Food Stamp Program: States Have Made Progress Reducing Payment	 
Errors, and Further Challenges Remain (05-MAY-05, GAO-05-245).	 
                                                                 
In fiscal year 2003, the federal Food Stamp Program made payment 
errors totaling about $1.4 billion in benefits, or about 7	 
percent of the total $21.4 billion in benefits provided to a	 
monthly average of 21 million low-income participants. Because	 
payment errors are a misuse of public funds and can undermine	 
public support of the program, it is important that the 	 
government minimize them. Because of concerns about ensuring	 
payment accuracy GAO examined: (1) what is included in the	 
national food stamp payment error rate and how it has changed	 
over time, (2) what is known about the causes of food stamp	 
payment errors, and (3) what actions the Food and Nutrition	 
Service (FNS) and states have taken to reduce these payment	 
errors. To answer these questions, GAO analyzed program quality  
control data for fiscal years 1999 through 2003 and interviewed  
program stakeholders, including state and local officials from	 
nine states.							 
-------------------------Indexing Terms------------------------- 
REPORTNUM:   GAO-05-245 					        
    ACCNO:   A23458						        
  TITLE:     Food Stamp Program: States Have Made Progress Reducing   
Payment Errors, and Further Challenges Remain			 
     DATE:   05/05/2005 
  SUBJECT:   Accountability					 
	     Erroneous payments 				 
	     Food relief programs				 
	     Internal controls					 
	     Performance measures				 
	     Program management 				 
	     Quality control					 
	     Risk assessment					 
	     Risk management					 
	     Eligibility determinations 			 
	     Errors						 
	     Food Stamp 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-05-245

United States Government Accountability Office

GAO

                       Report to Congressional Committees

May 2005

FOOD STAMP PROGRAM

States Have Made Progress Reducing Payment Errors, and Further Challenges Remain

GAO-05-245

[IMG]

May 2005

FOOD STAMP PROGRAM

States Have Made Progress Reducing Payment Errors, and Further Challenges Remain

                                 What GAO Found

The national dollar payment error rate for the Food Stamp Program, which
combines states' overpayments and underpayments to program participants in
all states, has declined by almost one-third over the last 5 years to a
record low of 6.63 percent. This decline has been widespread; the rate
fell in 41 states and the District of Columbia, and rates in 18 of these
states fell by at least one-third. However, despite this decrease, some
states continue to have relatively high payment error rates. For example,
in 2003, 7 states had payment error rates of more than 10 percent.

Food Stamp Payment Errors Have Dropped over the Last 5 Years

Percent

12

9.86

10

8

6

4

2

0

1999 2000 2001 2002 2003

Fiscal year

Underpayments Overpayments Combined

Source: the Food and Nutrition Service.

Almost two-thirds of food stamp payment errors are caused by caseworkers,
usually when they fail to keep up with reported changes or make mistakes
applying program rules, and one-third are caused by participant failure to
report required, complete, or correct information, such as household
income and composition. State officials said program complexity and other
factors, such as the lack of resources and staff turnover, can contribute
to these errors. In fiscal year 2003, states referred about 5 percent of
all cases identified with errors for suspected participant fraud
investigation.

To increase food stamp payment accuracy, FNS and the 9 states GAO reviewed
took many approaches that parallel good internal control practices. These
efforts include increasing the leadership and accountability in the
program, performing risk assessments to identify problem areas,
implementing various program and process changes in response to the
findings from risk assessments, and monitoring and promoting improved
performance. The states are using a combination of approaches to improve
payment accuracy, making it difficult to tie error rate improvements to
specific practices. However, state officials point to their improved state
error rates as evidence of a collective impact.

                 United States Government Accountability Office

Contents

Letter

Results in Brief
Background
The Food Stamp Error Rate, Which Combines Overpayments and

Underpayments, Has Declined by Almost One-Third over the Last 5 Years
Caseworkers Cause about Two-Thirds and Participants Cause

about One-Third of Payment Errors FNS and States Have Taken Steps to
Increase Payment Accuracy Concluding Observations Agency Comments

                                       1

                                      3 5

10

16 24 36 37

Appendix IMethodology for Determining the Causes of Food Stamp Payment
Errors for Fiscal Years 1999 through 2003

Appendix IIFood Stamp Combined Error Rates by State for Fiscal Years 1999
to 2004

Appendix III GAO Contacts and Acknowledgments

Related GAO Products

Tables

Table 1: Changes in Payment Error Rates for States Providing the Largest
Amount in Food Stamp Benefits, Fiscal Year 2003 14 Table 2: Caseworker
Errors Most Often Resulted in Incorrect Household Income or Deductions,
Fiscal Year 2003 18 Table 3: Participant-Caused Errors Most Often Resulted
in

Incorrect Household Income Determinations, Fiscal Year

2003 23

Figures

Figure 1: Food Stamp Recipiency Has Increased Sharply in the Last 3 Years,
Following a Substantial Decline 6 Figure 2: National Payment Error Rate
for the Food Stamp Program, Fiscal Years 1999 to 2003 12 Figure 3: Map of
State Error Rate Changes from Fiscal Year 1999 to

2003 13 Figure 4: Map of State Error Rates for Fiscal Year 2003 15 Figure
5: Caseworker- and Participant-Caused Errors in Fiscal Year

2003 17 Figure 6: California Used a Combination of Internal Control
Practices to Reduce Payment Error 34

Abbreviations

EBT Electronic Benefits Transfer
FNS Food and Nutrition Service
QC quality control
TANF Temporary Assistance for Needy Families
USDA U.S. Department of Agriculture

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

May 5, 2005

The Honorable Saxby Chambliss
Chairman
The Honorable Tom Harkin
Ranking Minority Member
Committee on Agriculture, Nutrition, and Forestry
United States Senate

The Honorable Bob Goodlatte
Chairman
The Honorable Collin C. Peterson
Ranking Minority Member
Committee on Agriculture
House of Representatives

In fiscal year 2003, the federal Food Stamp Program, administered by the
U.S. Department of Agriculture's (USDA) Food and Nutrition Service
(FNS), reported it made payment errors totaling about $1.4 billion in
benefits. This sum represents about 7 percent of the total $21.4 billion
in
benefits provided each year to a monthly average of 21 million low-income
program participants. The program is intended to help low-income
individuals and families obtain a better diet by supplementing their
income
with benefits to purchase food. However, payment errors reflect the
misuse of public funds and may undermine public confidence in the
program.

The Food Stamp Program is jointly administered by FNS and the states.
State caseworkers must determine an applicant's eligibility and benefit
levels based on a complex formula that takes into account the members of
the household, their assets, and net monthly income; households must
report changes in their circumstances that may affect their eligibility
and
benefit levels; and caseworkers periodically recertify eligibility.
Depending
on household circumstances, some cases may require more adjustments
than others. For example, households with earned income may be
required to report more income changes to caseworkers than households
without earned income, such as those that are dependent solely on
retirement benefits. In addition, some food stamp participants receive
benefits from other programs, such as Medicaid or the cash assistance
program Temporary Assistance for Needy Families (TANF). Although the
caseworkers who process these eligibility determinations may be the same
as those who administer the Food Stamp Program, the rules for the

various programs can differ. These differences add to the complexity of
determining and recertifying eligibility and program benefits.

FNS's quality control (QC) system measures payment accuracy and monitors
how accurately states determine food stamp eligibility and calculate
benefits. Under FNS's QC system, states participate in the calculation of
their payment errors by reviewing a sample of cases to examine whether
eligibility was correctly determined and whether participating households
received the correct benefit amount. FNS validates the sample and the
accuracy of the state review.

Because the government must make the best use of funding, it is important
to minimize payment errors. Due to concerns about ensuring payment
accuracy, we examined (1) what is included in the national food stamp
payment error rate and how has the rate changed over time, (2) what is
known about the causes of food stamp payment errors, and (3) what actions
USDA and states have taken to reduce these payment errors.

To determine what is included in the payment error rate, how it has
changed over time, and the causes of payment error, we analyzed FNS's QC
data for fiscal years 1999 through 2003.1 We determined that the QC data
were reliable for the purposes of our work by reviewing our past reports,
FNS and external evaluations of the QC system, and related documents. We
also met with knowledgeable FNS officials to discuss issues of the QC
system's accuracy and completeness. To understand the causes of payment
errors and what actions have been taken to reduce them, we conducted
interviews with program stakeholders from FNS headquarters, each of FNS's
seven regional offices, and the USDA Office of the Inspector General. In
addition, we interviewed food stamp officials from 9 states, officials
from the state auditor's office in each of these 9 states, food stamp
officials from the local office within 8 of these 9 states with the
largest food stamp caseload, and food stamp researchers and
representatives from special interest groups. The 9 states we selected
were California, Michigan, Mississippi, New Jersey, New York, Oregon,
South Dakota, Texas, and Wisconsin. We chose these states for the
diversity in their locations, number of Food Stamp Program participants,
and payment accuracy performances. We included 3 states with consistently
low error rates, 3 states with consistently high error rates, and

1See appendix I for a detailed explanation of the methodology we used to
analyze FNS's data.

