California Department of Social Services, DAB No. 816 (1986)

DEPARTMENTAL GRANT APPEALS BOARD

Department of Health and Human Services

SUBJECT: California Department
of Social Services

Docket No. 85-159
Audit Control No. 50260-09
Decision No. 816

DATE: December 5, 1986

PARTIAL DECISION

The California Department of Social Services (California, State)
appealed a disallowance by the Office of Refugee Resettlement (ORR) of
$30,157,810, for payments to refugees residing in California, for the
period April 1, 1981 through September 30, 1982. ORR later reduced the
disallowance to $28,163,299. Transcript (Tr.) 7. The costs were
disallowed primarily because of allegedly inadequate documentation and
eligibility determinations.

Below, we uphold ORR's determination on the following elements at issue
in this case: the use and choice of statistical methodology, improper
determinations of eligibility for refugee aid, and the general right of
ORR to seek recoupment for improper expenditures without use of a
"tolerance" level such as that found in the Aid to Families with
Dependent Children (AFDC) quality control program. However, we have
reserved ruling on the validity of ORR's determinations on other
disputed individual cases in the sample underlying the disallowance,
because we believe the record needs further development (primarily on
the issue of the relationship of ORR's actions to disallowance policies
in the AFDC program). We are sending questions separately to aid the
prompt development of this part of the case.

Background

The Refugee Act of 1980, Public Law 96-212, established ORR and
authorized assistance and services to refugees residing in the United
States.

Under the refugee resettlement program, cash assistance, medical
services, and social services were generally provided to refugees under
existing federally funded programs such as AFDC and Medicaid. For
refugees who were eligible for these programs, ORR reimbursed the State
for the non-federal share of the AFDC and Medicaid program costs. If
the refugees did not meet the eligibility . -
2 -

requirements under the existing federal programs, they were provided
assistance under ORR's Refugee Cash Assistance (RCA) and Refugee Medical
Assistance (RMA) programs. The cost of the RCA and RMA programs was
funded entirely by ORR. Under ORR regulations, the RCA and RMA 100
percent federal funding could also be used for state and local general
assistance payments to refugees.

The California Department of Social Services was responsible for
administering refugee aid in California (the California Department of
Health Services administered medical services to refugees under a
contract with the Department of Social Services). 1/ During the period
April 1, 1981 through September 30, 1982, the State claimed a total of
$442,322,390 under the refugee program. Of this amount, $256,441,911
was claimed for cash assistance payments. 2/

The disallowance in this case is based on an audit by the Office of
Audit, Office of the Inspector General. The auditors discovered a
number of different kinds of errors with financial consequences; the two
most prominent errors were inadequate initial eligibility determinations
and inadequate updating of continuing eligibility status. Since the
individual case determinations were voluminous, the auditors used
statistical sampling techniques in lieu of examining all records to
establish the amount of the disallowance, an approach upheld in
principle by courts and this Board before.

The auditors basically used samples of two sets, or universes, of
payments to arrive at the dollar value of the cash assistance payment
errors. One universe consisted of about 893,000 payments made in
Alameda, Orange, Sacramento, San Diego, San Francisco and Santa Clara
counties. The second sample was selected from 315,491 payments made in
Los Angeles County (not included were certain AFDC payments for which
records were unavailable). The counties in these two universes
represented 85 percent of the State's cash assistance payments during
the period.

1/ Eligibility was determined by the local county welfare departments,
which were also responsible for issuing the cash assistance payments.

2/ Of the remainder, $101,879,819 was for medical assistance,
$26,701,099 for social services, $11,942,879 for Supplemental Security
Income, and $45,356,682 for administration costs.

- 3 -

Summary of Decision

The issues in this case fall into two categories: issues concerning the
validity of ORR's sampling process, and issues concerning whether ORR
correctly determined that California erred in using refugee funds for
the cases in the samples. Our overall determinations are as follows:

o We uphold ORR's statistical approach. Below, we examine that
methodology and find it sound and reasonable. California attacked the
methodology primarily on two bases. First, the State alleged that ORR's
use of a "probe" sample as part of the ultimate sample biased the
sampling process. We find that the record does not support California's
argument and, indeed, that the use of the "probe" sample was quite
reasonable in the facts of this case. Second, California proposed an
alternative methodology which it alleged would produce a lower
disallowance amount. We find that on balance, ORR's methodology was the
sounder approach, and that the lower result reached by California could
be attributable to coincidence or factors which, in any event, do not
impeach the reliability of ORR's methodology.

o We uphold ORR's position that the State has a burden of documenting
eligibility of individuals receiving refugee payments. California
conceded that there was a lack of documentation in certain cases, and
that documentation for other cases was insufficient. We disagree with
the State that the disallowance should be reversed based on lack of
publication of an ORR policy making AFDC provisions (including one on
monthly reporting by recipients) applicable to the refugee program.

o Concerning ORR's determinations on individual cases for which the
State alleged its documentation was sufficient, we have determined that
further record development is needed. Our basic concern relates to
whether the disallowance would have been authorized in similar
circumstances under the AFDC program, since ORR has, in effect, linked
its determinations to policy requirements adopted from that program. We
have separately addressed questions to ORR to develop the record on this
issue.

