Full text
United States Government Accountability Office
Congressional Requesters
UNEMPLOYMENT
September 2023
INSURANCE
Estimated Amount of
Fraud during
Pandemic Likely
Between $100 Billion
and $135 Billion
GAO-23-106696
September 2023
UNEMPLOYMENT INSURANCE
Estimated Amount of Fraud during Pandemic Likely
Between $100 Billion and $135 Billion
Highlights of GAO-23-106696, a report to
congressional requesters
Why GAO Did This Study What GAO Found
The UI system has faced long-standing Based on statistical sampling and imputation techniques, GAO estimates that the
challenges with program integrity, amount of fraud in unemployment insurance (UI) programs during the COVID-19
which worsened during the COVID-19 pandemic was likely between $100 billion and $135 billion. This is about 11
pandemic. In response to historic percent and 15 percent, respectively, of the total amount of UI benefits paid
pandemic job losses, Congress during the pandemic. GAO’s estimate is for the period from April 2020 (first full
created new temporary UI programs to month of payments from all UI programs) to May 2023 (end of the public health
provide relief for the unemployed. The emergency). This estimate covers all 53 states that participated in the regular
unprecedented demand for benefits and temporary UI programs. The full extent of UI fraud during the pandemic will
and need to quickly implement the new
likely never be known with certainty. In commenting on a draft of this report, the
programs increased the risk of fraud.
Department of Labor (DOL) expressed concerns about GAO’s fraud estimation
Due to this and other challenges, GAO
added the UI system to its High Risk
methodology and stated that the resulting estimate was likely overstated. GAO
List in June 2022. disagrees and explains in the report the steps taken to estimate the range of
fraud. These steps include using (1) a 95-percent confidence interval to account
This report (1) provides an estimate of for sample design and size, and (2) multiple data sources and validity checks to
fraud within UI programs during the account for uncertainty associated with identifying potential fraud in its sample.
pandemic; (2) identifies the assistance During the pandemic, DOL provided states with assistance to improve UI
DOL provided to states; and (3) systems and processes. As of July 2023, DOL reported allocating grants totaling
presents amounts that states reported
about $1.4 billion to states for initiatives including fraud prevention, detection,
in UI overpayment recoveries and
investigation, and recovery. Officials from selected states confirmed that they
waivers, among other amounts. GAO
used data from multiple sources to used this assistance for fraud detection and prevention, including verification
produce an estimated range of the software and improvements in payment timeliness.
extent of fraud during the pandemic. In Coronavirus Aid, Relief, and Economic Security (CARES) Act and American Rescue Plan Act
determining the estimate, GAO (ARPA) Unemployment Insurance (UI) Financial Assistance
reviewed DOL data, selected and
reviewed a sample of payments,
matched samples to other federal data
bases, developed an econometric
model on claims and economic
conditions, and performed numerous
other analyses. GAO also reviewed
data on state-reported overpayments,
recoveries, and waivers. GAO
interviewed officials from 14 states,
selected based on fraud risk and other
factors.
What GAO Recommends
Since 2018, GAO has made 26
recommendations to DOL to improve As of May 1, 2023, states reported identifying about $55.8 billion in fraudulent
the UI system. However, DOL has not and nonfraudulent UI overpayments and recoveries of about $6.8 billion from
yet fully implemented 16 of these; March 2020 through March 2023. During this period, states reported identifying
doing so can reduce UI’s fraud fraudulent UI overpayments totaling $5.3 billion and recoveries of $1.2 billion.
vulnerabilities. States use several tools to recover overpayments, including direct repayment
View GAO-23-106696. For more information, and offsets. States can also write off overpayments as uncollectible. Further,
contact Seto Bagdoyan at (202) 512-6722 or states may waive their legal right to collect nonfraudulent overpayments. DOL
BagdoyanS@gao.gov, or Jared Smith at (202)
512-2700 or SmithJB@gao.gov.
rules do not allow states to waive fraudulent overpayments.
United States Government Accountability Office
Contents
Letter 1
Background 7
Estimated UI Program Fraud during the Pandemic Ranges from
$100 Billion to $135 Billion 17
DOL Has Allocated $1.4 Billion in Assistance to States and Is to
Track Funds through Quarterly Reporting 19
States Have Reported Billions in UI Overpayments, Recoveries,
Write-Offs, and Waivers to DOL 27
Agency Comments and Our Evaluation 34
Appendix I Detailed Information on the Methodology GAO Used to Estimate Fraud in
Unemployment Insurance (UI) Programs during the Pandemic 37
Appendix II GAO Unemployment Insurance-Related Recommendations to the
Department of Labor 49
Appendix III Financial Assistance Allocated and Awarded to States 52
Appendix IV Fraudulent Overpayments Recovered and Written Off in Unemployment
Insurance (UI) Programs 55
Appendix V Nonfraudulent Overpayments Recovered, Written Off, and Waived in
Unemployment Insurance (UI) Programs 59
Appendix VI Comments from the Department of Labor 63
Appendix VII GAO Contact and Staff Acknowledgments 66
Page i GAO-23-106696 Unemployment Insurance
Tables
Table 1: Coronavirus Aid, Relief, and Economic Security (CARES)
Act and American Rescue Plan Act (ARPA)
Unemployment Insurance (UI) Financial Assistance for
Fraud Prevention Efforts, July 2023 20
Table 2: Six Selected State Law Provisions for Recovering
Fraudulent Unemployment Insurance (UI) Overpayments 29
Table 3: Selected State Fraudulent Overpayment Write-Off
Criteria 31
Table 4: Six Selected State Law Provisions for Recovering
Nonfraudulent Unemployment Insurance (UI)
Overpayments 32
Table 5: Six Selected State Law Provisions for Waiving Recovery
of Nonfraudulent Unemployment Insurance
Overpayments 33
Table 6: GAO’s 26 Recommendations to the Department of Labor
(DOL) to improve the Unemployment Insurance (UI)
System, Status as of August 2023 49
Table 7: Coronavirus Aid, Relief, and Economic Security Act and
American Rescue Plan Act Financial Assistance Dollar
Amounts Allocated (as of July 2023) and Awarded (as of
May 2023) by the Department of Labor to States and U.S.
Territories 52
Table 8: Fraudulent Overpayments Recovered and Written Off in
the Regular Unemployment Insurance Program, March
2020 – March 2023 (as of May 1, 2023) 55
Table 9: Fraudulent Overpayments Recovered and Written Off in
the Pandemic Unemployment Insurance Programs, March
2020 – March 2023 (as of May 1, 2023) 57
Table 10: Nonfraudulent Overpayments Recovered, Written Off,
and Waived in the Regular Unemployment Insurance
Program, March 2020 – March 2023 (as of May 1, 2023) 59
Table 11: Nonfraudulent Overpayments Recovered, Written Off,
and Waived in the Pandemic Unemployment Insurance
(UI) Programs, March 2020 – March 2023 (as of May 1,
2023) 61
Figures
Figure 1: Examples of Tools That States Use to Recover
Overpayments, and Penalties for Fraudulent
Overpayments 15
Page ii GAO-23-106696 Unemployment Insurance
Figure 2: Department of Labor (DOL) Financial Assistance
Awarded to States, May 2023 22
Figure 3: Tiger Team Recommendations That Selected States
Have Reported Implementing That Align with the Three
Pillars 25
Figure 4: Total State-Reported Established Overpayment and
Recovery Amounts for all Unemployment Insurance (UI)
Programs, March 2020 – March 2023 (as of May 1, 2023) 27
Figure 5: Total State-Reported Established Fraudulent
Overpayment Amounts for Unemployment Insurance (UI)
Programs, March 2020 – March 2023 (as of May 1, 2023) 28
Figure 6: Total State-Reported Established Nonfraudulent
Overpayment Amounts for Unemployment Insurance (UI)
Programs, March 2020 – March 2023 (as of May 1, 2023) 32
Figure 7: Selected Steps Taken to Derive the Estimated Fraud in
Unemployment Insurance (UI) Programs during the
Pandemic 40
Page iii GAO-23-106696 Unemployment Insurance
Abbreviations
ARPA American Rescue Plan Act
BAM Benefit Accuracy Measurement
BPA Blanket Purchase Agreement
CARES Act Coronavirus Aid, Relief, and Economic Security Act
DMF Death Master File
DOL Department of Labor
ETA Employment and Training Administration
EVS Enumeration Verification System
FPUC Federal Pandemic Unemployment Compensation
FRA Fiscal Responsibility Act of 2023
IT Information Technology
MEUC Mixed Earner Unemployment Compensation
NDNH National Directory of New Hires
OIG Office of Inspector General
OMB Office of Management and Budget
PEUC Pandemic Emergency Unemployment Compensation
PUA Pandemic Unemployment Assistance
SSA Social Security Administration
SSN Social Security number
SWA state workforce agency
UI unemployment insurance
UIPL Unemployment Insurance Program Letter
This is a work of the U.S. government and is not subject to copyright protection in the
United States. The published product may be reproduced and distributed in its entirety
without further permission from GAO. However, because this work may contain
copyrighted images or other material, permission from the copyright holder may be
necessary if you wish to reproduce this material separately.
Page iv GAO-23-106696 Unemployment Insurance
Letter
441 G St. N.W.
Washington, DC 20548
September 12, 2023
The Honorable Mike Crapo
Ranking Member
Committee on Finance
United States Senate
The Honorable Jason Smith
Chairman
Committee on Ways and Means
House of Representatives
The unemployment insurance (UI) system has faced long-standing
challenges with effective program integrity. 1 In response to historic
pandemic job losses, on March 27, 2020, Congress enacted the
Coronavirus Aid, Relief, and Economic Security (CARES) Act. The act
created three new federally funded temporary UI programs that expanded
UI benefit eligibility, enhanced benefits, and extended benefit duration. 2
The temporary programs supplemented existing UI programs, known as
“regular” UI, which is a federal-state partnership that provides temporary
financial assistance to eligible workers who become unemployed through
no fault of their own. 3 The federal government directly funded the
administration of, and benefits for, the new pandemic UI programs and,
1The UI system includes UI programs that were established prior to the COVID-19
pandemic and programs established in response to the COVID-19 pandemic: Pandemic
Unemployment Assistance (PUA), Federal Pandemic Unemployment Compensation
(FPUC), Pandemic Emergency Unemployment Compensation (PEUC), and Mixed Earner
Unemployment Compensation (MEUC).
2Pub. L. No. 116-136, §§ 2102, 2104, 2107, 134 Stat. 281, 313-28. The Consolidated
Appropriations Act, 2021, created the MEUC program, which is an additional temporary
supplemental UI program. Pub. L. No. 116-260, div. N, tit. II, subtit. A, chap. 1, §
261(a)(1), 134 Stat. 1182, 1961.