  Results in Brief

3 states that reduced their error rate by more than 30 percent between
1999 and 2003. To guide our work on actions taken to reduce payment
errors, we used the key components of internal control as our framework.2
Finally, to learn about past work regarding Food Stamp payment error, we
reviewed previous GAO reports on the Food Stamp Program and FNS reports
concerning food stamp payment error. We conducted our work between May
2004 and April 2005 in accordance with generally accepted government
auditing standards.

The national payment error rate for the Food Stamp Program combines
states' overpayments and underpayments to program participants and has
declined by almost one-third over the last 5 years to a record low of 6.63
percent in a time of rising caseloads. Of the total $1.4 billion of errors
in fiscal year 2003, 76 percent were due to overpayments and about 24
percent were underpayments. The payment error rate has fallen each year
since 1999, when it was 9.86 percent. This decline in the payment error
rate has been widespread; the rate fell in 42 states and the District of
Columbia, and rates in 18 of these states fell by at least one-third.
However, despite the decrease in many state error rates over the past few
years, a number of states continue to have difficulties reducing payment
error. For example, in 2003, 7 states had payment error rates of more than
10 percent. Finally, in addition to measuring improper payments to program
participants, FNS also monitors households that were refused benefits. In
fiscal year 2003, about 8 percent of these cases were improperly denied,
suspended, or terminated. However, these cases are not part of a state's
error rate, and the amount of benefits these households would have
received is unknown.

Food stamp payment errors are caused primarily by caseworkers, usually
when they fail to keep up with new information or make mistakes when
applying program rules, and by participants when they fail to report
needed information. These causes can be linked, in part, to how frequently
changes must be reported and the complexity of program rules. Almost
two-thirds of all payment errors occur when state food stamp caseworkers
fail to act on reported information or misapply complex rules in
calculating benefits. For example, the increase in the number of food

2See GAO, Internal Control: Standards for Internal Control in the Federal
Government, GAO/AIMD-00-21.3.1 (Washington D.C.: November 1999), and GAO,
Executive Guide: Strategies to Manage Improper Payments: Learning from
Public and Private Sector Organizations, GAO-02-69G (Washington D.C.:
October 2001), for more details.

stamp recipients who are low-wage workers and the changeable nature of
their income has made it more difficult for caseworkers to keep up with
changes, according to state officials. They also cited increased caseloads
and state fiscal problems, which resulted in staff reductions and
competing demands on workers, as contributing factors. In addition,
caseworkers may misapply the numerous eligibility requirements-such as
allowable deductions for shelter, utility, or child care-when calculating
a household's net monthly income. Moreover, state and local officials from
5 of the 9 states we contacted told us that it can be difficult for
caseworkers when they are responsible for multiple programs-such as TANF,
food stamps, and Medicaid-because the eligibility and reporting rules
among the programs often differ. Payment errors associated with
participants account for about one-third of all payment errors. This
generally is a result of participants not providing required information
to caseworkers, such as changes in household income and composition and
employment or of providing incomplete or incorrect information.
Participants may fail to provide this information either intentionally or
unintentionally. In 2003, states referred about 5 percent of all errors
for suspected participant fraud investigation. However, despite these
widespread challenges, states have continued to reduce their payment error
rates and remain concerned about continued improvement in the future.

FNS and the 9 states we reviewed have taken many approaches to increasing
food stamp payment accuracy, most of which are parallel with internal
control practices known to reduce improper payments. These approaches
include practices to improve accountability, conduct risk assessments,
implement program and process changes based on those assessments, and
monitor and promote improved performance. FNS's increased focus on the
error rate and the threat of increased financial penalties were cited by
several states as the impetus for state leaders and managers to make
payment accuracy a priority. Also, some states are holding their local
managers accountable for their error rates by setting overall local office
target rates or including target rates in the managers' contracts. FNS and
the states are also actively conducting risk assessments to identify the
types and sources of payment errors. For example, California, a state that
has reduced its error rate by over 50 percent since 2001, has increased
the number of cases sampled for its 19 largest counties as a way to assess
risk and identify the causes of errors at the county level. Once the
likely causes are identified, the states are adopting program and process
changes to address risk. For example, some localities have adopted
specialized units to respond to reported changes in case information to
address their failure to act on reported information errors. In addition,
most states we contacted have adopted a simplified

Background

reporting option, which is designed to reduce administrative burden and
promote higher participation. The option also helps reduce errors because
it reduces the frequency with which households must report changes. In
essence, unreported changes that might have caused errors in the past are
no longer required to be reported. Overall, states put into place a
combination of approaches based upon their available resources,
priorities, the nature of their errors, and other factors, making it
difficult to tie error rate improvements to specific practices. However,
state officials point to their improved state error rates as evidence that
collectively the practices are having an impact.

The federal Food Stamp Program is intended to help low-income individuals
and families obtain a more nutritious diet by supplementing their income
with benefits to purchase food. FNS pays the full cost of food stamp
benefits and shares the states' administrative costs-with FNS paying
approximately 50 percent-and is responsible for promulgating program
regulations and ensuring that state officials administer the program in
compliance with program rules.3 The states usually administer the program
out of local assistance offices that determine whether households meet the
program's eligibility requirements, calculate monthly benefits for
qualified households, and issue benefits to participants, almost always on
an Electronic Benefits Transfer (EBT) card. The local assistance offices
often administer other benefit programs as well, including TANF, Medicaid,
and child care assistance.

In fiscal year 2004, the Food Stamp Program issued almost $25 billion in
benefits, and in September 2004, almost 25 million individuals
participated in the program. As shown in figure 1, the increase in the
average monthly participation of food stamp recipients in 2004 continues a
recent upward trend in the number of people receiving benefits, with
caseloads increasing over 40 percent since 2001, but still below the level
in 1996.

3Following passage of the Personal Responsibility and Work Opportunity
Reconciliation Act of 1996, reimbursements for food stamp administrative
costs in 44 states are adjusted each year to subtract certain food stamp
administrative costs that have already been factored into these states'
TANF grants. As a result, these states receive less than 50 percent of
their administrative costs. See GAO, Food Stamp Program: States Face
Reduced Federal Reimbursement for Administrative Costs, RCED/AIMD-99-231
(Washington D.C.: July 23, 1999).

Figure 1: Food Stamp Recipiency Has Increased Sharply in the Last 3 Years,
Following a Substantial Decline

                 Average monthly participation (in millions) 30

Eligibility for participation in the Food Stamp Program is based on the
Department of Health and Human Services' poverty measures for households.
The caseworker must first determine the household's gross income, which
cannot exceed 130 percent of the poverty level for that year (or about
$1,654 per month for a family of three living in the contiguous United
States in 2003). Then the caseworker must determine the household's net
income, which cannot exceed 100 percent of the poverty level (or about
$1,272 per month for a family of three living in the contiguous United
States). Net income is determined by deducting from gross income expenses
such as dependent care costs, medical expenses, utilities costs, and
shelter expenses. In addition, there is a limit of $2,000 in household
assets, and basic program rules limit the value of vehicles an applicant
can own and still be eligible for the program.4 If the household owns a
vehicle worth more than $4,650, the excess value is included in
calculating the household's assets.5

4Households with elderly or disabled members are exempt from the gross
income limit and may have assets valued at $3,000.

5If a household has no other assets, its vehicle can be worth $6,650.

                                       25

                                       20

                                       15

                                      10 5

                                       0

1996 1997 1998 1999 2000 2001 2002 2003 2004 Fiscal year

Source: the Food and Nutrition Service.

                            Eligibility Requirements

After eligibility is established, households are certified to receive for
food stamps for periods ranging from 1 to 24 months depending upon
household circumstances. The average certification period is 10 months.
Once the certification period ends, households must reapply for benefits,
at which time eligibility and benefit levels are redetermined. Between
certification periods, households must report changes in their
circumstances-such as household composition, income, and certain
expenses-that food stamp agencies must consider to determine whether the
change affects their eligibility or benefit amounts. States have the
option of requiring food stamp participants to report on their financial
circumstances at various intervals and in various ways. States can
institute a type of periodic reporting system, or they can rely on
households to report changes in their household circumstances within 10
days of occurrence.6 Under periodic reporting, participants report
monthly, quarterly, or under a simplified system. The simplified reporting
system, available since early 2001, provides for an alternative reporting
option that requires households with earned income to report changes
between certifications only when their income rises above 130 percent of
the poverty level. This easing of program requirements was designed to
help increase the program access and participation of eligible working
families, an FNS goal, by making it easier for them to participate, as
well as to reduce the administrative burden on local food stamp offices.

                             Quality Control System

To ensure the accuracy of food stamp payments, FNS and the states have an
extensive quality control system. In fiscal year 2003, the states spent an
estimated $80 million to administer the system, and FNS spent and
estimated $9 million.7 According to FNS officials, each month a state's
food stamp QC staff selects a representative sample of the open food stamp
cases for review.8 The QC staff reviews each sample case to verify whether
the recipient's eligibility and benefit amount were determined correctly.
If the reviewer finds the benefit amount off by more than

6States can choose from a variety of change-reporting methods. They can
require households to report only when a member changes jobs, receives a
different rate of pay, or has a change in his or her work status, such as
from full-time to part-time or vice versa. States can also require
households to report only when there is a change in earnings of $100 or
more per month.

7FNS reimbursed the states half of the money they spend on QC. State QC
expenses include salaries for QC workers, the costs associated with
sampling and reviewing case files, office space, supplies, and travel.

