Optimal thresholds in the collection of food stamp program claims

              pi 319*.
*r
OoW-A-oi Aqr.2:-«n
USDA
United States
Department of
Agriculture
Food and
Consume'
Service
Office of
Analysis and
Evaluation
Optimal Thresholds
in the Collection of
Food Stamp Claims
June 1996
9?00457O &
Contract No. FCS 53-3198-3-022, Task 9
MPR Reference No. 8168-009
OPTIMAL THRESHOLDS IN THE COLLECTION OF
FOOD STAMP PROGRAM CLAIMS
June 5,1996
Author:
Myles Maxfidd
This study is sponsored by the Food and Consumer Service, U.S. Department of Agriculture. The
conclusions and opinions expressed in this report do not necessarily represent the po ition of Food
and Consumer Service.
Submitted to:
US Department of Agriculture
Food and Consumer Service
3101 Park Center Dr., Room 214
Alexandria, VA 22302
Attn.: Sharron Cristofar
Submitted by:
Mathematica Policy Research, Inc.
600 Maryland Avenue, S.W.
Suite 550
Washington, D.C. 20024
Under subcontract to:
TASCON
7101 Wisconsin Avenue
Bethesda,MD 20814
z
JL
Mffi mm
ACKNOWLEDGMENTS
The author would like to acknowledge the contributions of Sharron Cnstofar, James Porter, Abigail
Nichols, and Dan Wilusz, of the US Department of Agriculture, and Richard Hilton and Judy Barokas
ofTASCON, Inc
This project could not have been completed without the support and cooperation of FCS Regional
Office staff who served as project links to the participating states. Particular thanks go to Jeanene Cox of
the Southeast Regional Office, Jeanette Moms of the Southwest Regional Office, Donna Appleton of the
Mountain Plains Regional Office, and Pat Cruise of the Western Regional Office
The author would particularly like to express their appreciation for the cooperanon and support
provided by program staff in all of the participating states. Special thanks go to Charles Cook and Connie
Walters of the Alabama Department of Human Resources, Julia Sohn and Aldona Vaitkus of the Arizona
Department of Economic Security, Mike Boyd and Bob Verucchi of the Arkansas Department of Human
Services. Phil Passentino of the Passaic County Board of Social Services, and Brenda Hofer of the Utah
Department of Human Services
111
I MMM
CONTENTS
Chapter Page
EXECUTIVE SUMMARY ix
I INTRODUCTION 1
A. BACKGROUND 2
B. METHODOLOGY 6
C PLAN OF THE REPORT 7
D ELEMENTS OF CLAIM COLLECTION COST 9
A. AE CLAIMS 14
B ME CLAIMS 15
C. IPV CLAIMS 17
ED THEORY OF OPTIMAL THRESHOLDS 21
A BASIC CONCEPTS 21
B OPTTMALITY FOR AN INDIVIDUAL CLAIM 22
C DETERRENCE AND UNMEASURED COSTS 25
D ESTABLISHMENT THRESHOLD AND ITS INTERACTIONS
WITH THE COLLECTION THRESHOLD 26
IV THE OPTIMAL THRESHOLD LOCATION 29
V EMPIRICAL RESULTS 31
A COST AND CLAIMS DATA 31
1 Passaic County, New Jersey 31
2. Arizona 34
3. Utah 34
B OPTIMAL THRESHOLD LEVEL 35
1 Passaic County, New Jersey 35
2. Arizona 36
3. Utah 36
v
CONTENTS (continued) Page
Chapter
VI CONCLUSION 37
REFERENCE 39
APPENDS 41
\i
TABLES
Table Page
VII SUMMARY OF OPTIMAL THRESHOLD LEVELS 38
A.l PARAMETERS FOR PASSAIC (COUNTY, NEW JERSEY) 51
A.2 SUMMARY OF OPTIMAL THRESHOLD LEVELS 52
vn
FIGURES
Figure Page
E 1 GENERIC CLAIMS PROCESS 10
m 1 ILLUSTRATIVE OPTIMAL THRESHOLD LEVEL 24
viu
EXECUTIVE SUMMARY
This report presents a model of the optimal level ar.d position of the Food Stamp Program
threshold for processing claims against clients A claim results from an ovenssuance of food stamps to
a client, who is obliged to refund the value of the ovenssuance to the state food stamp agency (FSA)
After the FSA detects an ovenssuance and establishes the ovenssuance as a claim, it undertakes a
sequence of activities designed to collect the claim while at the same time protecting the client's nghts
This is refened to as the claims collection process
The Food Stamp Act of 1977 stipulates that every ovenssuance must be established as a claim,
that the FSA must attempt to collect every claim that is the result of an intentional program violation by
the client, that is, fraud, and that the FSA is required to attempt the collection of other types of claims,
that is, non-fraud claims, only when it is cost-effective to do so In 1978, the Food and Consumer
Service regulations specified that a claim whose amount is equal to or greater than $35 is cost-effective
to collect In recent years several FSA's have wondered whether a $35 claim amount is still an
accurate indicator of cost-effectiveness
Some FSA also raised a second associated issue Since the statute requires all ovenssuances to be
established as claims, including claims that the FSA knows will fail the $35 collection threshold,
uncollected and in some cases uncollectible, claims have accumulated on the FSA's books
Accumulated uncollectible claims represents a burden to FSA because of the effort required to justify
wnting off a claim as a bad debt, thereby getting the claim off of the FSA's books This burden has
motivated some FSAs to suggest the use of threshold pnor to establishing the claim, as a substitute for
the existing threshold which is implemented after establishing the claim
This study estimates the optimal level of the current collection threshold in three FSAs -- Passaic
County NJ, Anzona. and Utah — based on a model of threshold cost-effectiveness The study also
investigates the feasibility of an establishment threshold, instead of the current collection threshold
The conclusions of the study are as follows:
OPTIMAL LOCATION
• From a cost perspecf.ve, the optimal location is close to the beginning of the claims
collection activity as possible, making the establishment threshold preferable to the
collection threshold
• An establishment threshold must be implemented in a manner that is consistent with the
Food Stamp Act of 1977, which requires all claims involving fraud to be pursued
• If an establishment threshold is not feasible, the existing collection threshold is at the
optimal location
ix
OPTIMAL LEVEL
The optimal level of a threshold at a given location is determined by the total downstream
cost that can be avoided for each claim that fails to pass the threshold and by the
proportion of the claim amount expected to be collected The threshold level which
ensures that every claim that passes the threshold is cost-effective is computed by
dividing total downstream costs avoided by claims that fail the threshold by the collecnon
rate
The opnmal collecnon thresholds for the states included in the study are uniformly many
times higher than the existing $35 collecnon threshold, and several nmes higher than the
$ 100 collecnon threshold used by several FSAs under FCS waivers
There is substannaJ FSA-to-FSA variation in the opnmal threshold level This suggests
that there is no uniform nanonal optimal threshold, and that FSAs be permitted to set
their own levels based on a cost study Perhaps this could be done through a waiver
process similar to that recently used by FCS in granting permission to several FSAs to
raise their collecnon thresholds to $100
Optimizing the claims process as a whole, rather than optimizing each claim individually,
results in threshold levels that are substantially below those resulting from opnmizing
individual claims In one FSA for which we can make such a comparison, the total
process optimization results in threshold levels that are approximately half as high as
those resulnng from individual claim optimization
I. INTRODUCTION
The Food Stamp Program (FSP), which provides benefits to low income households, is funded by the
U.S. Department of Agriculture and administered by state food stamp agencies (FSAs) The Food and
Consumer Service (FCS) of the U.S. Department of Agriculture (USDA) is responsible for the
development and implementation of national (FSP) policy This includes the promulganon of regulanon.
financial planning, review of and reimbursement for State expenditure, monitoring State FSP operations,
and evaluanng the Food Stamp Program
At the state level, administration of the FSP is the responsibility of State designated agencies (FSAs)
States have flexibility- within the constraints of the law and regulations to develop their own procedures in
operating the Program and. when the need arises to obtai" waivers of regulatory provisions from the
Department As a result the States vary considerably on many aspects of program operations, their cosib,
and their effecnveness
In September of 1993, as pan of it's ongoing effort to refine program operanons, reduce error and
deter fraud, FCS contracted with TASCON & Mathematica Policy Research to conduct an in-depth
examination of selected components of the claims process by collecting information on the effective State
management of Food Stamp recipient claims In order to do this, data was systemancally gathered and
analyzed in five sites - Alabama, Arizona, Arkansas, Utah and Passaic County, New Jersey
The study had three main objectives The first objective of this study was to evaluate the cost-effecnveness
of the $35 ovenssuance rule for the initiation of a collection action on an established non-fraud
claim Based on the outcome of that evaluation, we were to determine the appropriate collecnon
threshold.