3We refer to the UI program—excluding both the temporary UI programs created by the
CARES Act and other legislation, as well as the Extended Benefits program—as the
regular UI program and the benefits paid under the program as regular UI benefits. For
purposes of this report, regular UI benefits are benefits paid by the state under state UI
law, Unemployment Compensation for Federal Employees, and Unemployment
Compensation for Ex-Service Members programs. The Extended Benefits program, which
existed prior to the pandemic, provides up to 13 or 20 additional weeks of benefits when a
state is experiencing specific levels of high unemployment. We refer to the four temporary
UI programs created by the CARES Act and the Consolidated Appropriations Act, 2021,
as pandemic UI programs.
Page 1 GAO-23-106696 Unemployment Insurance
as directed by statute, relied on state workforce agencies (SWA) to
process claims and issue benefits to individuals. 4 From April 1, 2020,
through May 31, 2023, expenditures across the UI system totaled
approximately $900 billion, according to Department of Labor (DOL)
data. 5
The unprecedented demand for UI benefits and the need to quickly
implement the new programs during the pandemic increased the risk of
fraud. 6 Findings from the DOL Office of Inspector General (OIG) and prior
GAO reports, and the urgent need to address persistent issues in the UI
system—including service delivery inefficiencies and outdated IT
systems—led us to designate the UI system as high risk in June 2022. 7
This designation is intended to help spur progress in resolving persistent
issues by shining a spotlight on such issues and ways the federal
government can lead efforts to find solutions.
The increased significance of the UI system during the pandemic drew
attention to its vulnerabilities and susceptibility to fraud, waste, abuse,
and mismanagement. In our prior work, we reviewed existing measures
and estimates of fraud and found evidence of substantial levels of fraud
and potential fraud in UI programs during the pandemic. However, we
concluded that available measures and estimates were incomplete and
4Fifty-three SWAs administer UI programs across the 50 states, the District of Columbia,
Puerto Rico, and the U.S. Virgin Islands. In addition to the 53 SWAs, DOL made CARES
Act funding available to other territories and freely associated states that do not operate
regular UI programs—American Samoa, the Commonwealth of the Northern Mariana
Islands, the Federated States of Micronesia, Guam, Republic of the Marshall Islands, and
Republic of Palau—to operate PUA and FPUC programs. For purposes of this report,
when we refer to states’ administration of the UI program, we include states, territories,
and freely associated states. SWAs are responsible for administering unemployment
insurance programs, among other things. State unemployment tax revenues are held in
trust by the Secretary of the Treasury and are used by the states to pay for weekly regular
UI benefits.
5This amount includes about $230 billion in expenditures under the regular UI and
Expanded Benefits programs and about $670 billion in expenditures under the pandemic
UI programs that expired on September 6, 2021. However, 24 states ended their
participation in at least one of the pandemic UI programs before the programs expired.
6Fraud involves obtaining something of value through willful misrepresentation.
7The High Risk List highlights federal programs and operations that we have determined
are in need of transformation. It also names federal programs and operations that are
vulnerable to waste, fraud, abuse, and mismanagement. GAO, Unemployment Insurance:
Transformation Needed to Address Program Design, Infrastructure, and Integrity Risks,
GAO-22-105162 (Washington, D.C.: June 7, 2022); and High-Risk Series: Efforts Made to
Achieve Progress Need to Be Maintained and Expanded to Fully Address All Areas,
GAO-23-106203 (Washington, D.C.: Apr. 20, 2023).
Page 2 GAO-23-106696 Unemployment Insurance
did not fully reflect the extent of fraud and potential fraud in UI programs
during the pandemic. 8
SWAs endeavored to implement new temporary UI programs and
process unprecedented claims volumes during the pandemic. A key
challenge facing those SWAs was simultaneously ensuring that UI
benefits were paid to only those individuals eligible under program
requirements and were paid in the correct amounts. Accurate initial
determinations of eligibility were critical to ensuring that benefits were
granted only to those intended by the programs. This also included
ensuring that program monitoring over the use of funds was sufficiently
designed and accurately reported at the state and federal level. To help
assist SWAs, the CARES Act and the American Rescue Plan Act (ARPA)
contained provisions and, starting in March 2021, also provided additional
funding, for DOL to provide financial and technical assistance to states to
improve UI systems and processes. 9
UI overpayments—payments to ineligible recipients or payments in the
incorrect amounts—can be the result of error on the part of the employer,
claimant, the SWA, or a combination of these parties, or the result of
fraud. SWAs report identified overpayments—including fraudulent
overpayments—and recoveries to DOL.
You asked us to continue our work to develop a more comprehensive
estimate of UI fraud; review DOL financial and technical assistance
provided to states during the pandemic; and identify the extent to which
states have recovered, written off, or waived UI overpayments. This
report addresses (1) the estimate of fraud (lower and upper range) within
UI programs during the COVID-19 pandemic; (2) how much financial and
technical assistance DOL has allocated and awarded to states under the
CARES Act and ARPA, and how DOL tracks use of this assistance; and
(3) how much states have reported in UI overpayments and related
recoveries, write-offs, and waivers.
To address our first objective to develop an estimate of fraud within UI
programs during the pandemic, we combined information from multiple
sources to produce an upper and lower range on the extent of UI fraud
8GAO, Unemployment Insurance: Data Indicate Substantial Levels of Fraud during the
Pandemic; DOL Should Implement an Antifraud Strategy, GAO-23-105523 (Washington,
D.C.: Dec. 22, 2022).
9Pub. L. No. 116-136, §2102(f)(2)(B), 134 Stat. at 316; Pub. L. No. 117-2, §9032, 135
Stat. at 121.
Page 3 GAO-23-106696 Unemployment Insurance
during the pandemic. 10 Specifically, we combined separate estimates of
fraudulent payments associated with (1) the regular UI program,
Pandemic Emergency Unemployment Compensation (PEUC), Mixed
Earner Unemployment Compensation (MEUC) payments, Extended
Benefits, and the portion of Federal Pandemic Unemployment
Compensation (FPUC) payments not associated with PUA; and (2) the
PUA program, including FPUC payments associated with PUA claims.
The DOL OIG reported in October 2020 that the PUA program in
particular was at high risk for fraud due to its unique program rules and
eligibility requirements. 11 We developed separate procedures for the PUA
program because of the program’s unique fraud risk profile.
The scope of our review was from April 2020—the first full month of
pandemic UI program payments—through May 2023—the end of the
COVID-19 public health emergency. We estimated the extent of fraud
across all 53 SWAs for the regular UI and pandemic UI programs.
Throughout this report, we use the phrase “fraud estimate” or “estimate of
fraud” to refer to estimates that attempt to quantify the extent of fraud,
regardless of whether such fraud has already been detected and
adjudicated.
To derive the estimated fraud in the UI programs, excluding PUA, we
used data from DOL’s Benefit Accuracy Measurement (BAM) program—
which DOL uses to estimate the amount and rate of improper payments,
including those caused by fraud—from April 2020 through December
2022. Using the BAM program estimates, we developed a statistical
10We define the pandemic period as from April 2020 through May 2023. We selected April
1, 2020, as the beginning date for this range to reflect the period when all pandemic UI
program payments were being paid and to align with DOL’s quarterly reporting on
estimated fraud rates. While the pandemic UI programs expired in September 2021, the
COVID-19 public health emergency ended in May 2023. Therefore, we selected May 31,
2023, as the end date for this range.
11Department of Labor, Office of Inspector General, COVID-19: States Cite Vulnerabilities
in Detecting Fraud While Complying with the CARES Act UI Program Self-Certification
Requirement, Report No. 19-21-001-03-315 (Washington, D.C.: Oct. 21, 2020.)
Page 4 GAO-23-106696 Unemployment Insurance
model to impute the regular UI program fraud rate for the first 3 months of
the pandemic when BAM was suspended. 12
For the PUA program, we obtained the generalizable sample of 2,540
PUA payments that DOL selected as part of its improper payment
estimation effort. We then selected a subsample of 260 PUA payments
for further review. The DOL OIG used data analytic procedures to identify
the presence of fraud indicators in the sample of 2,540 PUA payments
and provided them to us. 13 To identify the presence of additional fraud
indicators, we cross-matched our sample with the Death Master File to
identify potentially deceased individuals and with the National Directory of
New Hires (NDNH) to identify claimants’ unreported wages. 14 For the
sample of 260 PUA payments, we then followed up on matches by
reviewing the state case files; discussing cases with the DOL OIG; and
reviewing publicly available information, when applicable, to determine
the risk of fraud on those matches. 15 We also matched information from
12DOL uses its BAM program to estimate the amount and rate of improper payments,
including those caused by fraud. The BAM program only includes testing of regular UI
claims. For fiscal years 2021 and 2022 improper payment reporting, DOL applied the
estimated improper payment rate from the BAM program testing of regular UI claims to
calculate the estimated improper payment amounts for FPUC and PEUC. Thus, the
estimated improper payment amounts for these two programs were incorporated into the
overall UI estimated improper payment amount reported for fiscal years 2021 and 2022.
However, this overall estimated improper payment amount for UI did not include an
estimate for PUA. According to DOL, it did not include PUA in the extrapolation of the
BAM estimated improper payment rate because the PUA program served a different
population of workers and had different eligibility requirements. To impute is to assign a
value to something by inference. Extrapolation is a technique that can offer a rough or
notional estimate of fraud or potential fraud, even if data on a specific measure or rate are
unavailable, but may have limitations related to validity, accuracy, and completeness.
13Fraud indicators are characteristics and flags that serve as warning signs suggesting a
potential for fraudulent activity. Indicators can be used to identify potential fraud and
assess fraud risk but are not proof of fraud, which is determined through the judicial or
other adjudicative system. The DOL OIG provided 18 indicators including, for example,
multistate claims and shared or suspicious email addresses.
14NDNH is a national repository of new hire, quarterly wage, and unemployment insurance
information reported by employers, states, and federal agencies. NDNH is maintained and
used by the U.S. Department of Health and Human Services for the federal child support
enforcement program, which assists states in locating parents and enforcing child support
orders. DOL does not have access to NDNH wage data; however, states have access to
NDNH wage data.
15In many cases, a fraud indicator may be explained by events other than fraud. The goal
of the manual review was to account for alternative explanations of the observed fraud
indicators. For example, an address may have a large number of claims because it is a
multiunit dwelling and so, when assessing fraud risk associated with individual addresses,
we examined the size of the dwelling and whether it was multiunit.
Page 5 GAO-23-106696 Unemployment Insurance
the case files against the Social Security Administration’s Enumeration
Verification System to identify claimants with invalid personal information.
The above steps produced manually adjusted fraud risk scores for the
sample of 260 PUA payments and programmatically generated fraud
indicators for the DOL sample of 2,540. We used a statistical procedure,
known as multiple imputation, to estimate manually adjusted fraud risk
scores for the sample of 2,540 payments, given our more detailed review
of the 260 payments. 16 We used the DOL sample design and sampling
weights to extend the sample results to the full population of PUA
payments.