8Sample sizes range from under 400 in smaller states to over 1,500 cases
in others.

$25, it is counted as an error. The statewide sample produces a valid
statewide error rate, although in most cases, it does not include
sufficient cases to generate error rates for local offices.9

FNS plays a significant role in monitoring and validating the state's
review. The FNS regional offices approve the states' sampling plans;
validate the states' samples, totaling 56,557 in fiscal year 2003; and
review one-third of these sample cases to ensure accuracy. They also
handle informal arbitration of disputes resulting from differences between
the state and FNS review outcomes. Disputes that are not resolved at the
regional office can be appealed to FNS headquarters for formal
arbitration. In fiscal year 2003, regional reviews found 151 cases where
the regional offices' finding or error amount was different from the
states' finding or error amount. According to FNS officials, this
constitutes less than 1 percent of the cases reviewed by the regions, and
each year between 20 and 30 of these unresolved disputes between the state
and the regional office are appealed to FNS headquarters for formal
arbitration. According to FNS officials, upon the completion of the
regional office's review and error disagreement processes, the regional
office adjusts error rates to reflect the final results.

Once the error rates are final, FNS is required to compare each state's
performance with the national error rate and imposes penalties or provides
incentives according to specifications in law. Prior to fiscal year 2003,
penalties were levied each year a state's payment error rate was above the
national average. In addition, states with error rates above 6 percent,
other than for good cause, were required to develop corrective action
plans that are monitored by the FNS regional offices. FNS can negotiate
with the states the amount of the penalty that will be paid to FNS, the
amount that will be reinvested into the program, and the amount of money
that will be collected if the state does not improve its error rate to an
agreed-upon amount. In order to encourage program improvement, FNS also
provided enhanced funding to states that with a payment error rate less
than or equal to 5.90 percent according to a formula set in law. During
this period of time, the states were held accountable only for their error
rate and no other performance measure.

9The state's error rate is determined by weighting the dollars paid in
error divided by the state's total issuance of food stamp benefits.

The Farm Security and Rural Investment Act of 2002 (the 2002 Farm Bill)
made significant changes to the way penalties and incentives are
calculated and awarded.10 States will not be penalized until their error
rate exceeds the national error rate threshold for 2 years in a row. The
error rate threshold changed so that states are not penalized unless there
is a 95 percent statistical probability that their error rate exceeds 105
percent of the national average for 2 consecutive years. If a state's
error rate exceeds the threshold for 2 years in a row, a penalty will be
established that is equal to 10 percent of the cost of errors above 6
percent.11 In addition to establishing the new penalty system, the 2002
Farm Bill instructed FNS to create new criteria for performance bonuses
that award states with high or most improved performance for actions taken
to correct errors, reduce error rates, improve eligibility determination,
and other indicators of effective program operations.

FNS and the states also conduct fraud prevention activities to detect and
prosecute food stamp fraud by retailers and participants. In fiscal year
2002, the states spent $229 million on their fraud control activities and
reported that they completed 834,000 client investigations resulting in
12,000 state prosecutions and 61,000 ineligibility rulings. As a result of
these fraud control activities and following up on overpayments identified
through the QC process and during regular case processing activities, the
states established almost $26 million in fraud claims, $176 million in
household error claims, and $59 million in agency error claims. States
also reported they collected $209 million on previously established
claims. FNS's payment error statistics do not account for the states'
results in recovering overpayments.

10The 2002 Farm Bill also gave states the option of adopting provisions
that could simplify program administration and possibly reduce error
rates. These options include simplifying income and resources, housing
costs and deductions, reporting requirements, and utility allowances. See
GAO, 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 2004).

11Of that amount, USDA may waive all or part, and/or require up to 50
percent to be reinvested in corrective action programs and/or require up
to 50 percent to be set aside for possible recovery in the third year. If
a state's error rate exceeds the threshold for 3 consecutive years, the
state is responsible for paying the second year at-risk amount, and USDA
will again require up to 50 percent of the liability amount to be
reinvested in corrective action programs and up to 50 percent be set aside
for possible recovery in the following year if the state again exceeds the
threshold for that year.

Payment errors can typically be traced to a lack of or a breakdown in
internal controls, which are an integral component of an organization's
management. Internal control is not one event, but a series of actions and
activities that occur throughout an organization on an ongoing basis.
Therefore, to guide our review of FNS and state actions taken to reduce
payment errors, we used the key components of internal control as our
framework. These components include creating a work environment that
promotes accountability and the reduction of payment error, analyzing
program operations to identify areas that present the risk of payment
error, making policy and program changes to address the identified risks,
and monitoring the results and communicating the lessons learned to
support further improvement.

  The Food Stamp Error Rate, Which Combines Overpayments and Underpayments, Has
  Declined by Almost One-Third over the Last 5 Years

The national Food Stamp Program payment error rate combines overpayments
and underpayments to participants, and has declined by about one-third in
recent years from 9.86 percent in 1999 to a record low of 6.63 percent in
2003.12 In dollars, this means if the 1999 error rate was in effect in
2003, the program would have made payment errors totaling over $2.1
billion rather than the $1.4 billion it experienced. Most states have
enjoyed a recent reduction in payment error, with error rates falling in
41 states and the District of Columbia. However, some states continue to
struggle with relatively high payment error rates. In addition to
measuring the accuracy of benefits paid, about 8 percent of the decisions
to deny, suspend, or terminate benefits were also made in error. However,
the amount of benefits these households would have received is unknown and
is not part of a state's payment error rate.

    Food Stamp Payment Error Rate Combines Benefit Overpayments and
    Underpayments

The national food stamp payment error rate combines overpayments and
underpayments made to benefit recipients in all states. Of the total $1.4
billion in payment error in fiscal year 2003, $1.1 billion, or about 76
percent, were overpayments, which represent a financial loss to the
federal government. Overpayments occur when eligible persons are provided
more than they are entitled to receive or when ineligible persons are
provided benefits. Underpayments, which occur when eligible persons are
paid less than they are entitled to receive, totaled $340 million, or
about 24 percent of dollars paid in error, in fiscal year 2003.

12In contrast, USDA's strategic plan for fiscal years 2000 to 2005 set a
target error rate of 9.2 percent by fiscal year 2005.

Underpayments represent unintentional financial savings to the federal
government.

Studies have reviewed the effects of payment errors on household income.
An analysis of fiscal year 2003 QC data conducted by Mathematica Policy
Research, Inc., for FNS found that typical overpaid eligible households
received an average of $97 too much in monthly benefits and underpaid
eligible households received an average of $78 too little in monthly
benefits.13 As a result, overpaid households' purchasing power, which
includes household gross income and food stamp benefits, rose by 8
percentage points, from 94 percent of the federal poverty level to 102
percent of the federal poverty level. Underpaid households' purchasing
power decreased by 6 percentage points from 80 percent of the federal
poverty level to 74 percent of the federal poverty level. More than 98
percent of households receiving food stamps were eligible for the program.
Ineligible households receiving food stamp benefits saw their purchasing
power rise from 118 percent of the federal poverty level to 132 percent of
the federal poverty level.

    The National Error Rate Declined by One-Third in the Last 5 Years, Driven by
    States Providing the Largest Amount of Food Stamp Benefits

The national Food Stamp Program payment error rate has declined by about
one-third over the last 5 years. The rate has declined each year, from
9.86 percent in 1999 to a record low of 6.63 percent in 2003, as shown in
figure 2. If the 1999 error rate had been in effect in 2003, the program
would have made payment errors totaling over $2.1 billion rather than the
$1.4 billion it experienced. In addition, the state-reported error rates
for fiscal year 2004 suggest that the overall error rate has continued to
decline. These error rates have not yet been validated by FNS, which
usually produces slight adjustments to these state-reported rates.

13Memorandum: Carole Trippe and Daisey Ewell, Size and Impact of Food
Stamp Payment Errors Based on FY 2003 FSPQC Unedited Database (Prepared by
Mathematica Policy Research, Inc., for the Food and Nutrition Service,
USDA, Alexandria, Va.: January 2005).

Figure 2: National Payment Error Rate for the Food Stamp Program, Fiscal
Years 1999 to 2003

Percent 12

9.86

10

8

6

4

2

0 1999 2000 2001 2002 2003

Fiscal year

Underpayments Overpayments Combined

Source: the Food and Nutrition Service.

Error rates fell in 41 states and the District of Columbia, and 18 states
reduced their error rates by one-third or more, as shown in figure 3. See
appendix II for more information on individual states' error rates over
time.

table 1. The changes in these states have a large effect on the national
error rate because of the way the rate is calculated.14

Table 1: Changes in Payment Error Rates for States Providing the Largest
Amount in Food Stamp Benefits, Fiscal Year 2003

                                                                   Percentage 
                                           1999       2003    change in error 
                         2003 benefit     error      error      rates between 
                 State     payments        rate       rate      1999 and 2003 
              New York  $1,676,508,940    10.47       5.88 
               Florida    987,926,276      9.43       8.00 
              Illinois   1,052,739,082    14.79       4.87 
                 Texas   1,880,851,630     4.56       3.29 
            California  $1,807,987,279    11.34       7.96 

Source: GAO analysis of FNS data.

