Recoupment in the Food stamp program

              •
..
"
RECOUPMENT
in the
FOOD STAMP PROGRAM
United States Department of Agriculture
Food and Nutrition Service
Office of Policy, Plannin·g and Evaluation
•
Acknowledgments
This report was prepared by the Office of Policy, Planning and Evaluation
of the Food and Nutrition Service. Useful advice, assistance, and comments
were received from several interested groups and individuals.
Within FNS, members of the Program Development Division and Performance
Reporting Division (both of Family Nutrition Programs), the Financial
Policy and Systems Division of Financial Management, Regional Operations
and the Regional Offices all made important contributions. Alberta Frost,
Acting Deputy Administrator for Family Nutrition Programs, reviewed a draft
of the report and suggested several improvements.
In the Treasury Department, Nelson McClung, Assistant Director of Personal
Taxation Staff, Office of Tax Analysis, provided critical review and
assistance. Gordon Wilson of his staff provided the recoupment estimates
from a merged data file. Tom Tiffany of the Legislative Analysis Division
of the Internal Revenue Service authored their analysis of the effect of
recoupment.
David Lindeman of the Department of Health, Education and Welfare provided
critical comments on the research design and draft report. Staff at the
Social Security Administration provided information on Social Security
number validation procedures.
Staff from the following State. food stamp offices provided considerable
expertise on administrative matters: Florida, Maryland, Oklahoma, Louisiana,
Colorado, Washington, West Virginia, Minnesota, California, Connecticut and
Massachusetts.
Mark WorthingtonofUrban Systems Research and Engineering, Inc., made a
substantial contribution to the analysis.
Staff of Mathematica Policy Research under the direction of Harold Beebout
developed the simulation model from which the recoupment estimates were
developed. Ricardo Springs, Margorie Odle and Diane Hollenbeck were the
primary contributors on that part of the analysis.
The principal investigator of the study was Christy Schmidt of the Office
of Policy, Planning and Evaluation. Ellen Goldberg was the primary
researcher for the administrative implications and costs portion of the
report. Other office members who made important contributions included
Jane Ross, Todd Torrance, Kathy Bishop, Judi Reitman and David de Ferranti,
Director of the Office. Jean Osterling, Toni Walls, Jacqueline Eaton,
Yvetta Evans and Dolly Harrington patiently and expertly typed the several
drafts and final version of the report.
i
HIGHLIGHTS
In the Food Stamp Reform Act of 1977, P.L. 95-113, Congress directed the
Department of Agriculture to perform a study of a recoupment proposal
sponsored by Representative James Jeffords. The Jeffords plan would
require food stamp recipients to pay back some or all of the food stamp
benefits they receive during a calendar year if their adjusted gross
annual income for that year exceeds twice the poverty line. The amounts
that recipients must repay would be collected through the Federal income
tax system. Highlights from the study's findings follow.
The Jeffords plan would reduce the Federal cost for Food Stamp Program
benefits, but increase State and Federal program administrative expen­ses.
Internal Revenue Service expenses for processing income tax
returns also would incr~ase. Overall, there would be a small net
savings at the Federal level, but only after an initial two-to-three
year period of higher net Federal and State spending for start-up.
-Start-up costs would be $10 million.*
- After start-up, the net Federal savings would be $48 million
annually. State costs would be up by $27 million annually.
Overall at all levels, there would be a $21 million savings
annually.
- These amounts could be higher or lower depending on economic
conditions, details of the final legislation,and other factors •
• If the plan was enacted prior to April 1980, recovetyof benefits
would begin in April 1983. This means start-up would extend through
1982, and there would be no savings until 1983.** Faster implementation
is infeasible because:
- Recoupmen~ must be geared to the tax year. To recover benefits
on April i983 tax returns, new data collection procedures must
be fully operational by January 1, 1982.
- Food stamp agencies and the Internal Revenue Service must develop
new regulations and undertake major changes in forms, procedures,
files, and computer systems. All this will take at least twenty
months.
*Figures are in 1980 dollars.
**In Calendar Year 1982 which is after start-up is completed but before benefits
are being recovered, ~he Jeffords plan would result in net increased spending of
$35 million at the Federal level and $27 million for states.
ii
•
•
•
•
The plan would affect a relatively small number of food stamp
recipients and recover a modest amount of benefits. The plan would be
costly to administer.
Five percent of all households rece1v1ng food stamps at any time
during a year period would be subject to recoupment.
- They would owe back 1.5 percent of the total food stamp benefits
paid in that year •
- The plan would add $53 million to Federal and State costs for
administering the Food Stamp Program (excluding start-up). It
would add a further $5 to 12 million to Internal Revenue
Service costs.
• Recovery of benefits would be accomplished partly through reductions
in tax refunds and partly through filer payments.
- 48 percent of the amounts that recipients owe could be taken
from their tax refunds.
- More than half of all households subject to recoupment would
owe less than $200 a year.
- Some of the amounts owed would not be recovered due to filer
error, non-compliance, and waivers under a hardship provision.
Recoupment would affect primarily working households that participate
in the Food Stamp Program for relatively short periods. The typical
household subject to recoupment:
- would be either a two-parent family of four or a single person,
-would receive food stamps for fewer than three months of the year,
and have at least one earner who is employed six months or more,
and
-would have a gross annual income between 2.0 and 2.25 times the
poverty line (or $14,300-16,087 for a family of four and $7,340-
8,257 for a single person in fiscal year 1979 terms).
• Recoupment could discourage employable food stamp recipients from
looking for or accepting jobs that would increase their income. In
some cases, this might induce recipients to stay on the Food Stamp
Program longer than they would otherwise, which would increase program
costs.
- For every additional dollar a recipient earns (within certain limits)
the Jeffords plan would take back a dollar. In addition, the
recipient would lose a further 20 cents or more in income taxes,
social security taxes, and reductions in food stamp benefits.
Overall, the recipients could be financially worse off working
than not working.
iii
- If the plan were altered to take back less than a dollar for
every dollar earned, there would be less work disincentive. The
savings from recovering benefits would be smaller, but would
decline less than proportionally with the recoupment rate on each
dollar earned •
• The Food Stamp Program and the income tax system use different units:
the household and the tax filing unit, respectively. Because some
households contain two or more tax units, certain new inequities
could arise.
- Some households would be subject to recoupment even though their
gross incomes are below the threshold (twice the poverty line),
contrary to the original intent. Other households would not be
subject to recoupment even though their incomes are above the
threshold.
- Some households would be subject to recoupment even though they
legitimately receive food stamps in all twelve months of the
year. Because these households receive food stamps for a longer
period of time than others subject to recoupment, they bear a
disproportionately large share - 27 percent - of the total amounts
owed. Many of these households would have to pay back $300-$800
annually.
Translating from households to tax units also would have major
administrative implications. An allocation scheme would be required
to apportion a household's benefits among its members. Four alterna­tive
schemes were studied.
The allocation scheme proposed in the Jeffords plan would
apportion benefits on the basis of individuals members' annual
incomes and estimates of the cost of "maintaining the household"
over a full calendar year period. This would require in-depth,
year-end interviews of all potentially recoupable households-­the
equivalent of a major multi-month survey that would have to
be completed in two or three weeks each January. This is probably
not feasible.
- A modified Jeffords allocation scheme would apportion benefits
by certification periods rather than on a calendar year basis.
While this would not require the year-end survey, it would involve
collection and verification of far more extensive and detailed
income and expenditure data than either food stamp offices or
the IRS do at present.
- A third possible allocation scheme would apportion benefits
according to the sizes of the tax units within a household.
To administer this scheme, food stamp caseworkers would need
to have extensive knowledge of income tax regulations. Also,
the assumptions that caseworkers would have to make abo.ut a
iv
•
•
•
•
household's tax unit composition would not match actual filing
patterns in many cases where changes in households occur during
the year.
- A fourth scheme would apportion benefits on a strictly per capita
basis within the household. Although requiring less data than the
other three schemes, per capita allocations would be furthest
from the original intent of the Jeffords plan and would require
costly and time-consuming validation of Social Security numbers.
Savings from recovery of benefits would probably be smaller than:~ under
the original Jeffords allocation scheme •
• Recoupment also would have important administrative implications for
the Internal Revenue Service.
- Insufficient information would be available to audit fully the
recoupment sections of some tax returns (e.g., when a household
changes residences during a year on when a tax filer should
report income from dependents on his or her tax return).
- In cases of non-compliance (whether due to misunderstanding or
delinquency), the amounts owed would often be less than the
cost of collecting them. While more than half of all households
would owe less than $200 a year and more than a quarter would owe
less than $100, IRS collection costs would average more than
$113 per case.
• There could be difficulties coordinating the timing of recoupment
reports with the rest of the tax filing process.
- Food stamp agencies would need to send year-end reports (like
W-2 forms) to recipients and the IRS by the end of January.
Some States do not complete the necessary reconciliation of
participation and issuance data for the entire calendar year
until mid-February, and would have difficulty mailing out the
year-end reports before the end of February.
- If some recipients consequently filed tax returns before
receiving their W-2 forms, no benefits would be recovered
from them unless they voluntarily send in corrected forms
later or the IRS undertook special collection efforts •
v
SUMMARY
This is a summary of the findings from a study on a recoupment proposal for
the Food Stamp Program. The proposal originates from an amendment offered
by Representative James Jeffords to the Food Stamp Act of 1977, P.L. 95-113.
Although the amendment did not pass, Congress mandated that the present
study be performed to assess the implications of recoupment for consideration
by Congress in subsequent years.
"The summary is divided into six sections, discussing (1) what recoupment
is and how it would work, (2) its impact on Food Stamp Program costs and
overall government spending, (3) the types of households it would affect,
(4) equity and incentive issues, (5) administrative issues, and (6) compari­sons
with earlier estimates.
WHAT RECOUPMENT IS AND HOW IT WOULD WORK
The Jeffords plan, if enacted, would require food stamp recipients to pay
back some or all of the benefits they receive during a calendar year if
their adjusted gross incom~ for that year exceeds twice the poverty line.
, Under the 1977 Act, households are eligible for food stamps if their net
income (after various deductions from gross income as prescribed by the
Act) is less than the appropriate poverty line for their household size.
Eligibility is determined on a monthly basis. It is possible, therefore,
for a household with a low net, monthly income during part of a year to be
legally entitled to food stamps for that period, and yet have a gross,
annual income for the entire calendar year that is above twice the poverty
vi
•
•
line. For example, a household with highly seasonal earnings or spells
of unemployment could be eligible for food stamps in some months but
have a yearly income above twice the poverty line.
Recoupment is a way of accounting for these variations in a household's
income during a year. From a broader perspective, recoupment is one of
several possible approaches to the issue of how to enable a program such
as Food Stamps, which is based on a short (i.e., monthly) accounting
period, to take into consideration applicants' income streams over a
longer (i.e., annual) horizon. Other approaches to this issue have
included schemes involving summing together several prior months' income
at the time of eligibility determination.
Recoupment, unlike many alternative schemes, would retain the short
accounting period, allowing the program to continue to be. immediately
responsive to sudden loss of income. A household that experiences an
abrupt worsening of its financial circumstances would be able to begin
receiving food stamps as quickly as under the current program.
It can also be argued that with the institution of recoupment the program
would be more target efficient, in so far as recovering benefits from
relatively higher income recipients is interpreted to imply that the
program is targeted more precisely on the groups it was meant to serve
and does not provide benefits to those it was not meant to serve. However,
this conclusion depends on one's opinion of what the target population
vii
should be, and whether a · monthly or annual perspective is adopted. Some
opinions hold that recoupment would deny benefits to certain recipients
meant to have them.
Other issues discussed in the present analysis include: horizontal
equity (to what extent would program participants in similar circumstances
be treated the same?); vertical equity (to what extent would participants
with greater needs ~lways receive more benefits?); and work incentives
(would employable participants find it more or less financially rewarding
to increase their work effort?).
The Jeffords plan also specifies exactly how recoupment would be carried
out. A key element of the plan is a tie-in between Food Stamp Program
administration and the Federal income tax system. The Internal Revenue
Service's annual income tax forms would be amended to include questions
on food stamps obtained during the year. Recipients would have to r e im­burse
the IRS for the smaller of (1) the full amount of the benefits they
received or (2) the amount by which their adjusted gross annual income,
as defined for tax purposes, exceeded twice the poverty line. This would
mean that their "recoupment liability," or the amount they owe, would be
the difference between their actual income and the "recoupment threshold"
(twice the poverty line), up to, but not exceeding, the food stamp benefits
received. When preparing their tax returns, recipients would subtract
their recoupment liability from their tax refund (if they have one). It
was expected that in many cases this would take care of the full amount
owed. In other cases, after the refund was reduced to zero, there would
viii
remain a payment due to IRS. An additional provision would allow a
household to either defer this payment if it is still receiving food
stamps at the time tax returns are due, or apply for a waiver of the
liability if payment would create a hardship.
