced by the modeling process or by
administrative data. We assessed but did not quantify these errors.

cThe EITC participation rate shown has a sampling error that does not
exceed plus or minus 2.7 percentage points. The number of eligible tax
filers used in this estimate was adjusted to exclude those who received
EITC benefits in error. The participation rate is a conservative estimate
based on 32 percent of EITC dollars being claimed in error; the IRS
estimated between 27 and 32 percent of these tax dollars were paid in
error in 1999. Because more recent information about EITC payment errors
is not available, we were unable to provide a more recent estimate.
However, Congress has enacted new tax laws and the IRS has taken steps to
improve compliance. For more details, see GAO, Earned Income Tax Credit
Participation, GAO-02-290R (Washington, D.C.: Dec. 14, 2001).

dWhile the Urban Institute and Mathematica each generate food stamp
participation rate estimates using data from the CPS, their estimates
differ slightly. The differences arise primarily from differences in
imputation methodology for data missing from the CPS-such as assets,
immigration/refugee status, and information on which household members buy
and prepare food together.

e,fConfidence intervals are not available. These estimates are from
Mathematica Policy Research, Inc., Trends in Food Stamp Program
Participation Rates: 1999 to 2002. (Washington, D.C.: Sept. 2004).

gEstimates do not include individuals who are institutionalized. Estimates
do account for variation in state eligibility rules and for people who may
be eligible for only part of the year.

hWe were unable to estimate a participate rate for the Pell Grant program
because we were unable to assess the reliability of data on the family
income of students that did not apply for federal financial assistance.

iNo reliable national survey data exist on individuals who meet the SSI
disability criteria. Therefore, researchers made assumptions about a
person's disability based on self-reported information on inability to
work due to illness or other disability, lack of work activity in the
prior year, and receipt of disability income. Some of those assumed
eligible for SSI based on disability may not meet the criteria and vice
versa. As a result, the models may overestimate or underestimate the
number of eligible people. This estimate also does not include
institutionalized individuals and disabled children.

These estimates indicate that between half and three quarters of those
eligible are participating in four of the entitlement programs. Although
differences in years, data sources, and estimation methodologies make it
inappropriate to compare participation rates across programs, these
estimates provide a general sense of the extent to which programs are
reaching those eligible for benefits. Specifically, we found:

o For the EITC program, we estimated in prior work that about 75 percent
of eligible households took advantage of this credit in 1999. This
estimate was adjusted to account for tax dollars that were distributed in
error that year.7 However, Congress has enacted new tax

laws and the IRS has taken steps to improve EITC program

compliance. o Two organizations, the Urban Institute and Mathematica
Policy Research, Inc., have estimated food stamp participation rates using
different methodologies but both found that about 48 percent of eligible
households participated in this program in 2001 and 2002, respectively. In
addition, Mathematica found that 54 percent of eligible individuals were
enrolled in 2002.

o An estimated 66 to 70 percent of eligible, noninstitutionalized people
were enrolled in Medicaid in 2000. This estimate does not include
institutionalized people, such as those in nursing homes.

o We were unable to estimate Pell Grant participation rates because we
were unable to assess the reliability of available data on the family
income of enrolled students who did not apply for federal financial aid.

o For SSI, an estimated 66 to 73 percent of eligible adult individuals
and married couples received benefits in 2001. This estimate also does not
include institutionalized people or disabled children.

Participation rates also can vary substantially even within a single
program. Disaggregating program participation rates by targeted
subpopulations can help us to better understand the implications of
overall program participation rates. In table 5, we provide participation
rate estimates for selected subgroups within the four entitlement

7For more details, see GAO, Earned Income Tax Credit Participation,
GAO-02-290R (Washington, D.C.: Dec. 14, 2001).

programs for which we have participation rate estimates. As in table 4,
these estimates are expressed in ranges to reflect potential errors that
result from using sample survey data. (See the footnotes to table 5 for
more details.)

Table 5: Available Subgroup Participation Rate Estimates for Entitlement
Programs for the Most Recent Year Data Were Available

Participation rate
estimates b
Program (year)a Subgroup (percent)
EITC (1999) cAll Households 75d
Households with No children 45e
Households with 1 Child 96r
Households with 2 Children 93g
Households with 3 or More 63h
Children
Food Stamps (2001) iAll Households 46-48
Households with Children 55-57
Households with Elderly 27-28
Members
Medicaid (2000) All Individualsj 66-70
Adults 56-64
Children 74-79
Elderly 40-43k
SSI (2001) lAll Individuals and Married 66-73
Couples
Elderly 61-68m

Source: GAO's analysis of CPS and IRS survey data and the Urban
Institute's TRIM3.

aAll estimates are for the calendar year mentioned.

cThe number of eligible tax filers used in these estimates were adjusted
to exclude those who received EITC benefits in error. The participation
rate is a conservative estimate based on 32 percent of EITC dollars being
claimed in error; the IRS estimated between 27 and 32 percent of EITC
dollars claimed in 1999 were paid in error. Because more recent
information about EITC payment errors is not available, we were unable to
provide a more recent estimate. However, since 1999, Congress has enacted
new tax laws and the IRS has taken steps to improve compliance. For more
details, see GAO, Earned Income Tax Credit Participation, GAO-02-290R
(Washington, D.C.: Dec. 14, 2001).

dThe EITC participation rate for all households has a sampling error that
does not exceed plus or minus 2.7 percentage points.

eThe EITC participation rate for households with no children has a
sampling error that does not exceed plus or minus 3.9 percentage points.

fThe EITC participation rate for households with one child has a sampling
error that does not exceed plus or minus 7.7 percentage points. The actual
participation rate cannot exceed 100 percent.

gThe EITC participation rate for households with 2 children has a sampling
error that does not exceed plus or minus 8.3 percentage points. The actual
participation rate cannot exceed 100 percent.

hThe EITC participation rate for households with 3 or more children has a
sampling error that does not exceed plus or minus 8.3 percentage points.

iEstimates in the table are from the Urban Institute. Mathematica has also
estimated participation rates for these subgroups and found that about 66
percent of eligible households with children and 28 percent of those with
elderly members participated in fiscal year 2002. While the Urban
Institute and Mathematica each generate food stamp participation rate
estimates using data from the CPS, their estimates differ slightly. The
differences arise primarily from differences in imputation methodology for
data missing from the CPS-such as assets, immigration/refugee status, and
information on which household members buy and prepare food together.

jEstimates do not include individuals who are institutionalized. Estimates
do account for variation in state eligibility rules and for people who may
be eligible for only part of the year.

kEstimates include those who are dually enrolled in Medicare and Medicaid
who thus qualify for the full Medicaid benefit and those who only receive
Medicaid assistance for their Medicare cost sharing. Estimates do not
include eligible elderly people who are institutionalized. The
participation rate may be lower for this group in part because those over
the age of 65 may have Medicare coverage.

lNo reliable national survey data exist on individuals who meet the SSI
disability criteria. Therefore, researchers made assumptions about a
person's disability based on self-reported information on inability to
work due to illness or other disability, lack of work activity in the
prior year, and receipt of disability income. Some of those assumed
eligible for SSI based on disability may not meet the criteria and vice
versa. As a result, the models may overestimate or underestimate the
number of eligible people. This estimate also does not include
institutionalized individuals or disabled children.

mEstimates do not include eligible elderly people who are
institutionalized.

As shown, participation in low-income programs varies markedly across
subgroups. For example, a smaller share of elderly people eligible for low
income programs tend to participate in the Food Stamps and Medicaid
programs than among the total eligible population. The elderly may
participate in the Food Stamp Program at a lower rate than other
households because most elderly households receive Social Security and are
eligible for relatively small food stamp benefits. For example, in 2000,
44 percent of all households with elderly members eligible for food stamps
were eligible for a monthly benefit of only $10 or less, the minimum
benefit for households of one or two persons. In comparison, 12 percent of
households without elderly members eligible for food stamps receive
benefits that low. Similarly, Medicaid participation among the elderly may
be low because Medicare also covers many elderly people who are eligible
for Medicaid. In contrast, families with children tend to participate in
food stamps, Medicaid, and the EITC at higher than average rates than
those

households without children.8 Although we were unable to disaggregate
participation rates by geographic area, program participation rates can
also vary by state or locality.

Information on trends in program participation rates over time can also
help in interpreting participation rate estimates, but we were only able
to provide information on trends in participation for one program.
Mathematica Policy Research, Inc., estimates food stamp participation
rates annually under contract with USDA and has taken steps to ensure that
estimates are comparable over time. As shown in figure 1, food stamp
participation rates remained fairly stable between fiscal years 1999 and
2002, declining slightly from 52 percent in fiscal year 1999 to 48 percent
in fiscal year 2002 among eligible households and 56 percent in fiscal
1999 to 54 percent in fiscal year 2002 among eligible individuals.

8We estimated a food stamp participation rate for another subgroup,
eligible individuals in working families, in prior work and found that 52
percent of these individuals were enrolled in 2001. See GAO, Food Stamp
Program: Steps Have Been Taken to Increase Participation of Working
Families, but Better Tracking of Efforts Is Needed, GAO-04-346
(Washington, D.C.: Mar. 5, 2004).

Figure 1: Food Stamp Participation Rate Estimates for Fiscal Years 1999 to
2002

Percent 60

0 1999 2000 2001 2002

Fiscal year

Participation rate for households

Participation rate for individuals

Source: Mathematica Policy Research, Inc., Trends in Food Stamp Program
Participation Rates: 1999 to 2002. (Washington, D.C.: Sept. 2004)

We were not able to provide information on trends in participation rates
for the Medicaid and SSI programs because the most recent participation
rate data available for these programs is not comparable to prior year
estimates. While prior year estimates are available, because of changes
from 1 year to the next in the methodology used to estimate participation
rates, estimates for prior years are not perfectly comparable with the
most recent estimates and are therefore not shown in this report. The
participation rate estimate for the EITC was available for only 1 year.

For Non-Entitlement Programs, the Proportion of Those Eligible Who Were
Enrolled Ranged from Less Than 10 Percent to about 50 Percent

Among the seven non-entitlement programs we reviewed, the share of those
eligible who participate ranges from less than 10 percent to about 50
percent. These programs are generally not funded to serve all eligible
applicants, and many have other design features, such as the
prioritization of certain subgroups and eligibility criteria specific to a
state or locality, that would not necessarily allow for all eligible
people who apply to receive benefits.9 Consequently, we refer to these
estimates as coverage rates rather than participation rates. Table 6
provides estimated coverage rates for each of these programs, expressed in
ranges to reflect potential errors in survey data. The actual coverage
rates may be within a broader range of estimates than shown because other
errors that may have resulted from calculating these estimates could not
be quantified. Estimates for the CCDF, SCHIP, and TANF cash assistance
programs were developed by the Urban Institute under contract with HHS. We
also include a WIC estimate developed by the National Research Council,
and another WIC estimate developed by the Urban Institute under contract
with us. We estimated the Head Start and housing programs using
administrative and national survey data.

9In recent years WIC has been funded at levels that allow the program to
serve all or nearly all applicants who have sought services and meet
program eligibility criteria. However, WIC is a non-entitlement program
whose funding is determined annually through the congressional
appropriations process. In addition, because the SCHIP and TANF programs
provide grants directly to states and allow states substantial flexibility
in determining program eligibility and implementation, funding constraints
contribute to, but do not determine coverage rates to the extent that they
do for the remaining programs.