In addition to contributing to the downward trend in the payment error
rate, an increasing number of states had error rates below 6 percent in
2003.15 However, payment error rates vary among states. For example, 21
states had error rates below 6 percent in 2003 (see fig. 4 for states'
error rate performance); this is an improvement from 1999, when 7 states
had error rates below 6 percent. Despite the decrease in many states'
error rates over the past few years, some states continue to have high
payment error rates. For example, 7 states had payment error rates of 10
percent or higher in 2003. These states are also making progress, however,
and are expected to have reduced their error rates in 2004.

14The national food stamp payment error rate is the average of the states'
food stamp payment error rates weighted by each state's proportion of all
food stamp benefits issued during the fiscal year. Therefore, a state
issuing a higher proportion of the food stamp benefits plays a larger part
in determining the national food stamp payment error rate compared with a
state issuing a smaller proportion of the food stamp benefits.

15FNS does not require states with payment error rates under 6 percent to
develop and implement corrective action plans to reduce payment errors.

  Caseworkers Cause about Two-Thirds and Participants Cause about One-Third of
  Payment Errors

error. In fiscal year 2003, FNS reported that about 8 percent of the
decisions to deny, suspend, or terminate benefits were made in error.
However, the amount of benefits these households would have received had
this error not occurred is unknown.

Almost two-thirds of the payment errors in the Food Stamp Program are
caused by caseworkers, usually when they fail to act on new information or
make mistakes when applying program rules, and one-third are caused by
participants, when they unintentionally or intentionally do not report
needed information or provide incomplete or incorrect information (see
fig. 5). Program complexity and other factors, such as the lack of
resources and staff turnover, can contribute to caseworker mistakes.
Despite the decrease in error rate in recent years, these factors have
remained the key causes of payment error over the last 5 years.

Figure 5: Caseworker-and Participant-Caused Errors in Fiscal Year 2003

4%

Incomplete/incorrect information provided

Incomplete/incorrect information provided and case referred for potential
fraud investigation

Failure to act on reported information Information not reported

Policy incorrectly applied

Failure to follow up on or verify information

Other

                         Participant caused error (35%)

Caseworker caused error (65%)

Source: the Food and Nutrition Service and GAO analysis.

    Caseworkers Failing to Act on Reported Information and Misapplying Program
    Rules Cause Most Caseworker Errors

Caseworkers Fail to Act on Changes

Almost two-thirds of all payment errors are made by state food stamp
caseworkers, according to our analysis of FNS QC data.16

Errors can occur when caseworkers have difficulty keeping up with reported
changes in household circumstances, according to officials from all of the
states we reviewed. Caseworkers are required to review reported changes
and assess their effect on a household's eligibility and benefit levels.
In addition, caseworkers regularly receive information from data matches
and other sources that should be assessed and verified, and the failure to
do so is another important cause of error. In previous work, we have found
that the risk of improper payments increases in programs with a
significant volume of transactions. When caseworkers fail to keep up with
changes, the errors usually are reflected as incorrect household income or
deductible expenses, as shown in table 2.

Table 2: Caseworker Errors Most Often Resulted in Incorrect Household Income or
                          Deductions, Fiscal Year 2003

                       Percentage         Percentage of Percentage Percentage 
                           of             Percentage of of                 of 
         Reason          income   deduction errors        other         total 
                         errors   nonfinancial errorsa    errors       errors 
    Failure to act on                                              
        reported                                                   
       information          10.36             8.86 4.08        .03      23.33 
Policy incorrectly        8.35             6.99 3.99        .81      20.14 
         applied                                                   
    Failure to verify                                              
     information or          6.81             4.65 1.54        .03      13.03 
        follow up                                                  
Other agency error        5.22               2.72 .4        .66 
          Total                                                    
    caseworker-caused                                              
          error             30.74           23.22 10.01       1.53      65.5b 

Source: GAO analysis of FNS data.

aNonfinancial errors refer to factors considered in determining a
household's eligibility, such as household composition, citizenship, and
student status of household members.

bThe caseworker errors in table 2 and participant errors in table 3
account for all errors. The two totals may not add to 100 percent because
of rounding.

16The percent of all errors due to caseworker mistakes is 65.5 percent.
The margin of error associated with this estimate is plus or minus 1
percent at the 95 percent level of confidence. GAO's analysis results
could differ slightly from those reported by FNS because of small
variations in the databases. FNS utilized the raw QC database for its
analysis while GAO's database omitted some cases with incomplete case
data.

Caseworkers Incorrectly Apply Program Rules

Food stamp officials in 8 of the 9 states told us that increasing
caseloads have contributed to payment errors, making it more difficult for
caseworkers to attend to all of the reported changes. In recent years, FNS
and several states have made it a priority to reach out to likely eligible
households that are not yet participating in the program, in addition to
focusing on minimizing payment error. At the same time, the nation
experienced an economic downturn, which contributed to an increase in the
number of families who had a need for food assistance. As a result of
these and other factors, nationally, the number of food stamp participants
has increased by more than 30 percent since February of 2001.

Moreover, as states across the country have faced fiscal challenges due to
the overall slowdown in the economy, some responded by reducing their
staff, offering early retirements, or imposing hiring freezes. This also
has contributed to rising caseloads per worker. For example, food stamp
officials in Michigan said state fiscal problems resulting in staff
reductions, increased caseloads per worker, and competing demands on
workers made it difficult for caseworkers to act on all reported changes
because of high caseloads. Oregon state officials also attribute their
difficulties with payment accuracy to a 40 percent increase in the number
of food stamp cases in the state between 2001 and 2003 as well as state
financial problems that led to staff cuts and a hiring freeze. FNS
officials informed us that there is no central collection of comparable
data on caseload per worker among states.

Further, the recent outreach efforts included a focus on increasing
participation among working families. State and local officials from 8 of
the 9 states we interviewed said managing cases with earnings contributes
to payment error in part because caseworkers may find it difficult to keep
up with the frequent changes reported to them.17 For example, Michigan
food stamp officials told us that they experienced an increase in
overpayment errors because caseworkers were failing to act on the frequent
wage and salary changes reported by working participants.

The complexity of the eligibility criteria for the Food Stamp Program
contributes to caseworker errors. In previous work, we found that the risk
of improper payments increases in programs with complex criteria for

17GAO previously reported that FNS and some states and localities have
taken several steps to help working families participate in the program.
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.: March 2004) for more details.

computing eligibility and payments.18 Caseworkers may miscalculate a
household's eligibility and benefits, in part because of the program's
complex rules for determining eligible household members and for
calculating the household's financial status. Our analysis of QC data
found that caseworker mistakes often involve incorrectly determining
household income, followed by mistakes related to income deductions, and
nonfinancial issues, such as determining household composition. Although
the error rate has declined in recent years, these three types of mistakes
have remained the major sources of error over the last 5 years.

To determine household gross income, caseworkers must decide which types
of income to include. Households may have income from a number of
different sources, and rules require that some of this income be counted
and some not. Further, the fluctuations in earnings for low-income working
participants can increase the likelihood of error simply because they
result in a higher volume of case reviews and adjustments.

Payment errors also occur when caseworkers misapply one or more of six
allowable deductions when determining net income. Caseworkers calculate
and deduct expenses such as dependent care costs, medical expenses,
utilities costs, and shelter expenses-each of which have their own set of
eligibility criteria.19 For example, caseworkers can provide households an
excess shelter expense deduction if their shelter expenses exceed 50
percent of monthly household income after applying other deductions. As
part of that process, caseworkers must determine whether the household is
entitled to a standard utility allowance.20

Other common caseworker errors involve nonfinancial factors, such as
misapplying the program's complex rules for determining the members of the
household. Although individuals may be living in the same home, they may
be treated as different households for eligibility and benefit purposes,
depending on whether they customarily purchase food and prepare meals

18See GAO, Executive Guide: Strategies to Manage Improper Payments:
Learning from Public and Private Sector Organizations, GAO-02-69G
(Washington D.C.: October 2001).

19The six allowable deductions are a standard deduction, an earned income
deduction, a dependent care deduction, a medical deduction, a child
support deduction, and an excess shelter cost deduction.

20Not all households are eligible for the standard utility allowance.
Exceptions include households sharing a living space with others and not
eligible for the full value of the allowance and public housing residents
who were charged only for excess utility costs.

together. However, this is sometimes difficult to determine. Food stamp
officials in Michigan told us that given the variety of household
circumstances and arrangements caseworkers face, determining household
composition can be confusing. For instance, officials said it can be
difficult to determine how to treat a youth over age 22 who moves in and
out of the parents' home or households that contain multiple generations
of family members. In addition, officials from 5 of the 9 states we
contacted told us that having caseloads with legal noncitizens was a
challenge to reducing payment error, in part because of the numerous
policy changes in recent years that affect the eligibility of various
segments of this population.21

Correctly determining food stamp eligibility and benefits can be
complicated by differences between Food Stamp Program rules and the rules
governing other assistance programs. Officials from 5 of the 9 states we
interviewed told us that minimizing payment error is difficult for
caseworkers when they are responsible for multiple programs, such as TANF,
food stamps, and Medicaid, because the eligibility and reporting rules
among the programs often differ.22 For example, local officials from Texas
told us that because of the way the state chose to implement the
simplified reporting option, caseworkers are held responsible for failing
to act on a change when a birth is reported to the Medicaid program, even
though participants are not required to report the change to the Food
Stamp Program, according to a recently approved policy option. Oregon
state and local officials also told us that it is challenging for
caseworkers to attend to food stamp payment accuracy when they have to
determine eligibility and recertify households for other assistance
programs.