Because the tax system would be used, the Jeffords plan would need some
procedure for reconciling differences in the definition of a food stamp
household and a tax filing unit. In the Food Stamp Program, all persons
living together and customarily purchasing food and preparing meals
together for home consumption are considered a single household and
must apply together for food stamps. Income tax filing units, however,
normally consist only of related individuals, and even related persons
may file separately. A food stamp household may, therefore, contain
more than one tax unit. (In 25 to 33 percent of all food stamp households,
the household and the tax unit are not the same entity, according to this
study's data.) As a result, the benefits received by a household must be
apportioned to its component tax units if recoupment is to be effected
through the tax system. The Jeffords plan would deal with this issue by
assigning the household's benefits to the individual (or married couple)
in the household whose income provided at least 80 percent of the cost of
maintaining the household during the calendar year. If no individual or
couple accounted for 80 percent of this cost, the benefits would be pro­rated
among members of the household according to their separate contri­butions.
Thus, it would be possible for a food stamp household to have
its benefits, and its potential liability for recoupment, divided among
two or more tax units.
Variations on theJeffordsplan can be devised by altering one or more of
three critical design variables: the threshold income level, the method
of apportioning benefits among household members, and the recoupment rate.
The recoupment rate is the rate at which liability for benefits is phased
in at incomes higher than the threshold. A fourth variable, the definition
of income, is also important. Several alternative combinations of these
variables are examined in this study.
IMPACTS OF FOOD STAMP PROGRAM COSTS AND INTERNAL REVENUE SERVICE EXPENSES
The Jeffords plan would reduce Food Stamp Program benefit costs at the
Federal level, but would increase State and Federal program administrative
expenses. Internal Revenue Service expenses for processing income tax
returns also would increase. Overall, there would be a small net savings
at the Federal level, but only after an initial two-to-three year period
of higher spending for start-up.
The magnitudes of these impacts would depend on several factors, including
• the amounts that recoupment households would owe (their
"recoupment liabilities")
• the extent to which these amounts actually would be
collected, and
• the particulars of the administrative procedures adopted.
These factors in turn would depend on economic conditions, the extent
and nature of IRS collection efforts, the final provisions of legislation
instituting recoupment, and any future reforms in the Food Stamp Program
that may be enacted.
X
'"
For each component of the costs and savings, a range of estimates was
derived and then a best single estimate was determined. The best
estimates imply that:
• Start-up expenses would run approximately $5.6 million
at the Federal level and $4.4 million for States·;*
• After start-up, there would be a net savings at the
Federal level of about $48 million annually and a
net increase in States' administrative expenses of
nearly $27 million annually. Overall at all levels,
there would be approximately $21 million net savings
annually.*
If the plan were enacted in the fall of 1979 or the winter of 1980,
recovery of benefits could begin in April 1983.** The year-by-year
impacts (in millions of 1980 dollars) would be:
!
Federal Cost (+) and Savings (-)
Food Stamp Program State
Fiscal Administrative Benefits IRS Total Administrative Total
Year Expenses Recovered Costs
1980 + .3 - - + .3 + .3 + .6
1981 + 4.1 - +1.2 + 5.3 + 4.1 + 9.4
1982 +26.6 -· +8.4 +35.0 +26.6 +61.6
1983(and +26.6 -83.4 +8.3 -48.4 +26.6 -21.8
annually
there-after)
*Figures are in 1980 dollars and do not include savings or costs for Puerto
Rico or other territories. Those areas currently account for 12 percent of
Food Stamp Program benefit costs annually, but cannot implement the Jeffords
plan as currently constituted because territories are not subject to the
Federal income tax.
Also, star·t-up figures do not include $62 million in regular administrative
expenses the first year of operation, when no benefits are recovered. See
table above for details .
**See "Administrative Issues" section below on timing of implementation.
xi
A breakdown of these figures follows.
Recovery of Benefits
The estimated $83.4 million annual savings from recovery of benefits assumes
that households subject to recoupment would have a total recoupment liability
of $105 million a year, and that 79 percent of this would actually be
collected.
The $105 million in total recoupment liability is the midpoint in a range
extending from $87 million to $124 million. These latter figures are the
best estimates obtained from two different data bases.* The $124 million
is almost certainly too high due to a bias involving the income data. The
$87 million has no known bias; although some of its underlying assumptions
may put it slightly on the low side, others have an offsetting effect.
However, to be sure that the savings from recoupment would not be under-estimated,
the midpoint between the two estimates was selected, yielding
$105 million.
The above estimates were derived in part from survey data reflecting 1975
economic conditions.** Because unemployment was exceptionally high that
year (8.5 percent, as compared to 6.0 percent in August 1979), food stamp
*The data bases are, respectively, (1) the Survey of Income and Education
and (2) the Survey of Income and Education matched with the Survey of Income,
a compilation of Treasury Department tax records. Both estimates are lower
than had been generally anticipated in 1977. The reasons for this are
discussed in the concluding section of the Summary. On the possibilities
for bias in the estimates, see the main text for details.
**See main text for details on the estimation procedures and the technique
used to project to 1980.
xii
•
rolls included a greater than normal number of households likely to be
subject to recoupment (i.e., earners who needed food stamps when they
were out of work temporarily, but finished the year with income exceeding
twice the poverty line). As a result, the $105 million estimate is higher
than can be expected for most years, when unemployment is below the 1975
level.
Also, the $105 million is based on the new eligibility rules and income
limits mandated by the 1977 Act. These reforms eliminated from the program
many participants who would have been subject to recoupment, and reduced
the benefits received by many others. Comparison of estimates derived
under the old and new rules shows that the reforms have reduced the total
recoupment liability by 7.1 percent.
More recently, further changes in the program enacted in July 1979 will
allow individuals over 60 or receiving Supplemental Security Income or
Social Security disability benefits to deduct medical expenses over $35
a month from their income when their food stamps are computed, and will
remove the ceiling on shelter cost deductions for households containing
such persons.
The 1979 changes may increase the recoupment liability slightly. However,
only a very minor increase is likely because most elderly and disabled
households covered by the new provisions do not have significant earnings
and would not be subject to recoupment. At the same time, for that small
percentage of elderly and disabled households with high medical and/or
xiii
shelter expenses who woyld be subject to recoupment, a somewhat anomalous
situation cpuld arise. Some of these households could receive food stamp
benefits all 12 months of the year due to the new amendments, and then
have to pay back some of their benefits the following April.
Collectibility
The extent to which recoupment liabilities would actually be collected
depend? on a number of factors, including the degree to which liabilities
can be offset against tax refunds, how many recipients would file incomplete
or incorrect tax returns (by mistake or in deliberate non-compliance), and
what would be done about such problems. In addition, the collectibility
of the amounts owed is limited by the waiver and deferral provisions.
Several findings are important in this regard:
• 48 percent of the total liability could be recovered
from recipients' tax refunds if their returns are all
complete and correct;
• The remaining 52 percent would be payable by check or
cash when returns are filed; and
7 percent of the above would qualify for deferral.
Furthermore,
• More than half of all households subject to recoupment
would owe less than $200 for the year. More than 25
percent w9uld owe less than $100 .
• Collection costs would exceed $113 per case on average.
Where special enforcement activities are required (e.g.,
investigation and prosecution), the cost would be greater.
xiv
..
..
..
Because so much depends on voluntary compliance, collectibility is
difficult to estimate in advance. Nevertheless, in light of the Internal
Revenue Service's experience with other kinds of income tax liabilities,
it can be expected that the amount recovered would not exceed 79 percent
of the total owed. This leads to the best single estimate of about $83.4
million (out of the $105 million owed). The $83.4 million assumes that
(1) all amounts obtainable from tax refunds (48 percent of the total
liability) would be recovered, (2) 10 percent would not be recovered
due to waivers and deferrals, and (3) three quarters of the remainder
(three quarters of 100-48-10=42 percent) would be recovered by check or
cash.
Actual collections could fall short of $83.4 million for any of several
reasons. The amounts obtained from refunds conceivably could be less
than the full potential in so far as tax returns are not all complete
and correct. Moreover, with so many households owing only small amounts
(100 or $200 or less), some may be prone simply to ignore their liability.
With collection costs often exceeding the amount to be collect~d, the
Internal Revenue Service would be hard pressed to justify vigorous
enforcement activities in these circumstances. Finally, certain house­holds
(e.g., some that change residences during the year; see the
"Administrative Issues" section below for details) would be untraceable,
either for sending them the information they need to compute their
liability or for checking up on them.
XV
I
To the extent that any of these points would be significant, $83.4
million would be on the high side and the savings from recoupment
would be less than previously indicated.
Administrative Costs
The administrative cost estimates given in the table above ($62 million
a year in State and Federal costs, including IRS expense~were derived
from task...:.by-task analyses of the detailed responsibilities of Federal,
State, and local Food Stamp Program authorities and the Internal Revenue
Service in both the start-up and operating phases. Extensive data were
obtained from eleven States and several Federal agencies. Costs were
calculated for the additional burden of recoupment activities (i.e.,
net of the level of effort already required by existing legislation),
and reflecting recent per-case costs and the number of recoupment cases
anticipated.
For the Food Stamp Program, the principal source of increased administra-tive
costs would be for the following new functions required for State
and local agencies:
• collecting sufficient income and household expenditure
information at certification to apportion food stamp
benefits each month among individual household members,
tax units within households, or primary household
supporters, depending on the allocation plan selected,
maintaining a cumulative record of food stamp benefits
received by individual, tax unit, or household supporter,
preparing and sending out year-end reports (like W-2 forms)
showing the total annual food stamp benefits allocated to
every individual, tax unit, or household supporter on the
program at any time during the year.
xvi
In order to carry out these functions, the data processing, storage,
and retrieval capabilities of food stamp offices would have to be
greatly expanded. Computer programs and office procedures would have
to be revamped, and case filing systems would have to be completely
reorganized so that data accumulated month by month would be readily
usable on a case by case basis at the year's end. In addition,
certification interviews would be longer and take more staff time,
thereby also requiring the hiring of some additional certification
personnel.
For the Internal Revenue Service, the principal sources of increased
costs would be revisions of the basic 1040 and 1040A tax forms,
additional auditing, development and monitoring of deferral plans,
and collection and enforcement activities.
The revisions to the basic 1040 and 1040A tax forms would depend on the
final design of the plan, but at a minimum would include adding two or
mor e lines to the 1040 and requiring many filers to switch from the short
1040A to the longer 1040. One way or another, all tax returns would become
longer, even those not subject to recoupment. This would substantially
increase the amount of information that Internal Revenue Service computers
must process, store, and check for errors.
CHARACTERISTICS OF HOUSEHOLDS SUBJECT TO RECOUPMENT
Approximately five percent of all households that ever receive food stamps
during a year would be subject to recoupment. In 1980, that is expected
xvii
to be about 425,000 households. Recoupment households would receive two
percent of the total benefits paid out over an entire year.
The data indicate that the "average" recoupment household 'would have the
following distinctive features:
it would be either a two-parent family of four or a
single person,
earnings would be the primary source of income,
• the primary earner would be a male, between ages 18
and 44, who is employed for at least six months but
unemployed for at least one month during the year,
the household would participate in the Food Stamp
Program fewer than three months in the year,
• its gross annual income would be between 2.0 and
2.25 times the poverty line (i.e., in 1979 terms,
between $14,300 and $16,087 for a family of four
and $7,340 to $8,257 for a single person),
its recoupment liability would be less than $200.
In general, recoupment households would be distinctively different as a
group from other food stamp participants .. They would be in their prime
earning years by and large, with higher earnings and fewer spells of
unemployment than their non-recoupable counterparts. Almost by definition,
they would have much higher incomes and hence receive much lower monthly
benefits. And they would include almost none of the programs' elderly
and Supplemental Security Income recipients, and relatively few of its
AFDC families. Only with respect to their region of residence would they
be broadly similar to the rest of the food stamp caseload.
xviii
•
The fact that recoupmen~ households would be among the most employable
and highest earning food stamp participants suggests that many of them
may be relatively well established in the labor market. In addition,
the fact that the majority of recoupment households receive food stamps
for fewer than three months of the year (and 80 percent are participants
for fewer than six months) indicates that they are not chronic dependents
on public assistance. All this is hardly surprising, but underscores
that recoupment mostly would affect working households who (1) have
annual incomes below the national average for their household size, but
above the lowest quartile, and (2) use food stamps for comparatively
short periods.