The second objective was to estimate a dollar threshold for the establishment of food stamp recipient
claims and determine if the FCS should implement an establishment threshold
1
The third objective was to identify effective and cost-efficient alternative follow-up acnvines
(collection tools) on claims for which recoupment is not an option This included claims from over
issuance for active cases due to State Agency Error and for any delinquent inacnve cases
This report presents findings from objectives 1 and 2 which focus on estimating dollar thresholds for
initiating collecnon acnvines; and establishment of non-fraud claims It is a companion to Hilton and
Barokas (1995) which presents the findings from Objecnve 3 which addresses altemanve follow-up
acnvines on delinquent claims (collecnon tools)
A. BACKGROUND
Food Stamp benefits are issued to participating households each month The amount of food stamp
assistance provided to an eligible household is computed using a benefit formula, which takes into account
the household's size, income, assets, and specific types of expenditures This formula is applied when the
household applies to the FSP and at periodic recernficanons ' Occasionally the calculated benefit amount
is incorrect, that is, it is not equal to the benefit that results from the correct applicanon of the benefit
formula to the household's true circumstances Usually, such errors re ult in more food stamps being
issued to the household than is appropriate Many ovenssuances continue for several months since most
participating households are not recertified each month When an ovenssuance is detected, the Food
Stamp Agency (TSA) determines whether the ovenssuance occurred in past months, and computes the
cumulanve ovenssuance Ovenssuances are caused by tiiree types of errors:
The FSA applies the formula incorrectly or fails to act on information provided by the
household, which results in an agency error (AE).
'The purpose of recemficanon is to ensure that the size of the FSP benefits is appropriate to the
household's current circumstances, which may have changed since the time of application. A household
is usually recertified every six or twelve months, depending on the type of household. For example,
earners have a shorter certification period and SSI recipients a longer period. In some states, certain
households report their circumstances every month.
• The household makes an unintentional error in reporting its size, income, assets, or
expenses, which results an inadvertent household error (IHE)
• The household makes an intentional error in reporting its size, income, assets, or
expenses, which results in an intennonal Drogram violation (IPV)
The Food Stamp Act of 1977 (the Act) established that the recipient of an ovenssuance ;s obliged to
repay the amount of the ovenssuance to the FSA, regardless of the cause of the ovenssuance If the
ovenssuance is due to an AE or an IHE, the household is only liable for the amount of the ovenssuance
and usually is not subject to any other penalties. If the ovenssuance is due to an IPV, the household is
liable for the amount of the ovenssuance, the individual responsible for the violanon is disqualified from
the FSP for a specified number of months, and may be subject to addmonal penalties under the fraud
statutes of the state The number of months of disquahficanon depends on the seventy of the violanon and
whether the individual committed previous IPVs.
Section 13 of the Act requires FSAs to establish a claim against every household that receives an
ovenssuance The FSA is required to collect an established claim unless the FSA can demonstrate that
it is not cost-effective to do so Collecting a claim is cost-effective if the amourI collected exceeds the cost
of collection
Federal FSP regulanons specify some circumstances under which collection is deemed to be cost-effective
If the claim is caused by an IPV or an IHE and if the household is currently participating in the
FSP, coUecnon is deemed to be cost-effective The ranonale for this is that, if the household is currently
participating, the FSA can recover the claim by reducing the household's FSP benefit in the cunent and
future months Collection is considered to be virtually costless to the FSA, so collecting all such claims
is viewed as cost-effective.
If the claim is caused by an AE or an IHE and the household is not currently participanng in the FSP,
the claim is deemed to be cost-effecnve if its amount is $35 or more 2 The regulanons permit FSAs to
accumulate mulnple smaller claims against the household until the total is $35 or more This cost-effectiveness
test is known as the "collection threshold." Many claims in the FSAs reviewed in this study-are
uncollectible The most common reason for being uncollectible is that the FSA is unable to locate the
household Another common reason is that the household does not have enough income or assets from
which to collect the claim.3
Since these regulanons were first promulgated several FSAs have presented the Food and Consumer
Service (FCS) of the U.S. Department of Agriculture with two issues The first is that the $35 level at
which the collection threshold is set may no longer equal the cost of collecting an AE or IHE claim This
contennon is based on two beliefs The first is that the $35 figure never was an accurate amount of the
costs of collecting claims even when it was adopted in 1978 Second, the $35 figure has not been changed
in 17 years During this period, the average cost of collection has probably risen Thus, even if the
threshold level were appropriate in 1978 (which it may not have been), it is unlikely to still be appropnate
in 1995
This concern leads to the first objective of this study—to determine the appropriate level of the
collection threshold
The second issue raised by FSAs is that large numbers of uncollected claims have accumulated over
the years All overpayments must be established as a claim, even if they are predicted to fail the collecnon
threshold FSAs must track and account for each claim Each quarter, the FSAs must report to FCS the
2The FCS of the U.S. Department of Agriculture has recently permitted several FSAs to increase the
$35 limit to $100
Several FSAs define financial resources in terms of the household remaining independent of the
welfare system If collecting the claim would force the household back onto the FSP or other welfare
programs, the FSA considers the household's resources to be insufficient to collect the claim.
total number of claims, the number of new claims added to the books, and the number of claims that have
been closed, terminated, or compromised. When a household pays the claim in full, the claim is said to
be closed When the FSA determines that the claim is uncollect.ble and is expected to remain
uncollectible, the claim is first suspended and then terminated A suspended claim is one for which the
FSA connnues to carry the debt on its books, but does not actively try to collect A terminated claim is
written off the FSA's books as a bad debt.4 Occasionally the household pays a porcion of the claim but
is not expected to pay the outstanding portion. If the FSA determines that the remauang balance is
uncollecnble, the unpaid balance is considered a bad debt and the claim is said to be compromised
FSP regulations require that claims that are not closed (that is, not paid by the recipient) be kept open
for a minimum of three years to allow sufficient time for the household to be located or for the household's
financial circumstances to improve.5 Some FSAs hold these claims open for longer periods The FSA is
required to document in writing each determination that a claim be terminated or compromised Some
FSAs have found this process burdensome and prefer to leave the claim open
As a result, large numbers of uncollected claims have accumulated on the books of many FSAs
FSAs are faced with the choice of devoting scarce agency resources to either tracking and reporting
uncollected claims quarter after quarter or justifying terminations Several FSAs have suggested the
possibility of an establishment threshold While a collection threshold is imposed after a claim is
established and after the claim is determined to be caused by an AE or HE, an establishment threshold
is imposed before the claim is established Thus, the establishment threshold occurs upstream in the claims
process from the collection threshold With an establishment threshold, overissuances determined to be
too small for cost-effective collection, or known to be uncollectible, would not be established as claims
A terminated claim may be reactivated if the FSA receives relevant information after the claim is
terminated
Occasionally special circumstances permits an FSA to terminate a claim less than three years after it
was established
This approach would keep claims that are not cost effective from getting onto the books, thereby reducing
the admimstranve burden of tracking and justifying the terminanon of uncollected claims
This concern leads to the second objective of the study—to determine if the FCS should implement
an establishment threshold, and ifso. the appropriate level
B. METHODOLOGY
These issues were investigated in four State FSAs--Arizona (Western Regional Office), Arkansas
(SouthWest Regional Office), Alabama (SouthEast Regional Office), and Utah (Mountain Plains Regional
Office)-and one county FSA-Passaic County, New Jersey (Mid-Atlantic Regional Office) These FSAs
were selected based on the following criteria:
• Having a relatively high collecnon rate, the proportion of the total dollar amount of
claims that are collected
• Having a cust per claim, as proxied by the financial figures reported in the quarterly
FCS Form 269, that is near the median
• Using at least three of the following collection tools State tax intercept, wage
garnishment, property liens, and small claims court
In addition to these criteria, we attempted to have a representative distribution of State across FSP
Regional Offices Passaic County New Jersey was selected because it contains an urban area Arizona
was selected because it actively volunteered to participate in the study All five FSA integrated the FSP
claims collection process with that ofAFDC and other social programs Several of the selected FSAs used
Federal tax refund intercepts to collect FSP claims.6 Details of the claims processing in each FSA are
presented in TASCON, Inc (1995)
After determining that the none of the selected FSAs could supply the necessary cost data directly
from their accounting systems, we requested that each FSA estimate the costs of the major steps in the
bAll of the selected FSAs use Federal tax refu.id intercepts in 1995
6
1
United States
Department of
Agriculture
Food and
Consumer
Service
3101 Park Center Drive
Alexandria, VA 22302
SUBJECT: Release of "Guidelines for Use of Alternative Tools in the Collection of
Recipient Claims in the Food Stamp Program"
Attached is a copy of the final report, "Guidelines for Use of Alternative Tools in the
Collection of Recipient Claims in the Food Stamp Program."