o We uphold ORR's decision that the State was not entitled to retain
federal refugee funding for payments made to persons incorrectly found
ineligible for AFDC cash assistance on the basis of the initial

- 4 -

Discussion

A. The disallowance was properly estimated from a random sample using
statistical techniques shown to be valid.

The State argued that the disallowance should be reversed, at least in
part, because the Agency (1) used a probe sample improperly in
combination with the full sample, and (2) used the allegedly less
precise method of extrapolating from unallowable rather than allowable
payments, which arbitrarily increased the disallowance. The Agency
countered that (1) the use of a small initial sample was to determine
the size of the full sample, not to decide whether to continue sampling,
and was validly combined with the larger sample; and (2) that the method
of extrapolating from unallowable payments was generally more precise,
was the method reasonably and customarily used by the auditors, and had
not been shown to be improper here.

1. The State did not show that the use of consecutive samples here
impeached the estimate on which the disallowance was based.

It is well-established in Board and court precedent that sound
statistical sampling methodology can be used to determine the amount of
costs properly charged to HHS programs. See Ohio Department of Public
Welfare, Decision No. 226, October 30, 1981; and California Department
of Social Services, Decision No. 524, March 29, 1984, and precedents
cited therein. Sampling typically is used when a claim for federal
funds is based on the sum of numerous cost items (each subject to proof
of allowability) because it is impossible, or at least costly and
impractical, to examine each item. In such cases, the reviewing
authority will take . - 5 -

a random sample of cost items, examine them, and extrapolate the
findings to the whole group. If done in accordance with the general
rules and conventions statisticians have developed, the extrapolated
finding has a high degree of probability of being close to the finding
which would have been produced if all the cost items had been
considered. The sampling result may even be more accurate than a
full-scale examination, since clerical and other errors can reduce the
accuracy of a 100 percent review. In simple terms, valid statistical
sampling will produce a result which has a high degree of probability of
being at least as good as a full-scale review. The basic issue before
this Board in cases such as the one before us concerns the validity of
the auditors' sampling process.

In estimating the amount of the disallowance from the sample in this
case, the auditors established a confidence interval containing a set of
values which they were 95 percent certain included the true dollar value
of erroneous payments. 3/ They used the lower limit, or bound, of this
confidence interval for the disallowance, even though statistically the
middle figure, or point estimate, represents the most likely true dollar
value of erroneous payments. Thus, the basis for the disallowance was
conservative in favor of the State; in fact, the probability was 95%
that the real (albeit unknown) error amount was higher than the figure
chosen by ORR.

The plan for the sampling exercise called for a small sample to be taken
initially in an effort to determine the size of the sample that should
be used. Appellant's Hearing Exhibit 2. The initial sample indicated
an optimal sample size of 3000. Id. Based on audit resources and time
limitations, a sample size of 639 was selected. 4/ It was undisputed
that

3/ Actually, the confidence level per se was 90 percent, using the
full confidence interval. However, because the auditors used the lower
bound, there was a 95 percent probability that the real error amount was
higher. The State's expert referred to this as a two-sided 90 percent
confidence interval, or lower (or upper) 95 percent confidence level
(Tr. 87), and ORR's expert agreed (Tr. 204). For convenience, we refer
to it in this Decision as a 95 percent confidence level.

4/ Appellant's Hearing Exhibit 2; Tr. 303-304; Respondent's submission
of March 12, 1986. Later, the sample was expanded again, by 167 more
payments. See discussion, infra.

- 6 -

the smaller sample size was still adequate statistically. 5/ Decreasing
the sample size most likely favored the State in this case, due to the
auditors' use of the lower bound. Since use of the smaller sample
increased variability somewhat, the lower bound of the confidence
interval (and therefore the disallowance amount) was probably a lesser
dollar amount than if the sample size had been 3000.

In the six-county sample and in Los Angeles, the auditors took a small
initial sample and combined it with a larger second sample for the final
sample size used to arrive at the amount of the disallowance. The State
objected that this procedure was improper. Essentially, the State's
position appeared to be that a two-step sample required more complete
and sophisticated preliminary explication to avoid the possibility that
samplers could manipulate the process unfairly.

The State's objection is not sound. At best, it might involve an issue
of fairness if ORR had evaluated the first sample and, finding it
unsatisfactory, decided to enlarge the sample until results more to
ORR's liking (such as, presumably, a larger disallowance) were obtained.
As discussed below, ORR did not evaluate the first sample in this
manner, but used it only to estimate the sample size needed for a
certain precision--as they had planned to do. But whatever the case, a
decision to use a larger sample generally would be justified per se
because the larger the sample, the less variability there is and thus
the results have greater precision. That a larger sample coincidentally
results in a larger disallowance does not necessarily make it less
justifiable. Given the resources, ORR could have chosen to audit a
larger sample yet, or even every case.

ORR did not dispute that the larger final sample probably did result in
a larger disallowance than if the smaller initial sample had been the
only basis, but it pointed out that its analysis of the initial sample
did not establish a lower bound, so it did not know what the initial
results would have been. ORR's statistical expert testified that the
lower bound was not manipulated in any sense because there was "no
decisionmaking in between." Tr. 260-262.