21The Personal Responsibility and Work Opportunity Reconciliation Act of
1996 (P.L. 104-193, Aug. 22, 1996) tightened food stamp eligibility
requirements, in part, by disqualifying most permanent resident aliens.
Subsequently, the Agricultural Research, Extension, and Education Reform
Act of 1998 (P.L. 105-185) restored eligibility to permanent resident
aliens who were lawfully residing in the United States as of August 22,
1996, and (1) were age 65 or older at that time or (2) are either disabled
or under age 18. The 2002 Farm Bill also partially restored food stamp
eligibility on certain dates to qualified aliens who are otherwise
eligible and meet criteria laid out in the legislation.

22GAO previously reported that changes in food stamp reporting rules aimed
to reduce program complexity and payment error introduced complications
for participants and caseworkers because the rules were not consistent
with the reporting rules of other assistance programs. See GAO, 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 2004) for more details.

Officials from all 9 of the states we interviewed stated that staff
turnover contributes to incorrect application of program rules. Food stamp
officials in Oregon said that half of the caseworkers in the Portland area
have less than 1 year of work experience because of high staff turnover,
which makes it difficult for the office to maintain a workforce trained in
making accurate eligibility decisions. Officials also told us that lack of
training can be a challenge in part because it is difficult for
caseworkers to learn the complex program rules and policies.

Similar factors also affect errors where benefits are improperly denied,
suspended, or terminated, according to officials from states we
interviewed. They cited caseworkers misapplying policies or miscalculating
income. For example, Michigan food stamp officials told us that these
errors sometimes occur when caseworkers temporarily suspend benefits
because participants are not complying with certain rules but then do not
review the case to complete it correctly. Mississippi officials told us
that these errors can also occur when caseworkers misapply a policy or
fail to add up wages correctly.

Participant Error Involves About 35 percent of all payment errors occur
because participants do not Failure to Report provide required, complete,
or correct information to caseworkers, either Required, Complete, or
unintentionally or deliberately (see table 3).23 Although applicants are
Correct Information to required to provide a variety of personal
information to the caseworker, Caseworkers failure to report income is the
most common cause of participant food

stamp errors.

23The percentage of all errors attributable to participant reporting is
34.5 percent. The margin of error associated with this estimate is plus or
minus 1 percent at the 95 percent level of confidence.

 Table 3: Participant-Caused Errors Most Often Resulted in Incorrect Household
                    Income Determinations, Fiscal Year 2003

                     Percentage Percentage   Percentage Percentage Percentage 
                         of             of           of of                 of 
        Reason       income     deduction  nonfinancial   other         total 
                     errors     errors     errors         errors       errors 
Information not        17.46       4.30         3.75        .13      25.64 
       reported                                                    
    Incomplete or                                                  
      incorrect                                                    
     information           1.63       1.76          .57            
       reported                                                    
    Incomplete or                                                  
      incorrect                                                    
     information                                                   
reported & case                                                 
     referred for          2.96       1.13          .83            
suspected fraud                                                 
        Total                                                      
  participant-caused                                               
        error             22.05       7.19         5.15        .13     34.52a 

Source: GAO analysis of FNS data.

aThe caseworker errors in table 2 and participant errors in table 3
account for all errors. The two totals may not add to 100 percent because
of rounding.

Program complexity may play a role in participants' failure to report
needed information because the participants may not understand the
reporting requirements, according to officials from 2 states we
interviewed. For example, California state food stamp officials told us
they believe that some participants do not report information because they
are unfamiliar with the reporting requirements or because of language
barriers. In addition, when participants receive assistance from multiple
programs, they may be confused about what to report to whom because the
requirements differ among the programs, including those for Medicaid and
TANF.24 When participants fail to report information, the result is
usually an incorrect determination of household income. Further,
participants may not report information to caseworkers because of the
perceived burden associated with reporting changes. For example, a food
stamp official in Wisconsin told us that because of the lack of staff at
the call center, participants calling to report changes may wait on the
line for up to 20 minutes, and as a result, some participants will hang
up.

Errors may also occur when the participant intentionally does not report
needed information or unintentionally or intentionally provides the
caseworker with false or incomplete information. Although the percentage
of payment errors that involve participants intentionally withholding
information is not known, food stamp workers from all of the states we

24See GAO, 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 2004) for more
details.

interviewed refer cases for investigation when they suspect fraud. For
example, Oregon food stamp officials explained that cases are referred for
suspected fraud when a participant consistently reports no income yet
seems to have the resources needed to live self-sufficiently. In 2003,
about 5 percent of all payment errors were referred for fraud
investigation. Data are not available, however, to determine what
percentage of these error cases resulted in disqualifying participants
because of fraud.

Despite the recent decrease in error rates, the program continues to face
these same causes of error over time. Over the last 5 years, caseworker
failure to act on reported information, caseworker misapplying program
policies and requirements, and participant failure to report key
information have remained the three largest causes of error. Moreover,
errors involving incorrect household income or deductions for expenses
continue to be the most common types of errors over the same period.

FNS and the states we reviewed have taken many approaches to increasing
food stamp payment accuracy, most of which are parallel with internal
control practices known to reduce improper payments.25 These include
practices to improve accountability, perform risk assessments, implement
changes based on such assessments, and monitor program performance. Often,
several practices are tried simultaneously, making it difficult to
determine which have been the most effective.

  FNS and States Have Taken Steps to Increase Payment Accuracy

    FNS's and States' Approaches Are Parallel with Good Internal Control
    Practices

Improving Accountability

Because payment errors can typically be traced to problems with internal
controls, we used the key components of internal control as our framework
to categorize the approaches taken to reduce payment errors. In doing so,
we found that both FNS and the states we reviewed were employing many of
the same practices recognized as being effective in reducing payment
errors.

Both FNS and states have taken steps to ensure that program officials
recognize their responsibility for payment accuracy. FNS has long focused
its attention on states' accountability for error rates through its QC
system by assessing penalties and providing financial incentives. The
administration of the QC process and its system of performance bonuses and
sanctions is credited or faulted by many as being the single largest
motivator of program behavior, and most of the states in our review

25See GAO-02-69G.

believe the QC system has helped increase payment accuracy. From fiscal
year 1998 to fiscal year 2002, FNS has assessed $327 million in penalties.
Of these penalties, FNS waived $93 million, approved $92 million for
reinvestment into state food stamp programs, collected almost $24 million,
and designated $118 million at risk for payment if the states did not
improve their error rates to agreed-upon targets. During this same period,
FNS awarded states almost $251 million of enhanced funding because of
their low error rates.

In fiscal year 2003, the first year under the 2002 Farm Bill changes to
the QC system, 11 states were found to be in jeopardy of being penalized
if their fiscal year 2004 error rates did not improve. This was a higher
number than was originally expected by some analysts because the error
rate had fallen much faster than in previous years, leaving more states
above the new error rate threshold. Some states have expressed concern
that they may improve their error rates and yet still be penalized because
the national rate continues to drop around them. In addition, under its
new performance bonus system, FNS awarded a total of $48 million to
states, including $24 million to states with the lowest and most improved
error rates and $6 million to states with the lowest and most improved
negative error rate.26

In addition to using the tools available under its QC system, FNS's
leadership has actively communicated the importance of accountability.
Establishing payment accuracy as a program priority is considered by many
to be the most important strategy for achieving program improvement. Since
the arrival of the current Undersecretary for Food, Nutrition, and
Consumer Services in 2001, FNS has put increased pressure on states to
reduce error rates. For example, the undersecretary and other FNS
officials visited states with particularly high error rates to discuss
payment accuracy. FNS also began to collect a higher percentage of
penalties. From fiscal year 1992 to 2000, FNS collected about $800,000 in
penalties. Since fiscal year 2000, FNS has collected more than $20 million
in penalties. Officials from one advocacy group active in food stamp
issues credits this official's active role as one reason for the drop in
the error rates in the larger states. The FNS regional administrators also
visit high

26The remaining $18 million was awarded for improvements not related to
error rates-the highest and most improved ratio of food stamp participants
compared with the number of persons in poverty and the highest percentage
of timely completed applications.

error rate states and emphasize payment accuracy as a major management
priority at regional meetings of state commissioners.

All the states we reviewed also reported taking steps to increase the
awareness of, and the accountability for, errors in their programs. Often,
this coincided with a change in state leadership and responded to
accumulating program penalties, bad publicity, or both. For example,

o  	Michigan state officials said that after their new governor took
office in 2003, error reduction became an issue for the governor and the
legislature because the state had paid more than $5 million in penalties
in 2003 and 2004. In response, the Food Stamp Program began producing
weekly internal reports and issuing regular reports to the governor and
the legislature. The state's error rate has dropped from 14.1 percent in
fiscal year 2002 to a state-reported error rate of 6.73 percent in fiscal
year 2004. As a result of the state's progress in reducing its error rate,
the governor has publicly recognized the program's efforts.

o  	Wisconsin's turnaround began in 2002 when state officials, with the
support of the governor, made it clear to local food stamp offices that
double-digit error rates and the penalties that go along with them were no
longer acceptable. Wisconsin had been assessed penalties totaling over $8
million for 2000, 2001, and 2002. The state's error rate has dropped from
13.14 percent in fiscal year 2001 to a state-reported error rate of 6.57
percent in fiscal year 2004.

o  	Penalties totaling over $5 million for 1998, 1999, and 2000, also
spurred New Jersey's human services director to appoint a special
assistant to focus on reducing the state's error rate. The state's error
rate has dropped from 12.93 percent in fiscal year 1999 to a
state-reported error rate of 2.62 percent in fiscal year 2004.

