Food Stamps: Substantial Overpayments Result From Prisoners Counted as
Household Members (Letter Report, 03/10/97, GAO/RCED-97-54).

GAO provided information on Food Stamp Program overpayments, focusing
on: (1) how many prisoners were included as members of households that
received food stamp benefits, hereafter referred to as prisoner
participation, and the estimated value of improper benefits that were
issued to the households; (2) how prisoner participation could take
place without detection; and (3) whether computer matching can be an
effective method for identifying prisoner participation.

GAO noted that: (1) despite federal regulations prohibiting inmates of
correctional institutions from participating in the Food Stamp Program,
GAO identified 12,138 inmates in the areas it examined who were included
in households receiving food stamps; (2) these households improperly
collected an estimated $3.5 million in food stamp benefits; (3) prisoner
participation goes undetected because agencies generally do not verify
the information on household membership provided by food stamp
applicants; (4) furthermore, according to officials of the Department of
Agriculture's Food Stamp Program, most state or local agencies
responsible for administering the program do not routinely collect and
review lists of individuals incarcerated in state and local facilities
to determine whether any of these individuals are being counted as
members of food stamp households; (5) given the program's reliance on
client-provided information, computer matching of lists of prisoners and
food stamp household members provides a straightforward and potentially
effective mechanism to accurately and independently identify prisoners'
participation; (6) while states have implemented various computer
matching routines, such as the Income and Eligibility Verification
System, which compares data on welfare clients with data on state and
federal wages and benefits, many states have not yet implemented a
computer matching program to identify prisoners participating in the
Food Stamp Program.

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

 REPORTNUM:  RCED-97-54
     TITLE:  Food Stamps: Substantial Overpayments Result From Prisoners 
             Counted as Household Members
      DATE:  03/10/97
   SUBJECT:  Food relief programs
             Program abuses
             Welfare benefits
             Prisoners
             Overpayments
             Families
             Computer matching
             Law enforcement information systems
             Eligibility determinations
             Locally administered programs
IDENTIFIER:  Food Stamp Program
             FNS Income and Eligibility Verification System
             California
             New York
             Florida
             Texas
             Massachusetts
             Missouri
             SSA Enumeration Verification System
             Los Angeles County (CA)
             Dade County (FL)
             New York (NY)
             Harris County (TX)
             
******************************************************************
** This file contains an ASCII representation of the text of a  **
** GAO report.  Delineations within the text indicating chapter **
** titles, headings, and bullets are preserved.  Major          **
** divisions and subdivisions of the text, such as Chapters,    **
** Sections, and Appendixes, are identified by double and       **
** single lines.  The numbers on the right end of these lines   **
** indicate the position of each of the subsections in the      **
** document outline.  These numbers do NOT correspond with the  **
** page numbers of the printed product.                         **
**                                                              **
** No attempt has been made to display graphic images, although **
** figure captions are reproduced.  Tables are included, but    **
** may not resemble those in the printed version.               **
**                                                              **
** Please see the PDF (Portable Document Format) file, when     **
** available, for a complete electronic file of the printed     **
** document's contents.                                         **
**                                                              **
** A printed copy of this report may be obtained from the GAO   **
** Document Distribution Center.  For further details, please   **
** send an e-mail message to:                                   **
**                                                              **
**                                            **
**                                                              **
** with the message 'info' in the body.                         **
******************************************************************


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


Report to Congressional Committees

March 1997

FOOD STAMPS - SUBSTANTIAL
OVERPAYMENTS RESULT FROM PRISONERS
COUNTED AS HOUSEHOLD MEMBERS

GAO/RCED-97-54

Prisoner Participation in Food Stamp Program

(150259)


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

  EVS - Enumeration Verification System
  FCS - Food and Consumer Service
  IEVS - Income Eligibility Verification System
  SSN - Social Security Number
  USDA - U.S.  Department of Agriculture

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


B-275943

March 10, 1997

Congressional Committees

During numerous hearings over the last several years, the Congress
has expressed its strong desire to reduce the level of fraud, waste,
and abuse in the Food Stamp Program.  In fiscal year 1995, for
example, the U.S.  Department of Agriculture (USDA) issued over $22
billion in food stamp benefits for about 26 million individuals.  In
that same year, USDA reported that approximately 15 percent of food
stamp cases contained errors resulting in as much as $1.7 billion in
overpayments nationwide. 