Table 6: Estimated Coverage Rates for Non-Entitlement Programs for the
Most Recent Year Data Were Available

Coverage rate
estimates
Program (year) Eligibility unit (percent)a
CCDF (2001) bChildren 18-19c
Head Start (2003) dChildren 44-54e
Housing Choice Vouchers (1999) fHouseholds 13-15
Public Housing (1999) Householdsg 7-9
SCHIP (2000)h iChildren 44-51
TANF - cash assistance (2001) Familiesj 46-50
WIC (1998): National Research Individuals 51l
Council
(2001): Urban Institute kInfants and Children 51-55m

Source: The Urban Institute's TRIM3 and GAO's analysis of data from the
CPS and HUD.

bThe estimates account for variations in state eligibility criteria and
are based on criteria as of October 2001. Estimates do not account for
participation in other programs that provide similar services.

cTo increase the precision of these estimates, 3 years of CPS data were
used to estimate the eligible population; other coverage rates and
participation rates estimated using TRIM3 only use 1 year of data. As a
result, the confidence interval based on the sampling error is smaller
than for the other estimates.

dHead Start participation is measured by the number of funded slots, not
the number of children in the program. Therefore, this estimate may
overstate or understate the actual coverage rate because more than one
child may fill a slot in a year and some slots may go unfilled. Estimates
do not account for participation in other programs that provide similar
services.

eThe coverage rate does not include enrolled children living above the
poverty threshold. Nonpoor children can comprise up to 10 percent of Head
Start enrollment slots.

fHUD's data on the number of eligible households was based on the American
Housing Survey. This survey may overestimate household income and
underestimate the number of households in poverty. These estimates also
may include households who do not qualify for the voucher despite their
lowincome status. Coverage rates are based on the number of households who
successfully leased units in 1999. A total of 1,649,645 vouchers were
authorized, but some of these were not used partially due to
voucher-holders' inability to find housing. Rates reflect only those
participating in the Housing Choice Voucher program; however, those
eligible may also be served by a number of other federal, state, and local
housing assistance programs.

gHUD's data on the number of eligible households was based on the American
Housing Survey. This survey may overestimate household income and
underestimate the number of households in poverty. Income-eligible
households may not be eligible for other reasons-for example, local public
housing authorities may deny assistance to people with habits and
practices that may be detrimental to other public housing tenants-and the
estimate does not account for this. Coverage rates are based on the number
of public housing units that were leased; a total of 1,235,229 units were
available but some were unleased for a variety of reasons, such as
resident turnover. Rates reflect only those participating in the public
housing program; however, those eligible may also be served by a number of
other federal, state, and local housing assistance programs. GAO used an
eligible population of those who earn less than 80 percent area median
income as per the eligibility criteria; however, nationally the vast
majority of those served by public housing make less than 50 percent of
the area median income.

hA new program established in 1997, SCHIP saw its enrollment increase 75
percent between fiscal years 2000 and 2003; a more recent estimate of the
SCHIP coverage rate would likely be higher than the range shown. In
addition, because states' implementation of their programs varied,
awareness of SCHIP may lag in states that created their programs more
recently.

iEstimates account for variation in state eligibility rules, and the
estimates are based on state eligibility rules in place in 2000. Children
who are eligible for Medicaid expansion SCHIP programs and separate state
SCHIP programs are included in the estimate. Children covered by private
and other public health insurance such as Medicaid are generally not
eligible for SCHIP and are, therefore, not included in the denominator for
the calculation; to the extent that some are included in the actual
caseload under "employer buy in" programs, the coverage rate will be
slightly overestimated.

jEstimates account for variation in state eligibility rules. All units
receiving TANF cash assistance are included, even though a separate state
program might fund their benefits. Families receiving noncash benefits
funded by TANF are not included, although an increasing proportion of TANF
funds are being used to provide non-cash assistance to families. Some
families or individuals that may appear eligible according to model may
not be participating because they have not complied with program
requirements, such as being involved in work activities. However, those
who have reached their state or federal TANF time limit are appropriately
excluded from eligibility in estimating the coverage rate.

kEstimates only include eligible infants and children, even though
pregnant, postpartum, and breastfeeding women are eligible for this
program, because the CPS does not have data to determine a woman's
pregnancy status or whether she is breastfeeding. According to the FNS,
women account for a quarter of WIC participants.

lA confidence interval is not available. This estimate is based on data
from National Research Council of the National Academies, Estimating
Eligibility and Participation for the WIC Program (Washington, D.C.: Sept.
2003). Estimates include pregnant women, infants, and children. It does
not include postpartum and breastfeeding women.

mEstimates are preliminary because they are partially based on the numbers
used to create Medicaid participation rates and SCHIP coverage rates. The
Medicaid participation rate and SCHIP coverage rate for 2001 had not been
completed in time to include in this report.

Generally less than half of those eligible participate in the seven
nonentitlement programs. Coverage rates also have limitations such as data
sources, and estimation methodologies that make it inappropriate to
compare the estimates across programs. However, they provide a general
sense of the extent to which these programs are reaching eligible people.
Specifically, we found:

o Almost 20 percent of children who meet state-defined eligibility
criteria

are receiving child care services through CCDF.10 This estimate does not
reflect participation in child care funded by other federal sources. HHS
estimates that about 26 percent of CCDF-eligible children are receiving
child care services through either CCDF (including the Child Care and
Development Block Grant, state CCDF funds or TANF

10Under CCDF, states may provide child care subsidies to families with up
to 85 percent of state median income, if they are working or involved in
education or training. States determine the income criteria up to this
maximum. See GAO, Child Care: Recent State Policy Changes Affecting the
Availability of Assistance for Low-Income Families,

GAO-03-588 (Washington, D.C.: May 5, 2003).

transfers to CCDF), or child care services funded directly through the

Social Services Block Grant (SSBG), TANF, or TANF state funds.11 o The
Head Start program funded enough slots to serve about 44 to 54 percent of
eligible 3-to 4-year-old low-income children in 2003.12

o An estimated 13 to 15 percent of households eligible for Housing
Choice Vouchers on the basis of income both received a voucher and were
able to successfully lease a housing unit in 1999, the most recent year
for which these data were available. In addition, less than 10 percent of
households eligible on the basis of income were served through the Public
Housing program in 1999. While HCV and public housing are the two largest
federal housing programs, there are other federal, state, and local
housing programs from which eligible households could receive assistance.
HUD estimated that in 1999 about a quarter of all households eligible for
any kind of housing assistance received assistance.

o In 2000, about 44 to 51 percent of eligible children participated in
SCHIP. Since 2000, SCHIP enrollment has increased by nearly 3 million
children, but the impact this has had on the coverage rate is not known
because information on how the eligible population may have changed over
this time period is not available. In addition, because states'
implementation of their programs varied, awareness of SCHIP may lag in
states that created their programs more recently.

o About half of all eligible households receive cash assistance through
TANF. This estimate does not account for families that receive other
services, such as transportation and child care, that are offered through
this program, but who do not receive cash assistance.13

o WIC participation rates were available from two sources. The National
Research Council using the Survey of Income and Program Participation
found that 51 percent of eligible infants, children ages 1 to 4 and
pregnant women were enrolled in WIC in 1998. The Urban

11HHS does not collect information on the number of children who receive
child care subsidies through the SSBG, TANF and TANF Maintenance of Effort
funding. Therefore, the agency estimated the number of children served
through those other funding sources by dividing the total amount spent on
child care services from each source by the average cost of serving a
child under CCDF.

12Even so, in some areas, Head Start programs are underenrolled. See GAO,
Head Start: Better Data and Processes Needed to Monitor Underenrollment,
GAO-04-17 (Washington, D.C.: Dec. 4, 2003).

13For more information on those who receive TANF-funded services other
than cash assistance, see GAO, Welfare Reform: States Provide TANF-Funded
Work Support Services to Many Low-Income Families Who Do Not Receive Cash
Assistance,

GAO-02-615T (Washington, D.C.: Apr. 10, 2002).

Institute, using the Current Population Survey, found that about 51 to 55
percent of eligible infants and children participated in 2001.

Coverage rates for these programs can also vary substantially by subgroup,
and they do vary within the TANF and WIC programs, the only
non-entitlement programs for which we have this information. Understanding
the extent to which different subgroups are covered by non-entitlement
programs can be particularly important precisely because these programs
are not necessarily funded to cover all those eligible. Table 7 provides
coverage rates for groups of people eligible for TANF cash assistance and
WIC.

Table 7: Coverage Rate Estimates for TANF Cash Assistance and WIC
Subgroups for the Most Recent Year Data Were Available

Coverage rate
estimates
Program (year) Eligibility units (percent)
TANF - cash assistance (2001)a All Families 46-50
bFamilies with Earners 37-44

Families with No Earners 51-56

Two-Parent Families 31-36

Families with Immigrants 34-43

WIC (2001)d All Individuals 51-55 Infants 79-93

Children Ages 1-4 41-45

Source: GAO analysis of estimates from the Urban Institute's TRIM3.

aEstimates account for variation in state eligibility rules. All units
receiving TANF cash assistance are included, even though a separate state
program might fund their benefits. Some families or individuals that may
appear eligible according to the model may not be participating because
they have not complied with program requirements, such as being in
involved in work activities. However, those who have reached their state
or federal TANF time limit are ineligible and are not included, unless
receiving assistance through a separate state program.

b,cEstimates only include families with one parent or no parents present.

dEstimates are preliminary because they are partially based on the numbers
used to create Medicaid participation rates and SCHIP coverage rates. The
Medicaid participation rate and SCHIP coverage rate for 2001 had not been
finalized during the time frames of this report.

eEstimates only include eligible infants and children, even though
pregnant, postpartum, and breastfeeding women are eligible for this
program, because the CPS does not have data to determine a woman's
pregnancy status or whether she is breastfeeding. According to the FNS,
women account for a quarter of WIC participants.

As shown, families with no income earners tend to receive TANF cash
assistance at a higher rate in the TANF program than families with earners

or two-parent families. The data also suggest that families with
immigrants have a lower coverage rate than families without immigrants. As
noted earlier, some families within these subpopulations may be receiving
other services funded by TANF, although data are not available on this.
The participation rate among infants eligible for WIC is higher than among
children ages 1 to 4.

We were unable to provide estimates of changes in coverage rates over time
for most of the non-entitlement programs covered in this review, but we
did find an increase in the coverage rate for the Head Start program over
the past several years. Since 1997, Head Start coverage rates have
increased from about 40 percent to 50 percent. However during this time
period, the coverage rate increased between 2000 and 2001 from about 50
percent to 58 percent and then gradually declined. The coverage rate
increased by this amount within 1 year because slots increased by nearly
40,000 for the Head Start program while the number of children eligible
decreased by about nearly 130,000. Since 2001 the number of funded Head
Start slots has increased, but the coverage rate fell because the child
poverty rate, and thus the number of eligible children, increased. Figure
2 shows Head Start coverage rate estimates for 1997 to 2003.

Figure 2: Head Start Coverage Rate Estimates for 1997 to 2003

Percent
60

0
1997 1998 1999 2000 2001 2002 2003

Year

Source: GAO's analysis of CPS and HHS data.

Because of changes in the methodology used to estimate coverage rates from
1 year to the next, we cannot reliably compare the most recent coverage
rate estimates for SCHIP and TANF to those of prior years, but there is
compelling evidence that coverage rates for these two programs changed
significantly over time. SCHIP was first implemented in 1997 and the
number of children in the program grew from 660,351 in fiscal year 1998 to
over 3 million in fiscal year 2000. As states continued to reach out to
eligible children in subsequent years, participation continued to increase
to nearly 6 million in fiscal year 2003. This eightfold increase in the
number of children enrolled in SCHIP since fiscal year 1998 almost
certainly had a significant impact on the coverage rate. Meanwhile, as
states implemented welfare reforms during the strong economy of the late
1990s and other changes occurred in programs serving low-income families,
the number of families receiving TANF cash assistance declined, falling by
about 50 percent between fiscal years 1997 and 2003.14 This also likely
had a significant impact on the program coverage rate. HHS has reported
that about 70 percent of families eligible for TANF cash assistance were
enrolled in the program in 1997, compared to less than 50 percent in 2001,
but we were unable to quantify how much of this change was caused by
changes in estimation methodology.15 It is also important to note that
states' TANF programs have changed over this time. States have more
flexibility in the types of non-cash assistance they may provide families,
and some families eligible for cash assistance may be receiving other
forms of aid instead. In addition, some families or individuals that may
appear eligible according to the model may not be participating because
they have not complied with program requirements, such as being involved
in work activities.

14For more information on recent program changes see GAO, Supports for
Low-Income Families: States Serve a Broad Range of Families Through a
Complex and Changing System, GAO-04-256 (Washington, D.C.: Jan 26, 2004).

15While the exact effect of the methodological changes has not been
quantified, Urban Institute staff and HHS do not believe that the
methodological changes are driving measured changes in participation
rates. See U.S. Department of Health and Human Services, Indicators of
Welfare Dependence: Annual Report to Congress, 2004, (Washington,
D.C.:2004).

The Estimated Costs of Serving Eligible Nonparticipants Show That in Some
Programs, Those Eligible for the Largest Benefits Are More Likely to be
Enrolled

Our estimates of the costs of providing benefits to eligible
non-participants show that for some programs, those currently
participating in the programs are generally eligible for a greater benefit
amount than those not participating. For these programs, the estimated
costs of serving those eligible but not currently receiving benefits may
be less than one might expect based on the participation or coverage rates
we estimated. This is not surprising, as it makes sense that, for example,
a person eligible for a $10 food stamp benefit might be more likely to
forgo that benefit than a person eligible for a $150 benefit. More
specifically, Mathematica found that the 48 percent of households
participating in the Food Stamp Program in fiscal year 2001 received about
62 percent of the total amount of benefits that would be paid out if all
eligible households received benefits.