ORR explained that the decision to conduct the larger sample was already
made before the smaller sample was taken and that the only purpose of
the smaller sample was to help the auditors decide how large the sample
should be. The auditors knew in advance that the initial sample would
not

5/ Office of Audit policy requires a sample size of at least 200. Tr.
301. - 7 -

give them a usable figure for the dollar value of the errors. ORR also
noted that much of the testimony of both the State's and ORR's experts,
relied on by the State, was about the reasons why one should not use a
probe sample to test for attributes (i.e., why and at what rate the
payments were erroneous) and did not apply to estimating variables
(i.e., the dollar value of the erroneous payments, the case here).

We agree that use of an expanded sample, as described above, was
perfectly appropriate and sound statistical practice. The State has
produced no evidence that the sample was analyzed for any purpose other
than the goal of ascertaining an adequate sample size or that the choice
of sample size was manipulated to produce a result favoring ORR. ORR's
explanation of what it did was reasonable, and we find that it was
statistically justifiable.

While we find nothing wrong with the use of an expanded probe sample
generally, there was a particular circumstance involving Los Angeles
which requires further analysis. The State alleged that in this
particular instance, the auditors expanded the sample because the sample
showed a negative value (minus) for the lower bound. The State argued
in effect that even if a two-sample approach were valid in determining
sample size, it is not valid where the decision to expand the sample is
based on the fact that the initial sample size indicated a lack of
payment errors (by yielding a negative-value lower bound).

The following facts are undisputed. Because the number of assistance
payments was known only for Los Angeles, the auditors had to divide
their review into separate samples for Los Angeles County and for the
six other counties. The other six counties maintained their records by
recipient. Los Angeles County was further separated into three strata:
AFDC; RCA and general assistance (GA); and retroactive payments to
non-Indochinese refugees (LA RETRO). After the small initial sample, a
sample size of 395 was selected from Los Angeles County, which included
57 for the LA RETRO stratum. Later, the LA RETRO stratum was expanded
to 224. The sizes of the samples for the other two strata were not
increased after the initial draw.

Thus, of the three initial sample strata in Los Angeles, only LA RETRO
was subjected to the extra sampling discussed here. California's
position essentially was that this extra sampling implied that ORR was
inappropriately reaching beyond initial evidence (in the form of the
negative value) that no money was owed, until ORR produced evidence that
some money was owed.

- 8 -

ORR's response essentially was that its auditors did no more than
increase the sample size to avoid producing an absurd result. ORR
denied that the LA RETRO stratum was actually evaluated any differently.
Tr. 306-311. ORR argued that the obvious statistical remedy for a
negative lower bound (when the point estimate was a positive value) was
to increase the size of the sample, not to eliminate the stratum. ORR
submission of March 12, 1986.

We have determined that the decision to expand the LA RETRO stratum was
appropriate. ORR was sampling to establish the dollar value of the
payment errors, not to discover the error rate. In effect, there was an
assumption that there were errors (an assumption not contested in
principle, but only in amount, by California). If there were no errors,
the dollar value of the lower bound would never exceed zero, and
California could never be prejudiced by an expansion of the sample.
Since there was, however, a reasonable and . uncontested assumption that
there were some errors, ORR was justified in expanding the sample size
to overcome the negative value. A true negative value was clearly
impossible in this case: auditors gave a value of zero to each proper
payment and a positive dollar value to each erroneous payment, so that
the negative statistical value apparently was produced only because the
original sample was so small that the variability of the distribution
values got pushed out, at the edges, into a statistically theoretical,
but practically impossible, area of negative values. Clearly, the
auditors were justified in correcting this absurd result by the simple
expedient of expanding the sample size. The overall context must be
kept in mind as well: expanding the sample size, assuming measurement
error is minimized, can only decrease variability and increase accuracy.

The authorities relied on by the State to support the alleged defects in
so-called two-step sampling including ORR's expert were clear only as to
the objections to using such a procedure in attributes (i.e., error
rate) sampling. The State's expert was not convincing when he tried to
apply the same objections to variables (i.e., dollar value) sampling,
the sampling at issue here, for the reasons discussed above.

ORR also cited, in support of its sampling and estimation methods in
this case, a report by the General Accounting Office (GAO). ORR
prehearing brief, Exhibit 5. The report notes that one of California's
criticisms was that the auditors had increased the size of the LA RETRO
sample after the auditors had already examined a sample of the payments.

- 9 -

GAO concluded that the expansion of the sample "was not inappropriate as
the sample's randomness was not affected." Report, p. 5. 6/ GAO's review
appears to be comprehensive and objective and reinforces our conviction
that ORR followed a proper sampling methodology when it expanded the
sample in this case.