The Food Stamp Program covers certain categories of households who are
often not covered by other public assistance programs - childless couples,
single persons and intact families with children. It is precisely these
types of families that would be the target of recoupment.
The data indicate that most recoupment households would be only slightly
above the recoupment threshold (twice the poverty line), and have relative­ly
small recoupment liabilities (e.g., under $200).
One of the reasons for this concentration just above the threshold is
that the number of households participating in the Food Stamp Program
drops off very quickly at higher income levels. Also, as has been noted,
xix
the 1977 Act eliminated many of these households by tightening eligibility
requirements and effectively setting a ceiling on gross income. As a
result, the drop-off in the number of households with annual incomes above
twice the poverty line is even more pronounced now than before.
EQUITY AND INCENTIVE ISSUES
Recoupment would alter various incentive pressures on low income households
and affect program equity. The principal issues in this regard are as
follows.
Consequences of Differing Definitions of Income
The Food Stamp Program uses a more inclusive definition of income than the
income tax system. In the Food Stamp Program,unemployment compensation,
social security and welfare are counted in benefit calculations, but in
the tax system these sources are nontaxable. As a result, two households
with the same total annual income could fare differently under recoupment,
depending on their income sources. Households with mostly earnings would
be at a disadvantage relative to those with mostly transfer income. This
conceivably could have the effect of discouraging some households from
seeking to substitute earnings for transfer income.
Consequences of Differences Between Households and Tax Filing Units
Because the Food Stamp Program and the tax system deal in different units
(households and filing units respectively), some individuals would be sub­ject
to recoupment who apparently should not be, while others would not be
who apparently should. The following examples illustrate this.
XX
Households A and B each contain five members; and the
relevant household recoupment threshold is $16,620.
At first glance, household A, with an annual income
of $11,000, should not be subject to recoupment whereas
household B with an annual income of $18,000 should.
However, household A contains two tax units: a single
unrelated individual with $8,000 in earnings and a four­person
tax unit with $3,000 in transfer income. Since
the single filer's income exceeds twice the poverty line
for an individual, he or she would be subject to recoup­ment.
Thus, after all, household A is recoupable.
Household B also contains one four-person and another
one-person tax unit. However, the single filer has
earnings of $7,200 and the four-person tax unit has
earnings of $10,800. Both tax units have incomes below
twice the poverty line for one and four persons,
respectively. As a consequence, no one in household B
would be subject to recoupment, contrary to initial
appearances.
From the standpoint of current food stamp .law, household A should be
entitled to keep all the benefits it receives, but household B should
be subject to recoupment. However, from the Internal Revenue Service's
perspective, it should be the reverse: household A should repay some
benefits and household B should retain everything.
As a result of this noncoincidence of definitions, one of every ten
recoupment households is actually legally entitled to food stamp benefits
throughout all 12 months of the year, yet would have to pay some or all
of its benefits back simply because it is comprised of two or more tax
units. Household A (above) is an example of this situation. Since these
households, who are subject to recoupment because of the discrepancy
between the food stamp household unit and the IRS filing units, can
xxi
participate in the Food Stamp Program all twelve months, their recoupment
liabilities are very large. In fact, these households would owe over a
quarter of all recoupment liabilities, often in range of $300 to $800
per household annually.
Effects on Work Effort
Because recoupment would be a form of tax on earnings, it could create
a work disincentive. In certain circumstances, employable food stamp
recipients could find it in their interest to limit their earnings by
constraining the time they spend working or by not looking for or
accepting new jobs. The source of this disincentive is the Jeffords
plan's implicit 100 percent tax rate on income above the recoupment
threshold. The 100 percent tax rate means that when a household's
income reached or exceeded the threshold, any additional earnings would
be completely offset by increased recoupment, dollar for dollar. The
workers, and the household, would earn more, but also owe more to the
Internal Revenue Service, and in the end wind up no better off than
without the additional work effort. Only when the household's income
had risen enough so that all of its food stamp benefits had been
recouped by the IRS would further earnings not be entirely taxed away.
Furthermore, a household facing a recoupment tax rate of 100 percent
would have an overall tax rate of well over 100 percent, when all other
adjustments to earnings are included. Federal, State and local income
taxes and FICA deductions would increase the household's overall tax
rate to more than 120 percent. This means that, over a range of income,
xxii
•
the household would actually lose $1.20 or more for each additional dollar
earned. Benefit reductions in food stamps and other public assistance (A.FDC,
SSI, or general assistance) would raise the overall rate even higher. In
the end, the household could be worse off working than not working, as
the following example demonstrates:
A household consisting of a couple and their two children
has one earner, the male. Through August his earnings
are almost twice the poverty line. In August he loses his
job and in September he begins receiving unemployment
compensation ($300 per month) and food stamps ($146 month).
In early November he is offered a full-time job at $3.50
per hour. If he takes the job, the worker will earn $616
gross per month. Income and payroll taxes will take $123
and $37 per month, respectively, leaving him with $456 to
take home. He will lose his unemployment compensation
and his food stamps will be reduced to $45 per month, so
he will have a net monthly income of only $55 higher than
before. In addition, if he takes the job his annual
adjusted gross income will be high enough to trigger
recoupment of all food stamp benefits received during the
year, an amount equal to $382. By taking the job for the
last two months of the calendar year, then, his family
will end up with $272 less to live on than if he remains
unemployed. It thus benefits him financially to wait
until January to begin employment.
A similar result could occur for other types of households, such as a
household with a fully employed primary earner and a secondary earner
on the margin between working and not working. This may be particularly
true of secondary workers considering jobs in November or December, when
numerous temporary jobs become available. Taking such a job could reduce
actual household income. Under the 1977 Act's work registration provisions,
a recipient is in certain circumstances required, as a condition for being
xxiii
entitled to food stamps, to accept a job located by the state's employment
service. However, recipients still have an unconstrained choice regarding
jobs located by other means, which are likely to remain the majority of jobs
they consider.
To the extent that recoupment households do limit their work effort, their
// recoupment liabilities and hence the net savings for taxpayers would decrease
somewhat from the estimates discussed above. To reduce this disincentive,
the Jeffords plan conceivably could be altered to reduce the recoupment tax
rate to less than 100 percent. This need not eliminate most of the savings
from recovery of benefits, according to the present analysis. For instance,
cutting the rate to 50 or 25 percent reduces total liabilities subject to
recoupment by 7 and 16 percent respectively.
ADMINISTRATIVE ISSUES
As noted above, major changes in State and local food stamp offices'
operating procedures, filing systems, and computer programs would be required
to accommodate recoupment. In addition, there would be the following other
administrative issues.
Year-End Deadlines
In order for the Internal Revenue Service to carry out its responsibilities
in synchronization with the rest of the• tax filing process, food stamp
agencies would have to send year-end report (like W-2 forms) to recipients
and the IRS by no later than. the· end of January. . This would be
xxiv
difficult in many instances because the extensive work involved in preparing
the reports would have to be completed in a short period which is also the
time of year when casework activity is usually at its peak.
Furthermore, there could also be a serious reconciliation problem. The
year-end reports would need to be based on the value of the food stamps
actually issued to households, not just on the amount they are found to be
entitled to. Their Authorization-to-Participate records would thus have to
be reconciled with stamp issuance data before the reports could be prepared.
States generally reconcile monthly issuance data 45 days after the close of
the month. While some States may be able to do more this quickly, reconcilia­tion
for December could not be completed in many States much before February 15.
Yet a final accounting would be needed for preparation of the year-end report
by early or mid-January. With these timeframes, the deadline for the reports
could not be met unless December and in a few cases part of November were
omitted.
In some cases, recipients may already have filed their tax returns before
receiving their late-arriving food stamp W-2 forms. No benefits could be
recovered from these recipients unless they voluntarily send in corrected
returns or the Internal Revenue Service undertakes special collection
efforts. The magnitude of the benefits that would not be recovered on this
account is difficult to predict and hence has not been included in the cost
estimates above, but conceivably could be considerable.
Allocation Schemes
As has been discussed, the fact that the Food Stamp Program and the income
tax system use different units (the household vs. the tax filing unit)
would mean that some allocation scheme is needed to apportion a household's
benefits among its members. In addition to bther consequences already
mentioned, the need for and nature of the allocation scheme would have
important administrative implications. Four alternative schemes were
analyzed, each with its own implications.
The original Jeffords plan specifies a scheme* that would require collection
of detailed data on each household's annual expenditures and the annual
contributions of each member to cover those expenses. Because the data
would have to cover the entire calendar year and would be too complex to
obtain by mail or telephone, local food stamp offices would have to conduct
year-end, in-person interviews with all persons who were in households that
ever received food stamps during the year. This would include persons no
longer members of participating households and households no longer participat-ing
at the time the interviews must be held. The entire effort would be
the equivalent of a major multi-month Census survey that would have to be
completed in approximately two weeks in order to meet deadlines for sending
out the year-end W-2 type forms.
*Liability for the full amount of foodstampsreceived by a household is
assigned to the individual (or married couple) in the household who provided
at least 80 percent of the cost of maintaining that household during the
calendar year. If no individual or couple provided a full 80 percent,
the liability would be pro-rated among the household's members according
to their relative annual contribution to the cost of maintaining the
household.
xxvi
•
Since such an undertaking would be highly unworkable, three alternative
schemes were devised and evaluated. One scheme would be identical to
the original Jeffords formula in all respects except that the allocation
of benefits would be determined for each certification period rather than
the year as a whole, thus obviating the need for a year-end survey. This
"modified Jeffords allocation scheme" was used for the cost estimates pre-sented
above. Although more feasible than the original formula, the
modified scheme would still entail major new efforts to collect and verify
extensive data that is not currently needed for eligibility determinations,
and would add considerably to the complexity of the certification process.
Administrative procedures somewhat simplified by the 1977 Act would become
significantly more complex than they had been before the 1977 Act.
A third possible scheme would allocate benefits to tax units within a
household in proportion to the number of individuals within each tax unit.*
This "allocation by tax unit" scheme would involve far less data collect.ion
than the preceding two schemes, but would require food stamp caseworkers to
predict the tax unit composition of households far in advance of filing
time. Caseworkers would have to become skilled interpreters of tax code
regulations. Furthermore, the predicted tax unit composition of a household
might not match actual filing patterns in many cases where changes occur i n
households over the year. There would be no way to correct for all these
changes without the same kind of extensive year-end survey necessary for the
original Jeffords scheme.
*For example, if a household contained two tax units of three and two people
each, three-fifths and two-fifths of the benefits would be allocated to each
unit respectively.
A fourth and final scheme would apportion benefits on a strictly per
capita basis within each household. This scheme entails less data collection
than the previous three schemes, but would require every household member
to have a Social Security number that has been validated against Social
Security Administration files.* From the IRS' perspective, a "per capita"
scheme would be difficult or infeasible to administer, since the data provided
by food stamp agencies would not be organized on anything resembling a tax
unit basis. Also, the "per capita" scheme is the farthest from the intent
of the original Jeffords plan to allocate benefits according to contribution
to the household's maintenance, and might result in smaller savings from
benefits recovery than the Jeffords plan's allocation scheme.
Handling Households That Move
Households that change addresses during the year would not receive noti-fication
of the food stamp benefits their members received at their first
residence unless they have left a forwarding address. Some of these house-holds
would probably be untraceable.
Since households subject to recoupment receive food stamps for fewer than
three months a year on average, it is possible that a noticeable fraction
of them change residences during the year. How much of what they owe would
never be collected is unclear (and is not reflected in the estimates), but
may be important.
*1979 legislation provides for Social Security number enumeration of all food
stamp recipients. Validation would be a new requirement, involving checking
the accuracy of the numbers reported by recipients. The Internal Revenue
Service maintains that validation is necessary if Social ~ecurity numbers
are used for tax purposes.
xxviii
•
Impacts on Error Rates
The substantial increases required in data collection, storage, and process­ing
would create more opportunities for error. Incorrect determination of
recoupment liabilities could conceivably diminish the potential savings.
States that are unable to hire additional caseworkers due to personnel
ceilingscould face other problems. Since recoupment will add an average
of 15-25 minutes to certification interviews, the amount of time otherwise
spent in the interview on exploration and verification of household circum­stances
could be reduced. This could increase error rates in these States.