As shown in table 8, we were able to estimate the additional cost of
providing benefits to eligible nonparticipants for 6 of the 12 programs we
reviewed. These cost estimates take into consideration the characteristics
of non-participants and the benefits for which they would be eligible but
do not account for increases in administrative costs that could result
from participation increases. We were unable to estimate the cost of
providing benefits to eligible, non-participants in the six other
programs-CCDF, HCV, Medicaid, Pell Grants, Public Housing, and
SCHIP-because we did not have enough information to determine how the
characteristics of the non-participants would have affected the benefit
amounts they could have received.

Table 8: Potential Cost of Providing Benefits to Eligible Non-participants
for the Most Recent Year Data Were Available

Potential annual cost
of serving eligible
Program (year) non-participantsa,b

EITC (1999) $2-3.4c

Food Stamp Program (2001) 8.8-11.5

Head Start (2003) 3.8-5.6d

SSI (2001) 8-9.8

TANF-cash assistance (2001)e 8-9

WIC (2001)f 1.9-2.1

Source: The Urban Institute's TRIM3 and GAO's analysis of data from CPS,
HHS, and IRS.

Note: Cost estimates do not include increases in administrative costs
associated with greater program participation.

aThis range represents a 95 percent confidence interval based on estimated
sampling error in the national survey data. This interval does not take
into account errors introduced by the modeling process or by
administrative data. We assessed but did not quantify these errors.
Estimates are based on most of the same data used to estimate
participation and coverage rate estimates for the related programs and,
therefore, many of the same data and model limitations apply.

bEstimates do not account for policy or economic changes that would occur
if participation in these programs increased. All estimates are based on
the assumption that participation in other programs remains constant.

cEstimates are from GAO, Earned Income Tax Credit Participation,
GAO-02-290R (Washington, D.C.: Dec. 14, 2001).

dEstimate is based on an average cost per slot multiplied by the number of
non-participating eligible children. This estimate does not account for
local variation in the cost of services, and does not account for the
possibility of more than one child filling a slot.

eEstimates are only for the additional cost of providing TANF cash
assistance. The program provides other supports, such as child care and
transportation, to needy families, and the cost of these services would
increase the estimate if they were included.

fEstimates are only for the additional cost of providing food benefits to
infants and children who are not enrolled.

Although Medicaid is by far the largest program we reviewed, developing
reliable estimates of the costs of serving those eligible but not enrolled
in the Medicaid program is difficult for a number of reasons. These
reasons include a lack of information about the health status of eligible
nonparticipants, the many variables that affect health care costs, and the
open-ended nature of health care benefits.

The costs included in the table do not reflect the total costs that would
be incurred if all eligible people enrolled in these programs. Probable
increases in administrative costs that would be incurred are not included,
and we did not estimate how the cost of goods and services offered through
the Food Stamp, Head Start, and WIC programs would change if more
participants were demanding them. However, the costs do include some
additional costs that would be incurred by states in programs where states
are required to cover some of the program costs.

Recognizing the differences in goals, design, administration, and funding
in the 12 programs, it may be neither feasible nor desirable to provide
program benefits to all those eligible. Because alternative programs could
offer the similar benefits and services as the programs we reviewed, some
of these programs, especially those with limited funding, may be best
administered by only serving a certain subgroup of their eligible
population. For example, in localities where school systems offer
prekindergarten classes to 4-year-old children, Head Start programs may
want to focus more on enrolling 3-year-olds from low-income families.
Also, limited administrative resources may prevent federal, state, and
local

agencies from serving all those eligible. Full participation among
eligible people also may not be feasible because some choose not to enroll
in means-tested programs. Given these constraints, some programs may focus
on targeting their resources to those most in need of program services.

Finally, while participation and coverage rate estimates as well as
estimates of potential program costs provide important program information
that can be useful to both policymakers and program administrators, these
estimates must be interpreted carefully to take into consideration such
factors as survey data, research methodology, timeliness of estimates, and
the availability of alternative programs. These factors limit our ability
to compare rates across programs and would need to be considered in
efforts to define appropriate or desired participation or coverage rate
levels. For a detailed discussion of the factors that affect use of these
estimates, see appendix II.

Many factors influence access to means-tested programs, including the
benefits provided by the program, ease of access, misperceptions about
program requirements, and eligibility verification procedures put in place
to ensure program integrity, based on our literature review and agency
interviews. These factors influence participation by affecting the number
of people who participate in a program as well as the type of people who
participate. Entitlement programs may focus on addressing factors that
affect access in order to increase program participation. However, because
coverage rates in non-entitlement programs are determined primarily by
funding levels, non-entitlement programs are more likely to be concerned
about the impact of these factors on other aspects of program access, such
as the composition of the program caseload and how the program interacts
with alternative programs that provide similar benefits.

Many Factors Influence Participation but Their Impact on Entitlement and
Non-Entitlement Programs Differs

The Size and Type of Benefits Can Affect Participation

The size and type of program benefits can affect whether or not an
individual participates in a program. Numerous studies show that the size
or value of the benefit influences program participation, and this depends
largely on the structure of program benefits. Some programs, such as the
Food Stamp Program, determine benefit levels based on income, allowing a
three-person household to receive as little as $1 per month depending on
household income. Other programs, such as Medicaid, generally offer
comparable benefits to every resident of a state who meets eligibility
requirements. A recent National Bureau of Economic Research study

showed that across a range of programs and within programs, larger
benefits were associated with higher participation and coverage rates.16
In addition, our literature review, site visits, and analysis of
participation rates for selected programs showed that participation and
coverage rates tend to be higher among subgroups eligible for larger
benefits. For example:

o Elderly individuals are generally eligible for small food stamp
benefits; their participation rate in the program is lower than for other
groups.

o Families with infants participate in WIC at higher rates than families
with older children, in part because the value of the WIC voucher is
greater for families with infants. WIC officials in Connecticut reported
to us that 10 percent of families stopped participating in the program
when their youngest child reached age 1.17

o On average, families with children who are eligible for the EITC
receive a higher EITC benefit than individuals and families without
children, and they participate at a higher rate.

The type of benefits offered can also influence whether someone eligible
for a program participates. Some programs-such as EITC and SSI-offer cash
benefits while other programs-such as Head Start and Medicaid- offer
direct services, such as educational or health services. Our research on
participation in low-income programs generally shows that participation
and coverage rates are higher among programs that provide cash benefits
than among programs that provide direct services. One reason for this
difference is the flexibility cash benefits give to individuals. Although
the recipient may need the services provided by programs in which they are
enrolled, these types of benefits do not allow individuals to shift
resources away from program purposes toward what they might consider more
pressing needs. Our site visits confirmed that the type of benefits
influence participation and coverage rates. For example, we found that:

16Currie, Janet, The Take Up of Social Benefits, NBER Working Paper 10488,
(Cambridge Mass.: National Bureau of Economic Research, May 2004). This
study reviews literature on participation and coverage rates among social
programs in both the United States and the United Kingdom.

17This decline in participation is consistent with national data provided
by USDA. Nationally, there is about a 30 percent decline in enrollment
when the value of a child's benefit decreases. The average benefit
declines by about 60 percent after the child's first birthday.

o SSI recipients who participated in a pilot project that allowed them
to receive their food stamp allotment in cash expressed a strong
preference for this over traditional food stamp coupons.

o Some individuals who need subsidized housing may prefer the Housing
Choice Voucher program, which allows individuals to choose their housing
in neighborhoods that offer better educational and employment
opportunities or choose to remain in a place while paying less rent, over
the Public Housing program that does not give the individuals flexibility
to choose where they live.

Ease of Access to Program Benefits and Services Can Influence
Participation

Factors that influence the ease with which potential participants can
access a program-including office hours, program location, and the ability
of program participants to redeem or use their benefits-can also affect
the number and groups of people who participate in the programs. Several
of the programs we reviewed require applicants to visit the program office
to establish and maintain eligibility. For instance, local WIC, TANF, and
Food Stamp Program offices typically require face-to-face interviews
before individuals can receive benefits.18 Those programs that keep
traditional office hours-8:00 a.m. to 5:00 p.m.-pose a barrier to
potential applicants who work and would have to take time away from their
job in order to apply. Studies on child care and WIC programs identify
traditional office hours as a barrier for working families. Similarly,
Medicaid officials in California told us that one of their barriers to
participation is office hours that do not accommodate working families.
Although many program officials suggested that having flexible office
hours is important to participants, some offices are challenged to extend
their operating hours. WIC officials in Connecticut told us that they have
attempted to promote extended hours and have made some progress; however,
instituting extended hours remains problematic due to limited funding and
contracts governing the workforce.

The location of the program office can also affect participation and
caseload composition. For many individuals-like the elderly, individuals
with disabilities, and families living in rural areas-traveling to a
program office that is outside the limits of their available mode of
transportation makes it difficult for them to receive needed benefits. The
USDA published in its National Survey of WIC Participants that one of the
top barriers in the WIC program is transportation with 30 percent of
participants reporting that they missed appointments because they lack
transportation

18Most programs waive the requirement for a face-to-face interview in
hardship cases.

to the office. Studies on the Housing Choice Voucher, child care, Head
Start, and Food Stamp programs reported similar findings. Many state and
local officials we visited agreed that transportation is a barrier to
participation. A TANF, Food Stamp, and Medicaid official in Washington
County, Maryland, told us that public transportation is limited and
getting to and from the program offices poses a barrier to participation
for many individuals, especially those in rural communities. According to
officials, participants without vehicles have to seek rides from family
members or acquaintances to access services. For many of these
participants being seen going to the welfare office is difficult and
sometimes stigmatizing.

Some programs-including child care, Housing Choice Voucher, and
Medicaid-offer benefits that can be difficult for participants to use, as
some service providers-landlords, day care workers, or health care
providers-will not exchange their services for program benefits. For
example, our prior work found that the proportion of providers who will
accept child care subsidies varied widely by state, ranging from 23 to 90
percent. Even in cases where providers accepted subsidies, the number

of slots for children that used child care subsidies was limited.
Additionally, several state and local officials suggested that
participants are sometimes challenged to find health care providers that
accept Medicaid. According to officials in Chisago County, Minnesota,
participants have to drive up to 200 miles to visit a dentist that accepts
Medicaid.

Misperceptions about Program Requirements Discourage Some Individuals from
Participating

From our literature review and interviews with program officials, we also
found that many individuals do not participate in low-income programs
because they do not know that they are potentially eligible for benefits.
Numerous studies have documented instances where participation was
compromised because individuals were unaware of program benefits or had
misconceptions about eligibility. For example, several EITC, Medicaid,
Food Stamp, Head Start, and SSI studies indicate that one of the primary
barriers to participation is that individuals do not know that they are
eligible for these benefits. For some individuals-like the elderly and
non-English speakers-this unfamiliarity with program benefits is even more
widespread, creating a larger barrier to participation and an under
representation of these individuals in the caseload. Many program
officials

19GAO, Child Care: States Exercise Flexibility in Setting Reimbursement
Rates and Providing Access for Low-Income Children, GAO-01-894 (Washington
D.C.: Sept. 2002).

we visited agreed and explained that individuals also have misconceptions
about program eligibility. For example, some individuals do not believe
that they are eligible for benefits because they are employed while others
do not want to be attached to the perceived stigma associated with the
programs. In Los Angeles, Medicaid officials noted that rumors circulated
about the eligibility criteria for Medicaid prevent many potential
recipients from applying for the program. Particularly in non-English
speaking communities, information about social services is often received
through word of mouth; thus, when incorrect information circulates, it can
have a significant impact on participation.

Application Process and Eligibility Verification Requirements to Ensure
Program Integrity Can Have an Effect on Participation

Eligibility verification requirements-rules put in place to improve
program integrity by ensuring that only those eligible for program
benefits receive them-can also affect the overall number and
characteristics of people who participate in the program. Each low-income
program in this study has a set of verification requirements that
individuals must complete to establish and maintain eligibility. Common
requirements among the programs include completing application forms and
providing necessary documentation. In addition, some programs require
applicants to take additional steps, such as attending face-to-face
interviews, documenting parental or spousal assets, or participating in
orientation classes. According to our literature review, the complexity of
verification requirements can impact the number of enrollees, especially
for individuals with mental disabilities or who are homeless. One WIC
study reviewed participation rate differences among states and found that
states that required applicants to provide proof of income (before it was
federally mandated) and had stricter program rules had lower program
participation than states that did not require income documentation.20
Additionally, another study examined EITC participation rates across
states and found that differences in participation rates are due in part
to the applicant having help completing the complex tax forms.21 Many
state and local officials agree that strict and/or complicated
verification requirements can decrease participation, particularly among
certain groups of people. For example, we found:

20M. Bitler, J. Currie, and J. Scholz, WIC Eligibility and Participation,
Institute for Research on Poverty Discussion Paper 1255-02 (Madison, WI:
Institute for Research on Poverty, June 2002, revised March 2003).