To help assess the validity of our estimate of the extent of fraud in the
PUA program, we conducted an analysis of PUA benefit payments
volume over time from March 2020 to December 2021. Specifically, we
developed an econometric model to predict the level of PUA benefit
payments if all states were comprehensively implementing fraud
prevention tools or processes, using explanatory variables that captured a
broad range of state-level conditions, such as states’ COVID-19 disease
burden.
We combined our fraud estimate from the BAM program with our PUA
fraud estimate from our sample of PUA payments to estimate the upper
and the lower range of the amount of fraud in the UI programs during the
pandemic. See appendix I for additional details on our methodology for
calculating this estimate, and a full description of the limitations and
assumptions.
To address our second objective, we reviewed DOL data on the financial
and technical assistance provided to states from March 2020 through July
2023. This period covers the beginning of the pandemic through the most
recent month of data available at the time of our review. We reviewed
these data to determine the amount of financial and technical assistance
allocated and awarded to each state. We reviewed DOL guidance to
understand the grant reporting requirements to determine how the agency
oversees financial and technical assistance provided to states. We
conducted interviews with state officials from six selected states—
California, Florida, Kansas, Nevada, New York, and Washington—to
obtain information related to the assistance provided. These states were
selected based on a range of (1) the fraud risk level identified in our first
objective, (2) the amount of grant funding received, and (3) acceptance of
16In this report, we do not detail all the steps of our fraud scoring process so that potential
perpetrators of fraud do not become aware of fraud risks or exploit potential weaknesses
in the program.
Page 6 GAO-23-106696 Unemployment Insurance
DOL’s offer of financial and technical assistance. 17 Information from the
six selected states is not generalizable to all states.
To address our third objective, we reviewed the most recent data
available as reported by states as of May 1, 2023, to DOL through Forms
902P (PUA) and 227 (non-PUA) on the extent that states have recovered,
written off, and waived overpayments from March 2020 through March
2023. 18 These were the three most recent years available at the time of
our review. We conducted interviews with DOL officials and the six
selected states to obtain information related to (1) the identification of
fraudulent UI overpayments, (2) efforts to recover fraudulent UI
overpayments, (3) the criteria used to write off fraudulent overpayments,
and (4) waivers processed for nonfraudulent UI overpayments. Analysis
conducted for the six selected states is not generalizable. We conducted
various electronic tests to assess the reliability of the data. These tests
included identifying missing data, duplicate records, and values outside
our designated range. Based on the results of our electronic tests and our
review of reporting guidance, we determined these data to be sufficiently
reliable for the purpose of reporting amounts overpaid, recovered, written
off, and waived, as provided to DOL by the states.
We conducted this performance audit from January 2023 to September
2023 in accordance with generally accepted government auditing
standards. Those standards require that we plan and perform the audit to
obtain sufficient, appropriate evidence to provide a reasonable basis for
our findings based on our audit objectives. We believe that the evidence
obtained provides a reasonable basis for our findings based on our audit
objectives.
Background
17These six states were selected based on different criteria compared with the 14 SWAs
that were selected for interviews to obtain information about the operation of the pandemic
UI programs and help inform the estimation process. For the list of the 14 SWAs selected
for interviews to obtain information about the operation of the pandemic UI programs and
help inform the estimation process, see app. I.
18DOL’s Employment and Training Administration (ETA) 227 reports are used for the
regular UI, PEUC, FPUC, and MEUC programs. ETA 227 UI data are reported quarterly.
Additionally, reporting for the regular UI program includes totals for the Unemployment
Compensation for Ex-Service Members/Unemployment Compensation for Federal
Employees, and Extended Benefits programs.
Page 7 GAO-23-106696 Unemployment Insurance
Federally Funded UI The CARES Act created three new federally funded temporary UI
Programs in Response to programs that expanded UI benefit eligibility and enhanced benefits. 19
COVID-19 • PUA, which was generally available through September 6, 2021, and
authorized UI benefits to individuals not otherwise eligible for UI
benefits, such as the self-employed and certain gig economy workers,
who were unable to work because of specified COVID-19 reasons. 20
The total federal expenditure for PUA program benefits was $138
billion through May 31, 2023.
• FPUC, which generally authorized an additional $600 weekly benefit
through July 2020 and generally authorized a $300 weekly benefit for
weeks beginning after December 26, 2020, and ending on, or before,
September 6, 2021, for individuals eligible for weekly UI benefits
available under the regular UI program and CARES Act UI
programs. 21 According to DOL officials, the agency does not have a
breakout of how much FPUC money was distributed related to regular
UI or PUA. The total federal expenditure for FPUC program benefits
was $442 billion through May 31, 2023.
• PEUC, which was generally available through September 6, 2021,
and generally authorized additional weeks of UI benefits for those who
had exhausted their regular UI benefits. 22 The total federal
expenditure for PEUC program benefits was $90 billion through May
31, 2023.
In addition, the Consolidated Appropriations Act, 2021, created the MEUC
program, which was extended by ARPA and expired in September
19These programs were subsequently extended and amended by the Consolidated
Appropriations Act, 2021, as well as ARPA, and expired in September 2021. However, 24
states ended their participation in at least one of these programs before the programs
expired in September 2021.
20At the time of the program’s expiration in September 2021, PUA generally authorized up
to 79 weeks of benefits. Pub. L. No. 117-2, § 9011(a), (b), 135 Stat. 4, 118; Pub. L. No.
116-260, div. N, tit. II, § 201(a), (b), 134 Stat. 1182, 1950-1951 (2020); Pub. L. No. 116-
136, § 2102, 134 Stat. 281, 313 (2020).
21Pub. L. No. 117-2, § 9013, 135 Stat. 4, 119; Pub. L. No. 116-260, div. N, tit. II, § 203,
134 Stat. 1182, 1953; Pub. L. No. 116-136, § 2104 Stat. 281, 318.
22At the time of the program’s expiration, PEUC generally authorized an additional 53
weeks of benefits for claimants who were fully unemployed. Pub. L. No. 117-2, § 9016(a),
(b), 135 Stat. 4, 119-120; Pub. L. No. 116-260, div. N, tit. II, § 206(a), (b), 134 Stat. 1182,
1954; Pub. L. No. 116-136, § 2107, 134 Stat. 281, 323.
Page 8 GAO-23-106696 Unemployment Insurance
2021. 23 According to DOL, the MEUC program was intended to
supplement regular UI claimants whose benefits do not account for a
significant self-employment income. Consequently, these claimants may
have received a lower UI benefit than they would have received had they
been eligible for PUA. The total federal expenditure for MEUC program
benefits was $78 million through May 31, 2023.
UI Program Integrity The unprecedented demand for UI benefits and the urgency with which
states implemented the new programs during the pandemic increased the
risk of improper payments, including, but not limited to, those due to
fraud. 24 DOL uses its BAM program to estimate the amount and rate of
improper payments, including those caused by fraud. The BAM program
includes testing of regular UI claims.
For fiscal years 2021 and 2022 improper payment reporting, DOL applied
the estimated improper payment rate from the BAM program testing of
regular UI claims to calculate the estimated improper payment amounts
for FPUC and PEUC. 25 Thus, the estimated improper payment amounts
for these two programs were incorporated into the overall UI estimated
improper payment amount reported for fiscal years 2021 and 2022.
However, this overall estimated improper payment amount for UI did not
include an estimate for PUA. According to DOL, it did not include PUA in
23The MEUC program, which was voluntary for states, authorized an additional $100
weekly benefit for certain UI claimants who received at least $5,000 of self-employment
income in the most recent tax year prior to their application for UI benefits between
December 27, 2020, and September 6, 2021. Pub. L. No. 117-2, § 9013(a), 135 Stat. 4,
119; Pub. L. No. 116-260, div. N, tit. II, § 261(a)(1), 134 Stat. 1182, 1961.
24An improper payment is defined by law as any payment that should not have been made
or that was made in an incorrect amount (including overpayments and underpayments)
under statutory, contractual, administrative, or other legally applicable requirements,
including unknown payments. It includes any payment to an ineligible recipient, any
payment for an ineligible good or service, any duplicate payment, any payment for a good
or service not received (except for such payments where authorized by law), and any
payment that does not account for credit for applicable discounts. 31 U.S.C. § 3351(4).
When performing improper payment risk assessments and estimates, executive agencies
are required to treat as improper any payments whose propriety cannot be determined
due to lacking or insufficient documentation. 31 U.S.C. § 3352(c)(2).
25DOL did not calculate an estimated improper payment amount for the MEUC program,
according to officials, because the program only operated between January and
September 2021. Office of Management and Budget (OMB) guidance instructs agencies
to complete improper payment risk assessments for newly established programs after the
first 12 months of the program. If the agency determines that the program is susceptible to
significant improper payments as a result of the assessment, then, in the following year,
the agency should produce a statistically valid estimate of the program’s improper
payments. DOL officials explained that, because MEUC existed for less than one year,
DOL did not estimate or report improper payments for this program.
Page 9 GAO-23-106696 Unemployment Insurance
the extrapolation of the BAM estimated improper payment rate because
the PUA program served a different population of workers and had
different eligibility requirements. 26 In addition, the BAM program did not
cover the start of the pandemic due to a temporary 3-month suspension
of testing for claims filed from April 1, 2020, to June 30, 2020, in order to
allow BAM investigators to help process initial claims and adjudication in
operations. 27
DOL’s annual estimated improper payments in UI increased from $8.0
billion (9.2 percent estimated improper payment rate) for fiscal year 2020
to $78.1 billion (18.9 percent estimated improper payment rate) for fiscal
year 2021. For fiscal year 2022, DOL reported estimated improper
payments of $18.9 billion (22.2 percent estimated improper payment
rate). 28 Improper payments could suggest that a program may be
vulnerable to fraud. However, improper payments represent all
overpayments—including fraud—and underpayments resulting from any
type of intentional or unintentional error. This amount is not a valid
indicator of fraud in a particular program.
In the UI system, program integrity is a shared responsibility between the
federal and state governments. DOL provides general support and
technical assistance, and states assume responsibility for determining
eligibility, ensuring accurate benefit payments, and preventing fraud and
other improper payments. Under the BAM program, each state is to
review a number of randomly selected cases on a weekly basis and
reconstructs the UI claims process to assess the accuracy of the
payments that were made. A BAM investigator in the SWA is to review
each sampled claim and identifies errors and the causes of the error,
including those caused by fraudulent activity.
26While DOL planned to report a statistically valid national improper payment rate for PUA
by fall 2022, according to DOL, OMB requested that it conduct further analysis of the
outcomes recorded through the PUA case review process. Also, according to DOL, OMB
allowed additional time to conduct this analysis and report on PUA outcomes in fiscal year
2023. In August 2023, DOL released its estimate of improper payments made from March
2020 to September 2021 under the PUA program, concluding that the PUA program had a
total estimated improper payment rate of 35.9 percent. DOL noted that its analysis
focused on the broader universe of improper payments, does not isolate fraud, and should
not be considered a fraud estimate for the PUA program.
27According to DOL officials, BAM investigators helped process initial claims and
adjudication in operations because they had the experience in this area.