Furthermore, the States would have little incentive to administer recoupment
as efficiently and effectively as possible since all benefits recovered would
revert not to the States but to the Federal Government.
Extent of Caseload Covered
Although only a small portion of the food stamp caseload would be subject
to recoupment, all participating households would have to be individually
checked at the year's end to determine whether they are recoupable or not.
Therefore, all the information needed for determining recoupment liabilities
would have to be collected and filed on the entire caseload.
As an alternative, it might be thought preferable simply to exclude all
households consisting solely of AFDC or SSI recipients, on the grounds
that since these households are unlikely to be recoupable there is no
point to collecting all the necessary data on them. This, however, would
lead to certain inequities for households that move on or off AFDC or SSI
xxix
•
during the year. A household that went off AFDC or SSI during the year
could not later be made recoupable that year, since the necessary informa­tion
would not have been collected. This household would remain exempt
while another household that initially was not on AFDC or SSI, but later
came on, could be recoupable. Although the two households might have
received AFDC or SSI for the same number of months and otherwise be identical,
they would be treated differently.
Timing of Implementation
To implement the Jeffords plan, a period of start-up would be required
involving three principal activities: writing new Federal regulations,
redesigning and reprogramming States' computer systems, and changing local
office recordkeeping and data collection procedures.
New Federal regulations would need to be developed jointly by two agencies,
the Food and Nutrition Service and the Internal Revenue Service. This
would take eight months at a minimum, allowing 60 days for developing and
drafting proposed regulations, 30 days for clearance and publication, a
standard 60-day comment period, and 90 days for comment analysis, preparation,
clearance and publication of final regulations. This timeframe would be
exceptionally quick for development of join regulations between two
Departmentsto establish a new system of this degree of administrative intricacy.
Redesigning and reprogramming computer software sys.tems (and augmenting
hardware facilities where necessary) can begin only after the Federal
regulations have been completed, since until then the States would not know
XXX
the detailed tasks required. Once started, the redesigning and reprogramming
would take an average of ten months. Some States would need more or less
time, depending on their current computer capabilities and the nature of
the detailed tasks.
Local offices would need three months to convert their recordkeeping and
data collection procedures. The conversion could begin a month before the
computer reprogramming was finished, and would extend two months after then.
This is primarily because new information must be obtained from recipients
at their recertification interviews, which span several-month intervals.
In total, not less than twenty months would be required to implement the
Jeffords plan. Because of the tie-in with the tax system, data collection
would have to begin on the first day of a calendar year and recovery of
benefits would start some fourteen or more months later when .. the first
full year's tax returns were due. If legislation were enacted in the fall
or 1979 or the winter of 1980, the earliest Janpary 1 when data collection
could begin would be January 1, 1982. The first savings from recovery of
benefits would be realized in April of 1983.
COMPARISON WITH EARLIER ESTIMATES
At the time the Jeffords plan was under discussion in 1977, several
alternative estimates of its impacts on costs were prepared. Most of these
estimates were considerably higher than the figures here.
The principal explanation for this is that earlier estimates focused
exclusively on recipients' total recoupment liability. Np data were included
xxxi
on the collectibility iSsue or administrative costs. As a result, the
estimates did not show what the net effect would be.
In addition, earlier figures for the total recoupment liability were higher
than the range here of $87 million to $124 million for 1980, with a mid-point
of $105 million. There are several reasons for this.
First, earlier estimates concentrated exclusively on households, and did
not take into account the full implications of differences between households
and tax units. Tax units, the actual focus of recoupment, are after smaller
than the households containing them. Because of this, there can be a
"leakage" of liability as one moves from households to tax units. If house-holds
are used as the basis for figuring recoupment liability, a household's
entire food stamp benefits may be found recoupable. On the other hand, when
the household is broken down into its component tax units, some of the bene-fits
may be allocated to individuals who turn out not to be subject to
recoupment. The recoupable tax units that remain are accountable only for a
smaller amount of benefits, and thus have a smaller liability.*
Second, earlier estimates generally assumed that all of the benefits
received by individuals subject to recoupment would be recoupable. In other
words, if a tax unit receives $300 a year in food stamp benefits and is
*The gross annual incomes attributed to tax units also are smaller than
household incomes. This, in principle, can lead to either higher or lower
liabilities, since the smaller incomes are compared with the lower recoup­ment
thresholds appropriate for a smaller group of individuals (i.e., the
tax units vs. the households they belong to). Overall the data show that
liabilities wind up lower on the tax unit basis.
xxxii
subject to recoupment, the unit was assumed to be liable for the full
$300. In fact, the Jeffords plan holds the unit liable only for the lesser
of (1) the difference between the unit's income and the recoupment threshold
and (2) the full benefits. A number of recoupment tax units, the present
findings indicate, have incomes only slightly above their threshold, and
thus would not be liable for their full benefits.
The significance of these two observations is underscored by the present
finding that recoupment tax units would include only 86 percent of the
individuals in the households they belong to. Furthermore, the recoupment
liabilities of these tax units under the Jeffords plan would be only 67
percent of total benefits their households receive. If this is so, the total
recoupment liability found in earlier estimates is 38 percent too high,
considering only the two reasons discussed so far.
A third consideration concerns the source of the benefits data used in
earlier estimates. Unlike the procedure followed in this study, the earlier
estimates did not estimate benefits by simulating households' entitlements
on a month-by-month basis over a calendar year. Instead, some of the
estimates assumed that the average monthly benefit received by recoupment
households was the same as the average received by all food stamp households
of the same household size. However, the present data suggest that recoup­ment
households receive lower average monthly benefits. This is because their
incomes are higher, sustained in part by transfer income (such as unemploy­ment
benefits) and secondary earners' wages as supplements to the primary
earner's wages. (The annual incomes of recoupment households are higher ~y
virtue of the fact that they are recoupable; the additional point being made
xxxiii
here is that their incomes during the months they receive food stamps also are
higher, relative to nonrecoupment food stamp households.
Yet another reason earlier estimates may have been too high is that they did
not take into account the effect of the tightening provisions of the 1977
Act. As has been noted, these provisions result in a reduction of the
total recoupment liability by 7.1 percent.
Finally, earlier estimates apparently made no adjustment for the fact that
recoupment probably would not be instituted in Puerto Rico or other
territories, where different tax procedures apply. As has been noted,
those areas currently account for 12 percent of Food Stamp Program costs
annually •
•
xxxiv
TABLE OF CONTENTS
PAGE
ACKNOWLEDGEMENTS • • • • . • • • • • . . . • • . . • • . • . . . • . . . . . • • . . . . . . . . . . . . . • . • . • . . . . . . . . . . i
HIGHLIGHTS . . . . . . . . . . . . . . . . • . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . i i
EXECUTIVE S~Y. • . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . • . . . . . . . . . . . . vi
TABLE OF CONTENTS . • • • . • • • • • • . • . • • • • • • . . • . • • • . . • • • • . • • • • • • . . . . . . . • . . . • . . . . . • . XXXV
LIST OF TABLES AND EXHIBITS .•.•.•.••.•..•••..........•..•.•........•.••••. xxxviii
CHAPTER 1.
1.1
1.2
1.3
1.4
CHAPTER 2.
2.1
2.2
2.3
2.4
CHAPTER 3.
3.1
3.2
3.3
3.4
CHAPTER 4.
4.1
4.2
4.3
4.4
•.
INTRODUCTION ••.••...•••......•......••...•..•...
Legislative History •••••.....•..••.•••••..•••..
The Rationale for Recoupment ••...••...••.••.•.•
Underlying Issues and Problems ••..•.•.•..•...•.
Presentation of the Report ..•.•••..••..•.......
WHAT RECOUPMENT IS AND HOW IT WOULD WORK ...... .
Introduction ................................... .
Further Details of the Jeffords Plan .••.•..•..•
Alternative Allocation Schemes .•.••••..........
2.3.1 The Jeffords Allocation Scheme-
1
1
3
8
11
13
13
14
19
Pro Rata on an Annual Basis............. 21
2.3.2 Modified Jeffords Allocation Scheme-
Pro Rata by Certification Period ••..••..
2.3.3 Allocation by Tax Unit ..•..•••..•...•.. .
2. 3. 4 Per Capita Allocation ••..•..••.....••••.
Alternative Recoupment Rates •.•.•..•••.•••••...
RECOVERY OF BENEFITS ......•.. . .................
Estimation Procedure .•••..••.•..•.•.....••.....
3.1.1
3.1. 2
3.1. 3
Data Base Employed •...•.•.....••....•...
Simulation Technique •.•••.••••....•..•..
Baseline Estimates of the Food
. Stamp~ ·· Program . ......................... .
The Jeffords Plan •••••••••••••••..• : ..•.......•
3.2.1 The New Law Without Increased
Participation Due to EPR, 1975 .••.......
3.2.2 Comparison with Recoupment Under
"Old Law" Rules ••••.......•• • . • · · · • · · • · ·
3.2.3 The Impact of Recoupment in 1980 ••.•.•..
Alternative Recoupment Plans .•.•.•..•...•.•..•.
3.3.1 Variations in the Recoupment Threshold ••
3.3.2 Variations in the Recoupment Rate ...... .
3.3.3 Variations in the Allocation of Food
Stamp Benefits •....•.•.••.•.•••••......•
Implications of Estimation Procedures .•..••....
CHARACTERISTICS OF RECOUPMENT HOUSEHOLDS ...... .
Household Size . ............................... .
Household Income ••..•.••....••..••••.....•.....
Recoupment Amounts •••.••••.••..... · ....•••......
Months of Program Participation .....•...••.•..•
XXXV
23
. 26
27
28
33
34
34
37
40
52
52
55
57
58
63
65
71
72
74
77
81
4.5
4.6
4.7
4.8
CHAPTER 5.
5.1
5.2
CHAPTER 6.
6.1
6.2
6.3
6.4
6.5
6.6
Household Composition •.•...••.•.....•.•••..•.•.
Sources of Household Income ••..•..••...•••.••..
Employment Patterns ••..•..•.•••......•.....•.••
The Characteristics of Recoupment Households:
A Profile ..................................... .
EQUITY AND INCENTIVE ISSUES .............•....•..
Equity
5.1.1.
5.1.2
5.1. 3
5.1.4
5 .1.5
Issues ................................. .
Definition of Income •••...•.•.•.•.•••.•••
The Incongruence between Tax
Units and Households .•.........•....•.••
Collections, Referrals, and Waivers •.••••
Allocation Methods .••......•...•••••...•
Calendar Year Accountin& .•..••........•.
Incentive Issues .. .... ........................ .
5.2.1
5.2.2
5.2.3
5.2.4
Work Effort ............................ .
Tax Compliance ......................... .
Program Participation ••••...••.•.•...•••
Income Reporting. ....................... .
ADMINISTRATIVE ISSUES AND COSTS: STATE
FOOD STAMP AGENCIES ....••..•......•.••..•.••..•
Methodology . .................................. .
Alternative Allocation Plans •..•••....•. ~ •••.••••
The Recoupment Process ........•.•..••.•••••••...
6.3.1 Certification ••..•.....••......•..•••••.
6.3.2 Information Processin& .....••.•.•.••.•••
6. 3. 3 End of Year Forms Generation •.........•.
Implementation ................................ .
Costs
6.5.1
..........................................
Per Capita Allocation ••••.••••••••.••.••
6.5.1.1 Start-up Costs ..•••••••.•••••••
6. 5 .1. 2 Operating Costs .••...•..•••••..
6.5.2 Tax Unit Allocation .••.••••.•••••••••••.
6.5.2.1 Start-up Costs ..••••.••.•.•.•.•
6.5.2.2 Operating Costs •••.••.•..•.••.•
6.5.3 Modified Jeffords Plan ..•••••..•••••....
6.5.3.1 Start-up Costs ..••.••...•.••..•
6.5.3.2 Operating Costs .•...•.••.•..••.
Administrative Issues .......................... .
6.6.1 Error rates .•...••.•.•••.•••.••.•.••••..
6.6.2 Social ~ Security ........................ .
6.6.3 Changes in Household Composition
6.6.4
6.6.5
or Circumstances ....................... .
Household Mobility •....•.••.••..••....•••
Excluding Some Participants from
Recoupment . ............................. .
6.6.6 Deferrals .............................. .
xxxvi
86
90
90
95
99
99
99
"'
100
102
105
105
107
107
112
113
114
117
118
121
124
125
129
133
136
140
142
142
145
151
151
152
155
155
156
159
159
161
164
167
169
170
CHAPTER- 7
7.1
7.2
7.3
APPENDIX A
APPENDIX B
APPENDIX C
l<PPENDIX D
•
..