In addition, states we reviewed understood the need to communicate the
importance of payment accuracy to individuals working at all levels of the
program. Of the states we studied, California, Michigan, New Jersey, and
Oregon have begun to set error rate targets for their local offices and
have supplemental quality assurance processes in place to produce local
error rates or error rates for their largest offices. Oregon and Texas
also include payment accuracy goals in the expectations for their managers
and workers, making payment accuracy one of the bases for their
evaluations. California, New York, and Wisconsin have shared the
accountability for poor performance by passing on a portion of their
state's financial

                               Identifying Risks

penalties to their largest counties. New Jersey, South Dakota, and Texas,
on the other hand, have shared the enhanced funding they have received for
good performance with their local food stamp offices.

Both FNS and states have taken steps to analyze program operations to
identify where risks exist. For example, through its QC system, FNS
determined that working families receiving benefits were error prone
because of frequent changes in their income and deductions. In addition,
officials from our 9 review states said they analyze the QC data to
identify the sources and causes of food stamp payment error in their
states. New Jersey officials used the QC data to identify salaries and
wages as the largest sources of error in their state. In most cases,
however, the QC samples are not large enough to produce valid error rates
or to identify specific problem areas for most counties or local offices.
In order to be able to obtain this information, California, Michigan, New
Jersey, New York, and Oregon have developed their own quality assurance
systems to produce monthly error rates for their counties or local
offices. For example, in January 2003, Oregon instituted a targeted case
review process that requires officials in local offices to review between
35 and 100 cases per month to identify errors. State officials say the
reviews provide better information to local-level officials on the causes
and sources of payment error at their site so they can plan corrective
action. Oregon's payment error rate dropped from 13 percent in fiscal year
2003 to a state-reported error rate of 7.81 percent in fiscal year 2004.

California, New York, Wisconsin, and Michigan targeted their largest and
most error-prone offices for special risk assessments. In Wisconsin, for
example, the state focused its approaches on Milwaukee because it is the
largest metropolitan area in the state, accounting for 47 percent of the
state food stamp caseload. Because it had the highest error rate, it had
the most significant influence on the state's error rate. The state
brought in a contractor that conducted an assessment of payment accuracy
and the service delivery model used in Milwaukee. The contractor
recommended that Milwaukee adopt a number of policy, program, and case
review changes. In response, Wisconsin and the city of Milwaukee conducted
a one-time find-and-fix case sweep between March and September 2004. State
and county case readers reviewed 14,000, or almost 25 percent of , their
food stamp cases to identify and correct potential errors. The information
gained from this exercise identified certain risks and errorprone cases
that county officials have used to implement other changes. As a result,
Milwaukee County officials said their error rate dropped from 12.2 percent
in March 2004 to 7.7 percent in June 2004.

Responding to Findings from Risk Assessments

Once the QC review process is completed, penalties are assessed by law to
high error rate states and FNS works with the states to correct the
problems. Staff from the FNS regional offices work with the states on the
development and implementation of reinvestment and corrective action plans
that address specific threats and risks identified in risk assessments.
These plans can vary depending upon the state's systems and
characteristics. Examples of activities included in the plans include
training to address errors identified from QC and quality assurance
reviews, developing online training curricula, and correcting errors
generated by automated systems.

States have also adopted practices to prevent, minimize, and address
payment accuracy problems in response to the sources of error identified
in risk assessments. States chose their varied practices in response to
their unique characteristics, resources, and risks.

o  	Automated system changes. Michigan implemented changes in its
automated system to help deal with problems resulting from failure to
collect complete case information, particularly household income, during
the application and recertification processes. The state's automated
system now prompts workers to obtain complete income documentation for
cases with earned income

o  	Specialized change units. In June 2002, Los Angeles established 30
specialized change units for its 30 district offices to address their
failure to act on reported information, which was one of their largest
sources of errors. FNS supports the adoption of change centers such as
these based upon their reported outcomes in other states. Los Angeles
County officials said the change unit workers now act upon reported case
changes that previously had not been acted upon by caseworkers because of
their large caseloads.

o  	Outreach to more stable food stamp population. New York has
implemented a program to automatically certify eligible nonparticipating
elderly Supplemental Security Income recipients for food stamps for 4
years.27 In addition to reaching an underserved population without adding
undue administrative burden on the local offices, officials believe that
increasing the participation of these recipients could help reduce the

27The Supplemental Security Income Program is designed to provide aged,
blind, and disabled people who have little or no income with cash to meet
basic needs for food, clothing, and shelter.

state's error rate because this group is less error prone because of its
stable income and circumstances.

States also adopted various case review practices that would help them
address a wide range of risks and problems.

o  	Supervisory review of cases. Several states have begun to require
local supervisory reviews of cases to detect and correct errors caused by
misapplication of food stamp policies or workers failing to act on
reported information. Some states require that all cases be reviewed,
while others target error-prone cases or a certain number of cases per
worker.

o  	Targeted local office reviews. Some states have used contractors or
have established their own teams to target high error rate offices for
improvement. Michigan recently started using technical assistance teams to
observe the local office's processes and make recommendations for
improvement.

o  	Error review panels. Some of our review states have also established
panels to review errors discovered through the QC process. New Jersey
established such a panel, consisting of system, policy and QC staff. This
panel reviews all errors, challenges some that it believes have been
inaccurately classified and develops corrective actions to address the
root causes of the errors. The results of the reviews can then be
communicated to all local offices. For example, as a result the panel's
finding that computing utility bill deductions was a source of payment
errors, the state implemented a mandatory standard utility allowance
policy to reduce this type of error.

Many of the error reduction practices employed by the states in our review
focused primarily on agency-caused rather than client-caused errors. Many
state officials we spoke with believe that states should not be held
accountable for participant-caused errors, such as failure to report
information, because the state cannot control participants' behavior.
However, FNS officials believe that states can reduce participant-caused
errors by better using computer matching of state data sources and other
outside sources of data, improving interviewing techniques to collect all
relevant information and identify discrepancies, and educating clients
about their responsibilities.

In addition to taking the above steps focused specifically on decreasing
the error rate, FNS has made and advocated for a number of program and
policy changes designed primarily to address other issues, such as

program participation, which have also helped reduce payment errors. FNS
believes that serving eligible low-income families, particularly working
poor families, is imperative to the success of welfare reform and the
nutritional well-being of eligible persons. However, because the income
and deductions for working poor families tend to be volatile, these
households are more error prone, and their participation could increase
the error rates of states trying hardest to serve them and thus discourage
states from reaching out to these families. In response, FNS raised the
error tolerance level in fiscal year 2000 from $5 to $25 for monthly food
stamp payments for all cases. This change exempted smaller errors that had
been counted in the past. FNS estimated that this change would have
reduced the nationwide error rate by 0.66 percentage points if it had been
implemented in the previous fiscal year.

In addition, FNS and Congress have made several options available to the
states to simplify the application and reporting process. These
simplification measures are designed, in part, to reduce the
administrative burden on both caseworkers and participants and thus
promote higher participation in the program. One option in particular
reduces the frequency with which households with earned income must report
changes. Prior to this simplified reporting option, participants were
required to frequently report changes in their circumstances. Under the
simplified reporting rule issued in November 2000, most households need
only report changes between certification periods if their new household
income exceeds 130 percent of the federal poverty level.

This simplified reporting option can reduce a state's error rate as well.
Absent simplified reporting, certain unreported or undetected changes
between certification periods would be considered an error. Minimizing the
number of income changes that must be reported between certifications can
help reduce errors associated with caseworker failure to act as well as
participant failure to report changes, and income-related errors account
for more than half of all payment errors.

Essentially, this simplification option redefines the threshold for what
is considered an error. This type of change can result in an increase in
program benefits paid out, such as when participants experience an
increase in income between certification periods that need not be reported
until the next certification under the simplified requirements. In 2000,
FNS estimated the additional cost to the program to be approximately $51
million in fiscal year 2004 affecting nearly 1.5 million households per
month. By expanding this option in the 2002 Farm Bill beyond earned income
households to any and all households that can be asked to report

Monitoring and Promoting Performance

periodically, an FNS official said Congress had endorsed the idea of
making the program more user friendly to working families. Since the 2000
estimate, program participation has grown significantly, but FNS has not
completed a more recent estimate of the additional cost. Moreover, the
possible savings and efficiencies gained in program administration have
not been quantified.

Most of our review states have adopted some form of simplified reporting
to help them better serve working families, permit greater program
participation, and address the errors associated with frequent change
reporting. Nationwide, FNS reported that as of September 2004, 41 states
and the Virgin Islands had adopted some form of simplified reporting.