In this context, we examined the computerized records of inmates of
correctional institutions, who are ineligible for food stamps, to
determine whether they were being included as members of households
receiving food stamp benefits.  The value of food stamp benefits that
a household is entitled to receive is partially determined by the
number of eligible household members.  Prisoners are not to be
counted as part of a household when benefits are calculated.  We
determined (1) how many prisoners were included as members of
households that received food stamp benefits (hereafter referred to
as prisoner participation) and the estimated value of improper
benefits that were issued to the households, (2) how prisoner
participation could take place without detection, and (3) whether
computer matching can be an effective method for identifying prisoner
participation. 

We conducted a computer match of the 1995 food stamp rolls with state
prison data in four states (California, Florida, New York, and
Texas)\1 and with local jail data in four metropolitan areas (Los
Angeles County, California; Dade County, Florida; New York City, New
York; and Harris County, Texas). 


--------------------
\1 California food stamp data are limited to only those Los Angeles
County participants who were on the rolls as of August 1995.  We
conducted the computer match for the three states and Los Angeles
County.  Together, these jurisdictions account for 27 percent of food
stamp program participants nationwide. 


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

Despite federal regulations prohibiting inmates of correctional
institutions from participating in the Food Stamp Program, we
identified 12,138 inmates in the areas we examined who were included
in households receiving food stamps.  These households improperly
collected an estimated $3.5 million in food stamp benefits. 

Prisoner participation goes undetected because agencies generally do
not verify the information on household membership provided by food
stamp applicants.  Furthermore, according to officials in USDA's Food
Stamp Program, most state or local agencies responsible for
administering the program do not routinely collect and review lists
of individuals incarcerated in state and local facilities to
determine whether any of these individuals are being counted as
members of food stamp households. 

Given the program's reliance on client-provided information, computer
matching of lists of prisoners and food stamp household members
provides a straightforward and potentially effective mechanism to
accurately and independently identify prisoners' participation. 
While states have implemented various computer matching
routines--such as the Income and Eligibility Verification System,
which compares data on welfare clients with data on state and federal
wage and benefits--many states have not yet implemented a computer
matching program to identify prisoners participating in the Food
Stamp Program. 


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

The Food Stamp Program is designed to promote the general welfare and
to safeguard the health and well-being of the nation's population by
raising the nutrition levels of low-income families.  Recipients use
their food stamp benefits to purchase allowable food products from
authorized retail food merchants. 

Eligibility for food stamp benefits is determined on a household
basis.  A household can be either an individual or a family or other
group that lives together and customarily purchases and prepares food
in common.  The value of food stamp benefits for a household is
determined by the number of eligible household members and their
income, adjusted for assets and such costs as shelter and utilities. 
The household's monthly food stamp allotment increases with each
additional member, provided income limits are not exceeded. 
Household members who are incarcerated and fed by a correctional
facility are not eligible for food stamp benefits and are not to be
included in the household for purposes of calculating the food stamp
benefit.  Households that receive food stamps are required to report
changes in household membership, such as a member's incarceration, to
the administering state or local agency. 

Within USDA, the Food and Consumer Service (FCS) administers the Food
Stamp Program through agreements with state agencies.  FCS is
responsible for approving state plans for operation and ensuring that
the states are administering the program in accordance with
regulations.  States are required to establish a performance
reporting system to monitor the program, including a quality control
review process to help ensure that benefits are issued only to
qualifying households and that the benefit amounts are correct. 
State agencies are responsible for imposing penalties for violations
of program requirements and for recovering food stamp overpayments.\2

The program is administered at the local level by either a state
agency or a local welfare agency, depending on the state.  In
California, county agencies operate the program at the local level,
while in New York State, districts operate the program.  The state
agency supervises operations in both states.  In Florida and Texas,
state agencies operate the program through district and regional
offices, respectively.  Whatever the administering authority, local
service centers work directly with clients to certify household
eligibility and determine benefit amounts at the time of application
and at least annually thereafter. 

To identify prisoner participation, we performed a computer match
comparing 1995 food stamp rolls with inmate rolls.  To ensure that
our analyses resulted in valid matches, we (1) verified the
prisoners' social security numbers through the Social Security
Administration's verification system, (2) used only those matches
showing dates of incarceration that coincided with the dates that
food stamp benefits were issued to the household, and (3) used only
those matches showing that the prisoner had been incarcerated for at
least a full month and that sufficient time had elapsed for the
household to notify the state of the change and for the state to take
action.\3 The food stamp rolls covered three large states (Florida,
New York, and Texas) and one large county (Los Angeles, California). 
(See app.  I.) The inmate rolls covered the state prison population
in the four states and the jail population in large metropolitan
areas of each state, that is, Los Angeles County, California; Dade
County, Florida; New York City, New York; and Harris County, Texas. 
Our detailed methodology is discussed in appendix II. 