By law, State agencies must establish claims for overpayment to participants and
attempt to collect every fraud claim, and all other claims when cost-effective. The
recipient of an overissuance must repay the amount regardless of its cause. Three
primary methods for paying claims are available: lump sum payment, installment
payments, and reduction of the monthly food stamp allotment (recoupment).
Households may choose one or a combination of methods.
States have flexibility within the constraints of the law and regulations, to develop their
own procedures to collect delinquent claims. This has resulted in the development of a
diversity of administrative practices. This report describes the strategies used to
manage delinquent Food Stamp recipient claims in five study sites - Alabama, Arizona,
Arkansas, Utah, and Passaic County, New Jersey.
The report was initiated and written under the direction of Dr. Sharron P. Cristofar. If
you have any questions or need more information, please contact her at the Office of
Analysis and Evaluation (703) 305-2131.
t*
claims collection process based on the amount and level of the direct labor required to perform each step
The steps were pre-establishment activines, post-establishment activities of the FSA, excluding activities
of other agencies and courts. Fair Hearings and Administrative Disqualificanon Hearings, criminal
invesnganons by agencies other than the FSA, and court All five FSAs provided a detailed description
of their claims collection process, and three FSAs (Arizona, Utah and Passaic County) provided cost data
sufficient to esnmate optimal thresholds
C. PLAN OF THE REPORT
Chapter n is an overview of the claims collection process Chapter III presents a theoretical model
for determining the appropriate level of a threshold Chapter TV presents a discussion of the opnmal
location, as opposed to level, of a threshold. Chapter V reviews the data on the costs of collecting claims
and the resulting collections, outlines the empirical model, and presents the estimates of the opnmal
threshold levels Chapter VI summarizes the report and provides conclusions.
II. ELEMENTS OF CLAIM COLLECTION COST
This chapter outlines the process by which the FSA and other State agencies handle FSP claims 7 The
steps represents the units in which claims collection costs are incurred There is wide state-to-state
vananon in these steps, so there is no single claims process Therefore, we have abstracted the general
features of the claims process found in the five States participating in this study The resulting genenc
claims process is illustrated in Figure II 1 For simplicity, we have limited Figure II 1 to IHE and IPV
claims
The claims collecnon process begins with the detecnon of a potennal ovenssuance Once the
caseworker detects an overpayment, the amount of the suspected overpayment is computed and verified
The amount of the overpayment is computed by entering the new household composinon, income, assets,
or expense data into the FSP benefit formula The new computed benefit amount is compared to the
benefit actually provided to the recipient, and the difference is considered to be the overpayment
Ovenssuances are typically identified by the FSP caseworker by comparing informanon on the
recipient's household composinon, income, assets, or expenses from several sources other than the
recipient to the analogous information provided by the recipient During the application and recertificanon
processes, the caseworker attempts to confirm the data reported by the recipient by asking the client for
documentation verifying the information, and occasionally by contacting the recipient's employer, landlord,
and bank The Income Eligibility and Verification System (IEVS) provides the caseworker with
employment and financial data on the recipient from other federal databases, including the Unemployment
C ompensanon data on earned income, federal tax return data on unearned income, and benefits from the
7This chapter presents the claims collection process as described by the five FSAs participating
in the study The procedures described herein do not necessarily reflect FCS policy or FSP
regulations FCS does not condone or advocate any deviations from policy or regulations
FIGURE Dl
IHE AND IPV CLAIMS PR CESS IN THE STUDY FSAf
Detect the
overpayment
1
Verify the information
leading to tbe claim, and
compute tbe amount of
the claim
< '
Establish the claim
[civil DOD-W—
Vfriudy
No-
Yti -Wcrtmiail) vex
10
FIGURE D-l (CONTINUED)
D3E AND D7V CLAIMS PROCESS IN THE STUDY FSAs
Demand letters
Fair Hearing
(Option of the
household)
Suspension and
termination
Recoupment
Suspension and
termination
Tax Intercept
Prepare a case -Y«- •No-
Court preceding
and court action
Wage
garnishment
FSA monitors
compliance
11
FIGURE D-l (CONTINUED)
ra£ AND D7V CLAIMS PROCESS TS THE STUDY FSAs
£
Referral to criminal
investigation agency
Criminal investigation
Criminal court
proceding and court
action
Criminal penalty
Administrative
Disqualification
Hearing investigation
Administrative
Disqualification
Hearing
Disqualification from
theFSP
Administrative
Disqualification
Hearing investigation
Administrative
Disqualification
Hearing
Disqualification from
theFSP
Civil non-fraud
collection activities
12
Supplemental Security Income (SSI) and Social Security programs The caseworker may also access
informanon on benefits from state programs, such as General Assistance
To verify the information that forms the basis for the overpayment, another caseworker or a claims
worker checks the calculations and sometimes collects additional information both from the recipient and
from other entities To compute the total amount of the overpayment, the caseworker idennfies the months
in which the recipient's information was not correct. The FSP benefit formula operates on a monthly
accounting period, using monthly income, household composition, and expenses as input, and computes
a monthly benefit The initial overpayment computation is a monthly figure The date of onset of the
monthly overpayments is used to compute the total overpayment. Because they are labor-intensive steps,
the verification step and the computanon of the claim amount can consume half of the resources required
to complete the entire claims process In one of the five states in the study, as many as half of the originally
detected overpayments were not established as claims following these steps.
There was significant variation across FSAs regarding the organizational units responsible for
venficanon and computation of amount Some FSAs used local FSP office personnel, and others used the
unit that is responsible for collecting claims 8 If these two steps were performed outside the claims unit,
the overpayment was then transferred to the jurisdiction of the claims unit.9
After these two steps are completed, the overpayment is established as a claim Establishing the claim
consists of entering the overpayment data int.. the (typically automated) claims tracking system From a
legal perspective, establishing the claim constitutes a statement that the agency believes the recipient is
For details on the claims processing of our individual study FSAs, see TASCON (1995)
Typically, the claims unit was responsible for collectin0 claims from recipients of all the welfare
programs in the state, and was housed in the state department ofhuman services. If the lrur.r' invesngation
was performed outside the claims unit, the claims unit may verify the information underlying the claim a
second time
13
indebted to the agency for the amount of the overpayment From a regulatory perspective, establishing the
claim makes it subject to federal regulations of uV claims collection process.
At this point, the FSA determines the cause of the overpayment -- an error by the FSA, or agency
error (AE); an error by the client, or inadvertent household error (IHE), or fraudulent behavior by the client,
or intentional program violation (PV). Suspected IPV claims are further subdivided into suspected
criminal fraud claims and non-criminal fraud claims
A. AE CLAIMS
There is significant variation in the way States handle AE claims (not shown in Figure II-1) Some
States limit their efforts to a demand letter to the client requesting a voluntary repayment If the client
refuses, no further action is taken Other States treat AE claims as if they were IHE claims In the demand
letter, the household is offered the opportunity to request a Fair Hearing At a Fair Hearing, the household
and the FSA will review the claim, typically, before an administrative law judge (ALT)10 The purpose of
such a hearing is to give the client the ability- to appeal the FSA's decision The client may contend, for
example, that the FSA's information about earned income, which was the basis for computing the
overpayment, is incorrect As a second example, the client might contend that the household never
received the overissued food stamps If the hearing authonty finds that the claim is invalid, the agency sets
the amount of the claim to zero and closes the claim The outcome of the Fair Hearing is that the claim is
judged to be legitimate, or to have no basis Further, the Fair Hearing determines the amount of the claim,
taking into account any information presented by the household at the hearing
TSP regulations do not require the presiding officer at a Fair Hearing to be an AU The Fair Hearing
process is used by many programs other than the FSP, some of which require an ALJ
14
B. IHE CLAIMS
The household is sent up to three or four demand letters explaining the nature of the claim and the
consequences of not making payment The client is advised of the right to a Fair Hearing in the lninal
demand letter " If the claim is found to be valid, either because the administrative law judge found in
favor of the agency or because the client waived the right to a Fair Hearing, the agency's next step is
usually to recoup the claim if the household is an active FSP participant. If the client is a current recipient,
the claim may be recouped, that is, recovered from future benefit payments. The recipient's benefit
amount is reduced by 10 percent for a nonfraud claim.12 Recoupment continues until the entire claim is
paid or the client leaves the FSP.