THE INTERNAL REVENUE SERVICE:
ADMINISTRATIVE ISSUES AND COSTS .•.........•..••
The Role of the IRS ••..•..............•.•••.•..
7.1.1 Verification, Audits and Queries .......•
Administrative Issues ......•.......•.••.•.•.•.•
7. 2. 1
7.2.2
7.2.3
7.2.4
7.2.5
The Collection of Amounts Due ..•..•.....
The Earned Income Tax Credit (EITC) .... .
Liability for Interest and Penalties ... .
Income Verification and Audits .•...•....
Social Security Numbers .•••...•....•....
Administrative Costs .....••.•.....•.......••...
7.3.1
7.3.2
7.3.3
7.3.4
7.3.5
Changes in Current Forms .•.••....•......
Increased Processing Costs .•............
Cost of Increased Inquiries ....•........
Collection and Enforcement ............. .
Cost Summary ..................•.......••
The 1979 Jeffords Amendment .........•.......•.•
Food Stamp Recoupment Through the Tax System ...
Technical Appendix: The Simulation ...•......••
The 1977 Jeffords Amendment. ............••...••
xxxvii
173
175
178
179
181
183
184
185
186
187
189
190
190
191
196
Al
Bl
Cl
Dl
Table 1-1
Exhibit 2-1
Table 3-1
Table 3-2
Table 3-3
Table 3-4
Table 3-5
Table 3-6
Table 3-7
Table 3-8
Table 3-9
Table 3-10
Table 4-1
Table 4-2
Table 4-3
Table 4-4
Table 4-5
f
LIST OF TABLES AND EXHIBITS
PAGE
Households C, D, and E ................................. . 7
Alternative Recoupment Schemes ••••••••.••••••••.••.•.••• 31
Recoupment Amounts: Summary Table (1975)
(New Law With No EPR-Induced Participation) .•••.•••••... 42
Recoupment Amounts: Summary Table/Results
From Merged SIE/SOI Data Base (1975) (New
Law With No EPR-Induced Participation) ••••.•..•.•..•.••. 45
Distribution of Recoupment Amounts and Tax
Refund Offsets.......................................... 47
Amounts To Be Collected That Are Not Offset
by Refunds.............................................. 49
Recoverability of Recoupment Dollars Including
Deferrals............................................... 51
Amounts Collected Over Time (Millions of
Dollars)................................................ 53
Comparison of Recoupment Amounts Under
Alternative Assumptions (1975).......................... 54
Recoupment Amounts: Summary Table (FY 80)
(New Law With EPR-Induced Participation
Increases)..... • • • • • . • . • • . • . . • • . • • . . . . • • . . • . • • • . . . • • • • . . 56
Recoupment Amounts: Alternative Recoupment
Rates (New Law With No EPR-Induced Participation) •••••.. 59
Recoupment Amounts: Alternativ- Recoupment
Rates (New Law With No EPR-Induced Participation) •••.••• 61
Characteristics of Recoupment Households:
Size of Recoupment Households and Tax Units ••••••..•.••. 73
Characteristics of Recoupment Households:
Size of Household/Amount Recoupment Dollars .•.••.•.••... 75
Characteristics of Recoupment Households:
Per Capita Annual Income. • • • • • . • • • • • • • • • • • • • . • • • • • • • • . • • 7 6
Characteristics of Recoupment Households:
Annual Adjusted Gross Income ...................... · ..... . 78
Characteristics of Recoupment Tax Units:
Percent AGI in Excess of Threshold •••••••••••••.••••• .••• 79
xxxviii
...
Table 4-6 Characteristics of Recoupment Households:
Amounts of Recoupment Liability. . . . . . . . . . . • . • • • . . . • • • • • • 80
Table 4-7 Characteristics of Recoupment Households:
Months of Participation in the Food Stamp
Program. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
Table 4-8 Recoupment Households by Months Participation
In Program and Recoupment Liability ..•...••.....••.....• 85
Table 4-9 Characteristics of Recoupment Households:
Household Composition................................... 87
Table 4-10 Characteristics of Recoupment Households:
Primary Earner. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
Table 4-11 Characteristics of Recoupment Households:
Sources of Income. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
Table 4-12 Characteristics of Recoupment Households:
Unemployment of Primary Earner ...........•..•.••..••..•• 92
Table 4-13 Characteristics of Recoupment Households:
Number of Secondary Earners............................. 94
Table 6-1 Recoupment Information Record-Keeping ...•••.•.••.•.•.... 126
Table 7-1 Sunnnary of IRS Cost Estimates........................... 193
xxxix
•
CHAPTER 1
INTRODUCTION
Recoupment is a partly new, partly old idea. The specific recoupment pro-posals
examined in this report are, in a narrow sense, relatively new, having
first received close public scrutiny at the time of deliberations on the
Food Stamp Act of 1977. In a broader sense, the underlying policy issues
and problems raised by the proposals are as old as public assistance programs
themseleves.
This introductory chapter reviews the legislative history of the proposal,
discusses the rationale for recoupment, and briefly outlines some of the
key policy questions. Against that background, Chapter 2 then describes
the details of the proposal.
1.1 LEGISLATIVE HISTORY
When the proposed Food Stamp Act of 1977 first came up for consideration
by Congress, it did not contain a recoupment provision, nor did the program
itself at that time have any such provision. During the House Agriculture
Committee's debate of the bill, Representative James Jeffords from Vermont
offered an amendment with a recoupment plan involving, among other distinc-tive
features, a link with the Internal Revenue Service. As the Committee
Report later noted, the Jeffords Plan
••• would have required food stamp recipients whose adjusted (tax)
gross income exceeded twice the poverty line in a calendar year
to pay back in cash some, if not all, of the food stamp benefits
they received in that year. They would have to report their
year's cumulation of coupon allotments in a special box on their
tax form and reimburse the Treasury for that full amount by which
1
their adjusted (tax) gross income exceeded twice the poverty
line for a household of their size. In many instances, the
sum they owed would simply be subtracted from their tax
return ••• tc Committee Report, p. 365
For others, the recoupment amounts would have to be paid in cash. In
order to mitigate potential hardship, any household still receiving food
stamps when it filed its tax return could defer the payment of any recoup-ment
liability in excess of its tax refund. Also, the Department of
Agriculture in consultation with the Treasury Department could waive any
collection that would result in undue hardship.
The Jeffords plan raised numerous questions that could not be completely
answered from the data then available. How much money conceivably could
be recouped? How much in fact could be collected? How many households
would be subject to recoupment, .what would be their general characteristics,
and how, if at all, would they .differ from other food stamp participants?
Would recoupment be difficult and costly to administer? Would the link with
IRS be workable? To what extent would the filing of income tax returns be
complicated not only for food stamp households but also for more than 90
million other tax filers nationwide? Would the savings in recouped benefits
outweigh the administrative costs and problems? Would the prospect of recoup-ment
create major work disincentives?
In the end, the House Agricultu~e Committee rejected the amendment on a
recorded vote of 21 to 23. A similar amendment introduced on the House floor
was defeated by 149 to 262. The Senate took no action on recoupment.
*The text of the Amendment is included in Appendix D to this report. ·
2
The House did agree, however, to mandate USDA to perform a study "of
the feasibility, alternative methods of implementation, and the effects
of a program to recover food stamp benefits from members of eligible
households in which the adjusted gross income ••• may exceed twice the
income poverty guidelines ••• " (Sec. 17 (d) of the 1977 Food Stamp Act).
The conference committee retained the study provision in the final bill
enacted on September 29, 1977. This report presents the study's find­ings.
1.2 THE RATIONALE FOR RECOUPMENT
Underlying the 1977 deliberations on recoupment was a general concern
that food stamps should be available only to those who truly need them,
and only to the extent of the actual need. It was noted that certain
households in special circumstances could legally receive food stamps
even though their gross annual income seemed high relative to the net
monthly income limits used for eligibility determination.
Under the old program rules of the 1964 Act in existence at the time,
these special cases could arise for either (or both) of two primary
reasons. First, a household might qualify for large deductions reducing
a comparatively high gross income in a given month to a much lower net
figure for that month. Second, a household's income might vary greatly
over the course of a year, so that even though eligible for food stamps
during some months when its income is lowest, the household might still
have a relatively high income for the year as a whole.
3
When the Jeffords Amendment was introduced, many thought that recoupment
was an answer to both these problems. This was true as long as the old
program rules were retained. However, the 1977 Act as finally enacted
completely revamped the types and amounts of deductions allowable, and
greatly limited the extent to which gross income could exceed net income.
Households that would have been subject to recoupment under the Jeffords
plan because they qualified for very large deductions were made ineligible
for food stamps under'the new Act. In short, the first rationale for recoup-ment--
large discrepancies between gross and net income within a given month--
was effectively eliminated by other tightening provisions of the new Act.
The change in the deductions procedure mainly involved switching from
itemization to the use of uniform standards. Under the 1964 Act, itemized
deductions were figured for each household individually, reflecting its
actual medical bills, withholding tax, FICA, work related child care expenses,
and shelter expenses (in excess of 30 percent of income net of all other
deductions). There was no limit, explicit or implicit, on the total deduc-tions
permissible, and hence no limit on how high gross income could be.
The 1977 Act set a limit on child care and shelter deductions* and replaced
the rest with a standard deduction and a uniform work expense allowance.
The standard deduction, recently raised to $70 per household per month
because of inflation, covers miscellaneous needs formerly treated separately.
The work expense allowance reduces earnings by 20 percent to allow for taxes,
FICA, union dues, etc.
*The limit was originally set at $75 per household per month, but is now
$90 due to adjustments for inflation. Both deductions singly, and their
sum, must remain within the limit. Also, the 1977 Act changed the shelter
expensededuction to count shelter costs in excess of 50 percent, rather than
30 percent, of net income.
4
In effect, these changes establ~sh a ce~l~ng on gross inco~e. For a
family of four, for example, the maximum standard and itemized deductions
permit eligibility of households with a maximum gross monthly income of $945,
or $11,340 annually.* For other household sizes, the ceiling is different,
but in all cases it is well below twice the poverty line. Since households
are recoupable under the Jeffords plan only if their gross (annual) income
exceeds twice the poverty line, it follows that no household eligible for
benefits under the 1977 Act can also be subject to recoupment solely due to
the difference between gross and net income within a given month.
The second possible reason for recoupment--variation in income over the
course of a year--was not eliminated by the program changes incorporated
in the 1977 Act. It was relevant under the 1964 Act's rules and has remained
no less so since. The basic issue involved was succinctly set forth during
the House Agriculture Connnittee's deliberations by means of the following
example:
Household A earns their income at the rate of $1,000 per month
for 12 months. They are not eligible for food stamps, nor did
they have taxes overwithheld by their employer. Household B
receives $1,200 for 10 months and is eligible to receive maximum
food stamp benefits for the other two months. Household B is
also eligible for a tax refund of $359 because of overwithholding.
With a recoupment system, Household B's tax refund would be
reduced by $340--equal to their two month's coupon allotment value-­reinstating
equality with Household A. Connnittee Report, p. 365.
* $945 - (}70 (Standard) + $189 (Work-related expenses) + $90 (maximum
shelter cost and/or . c~;i,ld C?~e deductioti) = $596- and qualifies tl1e house~
hold for a $26 food stamp benefit. "Note that households containing elderly
members can have higher gross monthly incomes because of unlimited shelter
and medical deductions.
5
The figures in this example are hypothetical. Nevertheless, even when
numbers reflecting the actual provisions of the 1977 Actare used, the
same issue arises.
Households C, D and E (see Table 1-1) each have four members and an
annual gross income of $l.f,_900. Since the poverty line for house­holds
of this size is $•7 ,150 these households all are at an income
more than twice the poverty fine. Household C earns its income
regularly during the year and is never eligible for food stamps.
Earned income in households D and E varies over the year and, as
a consequence, these households are eligible for food stamps in
some months.
The principal earner in household D has a two month unemployment
spell and collects $3 60 in Unemployment Insurance benefits for
each month of unemployment. In those months, the household's
net monthly income under the 1977 Act is $ 200 and the household
receives $14 4 of food stamp benefits. The net monthly income is
derived as follows: $3 60minus the $ 70 standard deduction is
$ 290 the household 1 s shelter costs exceed one-half this amount
and the household qualifies for the full $ 90 shelter cost deduc­tion;
and so net monthly income is $ 200.