--------------------
\2 The Personal Responsibility and Work Opportunity Reconciliation
Act of 1996 doubled the disqualification period for intentional
program violations from 6 months to one year for the first violation,
and from one year to 2 years for the second violation. 

\3 Program regulations generally give a household 10 days to notify
the state of changes in household composition and the state 10 days
from the date of notification to adjust benefits. 


   PRISONERS INCLUDED AS HOUSEHOLD
   MEMBERS COST MILLIONS IN
   OVERPAYMENTS
------------------------------------------------------------ Letter :3

During calendar year 1995, about $3.5 million in food stamp benefits
were issued on behalf of state prison and county jail inmates claimed
as household members in the locations we examined.  (See table 1.) Of
this total, nearly 9,500 state prison inmates included as household
members accounted for an estimated $2.6 million in benefits.  About
2,700 county jail inmates accounted for over $900,000 in benefits. 



                                Table 1
                
                  Months of Prisoner Participation and
                  Estimated Overpayments, by Prison or
                              Jail System

                         (Dollars in thousands)

                                         Total
                                        inmate   Months of   Estimated
                                    participan  participat  overpaymen
Jurisdiction examined                       ts       ion\a          ts
----------------------------------  ----------  ----------  ----------
State prison system
----------------------------------------------------------------------
California \b                              913       2,814        $194
Florida                                  1,167       4,523         353
New York                                 1,670       7,445         587
Texas                                    5,690      20,081       1,470

County jail system
----------------------------------------------------------------------
Los Angeles\b                              277         494          34
Dade                                     1,712       9,450         738
New York City                              604       1,700         134
Harris                                     105         267          19
======================================================================
Total                                   12,138      46,774      $3,529
----------------------------------------------------------------------
\a Months of participation is the number of months for which
households received benefits while household members were in prison. 

\b Includes only the Los Angeles County food stamp recipients who
were on the rolls as of Aug.  1995. 

Sources:  For California, California Department of Corrections and
Los Angeles County Department of Public Social Services; for Florida,
Florida Department of Corrections and Florida Department of Health
and Rehabilitative Services; for New York State, New York State
Department of Correctional Services and New York State Department of
Social Services; for Texas, Texas Department of Criminal Justice and
Texas Department of Health Services; and USDA's Food and Consumer
Service. 

The inmate participants that we identified in our match were members
of households of varying sizes, some with multiple members and some
with a single member--the prisoner was the household.  For households
with multiple members, the household continued to receive its monthly
benefits, which were calculated on the presumption that the prisoner
was present in the home.  For single-member households, someone other
than the prisoner was issued the benefits.  The stamps could have
been issued either to a person designated as the prisoner's
authorized representative\4 or to someone who fraudulently
represented himself or herself as the prisoner to receive the
benefits. 

Food stamp benefits are issued either as coupons or via electronic
benefit transfer systems.  For coupons, issuance procedures require
that the client presents various items of identification, such as
Food Stamp Program cards bearing the client's signatures, in order to
pick up food stamps from a service center or other outlet.  A small
number of clients receive their coupons through the mail.  Under
electronic benefit transfer systems, the state agency issues access
cards (similar to credit cards) and personal identification numbers
to clients who obtain benefits through point-of-sale terminals in
stores.  However, the effectiveness of the issuance procedures to
ensure that only eligible participants receive benefits depends on
how rigorously the procedures are implemented by the responsible
staff. 


--------------------
\4 All food stamp clients have the right when they apply to specify
an authorized representative to act on the their behalf, including
receiving their stamps. 


   LACK OF VERIFICATION ALLOWS
   PRISONER PARTICIPATION TO GO
   UNDETECTED
------------------------------------------------------------ Letter :4

Prisoners are able to participate in the Food Stamp Program because
local welfare agencies seldom verify the composition of a household. 
Instead, most agencies rely on food stamp applicants to provide
accurate household information and to report subsequent changes, such
as the incarceration of a household member.  Most agencies do not,
for example, routinely compare lists of prison or jail inmates with
lists of household members. 