21SB/SE Research, Participation in the Earned Income Tax Credit Program
for Tax Year 1996 (Greensboro, N.C.: January 2002).

o According to officials, the Pell Grant application is very long and
complicated and can be difficult for students and their parents to
complete. Staff from an organization in St. Paul, Minnesota, dedicated to
assisting low-income students to access higher education told us that many
low-income students in the area would not be able to complete the forms
correctly without their assistance.

o CCDF Officials in Connecticut and Georgia told us that the application
process can be complex and difficult for applicants and, as a result, many
applicants submit their applications late or incomplete. Because there is
very high demand for child care subsidies, these families are often denied
benefits or removed from the waiting list.

o Local SSI officials told us that participants face many life changes
such as frequent hospitalization and institutionalization, moving into
different family households, and changes in the hours that they work. To
remain eligible for program benefits, SSI recipients must report these
types of changes within 10 days from the end of the month of the change.

Finally, while each program has its own set of verification requirements,
these requirements become even more challenging when participants are
applying or maintaining eligibility for multiple programs. Program
officials in Maryland told us that one of the biggest barriers to
accessing multiple means-tested programs is the many different eligibility
criteria and reporting requirements of the various programs. According to
state officials in Maryland, the TANF and Food Stamp Programs are
relatively well aligned, but Medicaid has not coordinated as much with the
other programs. According to officials, this can be problematic because a
participant who is turned down for one program might assume that he or she
is ineligible for other low-income programs and may fail to apply for
benefits for which they are eligible.

Factors That Influence Participation Rates Can also Impact Other Aspects
of Program Access Such As Targeting and Program Interactions

Factors that influence participation also impact other aspects of program
access such as targeting and program interactions among all low-income
programs, but particularly among non-entitlement programs whose enrollment
levels are determined largely by funding levels. The nonentitlement
programs covered in this review-CCDF, Head Start, Housing Choice Vouchers,
Public Housing, SCHIP, and TANF-have funding limitations that could
restrict the total number of people who could potentially participate in
the programs and, as a result, coverage rates for these programs may be
determined, at least in part, by funding. However, among programs that
allow states substantial flexibility in determining program eligibility
and implementation such as CCDF, SCHIP, and TANF,

funding constraints contribute to, but do not determine coverage rates to
the extent that they do for programs that allow less state and local
flexibility. Given funding constraints for non-entitlement programs,
agencies are challenged to allocate scarce resources appropriately by
targeting their benefits to specific groups of people and planning
strategically to ensure that the program complements a broader range of
programs administered by the same agency or that provide similar services
or benefits to the program's eligible population.

Because factors that influence participation often affect the composition
of a program's caseload, some agencies that administer programs with
limited funding have put measures in place to ensure that their caseloads
reflect program priorities-whether that involves targeting certain groups
or ensuring that benefits do not disproportionately favor or exclude
certain subpopulations. For example, although the public housing program
generally does not target benefits to households with the lowest incomes,
it allows priority to be given to elderly and disabled recipients. Agency
officials reported that they track participation of these groups to ensure
that their share of the total caseload does not change dramatically from 1
year to the next. Similarly, state child care officials told us that funds
are generally provided to recipients based on priority groups that favor
families receiving public assistance and those with the greatest financial
need.

Additionally, many agencies are cognizant of how their programs interact
with and complement other programs that serve the same population. For
example, some child care programs pair with half-day Head Start programs
to provide continuous care to children of working parents; Medicaid and
SCHIP programs within some states coordinate their enrollment processes to
facilitate access to health insurance for individuals in the state; and
between 1998 and 2003, HUD responded to a decrease in the number of public
housing units available for low-income households by increasing resources
available through its voucher program to maintain coverage for families.
Likewise, in allocating TANF resources, programs often coordinate with
local workforce development agencies, child care programs, and other
programs that provide similar services to best meet the needs of their
clients without duplicating an existing effort. In these ways, programs
acknowledge and respond to access issues and ways in which factors affect
access other than participation levels and rates.

Program Administrators Have Strategies That Improve Both Access and Integrity,
but Federal Agencies Have Generally Focused More on Measuring Program
Integrity Outcomes

Program administrators have implemented many strategies to achieve desired
program outcomes, including those related to program access and program
integrity, but while agencies have generally taken steps to monitor and
disseminate information on program integrity, few track and report on the
extent to which they are serving their eligible populations. The programs
we reviewed have many goals and objectives of varying degrees of
importance to the agencies that administer them, among them program access
and program integrity. Strategies implemented at the federal, state, and
local level to address these two issues include information systems, data
sharing, and technological innovation, changes to the application and
eligibility verification process, and outreach and coordination with other
programs. Although federal agencies do not have direct control over many
of the strategies we identified-including those mandated by law and those
initiated at the state and local level-federal managers can play a role in
encouraging and facilitating such strategies by emphasizing program access
and program integrity as federal agency priorities. In response to the
Improper Payments Information Act of 2002, all of the federal agencies we
reviewed have taken at least some steps to identify and begin reporting
information to Congress and others on the extent of improper payments. We
also found, however, that for several of the programs we reviewed, federal
agencies have not taken steps to identify and use participation or
coverage rate information in managing their programs. Having up-to-date
information on the extent to which their programs reach eligible
individuals can help agencies plan strategically, set priorities, and
ensure that agencies are reaching eligible families.

Programs Have Found Ways to Improve Access by Reducing the Burden on
Applicants While at the Same Time Improving Program Integrity

Information Systems, Data Sharing, and Technological Innovation

The state, local and federal program officials we spoke with identified
several strategies that they believe may improve either of the two
objectives-program access or program integrity-without harming, and in
many cases improving, the other. As mentioned, some efforts to increase
program integrity and prevent fraud, such as increasing the amount of
eligibility verification required, may actually hurt access to the program
by deterring even those eligible from applying. However, program
administrators told us of several strategies that increase access while
maintaining and even improving integrity. The complementary strategies we
identified are enabled by information systems, data sharing, and
technological innovations, changes in the application and eligibility
verification process, and outreach and coordination with other programs.

Improved information systems, sharing of data between programs, and use of
new technologies can help programs to better verify eligibility and make
the application process more efficient and less error prone. These
strategies can improve integrity not only by preventing outright abuse of

programs, but also by reducing chances for client or caseworker error or
misunderstanding. They can also help programs reach out to populations who
may face barriers. One strategy involves sharing verified eligibility
information about applicants across programs. Data sharing prevents
applicants from having to submit identical verification to multiple
programs for which they may be eligible, and it can also speed up the
sometimes-lengthy application process. In addition, data sharing allows
programs to check the veracity of information they receive from applicants
with other databases.

In Minnesota, WIC caseworkers are able at the time of application to
determine participant eligibility by accessing basic eligibility
information via technology systems from other means-tested programs, such
as Medicaid, food stamps and TANF. SSI administrators told us that during
the application interview, income data is entered in their computer
system. This information is then checked with a number of other databases
including those of the IRS, the Department of Veterans Affairs, and the
state Civil Service Administration to quickly verify earnings and other
eligibility requirements and inconsistencies are flagged. Similarly,
Georgia has an interactive computer system that aggregates client
information for TANF, Medicaid, and the Food Stamp Program and calculates
benefits for all three programs. Additionally, when clients need to make a
change in their eligibility information, these data are automatically
changed for all three programs, so that clients can receive the
appropriate level of benefits based on their most current eligibility
information. Furthermore, the data reliability matches may include
information from the Department of Labor wage match, SSI benefits, prison
information, and information on a person's death. If any of the matched
information conflicts with the information provided by the client, a
caseworker is to ask additional questions of the client. In all these
ways, caseworkers can accelerate the application process while
capitalizing on verified eligibility information already provided to other
programs.

Some administrators told us that such data sharing, however, can be
complicated by concerns over ensuring privacy and by insufficient
technological capacity. Local program administrators told us that
eligibility data for Medicaid and SCHIP is tightly controlled because of
Health Insurance Portability and Accountability Act (HIPAA)-a law that
limits access to an individual's health information-but that its data
might be shared, if warranted, with proper consent. Data sharing can be
impeded by technological limitations, such as computer systems that cannot
communicate directly with other systems over the Internet or some other
network, or a lack of software needed to translate information into

Application and Eligibility Verification Process Changes

formats that other computers can understand. Our prior work on data
sharing identifies strategies agencies have implemented to address
concerns about privacy and technological capacity issues.22

Some agency officials told us of ways in which on-line applications can
improve both access and integrity. As stated, program integrity is
compromised not only by willful misrepresentation, but also by applicant
and caseworker error. Web-based or partially web-based applications can
automatically check that all required fields are completed before an
application is submitted for review and automatically calculate benefits
thereby reducing error. Moreover, having the option of an on-line
application can increase access to those who are eligible for benefits but
who may have difficulty physically getting to program offices. Minnesota
has instituted an on-line interactive application for Medicaid that
program administrators believe will make the application process easier
for both participants and caseworkers. In addition to making the process
easier, the system will also check verification information with a variety
of other data sources and thus reduce the likelihood of errors as well. In
this way, a potentially burdensome application process becomes simpler,
more accessible, and less error prone. Similarly, eight states accept food
stamp applications through the Internet. Of course, this strategy is less
viable or effective where access to computers is limited or where the
on-line environment cannot be adequately secured.

Program administrators told us about several changes in the application
and eligibility verification process that have improved their ability to
promote access and ensure integrity. Efforts to increase the accuracy of
eligibility and benefit determination can increase access by reducing the
number of applicants who were incorrectly denied benefits or whose benefit
determinations resulted in underpayments. For example, streamlining or
simplifying the application and providing applications in other languages
can increase access, but also promote program integrity by reducing errors
or misunderstanding in the application process. Simplifying applications
by clarifying confusing wording and removing redundant questions as well
as providing applications in an applicant's native language can reduce the
possibility of these types of errors. According to federal officials, most
states use a simplified application

22See GAO, The Challenge of Data Sharing: Results of a GAO-Sponsored
Symposium on Benefit and Loan Programs, GAO-01-67 (Washington, D.C.: Oct.
20, 2000), and GAO, Benefit and Loan Programs: Improved Data Sharing Could
Enhance Program Integrity,

GAO/HEHS-00-19 (Washington, D.C.: Sept. 13, 2000).

form for the SCHIP program to remove application barriers for families.
For example, the SCHIP program applications in both Connecticut and
California have been reduced from 17 and 27 pages respectively to only 4
pages each without sacrificing required information, according to state
officials. In many cases, the SCHIP streamlined application forms caused
corresponding changes in state Medicaid applications as well. In addition,
many of the program administrators with whom we spoke told us that their
applications and forms were available in many languages. For example, the
Minnesota Department of Revenue now has state EITC tax forms and fact
sheets directing them to the federal EITC in eleven languages.

Another way in which programs can streamline the application process is
through categorical eligibility with other programs, which can be set at
the federal or state level depending on the program. Categorical
eligibility, also known as adjunctive eligibility, allows an individual
who has been found eligible for one program to be automatically granted
eligibility for another program, typically one with less stringent
eligibility criteria. For example, if an individual is receiving SSI,
federal legislation generally requires that that person be automatically
eligible for Medicaid.23 Similarly, if a pregnant woman is receiving Food
Stamps, Medicaid, or TANF, then she automatically qualifies for WIC
benefits-and thus does not need to provide income verification once again
when applying for the program. States also have flexibility to grant
automatic eligibility for other programs. A state WIC program, for
example, may grant automatic eligibility for families receiving Free and
Reduced Price School lunches. These adjunctive or automatic eligibility
policies allow ultimately for simpler applications, which may enhance
access and reduce error. In fact, federal administrators at USDA noted
that adjunctive eligibility is one of the most important tools now used to
address program integrity and access issues in a way that cuts across
programs.