FNS has taken many actions to track the success of improvement initiatives
and to provide the information needed to facilitate program improvement.
FNS managers use data generated from the QC system as well as the results
of their own monitoring activities to track the states' performance over
time. FNS regional offices annually review state agency operations to,
among other things, confirm that problems in program operations are being
identified, properly analyzed, and resolved. Where applicable, the
regional office also monitors the states' implementation of corrective
action plans. FNS, in turn, requires states to perform management
evaluations to monitor whether adequate corrective action plans are in
place at local offices to address the causes of persistent errors and
deficiencies. To monitor corrective actions identified through the
management evaluations, FNS suggests that states review a sample of case
records containing actions that are error prone.

In addition, in November of 2003, FNS created a Payment Accuracy Branch at
the national level to work with FNS regions to suggest policy and program
changes and to monitor state performance. The branch facilitates a
National Payment Accuracy Workgroup with representatives from each FNS
regional office and headquarters who use QC data to review and categorize
state performance into one of three tiers.28 FNS has recommended a
specific level of increasing intervention and monitoring approaches for
each tier as error rates increase, and the FNS regional

28Tier 1 states have an error rate under 6 percent, and tier 2 states have
an error rate of 6 percent or greater but do not fall into tier 3. States
are assigned to tier 3 when the lower limit of their error rate estimate
at the 90 percent confidence level is higher than 105 percent of the
national error rate estimate.

offices report to headquarters on both state actions and regional
interventions quarterly.

FNS also provides and facilitates the exchange of information gleaned from
monitoring by

o  	publishing a periodic guide to highlight the practices states are
using to address specific problems;29

o  	sponsoring national and regional conferences and best practices
seminars;

o  training state QC staff;

o  	providing state policy training and policy interpretation and
guidance; and

o  supporting adoption of program simplification options.

Once promising state practices have been identified, FNS also provides
funding to state and local food stamp officials to promote knowledge
sharing of good practices. Oregon officials said FNS provided state
exchange funds for them to visit Kentucky, Indiana, and Arizona-three
states that had effective systems for monitoring performance at the local
management and worker level. FNS also provided state exchange funds for
Oregon officials to meet several times with officials from Idaho and
Alaska to discuss common problems they faced trying to reduce payment
errors and to generate solutions. In fiscal year 2004, FNS provided
$612,000 for states to conduct state exchange visits. Officials from most
of our review states found this program to be particularly helpful to
their efforts to improve program performance.

States are also using information generated by the QC system to track the
results of their policy and program changes over time and communicate
timely operational information to local offices. Information gleaned from
monitoring can help inform their ongoing risk assessments. States are also
promoting knowledge sharing of promising practices. These practices
include

o  	preparing reports detailing causes and sources of errors for the local
offices and publishing and distributing monthly error rates for all local
offices;

29U.S. Department of Agriculture, Food and Nutrition Service, Payment
Accuracy in the Food Stamp Program (Alexandria, Va.: September 2004).

o  	transmitting the results of statewide error review panels on the
source and causes of errors to local offices, along with suggested
corrective actions;

o  	sponsoring statewide QC meetings and state best practices conferences
for local offices to discuss error rate actions taken and common problems;
and

o  sponsoring local office participation in FNS regional conferences.

Despite FNS and state mechanisms used to track the initiatives and share
promising practices, there are no data available on which initiatives are
most cost-effective. FNS's primary focus has been on monitoring progress
in reducing error rates, which can help ensure eligible households receive
the correct benefits and maintain public support for the program. Even so,
from fiscal year 2001 to 2004, the annual administrative cost per
participant has fallen from $129 to $99 per participant while program
participation has increased. It is possible that some states gained
efficiencies from simplified reporting. However, FNS has not studied the
cost-effectiveness of this or other measures and thus cannot share this
type of information with the states.

    States Are Using a Combination of Approaches to Address Payment Errors,
    Making It Difficult to Determine the Effectiveness of Specific Practices

Every state we surveyed has put into place a combination of approaches to
address the key components of internal control, and the practices states
adopted under each approach varied among them. For example, in California,
state and local officials employed a combination of practices under each
internal control component over the last several years to bring about
their improved error rate (see fig. 6).

Figure 6: California Used a Combination of Internal Control Practices to
Reduce Payment Error

Source: GAO analysis.

Because many states have adopted multiple error reduction practices,
officials we spoke with said it is difficult to isolate the results of
individual practices, particularly when other program and economic changes
are occurring simultaneously. State officials point to their low or
dropping error rates as evidence that, collectively, their new practices
are having a positive impact. However, they have little data to determine
which practices have been most successful or cost-effective.

Despite the lack of data, state officials citied various practices that
they believe have worked well in their state. For example, officials in
Michigan and New York believe new automated processes are their most
effective practices. Michigan food stamp officials cited targeted local
office reviews as another effective strategy for error reduction.
Mississippi food stamp officials believe their required supervisory review
of cases has been the most effective practice. California, South Dakota,
and Texas also cited supervisory reviews as one of their most effective
practices.

As a result of unique circumstances in each state, some practices that may
prove effective in one state would not be effective or feasible in
another. For example, New Jersey food stamp officials credit their 2001
implementation of the simplified reporting option for earned income cases
with being the most significant reason for the decline in their error
rates. However, officials in South Dakota continue to require monthly
reporting because they have been able to keep up with the reported
changes. They believe this requirement is primarily responsible for its
error rate, which is the lowest in the nation. Monthly reporting requires
participants to report, and caseworkers to act, on case changes once per
month, rather than relying on participants to report key changes and
workers to react to the reported change. Monthly reporting requires
significantly more work for both the caseworkers and participants, and
other states with larger caseloads have said they do not have adequate
resources to sustain this more labor-intensive approach.

The success of new practices, however, can be undermined if the changes do
not receive adequate management attention or are not effectively
implemented. For example, Los Angeles established 30 specialized change
units. County officials said these units helped reduce one of their
largest sources of errors, caseworkers' failure to act. On the other hand,
Milwaukee's change units have not been as effective in reducing the error
rate as officials hoped because they have not been able to staff the
center appropriately, according to county officials. They designed their
change units on a model implemented in Atlanta, Georgia. The Atlanta model
calls for 10 staff per 10,000 calls, and Milwaukee has about 7 staff per

20,000 calls. As a result, clients wait on the phone for up to 20 minutes,
and some hang up before their changes can be reported.

Similarly, Wisconsin state and Milwaukee food stamp officials said their
find-and-fix case sweep program conducted between March and September 2004
was a particularly effective practice for reducing payment errors.
Milwaukee officials believe the case sweep was largely responsible for
their error rate dropping from 12.2 percent in March 2004 to 7.7 percent
in June 2004, and they expect to see long-term effects as a result of
their workers learning from the errors identified using this practice.
However, Michigan tried a similar program but did not have comparable
results. State officials said using this method did not reduce their error
rate because the state and counties did not have enough staff to conduct a
sufficient number of reviews. Los Angeles County officials said they also
tried and abandoned a similar approach in 2001 because they did not have
sufficient staff to correct the errors that were identified.

Concluding Observations

The Food Stamp Program has seen a significant decline in the national
error rate to a record low in 2003. If the 1999 error rate was in effect
in 2003, the program would have made payment errors totaling over $2.1
billion rather than the $1.4 billion it experienced. Despite the many
challenges states identified, a number of them have significantly lowered
their error rates even while caseloads have continued to rise. However,
some states are having more difficulty lowering their rates, and improper
food stamp payments continue to account for a large amount of money- $1.4
billion in 2003.

It is not completely clear why some states have been more successful at
lowering their error rates than others. Rather than implementing one
specific strategy, the nine states we reviewed have each implemented a
package of changes in response to the unique circumstances in the state.
Even those states we selected because of consistently high error rates
have implemented multiple strategies and expect to see error rate
decreases this year. However, although it is difficult to determine which
actions are most likely to succeed in particular circumstances, we found
examples of strategies that did not succeed because they lacked adequate
management attention or were not effectively implemented.

Future similar error rate reductions may prove challenging. The three
major causes of errors have remained the same over time and are closely
linked to the complexity of program rules and reporting requirements. As
long as eligibility requirements remain so detailed and complex, certain

caseworker decisions will be at risk of error. Moreover,
participant-caused errors, which constitute one-third of the overall
national errors, are difficult to prevent and identify.

Attention from top USDA management as well as continued support and
assistance from FNS will likely continue to be important factors in
further reductions. In addition, if error rates continue to decrease, this
trend will continue to put pressure on states to improve because penalties
are assessed using the state's error rate as compared with the national
average. However, given the size of the Food Stamp Program, the costs to
administer it, and the current federal budget deficit, achieving program
goals more cost-effectively may become more important. FNS and the states
will continue to face a challenge in balancing the goals of payment
accuracy, increasing program participation rates, and the need to contain
program costs.

We provided a draft of this report to the U.S. Department of Agriculture
for review and comment. On April 7, 2005, we met with FNS officials to get
their comments. The officials said they agreed with our findings and
conclusions. FNS also provided us with technical comments, which we
incorporated where appropriate.

  Agency Comments

We are sending copies of this report to the Secretary of Agriculture,
appropriate congressional committees, and other interested parties. We
will also make copies available to others upon request. In addition, the
report will be available at no charge on GAO's Web site at
http://www.gao.gov. Please contact me at (202) 512-7215 if you have any
questions about this report. Other major contributors to this report are
listed in appendix III.