In general, the Food Stamp Program has to balance the issues of
client convenience, administrative simplicity, and payment accuracy;
consequently, controls over such eligibility factors as household
composition are not rigorous.  A household that wishes to receive
benefits must present an application listing members and provide
information about their income and other eligibility factors. 
Caseworkers review this information, interview a household
representative, and certify eligibility.\5 In addition, they
recertify the household at least annually.  However, at no time are
all household members required to appear and present identification. 
Furthermore, clients are responsible for identifying changes in
household composition.\6

According to FCS' 1995 quality control review, which identified error
rates for each state by reviewing a random sample of cases, client
errors or misrepresentations contributed significantly to incorrect
benefits, particularly when an overpayment occurred.  FCS reported
that overpayments occurred in about 15 percent of the cases reviewed
nationwide and that 62 percent of the dollar value of overpayments
was attributable to inaccuracies in client-provided information. 

Nevertheless, FCS' regulations do not require verification of
client-provided information on household composition, unless the
caseworker deems the information "questionable." The regulations
allow each state agency to develop guidance for identifying
questionable information.  In the states we visited, the guidance
defined questionable information as applicants' statements that were
contradictory or did not agree with information that was in the case
record or otherwise available to the caseworker. 

When the caseworkers in the states we visited suspected fraudulent
information, they could refer the application to investigators before
granting aid.  Investigators in each state told us that they
attempted to verify questionable information on household composition
by visiting homes and making collateral contacts to confirm
information with friends, neighbors, or landlords.  According to the
investigators, these techniques were hit-or-miss, time-consuming, and
costly undertakings, and provided information that was only as
reliable as its source.  Furthermore, investigative resources were
generally very limited; for example, the Miami area, which contains
about 26 percent of Florida's food stamp recipients, had just one
field investigator to conduct household visits. 

Some agencies have employed computer matching as a means of
identifying ineligible recipients, such as prisoners, but the
practice does not appear to be widespread.  According to FCS, four
states (Florida, Massachusetts, Missouri, and New York) currently
perform a monthly computer match between state prisons' inmate
records and food stamp rolls; two states were in the process of
developing such a match; and one state performed an annual match. 
FCS' regional offices identified only one local agency that compared
food stamp recipients with county jail inmates.  However, our
discussions with officials in the states we visited indicated that
the actual number of local agencies conducting such matches was
larger.  For example, in California, the state agency reported that
14 of the state's 58 county agencies collected and reviewed data on
local jails' inmates at least once a week. 

Of the states we visited, Florida and New York operated matching
programs, Texas was in the process of establishing a program, and
California had plans to implement a program at some future date. 
While Florida and New York conduct routine matching programs, we
identified prisoner participation in the Food Stamp Program in these
states because (1) our matches covered a time period not covered by
the states' matches and (2) we used prisoners' social security
numbers, which were verified by the Social Security Administration, a
step the states had not taken. 


--------------------
\5 Sometimes the caseworker fills out the application on the basis of
the interview. 

\6 Almost all households in California must submit monthly change
reports.  Households in Florida and Texas are required to report
changes in circumstances, including changes in household composition,
within 10 days of becoming aware of the change.  Households with
earned income in New York State must report changes quarterly; all
others must report within 10 days of a change. 


   COMPUTER MATCHING REPRESENTS A
   POTENTIALLY EFFECTIVE TOOL FOR
   DETECTING PRISONER
   PARTICIPATION
------------------------------------------------------------ Letter :5

Although computer matching of inmate data is not used often, our test
in four states demonstrates that it can be a useful technique for
identifying households that improperly include prisoners as members. 
A study by an FCS contractor of other computerized information
verification processes in place at state agencies demonstrated that
such matches are cost-effective, particularly when properly
targeted.\7 Ongoing and developing state matching programs could
benefit from use of targeted matching and from sharing experiences. 

Officials in the four states we visited viewed the matching of
prisoner data with food stamp data as a fairly straightforward,
effective process.  These officials said that they did not encounter
or foresee any privacy issues that precluded such matching. 
Furthermore, while they were unable to provide detailed cost or
savings information regarding their prison match programs, the two
states we visited that had implemented such programs believed that
they were beneficial.  New York State did not track implementation
costs but calculated savings in the Food Stamp Program of over
$900,000 from August 1995 to April 1996.  Because Florida was
legislatively mandated to implement computer prison matches, the cost
of implementation was not a major concern and therefore was not
tracked.  Florida has yet to calculate savings in the Food Stamp
Program. 