28DOL’s fiscal years 2020, 2021, and 2022 improper payment estimates do not include
PUA claims.
Page 10 GAO-23-106696 Unemployment Insurance
Fraud and Fraud-Related Fraud involves obtaining something of value through willful
Estimates misrepresentation, and it is a subset of overpayments. Whether any given
claim is fraudulent is determined through the judicial or other adjudicative
systems. The DOL OIG reported in November 2021 that fraud—
specifically claimants who received UI benefits through fraudulent
schemes, such as those perpetrated during the COVID-19 pandemic—
was one of the leading causes of improper payments. However, it did not
report a specific amount of fraud. 29
In October 2021, we reported that the amount of fraudulent and
potentially fraudulent activity in UI programs increased substantially after
implementation of the pandemic UI programs, relative to the amount of
such activity in the regular UI program before the pandemic. 30 For
example, the increased amount of benefits awarded and the PUA
program’s initial reliance on self-certification gave criminals incentive and
opportunities to commit fraud. DOL officials also identified other factors—
including significant increases in claims workload, new and inexperienced
staff and contractors, and quick implementation of new programs—that
provided additional opportunities for exploitation of program and system
vulnerabilities. In addition, DOL officials stated that the UI programs
during the pandemic were a key target for fraud because fraudsters could
receive a large amount of money in one payment because certain UI
claims could be backdated to the beginning of the eligibility period.
In December 2022, we found that federal and state fraud measures and
estimates indicated substantial fraud and potential fraud in UI programs
during the pandemic but did not fully reflect the extent of fraud. 31 While
federal and state entities have produced several fraud and fraud-related
measures and estimates of UI fraud during the pandemic, no estimate or
combination of estimates fully covers the extent of fraud in UI programs
29Department of Labor Office of Inspector General, Top Management and Performance
Challenges Facing the U.S. Department of Labor (Washington, D.C.: November 2021).
30GAO, COVID-19: Additional Actions Needed to Improve Accountability and Program
Effectiveness of Federal Response, GAO-22-105051 (Washington, D.C.: Oct. 27, 2021).
We define fraudulent activity as activity that has been confirmed to be fraudulent via an
adjudicative or other formal determination process. We define potentially fraudulent
activity as activity that has indicators that may suggest fraud.
31GAO-23-105523. We use the phrase “fraud measure” to discuss counts related to
proven fraud, such as adjudicated cases of fraud. We use the phrase “fraud estimate” to
discuss estimates that attempt to quantify what could be determined to be fraud—or the
extent of fraud—although such cases have not yet been, and may never be, proven.
Finally, we use the phrases “fraud-related” and “potential fraud” to describe measures and
estimates that attempt to quantify the extent of fraud indicators but do not suggest a
potential or actual determination of fraud.
Page 11 GAO-23-106696 Unemployment Insurance
during the pandemic. As discussed later in this report, we have developed
our own estimate, including a lower and upper range, of the total extent of
UI fraud during the pandemic.
DOL Assistance to States DOL provides administrative funding and support to SWAs as a function
of the regular UI program. Additionally, the CARES Act authorized DOL to
provide funding to states to administer the pandemic UI programs. 32 DOL
has used its authority under the CARES Act to provide states with
additional administrative funding to include the following for the pandemic
UI programs:
• prevent, detect, and investigate fraudulent overpayments;
• recover overpayments; and
• support identity verification and prevent identity fraud.
In March 2021, the President signed ARPA into law. The law created a
new section of the CARES Act and provided $2 billion in funding to DOL
to detect and prevent fraud, promote equitable access, and ensure the
timely payment of benefits to eligible workers with respect to the
unemployment compensation programs. In August 2021, DOL announced
initial funding to states to carry out work on four tracks to address
systemic shortcomings in access:
• sending expert “Tiger Teams” directly to states to help identify
process improvements that can speed benefit delivery, address
equity, and fight fraud;
• providing tools to help address immediate fraud concerns by
facilitating more effective identification verification processes;
• developing IT solutions that can be adopted by states to modernize
antiquated state technology; 33 and
• announcing funding opportunities to help states ensure timely
payment of benefits, promote equitable access, and combat fraud.
32The CARES Act did not specify the amount for DOL funding to states to administer the
pandemic UI programs.
33Many states rely on outdated legacy IT systems to operate their UI programs. The DOL
OIG and GAO have reported on the risks and challenges that legacy systems pose for
state UI programs, which have led to, among other things, reduced efficiency and
effectiveness. Legacy IT systems have led to slower payment processing, an inability to
detect and recover fraudulent overpayments, reporting difficulties, security vulnerabilities,
staffing challenges, and increased administrative costs. See GAO, Unemployment
Insurance: DOL Needs to Further Help States Overcome IT Modernization Challenges,
GAO-23-105478 (Washington, D.C.: July 10, 2023).
Page 12 GAO-23-106696 Unemployment Insurance
In June 2023, the Fiscal Responsibility Act of 2023 (FRA) was signed into
law. 34 This law rescinds total ARPA funding for UI programs that had not
been awarded and reduced the total ARPA funding for UI programs from
$2 billion to $1 billion. In July 2023, DOL announced updated financial
assistance amounts for states, which reflected the FRA rescission.
UI Overpayment States report to DOL on UI overpayments; recoveries; write-offs; and
Recoveries, Write-offs, waivers, among other information. An overpayment occurs when
individuals receive benefits to which they are not entitled. Once a state
and Waivers
determines that an overpayment has been made, the state must take
actions to recover the amount overpaid. If states have exhausted efforts
to collect an overpayment, the state may remove (write off) the amount
for accounting purposes within the authority of state law. A write-off does
not limit the legal authority of the state to collect the overpayment, should
the opportunity arise. Under the pandemic UI programs, a state can waive
the legal right to recover the overpayment, in limited circumstances,
where the individual is not at fault and repayment would be contrary to
equity and good conscience.
State laws provide methods for the recovery of benefit overpayments,
including fraudulent overpayments. 35 States use several tools to recover
overpayments, such as direct repayment, offsetting future UI benefits,
and assessing penalties. Under federal law, states must recover certain
types of overpayments by offsetting an individual’s federal income tax
refund payment through the Treasury Offset Program, including
overpayments due to fraud and overpayments due to misreported work
and earnings. 36 Similarly, states may also offset overpayments with
monies owed to the individual from state tax refunds or lottery winnings,
or the state can compel repayment by pursuing civil action in state court.
Some state laws may also include provisions for denying or suspending
professional licenses of persons owing an overpayment of UI benefits.
For overpayments due to fraud, states may bring criminal charges, which
can lead to fines and prison sentences. Federal law requires a mandatory
penalty assessment for fraudulent claims of not less than 15 percent of
the amount of the erroneous payment against claimants committing fraud
34The Fiscal Responsibility Act of 2023, Pub. L. No. 118-5, 137 Stat. 10 (2023), rescinded
$1 billion of the unobligated balance.
35Because states may use different definitions for categorizing an overpayment as
fraudulent, an overpayment that is classified as fraudulent in one state might not be
classified as fraudulent in another state.
3642 U.S.C.§ 503(a)(5).
Page 13 GAO-23-106696 Unemployment Insurance
in connection with states’ or federal UI programs. 37 Figure 1 illustrates
examples of how states recover overpayments and penalties for
fraudulent overpayments.
37Although UI benefit fraud typically involves an individual’s attempt to obtain or increase
benefits, it also includes employers who attempt to prevent or reduce benefits to eligible
individuals, and employers who help an individual attempting to fraudulently claim
benefits. 42 U.S.C. § 503(a)(11).
Page 14 GAO-23-106696 Unemployment Insurance
Figure 1: Examples of Tools That States Use to Recover Overpayments, and Penalties for Fraudulent Overpayments
States can write off fraudulent overpayments after exhausting all options
to recover them and deem them unrecoverable. After exhausting options
to recover overpayments, most states will permit the SWA to write off
Page 15 GAO-23-106696 Unemployment Insurance
certain types of overpayments—meaning that the SWA will remove the
debt from its books as being uncollectible. Writing off an overpayment is
not the same as a state waiving recovery of an overpayment. Writing off
an overpayment is an accounting procedure and does not impact the
state’s legal right to collect an overpayment, should the opportunity arise.
States may choose to write off overpayments based on how long the
overpayment has been outstanding (i.e., age of the overpayment), or in
cases of bankruptcy or death of the individual.
Under certain circumstances, the state may waive the recovery of the
overpayment. To waive the recovery of pandemic UI benefit
overpayments, states must determine that the individual is not at fault and
that overpayment repayment would be contrary to equity and good
conscience. For example, states may waive the recovery of
overpayments when the overpayment is due to an agency or employer
error. 38 The CARES Act provides authority for all states to opt to waive
recovery of certain nonfault overpayments of pandemic UI benefits in
cases where the recoupment would be against equity and good
conscience. 39 Additionally, there are limited circumstances under the
CARES Act in which states may use blanket waivers of certain nonfault
overpayments. 40 Waiving recovery of an overpayment involves the state
waiving its legal rights to collect an overpayment.
Prior GAO Since 2018, GAO has made 26 recommendations to DOL to improve the
Recommendations UI system. As of August 2023, DOL has implemented ten of those
recommendations. However, 16 recommendations—including four
Intended to Improve the UI
involving fraud risk management—have either not been implemented or
System only partially addressed.
• Three recommendations—including one priority recommendation and
two other recommendations related to fraud risk management—have
been partially addressed, meaning that the agency has completed
38According to DOL documentation, the following 11 states do not have waivers from
overpayment recovery: Delaware, Kentucky, Mississippi, Missouri, Nebraska, New
Mexico, New York, Oklahoma, Puerto Rico, Texas, and West Virginia.
39Pub. L. No. 116-136, §§ 2104(f)(2), 2105(f), 2107(e)(2), 134 Stat at 319-327; Pub. L. No.
116-260, div. N, tit. II, § 201(d), 134 Stat. 1182, 1952.
40DOL’s Unemployment Insurance Program Letter (UIPL) No. 20-21, Change 1, provided
guidance to states on the permissible use of blanket waivers.
Page 16 GAO-23-106696 Unemployment Insurance
action(s) that contribute to implementation but has not yet completed
all actions to fully implement the recommendation. 41
• Thirteen recommendations—including two related to fraud risk
management—have not yet been implemented. Three of these are
priority recommendations.
GAO continues to monitor the implementation status of these
recommendations. See appendix II for a list of the 26 GAO
recommendations.
We estimate that the fraud in UI programs during the pandemic—from
Estimated UI April 2020 through May 2023—was likely between $100 billion and $135
Program Fraud during billion. This represents about 11 percent and about 15 percent,
respectively, of the total amount of UI benefits paid during the pandemic.
the Pandemic
This estimate covers the period from April 2020 (first full month of
Ranges from $100 payments from all UI programs) to May 2023 (end of the public health
Billion to $135 Billion emergency) and all 53 states that participated in the regular UI and
pandemic UI programs.