2. The State did not prove that the disallowance should not be based
on an extrapolation from unallowable payments.

In calculating the disallowance from the results of the sample, the
auditors based their estimate on the dollar value at the lower bound of
the confidence interval of unallowable payments. The State argued that
ORR should not base its disallowance on the estimate thus obtained
(known

6/ During oral argument, counsel for the State alleged that GAO had
not had the audit papers and did not engage in extensive discussions
with ORR as the State's expert had done, and characterized the basis for
GAO's approval as "very superficial." Transcript of oral argument, July
22, 1986, p. 38. GAO described the extent of its review:

In conducting our review we interviewed ORR program officials
and reviewed ORR records to obtain (1) an understanding of the
Refugee Cash Assistance Program and (2) data on refugee funds
claimed by and awarded to California since the enactment of
the Refugee Act of 1980. We also interviewed officials of the
IG's office, in Washington and Sacramento, and reviewed the
auditors' workpapers to obtain an understanding of the
sampling methodology used and the basis for the conclusions
reached. To clarify possible legal issues, we met with
attorneys in HHS' Office of General Counsel. Additionally, in
order to fully understand the State's position, we interviewed
officials of California's Department of Social Services (DSS),
including the department's methodology consultant, and members
of a public interest group representing the involved counties.
Our review was made in accordance with generally accepted
government audit standards. [Emphasis added.]

Report, p. 3. Clearly, the GAO review was thorough. The State did
not offer any proof to support its contrary conclusions, and we
reject them.

- 10 -

as the difference estimate) because other methods suggested by the State
were better. The State did not contend that the method the auditors
used was incorrect, however.

The State suggested that better methods, such as the direct projection
method (resulting in a lower disallowance), involved extrapolating from
the allowable payments. 7/ In the case of Los Angeles, the State
contended that the direct projection method was more precise because the
sampling error was smaller than the sampling error in the method the
auditors used. 8/

The State also noted that an extrapolation of the total funds claimed by
the State for the six counties, derived from the auditors' difference
estimate, would be approximately $149.5 million, $9.5 million more than
the actual "book value" (i.e., the actual total claim). The total funds
claimed estimated by the State's direct projection. method was said to
be approximately the same as the actual book value. 9/ Appellant's
Hearing Exhibit 1. The State argued that ORR could not validly base a
disallowance on a method which estimated the total value of the State's
claim (in the six counties) $9.5 million higher than it actually

7/ The lower bound of the confidence interval in the direct projection
estimate was $21,416,652. This compares with the lower bound of
$30,157,810 in the difference estimate. Appellant's Hearing Exhibit 1.

8/ We note that in the six-county sample the result was just the
opposite. The sampling error derived from the direct projection was not
only considerably larger than the sampling error obtained by the
auditors' methodology, but also contained a negative value ($8,675,666)
at its lower bound. The State used instead the value of the errors
actually found in the sample ($2,791). Appellant's Hearing Exhibit 1.

9/ The extrapolated amount of total funds claimed in Los Angeles
County, using the difference estimate of the auditors, was approximately
$3 million more than the actual book value. State's submission of
February 7, 1986, first attached summary table. However, the State's
direct projection estimate of total funds claimed in two of the three
strata in Los Angeles County was approximately $24 million less than the
actual book value. Appellant's Hearing Exhibit 1. The State's expert
indicated that there was a "50/50" chance that a direct projection
estimate of total funds would be higher or lower than the book value.
Tr. 72.

- 11 -

was, particularly when there was at least one alternative method which
estimated the State's claim at approximately the same as the book value.
10/

ORR argued that the auditors' difference estimation method was usually
more precise than the direct projection method, a point on which both
the State's and ORR's experts agreed (Tr. 65, 265), and that its
auditors as a rule used difference estimation (for variables sampling,
as here). ORR also contended that the State's direct projection method
was subject to measurement errors in using the larger figures
constituting the allowable payments and California did not dispute this.
ORR also said that the State's method was subject to "disallowance
errors" in determining eligibility, payment amount, and funding source.
11/

We uphold the disallowance as based on the auditors' difference
estimation methodology. The State did not (and indeed made virtually no
attempt to) impeach the validity of the auditors' methodology per se: it
was agreed that the auditors' methodology generally was more precise and
less subject to measurement errors; and, in the facts of this case,
anomalies in extrapolated figures which California sought to exploit as
collateral evidence are accompanied by countervailing anomalies (and, in
any event, appear more attributable to coincidence or happenstance than
any factor which makes them more persuasive than the auditors'
findings). The State's own expert testified that the auditors' method
and its results in this case were statistically valid and emphasized
that his findings did not impeach the auditors' methodology. Tr. 66,
74, 76, 106. Both parties tried, unsuccessfully, to explain why the
difference estimates for the total amount claimed for the six counties
and for Los Angeles County were so far apart from the direct projection
of those amounts. We conclude, as suggested by

10/ The State also briefly suggested the use of the "ratio estimate"
method. This method yielded an estimated value of funds claimed for the
six counties some $10 million less than the actual claim, and for the
combined seven counties, some $35 million less. Appellant's Hearing
Exhibit 1. This method, which also relies on the allowable, rather than
unallowable, payments, also produced a disallowance less than that of
the difference estimate method (the lower bound was $28,576,308). Id.
We discuss this below at fn. 14.