Although detailed data supporting the cost-effectiveness of a
computer prison match is not available from the states we visited,
strong evidence exists that such a match, particularly when properly
targeted, is cost-effective.  The Income and Eligibility Verification
System (IEVS) compares wage, benefit, and other payment information
reported by food stamp clients with records in six databases,
including those maintained by the Social Security Administration, the
Internal Revenue Service, and state unemployment insurance agencies. 
After this matching program was implemented, some caseworkers charged
that much of the information provided in the IEVS matches did not
lead to savings in the Food Stamp Program.  The problems most often
cited were (1) out-of-date information, (2) lack of agreement in the
time periods covered by data sources, and (3) duplicate data. 

In response, in 1991, FCS engaged a contractor to evaluate the
cost-effectiveness of the IEVS system in two sample states, Arizona
and Michigan.  Various targeting criteria, such as beneficiaries over
a specific age or matches when specific dollar thresholds were
exceeded, were used to select cases for follow-up.  All of the
targeted IEVS matching programs reviewed in the study were found to
be cost-effective.  The study determined that the largest cost of the
IEVS matching program is the time spent by caseworkers on follow-ups,
approximately $5 to $7 per follow-up.  Data-processing costs averaged
2 cents per case, and Arizona spent approximately $104,000 to develop
its software.  Every match had a cost-effectiveness ratio (program
savings compared with the costs of the match, targeting, follow-up
and claims collection) greater than 1, indicating that every dollar
spent on IEVS returned more than a dollar in savings to the program. 
In addition, each match was found to have positive net savings for
the program, with the more narrowly targeted matches yielding the
largest net savings, since they focused follow-up actions on the more
egregious situations. 

The states we visited were implementing their prison matches in a
manner that was very similar to that reported in the study.  Matches
were sent to local offices, where caseworkers, specialists, clerical
staff, and fraud investigators could participate in the process.  The
case file information was reviewed, the client was contacted, and the
discrepancy was verified or refuted.  If the discrepancy was
verified, the client's eligibility and benefits were redetermined
and, as appropriate, overpayments could be recovered and fraud
investigations conducted. 

Our test showed that developing the computer programs to identify
prisoner participation did not require a large investment of a
programmer's time.  Our programmer required an average of about 20
days to develop a series of substantially different programs for each
state.  The 20 days included time to become familiar with the data as
well as to write, test, and execute the programs.  State programmers
may require less time because they are already familiar with the food
stamp data. 

As in the IEVS study, we used some targeting criteria to enhance the
effectiveness of the matching process.  Before using the inmate data,
we sent the information to the Social Security Administration for
verification of the prisoners' social security numbers (the
identifier common to all major federal databases on individuals) to
ensure that our cases did not include incorrect numbers that would
render the match invalid.  None of the states we visited with
computer matching programs submitted inmate social security numbers
to the Social Security Administration for verification.  (The agency
performs this service for government customers at no charge.) We
matched only those social security numbers that had been verified by
the Social Security Administration.  (See app.  III.) The majority of
the inmate participants we identified in our match occurred as a
result of the verification process.  (See table 2.)



                                Table 2
                
                    Total Prisoner Participation and
                 Participants Identified From Corrected
                Social Security Numbers Provided by the
                     Social Security Administration

                                        Participants     Percentage of
                                          identified      participants
                                                from   identified from
                          Total inmate     corrected         corrected
State prison system       participants       numbers           numbers
------------------------  ------------  ------------  ----------------
California\a                       913           348              38.1
Florida                          1,167           283              24.3
New York                         1,670           746              44.7
Texas                            5,690         4,473              78.6
======================================================================
Total                            9,440         5,850              62.0
----------------------------------------------------------------------
\a Includes only the Los Angeles County food stamp recipients who
were on the rolls as of Aug.  1995. 

Sources:  For California, California Department of Corrections and
Los Angeles County Department of Public Social Services; for Florida,
Florida Department of Corrections and Florida Department of Health
and Rehabilitative Services; for New York State, New York State
Department of Correctional Services and New York State Department of
Social Services; and for Texas, Texas Department of Criminal Justice
and Texas Department of Health Services. 

By selecting only prisoners (1) whose dates of incarceration matched
the dates that food stamp benefits were issued to their household and
(2) who had been incarcerated at least a full month, we avoided some
of the pitfalls that have been or could be encountered by states
implementing matching programs.  For example, case analysts in
Florida told us that they could not take any action on many of their
matches because the prisoners had been incarcerated for only a few
days or benefits had not actually been issued to the household during
the period the prisoner was incarcerated.  In other cases, an
unverified social security number in the prison records resulted in a
match with an eligible food stamp recipient.  Our analysis of results
reported from Florida's match of June 1996 for Dade County indicates
that of 674 matches, 423 resulted in no action taken by the
caseworker; for 41 matches, the record did not indicate any review. 
On the positive side, 210 matches resulted in a case closure
(household dropped from Food Stamp Program), removal of the
participant from a case (individual dropped from household membership
and benefits recalculated), a referral for fraud, or some combination
of those actions.  Florida officials acknowledged weaknesses in their
matching process and stated that they intend to review and improve
the process to better identify cases for which caseworkers could take
action. 