The principal earner in Household E earns a gross monthly income
of $goo per month throughout the year. In the fourth month of
the year, the spouse of the principal earner takes a job and has
gross monthly earnings of $525. The combined gross earnings of
$1.425 render Household E ineligible for food stamps for 8 months
of the year. However, in the first four months of the year the
household is eligible for $36 in food stamps each month, since it
has a net monthly income of $560 . The net income is derived by
substracting the following deductions: $ 70 (standard deduction),
$ 180 (work related expenses or 20% gross monthly earnings) and
$ 90 (shelter cost deduction).
Recoupment would affect households with varying incomes, like B, D, and E,
but not households with constant incomes. In addition, to be subject to
recoupment, a household's income must not only vary over the course of a
year but also be relatively high for the year as a whole: under the Jeffords
plan, it must be more than twice the poverty line. As a result, recoupment
would concern a comparatively small and atypical segment of the food stamp
recipient population.
6
Table 1-1
Household C Household D Household E
Gross Gross Gross
Monthly FS Monthly FS Monthly FS
Month Income allotment Income allotment Income allotment
Jan. 1250 0 1500 0 900 36
Feb. 1250 0 1500 0 900 36
Mar. 1250 0 1500 0 900 36
Apr. 1250 0 1500 0 900 36
May 1250 0 360 144 1425 0
June 1250 0 360 144 1425 0
July 1250 0 1500 0 1425 0
Aug. 1250 0 1500 0 1425 0
Sept. 1250 0 1500 0 1425 0
Oct. 1250 0 1500 0 1425 0
Nov. 1250 0 1500 0 1425 0
Dec. 1250 0 1500 0 1425 0
Annual Adjusted
Gross Income (AGI) 15,000 15,000 15,000
Amount ·:.b~ which
annual AGI exceeds
.. $14,300 (twice the poverty line)* 700 700 700
Annual FS
Benefit 0 288 144
*Poverty line as of July 1, 1979
7
Proponents of recoupment generally note that although these households may
have been in need of aid at the time they received it, their overall annual
circumstances enable them to repay some or all of it later. There is thus
a presumption that such households should continue to be entitled to
assistance during their period of need, but that they also should "settle
accounts" annually. As Representative Jeffords put it,
The proposal acknowledges that a household with a relatively high
income may have a temporary need. For such families, Food Stamp
benefits would simply be considered as an interest-free loan, to
be paid back in a painless manner through deduction from the family's
income tax refund, or on an interest free, penalty free schedule,
after the household is off food stamps. Committee Report, p. 837
The ultimate aim is to correct some of the perceived longer term inequity
problems in the program while possibly also saving taxpayers' dollars.
However, this simple objective involves many complex issues, as the next
section and following chapters discuss.
·1.3 UNDERLYING ISSUES AND PROBLEMS •
Recoupment would have the effect of reducing the benefits that certain
households receive in the long run, after annual repayments have been
settled. Since most of these households would be near the upper end of the
income distribution of food stamp participants, the result would be to con-centrate
a larger share of benefits on poorer households. This raises the
issue of how recoupment would affect the target efficiency of the program.
A public assistance program is target efficient if it can reach all those
de:emed truly in need (the target population) and at the same time exclude
all those not intended to be aided. Few disagree that target efficiency
is desirable, but there can be differing views of what the target popula-tion
is. For those who feel the target population of the Food Stamp
8
•
Program should not in9lude households with gross annual incomes more
than twice the poverty line, recoupment would be a move toward improving
the program's target efficiency. For those who believe such households
ought to be eligible for temporary assistance, recoupment would recover
benefits rather than remove deserving participants •
Other fundamental goals of the Food Stamp Program besides the degree of
target efficiency would also be affected by recoupment. These program­matic
features are discussed in subsequent chapters and can be briefly
summarized as follows:
Work Incentives. To what £Xtent do those who are able to work find it in
their interest to do so? ·In particular, for each dollar they earn, how
much are their food stamp benefits reduced, and thus how is their net
overall financial situation altered?
Horizontal Equity. To what extent are people in similar circumstances
treated the same?
Vertical Equity. To what extent do people with greater needs always receive
more benefits? This is related to, but not the same as, target efficiency.
Responsiveness. To what extent do benefits adapt quickly to reflect changes
in recipients' circumstances?
Adequacy. To what extent are benefits adequate according to some standard
of a minimum support level?
9
Administrative ~easibility. To what extent can a new policy (such as
recoupment) be implemented without undue administrative burden?
Clarity and Simplicity . To what extent can programs and policies be
easily understood by everyone?
Thes~ general programmatic goals frequently conflict with one another--a
fact that has been a continuing source of difficulty for welfare (and food
stamp)reform efforts over the years. For instance, perfect target efficiency
would require that no household receive more benefits than the minimum
needed to maintain it at the program standard for households of that type
as defined by a set of program criteria such as household size and income.
This would mean that benefits ought to be reduced one dollar for each addi­tional
dollar of income the household earns, or in other words, that the
program should have a 100-percent benefit reduction rate.
However, a 100-percent benefit reduction rate provides no inducement to
work. From the perspective of creating a work incentive, the benefit reduc­tion
rate should be as much lower than 100 percent as possible. A very low
r ate, though, would undercut yet another goal--vertical equity. When deter­mining
benefit reduction rates or any other aspect of the program, indivi­dually
desirable goals are not always mutually attainable.
Much hinges on the distinction between the monthly accounting period as the
basis for eligibility determination and the yearly perspective of recoupment.
Perceptions of recoupment sometimes differ according to which time frame is
stressed. Behind this lies a fundamental dilemma that confronts not only
10
-~· .
•
•
food stamps but all public assistance programs. If short accounting
periods--like a month--are used, longer term trends in.,individuals'
incomes cannot easily be taken into account. On the other hand, if longer
periods--like a year--are used, it is impossible to respond to sudden
changes in individual circumstances that create severe short-term needs .
Recoupment is one of several classes of strategies for attempting to
satisfy both sets of concerns simultaneously.
Finally, an important factor in the recoupment debate is that almost
nothing has been known in the past about the characteristics of the house­holds
that would be subject to recoupment. Would most of them be receiving
benefits for only a few months a year or for many months? Would most of
them have short spells of unemployment, serious unemployment problems or
be migrant workers? Some of these questions can be answered with the data
presented in this report and some cannot. Clearly, though, information
of this kind may have a considerable effect on the decision that must
ultimately be made.
1.4 PRESENTATION OF THE REPORT
Against this background of the trade-offs among worthy and competing pro­gram
objectives, different approaches to achieving target efficiency, and
the experience of recent reform efforts, this report will examine the pro­posal
for recoupment of food stamp benefits from households with relatively
high annual incomes.
Chapter 2 describes in greater detail the recoupment plan proposed to the
House of Representatives, and it analyzes the general strengths and weak­nesses
of that plan. In addition, it will discuss alternative specifications
11
and methods of implementation that might mitigate some of the problems
with proposed recoupment schemes.
Chapter 3 presents estimates of the amounts that might be recouped through
the proposed plan and through various alternative recoupment plans. The
focus of this chapter is on the pecuniary · benefits of recoupm~nt, the
amounts collected from recipients or offset from their tax returns.
Chapter 4 looks at the food stamp recipients (and potential recipients)
affected by recoupment. It describes recoupment households and the distri­bution
of recoupment amounts across households.
Chapter 5 presents the various equity and incentive issues surrounding
recoupment: the equity of benefit allocation plans, possible work dis­incentives,
issues of tax compliance and potential deterrence to partici­pation
and fraud.
Chapter 6 discusses the effects of recoupment on State and local welfare
agencies and presents some preliminary estimates of the feasibility and
costs of performing the tasks required of these agencies by a recoupment
scheme.
Chapter 7 addresses the feasibility of implementing recoupment through the
Internal Revenue Service. It discusses the administrative procedures, and
the consequent costs and effort, necessary to implement an annual recoupment
scheme through the personal income tax system.
12
•
•
•
CHAPTER 2
WHAT RECOUPMENT IS AND HOW IT WOULD WORK
This chapter elaborates on the brief description of the Jeffords plan in the
•
previous chapter, discusses some possible variations on the original plan,
and examines several underlying features shared by all recoupment plans •
2.1 INTRODUCTION
Recoupment is a way of allowing a program such as food stamps, which is based
on a short (i.e., monthly accounting period), to take into consideration
applicants' income streams over a longer (i.e., annual) horizon. In the design
of any recoupment plan, three parameters are critical: the allocation
scheme, the threshold income level, and the recoupment rate •
• The allocation scheme is the method by which benefits
are attributed to members of the household for the
purpose of recoupment. It determines which house­hold
members, and therefore how many dollars, are
subject to recoupment as administered through the
federal tax system.
. The threshold income level is the income at which
recipients become liable for repayment of some or
all of the benefits they have received .
• The recoupment rate is the rate at which the benefit
amount is recouped for each dollar of income over
the threshold. For example, if the recoupment rate
is 100 percent, then a household with a gross annual
income that is $200 higher than the threshold will
pay back $1 in benefits for each dollar of income
over the threshold up to the full amount of benefits
or the full $200 whichever is less .
13
A fourth element of a recoupment plan, the definition of income, is also
important •
• The definition of income subject to recoupment
determines which sources of income are considered •
and which are excluded.
Together, these basic parameters define the target population and determine
the effects that recoupment would have on cost savings, on recipients, and
on administrative burden. The assignment of administrative responsibilities
are also important, not only because administrative costs affect the net
savings from recoupment, but also because the administrative arrangements
affect the definitions of filing units' income.
2.2 FURTHER DETAILS OF THE JEFFORDS PLAN
The Jeffords plan would recover benefits from recipient households with
annual incomes over twice the poverty line, that is, twice the program's
annualized net eligibility limits. Food stamp benefits received during the
year would be reported with other income on tax forms, and recipients would
have to reimburse the Treasury of the amount by which their adjusted gross
income exceeded twice the poverty line, up to the full amount of benefits
received. Thus, in the Jeffords plan the threshold income level is twice
the poverty line and the recoupment rate is 100 percent.
Under this plan, monthly benefits would continue to be calculated on the
basis of anticipated monthly income; however, for those recipients who have
14
..
..
relatively high incomes over the course of the year, the benefits would
become an interest-free loan to help them during periods of temporarily
low income. State and local welfare agencies would maintain a record of
the monthly benefits received by each household during the year.* In
January of every year they would prepare and send out year-end reports
like W-2 forms (hence referred to hereafter as "W-X forms") bearing the
amount of the annual allotment, so it can be reported on and submitted with
Federal tax returns. They would also provide the same information to the
Internal Revenue Service on computer tapes.
Because of the significant differences between the food stamp unit and income
tax filing unit, some provision would have to be made for apportioning benefits
among household members. Under the Jeffords plan, liability for the full
amount of food stamps received by a household would be assigned to the
individual (or married couple) in the household who provided at least 80
percent of the cost of maintaining that household during the calendar year.
If no individual or couple provided a full 80 percent of such costs, the
liability would be prorated among members of the food stamp household according
to their relative contributions to the cost of maintaining the household.
Thus, in the Jeffords plan the allocation scheme is a pro-rata procedure
figured on an annual basis. The procedure entails having some food stamp
households divide their benefits, and their potential liability for recoupment,
among two or more tax units.
*In States where authorization to participate (ATP) cards are issued, welfare
agencies would reconcile these issuance amounts with the cards that were
redeemed for food stamp coupons and correct records to reflect actual benefits
received.
15
The Jeffords plan further proposes that the collection of recoupment
liabilities would be administered by the Internal Revenue Service, which
already collects annual income information. Each tax unit that receives
benefits during the year would calculate its recoupment liability either on
the Form 1040 (or 1040A) or on a special form to accompany the 1040. The
taxpayer would look up its recoupment income threshold, equal to twice the
poverty line for the appropriate tax unit size, on a table included with the
tax filing instructions. If the tax unit's adjusted gross income (AGI) as
reported on the tax form were less than the relevant threshold, there would
be no recoupment liability. If its adjusted gross income were higher than the
threshold, the tax unit would have to pay back some or all of its benefits.
The resulting recoupment liability would be equal to the excess of AGI
over the threshold, or the full amount of food stamp benefits received during
the year, whichever were less. This liability would be subtracted from any
tax refund otherwise due the taxpayer. Any excess of the recoupment liability
over a refund would be payable to the Treasury. However, payment of this
balance would be deferred if the taxpayer were a member of a household
receiving food stamps "at the time prescribed by law for the payment of
Federal income tax." At the point when the household is no longer receiving
benefits, its liability would be due. The Internal Revenue Service could
arrange a schedule bf payments for these liable units when appropriate.