The provision of more comprehensive in-person assistance in filling out
applications can encourage a candidate eligible for benefits to complete
an otherwise lengthy, complicated, or intimidating application process.
Such personal assistance can also prevent errors that result from

23Individuals receiving SSI are assured eligibility for Medicaid in 39
states and the District of Columbia. The remaining 11 states may use
different or more restrictive standards for disability, income or assets;
thus, SSI beneficiaries in these 11 states may not have assured
eligibility for Medicaid. These 11 states are known as 209(b) states based
on section 209(b) of the Social Security Amendments of 1972.

misunderstanding of application questions or requirements or even fraud by
reminding applicants in person that fraudulent claims are punishable by
law. For example, to increase access to the program, California program
officials told us that the SCHIP program developed a training process that
certified staff from other entities, such as schools and community-based
organizations, as "application assistants." These "application assistants"
are also trained to recognize when applicants seem to be submitting false
information and are prohibited from assisting families who have previously
been found to have committed fraud. In a Maryland SSA office, SSI program
administrators told us that they had abbreviated their application form so
that most information is now collected in an interview. This process helps
improve access for those with limited literacy or those who may be
discouraged by a long paper application. Officials also noted that this
process helps program integrity because the interviews allow caseworkers
the opportunity to remind clients that fraudulent claims are punishable by
law. Similarly, EITC officials in Minnesota told us that free tax
preparation assistance could make the process less complicated for filers
and could also cut down on fraud from tax preparers. As one official
explained, many low-income families live near a high concentration of paid
tax preparers, who have incentives to fraudulently file for a high credit.
While more in depth application assistance has these benefits, it can pose
challenges to programs in ensuring the quality of the assistance offered
and it can be costly. According to California SCHIP officials, funding to
train "application assistants" for the SCHIP program in California was
ended in May 2002 due to budget constraints.

Lastly, some policies put in place to prevent fraud at the time of
application can also help with program access. Connecticut Department of
Social Services has begun a fraud prevention program, termed FRED (Fraud
Early Detection), which was instituted to address an increase in the
amount of fraud and error detected in the program. The purpose of FRED is
to identify potential fraud cases before benefits are paid out to prevent
improper payments rather than identifying fraud after the payments have
been made. Under FRED, cases that meet 1 of 12 risk criteria are to be
referred to a fraud investigator who is to visit the home within 10 days
of the referral. While the investigator is visiting the home, he/she is to
interview the client about aspects of the risk criteria, and make a
recommendation regarding eligibility. While the primary purpose of the
visit is to prevent fraud, in practice investigators have found that their
role is also to explain the eligibility criteria more clearly to
applicants and to explain what steps they need to take in order to be
eligible for program benefits. In addition, investigators are trained to
provide additional information to the family on other resources available
to them

Outreach and Coordination with Other Programs

and to assist them with referrals, thereby increasing access to an array
of related social services.

Targeted outreach to populations that might have more barriers to
accessing the program has the potential to increase access without
necessarily compromising integrity. Many program officials we spoke with
told us that they had outreach strategies to reach out to populations-such
as rural-dwellers, the elderly, or those with limited English proficiency-
that have had trouble accessing the program. Such strategies include
information sessions or pamphlets in ethnic community centers or
partnering with advocacy groups. In addition, cross agency referrals for
applicants already found to be eligible and more overall agency
coordination can improve access without sacrificing integrity.

Many agency officials with whom we spoke told us that they coordinated
with other programs to reach potential clients. In Maryland, the
Department of Social Services invites the Volunteer Income Tax Assistance
program to work out of its office, so that those low-income earners
already in contact with social services can have better access to the
EITC. Some offices in California participate in an open door policy
program in which clients can access any social service program through
referrals at any state government office they walk in to. Some program
administrators told us that program co-location can help increase access.
This happens when an individual applying for one program in an office is
referred to and immediately serviced by another program in that same
location. Some examples we found had Food Stamps located in TANF offices
and a WIC clinic collocated with the state Children's Health Insurance
Program.

Federal Agencies Have Taken Steps to Track Improper Payments, but Few
Estimate Program Participation or Coverage Rates to Monitor Program Access

All of the programs we reviewed emphasize both program access and program
integrity, and as we have described, many federal, state, and local
program administrators have implemented strategies intended to achieve
access and integrity goals. However, not all agencies responsible for
administering these programs have established management practices, in
keeping with standards of internal control, that will consistently guide
them in achieving these outcomes. Among the standards that are
particularly relevant for agencies responsible for administering
meanstested programs are:24

o establishing and maintaining a control environment, a culture of
accountability that sets a positive and supportive attitude toward
achieving management objectives;

o monitoring program performance over time; and

o recording and communicating relevant, reliable, and useful information
to management and others who need it to carry out their internal control
and operational responsibilities.

To achieve desired program outcomes and safeguard federal funds, these
standards should be part of an agency's operational and management
practices. While the agencies we reviewed have taken steps to establish
internal controls for improper payments and program integrity, not all
have established similar controls for achieving desired program access
outcomes.

There are many ways in which federal agencies can establish a culture of
accountability, one that "sets the tone" for program administrators at all
levels of government to emphasize certain priorities. One way to do this
is to monitor progress toward achieving desired program outcomes, and
another is to effectively communicate program information in a broad
sense, with information flowing down, across, and up the organization.
Depending on the program, some aspects of an achievement-oriented
environment are codified in laws, regulations, and federal policies. For
example, the Improper Payments Information Act of 2002 requires those
programs and activities susceptible to significant erroneous payments to
calculate annual improper payment estimates and sets statistical sampling
confidence and precision levels for estimating those payments.

24See GAO, Standards for Internal Control in the Federal Government,
GAO/AIMD-00-21.3.1 (Washington D.C.: Nov. 1999), and GAO, Internal Control
Management and Evaluation Tool, GAO-01-1008G (Washington, D.C.: Aug.
2001).

As shown in table 9, some of the federal agencies covered by this review
have already begun to monitor improper payments for the programs they
administer and, as directed by OMB guidance, communicate this information
by including a measure of improper payments in their annual Performance
and Accountability Reports.25 Including measures of improper payments in
agency annual performance plans and accountability reports is one way to
provide officials with important information on areas of success and areas
that need improvement and to send a signal to all working on agency
programs that the measured goal is a priority. (For information on the
amounts of improper payments reported by agencies, see app.III.) To
varying degrees, the other agencies included in this review have begun
taking steps to identify and report on improper payments as well. While
agencies are taking steps, much remains to be accomplished before a
comprehensive picture is available of the amount of improper payments for
these programs.26

25Agency focus on program integrity might also be encouraged by other
agency-specific OMB guidance, the Presidents Management Agenda as well as
the GAO's High-Risk series-all strategies that encourage or require
agencies to focus on integrity.

26For more information on the implementation of the IPIA governmentwide,
see GAO, Financial Management: Fiscal Year 2003 Performance Accountability
Reports Provide Limited Information on Governmentwide Improper Payments,
GAO-04-613T (Washington, D.C.: Apr. 15, 2004).

Table 9: Status of Agencies Identification of and Reporting on the Amounts of
Improper Payments by Program

Agency included improper If not, has the
agency reported
payment information in taking some steps
fiscal to identify
Subject to year 2004 annual and monitor the
performance amount of
Agency Program the IPIA? and accountability improper payments?
report
USDA Food Stamps Yes Yes
WIC Yes No Yes
HHS/ACF TANF Yes No Yes

CCDF Yes No Yes

Head Start Yes No Yes IRS EITC Yes Yes HHS/CMSb Medicaid Yes No Yes

SCHIP Yes No Yes

SSA SSI Yes Yes
ED Pell Grant Yes Yes
HUD Public Housing Yes Yes
Housing Choice Yes Yes
Voucher

Source: GAO survey of federal agencies and interviews with agency
officials and agency performance and accountability reports.

aACF has estimated the percentage of "over-income" families served by Head
Start in excess of the 10 percent allowed by regulation.

bCMS has a performance goal in its annual performance plan related to a
pilot project to develop payment error rates for both SCHIP and Medicaid.

While an internal control framework is emerging for addressing financial
program integrity issues, few of the agencies responsible for
administering the programs we reviewed have established a similar
framework for achieving desired program access outcomes, although they
have implemented numerous strategies to increase program participation.
The 12 programs we reviewed have numerous program goals and objectives of
varying degrees of importance to the agencies that administer them.
Because the most fundamental purpose of these programs is to serve
lowincome individuals and families, all track measures that provide some
information about program participation and access. For example:

o All of agencies responsible for administering the programs we reviewed
track the number of people participating in their programs.

o Some federal agencies-such as HUD-measure the extent to which program
participants redeem or use benefits provided to them.

o Others use other measures related to access. CMS tracks the number of
individuals who are uninsured each year.

o Many agencies track participants and participant outcomes through
program effectiveness studies.27

Through these efforts, agencies have demonstrated their concern about
program access, yet, the ways in which most of these agencies measure
program participation do not inform administrators about the extent to
which they are meeting overall need and therefore do not provide all the
information they need to monitor access. As discussed above, access
encompasses not only the number of program participants, but also the
extent to which programs reach the total population eligible for benefits
and services, how well agencies are allocating resources among targeted
subpopulations, and, from an agencywide or even governmentwide
perspective, how well specific means-tested programs complement and
interact with programs that serve similar populations or provide for
similar needs. While concerns about the allocation of scarce resources and
interactions with other programs apply to all low-income programs, they
may be of particular importance to non-entitlement programs whose
participation is limited by funding constraints. Participation or coverage
rate information can guide program administrators in setting priorities
and targeting scarce resources, even among programs that were not intended
to serve everyone eligible for program benefits.

As shown in table 10, only one of the agencies we reviewed, USDA,
regularly estimates participation rates for the programs it administers
and communicates and disseminates information on these rates in a way that
promotes a culture of accountability supportive of achieving program
access outcomes. USDA has been estimating participation rates annually for
the Food Stamp Program since 1975. The agency uses this information to get
a better understanding of how well they are reaching certain priority
subpopulations such as the working poor, single-parent families with
children, and SSI recipients; and to identify states in need of outreach
assistance. In addition, USDA has established a performance goal to
increase participation rates in the Food Stamp Program and reports food
stamp participation rates in its annual performance report.28 In these

27Studies on program effectiveness typically do not provide information on
the number or characteristics of those eligible but not participating in
low-income programs.

28USDA also provides performance bonuses to reward states for their
performance in administering the Food Stamp Program. Awards focus on
payment error rates, negative error rates, participant access rates, and
timeliness of application processing.

ways, the agency uses this information for program management purposes.

Table 10: Status of Agencies Efforts to Use Information on Participation
or Coverage Rates in Managing Their Programs
If the Does If so, participation use this does it other
agency the does or coverage estimate rate include in their Agency Program coverage measure? key USDA Yes Yes Yes
estimates agency the rate, a information rate participation as a performance rate? program Food
a agency information or performance report or report?a Stamps

WIC Yes No bYes
HHS/ACF TANF Yes No Noc
CCDF Yes No dYes
Head Start Yes No No
IRS EITC Yese No Nof
HHS/ CMS Medicaid No
S-CHIP No
SSA SSI Nog
ED Pell Grant Noh
HUD Public Housing Noi
Housing Choice No
Voucher

Source: GAO survey of federal agencies, agency annual performance plans.

aKey program reports include routine reports that agencies must prepare
for Congress for the purpose of providing updated information on the
status of the programs they administer.

bUSDA's methodology for calculating the WIC coverage rate is being
revised. The agency has not yet started reporting in its annual program
reports but it has plans to do so.

cHHS includes a TANF coverage rate as part of their Indicators of Welfare
Dependence report.

dCCDF has published a coverage rate for child care in their Biennial
Report to the Congress, which includes SSBG and TANF funds.

eIRS has calculated a participation rate for the EITC, but only for
selected years.

fIRS has plans to use a participation rate as part of their Strategic
Plan.

gSSA has calculated a participation rate for the years 1991-1996, but only
for a subpopulation of those participating in SSI.

hED has calculated a take-up rate, which is the share of those who applied
and were found eligible for a Pell Grant who actually received a grant.

iHUD has calculated a voucher utilization rate for the housing choice
voucher program, and tracks the occupancy rates of public housing
authorities.

USDA has taken steps to develop a reliable participation rate measure for
the WIC program and plans to estimate the WIC participation rate annually.
Because the WIC program is not an entitlement and participation is
determined by the appropriated funds, FNS officials told us they do not
plan to establish a performance goal related to the WIC participation
rate. However, officials told us that they do think it is important to
monitor and report on the WIC participation rate for policy, planning, and
budgeting purposes, and the agency plans to use these data in managing the
program by highlighting the measure in the program's budget reports and
its annual report to Congress.

HHS plays a key role in estimating participation rates for several
meanstested programs. Since 1973, HHS's Office of the Assistant Secretary
for Planning and Evaluation has funded TRIM3 for the purposes of policy
development, analysis, and research on such topics as the interactive
effects of programs serving low-income families. The Welfare Indicators
Act of 1994,29 requires HHS to submit annual reports to Congress on
welfare receipt in the United States and key indicators and predictors of
welfare dependence. These reports, produced annually since 1997, include
participation and coverage rates from TRIM3 for TANF (and its predecessor
program, Aid to Families With Dependent Children), SSI, and food stamps.