As part of our work to calculate this estimate, we separated UI
expenditures by whether the expenditures were associated with the PUA
program, which had a unique fraud risk profile. Given the time frame of
this review, we were not able to obtain sufficient evidence about the PUA
program to report a separate statistical estimate for that program. Instead,
we designed our procedures such that when the total evidence of PUA
and non-PUA payments was considered together, the combined evidence
was sufficient to support an overall estimate of the extent of fraud in the
UI programs during our review. 42
Judicial or other adjudicative systems make final determinations of
whether any given UI claim is fraudulent. Fraudulent activities frequently
go undetected due to their deceptive nature and the limited resources
41Priority recommendations are those that GAO believes warrant priority attention from
heads of key departments or agencies. They are highlighted because, upon
implementation, they may significantly improve government operations, for example, by
realizing large dollar savings; eliminating mismanagement, fraud, and abuse; or making
progress toward addressing a high risk or fragmentation, overlap, or duplication issue.
42In this context, sufficiency depends on the precision of the estimate. The precision of our
overall estimate is captured by the width of our reported range, which accounts for the
statistical uncertainty associated with both the PUA and non-PUA payments at the 95
percent confidence level. We do not report our range at the 95 percent confidence level
because statistical intervals do not capture the uncertainty associated with identifying
which cases in the sample were fraudulent. To reduce this latter source of uncertainty, we
leveraged multiple data sources and review procedures. (See app. I for more details).
Page 17 GAO-23-106696 Unemployment Insurance
available to investigate and adjudicate fraud. We designed this range to
capture the extent of fraudulent activity, regardless of whether that activity
was previously detected or adjudicated. Because not all potential fraud
will be investigated and adjudicated through judicial or other systems, the
full extent of UI fraud during the pandemic will likely never be known with
certainty. Due to our use of statistical methods and the uncertainty
associated with estimating fraud without final adjudications, the actual
amount of fraud could be greater than or less than our estimated range.
See appendix I for more detailed information on the methodology we
used to estimate fraud in the UI programs during the pandemic, including
the limitations and assumptions associated with the analysis.
We have previously estimated the extent of fraud in the UI programs. In
December 2022, we estimated that at least $60 billion in fraudulent UI
payments were made to claimants by extrapolating the lower bound of
DOL’s 2021 estimated national fraud rate for the regular UI program to
total UI spending. 43 However, we concluded that the actual amount of
fraud in UI programs during the pandemic could be substantially higher
than the estimated $60 billion lower limit.
We now estimate that the amount of fraud was higher, with our new range
of $100 billion to $135 billion falling above the lower limit that we reported
in December 2022. To calculate our previous lower limit, we relied on
existing evidence about the extent of fraud in the UI programs. Our
current range extended this work through a substantial methodology
employing independent sampling and modeling work. 44 In addition, our
analysis supports the presence of higher fraud rates for PUA payments,
which matches our previous reporting about the increased fraud risk
associated with the PUA program. 45
In February 2023, the DOL OIG estimated that at least $191 billion in UI
payments during the pandemic could have been improper, with a
43GAO-23-105523.
44As described earlier, we relied on DOL’s BAM program to estimate the total fraud in the
regular UI program, PEUC, MEUC, Extended Benefits, and the portion of FPUC payments
that were not associated with PUA claims. To estimate fraud in the PUA program,
including FPUC payments associated with PUA claims, we selected a generalizable
sample of PUA payments and then reviewed the payments for fraud risk using multiple
public and nonpublic data sources. We also developed an econometric model to predict
the level of PUA benefit payments if all states were comprehensively implementing fraud
prevention tools or processes, using explanatory variables that captured a broad range of
state-level conditions, such as states’ COVID-19 disease burden.
45GAO-23-105523.
Page 18 GAO-23-106696 Unemployment Insurance
significant portion attributable to fraud. 46 We did not estimate improper
payments, but our findings are generally consistent with the DOL OIG’s
statement regarding the significance of fraud in the UI programs.
DOL Has Allocated
$1.4 Billion in
Assistance to States
and Is to Track Funds
through Quarterly
Reporting
DOL Allocated about $1.4 DOL initially allocated over $2 billion in CARES Act and ARPA funding to
Billion and Awarded $872 states for initiatives including fraud prevention; detection; investigation;
and overpayment recovery, among others. 47 As part of the FRA
Million in CARES Act and
rescission, in July 2023, DOL officials revised APRA spending plans and
ARPA Funds for Fraud issued updated funding allocations. ARPA funding was reduced by $639
Prevention Efforts million from the original allocation in the following three categories: (1)
Tiger Team funding was reduced by $86 million, (2) fraud prevention was
reduced by $100 million, and (3) IT modernization was reduced by $453
million.
After accounting for the rescission, DOL’s revised allocation of ARPA
funding resulted in about $1.4 billion in funding allocated to states. This
figure includes a total of $525 million in CARES Act funding to address
fraud in the pandemic UI programs and over $879 million in ARPA
funding to address fraud, increase equity, reduce backlogs, and
undertake other initiatives. 48 According to DOL guidance, any remaining
ARPA funds that have not been allocated for financial assistance could
eventually be used to procure identity verification tools for states. Table 1
provides a summary of financial assistance as of July 2023.
46Larry D. Turner, Inspector General, Department of Labor, Office of Inspector General,
testimony before the House of Representatives Committee on Ways and Means, 118th
Cong., 1st sess., February 8, 2023.
47Allocated amounts represent the maximum amount of funds available for states to apply
for in each of these grants.
48This amount reflects the partial impact of the FRA’s rescission of $1 billion from the
original ARPA grant funding amount.
Page 19 GAO-23-106696 Unemployment Insurance
Table 1: Coronavirus Aid, Relief, and Economic Security (CARES) Act and American Rescue Plan Act (ARPA) Unemployment
Insurance (UI) Financial Assistance for Fraud Prevention Efforts, July 2023
Funding Announcement Funding allocated (dollars) Examples of allowable UI programs
source date uses
CARES Act August 2020 100,000,000 • Prevent and Pandemic Unemployment
detect fraud and Assistance (PUA), Pandemic
identity theft Emergency Unemployment
• Recover Compensation (PEUC)
fraudulent
overpayments
• Investigate fraud
January 2021 100,000,000 • Identity PUA, PEUC
verification
• Prevent and
detect fraud
• Investigate fraud
• Recover
fraudulent
overpayments
August 2021 100,000,000 • Identity PUA, PEUC
verification
• Prevent and
detect fraud and
identity theft
• Recover
fraudulent
overpayments
July 2022 225,000,000 • Detect fraud PUA, PEUC, Federal
• Recover Pandemic Unemployment
overpayments Compensation (FPUC)
Total CARES Act 525,000,000
ARPA August 2021 260,000,000 • Remove access All programs
barriers
• Reduce backlogs
• Improve
timeliness
• Increase equity in
fraud detection
Page 20 GAO-23-106696 Unemployment Insurance
August 2021 140,000,000 • Identity All programs
verification
• Prevent and
detect fraud
• Improve data
management
• Improve
cybersecurity
• Recover
overpayments
November 2021 160,121,650a • Fund expert All programs
“Tiger Teams”
consultations
• Prevent and
detect fraud
• Recover
fraudulent
overpayments
• Promote
equitable access
• Reduce backlogs
• Improve
timeliness
December 2021 1,200,000b • Modernize UI All programs
information
technology
system pilot
program
January 2022 18,025,506 • Improve All programs
timeliness
• Help workers
navigate UI
application
process
July 2023 100,000,000 • Identity All programs
verification
• Prevent and
detect fraud
• Overpayment
recovery
July 2023 200,000,000c • Modernize UI All programs
information
technology
systems
Total ARPA 879,347,156
Total funding 1,404,347,156
Source: GAO analysis of Department of Labor information. | GAO-23-106696
Page 21 GAO-23-106696 Unemployment Insurance
aThe November 2021 funding includes $114 million in grants to states for implementing
improvements to their UI systems. The remaining $46 million has been used to fund the expert Tiger
Team consultations. Tiger Teams are UI experts that work with states to identify process challenges
and areas of improvement.
bDOL Announcement TEN [Training and Employment Notice]16-21, issued in December 2021,
initially announced up to $600,000 to selected states but awarded $1.2 million as an additional state
was selected for participation, as of May 2023.
cStates could apply for funding up to $11.25 million, regardless of size.
DOL allowed states to apply for and be awarded financial assistance up
to a fixed amount. For much of the financial assistance, DOL determined
the amount of financial assistance allocated to states based on either the
12-month average of UI-covered employment in the state or a
combination of covered employment and the number of first payments
made. 49 According to DOL Office of UI Modernization officials, not all
states applied for the financial assistance available. As of May 2023, DOL
awarded $872 million to states in financial assistance. Of the $872 million,
about $398 million was awarded from ARPA funding and about $475
million from the CARES Act. Figure 2 shows how much financial
assistance each state has been awarded.
Figure 2: Department of Labor (DOL) Financial Assistance Awarded to States, May
2023
49Covered employment refers to the number of employees covered by UI reported to the
states by employers.
Page 22 GAO-23-106696 Unemployment Insurance
According to officials from six selected states, they used CARES Act and
ARPA financial assistance to perform such things as
• administer fraud detection and prevention initiatives, including identity
verification and multifactor authentication software,
• improve and expand overpayment recovery efforts,
• improve the availability of translation services,
• reduce backlogs through process improvements and additional
staffing,
• improve the claimant experience with chatbots and enhanced self-
service portals, 50 and
• fund staff working on pandemic UI program reporting.
Appendix III provides a list of CARES Act and ARPA financial assistance
that DOL allocated and awarded to each state. 51
Tiger Team initiative. DOL provided technical assistance to states
separate from grants. Specifically, DOL provided expert assistance to
states through its Tiger Team initiative and established mechanisms to
facilitate state identity proofing. With funding provided by ARPA, DOL
allocated grant funds of up to approximately $114 million to support states
in improving UI systems and processes, which included addressing fraud
prevention and detection. As of April 2023, DOL had allocated $46 million
in additional ARPA funds via contracts for state Tiger Team consultations.
Through these contracts, multidisciplinary Tiger Teams are to analyze
state UI systems and work with states to identify process challenges and
areas of improvement. The Tiger Teams have been composed of staff
with expertise on UI systems from DOL, the National Association of State
Workforce Agencies, and a consulting firm. 52 Tiger Teams are to work
with states to identify ways to enhance their systems and processes by
making actionable recommendations. States utilize grant funding
50A chatbot is an interactive and automated system that can answer questions for
claimants and that frees up staff to assist claimants more efficiently.
51Awards represent the amount of funds that DOL approved states to receive after the
application process.