The design of the proposed plan attempts to concentrate recoupment efforts
on households with the highest incomes and minimize the necessity for
additional bureaucracy. It succeeds in certain aspects of the program,
16
•
but nevertheless would still require a substantial increase in Food
Stamp - ~rogram and IRS administrative effort, and . ~t would raise many of
the concerns inherent to any recoupment system as well.
The Jeffords plan would not affect program responsiveness. Since the
monthly accounting period would still be the basis for eligibility to
receive benefits, the program would be no less responsive to changes in
low-income households' situations (e.g., sudden loss of income) if the Jeffords
plan were adopted than otherwise. In this respect, the recoupment is usually
described as preferable to other conceivable approaches to accounting for
longer term (e.g., annual) income information into a monthly-based program,
since most other approaches would reduce responsiveness.
How the plan would affect the program's target efficiency is partly a matter
of opinion. In so far as recovering benefits from recipients with relatively
higher annual income is regarded as targetting the program more precisely on
the groups the program is meant to serve, then the Jeffords plan would improve
the program's target efficiency. However, this conclusion depends on one's
view of what the target population should be and over what time frame the
population's income should be measured. Those who hold that recoupment
households should be entitled to keep the benefits they legally receive in
any month will conclude that the Jeffords plan would not enhance target
efficiency.
17
Recoupment would improve horizontal equity by treating households that
have the same annual incomes more equally. However, there are limits to
this improvement for all households would not be treated exactly the same.
Because of the exclusion of transfer income from the IRS determination of
adjusted gross income, there would still be an advantage to higher income
households that re~eive unemployment compensation or welfare. And because the
recoupment threshold (at twicetheannual poverty level) would be higher than
the Food Stamp Program's monthly eligibility limits, there would still be
some advantage to uneven income flows during the year. Nevertheless, the
recoupment plan would mitigate inequities.
On the other hand, targetting on the households with the highest incomes
.
and using the mechanism of the Federal income tax does not eliminate all
the problems and concerns of recoupment. As with any recoupment plan,
concerns about hardship, inequity and work incentives remain. Moreover,
using the tax system may create new problems. The proposed use of tax
definitions of filing units and income poses significant administrative
obstacles and additional equity concerns. It is to some of these issues
that the chapter now turns.
18
•
•
2.3 ALTERNATIVE ALLOCATION SCHEMES
As has been discussed, the allocation scheme used to apportion a household's
benefits among its members in a critical design element of any recoupment
policy. The definition of a filing unit, and consequently the method by
which records are kept and calculations performed, differs between the
two agencies proposed to administer recoupment: food stamp agencies and
the IRS. For the Food Stamp Program, all persons living together and
customarily purchasing food and preparing meals together for home consump­tion
must apply together for food stamps. Income tax filing units, however,
normally consist only of related individuals, and even related persons may
file separately. The income tax filing unit is the individual; the food
stamp filing unit is the household, and it may contain several tax units.
Therefore, the benefits received by any household must be apportioned to its
component tax units -- in other words households must somehow be translated
into tax units -- if recoupment is to be effected through the tax system.
During the present analysis, it was found that the allocation scheme proposed
in the Jeffords plan would have several serious administrative problems.
Rather than simply stop there, an effort was made to develop and examine
alternative allocation schemes that would preserve the intent of the plan as
much as possible, but would be more feasible administratively. Three
alternatives were eventually selected. This section outlines them briefly,
leaving further analysis of them for subsequent chapters.
The alternative schemes were devised after consideration of several potential
criteria for evaluating a scheme's attributes: the amount of information
required to perform the allocation, the degree of accuracy in performing the
calculation itself, the accuracy of the data collected and the probability of
verifying those data, the equity of the allocation plan, the extent to which
19
the allocation maximizes or minimizes the amount of benefits subject to
recoupment, and administrative feasibility and cost. Just as there are
trade-offs in balancing various program goals such as target efficiency and
responsiveness, th~re are trade-offs to be made in the design of an
allocation scheme.*
The efficient performance of a task may be measured by the degree of error
associated with the result of the task, and the possibility of error rises
directly with the complexity of the task. Generally, errors arise from two
sources: when information is inaccurately reported and when agencies make
calculation errors. These two types of error increase with the number of
variables (or pieces of information) and the increase probability that
these variable change over time. For example, allocations based simply
on household size would be relatively straightforward while allocations
based on expenditures such as utilities or special consumption needs add to
the size and complexity of both client and agency responsibility. Thus it is
desirable to minimize the amount of information required by the allocation
formulation and to use readily accessible and verifiable information. The
above criteria become even more important in light of the fact that information
for benefit allocation must be collected for all households that ever
participate in the Food Stamp Program during the course of the year. The
vast majority of food stamp householdsdo notparticipate in the program
in all twelve months of a calendar year and whether a household is subject to
recoupment is a function of the income of the household members in those months
*See Chapter 6 for a more detailed discussion of the administrative
implications of alternative allocation schemes.
20
•
•
•
•
when the household does not participate. The food stamp agency does not have
information about the household for those non-participating months and has no
way of predicting whether a household receiving food stamps will ultimately
be subject . to recoupment at tax time •
These considerations have an important bearing on the design of allocation
schemes. Section 2.3.2 through 2.3.4 below describe the ~ree alternatives
to the original Jeffords scheme. To set them in context, Section 2.3.1
provides additional details on the original scheme.
2.3.1 The Jeffords Allocation Scheme - Pro Rata on an Annual Basis
Recall that under this scheme, liability for the full amount of food stamps
received by a household would be assigned to the individual (or married couple)
in the household who provided at least 80 percent of the cost of maintaining
that household during t he calendar year. If no individual or couple provided
a full 80 percent of such costs, the liability would be prorated among members
of the food stamp household according to their relative annual contributions
to the cost of maintaining the household.
This procedure would require collection of detailed financial data from food
stamp households, including each household's total annual expenditures and
the annual contributions of each member to cover those expenses. Neither
annual income nor expense data nor individual financial data are now collected
by the welfare agencies, The IRS, on the other hand, only collects annual
income of each tax unit. The kinds of intrafamilial data that are needed for
21
recoupment are not only difficult to collect; they are also difficult to
monitor and verify without substantial expenditures on the part of the agency
collecting the information.
The determination of annual contributions to household maintenance necessitates
collection of income data as well as expenditure data. In order to collect
this information on a retrospective calendar year basis, a survey would have
to be conducted of all persons who were in households that ever received
food stamp benefits during the year. Bill stubs and records would be needed
for verification as well as a supplement to memory as members of a household
attempt to reconstruct their "household maintenance account books" for the
year.
If the composition of the household changed during the year--for example,
if a son married and left the household--the data collection would become
more difficult. Any composition change would create the dilemma of determining
which members constituted the household for the year; i.e. whose income and
expenditures to include. It may be problematic to locate families that have
moved; certainly it would be costly and time-consuming, and in some cases
impossible. Furthermore, the detailed and confidential nature of the survey
would require in-person interviews--the most expensive of all data collection
techniques. The survey itself would have to be conducted at the close of
the calendar year and completed in a very short time frame to prepare and send
out the food stamp W-X forms prior to the earliest tax filing date of
January 31.
22
•
•
•
In this allocation scheme, the calculations performed by the food stamp
agency are complex, but particularly so in those cases where household
composition has changed or where no single member or married couple con­tributed
80 percent to household maintenance. As noted earlier, the complexity
of the calculations and the volume of information to be used open the possi­bility
for error in the administration of this plan. The relatively hi gh
probability of error is compounded by the speed with which data must be
gathered and allocations made to meet the IRS deadline.
Indeed, as discussed more fully in Chapter 6, the problems are severe enough
and expensive enough to warrant avoiding this method of allocation if at
all possible.
2.3.2 Modified Jeffords Allocation Scheme--Pro Rata by Certification Period
The modified Jeffords scheme, the first of the three alternatives developed
for this report, retains the basic elements of the original scheme, such -as
having the threshold at twice the poverty level and the recoupment rate at
100 percent. However, relative contributions to household maintenance
expenses are determined for each food stamp certification period rather than
for the year. It is easier for the applicants to supply this detailed informa­tion
when they are at the welfare office for their certification interview,
and also easier for food stamp agencies to collect and record it at these
periodic intervals. Each allocation is frozen for each certification period,
and totalled at the end of the year. No end-of-the year survey would be
required by this plan. However, detailed information on expenditures and
23
income would have to be collected and verified at each certification and
recertification. Unlike the original Jeffords plan, which is based on
retrospective information, this allocation plan is based on a prospective
estimation of contribution to maintenance. If any changes in contribution to
maintenance occurred during the certification period, the recipients would
be required to report the change and provide documentation before a change in
allocation is performed. The period after the change would be treated as a
new certification period.
Information collection on household expenditures was recently substantially
reduced and simplified by the institution of standard deductions through the
Food Stamp Act of 1977. It was accomplished not only to reduce the income
eligibility limits, but also to ease the administrative burden of gathering
all the specific, individual data as required in a personalized benefit
determination. Both of the Jeffords schemes counteract the second goal by
requiring substantial expansion of the information collected for the
computation of the benefit allocation. All of this information would then have
to be stored for several years under food stamp regulations in case the
allocations made were challenged.
A slight variation of the modified Jeffords scheme based on the relative
incomes (rather than contributions to maintenance) of household members is
also possible. This would simplify the allocation procedure somewhat.
Food stamp benefits would be deemed to each household member in proportion
to his or her income as a percent of the total income of the household.
24
•
•
This approach would be more consistent with the original determination of
benefits - which is based on total household income - and also more feasible
to implement on the basis of data now collected. The allocation could be
calculated and recorded for all food stamp households, and reported in
separate W-X forms for each person with income during the year. This would
mean a substantial increase in the income data to be stored and processed
by program agents, but would avoid the need to gather expenditure information,
which would make the greatest demands on computer capacity and costs. For
either of the pro rata allocation schemes, the original or the modified
Jeffords plan, unusual situations may arise. When the allocation is made
to the individuals who financially support the household, the problem of
assigning financial responsibility to minors with income must be faced.
To avoid expanding the amount of income information collected by the food
stamp offices to perform the allocation, the IRS could be made responsible
for this function. In order to do this, however, the IRS would essentially
have to translate its tax unit information back into household data. Each
tax unit would have to provide income data regarding each member of the
household, whether or not that person were in the tax filing unit. This
would allow the IRS to determine annual household income by cross-checking
the information provided on the tax returns of all component tax units to
verify consistent reporting of relative incomes, and therefore of relative
recoupment liabilities. This would be a formidable task for the IRS, and
one that would require extensive administrative effort and expense in an
area outside of their normal operations.
25
2.3.3 Allocation of Tax Unit
This plan divides household benefits in proportion to the relative sizes of
the component tax units. Thus if a household contained two tax units of
3 and 2 people each, three-fifths and two-fifths of the benefits would be
allocated to each unit, respectively. Again, benefits are allocated at
certification, the proportional division is frozen for that certification
period, and liabilities from each period are added together at the end of
the year. The plan involves a prospective determination of tax units.
Whether households or food stamp agencies can accurately predict the compo­sition
of their tax filing units so far in advance of tax filing time is a
serious issue. The best that could be done would be to use conservative
assumptions in assigning tax units: the household would be divided into the
maximum number of logical and feasible units according to guidelines
established for the program. Once the allocation by tax unit size is made,
it cannot be changed without involving extensive administrative problems.
Unfortunately, in practice, this allocation method may result in substantial
leakage and avoidance of recoupment liabilities. The amount of leakage
would probably be significant enough to make this allocation scheme
impractical and cost ineffective. Leakages would occur if the tax units(s)
identified within a household at the time of food stamp receipt were not
the same tax units that filed with the IRS at tax time.
26
2.3.4 Per Capita Allocation
This method of apportioning benefits would require fewer pieces of information
and simpler calculations than the previous three schemes and would more
closely parallel current IRS procedures dealing with individual income.
However, the results would be furthest from the intent of the original Jeffords
plan to allocate benefits according to individual contributions to household
maintenance.
For each certification period, the total benefits received by a household
would be divided by the number of household members. Each individual's
per capita share would be entered onto his or her file. At the end of the
year, the results from all certification period would be summed, reported on
a weparate W-X form for each member, and mailed to each individually. When
a tax unit forms, it would file the W-X's only for those household members
claimed as part of that unit. If a change in the composition of a household
occurs during the year, past benefits move with each individual and current
benefits are re-allocated simply according to household size.
Although simpler than the other schemes, this approach is not without problems.