HHS includes CCDF coverage rates30 in the biennial congressional report
and plans to include a performance measure related to participation rates
in its annual performance reports starting in fiscal year 2006. However,
HHS generally does not use participation rate estimates it has available
from TRIM3 in managing its other programs by highlighting the information
in its annual performance plans or congressionally mandated program
reports. For example, TANF participation rates are not included in the
agency's TANF Annual Report to Congress and Medicaid and SCHIP
participation and coverage rates are not included in any of the other
congressionally mandated reports disseminated by CMS, the agency that
administers these programs.

Only one of the other agencies we reviewed, the IRS, has plans to estimate
and report on participation rates for the programs they administer. IRS

29Pub. L. No. 103-432 (1994).

30CCDF's coverage rate also includes an estimate of the children served
through the TANF block grant and the Social Services Block Grant.

plans to begin estimating the participation rate in the EITC program on an
annual basis starting in fiscal year 2005 and to use the EITC
participation rate as a performance measure in its annual performance
report. However, the other agencies we reviewed, including SSA, HUD, and
ED, do not estimate participation or coverage rates for the SSI, Housing
Choice Voucher, Public housing, or Pell Grant programs and reported no
plans to do so.31

There are several reasons agency officials cited for not collecting
participation rate information. For some programs, information on certain
eligibility criteria can be difficult or costly to obtain. For example,
reliable data on the disability status of children who do not receive SSI
is not readily available from national survey data. We were unable to
provide a participation rate estimate for this group for this reason.
Absent a legislative or other mandate to estimate participation and
coverage rates, agency officials may not consider conducting such
estimates a priority. However, some agency officials that do not regularly
estimate participation or coverage rates indicated they would find such
estimates useful. For example, HUD officials indicated that the agency
finds its estimates of worst case housing needs, which include some
coverage rate measures for all housing programs, useful for policy and
planning purposes. Similarly, while Head Start officials do not publish or
report coverage rate estimates in any agency documents, they indicated
that they occasionally estimate participation rates informally for
internal use in policy, planning, and budgeting.

Conclusion Federal agencies responsible for administering means-tested
programs are charged with ensuring that those eligible for program
benefits and services are able to access them and with ensuring prudent
oversight of taxpayer dollars by maintaining program integrity, including
preventing improper payments. Program administrators may find it
challenging to achieve these

31ED monitors the share of student financial aid applicants found eligible
for a Pell Grant who receive a grant. We would refer to this as a
"take-up" rate rather than a participation rate as we have defined it,
because it does not provide information on individuals who are eligible
but do not complete an application. Similarly, HUD monitors the
utilization rate in its housing choice voucher program. This measure
tracks the extent to which voucher recipients are able to redeem their
benefits by leasing apartments. Like the Pell Grant "take-up" rate, this
rate provides program managers with important information they need to
manage their programs, but it does not provide them with a sense of the
extent to which their programs are reaching all eligible households and
is, therefore, not comparable to a participation or coverage rate for
monitoring program access.

goals in part because they may conflict-ensuring program integrity can
potentially come at the expense of program access. While we identified a
number of strategies that have been implemented by federal, state, and
local agencies that could potentially achieve both of these goals, unless
agencies track progress toward meeting these goals in a way that signals
to program administrators at all levels of government that both access and
integrity are federal priorities, there is a risk that one of the goals
may be compromised. Estimating participation rates for means-tested
programs is difficult-and more difficult for some than for others-and
cannot be considered "an exact science." Nevertheless, when done
rigorously-in a way that ensures that estimates are comparable over time
and quantifies the errors that result from the estimation
methodology-these estimates can provide a basic understanding of the
extent to which needy populations are being served by federal programs and
can be a key component of a federal agency's efforts to establish a
culture of accountability that emphasizes the importance of program
access. Federal agencies responsible for administering only four of the
programs we covered in this review-CCDF, food stamps, WIC, and EITC-either
currently collect and report information on the extent to which they are
reaching their target populations in key performance and program reports
or plan to do so. As a result, the federal agencies that administer the
other nine programs covered by this review may lack data that could inform
their budgetary and programmatic decisions, assist them in ensuring that
their programs are coordinated with and appropriately complement other
programs available to their target populations, help them to manage their
programs effectively, and enable them to provide policymakers with the
information they need to make decisions.

Recommendations to Executive Agencies

Recommendations to the As the department moves forward with participation
rate estimates for the Secretary of the WIC program, we recommend that the
Secretary:

Department of Agriculture

1. take steps to clarify to users the limitations of the estimates; and

2. comparability between estimates by reanalyzing data as estimation
methodologies change, so that consistent methods are applied over time.

Recommendation to the To help ensure that agencies have information on
program access, we Secretaries of the recommend that the Secretaries of
Education and HUD study the Department of Education feasibility of
calculating participation or coverage rates and including and the
Department of them in key program management reports.

Housing and Urban Development Recommendations to the Secretary of the
Department of Health and Human Services

We recommend that the Secretary of HHS consider making some improvements
to the participation and coverage rate information produced for the CCDF,
Medicaid, SCHIP, and TANF programs, such as:

1. quantifying errors that result from calculating these estimates to
help users better understand the accuracy of the data; and

2. ensuring, to the extent possible, that estimates are comparable over
time.

We also recommend that the Secretary include participation and coverage
rate estimates in key program reports for these four programs.

We also recommend that the Secretary study the feasibility of estimating
the coverage rate for the Head Start program on a regular basis and to
include these estimates in key Head Start program reports.

Recommendation to the As the IRS moves forward on developing participation
rate estimates to Commissioner of the use as a program performance measure
for the Earned Income Tax Credit, Internal Revenue Service we recommend
that the Commissioner of the Internal Revenue Service

take steps to quantify errors that may result from estimating these
participation rates to help users better understand the accuracy of the
data and ensure that estimates will be comparable over time.

Recommendation to the To help ensure that the agency has information on
program access, we Commissioner of the recommend that the Commissioner
consider the feasibility of providing Social Security participation rate
information (from the existing source or another source) Administration in
SSA program management reports.

Agency Comments

We provided drafts of this report to and received comments from the
Departments of Agriculture, Education, Health and Human Services, and
Housing and Urban Development; the Internal Revenue Service; and the
Social Security Administration. They generally agreed with our conclusions
and recommendations. Agency written comments appear in appendixes IV
through VII.

On February 3, 2005, we met with FNS officials to discuss their comments.
The officials said they agreed with our conclusions and recommendations.
They stated that they recognize the importance of tracking both program
access and program integrity and their ongoing efforts to estimate
participation rates in the food stamp program reflect their commitment to
balancing the goals of access and integrity. They stated that as they move
forward in developing a participation rate estimate for the WIC program,
they plan to take steps to clarify to users the limitations of the
estimates and to ensure that the estimates will be comparable over time.
Agency officials also provided technical comments, which we incorporated
in the report where appropriate.

The Department of Education agreed that data sources now available would
not provide sufficient information to develop a reliable participation
rate estimate for the Pell Grant program. However, the department cited a
study currently being conducted by the National Center for Education
Statistics that will determine the feasibility of implementing a system of
collecting data on all postsecondary students. This study will also
provide information that could be used to examine the feasibility of
estimating participation rates for the Pell Grant program.

The Department of Health and Human Services provided comments related to
each of the five programs under its jurisdiction. The department noted
that it has taken steps to measure program access in the TANF, CCDF,
Medicaid, and SCHIP programs through their funding of TRIM3, although they
do not always report these estimates in key program reports. With regard
to CCDF, HHS recognized the importance of coupling efforts to control
improper payments with efforts to provide effective service delivery, and
therefore access. The department noted that it has recently established a
performance measure related to coverage rates for the CCDF program. We
incorporated this information into the report. HHS also thought that we
overstated the differences between CCDF and TANF in the impact of state
flexibility on coverage rates so we changed that section of the report.
With regard to TANF, HHS raised the concern that increasing coverage rates
may conflict with other TANF program goals and consequently, coverage
rates are not an appropriate performance

measure for the TANF program. We recognize that increasing the coverage
rate is not an explicit goal of TANF and believe this concern is
consistent with the report's conclusion that increasing program access may
not be feasible or desirable for some programs. Nevertheless, we continue
to believe that including program participation and coverage rates in key
program reports can help inform policymakers and program administrators at
all levels of government and assist them making budgetary and programmatic
decisions. In its discussion of SCHIP and Medicaid, HHS indicated that the
report reinforced the agency's approach to promoting both access and
integrity. The department noted that CMS is working to improve its
coverage rate estimates for SCHIP and meanwhile, it continues to measure
progress toward increasing coverage among children under 200 percent of
poverty, a proxy for the program coverage rate.

HHS also highlighted several other issues that we believe are already
sufficiently discussed in the report. These include the fact that not all
those eligible and not participating need program services; that some
programs allow states considerable flexibility in implementing programs
and determining eligibility; that participation or coverage rates can be
challenging to estimate accurately; and that uncertainties and limitations
surrounding participation and coverage rate estimates should be made clear
to users of this information. HHS also reiterated some of the limitations
of our cost estimates and our Head Start coverage rate estimates. Agency
officials also provided technical comments, which we incorporated in the
report where appropriate.

Department of Housing and Urban Development officials provided us with
technical comments, which we incorporated in the report where appropriate.

The Internal Revenue Service agreed with our conclusions and
recommendation. The IRS recognizes the interactions between efforts to
improve program access and those intended to improve program integrity,
and as part of their efforts to utilize a balanced approach that
emphasizes both goals, the IRS has recently established the EITC
participation rate as a GPRA measure. As it moves forward in developing
the EITC participation rate, the IRS plans to take steps to quantify
errors that may result from estimating these participation rates to help
users better understand the accuracy of the data and ensure that estimates
will be comparable over time. Agency officials also provided technical
comments, which we incorporated in the report where appropriate.

The Social Security Administration agreed with our conclusions and
recommendation. The agency noted that it has considered the feasibility of
estimating participation rates in the past and has calculated
participation rates among the elderly. However, the agency also noted that
limitations of the national survey data make such estimates challenging
for the SSI program. The agency agreed that estimating participation rates
on a regular basis could provide an important management tool.

As agreed with your office, unless you publicly announce its contents
earlier, we plan no further distribution of this report until 30 days
after its
issue date. At that time we will send copies of this report to the
Secretaries
of the Departments of Agriculture, Education, Health and Human Services,
and Housing and Urban Development; the Commissioners of the Internal
Revenue Service and the Social Security Administration; relevant
congressional committees; and others who are interested. Copies will be
made available to others upon request, and this report will also be
available on GAO's Web site at http://www.gao.gov.

If you or your staff have any questions about this report, please contact
me
on (415) 904-2272. Additional GAO contacts and acknowledgments are
listed in appendix VIII.

Sincerely,

David Bellis, Director
Education, Workforce, and Income Security

Appendix I: Objectives, Scope, and Methodology

We designed our study to provide information on (1) the proportion of
those eligible who are participating in 12 selected low-income programs;
(2) the factors that influence participation in those programs; and (3)
strategies used by federal, state, and local administrators to improve
both access and integrity, and whether agencies monitor access by
measuring participation rates. To obtain information on these issues, we
compiled estimates of participation and coverage rates for each of the
programs covered by the review for the most recent year for which data
were available, we reviewed literature on participation in programs for
low-income people, and we gathered information directly from both federal
program officials and state and local administrators. To determine the
completeness and accuracy of data obtained from various sources, we took
several steps, described below, to assess the data and determined that the
data were sufficiently reliable for use in this report.

We performed this work between November 2003 and January 2005 in
accordance with generally accepted government auditing standards.

Participation and Coverage Rates

To obtain information on the proportion of those eligible who are
participating in 12 selected low-income programs, we used information
developed by the Urban Institute under contract with HHS about the
participation and coverage rates they estimated for 7 programs-CCDF,
Medicaid, TANF, food stamps, SCHIP, SSI, and WIC. Also, we estimated the
coverage rates for the Head Start program using administrative data and
CPS data and for the Housing Choice Voucher and Public Housing programs
using data from a HUD report on worst case housing needs in 1999. The
participation rate estimate for the EITC program was from our previous
work. In addition, we performed a literature review to obtain information
about program participation and coverage rates that were estimated by
other organizations. When possible, we obtained rates from 1997 to the
most recent year available, but as mentioned in the report, we did not
present many estimates of rates prior to the most recent year available
because changes in methodologies did not allow us to determine whether
changes in the rates over time were actual changes in participation rates
or resulted from changes to the estimation methodology. We were unable to
provide a Pell Grant program participation rate because we were unable to
assess the reliability of available data on the family income of students
who did not apply for federal financial assistance, which is necessary for
determining program eligibility. Table 11 summarizes the methodologies
used to calculate participation and coverage rates for the 12 programs
covered in this review.