52Each Tiger Team should be comprised of experts, including a fraud specialist,
equity/customer experience specialist, UI program specialist, business intelligence
analysts, computer systems engineer/architect, and project manager. See Department of
Labor, Grant Opportunity to Support States Following a Consultative Assessment for
Fraud Detection and Prevention, Promoting Equitable Access, and Ensuring the Timely
Payment of Benefits, including Backlog Reduction, for all Unemployment Compensation
(UC) Programs, UIPL No. 02-22 (Washington, D.C.: Nov. 2, 2021).
Page 23 GAO-23-106696 Unemployment Insurance
provided by DOL to make near-term improvements recommended by the
Tiger Teams. The recommendations that states receive, and the
corresponding improvements that states make with the funding, must
align with three pillars of ARPA:
• Equitable access
• Fraud prevention, detection, and recovery
• Payment timeliness and backlog reduction
The Tiger Team initiative has two phases: (1) a consultative assessment
and recommendations phase and (2) a subsequent funding and
implementation of potential solutions phase. During the Tiger Team
consultative assessment and recommendations phase, SWAs are to work
directly with the Tiger Team to identify areas of improvement within the
state UI system. Once the recommendations are finalized, the Tiger
Teams are to negotiate with the SWA to determine which
recommendations to fund and implement. SWAs then apply for grant
funding from DOL to implement the agreed-upon recommendations.
According to DOL, both phases of the Tiger Team initiative are underway.
For consultative assessments and recommendations, DOL officials stated
that as of May 2023, 45 states had applied for Tiger Team consultations;
of these 53
• five states had Tiger Team consultations in process;
• 11 states have not started the Tiger Team consultations; 54 and
• 29 states have completed their Tiger Team consultations, which
resulted in 301 recommendations to states.
The Tiger Teams are to develop recommendations aligned with the three
pillars of ARPA, with a focus on recommendations that states can more
readily implement. Of the six selected states, three are reported to be in
the process of funding and implementing Tiger Team recommendations
and have aligned their funding based on the three pillars. Tiger Team
recommendations to these selected states included developing plain
language materials and collecting additional data from applicants to
53From UIPL No. 02-22, Change 2, the deadline for states to express their interest in
participating in this initiative was March 31, 2023.
54DOL stipulates a period of performance for Tiger Team funding, but DOL officials said
that states could request additional time to use the funds. As a result of the FRA, states
that did not begin the consultative assessment as of June 30, 2023, are not eligible for
implementation grant funding.
Page 24 GAO-23-106696 Unemployment Insurance
analyze barriers to access. 55 In addition, to improve fraud prevention,
detection, and recovery efforts, recommendations have been made to
improve identity authentication software and UI system security, including
creating a fraud data warehouse. Finally, recommendations have been
made to improve payment timeliness and reduce backlogs, which
included automating internal UI system processes and implementing
dynamic fact-finding for employment separations. 56 Of the three
remaining selected states without recommendations:
• one is currently engaged in the consultative assessment and
recommendations phase,
• one has expressed interest but is not yet participating, and
• one chose not to participate in the initiative.
See figure 3 for the recommendations that selected states have reported
implementing that align with the three pillars.
Figure 3: Tiger Team Recommendations That Selected States Have Reported Implementing That Align with the Three Pillars
Note: Tiger Team recommendations may fit under multiple program pillars, and the figure is not a
comprehensive representation of what applies under each pillar.
Identity-proofing mechanisms. Identity fraud was a major contributor to
UI fraud during the pandemic. DOL provided states with two channels to
acquire identity-proofing software to help reduce fraud. First, the agency
established a Blanket Purchase Agreement (BPA) for states to procure
55Tiger Team documents suggest that plain-language materials may also help reduce
improper payments through (1) increased compliance with UI policies and (2) reduced
claimant errors.
56Dynamic fact-finding is an automated process that asks claimants a series of questions
to determine the reason for their employment separation (i.e., fired, laid off, quit).
Page 25 GAO-23-106696 Unemployment Insurance
identity-proofing services. 57 It competitively awarded BPAs to three
identity-proofing vendors: LexisNexis, V3Gate, and TransUnion. DOL
officials told us that they spent roughly $285,000 to test the identity-
proofing services integration with state UI systems. As of May 2023, no
states have utilized the BPA to procure identity-procurement services.
Officials from our six selected states said that they did not use the BPA
for a variety of reasons. For example, officials from one state said that
they had to comply with their own state’s procurement process. Officials
from another state said that their state had significant buying power and
did not need to use the BPA. Officials from the other selected states said
that they had already acquired identity-proofing services or had existing
relationships with vendors.
Second, DOL also collaborated with states to provide identity verification
services through the General Service Administration’s Login.gov and the
U.S. Postal Service to provide states with both online and in-person
identity proofing. Arkansas was the first state to implement Login.gov
integration through a pilot initiative from November 2021 through June
2022. In July 2023, DOL offered to provide online and in-person identity
proofing to states through Login.gov and the U.S. Postal Service. DOL is
planning to use ARPA funds to cover two years of transaction costs for
these services, depending on the availability of funding.
DOL Is to Track the Use of According to DOL officials, the agency’s Employment and Training
Financial and Technical Administration (ETA) is to track states’ use of financial and technical
assistance through quarterly reporting. States are required to provide
Assistance through
ETA with quarterly narrative progress and financial reports. DOL officials
Quarterly Reports said that regional offices are to review these reports and monitor the use
of these funds. According to DOL’s guidance, ETA is to use the quarterly
progress reports to track each state’s progress in implementing the
agreed-upon Tiger Team recommendations with the funds that DOL
provided, along with other funded projects. The quarterly reports should
identify the specific agreed-upon recommendations being implemented
and specific outcome metrics as they relate to those activities, as well as
ensuring that the state’s use of funds is consistent with the allowable use
of funds. The reports should also contain updates on all grant obligations
and disbursements made by the states. The data collected from the
reports are to be used by ETA to assess the effectiveness of programs,
monitor compliance with statutory limitations, and analyze financial
activity. In addition to quarterly reporting, officials from a selected state
57A BPA is an agreement established with a supplier to fill a repetitive need for services.
BPAs streamline the ordering process. Federal Acquisition Regulation § 13.303.
Page 26 GAO-23-106696 Unemployment Insurance
also said that ETA regional office staff have frequent discussions with
states regarding the use of funds.
Based on the most recent data available, as of May 1, 2023, states have
States Have reported identifying about $55.8 billion in established fraudulent and
Reported Billions in nonfraudulent overpayments across all UI programs from March 2020—
the beginning of the pandemic—through March 2023. States also
UI Overpayments, reported recoveries of about $6.8 billion, which is approximately 12
Recoveries, Write- percent of overpayments identified during the same period. All identified
overpayments are reported during the period in which they are
Offs, and Waivers to established, and recovered funds are reported as they are collected.
DOL However, recoveries can take many years to collect, and states can
modify recovery figures daily, which makes comparisons between
overpayments and recoveries difficult. 58 Figure 4 illustrates the total
amount of UI overpayments and recoveries that states reported for this
period.
Figure 4: Total State-Reported Established Overpayment and Recovery Amounts for all Unemployment Insurance (UI)
Programs, March 2020 – March 2023 (as of May 1, 2023)
Note: All overpayments are reported during the period in which they are established, and recovered
funds are reported as they are collected. However, recoveries can take many years, which makes
comparisons between overpayments and recoveries difficult. Additionally, some states did not report
totals for the pandemic UI programs during our review time frame. Specifically, three states did not
report information for Pandemic Unemployment Assistance (PUA), three states did not report
information for Pandemic Emergency Unemployment Compensation (PEUC), two states did not
report information for Federal Pandemic Unemployment Compensation (FPUC), and 19 states did not
report information for Mixed Earner Unemployment Compensation (MEUC).
58In ongoing work, we are reviewing agency COVID-19 overpayment recovery efforts,
including those for UI systems.
Page 27 GAO-23-106696 Unemployment Insurance
States Have Reported As of May 1, 2023, states reported identifying about $5.3 billion in
Fraudulent UI established fraudulent UI overpayments for regular and pandemic UI
programs from March 2020 through March 2023. States also reported
Overpayments,
recoveries of about $1.2 billion, which is approximately 23 percent of
Recoveries, and Write- fraudulent UI overpayments identified during this period. Specifically, as
Offs related to fraudulent overpayments and recoveries, states reported
identifying about
• $1.4 billion in fraudulent overpayments and about $1 billion in
recoveries for the regular UI program, and
• $3.9 billion in fraudulent overpayments and $214 million in recoveries
for the pandemic UI programs.
Figure 5 provides the fraudulent UI overpayment amounts and recoveries
for the regular and pandemic UI programs reported by states during this
period. Appendix IV provides amounts of fraudulent overpayments and
recoveries reported by each state for both the regular and pandemic UI
programs.
Figure 5: Total State-Reported Established Fraudulent Overpayment Amounts for Unemployment Insurance (UI) Programs,
March 2020 – March 2023 (as of May 1, 2023)
Note: All overpayments are reported during the period in which they are established, and recovered
funds are reported as they are collected. However, recoveries can take many years, and some of the
overpayment recoveries for the regular UI program are from overpayments established prior to the
pandemic, which makes comparisons between overpayments and recoveries difficult. Additionally,
some states did not report totals for the pandemic UI programs during our review time frame.
Specifically, three states did not report information for Pandemic Unemployment Assistance (PUA),
three states did not report information for Pandemic Emergency Unemployment Compensation
(PEUC), two states did not report information for Federal Pandemic Unemployment Compensation
(FPUC), and 19 states did not report information for Mixed Earner Unemployment Compensation
(MEUC).
Potential overpayments may be identified through a variety of methods,
such as cross-matches or fraud hotline tips. For example, states may
review interstate benefit matches to identify duplicate claims filed in other
Page 28 GAO-23-106696 Unemployment Insurance
states and under other UI programs. States must conduct an investigation
before issuing an official determination that an overpayment has been
made. Also, in the course of the investigation, states may determine that
the overpayment was due to fraud.
According to DOL officials, states do not report fraudulent UI
overpayments until investigations are complete and fraud has been
confirmed, which may take a long time to establish. States must ensure
that individuals that have the overpayment receive an opportunity to
respond and to present evidence before making an overpayment
determination.
Officials from the six selected states said that they have taken a variety of
actions to identify and recover fraudulent overpayments. For example,
officials from four states reported using a combination of data analytics,
such as cross-matching with a variety of databases and a manual review
process to determine whether an overpayment was due to willful
misrepresentation. Officials from another state reported using a manual
review process to establish identity theft cases that had not been flagged
by their system. Further, officials from one state reported that they review
evidence provided by third parties as part of their fraud determination
process. Lastly, officials from one state reported that they focused their
recovery efforts on the 75 banks that managed the majority of fraudulent
claims.
Table 2 provides information about how the six selected states recover
fraudulent overpayments through benefit offset or withholding state
income tax refunds.