If the claim becomes delinquent and recoupment is not an option, the next step is usually to intercept
the client's income tax refund In 1994, twenty-one FSAs participate in the Federal income tax refund
intercept program, and the majority of states operate state income tax refund intercept programs The
Federal program has operated for three years and is expanding rapidly because of the advantages it
provides to FSAs it is inexpensive, it locates many clients who have moved to other states, and it results
in large amounts of repayments. The principal limitation of the tax refund intercept program is that, for
an individual claim, it is effective only once per year and only for individuals owed a refund.
If the claim remains open after a tax intercept is attempted, the agency must decide whether to take
the claim to municipal, county, or state court. The primary criterion used in this decision is the size of the
claim, but other factors may also be used, including the amount of the client's income and assets from
which the claim could be paid, the likelihood that forcing the client to pay the claim would cause him/her
to return to the FSP, subjective factors related to the likelihood that the court will find in favor of the
"At any point in the claims process after the household receives a demand letter, the household can
pay the claim Upon being paid, the FSA closes the claim.
'"If the recoupment amount computed in this way is less than $10, the amount is set to $10
15
agency, and the current workload of the staff responsible for preparing the case for court In die process
of notifying the cbent of the impending court session, some states allow the client to respond to the threat
of court action by paying the claim prior to the court session
If the agency decides to take the claim to court, the agency arranges for its legal representanon and
prepares the case against the client If the court finds in favor of the agency, the judge typically specifies
a payment anangement Such arrangements can include a simple payment schedule; wage garnishment,
if the client is employed, or a lien against, and seizure of, the client's property, typically the client's vehicle
While the agency can inform the court of the client's resources, the court, rather than the agency, chooses
the payment arrangement Even though the court specifies the payment arrangement, payments are
typically made direcdy to the FSA rather than the court If a client subject to a court order fails to pay the
claim on schedule, he/she is in contempt of court and may be subject to additional court acnon
If the agency decides not to send the claim to court, its only options are sending addinonal demand
letters and intercepting the client's income tax refund the following year After exhausting these opnons,
the agency may determine that the claim is uncollectible This determination is usually made because the
cbent cannot be located, the cbent does not have, and is not expected to have, the financial resources to pay
the debt, or the client refuses to pay the claim, but the ageno decides not to take the matter to court
At this point, the agency may follow any one of three strategies: leave the claim in an officially open
status but not acnvely pursue payment, suspend the claim by officially placing it in an open but inacnve
status, or terminate the claim by writing off the remaining balance as a bad debt. Federal regulations
require the minimum penod of suspension to be three years. Some FSAs adopt the first or second strategy
for many uncollecnble claims, because of the burden of documenting terminations resulting in the
accumulanon of numerous old, uncollectible claims.
16
C. IPV CLAIMS
The process for a non-cnminal IPV claim is similar to that of an IHE claim The primary difference
is that the hearing is an Administrative Disqualification Hearing (ADH) rather than a Fair Hearing While
a finding for the agency by a Fan Heanns' is ° determination that the claim is legal and valid, a finding
for the agency by an ADH also -sulu. 1 .^qualification of the client from the FSP for a specified
period of time As is the case for IHE claims, the individual has the opnon to waive the ADH In addition
to disqualifying the client through an ADH, the FSA may use, and generally does use, the same strategies
for settling the noncnrrunal IPV claim that are used for IHE claims
The process for a criminal IPV claim is quite different from that of any other type of claim Each
state has a unique set of criteria for determining if the claim is criminal. Typically, the claim must be the
result of fraudulent behavior of the client and must represent a large sum of money The definition of a
large claim is generally established in state criminal statutes and varies from state to state The state may
also use more subjective criteria to classify a claim as criminal fraud, such as the likelihood of success in
criminal court For example, a claim against a 65 year-old widow may not be treated as criminal, even if
the claim is large and the agency has evidence of fraud. On the other hand, a relatively small fraud claim
may be treated as criminal if the client is a repeat offender.
The client is subject to two types of penalties for committing criminal fraud—disqualification from the
FSP through an ADH, and criminal penalties determined by criminal prosecution.'3 Criminal claims are
typically referred to a different state agency, which prepares the criminal case for prosecution. The
criminal investigation agency reviews the information about the claim provided by the unit making the
referral and decides whether to recommend the case for criminal prosecution This decision is based on
the nature of the alleged fraud and may include some of the same factors used to determine whether the
l3Sometimes the FSA does not hold an ADH, and the criminal court will include a disqualification from
the FSP as a part of its judgement
17
collections unit takes noncriminal claims to court, including (1) the size of the claim; (2) the amount of the
client's resources, including both income and assets, from which he/she could pay the claim; (3) subjective
fa. on related the likelihood that the court will find in favor of the agency, and (4) the current workload
of the staff responsible for preparing the case for court. Claims that the criminal investigation agency
decides not to take on are returned to the collections unit, which processes them thereafter as noncnminal
IPV claims
If the criminal investigation agency accepts the claims, that agency conducts an investigation into the
criminal aspects of the case. In some states, while this investigation is occurring, the FSA is
simultaneously attempting to collect such claims through civil means, such as recoupment, letters, and tax
refund intercepts. In other states, no attempt to collect the claim is made through civil procedures. At the
conclusion of its investigation, the criminal investigation agency decides whether it believes the client
committed criminal fraud If the conclusion is that the client did not commit criminal fraud, the claim is
referred back to the FSA, which may process it as a noncnminal IPV claim
If the investigation concludes that the client committed fraud, the criminal investigation agency takes
the case to the state's attorney general, who prosecutes the criminal case in municipal, county, or state
court If the client is convicted, he/she is required to participate in an arrangement to pay the claim in
conjuncnon with the judgement Payment arrangements are generally the same as those in civil court: a
simple payment schedule; wage garnishment, if the client is employed; or a lien against, and seizure of,
the client's property. If the client is acquitted, the claim is referred back to the FSA, which may process
it as a noncnminal IPV claim.
Although the preceding description views the client as an individual, the Food Stamp Act has a more
complex definition of client. Food stamp benefits are computed for, and issued to, the household The
household is generally defined as all persons in the dwelling who pool their resources for the purchase and
preparation of food and who regularly eat together. The household can include more than one adult The
18
Food Stamp Act establishes that all adults in the household are jointly and severally liable for an FSP claim
This means that the collection agency can collect payment of a claim from any adult, or combination of
adults, in the household, even if that adult has left the original household and become a member of another
household. Typically, the agency initiates collection with the individual identified as the head of the
household or who signs the FSP application and recertification forms If that attempt fails, some FSAs
make similar efforts to coLect from other adult household members. These efforts can include tax refund
intercepts or recovery of the FSP benefit of another household that a liable adult has joined
19
/d
mmMH
III. THEORY OF OPTIMAL THRESHOLDS
This chapter presents the theoretical framework within which we estimate the cost-effectiveness of
both collection and establishment thresholds The development of the framework begins as we define the
cost-effectiveness of the entire claims process. Section A presents the basic concepts for the analysis of
both the optimal level and location of a claims threshold Section B presents a model of the opnmal
threshold level from the perspective of an individual claim Section C discusses deterrence and
unmeasured costs. Section D presents the interactions between a collections and an establishment
threshold
A. BASIC CONCEPTS
We begin by imagining a system of activities by which claims are processed. A claim can proceed
along any one of a variety of paths through the process The particular path the claim takes through the
process is determined by the characteristics of the claim, whether the client is continuing to receive food
stamps, and whether the client voluntarily settles the claim Some claims take short, simple paths, resulting
in little cost Others will take longer, complex paths, resulting in greater costs The cost-effecnveness of
an individual claim is the ratio of the savings to the program resulting from processing the claim to the cost
of processing the claim The savings to the program is the amount of the claim that is collected The cost
is the sum of the costs of each activity in the claim's path through the collection process In this study, we
measure both costs and savings from the perspective of the combined Federal and State budgets, and we
do not adopt the broader cost-benefit perspective of society as whole.