11/ ORR did not develop the "disallowance errors" point, apparently
because it concluded that "disallowance errors can be accounted for."
ORR Post-Hearing Brief, p. 20. - 12 -

ORR's expert (Tr. 278), that there is no reason to expect that the two
calculations should come out the same, noting the contrasting methods.
12/ Recalling from fn. 9 that the total funds claimed for Los Angeles
County calculated by the State's direct projection method was
approximately $24 million off the alleged book values, it is surely
happenstance that the direct projection in the LA RETRO sample and in
the six county sample produced estimates for total funds claimed so
close to the alleged book values. For that matter, we do not know that
the figure touted as the book record value was accurate, as it, too, was
subject to measurement error. We do not think these erratic
fluctuations resulting from the State's methodology provide a sufficient
basis for overturning an audit result based on an otherwise unimpeached
and statistically valid procedure. 13/ We note that GAO also reviewed
this criticism and concluded, after examining the data, that the
auditors had used an appropriate method.

12/ To obtain a difference estimate, one multiplies the sample mean, or
average dollar value of payment errors, by the number of cases. In
direct projection, one divides the dollar value of the amounts claimed
by the number of cases. Tr. 71, 205, 206. The ratio estimate method
was not explained in detail in this record, but we think the parties
would agree it is basically a method in which a ratio formed between the
allowable payments in the sample and the total payments in the sample is
applied to the total population. See, Audit Sampling, a publication of
the American Institute of Certified Public Accountants, p. 89.

13/ We think this erratic display makes this case distinguishable from
the situation in New York State Department of Social Services, Decision
No. 522, March 12, 1984, where the Board said, in dicta (p. 101):

We believe SSA is quite wrong in arguing we should give
deference to an unexplained choice of alternative approaches
which could result in a difference of millions of dollars . .
. It would be indefensibly arbitrary to shield a huge
differential impact behind the merely incidental fact that the
methodologies used are both statistically sound.

We stand by this dicta, but note that it does not justify a state
merely showing that it can use statistics to produce a lower
disallowance. The (continued on the next page)

- 13 -

For all the foregoing reasons, we find that California failed to provide
any substantial evidence that the use of ORR's statistical methodology
as a basis for the disallowance was improper. We now proceed to a
consideration of issues concerning the cases in the samples. 14/

13/ (continued from the previous page) ultimate issue is the relative
validity, precision, and reliability of the federal methodology and
its implementation. Here, ORR has succeeded in persuading us that
its methodology was the more precise one, while the State has done
no more than use a less precise methodology to produce an
extrapolated figure from one part of the sample to produce a figure
which, for unknown reasons and apparently by happenstance, more
accurately approximates the total claim figure than. ORR's
extrapolation. In fact, the total claim figure itself could be
inaccurate because the State over- or under-claimed. In short, the
State's position relies only on a statistical curiosity, while
ORR's measurement of errors is statistically valid.

14/ Although our discussion focuses on the direct projection method as
a suggested alternative to the difference estimation used by the
auditors, we also reject the ratio estimate method for similar reasons.
Furthermore, ORR's contention that this method was biased (i.e.
inaccurate) in this case was not disputed by the State. Thus, in
addition to our other reasons, we do not think ORR should be required to
base a disallowance (or any part of it) on a method which is
statistically less acceptable than the one it used.

We note also that in our decision upholding the use of the
auditors' difference estimate methodology, we have not found it
necessary to rely on the declaration of an HHS auditor attached to
ORR's posthearing brief. The declaration attempted yet another
speculative explanation of the $9.5 million difference between the
total amount claimed estimated by the auditors' methodology and the
alleged book record value. The State objected to the declaration
because it was submitted after the hearing and without giving the
State an opportunity to cross-examine.

- 14 -

B. ORR validly disallowed payments for which the State conceded that
there was not sufficient documentation to establish eligibility.

The auditors found that for some of the sample cases reviewed, the file
did not contain a monthly report form (CA-7). The auditors determined
that the CA-7 was required for refugee cases based primarily on an
action transmittal issued by ORR. That transmittal said:
"Recertification and any other State eligibility reports shall be
required for refugee assistance recipients at least as frequently as for
AFDC recipients." SRS-AT-76-160, October 22, 1976, State's prehearing
brief, Exhibit 3. Although AFDC regulations in effect at the beginning
of the audit period make monthly reporting optional for no-income cases
(45 CFR 233.28 (1980)), the auditors found that the State required
monthly reporting for all recipients of AFDC and refugee assistance.
Specifically, the State's Manual of Policies and Regulations for the
Refugee Program, section 69.205.3, provides: "A periodic
redetermination of eligibility shall be made in accordance with (EAS)
Section 40-181 as it applies to the AFDC program." The Eligibility and
Assistance Standards (EAS) Manual requires counties to use the CA-7 for
the AFDC program.

The State argued that monthly reporting was not a valid federal
requirement for the refugee program because the action transmittal
provision was not published in the Federal Register. The State relied
on section 552(a)(1)(D) of the Freedom of Information Act (FOIA) and
section 553 of the Administrative Procedure Act (APA) for this
proposition and also cited to court decisions requiring publication of
rules which would have a "substantial impact." See. e.g., United States
Dept. of Labor v. Kast Metals Corp., 744 F. 2d 1145 (5th Cir. 1984); and
Herron v. Heckler, 576 F. Supp. 21 (N. D. Cal. 1983).