The states we visited that had or were developing matches were acting
with little or no knowledge of the matching efforts of other states. 
As a result, each state started without any information, rather than
building on the experiences of others.  Thus, any cost or time
savings that could have arisen from the sharing of information were
not realized. 


--------------------
\7 The Income and Eligibility Verification System (IEVS) Targeting
Demonstration, Findings and Guidelines for State Food Stamp IEVS
Programs, Final Report, USDA, Food and Consumer Service, Office of
Analysis and Evaluation, Apr.  1, 1995. 


   CONCLUSIONS
------------------------------------------------------------ Letter :6

The participation of ineligible individuals undermines the
credibility of the Food Stamp Program and results in overpayments. 
Conventional methods state agencies have used to verify the
membership of food stamp households have not prevented households
from including ineligible individuals, such as inmates in local jails
and state prisons.  Prisoners' participation in the Food Stamp
Program resulted in overpayments of $3.5 million for the locations
where we conducted matches.  A computer match of data on states' food
stamp participants and verified inmates could be a cost-effective
method for identifying a prisoner's participation in a food stamp
household and thus provide the evidence needed to remove the prisoner
from the calculation of a household's eligibility and benefits.  Some
states have recognized that matching is a cost-effective way to
reduce overpayments.  Sharing of information on effective matching
practices, such as methods of targeting the most productive cases,
would benefit states. 


   RECOMMENDATIONS
------------------------------------------------------------ Letter :7

To identify state and county prisoners who are included as members of
households receiving food stamps, we recommend that the Secretary of
Agriculture actively encourage states to implement periodic computer
matches of data on state and local prison inmates with data on
participants in the Food Stamp Program.  To facilitate this effort,
we recommend that the Secretary of Agriculture direct FCS to (1)
collect from the states that conduct matches information on the
policies and procedures used to implement their matches and (2)
evaluate, summarize, and disseminate to the states the policies and
procedures that represent best practices, such as the verification of
prisoners' social security numbers with the Social Security
Administration. 


   AGENCY COMMENTS
------------------------------------------------------------ Letter :8

We provided copies of a draft of this report to FCS for review and
comment.  In commenting on the draft report, FCS agreed with the
report's findings, conclusions, and recommendations.  These comments,
which appear as appendix IV, contained suggestions regarding the
phrasing used in the report that we incorporated as appropriate. 


---------------------------------------------------------- Letter :8.1

We conducted our work from March 1996 through February 1997 in
accordance with generally accepted government auditing standards. 
Our detailed methodology is presented in appendix II. 

We are providing copies of this report to appropriate congressional
committees, interested Members of Congress, and other interested
parties.  We will also make copies available to others on request. 

If you have any questions about this report, please contact me at
(202) 512-5138.  Major contributors to this report are listed in
appendix V. 

Robert A.  Robinson
Director, Food and
 Agriculture Issues


List of Addressees

The Honorable Richard G.  Lugar
Chairman
The Honorable Tom Harkin
Ranking Minority Member
Committee on Agriculture, Nutrition,
 and Forestry
United States Senate

The Honorable Thad Cochran
Chairman
The Honorable Dale Bumpers
Ranking Minority Member
Subcommittee on Agriculture, Rural Development,
 FDA, and Related Agencies
Committee on Appropriations
United States Senate

The Honorable Robert F.  (Bob) Smith
Chairman
The Honorable Charles W.  Stenholm
Ranking Minority Member
Committee on Agriculture
House of Representatives

The Honorable Joe Skeen
Chairman
The Honorable Marcy Kaptur
Ranking Minority Member
Subcommittee on Agriculture, Rural Development,
 FDA, and Related Agencies
Committee on Appropriations
House of Representatives


INFORMATION ON THE COST OF FOOD
STAMP BENEFITS AND STATE PRISON
POPULATION
=========================================================== Appendix I

Nationwide, California, Florida, New York, and Texas represent almost
36 percent of the cost of Food Stamp Program benefits and
approximately 39 percent of the states' prison population.  The
prison data in this table are based on the prison population as of
June 30, 1995. 