Somewhere near the beginning of the process, or "upstream," a threshold is implemented A
threshold separates the flow of new claims into two groups, those that pass the threshold, and those that
fail the threshold Claims that fail the threshold are sent down a relatively short path, and the FSA incurs
21
relatively small costs and collects none of the claim Claims that pass the threshold are sent down a
relatively long path, and the FSA incurs relatively large costs and may collect some or all of the claim
By assigning a claim to the "failure" group, the FSA forgoes any chance of collecnon in trade for a
reduced cost of processing the claim The threshold can be any rule that assigns each claim to one of the
two groups based on some characteristic, or combination of characteristics, of the claim Both the existing
collecnon threshold and the establishment threshold are based on the claim amount Claims whose
amounts are less than the threshold fail, and claims whose amounts are equal to or greater than the
threshold pass
Avoided cost is the difference between the downstream cost for claims that pass the threshold and the
downstream cost for claims that fail the threshold
B. OPTIMALITY FOR AN INDIVIDUAL CLAIM
This section presents a model of the opnmal threshold level for an individual claim A generalizanon
of this to a model of the optimal threshold level for the population of claims is presented in the Appendix
There are four altemanve definitions of optimality The threshold could be set at the level which:
1 Ensures every single claim is cost-effective, that is, the expected savings exceeds the expected
cost of each individual claim This is the definition of optimality used in this chapter
2 Ensures that the entire claims collection process has a cost-effectiveness equal to 10 This is the
definition of optimality used in the Appendix
3 Maximizes the cost-effectiveness ratio of the ennre claims collecnon process
4 Maximizes the "profit" of the claims collection process, that is, maximizes the surplus of total
savings over total costs of the process.
The advantage of the first definition is the simplicity of the resulting model. Its disadvantage is that it does
not optimize the cost-effectiveness of the entire claims process, but rather treats each claim individually
The remaining three definitions avoid this disadvantage by optimizing the process as a whole The
22
disadvantage of the third definition is that the cost-effectiveness rano may be maximized when the
threshold is set at an extremely high level and very few claims, perhaps a single claim, are pursued
The second and fourth definitions have no such disadvantages We have chosen the second definition
for the following reasons. First, the process under analysis is that of public agency, which is not required
by competitive pressure to maximize profits Second, FCS has expressed a preference for the threshold
to be set as low as possible short of causing the cost-effecfiveness ratio to fall below 10 This preference
appears to be based on a recogmnon of the unmeasured deterrent effect of vigorous claims collecnon.
We begin by considering the existing collection threshold The collecnon threshold is based on the
claim amount, and requires the following assumpnons:14
• The amount collected, S, from a claim is an increasing funcnon of the amount. A, of the
claim.
• S = 0 for claims that fail the threshold
• The cost, C, of processing a claim is greater for claims that pass the threshold than for
claims that fail the threshold
• The cost, C, of processing a claim that passes the threshold is not a funcnon of the size
of the claim (or at least does not increase as fast as S with increases in A)
These assumptions are illustrated in Figure IH1 The figure shows a hypothencal plot of both collecnons,
S. (the thick line) and cost, C, of claims that pass the threshold (the thin line) by the size of claim
Collections and cost are measured along the Y axis, and the size of the claim is measured along the X axis
Collecnons are shown to increase as the size of the claim increases. Each of the five FSAs
participating in this study' expects, when a claim is established, to collect a certain percentage of the claim
l4For clarity, both savings and costs are presented as if they are scalars In fact, each is an expected
value of the probability density funcnon of savings and costs
23
FIGURE m.l
ILLUSTRATIVE OPTIMAL THRESHOLD LEVEL
24
BEST COPY AVAILABLE
amount, known as the collection rate With such a rule, larger claims result in larger collecnons In
contrast, costs along the horizontal thin line do not vary with claim size.15
Costs equal collections if both are equal to C at claim size K, as shown in Figure ID. 1 If a claim is
smaller than K (to the left of K), costs exceed collections This makes attempting to collect the claim not
cost-effective If the claim is greater than K (to the right of K), collections exceed costs and the claims
process is cost-effective If the threshold is set equal to K, then claims that are not cost-effecflve will fail,
and claims that are cost-effective will pass This makes K the optimal threshold level An algebraic
version of this model is presented in the Appendix
C. DETERRENCE AND UNMEASURED COSTS
Deterrence occurs when a client who is considering committing a program violation takes the
likelihood of getting caught into account The more intensive is the process of detecting and collecting,
the greater is the likelihood of getting caught, and the less likely the client is to commit the IPV By
preventing some violations from being committed, the claims process produces a saving to the FSP that
is not captured in the model presented in Section B It is also possible that an active claims detection and
collection process "deters" some inadvertent errors. For example, if the household discovers its error, it
may be more likely to report it in order to avoid a claim.
Unfortunately, in a situation in which laws or regulations are being enforced, it is generally not
possible to measure the amount of deterrence resulting from that enforcement Measuring deterrence
requires that the inclination of clients subject to the laws and regulations about FSP claims to be compared
to statistically equivalent clients who are not subject to such laws and regulations Since forming such a
control group is not possible, the deterrent effect of the claims process cannot be measured
In practice, larger claims are more likely to go to court, thereby increasing the cost of collecting them
This fact is not represented in Figure ID. 1
25
Deterrence enters the model of the optima] threshold level as an indirect saving to the FSP beyond the
direct saving from claims collections Adding indirect savings to direct savings shifts the savings line in
Figure III 1 upward, which causes the intersection of the savings line and the cost line to shift to the left
Thus, adding deterrence to the model results in the optima] threshold being reduced.
Similarly, as detailed below in Chapter V, it is not feasible to measure all the costs associated with
collecting claims While measuring the costs borne by the FSA is straightforward, some costs are borne
by other public agencies, such as municipal and county courts and the state agency that is responsible for
criminal investigations and prosecutions Adding unmeasured costs to measured costs shifts the cost line
of Figure ID 1 upward, causing the intersection of the cost line and the savings line to move to the right
Thus, accounting for unmeasured costs results in the opnmal threshold to be increased
The impacts of deterrence and unmeasured costs on the optimal threshold level are offsetting One
can only speculate about the direction and magnitude of the net impact of both factors Both factors apply
only to a minority of claims The deterrence factor applies only to IPV claims, and the unmeasured cost
factor applies only to claims that go to court and/or require a criminal investigation
D. ESTABLISHMENT THRESHOLD AND ITS INTERACTIONS WITH THE COLLECTION
THRESHOLD
An establishment threshold is similar to the existing collection threshold except that it is implemented
at an earlier step in the claims process, that is, before the overpayment is established as a claim This
upstream posinon has the following impacts, relative to the collection threshold:
• The cost associated with claims that fail the threshold is smaller because the claim does
not have to be suspended and terminated.
• The cost of activmes downstream from the threshold is larger because there are more
activines downstream from the threshold
• The expected savings for claims that pass the threshold is unchanged, which in rum
depend on the likelihood of collecting the claim.
26
These impacts mean that the optimal level of an establishment threshold is higher than that of a collection
threshold
Some have viewed the collection threshold and the establishment threshold as operating independently
of each other If this were true, the establishment threshold could be implemented along with the existing
collecnon threshold and the level of each set independently of the other This view is not correct
If the two thresholds are both implemented, they interact with each other in several ways First, the
level of the collection threshold depends on the level of the establishment threshold If the establishment
threshold is set higher than the collection threshold, the collection threshold will be ineffective since the
collection threshold is imposed downstream in the claims process The smallest claim to reach the
collecnon threshold would have a value equal to the establishment threshold. No claim would reach the
collection threshold with a value above the collection threshold but below the establishment threshold. In
effect, the establishment threshold is the lowest the collection threshold may be meaningfully set.
Second, the level of the establishment threshold depends on the level of the collecnon threshold
Setting the establishment threshold lower than the collection threshold results in a group of nonfraud claims
being established but not pursued because they fail the collection threshold The current problem of the
unnecessary administrative burden on FSAs associated with establishing claims that are never pursued,
discussed in Chapter I, would not be solved Avoiding such unpursued claims requires that the
establishment threshold be set at a level equal to or above the collection threshold.
In summary, the establishment threshold should not be lower than the collecnon threshold, and the
collecnon threshold should not be lower than the establishment threshold These conditions can be true
only if the two thresholds are set at the same level
The problem with implementing both thresholds and setting them at the same level is that the
thresholds have different opnmal levels, as demonstrated above Specifically, the opnmal level of the
establishment threshold is higher than that of the collection threshold. Therefore, if the two thresholds are
27
set at the same level, as we have shown they should be, one of the two thresholds must be set at a
suboptimal level
The solution to this problem is to view the two thresholds as substitutes rather than complements
One threshold should be implemented, not both ofthem In other words, evaluating thresholds first requires
identifying the best location for the single threshold within the claims process, and then determining the
best level of that threshold
28
IV. THE OPTIMAL THRESHOLD LOCATION
This chapter identifies the optimal location of a threshold, and examines the parameters that determine
that location.