Table 2: Six Selected State Law Provisions for Recovering Fraudulent Unemployment Insurance (UI) Overpayments
State Recovery of fraudulent overpayments through benefit offset State tax refunds
Percentage that can be withheld Number of years limited
from UI weekly benefit
California 100% 6 years from mailing the Yes
overpayment notice
Florida 100% Commenced within 7 years Not applicable – due to lack of state income tax
from date overpayment
established
Kansas 100% No years limited Yes
Nevada 100% 10 years from date Not applicable – due to lack of state income tax
overpayment established
New York 100% No years limited Yes
Washington 100% No years limited No
Source: GAO analysis of Department of Labor documents. | GAO-23-106696
Page 29 GAO-23-106696 Unemployment Insurance
DOL officials said that recovering fraudulent overpayments in the
pandemic UI programs has been more challenging than the regular UI
program because of the differences in the type of fraud primarily being
committed. Specifically, in the regular UI program, individuals most
commonly commit eligibility fraud by falsifying information on their
application in an effort to obtain benefits to which they are not entitled. In
contrast, the pandemic UI programs—such as PUA—experienced large
amounts of identity fraud in which unknown suspects used stolen
identities to receive unemployment insurance benefits. The identification
and recovery of overpayments lost to identity fraud requires a coordinated
effort with law enforcement partners, which also takes longer to recover
than traditional recoveries.
Additionally, a comparison between regular UI overpayment and recovery
amounts is difficult. Overpayments and recoveries are reported during the
period in which they are established. 59 States are required to report
overpayment and recovery data to DOL on a continuous, rolling basis for
regular UI and most pandemic UI programs throughout the reporting
quarter. 60 States can also amend prior period data reported—going back
many periods–at any time during the quarter, so overpayment and
recovery amounts reported can change from day to day. Additionally,
since the regular UI program can have established overpayments that
occurred prior to the pandemic, SWAs are able to include any recovered
UI amounts in their current reporting. As a result, states might report
instances where the amount recovered for the regular UI program
appears to exceed the amount reported in established overpayments for
a given period. The DOL OIG has also previously reported that not all
states have provided these reports or have reported accurate data. 61
According to DOL officials, data are not yet available to show if states
using ARPA grant funds to assist with overpayment recovery efforts have
improved recovery rates.
DOL allows SWAs to write off certain types of overpayments consistent
with DOL guidance and statute—meaning that the SWA will remove the
59An established case is defined as any single issue involving either a fraudulent or
nonfraudulent overpayment that has been determined for a claimant within a single
calendar month or quarter and for which a formal notice of decision is issued.
60States report activity for the PUA program to DOL each month, which includes activities
performed during the preceding calendar month.
61Department of Labor, Office of Inspector General, Employment and Training
Administration, Advisory Report, CARES Act: Initial Areas of Concern Regarding
Implementation of Unemployment Insurance Provisions, Report no. 19-20-001-03-315
(Washington, D.C.: Apr. 21, 2020).
Page 30 GAO-23-106696 Unemployment Insurance
debt from their books as being uncollectible. Based on our analysis of
state-reported data, states have written off approximately $110 million in
fraudulent UI overpayments from March 2020 through March 2023. Five
of our six selected states have write-off provisions for fraudulent
overpayments after a specified period has passed from the date of
establishing the overpayment. Two states write off overpayments once
claimants are deceased, and two states write off overpayments in certain
cases that have been determined to be noncollectible based on state law.
One state does not have a specified period after which amounts are
written off. Table 3 provides information on selected states’ write-off
criteria.
Table 3: Selected State Fraudulent Overpayment Write-Off Criteria
State Age of the overpayment Other criteria, if applicable
Californiaa,b 6 to 10 years from establishment and no repayment funds received through Immediately if overpayment is
collection activity in the previous 36 months less than $10
Floridaa 5 years from establishment Bankruptcy or death
Kansasa 10 years from last recorded transaction Death
Nevadab 3 years from establishment Not applicable
New Yorka 10 years from last action on overpayment and 20 years from date judgment is filed, Not applicable
extended by payment activity
Washingtonb No period specified No cost-effective means of
collecting
Source: GAO analysis of Department of Labor documents. | GAO-23-106696
aWrite-off provisions found in policy.
bWrite-off provisions found in law.
While states are permitted to write off fraudulent UI overpayments, DOL
rules do not allow states to waive recovery of fraudulent overpayments.
Officials from selected states confirmed that their states did not waive
recovery of fraudulent UI overpayments.
States Have Reported As of May 1, 2023, states reported identifying about $50.5 billion in
Nonfraudulent established nonfraudulent UI overpayments for regular and pandemic UI
programs, from March 2020 through March 2023. States also reported
Overpayments,
recoveries of about $5.6 billion, which is approximately 11 percent of
Recoveries, Write-Offs, overpayments identified during this period. Specifically, states reported
and Waivers identifying about
• $11.3 billion in nonfraudulent overpayments and $2.6 billion in
recoveries for the regular UI program, and
Page 31 GAO-23-106696 Unemployment Insurance
• $39.3 billion in nonfraudulent overpayments and $3.0 billion in
recoveries for the pandemic UI programs.
Figure 6 provides the total nonfraudulent overpayment amounts and
recoveries for the regular and pandemic UI programs reported by states
during this period. Appendix V provides information on the amount of
nonfraudulent overpayments and recoveries reported by each state for
the regular and pandemic UI programs.
Figure 6: Total State-Reported Established Nonfraudulent Overpayment Amounts for Unemployment Insurance (UI) Programs,
March 2020 – March 2023 (as of May 1, 2023)
Note: Some states did not report totals for the pandemic UI programs during our review time frame.
Specifically, three states did not report information for Pandemic Unemployment Assistance (PUA),
three states did not report information for Pandemic Emergency Unemployment Compensation
(PEUC), two states did not report information for Federal Pandemic Unemployment Compensation
(FPUC), and 19 states did not report information for Mixed Earner Unemployment Compensation
(MEUC).
Table 4 provides information about how six selected states recover
nonfraudulent overpayments through benefit offset.
Table 4: Six Selected State Law Provisions for Recovering Nonfraudulent Unemployment Insurance (UI) Overpayments
State Recovery of overpayments through benefit offset
Percentage that can be withheld from UI weekly benefit Number of years limiteda
California 25% 6 years from mailing the overpayment notice
Florida 100% Commenced within 7 years from date overpayment
established
Kansas 100% No years limited
Nevada 50% 5 years from date overpayment established
New York 50% No years limited
Washington 50% (up to 100%, depending on claimant request) No years limited
Source: GAO analysis of Department of Labor documents. | GAO-23-106696
Page 32 GAO-23-106696 Unemployment Insurance
aThese are the state law provisions applicable to the regular UI program. The CARES Act, as
amended, statutorily limits benefit offsets for recovering Pandemic Emergency Unemployment
Compensation, Mixed Earner Unemployment Compensation, and Federal Pandemic Unemployment
Compensation overpayments to three years. This same limitation does not apply to Pandemic
Unemployment Assistance.
States also reported writing off about $849 million in nonfraudulent
overpayments during this period. As with fraudulent overpayments, most
states, after exhausting all options to recover overpayments, allow their
respective SWAs to remove certain types of overpayment debts from their
books as uncollectible.
For the same period, states reported waiving recovery of about $5.5
billion in nonfraudulent overpayments. DOL defines a waiver within the
regular UI program as a nonfraud overpayment for which the state
agency, in accordance with state law, relinquishes the obligation of the
claimant to repay. For the purposes of pandemic UI programs, waivers
are authorized when the overpayment was not the fault of the claimant
and requiring repayment would be against equity and good conscience or
would otherwise defeat the purpose of the UI law. 62 DOL officials told us
that states may waive recovery of nonfraudulent overpayment recoveries
for UI programs in accordance with their state law. For example, officials
from two states told us that the waiver determination process involves
supervisory review of documents, and a long-term hardship must be
established in order for an overpayment to be waived. Further, officials
from two states told us that they do not waive nonfraudulent
overpayments, although one of those states is planning to develop a
blanket waiver for overpayments made due to technical errors. 63 Table 5
identifies the types of provisions that the six selected states have in laws
regarding waiving recovery of nonfraudulent overpayments.
Table 5: Six Selected State Law Provisions for Waiving Recovery of Nonfraudulent Unemployment Insurance Overpayments
State Agency error Employer error Equity or good conscience Financial hardship
California — — —
Florida — — —
Kansas —
62Pub. L. No. 116-136, §§ 2104(f)(2), 2105(f),2107(e)(2) 134 Stat at 319-327; Pub. L. No.
116-260, div. N, tit. II, § 201(d), 134 Stat. 1182, 1952.
63SWAs may use blanket waivers for overpayments within the UI pandemic programs in
accordance with the specific scenarios set forth by DOL in UIPL No .20-21, Change 1
(Feb. 7, 2022).
Page 33 GAO-23-106696 Unemployment Insurance
Nevada — — —
New York^ — — — —
Washington — — —
Legend: (—) = legal provision not applicable; = state applies the legal provision; ^ = New York does not have an overpayment waiver of recovery provision.
Source: GAO analysis of Department of Labor and state workforce agency documents. | GAO-23-106696
We provided a draft of this report to DOL for review and comment. DOL
Agency Comments provided written comments, which are reproduced in appendix VI. It also
and Our Evaluation provided technical comments, which we incorporated as appropriate.
In its comments, DOL expressed concerns about the methodology we
used to estimate the range of UI fraud presented in the report.
Specifically, it noted that our estimate relied heavily on an analysis of
cross-matches and case records of a small sub-sample of PUA
payments. DOL stated that further analysis would be needed to determine
if a case is actually fraudulent. For this reason, DOL believes that our
range likely overestimates the level of fraud and that our estimate more
reflects an estimate of UI fraud risk, rather than UI fraud.
We disagree with DOL’s characterization of our methodology and our
estimate, which overlooks the totality of our estimation methodology. As
explained in the report and in greater detail in appendix I, our
methodology involved estimating a range by combining estimated
subpopulation fraud rates and expenditure information associated with
non-PUA and PUA programs. For the non-PUA subpopulation of UI
payments, we relied on existing estimates from the BAM program. Those
estimates are derived from thousands of payment reviews performed by
state investigators. For the PUA subpopulation, we relied on data analytic
testing along with a manual review of a statistically valid sub-sample of
PUA case files. We further validated this sampling work through the use
of econometric modeling. When calculating estimates from our PUA
sample, we used a two-sided 95-percent confidence interval, which
accounted for the uncertainty arising from our sample design and sample
size.
We agree with DOL’s comment that additional work would be required to
determine whether any given case in the sample is actually fraudulent.
Judicial and other systems would be needed to make such
determinations. Given that not all potential fraud will be investigated and
adjudicated through judicial or other systems, the full extent of UI fraud
during the pandemic will likely never be known with certainty. Therefore, it
is appropriate to rely on estimates, such as ours, to make more
Page 34 GAO-23-106696 Unemployment Insurance
comprehensive conclusions about the extent of fraud in the UI programs
during the pandemic. In presenting our estimate, we acknowledge the
inherent uncertainty associated with any estimate of fraud.