As Chapter 6 discusses, there would be several significant administrative
issues. Chief among these is the requirement that all recipients have
validated social security numbers. Legislation passed in the summer of 1979
provides for recording of social security numbers on all food stamp recipients.
However, validation of social security numbers would be an additional task
undertaken only for purposes of recoupment. Validation would involve
checking that the number provided by the recipient is accurate in relation
27
to the master files of 'the Social Securfty Administration. The Internal
Revenue Service maintains that validation is essential because social security
numbers would have a central role in identifying recoupable benefits with
the correct liable person. If a number is incorrect by even one digit, the
IRS would have no way of matching that individual's W-X form with the tax
return it receives. However, the added time and cost for filling out the
validation forms at certification are significant.
Tax unit alterations are possible with this plan by shifting dependents
among the household's tax units to minimize liability. Notwithstanding
this fact, benefits allocated to dependents are unidentifiable on tax
returns since the reporting of dependents' social security numbers is not
required under current IRS procedures. Therefore, no matter who claims the
dependents on tax returns, there is no way of checking whether dependents'
W-X forms are filed with those returns, resulting in the leakage of potentially
recoupable benefits.
2.4 ALTERNATIVE RECOUPMENT RATES
As Section 2.1 noted, the recoupment rate is the fraction of income above
the threshold that is liable for recoupment. Just as a recoupment plan
can have different allocation schemes, it can also have different recoupment
rates.
The recoupment rate determines the effect of the plan on marginal tax rates,
and consequently, the impact on work incentives and vertical equity. Also,
28
in combination with the threshold, the recoupment rate is a determinant of
the recoupment liability at any particular income level, the range of
income over which liabilities are phased-in, and the total amount of recovered
benefits. The higher the recoupment rate, the more the recoupment plan would
raise marginal tax rates, posing concerns about work incentives. On the
other hand, a higher recoupment rate would result in higher collections and
savings. Conversely, the lower the recoupment rate, the lower the adverse
impact of the plan on incentives. But a lower rate also reduces savings and
increases the range of income over which recoupment households are subject
to higher marginal tax rates by extending the income range over which benefits
are recovered.
Recoupment ·rates can be divided into three distinct classes.
All benefits could be recouped as soon as the income
of the household or filing unit exceeded the threshold.
This would create a "notch" at the threshold, where a
few dollars of earnings could result in a precipitous
drop in total income (net of the recoupment liability) •
• Benefits could be recouped at the rate of one dollar
for each dollar of income in excess of the threshold.
This would be a 100 percent recoupment rate, and it
would cause total marginal tax rates to go substan­tially
over 100 percent, at least 125 percent if the
earner is also subject to income and payroll taxes.
This means that for the entire income range over
which benefits are recouped, each extra dollar of
earnings will result in an absolute decline in
disposable income.
Benefits could be recouped more gradually, at a rate
lower than 100 percent. Thus, the recoupment rate
could be set such that an extra dollar of earnings
always resulted in higher disposable income.
29
Exhibit 2-1 displays examples of each of these possibilities graphically,
including two recoupment rates less than 100 percent--50 percent and
25 percent. The examples describe the impact of recoupment on the net income
(shown on the vertical axis) of a single-tax-unit household receiving $500
in food stamp benefits annually. The household has an annual gross income
(shown on horizontal axis) large enough to be subject to taxation. If the
household is only barely taxable, it would appear near the origin of the chart.
As its income increases, it moves up the solid line to point A, where its
annual gross income reaches twice the poverty line. Below point A, it is not
subject to recoupment. Its net income is less than its gross income by t.he
amount of its income taxes and social security taxes.
When the household's gross income exceeds "2 PL" (twice the poverty line)
on the horizontal axis, recoupment begins. If a "full recoupment at
threshold" policy is in effect, it drops immediately from point A to point B,
a decrease in its net income of $500 (the amount of food stamp benefits it
received). Thus, there would be a significant "notch" effect: a small
increase in gross income above twice the poverty line would lead to a sub­stantial
decrease in net income, due to recoupment. Beyond point B, if the
household's gross income rises further, it moves up line BF. On BF, the
relation between gross and net income depend on income taxes and social
security taxes, the same as on line OA (the two lines are parallel).
Alternatively, in the case of a 100 percent recoupment rate policy such as the
Jeffords plan, an increase in the household's gross income above twice the
poverty line carries it along line segment AC. There is not the severe
30
AGI*
minus
taxes
and
recoupment
w
1--'
2·PJ,
Tax
Threshold
*Annual Gross Income
! ,.
Exhibit 2-1
Effect of Alternative Recoupment Rates on a Single-Tax-Unit Household
with a 5.500 Recouome.nt T.iahi] ity
F
2 PL +500 +1000 +2000
~~e~~e~~ea--ee~ full recoupment
Q at threshold
100% rate
50% rate
X X X X X
25% rate
AGI*
notch effect of a "full recoupment at threshold" policy, but the household's
net income does decrease along line AC, as taxes and recoupment together
take away more than one dollar every time gross income increases by a dollar.
Once the household's income reaches point C ($500 above twice the poverty
line), it has paid back all its benefits, and thereafter moves up line CF.
Under a SO percent rate, the household followl line AD and then DF. In this ..
case, net income does not decrease as gross income increases within the range
where benefits are being repaid (AD), or "phase-in range." Taxes and
recoupment together take back less than a dollar from every dollar of
additional income. At the same time, though, the phase-in range is larger
than before: the household does not pay back all of its benefits until its
income reaches $1000 above twice the poverty line. Lower recoupment rates
(e.g., 25 percent) result in a higher increase in net income for every
additional dollar of gross income, and a larger phase-in range.
As has been said, a low rate may reduce the savings obtained from since less
of the total potentially recoupable benefits would actually be recovered.
However, whether the proportion of benefits not recovered would be large
or small depends on many factors, including, in particular (1) the amounts
of benefits households rec~ive (which, as the $500 does in the above examples,
affect the size of the phase-~n range) and (2) how far households subject
to recoupment are above the threshold (which determines where they lie within
' the phase-in range). If, for instance, households subject to recoupment
receive relatively small amounts of benefits or have incomes only slightly
above the threshold, it is possible that varying the recoupment rate would not
alter the proportion of the benefits recovered as substantially as one might
first think. This subject is explored further in Chapter 3.
32
CHAPTER 3
RECOVERY OF BENEFITS
This chapter presents the study's findings on the number of households and
tax units that would be subject to recoupment under the Jeffords plan, the
sums they would owe (their recoupment liabilities), and various aspects
of the collectibility of the amounts owed.
The principal findings, projected to fiscal year 1980, are:
• approximately five percent of all households* expected
to participate during the year would be subject to
recoupment; they would receive two percent of the
total benefits paid . out during the year, and only
slightly more than one half of those benefits
would be recoupable ;
• the total benefits subject to potential recovery
would be between $87 and $124 mi llion; **
• over half of all tax units subject to recoupment
would have small liabilities - less than $200;
and
• approximately one-half of the total recoupment
amount could be taken out of the recipients' tax
refund checks and the remainder would have to be
collected by the IRS.
This chapter begins with a description of the methodology and data base used
in the estimation of the effects of recoupment, and a discussion of the
strengths and weaknesses of the model and the estimate it produces.
*All figures in this study exclude Puerto Rico, Guam and the Virgin Islands.
Although together they account for about one-tenth of program participants,
they are outside the U. S. income tax system, and thus could not be recouped
from under any plan dependent on the IRS as the vehicle of collection.
**Gross savings exclusive of administrative costs, if all money subject to
recoupment is collected and there are no waivers or deferrals.
33
Against that background, estimates of the aggregate impact of the Jeffords
plan will be presented. These estimates include the number of households and
tax units that would be affected by recoupment, and the total amount of
recoupment liabilities that might be offset by tax refunds otherwise due.
Next, there will be a discussion of the impact of changing some of the basic
design parameters--particularly the recoupment rate and the allocation of
benefits among household members. Finally, the implications of assumptions
employed in the estimation will be analyzed.
3.1 ESTIMATION PROCEDURE
3.1.1 Data Bases Employed
In order to obtain the best possible estimates of potential recoupment
amounts and the number of households affected, a microsimulation model was
used to measure (simulate) the effects of recoupment had the proposal been
implemented in 1975. Since program records were insufficient to serve as
a data base, the recoupment simulations were made using the Survey of Income
and Education (SIE), while use of a merged data base, constructed by
statistically matching the SIE and the Statistics of Income (SOl) supplied
relevant tax information.
The SIE, conducted by the Bureau of Census for HEW, is a comprehensive study
of over 150,000householdsweighted to represent the national population.
In addition to supplying a wealth of demographic and economic information,
the survey includes data, supplied by the households surveyed, on participation
in the Food Stamp Program in 1975. These . data provide an excellent starting
34
•
point for estimating caseload, total benefits, and the impact of the
proposed recoupment policy.
In all data bases like the SIE, internal consistency checks are made on the
data supplied. For example, the SIE was examined for misreporting of income.
After checking with national income account totals, it was found that there was
little under-reporting of earnings. However, there were adjustments made for
under-reporting of transfer income, using data from programs such as AFDC,
SSI, and OASDI.*
The first step in the simulation was to examine the SIE to isolate a sub-sample
containing food stamp recipients. Total households who reported
having received benefits in 1975 constituted the food stamp caseload and
bonuses, and the shortfall--due to households failing to report food stamp
participation--was filled by households whose SIE-reported income qualified
them for the program. This sample was also modified to yield another sample
of program participants under the New Law (P.L. 95-113) eligibility changes
and for projections of the impact of the recoupment in 1980 with increased
program participation due to the elimination of the purchase requirement.
The SIE data base, however, does not contain some of the information needed
to answer all recoupment questions. For example, it does not have information
on tax liabilities or refunds. In addition, those surveyed are grouped by
household as in the Food Stamp Program itself, rather than by tax unit as
required by a recoupment plan. In order to supplement the SIE data, the
*Reporting of Transfer Income on the Survey of Income and Education: Initial
Correction of the Microdata for Underreporting, Harold Beebout, Mathematica
Policy Research - October 1977.
35
households were converted to tax units and then merging this file with the
SOl (Statistics of Income), compiled by the Treasury, which is an extremely
rich source of tax information for individuals. SOl data include the
presence and size of tax refunds, or additional tax liabilities, and income
as reported to the IRS--both importantvariablesin determining possible
returns from a recoupment policy. No adjustment for uriderreporting of income
was made on the SOl because this file contains information identical to that
used \in calculating tax liabilities as well as recoupment liabilities.
The use of merged data bases is a valuable tool for improving available
data for program analysis. For example, the Treasury Department has used
a merged file since 1973 to estimate the impact of changes in tax specifications.
In addition to developing the 1975 SIE/SOI merged file used for recoupment
estimates, the Treasury Department is currently constructing a 1977 merged
file using the March 1977 Current Population Survey (CPS) and the Calendar
Year (CY) 1977 SOl. The methodology employed by the Treasury Department to
create the merged SIE/SOI and the CPS/SOl is the same.
Because the SIE is so large, a representative subsample of 50,000 households
was chosen from the survey before matching with the SOl. Then, SIE households
were converted into the IRS filing units using a tested computer routine
which was originally developed for the Treasury Department in 1973 to merge
the SOl and CPS (Current Population Survey) data bases. This procedure
yielded approximately 80,000 tax units.
The merged file was constructed by computer by running each converted SIE
tax unit against SOl tax units in the same census area to find the SOl tax
36
unit which best matched the SIE unit. Matching was done on a set of
characteristics including adjusted gross income; age, race and sex of
tax unit head; tax unit size and schedule filed; and total wage, and salary
by source. An SOl tax unit record was used as many times as it was the
best potential match. Once a match was made, the two records were linked.
3.1.2 Simulation Technique
Microsimulation is currently used to assess the impact on costs and case­loads
of changes in transfer programs including Aid to Families with
Dependent Children, SSI, and food stamps, as well as to measure the impact
of changes in Federal income tax law. Simulation allows the application of
revised program rules against actual program information, in order that the
effect of these new rules can be measured.
This simulation of the Food Stamp Program replicates the eligibility condi­tions
for both the old (1964) and the new (1977) Food Stamp Program laws.
In order to simulate annual food stamp bonus and annual income for recoupment
purposes, it was vital that the measure of food stamp benefits be an
accurate reflection of what occurred during the preceding year. The annual
benefits had to be based on the sum of monthly benefits (and the associated
monthly income), not annual income observed at tax time. This allowed for
intrayear variation in income necessary for the potential recoupment of
benefits. Variation in household income from m