The model presented in Chapter III allows us to compute the optimal level of a collection threshold
and the optimal level of an establishment threshold. The optimal level of each is defined as that level at
which cosi-effective claims pass the threshold and claims that are not cost-effective fail the threshold Each
threshold is equally optimal.
In order to compare one threshold location (collection threshold) to the other (establishment threshold),
we must use another criterion One such possibility is the overall cost of the claims process The farther
upstream is the threshold, the smaller is the total cost of the claims process This is because fewer steps
of the process have been performed for claims that fail the threshold This reasoning leads to the
conclusion that the establishment threshold is preferable to the collection threshold. This finding is
demonstrated algebraically in the Appendix.
The feasibility of replacing the collection threshold with an establishment threshold depends on a legal,
rather than an economic, argument. In some FSAs in our study, the claim is determined to be fraud or
nonfraud after establishment. In those FSAs, the establishment threshold necessarily applies both to fraud
and nonfraud claims since it is applied before the fraud status is determined. This means that claims
.vhose amount is less than the threshold but which would ultimately be determined to involve fraud would
not be established, and therefore would not be collected. The Food Stamp Act of 1977 (Act) requires that
every claim involving fraud be pursued. If ovenssuances are viewed as possibly involving fraud, but the
FSA has not discovered the information required to make the determination, then an implementanon
threshold may not be consistent with the Act. On the other hand, if ovenssuances are assumed not to
29
involve fraud until future investigation reveals evidence of fraud, then there may be no inconsistency
between the Act and an establishment threshold l6
Another factor in the legal environment is the common practice among state agencies of applying an
informal cost-effectiveness criterion to criminal prosecutions Typically, the FSA refers a claim that it
suspects of involving a crime to another State agency such as the office of the attorney general That
agency reviews the claim and decides whether or not to prosecute That decision is based on several,
generally undocumented, factors, including the size of the claim State attorneys do not usually have
enough resources to prosecute all the cases referred to them Therefore, they must select specific cases
to prosecute In this selection process, large claims are typically given a higher priority than small claims
The selection is based on a cost-effectiveness criterion Generally speaking, small claims are seen as not
worth the effort to prosecute. Claims that are not prosecuted are returned to the FSA, which treats them
as non-cnrrunal fraud claims.
In summary, on economic, cost-effecnveness grounds, an establishment threshold is preferable to a
collection threshold, because it includes a larger portion of the total claims collecnon costs in the
opnmization process Second, there is a possible conflict between an implementanon threshold and the
Food Stamp Act of 1977 In the future, FCS will determine whether this is a real conflict Third, the
behavior of some other State agencies is consistent with the applicanon of a cost effectiveness threshold
to claims involving fraud
l6MPR does not offer a legal opinion in this matter. The issue of consistency between an
establishment threshold and the Act remains to be determined by FCS
30
tm
V. EMPIRICAL RESULTS
This chapter presents the empirical estimates of the optimal threshold level in Passaic County, New
Jersey, Arizona, and Utah
A. COST AND CLAIMS DATA
1. Passaic County, New Jersey
The Passaic County, New Jersey, Board of Social Services esnmated the cost of each major step ir
processing a claim The process includes the following steps.|7 After each step, we present the cost per
claim processed and the number of claims processed in a typical month The process begins with 239
overpayments detected within a typical month
Step I. Pre-establishment acnvines, including detecting the overpayment, reviewing the case file,
and verifying the basis for the overpayment These costs vary by the source of the refen-al
Each claim is processed in one of the following five acnvities.
A Referral from a Food Stamp/AFDC redeterminarion
1. Cost per claim18-$79
2 Number of claims - 60
B Referral from a Food Stamp-only redetermination
1 Cost per claim-$100
2. Number of claims- 37
These steps differ in minor ways from the generic description of the claims process presented in
Chapter JJ These differences are typical of the state-to-state or county-to-county vananon in the claims
process
1 8The cost per claim figures represented below are computed on the basis of the number of claims
processed in that step, rather than based on 239 claims
31
C. Referral from the Food Stamp/AFDC Income and Eligibility Venficanon System
1. Cost per claim - $75
2. Number of claims - 42
D Referral from the Food Stamp Income and Eligibility Verification System
1. Cost per claim-$81
2. Number of claims - 91
E. Referral from the fraud unit
1. Cost per claim-$291
2. Number of claims - 9
Step II. Establishment and initial collection activities, including calculating of the amount of the
claim, sending demand letters to the client, and setting up and maintaining the case tracking
records. In contrast to Step I, each claim is processed through all of the following acnvines
A Calculation of amount of the claim
1 Cost per claim - $317
2. Number of claims - 23 9
B. Demand letters
1. Cost per claim - $6
2 Number of claims - 191
C Maintaining the account
1. Cost per claim - $36
2. Number of claims -191
Step III. Federal income tax refund intercept
1 Cost per claim - $72
2. Number of claims - 48
32
Step IV. Process payments from the client
1 Cost per claim - $112
2. Number of claims -105
Step V. Hearings, including both Fair Hearings and Disqualification Hearings
1. Cost per claim19 - $1
2. Number of claims -15
Step VI. Disqualifying the client
1 Cost per claim - $44
2. Number of claims - 38
Step VII. Court action, including both civil and criminal
1. Cost per claim - $92
2 Number of claims -115
The cost figures are based on the FSA staff time spent on an activity for a claim, accounting for the
salary, the benefits, indirect costs, and general and administrative costs associated with that person Costs
are limited to those incurred by the FSA, costs incurred by the state agencies responsible for hearings and
for criminal investigations and prosecutions are excluded. The court costs in Step VTJ are limited to those
incurred by the FSA in preparing the case and representing the FSA in court. These costs do not represent
the ennre cost of a court proceeding
The Passaic County FSA did not provide cost figures for suspending and terminating claims. We
have assumed that each activity costs the same as Step DC activities.
These costs are so low because most of the costs of hearings are not borne by the FSA
33
2. Arizona
The cost data supplied by the FSA in Arizona was less detailed than those from New Jersey The total
cost in FT 1994 for processing overpayments/claims was $4,272,826, of which $2,735,383, or 64 percent,
reflected pre-establishment processing of 25,071 overpayments, and the remaining $1,537,443 reflected
post-establishment processing of 15,464 claims Thirty-eight percent of the detected overpayments were
not established as claims.
From these figures, we compute the average total cost per overpayment was $170, the average pre-establishment
cost per overpayment is $109, and the average post-establishment cost per established claim
is $99
The total amount of claims outstanding at the end of the year was $5,106,961. The total amount
collected during the year was $2,181,074 We use the ratio of collections during the year to the amount
of claims at the end of the year, 43 percent, as an estimate of the collection rate
Data on the number of claims that fail the collection threshold, and the cost of such claims, are not
available We assume that the percent of established claims that fail the threshold is the same as that of
Passaic County, 20 percent, and we assume the cost of suspending and terminating each failed claim is half
that in Passaic County, $20 Combining these assumptions with the post-establishment cost per claim
yields an est'mate of the post-establishment cost per claim that passes the threshold of $ 119.
3. Utah
The Utah FSA establishes 260 claims in the average month. The pre-establishment activities cost
$48,684 per month. Post-establishment processing, not including hearings or court costs, is $47,708 per
month In an average month, 47 DQH hearings and 11 fair hearings are completed, at a cost of $ 13,831
to the FSA In an average month, four claims are submitted to court, at a cost of $6,548 to the FSA The
Utah FSA spends a total of $116,773 per month on the claims process.