                               Table I.1
                
                1995 Food Stamp Program Benefit Cost and
                  State Prison Population for Selected
                               Locations

                                Food stamp benefit        State prison
Location                                      cost          population
------------------------------  ------------------  ------------------
California                          $2,472,936,680             131,342
Florida                              1,307,134,257              61,992
New York                             2,065,406,076              68,526
Texas                                2,246,359,764             127,092
======================================================================
Total                               $8,091,836,777             388,952
======================================================================
U.S. total                         $22,766,109,338           1,004,608
----------------------------------------------------------------------
Sources:  U.S.  Department of Agriculture's Food and Consumer Service
and the Department of Justice. 


OBJECTIVES, SCOPE, AND METHODOLOGY
========================================================== Appendix II

In response to the Congress' strong interest in reducing the level of
fraud, waste, and abuse in the Food Stamp Program, we reviewed food
stamp beneficiaries to determine whether prisoners, who are not
eligible for food stamps, were inappropriately included as members of
households receiving food stamps.  Specifically, we determined (1)
how many prisoners were included as members of households that
received food stamp benefits and the estimated value of improper
benefits that were issued to the households, (2) how prisoner
participation could take place without detection, and (3) whether
computer matching can be an effective method for identifying prisoner
participation. 

To determine if inmates of correctional facilities were included as
members of households that received food stamp benefits, and the
estimated value of benefits that were issued to the households, we
matched the food stamp records and state prison records of the four
states with the largest Food Stamp Program benefits and the largest
state prison populations.  We also matched food stamp records and
jail records in four metropolitan areas.  Specifically: 

  -- The Florida, New York, and Texas state welfare agencies provided
     us with computer files containing information on all members of
     households and the amount of household food stamp benefits
     issued during 1995.  In California, this information is
     maintained only at the county level, so we obtained information
     only for Los Angeles County beneficiaries, who account for
     approximately one-third of the benefits California issues.  The
     data provided personal identifiers, including name, social
     security number (SSN), date of birth, gender, and the months in
     which food stamp benefits had been issued to the household of
     which each individual was a member.  The state agencies had
     verified the SSNs for the data on food stamp beneficiaries
     through the Social Security Administration's Enumeration
     Verification System (EVS). 

  -- The state prison system in each state provided us with computer
     data on all prisoners incarcerated in a state facility for all
     or any part of 1995.  The data provided the same personal
     identifiers as we obtained for food stamp beneficiaries and
     listed the admission and release dates for each period of
     incarceration during the year.  To expedite the delivery of
     data, New York State simply listed each full month that a
     prisoner was incarcerated rather than providing specific dates. 
     We verified the prisoners' SSNs through the Social Security
     Administration's verification system. 

  -- Four large metropolitan county or local jail systems gave us
     permission to use data they had previously provided on our
     review of erroneous Supplemental Security Income payments to
     prisoners.\1 The systems, including one from each state in our
     review, were Los Angeles County, California; Dade County,
     Florida; New York City, New York; and Harris County, Texas.  The
     local jail system data included all prisoners who were
     incarcerated as of specific dates--these dates were selected by
     the jail systems and were based on their available resources. 
     The jail systems provided available personal identifiers, as
     listed above, and the date of incarceration.  The jail inmates'
     SSNs had been verified by the Social Security Administration's
     verification system during our previous review. 

We matched the verified SSNs of prisoners in each state or local
prison with the verified SSNs in the states' records of membership in
food stamp households.  For those prisoners identified as members of
households, we determined the periods in which food stamp issuance
and incarceration coincided.  We estimated the dollar value of food
stamps issued to households with participating prisoner members by
applying the state's average monthly issuance per individual
recipient from 1995 to each period where incarceration and issuance
coincided.  Food stamp benefits are calculated for households, not
for individuals.  As such, it is difficult to determine the exact
value of benefits issued to a prisoner participating in a household,
unless he or she is the only member of a household.  Even then, the
amount will vary from individual to individual, depending on factors
such as income, assets, and the cost of shelter.  Therefore, we
relied on the average monthly benefit issuance per person in the
locations we reviewed, which ranged from a high of $78.84 in New York
State to a low of $68.89 in Los Angeles County. 

In recognition of the notification and processing time frames that
allow 10 days for clients to report household changes and 10 days for
the state agency to take action, we did not consider any issuance in
the month of incarceration to be an overpayment.  Furthermore, if a
prisoner was admitted on or after the tenth day of the month, we did
not consider issuance in the following month to be an overpayment. 
We prorated the average monthly issuance to determine the overpayment
for days incarcerated in the month of discharge. 