Appendix I: Objectives, Scope, and Methodology

Table 11: Summary of Participation and Coverage Rate Methodologies
version 3 model
Participation Estimate The Urban TANF, of (TRIM3), Estimate TRIM3; of developed Estimate
or coverage Program methodology Data of Institute SCHIP, potentially developed of Mathematica potentially under EITC of percentage
rate source SSI, recipients Transfer Income WIC eligible under Food recipients Policy eligible contract eligible of
Medicaid, as a Microsimulation contract Stamp as a Research, Inc., with recipients potentially
CCDF, percentage model, with HHS. Program percentage microsimulation USDA. as eligible

GAO estimate based on administrative and CPS data and compliance
information provided by IRS, see GAO, Earned Income Tax Credit
Participation, GAO-02-290R (Washington, D.C.: Dec. 14, 2001).

Housing Choice Voucher, Estimate of eligible recipients as GAO estimate based on
administrative data and data from the

public housing

percent of potentially eligible HUD 1999 worst case needs report. Head Start
Estimate of number of funded slots as GAO estimate based on administrative and
national survey data

a percentage of 3- and 4-year-old children in families below poverty

from the U.S. Census Bureau's Annual Social and Economic Supplement to the
Current Population Survey (CPS).

Participation and Coverage Rates Estimated by The Urban Institute

Source: GAO.

Note: Some participation and coverage rate estimates are based on the
average monthly number of recipients and people who are eligible for a
program while others are based on the total number of recipients and
people who are eligible at any point in the year. The participation and
coverage rate for any program will differ by whether average monthly or
total numbers are used.

The Urban Institute estimated participation and coverage rates for CCDF,
Medicaid, TANF, food stamps, SCHIP, SSI, and WIC under a contract with the
Department of Health and Human Services (HHS). In order to evaluate the
reliability of these data and to determine whether it met our standards
for reporting quantitative information, we contracted with the Urban
Institute to obtain additional information about these estimates and the
methodologies used to calculate them, including the Transfer Income Model,
version 3 (TRIM3).

The TRIM3 microsimulation model was used to determine the number of people
who would be eligible for a program if they applied. This model simulates
the process that a caseworker would undergo to determine eligibility by
reviewing individual or household characteristics such as household
composition, income, disability, and other factors as appropriate for the
programs. TRIM3 primarily relies on the Census Bureau's Annual Social and
Economic Supplement to the Current Population Survey to estimate the
eligible population. For some programs, this data source does not include
all information needed to determine a person's or household's eligibility.
For example, the data do not include information on the value of certain
assets that might be considered during eligibility determination for some
programs. In these instances, the Urban Institute researchers used TRIM3
to assign eligibility-related

Appendix I: Objectives, Scope, and Methodology

characteristics to individuals included in the CPS who they believed were
likely to have them, a process referred to as imputation. The estimates of
the eligible population were used as the denominators of the participation
and coverage rates, and administrative data on the number of participants
were generally used as the numerator. In general, groups excluded from the
TRIM3 eligibility estimate used for the denominator (such as those who are
institutionalized) were also excluded from the administrative data used
for the numerator.

The CPS data limitations and the process of assigning characteristics may
have resulted in errors in the estimates of the eligible population and,
therefore, the participation and coverage rates for seven programs. These
include sampling error, non-sampling error, error related to the
imputation process, and error related to program implementation. GAO
statisticians and social science analysts reviewed the Urban Institute's
TRIM3 technical documentation, interviewed the Urban Institute
researchers, requested additional TRIM3 calculations from the Urban
Institute, and reviewed available academic and government literature on
the sources of error in microsimulation models of low-income program
participation to determine what errors existed. Although there are several
types of potential errors, not all the errors could be quantified.

o The CPS, like all sample surveys, is subject to sampling error. This
error reflects the difference between (1) the results that are obtained
from the members of the sample that are actually surveyed and (2) the
results that would be obtained if each member of the population were
surveyed. We contracted with the Urban Institute to calculate the sampling
error of CPS data used in TRIM3, and we verified that they used the
appropriate procedures to make these calculations. We used the estimates
of sampling error provided by the Urban Institute to calculate a range for
the most recent TRIM3 estimates at the 95 percent confidence level, and
these ranges are presented in the report. However, the ranges
underestimate the total error in TRIM3 estimates.

o The CPS data used in TRIM3 are also subject to non-sampling error that
has not been quantified. These errors can occur if survey respondents have
difficulty interpreting a particular question, lack information necessary
to answer a question, are uncomfortable with accurately reporting certain
sensitive information, or do not answer certain questions, among other
factors that affect data collection and measurement. Such non-sampling
errors are likely to have affected variables that are used by TRIM3, such
as labor force participation for

Appendix I: Objectives, Scope, and Methodology

Participation and Coverage Rates We Estimated

example. According to a 2002 Census Bureau report, the 1996 CPS covers a
smaller proportion of the Black population (84 percent) than it does of
the White population (94 percent). The Census Bureau reported that this
low coverage rate biases labor force estimates, because those who are left
out of the sampling frame are quite different from those who are included
in it. Yet, the full extent of non-sampling error in the CPS is unknown.

o The process of assigning characteristics when CPS data are not
available-the imputation process-also can produce error. In most
instances, efforts were not made to quantify the error associated with
these imputed data. However, a National Research Council (NRC) panel
attempted to quantify such error when reviewing methods for estimating
participation rates for WIC and found that estimates can vary noticeably
depending on whether imputed or reported monthly income data are used to
determine eligibility. During this review, the NRC panel compared
TRIM-adjusted CPS data, in which monthly income is imputed on the basis of
annual income, and Survey of Income and Program Participation (SIPP) data,
in which monthly income is collected directly from survey respondents.

o Specification error can result when the variables in a microsimulation
model incorrectly represent the process for determining program
eligibility that actually occurs between caseworkers and applicants. For
example, in its TANF module, TRIM3 does not consider certain aspects of
program eligibility, including specific work requirements, behavioral
requirements such as school attendance and immunizations, and sanctions
from failing to comply with work rules or child support rules. These
exclusions could cause TRIM3 to overestimate eligibility and therefore
underestimate participation and coverage rates. Similarly, while several
means-tested programs base their eligibility on the federal poverty
guidelines, there is no universal administrative definition of "family,"
or "household" that is valid for all programs that use the poverty
guidelines. To the extent that federal programs use administrative
definitions that differ somewhat from the statistical definitions used by
the CPS, estimates of the eligible population could be affected.

We estimated the participation and coverage rates for the Head Start,
Housing Choice Vouchers, Public Housing, and EITC (in prior work). These
estimates are also limited by the data sources and assumptions of who may
be eligible for these programs. The EITC and Head Start rates were
estimated using the CPS and, therefore, estimates are subject to sampling
and non-sampling errors mentioned earlier. In addition, because

Appendix I: Objectives, Scope, and Methodology

the CPS did not have complete information to determine who may be eligible
for the EITC, we made assumptions about whether children were in a
household long enough for them to be counted as a member for tax purposes.
We also assumed that the lack of data on certain assets would not
significantly affect participation rate estimates.1 For Head Start, the
administrative data did not allow us to determine the number of children
who are actually served in the program. The data were only on the number
of program slots funded, and more than one child can be served with each
slot over the course of a program year if there is turnover in
participation. However, for the purposes of calculating Head Start
coverage rates, we assumed that only one child filled a slot.

We used a HUD report on the households with worst case housing needs in
1999 and administrative data to estimate the Housing Choice Vouchers and
Public Housing coverage rates. HUD used data from the American Housing
Survey in this report.2 We assessed the reliability of the survey data and
determined that the data were sufficiently reliable for use in this
report. In conducting this assessment, we reviewed Bureau of the Census
studies on the survey. Census reported that the American Housing Survey
underestimates income and overestimates poverty for households, variables
used to determine the number of people with housing needs. We were unable
to quantify the error this may have produced in the estimates.

We determined that the administrative data used to estimate participation
and coverage rates were reasonably reliable for this purpose. As a part of
this assessment, we reviewed the data on the number of participants and we
discussed this information with federal agency officials. However, with
the exception of the EITC data, we did not adjust the number of
participants to account for those who may have received program benefits
and/or services in error. As a result, the administrative data could have
produced errors in the estimates due to potential inaccuracies in
reporting the number of eligible participants.

Reliability of Administrative Data

1For more details, see GAO, Earned Income Tax Credit Participation,
GAO-02-290R (Washington, D.C.: Dec. 14, 2001).

2For more details, see U.S. Department of Housing and Urban Development,
Trends in Worst Case Needs for Housings, 1979-1999: A Report to Congress
on Worst Case Housing Needs (Washington, D.C.: December 2003).

Appendix I: Objectives, Scope, and Methodology

Cost Estimates

To obtain the estimated cost of providing benefits to those eligible and
not participating in the Food Stamp Program, SSI, TANF cash assistance and
WIC, we contracted with the Urban Institute to provide us with these
estimates. The organization used the TRIM3 model to estimate the total
benefit amount for eligible non-participants, and these estimates were
based on the amount that each individual or household would receive based
on their characteristics. These estimates have similar limitations as
their related participation and coverage rates, and not all of their
errors could be quantified.

The EITC cost estimate was based on our prior work,3 but the Head Start
cost estimates were done during the course of this project. We estimated
the potential costs for Head Start using administrative data on the
average cost per slot and the number of eligible non-participating
children we estimated using the CPS. We assume that only one child would
be served with a slot, even though in actuality more one than can. We also
did not account for variation in program cost at the local level, and we
did not account for variation in local eligibility criteria.

All of the estimates have two major limitations. We did not include the
administrative costs of serving the non-participants; this would increase
the cost estimates for all programs. For some programs, federal, state,
and local governments would share these administrative costs. We also did
not factor in policy or economic changes that may occur if participation
increased in these programs. In addition, for all programs, we calculated
the estimates for each program individually and we did not account for how
changes in the participation for one program could influence benefit
amounts in another.

We were unable to estimate the cost of serving eligible, non-participants
in the five other programs-CCDF, HCV, Medicaid, Pell Grants, Public
Housing, and SCHIP-because we did not have enough information to determine
how the characteristics of the non-participants would have affected the
benefit amounts they could have received.

3For more details, see GAO, Earned Income Tax Credit Participation,
GAO-02-290R (Washington, D.C.: Dec. 14, 2001).

Appendix I: Objectives, Scope, and Methodology

Factors That Affect Participation and Strategies to Improve Access and
Integrity

To obtain information about factors that affect participation and
strategies to improve access and integrity for the 12 selected programs,
we conducted a literature review, obtained documentation from federal
officials, interviewed federal officials and interviewed state and local
officials in California, Connecticut, Georgia, Maryland, and Minnesota.
The literature review consisted of studies on factors that affect
participation and program integrity issues that we found through searches
of the Internet and social science research databases and recommendations
from federal agency officials. Among the documentation we reviewed for
this objective were the available fiscal year 2004 and 2005 annual plans
for each program and agency reports to Congress to determine whether the
relevant agencies had stated measures for program integrity and access
outcomes. We selected states to visit primarily based on their innovative
practices to address barriers to participation for multiple programs we
reviewed. We obtained information on innovative practices from federal
officials, federal agency publications and a database on state and local
initiatives complied by the Administration for Children and Families and
several non-profit, non-partisan agencies. We also considered state
efforts to ensure program integrity, program aspects that may create
barriers to access and demographic and geographic diversity in making our
selection. We were not able to address all the programs we reviewed during
every site visit, and information from these site visits cannot be
generalized to all states and localities and to all programs we reviewed.
In addition, we did not assess the effectiveness of the strategies we
presented.

Appendix II: Limitations Affecting Use of Participation and Coverage Rates

While participation and coverage rate estimates provide important program
information that can be useful to both policymakers and program
administrators, these estimates must be interpreted carefully to take into
consideration such factors as survey data, research methodology,
timeliness of estimates and the availability of alternative programs.
These factors limit our ability to compare rates across programs and must
be considered in efforts to define appropriate or desired participation or
coverage rate levels.

Limitations of National Survey Data

There are two national household level surveys conducted by the Census
Bureau that can be used to estimate the total number of people eligible
for low-income programs, the Annual Social and Economic Supplement to the
Current Population Survey (CPS) and the Survey of Income and Program
Participation (SIPP) and each has advantages and disadvantages. The CPS is
the larger of the two surveys and data from this survey are available
annually for the prior year. In addition, it has detailed information
about income, household composition, labor force participation, and
receipt of many federal benefits, but it does not include all the
information that may be needed to determine program eligibility. For
example:

o Although the value of assets such as automobiles and savings accounts
are considered when determining financial eligibility for some of programs
we reviewed, the CPS does not collect some of these data.

o The CPS also omits institutionalized individuals, some of whom may be
eligible for Medicaid and SSI.