34
B. OPTIMAL THRESHOLD LEVEL
Chapter ID shows thai a threshold level that ensures that each claim that passes the threshold is cost-effective
is the avoided downstream costs divided by the collection rate The empirical approach to
estimating the optimal threshold level is to measure the cost of processing a claim that passes the threshold,
the cost of processing a claim that fails the threshold, and measure the average proportion of the claim
amount that is collected
1. Passaic County, New Jersey
From the cost data provided by the Passaic County FSA we esnmate the average post-establishment
cost per claim of $541. To this figure we add the average pre-establishment cost per claim of $90,
producing a cost per claim for the entire process of $631 If an establishment threshold were implemented
in Passaic County the per claim cost of the steps downstream from the threshold would be $541
Before it can be used to compute the optimal threshold level this figure must be adjusted because it
reflects the influence of the existing $35 collection threshold. We assume that the 48 claims lost between
Steps IIA and IIB in the Passaic County cost data failed the collection threshold, and that if these claims
had not been screened out by the threshold, the> would have incurred the same average cost as the claims
that passed the threshold The resulting adjusted post-establishment per claim cost is $587
Using a 60 percent collection rate, the optimal level for an establishment threshold is $978
($587/0.60).20
The per claim cost of the steps downstream from the existing collection threshold is $251 Using the
same 60 percent collection rate, the optimal level for a collection threshold is $419 This level is less than
2(The algebraic rationale for dividing the cost by the collection rate is presented in the Appendix, pages
43 and 44 The collection rate should include not only the payments made by households, but also
foregone FSP benefits to households that disqualified from die program at an ADH
35
half of the optimal establishment threshold because calculating the amount of the claim occurs after
establishment and costs more than half of the post-establishment cost of processing the claim
2. Arizona
In Section A, we showed that the average cost to the Arizona FSA per ovenssuance is $170. In order
to correct this figure for the existing $35 collection threshold, we assume that 20 percent of the established
claims fail the collection threshold, and it costs $20 to process each of these after they fail Admimstranve
records show that the collection rate in Arizona is 43 percent A cost of $119 per claim that passes the
collection threshold means that the optimal collection threshold is $277 ($119/0.43). Unfortunately, the
available data from Arizona does not support the estimation of the optimal establishment threshold The
optimal establishment threshold lies between $277 and $428 Since relatively little happens to a claim
between establishment and the collection threshold in Arizona, since the verification, computation of the
claim amount, and determination of the type of claim occurs before establishment, the opnmal
establishment threshold will be only slightly above the $277 optimal collection threshold
3. Utah
In Utah, the average total cost per claim is $449. In order to correct this figure for the existing $35
collection threshold, we assume that 20 percent of the established claims fail the collection threshold, and
that it costs $20 to process each of these after they fail Administrative records indicate that the claims
collection rate in Utah is 68 percent These figures yield an optimal ovenssuance threshold level of $658
Using these same assumptions about claims that fail the existing collection threshold, we calculate the cost
downstream from the collection threshold per claim that passes the threshold at $322, yielding an opnmal
collecnon threshold level of $472
36
VI. CONCLUSION
This study identifies the optimal location and level of a threshold in the FSP State claims coliecnon
process To minimize the cost of the claims process, the optimal location is as far "upstream" in the claims
process as possible It is not feasible to apply a threshold much before establishing a claim because that
is the first point at which the FSA determines the claim is legitimate and computes the amount of the claim
Thus, from this economic perspective, an establishment threshold is preferable to a collection threshold
The Food Stamp Act of 1977 requires the FSA to pursue every claim involving fraud. This raises the
concern that an implementation threshold might screen out some claims that would be shown to involve
fraud by investigation conducted downstream from the threshold. By implementing the threshold at such
an early point in the process, that is, before the FSA identifies the type of claim (fraud/nonfraud), the
threshold would prevent the FSA from pursuing claims that, in fact, involve fraud. One possible resolution
of this problem might be for FCS to require FSAs to presume all overissuances are potennal nonfraud
claims, until later invesnganon reveals evidence of fraud.21
If no threshold may be applied to fraud claims, the earliest point that a threshold can be implemented
is after the claim has been established, after the fraud/nonfraud determination is made, and before any
coliecnon activities are performed. Thus, the existing collections threshold is at the opnmal locanon if a
threshold may not be applied to fraud claims
The opnmal level of a threshold at a given locanon is determined by the total downstream cost that
can be avoided for each claim that fails to pass the threshold and by the proportion of the claim amount
expected to be collected The threshold level which ensures that every claim that passes the threshold is
2IMPR is not offering a legal opinion in this matter. This issue can be resolved only by FCS
seeking legal counsel
37
cost-effective is computed by dividing total downstream costs avoided by claims that fail the threshold by
tne collection rate.
Several patterns emerge from the resulting estimates, summarized in Table VI 1 The optimal
collection thresholds are uniformly many times higher than the existing $35 collection threshold, and
several times higher than the $100 collection threshold used by several FSAs under FCS waivers This
means that either the existing collection thresholds are too low or that the deterrent value of the claims
process is much larger than the unmeasured costs of the claims process. As discussed in Chapter m, such
a circumstance results in lowering the optimal threshold level below the estimates presented in Table VI1
Since both deterrence and unmeasured collection costs are not quantified, we do not know which is larger
It seems unlikely that deterrence is so much larger than unmeasured collection costs that the true opnmal
is a small fracnon of the esnmated opnmal.
A second conclusion from Table VI1 is that there is substannal FSA-to-FSA vananon in the opnmal
threshold level This suggests that there is no uniform nanonal optimal threshold, and that FSAs be
permitted to set their own levels based on a cost study Perhaps this could be done through a waiver
process similar to that recently used by FCS in granting permission to several FSAs to raise their collecnon
thresholds to $100
TABLE VI. 1
SUMMARY OF OPTIMAL THRESHOLD LEVELS
Passaic County, NJ
Arizona
Utah
Collection Establishment
Threshold Threshold
$419 $978
277 NA
472 NA
38
REFERENCE
TASCON, Inc "Guidelines for Selecting and Supplementing Cost Effective Collection Tools for
Delinquent Claims " Bethesda, MD: TASCON, Inc., 1995
39
40 wm& m\
APPENDIX
¥K mm»
A MODEL OF OPTIMAL THRESHOLDS
In the Appendix we develop a model of optimality from the perspective of the entire population of
claims, rather than the individual claim We begin by presenting the model for individual claim optimality
A. OPTIMALITY FOR AN INDIVIDUAL CLAIM
For an individual claim that passes the threshold, let
S, • P, * A,
where S is the amount collected, P, is the collection rate, A is the amount of the claim, arid I is an mdex
of the claim For a claim that fails the threshold,
The cost of a claim that passes the threshold, A > T where T is the level of the threshold, is
where C is the cost and B2 is the constant cost per claim For a claim that fails the threshold, A < T,
c, = P,
where p3 < P2.
The claim amount A that generates collections equal to the cost of collecting the claim is
A = i-2 li Equation A.\
43
where the numerator is the cost avoided by a claim that fails the threshold, and the denominator is the
collection rate Setting the threshold at A guarantees that every claim that passes it will be cost effecnve
B. OPTIMAL THRESHOLD FOR THE PROCESS
Requiring every single claim that passes the threchtld to be cost-effective may be an unnecessarily
stringent standard. Claims whose amount exceeds.^ generate more collections than they cost to collect
When total collections and total cost are summed over all claims that pass the threshold, collections exceed
costs, ar.H the cost-effectiveness ratio of the entire process is above 1.0.
In this secion we derive an expression for the threshold that optimizes the claims process as a whole,
rather than optimizing each claim individually
Let us define the total gross savings from the entire population of claims to be
*« = P. • [ H • M) <H
A-T
where the amount, A. of the claim is distributed in the population normally with mean u and variance o*
The expression/^ is the density function of that normal distribution and T is the level of the threshold
The total cost of the process is
A--T
C„*P2 • [M)d4 +P3 • [AA)d4
-oo A'T
where the first term on the right side of the equation is the total cost of claims that fail the threshold, and
the second term is the total cost of claims that pass the threshold As we assumed in Section B, P3 < P:
The integral in the first term is the area under the probability density function of A 10 the left of T The
integral in the second term is the area under the probability density function to the right of T.
44
Having expressions for total savings and total costs, we now define optimality of the claims process
as a whole in two ways The first and most straightforward way is to set the threshold at the level at which
net savings, that is. the difference between savings and cost, is maximized, or
max (S - C)
This is the "profit maximizing" definition of optimality, in that it maximizes the excess of collections over
the costs of collecting claims. If the FSA were a profit maximizing firm, this would be the appropriate
definition of optimality
The claims collection component of the FSA, however, is a public agency performing a law and
regulation enforcement function, rather than a profit maximizing firm. Much of the value of public
expenditures on law and regulation enforcement is their deterrent effect. The knowledge that the public
agency might detect and prosecute an intentional program violation deters some FSP clients from
committing such a violation Although it is virtually impossible to measure the number of violanons that
were prevented by operating a claims collection process, we assume that the value of deterrence decreases
as the threshold is set at higher levels As the threshold is set at higher levels, the potential violator views
the likelihood of violation being pursued decreasing 22 As shown in Section D below, adding the value
of deterrence to the expression for Stot generally results in lowering the optimal threshold
Since the effect of adding deterrence is to lower the optimal threshold level and since the value of
deterrence is not measured, we approximate the effect of adding deterrence to the model by adopting an
alternative definition of the optimal threshold level. Instead of setting the threshold level such that S - C
is maximized, we set the level such that total savings equal total costs, or S - C = 0.