Because of the quality control program operated by USDA's Food and
Consumer Service (FCS) and the states' ongoing quality assurance
efforts, we accepted the computerized food stamp data as reliable. 
The prison data, such as dates of incarceration and release, would
have been very difficult to verify within the time frames of this
audit because these data are sensitive, dispersed within the states,
or not available in hard copy.  State prison officials attested to
the reliability of the admission and release data.  They said that
because these data are critically important, they are under the
constant scrutiny of the courts, law enforcement authorities, and
inmates.  In our previous study of prisoners receiving supplemental
security income, we verified a random sample of jail data and found
the data reliable. 

To determine why prisoner participation was not detected, we asked
FCS to identify state or local agencies that collect prison data and
compare that data with data on food stamp recipients to identify
prisoner participation.  To discuss and review policy and procedures
for verifying applicant data and any subsequent changes, we visited
state agency officials in Sacramento, California; Tallahassee,
Florida; Albany, New York; and Austin, Texas.  We discussed fraud
detection programs, quality control and assurance efforts, and
methods of food stamp issuance with state officials.  In addition, we
visited social service administrative and service centers in the four
large metropolitan areas we selected for review.  At each location we
observed and discussed the food stamp application, data verification,
certification and recertification process.  We discussed local fraud
detection efforts and observed the food stamp issuance process. 

To determine whether computer matching can be an effective method for
identifying prisoner participation, we discussed with agency
officials in each of the states we visited the cost, quality,
savings, and barriers to matching inmate data with state food stamp
data.  At the social service centers we visited, we discussed the
quality of the matches and observed the follow-up process.  To
identify the effort associated with data matching to identify
prisoner participation, we identified the time used by our programmer
to develop and implement the match programs and reviewed a cost study
performed for FCS regarding similar matching routines. 


--------------------
\1 Supplemental Security Income:  SSA Efforts Fall Short in
Correcting Erroneous Payments to Prisoners (GAO/HEHS-96-152, Aug. 
30, 1996). 


SOCIAL SECURITY ADMINISTRATION'S
VERIFICATION PROCESS SIGNIFICANTLY
INCREASED THE NUMBER OF VALID SSNS
========================================================= Appendix III

Verification of prisoners' SSNs by the Social Security
Administration's EVS significantly increased the number of valid SSNs
that we could use in our matches.  The state prison systems provided
us with available SSNs for all prisoners incarcerated in a state
facility for all or any part of 1995.  Therefore, this table contains
more prisoner data than table I.1, which contains data from one point
in time.  As shown in table III.1, over 60 percent of 522,525
prisoner SSNs were validated as accurate and usable as submitted. 
EVS identified an additional 120,525 valid SSNs for prisoners by
comparing submitted prison data (SSN if available, date of birth,
name, and gender) against information contained in Social Security
Administration records.  This comparison yielded numbers not
contained in the prison records, corrected transposition errors, and
substituted correct numbers for invalid numbers.  Because of the
Social Security Administration's confidence in, and the historical
reliability of the EVS process, we accepted these additional
validated SSNs for use in our match process.  Similarly, the SSNs for
local jail prisoners had been validated through EVS. 



                              Table III.1
                
                  Results of the EVS' Social Security
                      Number Validation, by State

SSN validation status   California   Florida  New York   Texas   Total
----------------------  ----------  --------  --------  ------  ------
Correct as submitted       110,013    60,774    51,287  97,765  319,83
                                                                     9
Correct SSN identified      41,496    12,536    20,880  45,613  120,52
                                                                     5
======================================================================
Subtotal                   151,509    73,310    72,167  143,37  440,36
                                                             8       4
Not validated               34,828     8,924    20,612  17,797  82,161
======================================================================
Total                      186,337    82,234    92,779  161,17  522,52
                                                             5       5
----------------------------------------------------------------------
Sources:  California Department of Corrections, Florida Department of
Corrections, New York State Department of Correctional Services,
Texas Department of Criminal Justice, and the Social Security
Administration. 




(See figure in printed edition.)Appendix IV
COMMENTS FROM THE U.S.  DEPARTMENT
OF AGRICULTURE
========================================================= Appendix III



(See figure in printed edition.)


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

Keith Oleson, Assistant Director
David Moreno, Project Leader
Brad Dobbins
Don Ficklin
Jon Silverman


*** End of document. ***