Researchers at the Urban Institute and Mathematica told us that, while
they were aware of these disadvantages, they continue to use CPS data in
their calculations because of the relatively large sample size and because
the CPS is conducted annually, which makes it a good source for yearly
snapshots of participation. To compensate for the limitations in the data,
researchers either make assumptions and assign information needed to
determine eligibility, such as asset values, to individuals in order to
determine their program eligibility or they omit certain subpopulations
such as institutionalized individuals or pregnant and postpartum women
from their participation rate estimates.

The SIPP is a smaller survey than the CPS, but because it was designed
specifically to better measure low-income program participation, it has
some advantages over the CPS. Specifically, it includes more detailed
information about individual characteristics, monthly labor force
activity, monthly income, and household assets that factor in program
eligibility

Appendix II: Limitations Affecting Use of Participation and Coverage Rates

determinations.1 However, the SIPP is designed as a panel study (following
the same individuals for 2 1/2 to 4 years) and thus may be more suited for
use in examining detailed characteristics of participants and reasons for
movement into and out of social programs. In addition, the sample faces
attrition over the survey period, and some data needed to determine
eligibility are available only for a limited number of years. The National
Research Council used the SIPP to estimate WIC participation rates. This
estimate included pregnant women in addition to infants and children, but
by 2003 when this estimate was published, the most recent estimate they
could create using SIPP was for 1998.

There are some low-income programs for which neither the CPS nor the SIPP
provide sufficient information to estimate participation rates. For
example, to estimate the proportion of enrolled students eligible for a
Pell Grant would require information on student enrollment status in
addition to family income information. The National Center for Education
Statistics within the Department of Education conducts a survey, the
National Postsecondary Student Aid Study, that includes both student
enrollment and family income information. However, concerns have been
raised about the reliability of the family income data for students who
did not apply for federal financial assistance. We were unable to assess
the reliability of this data element and thus were unable to determine
whether the data were sufficiently reliable for our purposes. As a result,
we did not estimate a participation rate for this program.

Errors from Research Methodology

Because no existing national data provide every piece of information
needed to determine the exact size of the eligible population for
lowincome programs, all participation rate estimates involve certain
assumptions and generalizations. These assumptions can create errors that
are difficult or impossible to quantify. Consequently, the margins of
error we present, which only capture errors that result from sampling,
could understate the true range within which participation rate estimates
are most likely to fall.

One way to better understand how these estimates are affected by the
research methodology is to compare them to alternative participation rate
estimates that have been calculated using different assumptions. While the

1The SIPP collects a detailed inventory of real and financial assets once
a year and conducts more frequent measures of assets relevant for
assistance programs.

Appendix II: Limitations Affecting Use of Participation and Coverage Rates

Urban Institute and Mathematica Policy Research, Inc., both estimate food
stamp participation rates based on CPS data, they use slightly different
assumptions to estimate the eligible population. For example, because the
CPS does not include information on which household members purchase and
prepare food together, model estimates had to make assumptions about
household food preparation habits. Despite differences in assumptions, the
participation rates estimated by the Urban Institute and Mathematica were
very similar.

Years for Which Estimates Are Available

The most recent participation rate data available varies by program, and
all estimates are at least 3 years old. As a result, changes in
legislation, state policy, program administration, the national economy,
and other factors that may have influenced the participation rate over the
past few years are not reflected in these estimates.

o The EITC participation rate estimate is based, in part, on an
estimated 32 percent of EITC claims being made in error in 1999. Since
that time, Congress has enacted new tax laws and the IRS has taken steps
to improve compliance.

o The total number of people receiving food stamps and Medicaid fell
after passage of the Personal Responsibility and Work Opportunity
Reconciliation Act of 1996 when participation in these programs was
"de-linked" from participation in TANF. Since 2000, however, states have
implemented policies to ensure that those eligible for food stamps and
Medicaid continue to receive benefits when they leave TANF.

o Since 2000, the national economy has emerged from a recession, but
economic growth, until recently, has been slow and uneven and marked by
slow job growth. Economic conditions have historically been correlated
with low-income program participation.

None of these factors are reflected in the numbers we presented in this
report.

Availability of Alternative When interpreting the participation and
coverage rate estimates we

Programs provided, it is important to keep in mind that those eligible
but not participating in the programs we reviewed may not be living
without assistance; many could be receiving assistance from alternative
programs. In some cases, people could be eligible for more than one
program we reviewed and able to receive similar services from these
programs. Those eligible could also have access to state and local
programs that address

Appendix II: Limitations Affecting Use of Participation and Coverage Rates

similar needs. This is particularly relevant for the non-entitlement
programs covered by our review.

The coverage rates we provide for the two housing programs in this review
are calculated in isolation, and as a result, some of the households that
appear eligible but not participating in the Public Housing program are
receiving Housing Choice Vouchers (HCV), and vice versa. In addition,
while HCV and public housing are two of the largest federal housing
programs, there are several other federal, state, and local housing
programs from which eligible households could receive assistance. HUD
estimated that in 1999 about a quarter of all households eligible for any
kind of housing assistance received assistance.

The availability of alternative programs could also affect coverage rates
for CCDF and Head Start. For example, state-financed pre-school programs
provide child care and educational development services for many
4-year-olds who appear eligible but are not enrolled in Head Start.
Children could be eligible for both programs but only receiving services
in one, but this would not be reflected in program coverage rate
estimates. HHS officials have recognized that other federal programs, such
as TANF and Social Services Block Grant (SSBG), also provide child care to
children eligible for CCDF. Because the CCDF participation rate does not
reflect these additional sources of federal child care funds, HHS has
produced a different coverage rate, estimating that about 26 percent of
CCDF-eligible children were receiving child care services through one of
four federal funding streams-CCDF, SSBG, TANF, and TANF state funds in
2001.2

Cost Considerations State and localities are able to set the eligibility
criteria for some of these programs, and the cost estimates do not account
for the policy changes agencies may make if the current criteria resulted
in increased participation. For example, if states found that current
eligibility limits may result in increased TANF participation, they may
change eligibility criteria or benefit levels to ensure state costs do not
increase. Either of these changes may reduce the total additional costs.

2HHS does not collect information on the number of children who receive
child care subsidies through the SSBG, TANF and TANF Maintenance of Effort
funding. Therefore, the agency estimated the number of children served
through those other funding source by dividing the total amount spent on
child care services from each source by the typical cost of serving a
child under CCDF.

Appendix II: Limitations Affecting Use of Participation and Coverage Rates

All cost estimates were calculated individually and, therefore, do not
account for how participation changes in one program may influence benefit
amounts in another. For example, Food Stamp Program costs may not be as
high if participation in the SSI program increased because some eligible
non-participants would qualify for lower food stamp benefits due to the
cash assistance. Also, as mentioned earlier, other federal, state, and
local programs may provide similar benefits that non-participants could be
receiving, and, therefore, they may not need or want the benefits from the
programs we covered. For example, food stamp eligible non-participants
with short-term nutritional needs may decide to receive assistance through
food banks. In addition, the TANF estimate we provide is only the cost of
providing cash assistance to eligible people who are not participating;
this program offers other supports to needy families, such as
transportation to work and child care, that is funded by the federal and
state government.

Appendix III: Agency Estimates of Improper Payments for Programs Reviewed

Table 12: Improper Payment Estimates Reported in Agency Fiscal Year 2003
Performance and Accountability Reports
Reported Food
amount Agency Program payments USDA Stamp $1,507,000,000 WIC NA HHS/ACF TANF NA CCDF NA Head NA IRS EITC 10,500,000,000 HHS/CMS Medicaid NA SCHIP NA SSA SSI 2,740,000,000 Education Pell 377,500,000 HUD Public 650,000,000 Housing $1,877,000,000
of Program Start Grant Housing Choice (Tenant- and
improper Voucher Project-Based)

Source: GAO, Financial Management: Fiscal Year 2003 Performance
Accountability Reports Provide Limited Information on Governmentwide
Improper Payments, GAO-04-631T (Washington, D.C.: Apr. 15, 2004). Data
from this report were based on agency fiscal year 2003 Performance and
Accountability Reports.

Appendix IV: Comments from the Department of Education

Page 68 GAO-05-221 Means-Tested Programs

Page 69 GAO-05-221 Means-Tested Programs

Page 70 GAO-05-221 Means-Tested Programs

Page 71 GAO-05-221 Means-Tested Programs

Page 72 GAO-05-221 Means-Tested Programs

Page 73 GAO-05-221 Means-Tested Programs

Page 74 GAO-05-221 Means-Tested Programs

Page 75 GAO-05-221 Means-Tested Programs

Page 76 GAO-05-221 Means-Tested Programs

Page 77 GAO-05-221 Means-Tested Programs

Appendix VI: Comments from the Department of the Treasury

Appendix VII: Comments from the Social Security Administration

Appendix VIII: GAO Contacts and Staff Acknowledgments

GAO Contacts

Staff Acknowledgments

Gale Harris (202) 512-7235, [email protected] Kathryn Larin (202) 512-5045,
[email protected]

Carolyn Blocker, Cheri Harrington, Danielle Giese, Amanda Mackison, and
Kim Siegal also made significant contributions to this report. Jay Smale,
Mark Braza, Doug Sloane, Ed Nannenhorn, and Beverly Ross provided
technical assistance in developing and evaluating the reliability of
participation rate and potential program cost estimates; Lise Levie
provided key technical assistance; and Johnnie Barnes, Carrie Watkins, and
Charles Wilson provided information on the two housing programs covered in
the report.

Related GAO Products

Food Stamp Program: Farm Bill Options Ease Administrative Burden, but
Opportunities Exist to Streamline Participant Reporting Rules among
Programs. GAO-04-916. Washington, D.C.: September 16, 2004.

Food Stamp Program: Steps Have Been Taken to Increase Participation of
Working Families, but Better Tracking of Efforts Is Needed. GAO-04-346.
Washington, D.C.: March 5, 2004.

SSA Disability Decision Making: Additional Steps Needed to Ensure Accuracy
and Fairness of Decisions at the Hearing Level. GAO-04-14. Washington,
D.C.: November 12, 2003.

Federal Student Aid: Expanding Eligibility for Less Than Halftime Students
Could Increase Program Costs, but Benefits Uncertain. GAO-03-905.
Washington, D.C.: September 10, 2003.

Social Security Disability: Reviews of Beneficiaries' Disability Status
Require Continued Attention to Improve Service Delivery. GAO-03-1027T.
Washington, D.C.: July 24, 2003.

Welfare Reform: States Provide TANF-Funded Work Support Services to Many
Low-Income Families Who Do Not Receive Cash Assistance. GAO-02-615T.
Washington, D.C.: April 10, 2002.

Food Stamp Program: States' Use of Options and Waivers to Improve Program
Administration and Promote Access. GAO-02-409. Washington, D.C.: February
22, 2002.

Medicaid and SCHIP: States' Enrollment and Payment Policies Can Affect
Children's Access to Care. GAO-01-883. Washington, D.C.: September 10,
2001.

Food Stamp Program: Program Integrity and Participation Challenges.
GAO-01-881T. Washington, D.C.: June 27, 2001.

Medicaid and SCHIP: Comparisons of Outreach, Enrollment Practices, and
Benefits. HEHS-00-86. Washington, D.C.: April 14, 2000.

Medicaid Enrollment: Amid Declines, State Efforts to Ensure Coverage after
Welfare Reform Vary. HEHS-99-163. Washington, D.C.: September 10, 1999.

Related GAO Products

Food Stamp Program: Various Factors Have Led to Declining Participation.
RCED-99-185. Washington, D.C.: July 2, 1999.

Medicaid: Demographics of Nonenrolled Children Suggest State Outreach
Strategies. HEHS-98-93. Washington, D.C.: March 20, 1998.

GAO's Mission

Obtaining Copies of GAO Reports and Testimony

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

The fastest and easiest way to obtain copies of GAO documents at no cost
is through GAO's Web site (www.gao.gov). Each weekday, GAO posts newly
released reports, testimony, and correspondence on its Web site. To have
GAO e-mail you a list of newly posted products every afternoon, go to
www.gao.gov and select "Subscribe to Updates."

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

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

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

To Report Fraud, Contact:

Waste, and Abuse in Web site: www.gao.gov/fraudnet/fraudnet.htm

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

Gloria Jarmon, Managing Director, [email protected] (202)
512-4400Congressional U.S. Government Accountability Office, 441 G Street
NW, Room 7125 Relations Washington, D.C. 20548

Public Affairs Paul Anderson, Managing Director, [email protected] (202)
512-4800 U.S. Government Accountability Office, 441 G Street NW, Room 7149
Washington, D.C. 20548

RECYCLED PAPER
*** End of document. ***