Sigurd R. Nilsen, Director Education, Workforce, and Income Security
Issues

Appendix I: Methodology for Determining the Causes of Food Stamp Payment Errors
for Fiscal Years 1999 through 2003

To determine the causes of food stamp payment errors for fiscal years 1999
through 2003, we analyzed the Food and Nutrition Service's (FNS) quality
control (QC) system data of active cases used in error rate calculations.
State officials draw monthly samples of cases-which are at the household
level-and review them to determine the extent to which the households
received benefits to which they were entitled. The results of these
reviews are included in FNS's QC database, and weighted analyses of these
data produce nationally representative results.

We constructed a database for each year from 1999 through 2003 that
contained a subset of the QC variables relevant to our analysis. For the
1999-2002 databases, we included the reason for error and type of error
variables from the database we obtained directly from FNS and the review
finding, amount of error, and weight variables from an FNS QC database
maintained by Mathematica Policy Research, Inc., and made available to the
public via Mathematica's Web site. For the 2003 data, we only used the FNS
QC database maintained by Mathematica and made available via its Web site
because it contained all the variables we needed. In addition, for each
data set, we created a new variable categorizing the numerous reasons for
error in the agency-or-client (1) variable for the most significant error
to reflect, on a very general level, whether the error was agency-
or-client caused. Likewise, we created a variable categorizing the
numerous types of error in the element (1) code variable as nonfinancial,
resources, income, deductions, or other for the most significant error. We
generated weighted frequencies for the reason, type, and review finding
variables for active cases that were used in calculating the error rate.
Sampling errors for these weighted tabulations were estimated using the
methodology provided in Appendix E of Characteristics of Food Stamp
Households: Fiscal Year 2003, FNS Report Number FSP-04-CHAR. We also
created weighted average dollar amounts of error by case review finding
(e.g., overissuance or underissuance) and weighted frequencies for the
intersection of reason for error and type of error.

To assess the reliability of the data we used, we worked with FNS staff to
obtain and understand the QC data and relied on FNS and Mathematica
documentation on the datasets, and FNS and Mathematica reports based on
these data. We ensured that we reliably downloaded the Mathematica QC data
from the Web and correctly read in FNS's raw QC data that FNS provided to
us by comparing the number of records in each database with the number of
records reported in FNS and Mathematica documentation. In addition, to
ensure the accuracy of the computer programs we used to create and process
the data, a review was made by a second GAO analyst.

Appendix I: Methodology for Determining the Causes of Food Stamp Payment
Errors for Fiscal Years 1999 through 2003

Through our assessment of the reliability of these data, we found that
some variability exists in how states interpret and code the reason for
error variable (i.e., whether error was client-or agency-caused). FNS
stated that no quantitative analysis of the differences across states has
been made. In 2003, FNS implemented guidelines to ensure greater
consistency in state interpretations of the reasons for error (i.e.,
whether the reason for error was client- or agency-caused). Prior to 2003,
interstate variation is believed to be greater than intrastate variation
in these interpretations. Consistency in the error amount is expected to
be a lesser problem since it is based on an established formula. We also
reviewed reports including previous GAO efforts that studied QC processes
and statistical properties. On the basis of the collective information and
findings of our reliability assessment, we determined the data are
sufficiently reliable for our analysis of the causes of food stamp payment
errors.

Appendix II: Food Stamp Combined Error Rates by State for Fiscal Years 1999 to
2004

            State              1999   2000     2001    2002    2003     2004a 
           Alabama            11.29  11.37     9.76    8.74    8.02   
           Alaska             15.94   7.24     9.69    10.99   13.88  
           Arizona             6.93   5.61     5.79    5.27    5.83   
          Arkansas             4.54   4.03     3.24    4.29    4.02   
         California           11.34  13.99    17.37    14.84   7.96   
          Colorado             9.02   7.77     8.53    9.66    7.40   
         Connecticut          13.90   9.31     9.86    11.70   8.77   
          Delaware            16.92  12.53    10.02    8.46    5.38   
    District of Columbia      12.12  10.62    11.38    8.75    8.97   
           Florida             9.43   9.40     9.80    9.61    7.93   
           Georgia            10.86   8.61     6.42    6.73    5.15   
            Guam              10.14  10.56     9.22    6.05    7.04   
           Hawaii              6.82   7.74     6.53    5.03    4.78   
            Idaho             10.94   9.71     7.41    9.04    11.31  
          Illinois            14.79   9.26     8.19    8.75    4.87   
           Indiana             8.11   6.86     6.83    8.31    10.00  
            Iowa               9.27   7.14     7.05    6.44    5.23   
           Kansas              8.98   9.88    10.37    11.70   10.45     4.65 
          Kentucky             7.72   5.81     7.53    7.71    6.32      5.39 
          Louisiana            7.35   5.66     5.78    5.78    5.79      4.74 
            Maine              8.79   9.25     8.49    6.26    13.29    10.38 
          Maryland            13.62  11.06     8.92    8.80    7.23      5.36 
        Massachusetts          9.34   8.63     8.50    8.40    4.99      4.58 
          Michigan            17.59  13.28    13.93    14.10   11.10     6.73 
          Minnesota            6.68   3.58     5.22    5.73    7.96      6.35 
         Mississippi           4.91   4.69     3.47    4.39    4.07      5.55 
          Missouri             8.58   8.06    10.21    9.77    6.75      7.16 
           Montana             8.10   8.48     8.15    8.18    5.78      4.33 
          Nebraska            14.22  10.16     8.44    7.02    7.24      5.48 
           Nevada              8.14   5.11     8.00    7.59    8.25      7.30 
        New Hampshire         12.86  10.26    10.99    12.03   7.52      6.98 
         New Jersey           12.93  12.88     7.97    4.08    2.43      2.62 
         New Mexico           10.39   8.11     6.65    6.71    6.16      5.41 
          New York            10.47  12.35     8.61    7.75    5.88      4.12 
       North Carolina          9.25   6.93     6.35    4.70    4.94      3.21 

Appendix II: Food Stamp Combined Error Rates by State for Fiscal Years
1999 to 2004

                        Wyoming 2.91 4.01 3.04 3.29 4.23
                 National average 9.86 8.91 8.66 8.26 6.63 n/a
State 1999 2000 2001 2002 2003 2004a North  8.03 7.04 5.96 6.14 4.85 4.09 Ohio 8.44 7.96 8.48 6.50 6.61 7.74 Oklahoma 11.88 7.05 8.23 7.94 8.98  Oregon 10.50 10.15 9.76 11.07 13.00  Pennsylvania 10.79 8.19 8.29 9.49 8.21  Rhode  7.05 8.74 5.56 10.21 8.94 12.60  South   5.79 4.47 4.62 4.40 4.94  South  2.19 1.18 2.11 2.12 1.16  Tennessee 8.64 5.71 6.22 7.02 7.20  Texas 4.56 4.14 3.73 4.85 3.29  Utah 12.55 14.43 9.04 6.60 5.00  Vermont 12.09 10.80 10.95 7.68 8.52  Virgin  5.85 6.50 4.70 5.72 6.88  Virginia 11.85 8.66 8.07 6.74 5.46  Washington 8.55 8.20 8.53 8.16 6.28    West   8.88 5.09 6.78 7.13 6.21                                         
                                     Dakota                                                                                                                                                                                   Island                                 Carolina                           Dakota                                                                                                                                                                     Islands                                                                                                    Virginia                           Wisconsin 13.42 12.72 13.14 12.69 9.32 

Source: the Food and Nutrition Service.

aThese are state-reported rates. FNS has not yet adjusted the rates to
reflect the final results of their review.

Appendix III: GAO Contacts and Acknowledgments

GAO Contacts 	Kay Brown, (202) 512-3674, [email protected] Kevin Kumanga
(202) 512-4962, [email protected]

Acknowledgments 	Cathy Roark and Luana Espana also made significant
contributions to this report. In addition, Carl Barden, Evan Gilman, and
Kevin Jackson produced our estimates of the causes of payment error, and
Corinna Nicolaou assisted in the message and report development.

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.

Welfare Reform: Information on Changing Labor Market and State Fiscal
Conditions. GAO-03-977. Washington, D.C.: July 15, 2003.

Food Stamp Employment and Training Program: Better Data Needed to
Understand Who Is Served and What the Program Achieves. GAO-03-388.
Washington, D.C.: March 12, 2003.

Financial Management: Coordinated Approach Needed to Address the
Government's Improper Payments Problems. GAO-02-749. Washington, D.C.:
August 9, 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.

Means-Tested Programs: Determining Financial Eligibility Is Cumbersome and
Can Be Simplified. GAO-02-58. Washington, D.C.: November 2, 2001.

Executive Guide: Strategies to Manage Improper Payments: Learning From
Public and Private Sector Organizations. GAO-02-69G. Washington, D.C.:
October 2001

Food Stamp Program: States Seek to Reduce Payment Errors and Program
Complexity. GAO-01-272. Washington D.C.: January 19, 2001.

Internal Control: Standards for Internal Control in the Federal
Government. GAO/AIMD-00-21.3.1. Washington, D.C.: November 1999.

Food Stamp Program: States Face Reduced Federal Reimbursements for
Administrative Costs. GAO/RCED/AIMD-99-231. Washington D.C.: July 23,
1999.

  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

                           PRINTED ON RECYCLED PAPER
*** End of document. ***