We disagree that the lack of publication of the action transmittal
provision provides a basis for reversing this disallowance. The State
did not deny that it had actual notice of the action transmittal; the
transmittal was addressed to state program administrators and such
transmittals are routinely sent to all states. Indeed, the State
implemented the provision by adopting monthly reporting for refugee
assistance, as it had for AFDC. Under the cited FOIA provision, actual
notice is sufficient. While the cited APA rulemaking requirement
(adopted for HHS grant programs in 1971) arguably might apply to the
transmittal provision, we do not think it provides a basis for reversal

- 15 -

under the circumstances here. 15/ Refugee program regulations require a
state plan to give assurances that the state will follow official
issuances of the Director, ORR. The State agreed in its State plan to
do this. State's prehearing brief, Exhibit 7, p. 49. In fact, with
respect to the RCA, the State explicitly agreed to be bound by action
transmittals and to use AFDC program policies and standards (with one
exception not relevant here).

We also note that the State's "substantial impact" argument was based on
the impact monthly reporting could have on program recipients. The
issue here is whether federal payments to the State can be disallowed,
not whether the State can recoup from the recipients.

We do not here resolve the issue of whether, based on the action
transmittal, a payment can be disallowed solely because Los Angeles
County did not require a monthly report or its equivalent for each of
these disputed cases. We are seeking more information from the parties
on this issue. But we do agree with ORR that (whether or not monthly
reporting is a substantive condition for federal funding) the State
generally bears a burden of documenting that the payments for which it
seeks federal funding were made to individuals who were eligible under
the federal program. See, e.g., California State Department of Health,
Decision No. 55, May 14, 1979.

Here, the State conceded that for approximately 20 cases, Los Angeles
County could produce no documentation showing that the individual was
eligible for the payment. Tr. 135. The State produced documentation
which it alleged was sufficient to show eligibility for the remaining
cases. The State later conceded that the documentation for some of
these cases was insufficient and ORR accepted the documentation for a
number of other cases. With respect to the 75

15/ ORR argued that the action transmittal provision was an
interpretative rule and therefore exempted from the APA requirement.
This argument, however, was based on the idea that the provision
implemented a statement in the legislative history of the refugee
program statute (rather than the statute itself) or that the provision
interpreted a refugee program regulation (45 CFR 400.62) which was not
in effect at the time. ORR also pointed to the monthly reporting
provisions applicable to AFDC; clearly the action transmittal was not
simply an interpretation of these provisions since it expanded the scope
of their applicability. Thus, we do not base our conclusions here on
ORR's analysis.

- 16 -

cases still in dispute, we have determined that we need more information
(see page 1 above). With respect to the cases for which the State
submitted no documentation or later conceded had insufficient
documentation, we uphold the disallowance.

C. ORR properly disallowed 100 percent reimbursement for refugees who
were found to be eligible for AFDC benefits subsequent to the initial
classification of the refugee.

ORR disallowed $10,527,334 for refugees who were initially determined to
be eligible under the Refugee Resettlement Program (RRP), but were later
found to have been eligible for AFDC benefits. 16/ The State was
required to first determine eligibility for AFDC and only if a refugee
was not eligible for AFDC could the State consider the refugee's
eligibility for cash assistance under the RRP. Action Transmittal
77-11, dated December 2, 1977; State's prehearing brief, Exhibit 5.

Cash assistance payments to refugees under these two programs could be
fully reimbursed by the federal government. If a refugee was determined
to be AFDC-eligible, the State was fully reimbursed, 50 percent from the
federal share of the AFDC program and 50 percent from the refugee
program. 17/ If a refugee was determined not to be eligible for AFDC,
but eligible for refugee aid, the State was reimbursed 100 percent from
refugee funds. Tr. 177. ORR disallowed 50% of payments for refugees
who initially had been found ineligible for AFDC (thus producing 100%
refugee funding for the State from the refugee program) where the same
refugees subsequently were found eligible for AFDC by the State (thus,
from ORR's perspective, removing eligibility for half the funding that
was provided).

The principal type of error in the initial determination of eligibility
was that the applicant was incorrectly found not to have had enough
"quarters of work" to be eligible for

16/ This was based on a 1982 reevaluation conducted by the State. See.
State's prehearing brief, Exhibit 1, p. 17. ORR also disallowed an
additional $780,682 based on a sample of refugees who were found in the
redetermination to have been eligible for another relief program for
Cuban-Haitian entrants. See, 22 U.S.C. 2601 and 45 CFR Part 401. The
State did not pursue an appeal from that part of the disallowance under
this issue, although the sample is included in the sampling issue,
discussed pp. 4-12, supra.