In a law and regulation enforcement setting, it is important that information about the threshold be
withheld from the public in general, and clients in particular. Withholding this type of information is
practiced by many similar Federal agencies, including the IRS
45
In order to evaluate the integrals in ST0T and CT0T, we must assume that claim amounts are
distributed in a particular way For each distribution of claim amounts from the three states for which we
have claim data-Alabama, Arkansas, and Passaic County, New Jersey-
Mean A > Median A > Mode A
We also know that the distnbution ofA cannot extend below A = 0. These characteristics suggest that the
probability density function of A is lognormal,
Z = In (A)
where Z is distributed N (ur, o.).
The mean of A, \xA, and standard deviation of A, aA, are measured from the claims data from the
three states listed above. Fromu^ and oA, u2 ando, for the analogous normal distribution are
computed:23
Oj = In
°A + »A
, ^ ,
1
u2 * In \iA - - In 't +
2\
,
2
»A 1
We next convert z into a standard unit normal distribution, u:
2 =
Z - n2
Z7 - ^2
Where Zj and ZT are the values of z and Z that correspond to the threshold T. Setting ST0T = CT0T
and making these substitutions yields
23These expressions are derived from equations 6 1 and 6 2 in Johnson and Kotz (1970), p 115.
46
P, • / (A • M) dz = P2 • F(zT) + P3(l - F(rr)) £^fl//o« A2
r=i
where ffz) is the standard normal probability function and F(z) is the cumulative standard normal function
The savings on the left side of the Equation A.2 can be expressed in terms of the average amount of claims
greater than the threshold times the proportion of claims above the threshold:
p, * TT * (1 - F(zT)) - [P2 * F(zT) ♦ P3(l - F(zT)j\ = 0 Equation A3
where AT is the mean value of A for claims that pass the threshold:24
AT = exp A*T)
1 - F(:,)
* o. + U2
Equation A. 3 is expressed in terms of the p.s, f^zj, F(zt), a,, and M. All of these terms are known
except z,. Equation A3 is solved for z, numerically, yielding z,. This standard normal term is converted
to Z, by:
ZT = (*T * °2> + U2
Finally ZT is converted to T by
T = eh
T is the threshold at which the cost-effectiveness of the claims process is 10
24Foi X distributed N(u, a) and truncated from below at T:
E(X) = AH
(1 - F(T))
* 07 + u
47
C UNMEASURED COSTS
As detailed in Chapter V, it is not feasible to measure all the costs associated with collecting claims
While measuring the costs borne by the FSA is straightforward, some costs are borne by other public
agencies, such as municipal and county courts and the state agency that is responsible for criminal
investigations and prosecutions Unmeasured costs enter the model as an addmonal term in the cost
equation:
cM = k • / yw U\ + [ft ♦ P,) • 1M) u]
-oo A -T
where P, is the unmeasured cost per claim that passes the threshold The addinon of unmeasured costs
to the model increases the optimal level of the threshold
D. OPTIMAL THRESHOLD LOCATION
Let us divide the claims process into two mutually exclusive and exhaustive portions, the pornon
occurring prior to the collecnon threshold, the upstream portion, and the portion occurring after the
threshold, the downstream portion. Through the model described in Chapter ITJ, the downstream portion
is optimized, such that downstream costs, CDS, equal collections, S, all of which are assumed to be
downstream of the threshold If the FSA incurs substantial costs upstream from the threshold, Cus. then
the enure claims process is not optimized, that is, S < CDS + Cus The FSA spends more to collect claims
than it collects from that process. The larger is Cus the farther from cost-effectiveness will be the claims
process
In order to bring total savings into equality with total cost, one can either raise the level of the
threshold or move the threshold upstream. Assuming for simplicity that A is distributed normally, rather
than lognormally, the former strategy involves finding the threshold level, T, such that
48
ST0T = p2 . Fit) ♦ p,.(l -F{t)) * Cus
where / = 1 and Cus is the cost of processing a claim before the threshold is applied to the
°A
claim. CkJ is not multiplied by a term involving the cumulative normal because all claims, both those that
subsequently pass and those that fail the threshold, result in these costs.
The second strategy is moving the threshold upstream to the beginning of the process This involves
setting Tsuch that
SWT = P2 • F(t) ♦ (P, ♦ Cus) - (1 - FiO)
Claims that fail the threshold result in costs of P2, and claims that pass the threshold result in costs of
P3 * Cm.
The optimal threshold level under each strategy equates savings to costs, producing a cost-effectiveness
rano of 1 0 Even though both strategies (raising the threshold and moving the threshold) are
equally opnmal from the perspecttve of their cost-effectiveness ratio, they are not equivalent in terms of
total costs. The second strategy (moving the threshold) results in fewer claims with the extra Cus costs
than does the first strategy (raising the threshold) Ifwe use the same threshold level for the two strategies,
the difference between the total cost under the first strategy, C^j.,, and the total cost under the second
strategy, Crar2,is
Thus, even though both strategies have a cost-effectiveness rano of 1.0, moving the threshold upstream
reduces the overall cost of claims collection, and is therefore preferable to leaving the threshold in its
original position and increasing its level This result holds if A is distributed lognormally
49
E. RESULTS
The threshold levels computed in Section V.B guarantee that every claim that passes the threshold is
cost-effective Such a threshold results in the cost-effecnveness rano of the ennre claims process
exceeding I 0 This is because all claims whose amount exceeds the threshold generate more collecnons
than they cost to collect Therefore the threshold can be set at levels below those computed in Secnon B
and still result in a cost-effecnve claims process The model presented in the Appendix generates the
threshold at which the cost-effectiveness of the entire claims process is 1.0
Passaic County, New Jersey is the only participating FSA that has provided both cost data, presented
in Section A, as well as claims data from which the mean and standard devianon of the distnbunon of
claim amount can be computed Trie parameters used to solve Equation A3 for the opnmal threshold are
listed in Table A 1
We have assumed the cost of suspending and terminating a claim that fails the threshold is $40 for
the collection threshold This cost is taken to be $0 for the establishment threshold, since one of the
50
TABLE A 1
PARAMETERS FOR PASSAIC COUNTY, NEW JERSEY
Parameter Collection Threshold Establishment Threshold
Cost of a Claim that Passes the
Threshold $251 $587
Cost of a Claim that Fails the
Threshold $40 $0
Collecnon Rate 60% 60%
Mean Claim Amount $339 $339
Standard Deviation of the
Claim Amount $543 $543
Opnmal threshold $204 $638
advantages of such a threshold is that claims that fail the threshold are never established as claims, and so
need not be suspended and terminated
The collection rate is set to the 1994 rate of 60 percent. The actual collecnon rate may be a function
of the claim amount, and indeed may be endogenous, depending on the level and locanon of the threshold
This complication has not yet been incorporated into the model. If the collection rate rises with the claim
amount, the true opnmal threshold is smaller than our estimate.
The mean and standard deviation of the claim amounts are measured from claims data provided by
Passaic County These figures are also endogenous As the threshold level changes, the distnbunon of
claim amounts change, affecting the mean and standard deviation of the claim amount. We have ignored
this endogeneity for this esnmanon.
51
The final row of Table A. 1 presents the optimal threshold level that solves Equanon A3 The opnmal
collection threshold is $204, and the opnmal establishment threshold is $638 For each type of threshold,
the threshold that makes the cost-effectivenesi; rano of the ennre claims process equal to 1.0 is about ha'f
of the threshold that ensures that every claim is cost-effective.
Table A.2 summarizes all the estimated optimal threshold levels. Two sets of esnmates are provided
for Passaic County, one opnnuzing each claim, and the other opnmizing the claims process as a whole
The esnmates for Arizona and Utah optimize each claim individually
TABLE A.2
SUMMARY OF OPTIMAL THRESHOLD LEVELS
Collection
Threshold
Establishment
Threshold
Passaic County, NJ
Individual Claim Opnmality
Total Process Opnmality
$419
204
$978
638
Arizona 277 NA
Utah 472 NA
52