17/ The State and the federal governments each pay 50 percent of AFDC
program cash assistance. - 17 -

AFDC. 18/ To be eligible for AFDC, an applicant must have established a
connection with the work force shown by six or more quarters of work
within any 13-quarter period ending within one year prior to the
application for assistance. A quarter of work means a period in which
the applicant received earned income of not less than $50. 45 CFR
233.100(a)(3). Of a 41-payment sample on which this part of the
disallowance was based, 34 involved situations in which work histories,
not obtained in the initial interview, but obtained afterwards when the
applicants were reinterviewed, revealed that they met the AFDC criteria
for the requisite work force connection. 19/

The State contended that the failure of its interviewers to obtain
correct work histories at the time of the initial interview was due to
cultural and language differences between the interviewers and the
refugees. The State alleged that by the time these refugees were
reinterviewed, they and their interviewers experienced less of these
differences and consequently more accurate work histories were obtained.
The State cited as examples cases in which interviewers had to determine
the value of "room and board" supplied to a "prisoner-physician in a
Communist slave labor camp" and of rice grown by refugees on farms in
Vietnam and Cambodia and bartered for eggs. These applicants allegedly
told the interviewers that they had not worked.

The State pointed out that if the refugees had been initially determined
to be eligible for AFDC, the State would have been fully reimbursed and
argued that a simple bookkeeping adjustment (replacing 50 percent of the
full reimbursement from the RRP with the 50 percent AFDC federal

18/ Actually, this is a requirement of the AFDC-U program of aid to
dependent children of unemployed parents. The references in this
Decision to the AFDC program also include the AFDC-U program, although
only the abbreviation AFDC is used.

19/ In four other payments, the initial work history was complete but
the number of qualifying work quarters was incorrectly evaluated. In
three other payments, the refugee's eligibility for AFDC was incorrectly
discontinued. Audit Report, p. 17, State's prehearing brief, Exhibit 1.
The State did not discuss these seven cases.

- 18 -

share) should be effected in lieu of a disallowance. 20/ The State cited
California Department of Social Services, Decision No. 245, January 12,
1982, in support of its contention that the Board was empowered to
direct ORR to perform the indicated adjustment.

ORR noted that the federal auditors had found no evidence that the
State's initial interviewers had "probed beyond the applicant's basic
statement," even where applicants had extended gaps for as long as four
years in their work histories or had stated that they never worked
during a time that they had a dependent spouse and children. 21/ Audit
Report p. 19, State's prehearing brief, Exhibit 1. ORR contended that
it was the State's responsibility to ensure at the time of application
that the refugees understood the eligibility criteria.

ORR also rejected the "bookkeeping adjustment" suggested.by the State
because the RRP and AFDC programs each were funded under separate
appropriations and separate regulations for claiming reimbursement. ORR
contended that the main holding in Decision No. 245 supported ORR's view
that the Board cannot cause funds to be transferred between two legally
separate and distinct programs and that the federal agency in Decision
No. 245 had agreed in the first place that the State was entitled to 100
percent federal reimbursement, unlike the case here.

20/ The State's program manager explained that it was not possible for
the State now simply to ask reimbursement from the AFDC program, because
to be reimbursed under AFDC, the State must have registered the refugee
at the time of application as a participant in the Work Incentive (WIN)
program, a companion program to AFDC, to prepare the applicant for
employment. Tr. 178 and see, 42 U.S.C. 630. A refugee aid recipient
does not participate in WIN. The manager alleged that California was
told that it could not seek retroactively to be reimbursed AFDC funds
for refugees initially determined eligible under the refugee program.

21/ Counsel for the State argued that accepting the "never worked"
statement without probing was defensible because the interviewers could
not know "how many people scrounged out a living" in Southeast Asia and
probably did know of situations in the United States where people with
dependent spouses and children had never worked. Transcript of July 22,
1986 oral argument, p. 41.

- 19 -

At the outset, we note that, except for a few specific examples cited by
the parties, we have little basis for evaluating the performance of the
State's interviewers in the initial eligibility determinations. The
State did not dispute the federal auditors' finding that there was no
evidence that the interviewers ever probed beyond the applicant's basic
statement. The cases which the State cited as examples were not
documented or described in sufficient detail to serve as a basis for our
directing ORR to estimate the disallowance without such cases in the
sample, much less the "equitable relief" of reversing this part of the
disallowance.

We are also not persuaded that the Board should, or could, reverse the
disallowance on the basis that ORR should simply arrange with the AFDC
program to transfer the 50 percent to which the State is no longer
entitled under the RRP. The agency charged with administering the AFDC
program is not before us. 22/ Even if it were, there might be legal
obstacles in their appropriation or in the program regulations or other
statutes and regulations which would prevent them from transferring AFDC
funds to the RRP. For example, such a transfer might be considered an
illegal augmentation of the RRP appropriation. See 31 U.S.C. 628 and
Greene County Planning Board v. F.P.C., 559 F.2d 1227 (2d Cir. 1977),
cert. denied, 434 U.S. 1086. Also, there is the WIN program
requirement, discussed earlier, which apparently already has resulted in
a refusal to retroactively give the State AFDC funds in such a
situation. See fn. 20.

The California decision, No. 245, is inapposite. There, the Board
upheld the disallowance because the costs in dispute were not allocable
to one program and could not be reasonably charged to the other.
California, supra, p. 4. Noting that the respondent agency had treated
claims against both programs together, the Board directed that another
amount which the agency auditors found was due the State be subtracted
from the disallowance instead of requiring the State to seek
reimbursement. Id. at 18. The "mere bookkeeping transactions" did not
involve parties not before the Board and apparently there was no legal
or administrative obstacle to treating the claims together. Id. at 18.
We do not think that what occurred in California, as holding

22/ We note that the State did not request that the AFDC program
administrator be joined as a party. Even if it had, we might not be
empowered to so extend our jurisdiction. See 45 CFR Part 16, Appendix
A. Certainly, nothing in this decision precludes California from
seeking relief from AFDC. - 20 -

or otherwise, is support for our directing ORR to effect a
cross-appropriation transfer on the basis of the record in this case.

D. ORR was authorized to disallow reimbursement.

The State alleged that there was no statutory, regulatory, or
contractual authority permitting ORR to recoup payments to recipients by
the State, and contended that in the absence of specific authority, ORR
could not issue this disallowance. The State argued also that there was
no authority to base the disallowance on a sample, particularly without
the use of a tolerance level as is done in the AFDC program quality
control system.

Among other supporting authorities, ORR cited 45 CFR 400.8, which makes
FFP available in reimbursement for cash assistance payments only under
terms and conditions approved by ORR; and Woods v. United States, 724
F.2d 1444, 1449 (9th Cir. 1984), a Food Stamp overpayment case which
recognized the federal government's "implied power to recover funds
spent in violation of the express terms and conditions of the grant."
ORR noted that the State acknowledged that ORR had not incorporated the
AFDC quality control regulations in its own regulations. ORR cited
federal court and Board decisions which recognize the use of statistical
sampling as a means of auditing and determining the amount to disallow
for erroneous payments.

Under the statute and implementing regulations, ORR was authorized to
fund only those State expenditures that met basic program requirements
relating to the eligibility of individuals for cash assistance. In
comparable circumstances, the Supreme Court has upheld the recoupment of
program funds by the federal government when states failed to expend
those funds in accordance with basic program conditions. See Bennett v.
New Jersey, 470 U.S. 632, 636 (1985); and Bennett v. Kentucky, 470 U.S.
656, 663 (1985). The general cost principles governing HHS grantees
declare as basic policy that grant funds "may be used only for allowable
costs." 45 CFR 74.170. Conversely, courts have held that the federal
government has a common law right to recoup monies paid in error unless
Congress has clearly manifested an intent to permit the funds to be
retained. See, e.g., United States v. Wurtz, 303 U.S. 414 (1938); and
Woods v. United States, supra. Here, the State acknowledged that by
execution of the State plan, a contract was created. State's prehearing
brief, p. 6. As noted before, the plan obliges the State to comply with
the rules and regulations of ORR. Thus, in addition to the common law
right, ORR has in effect a contractual right to issue a disallowance to

- 21 -

recover monies which the State erroneously paid contrary to the rules
and regulations of ORR and the State plan. Woods v. United States,
supra.

As noted above, we have previously held that an agency, such as ORR, may
use statistical sampling as an audit technique and base a disallowance
on an extrapolation from the sample, even in the absence of explicit
authorization in the regulations. The record does not indicate that ORR
had any policy limiting or prohibiting the use of statistical sampling
and extrapolation for disallowances. We conclude that ORR could validly
base this disallowance on an extrapolation from a sample.

The State cited only Maryland v. Mathews, 415 F. Supp. 1206 (D.D.C.
1976), to support its contention that ORR's adherence to a zero
tolerance level for erroneous payments was an abuse of discretion. In
the Maryland case, the court found that tolerance levels set by the AFDC
program agency at three and five percent were arbitrary and capricious
because the levels had been set without benefit of an empirical study.
The court observed that elimination of all erroneous payments from the
AFDC program was "totally unrealistic." Id. at 1212. Actually, this
characterization originated with the Secretary and was quoted by the
court. 40 Fed. Reg. 21737 (May 19, 1975).

In the Maryland case, the court recognized the principle that "payments
which are not made properly . . . are not to be matched by federal
funds." Id. at 1212. On the tolerance level issue, the Maryland case is
easily distinguished from this one. Maryland does not require tolerance
levels where there is no statutory requirement for them. The RRP is not
the AFDC program, and ORR did not undertake to set a tolerance level for
payment errors. The State alleged that the RRP is a more complex
program to administer than AFDC and that the states had been given
little federal guidance, but the State offered very little more than its
conclusions to support its allegations. Even if these conclusions were
accurate, we are not empowered to impose a tolerance level of our own
making. See California v. Settle, 708 F.2d 1380 (9th Cir. 1983),
upholding California Department of Health Services, Decision No. 170,
April 30, 1981. We see no reason in this record to overturn a
disallowance because of the Maryland holding. See Maryland Department
of Human Resources, Decision No. 358, November 28, 1982, p. 8.

Conclusion

For the reasons stated above, we uphold ORR's statistical sampling
methodology; its disallowance based on sampled

- 22 -

cases for which the State produced no documentation of eligibility or
conceded that the documentation was inadequate; its disallowance for
funding source errors; and its general authority to take such a
disallowance without establishing a tolerance system. We have reserved
ruling on issues related to other disputed individual case
determinations pending further development of the record in response to
questions which we are separately transmitting.

________________________________ Judith A. Ballard

________________________________ Donald F. Garrett

________________________________ Norval D. (John) Settle
Presiding Board