However, we disagree with DOL’s conclusion that this uncertainty means
our estimated range overstates the amount of fraud that occurred in the
UI programs during the pandemic. Given the high threshold we used for
identifying potential fraud, the risk of misidentifying nonfraudulent cases
as fraud is balanced by the counter risk of failing to identify all of the
fraudulent cases in the sample. Moreover, we took multiple steps to
reduce the risk of misidentifying nonfraudulent cases as potential fraud.
For example, except for deceased beneficiaries, we only treated
payments as fraudulent for the purpose of our estimate if multiple fraud
indicators were present. In addition, we subjected sampled payments to
multiple levels of manual review, which examined the fraud indicators in
conjunction with other available case and public information.
Finally, we do not agree that the term “fraud risk” provides a better
description of our estimated range of UI fraud. We designed our
estimated range to capture, as accurately as possible, the extent of
fraudulent activity that occurred in the UI programs, regardless of whether
that activity was previously detected or adjudicated. Our range would
have been higher if we were attempting to estimate the total dollar value
of payments that were at risk of fraud. For example, a large portion of
PUA payments have been previously identified by us and the DOL OIG
as higher risk of fraud due to the use of self-certification to determine
beneficiary eligibility.
As described, our methodology and presentation of the estimate
substantively accounts for the concerns raised by DOL. However, as
appropriate, we have incorporated clarifying language in this report.
We also provided a draft of this report to officials from the DOL OIG for
review. DOL OIG provided technical comments, which we incorporated as
appropriate.
In addition, selected excerpts of the draft report were provided to officials
from the Alaska, Arizona, California, Colorado, Florida, Georgia, Iowa,
Illinois, Kansas, Massachusetts, Michigan, Montana, Nevada, New
Jersey, New York, Oregon, Pennsylvania, Rhode Island, South Carolina,
Texas, Vermont, and Washington SWAs for review. We made technical
corrections or clarifications as needed based on the comments we
received from four SWAs. Nine SWAs indicated that they did not have
comments and nine SWAs did not respond.
Page 35 GAO-23-106696 Unemployment Insurance
We are sending copies of this report to the appropriate congressional
committees, the Acting Secretary of the Department of Labor, and other
interested parties. In addition, the report is available at no charge on the
GAO website at https://www.gao.gov.
If you or your staff have any questions about this report, please contact
Seto Bagdoyan, (202) 512-6722, BagdoyanS@gao.gov or Jared Smith,
(202) 512-2700, SmithJB@gao.gov. Contact points for our Offices of
Congressional Relations and Public Affairs may be found on the last page
of this report. GAO staff who made key contributions to this report are
listed in appendix VII.
Seto J. Bagdoyan
Director, Forensic Audits and Investigative Service
Jared B. Smith
Director, Applied Research and Methods
Page 36 GAO-23-106696 Unemployment Insurance
Appendix I: Detailed Information on the
Appendix I: Detailed Information on the
Methodology GAO Used to Estimate Fraud in
Methodology GAO Used to Estimate Fraud
Unemployment Insurance (UI) Programs
during the Pandemic
in Unemployment Insurance (UI) Programs
during the Pandemic
To develop an estimate of fraud (lower and upper range) within UI
programs during the COVID-19 pandemic, we combined separate
estimates of fraudulent payments associated with
• the regular UI program, Pandemic Emergency Unemployment
Compensation (PEUC), Mixed Earner Unemployment Compensation
(MEUC), Extended Benefits, and the portion of Federal Pandemic
Unemployment Compensation (FPUC) payments that were not
associated with Pandemic Unemployment Assistance (PUA) claims; 1
and
• the PUA program, including FPUC payments associated with PUA
claims. 2
1We refer to the UI program—excluding both the temporary UI programs created by the
Coronavirus Aid, Relief, and Economic (CARES) Act and other legislation, as well as the
Extended Benefits program—as the regular UI program and the benefits paid under the
program as regular UI benefits. Regular UI benefits are benefits paid by the state under
state UI law, Unemployment Compensation for Federal Employees, and Unemployment
Compensation for Ex-Service Members programs. The Extended Benefits program, which
existed prior to the pandemic, provides up to 13 or 20 additional weeks of benefits when a
state is experiencing specific levels of high unemployment. We estimate that 37 percent of
FPUC payments were made on top of PUA claims. FPUC benefits are additional
payments made on top of existing regular UI or pandemic UI program claims. The amount
of the FPUC payment changed throughout the pandemic. We estimated the percent of
FPUC attributable to PUA claims by multiplying the number of weekly PUA payments
made in a given month by an approximation of the FPUC amount that was relevant for the
period. The FPUC amount is an approximation because a claim paid in a given month
may be for the benefit from a previous period. For example, a weekly claim paid in April,
when the FPUC weekly benefit was $600, might be for a week in March, when the FPUC
weekly benefit amount was $0. To test the accuracy of our approach, we compared our
results with breakdowns provided by seven state workforce agencies (SWA). Our estimate
was close to the percentages reported by the states in all seven cases.
2For fiscal years 2021 and 2022 improper payment reporting, the Department of Labor
(DOL) applied the estimated improper payment rate from the Benefit Accuracy
Measurement (BAM) program testing of regular UI claims to calculate the estimated
improper payment amounts for FPUC and PEUC. Thus, the estimated improper payment
amounts for these two programs were incorporated into the overall UI estimated improper
payment amount reported for fiscal years 2021 and 2022. However, this overall estimated
improper payment amount for UI did not include an estimate for PUA. According to DOL, it
did not include PUA in the extrapolation of the BAM estimated improper payment rate
because the PUA program served a different population of workers and had different
eligibility requirements. DOL did not estimate improper payments for the MEUC program,
according to officials, because the program only operated between January and
September 2021. Officials explained that in accordance with Office of Management and
Budget (OMB) guidance, DOL is not required to estimate or report improper payments for
this program because it existed for less than one year. The total federal expenditure for
the MEUC program was $78 million through May 31, 2023. We included MEUC in our
estimate.
Page 37 GAO-23-106696 Unemployment Insurance
Appendix I: Detailed Information on the
Methodology GAO Used to Estimate Fraud in
Unemployment Insurance (UI) Programs
during the Pandemic
The scope of our effort to develop an estimate of fraud within UI programs
was from April 2020—the first full month of pandemic UI program
payments—through May 2023—the end of the COVID-19 public health
emergency. 3 We estimated the extent of fraud across all 53 state
workforce agencies (SWA) across the regular UI and pandemic UI
programs. 4 Throughout this report, we use the phrase “fraud estimate” or
“estimate of fraud” to refer to estimates that attempt to quantify the extent
of fraud, regardless of whether such fraud has already been detected and
adjudicated. For the purpose of estimation, additional uncertainty arises in
situations like the current one where the sampled cases have not yet
been adjudicated. We used multiple data sources, review steps, and
estimation procedures to reduce this uncertainty, but we cannot eliminate
it entirely.
As part of our work to calculate this estimate, we separated UI
expenditures by whether the expenditures were associated with the PUA
program, which had a unique fraud risk profile. 5 Given the time frame of
this review, we were not able to obtain sufficient evidence about the PUA
program to report a separate statistical estimate for that program.
Instead, we designed our procedures such that when the total evidence of
PUA and non-PUA payments was considered together, the combined
evidence was sufficient to support an overall estimate of the extent of
fraud in the UI programs during our period of review. 6
3The date range we used to develop the estimate of fraud within UI programs during the
pandemic is different compared with the date range we used to determine the financial
and technical assistance and the extent that states have recovered, written off, and
waived overpayments because, for the estimation of fraud, we used the first full month of
pandemic UI program payments rather than the date when the pandemic began.
4Fifty-three SWAs administer UI programs across the 50 states, the District of Columbia,
Puerto Rico, and the U.S. Virgin Islands.
5The DOL Office of Inspector General (OIG) reported in October 2020 that the PUA
program in particular was at high risk for fraud due to its unique program rules and
eligibility requirements. We developed separate procedures to calculate the estimate for
the PUA program because of the program’s unique fraud risk profile. Department of Labor,
Office of Inspector General, COVID-19: States Cite Vulnerabilities in Detecting Fraud
While Complying with the CARES Act UI Program Self-Certification Requirement, Report
No. 19-21-001-03-315 (Washington, DC: Oct. 21, 2020.)
6In this context, sufficiency depends on the precision of the estimate. The precision of our
overall estimate is captured by the width of our reported range, which accounts for the
statistical uncertainty associated with both the PUA and non-PUA payments at the 95
percent confidence level. We do not report our range at the 95 percent confidence level
because statistical intervals do not capture the uncertainty associated with identifying
which cases in the sample were fraudulent. To reduce this latter source of uncertainty, we
leveraged multiple data sources and review procedures.
Page 38 GAO-23-106696 Unemployment Insurance
Appendix I: Detailed Information on the
Methodology GAO Used to Estimate Fraud in
Unemployment Insurance (UI) Programs
during the Pandemic
Figure 7 describes key steps we took to derive the estimated fraud in UI
programs during the pandemic.
Page 39 GAO-23-106696 Unemployment Insurance
Appendix I: Detailed Information on the
Methodology GAO Used to Estimate Fraud in
Unemployment Insurance (UI) Programs
during the Pandemic
Figure 7: Selected Steps Taken to Derive the Estimated Fraud in Unemployment Insurance (UI) Programs during the
Pandemic
Page 40 GAO-23-106696 Unemployment Insurance
Appendix I: Detailed Information on the
Methodology GAO Used to Estimate Fraud in
Unemployment Insurance (UI) Programs
during the Pandemic
For the regular UI program, we used data from DOL’s Benefit Accuracy
Measurement (BAM) program—which DOL uses to estimate the amount
and rate of improper payments, including those caused by fraud—from
April 2020 through December 2022. Specifically, we compiled the
quarterly and yearly BAM estimated fraud rates and fraud total estimates
in the regular UI program by state.
We interviewed officials from 14 SWAs to obtain information about the
operation of the pandemic UI programs, potential limitations of the BAM
program, state definitions of fraud, and efforts to measure fraud in the
pandemic UI programs. We selected the 14 SWAs—Arizona, California,
Florida, Georgia, Illinois, Kansas, Michigan, Nevada, New Jersey, New
York, Pennsylvania, Rhode Island, Texas, and Washington—to reflect a
variety of BAM program fraud rates and state population sizes. 7
Information obtained through interviews was used to help assess the
reliability of the estimated fraud rate generated from the BAM program
and to better understand factors that SWAs identified as being associated
with fraud in both the pandemic and regular UI programs. The fraud rates
calculated as part of the BAM program depend on the state definitions of
fraud, which differ from state to state.
To determine the applicability of the BAM program fraud estimate, we
reviewed the scope of the BAM program. The BAM program includes 52
SWAs and the three major permanent state UI programs. 8 Our
engagement scope also includes the U.S. Virgin Islands, even though it is
exempt from operating a BAM program. Given the small size of the U.S.
Virgin Islands, the impact of any deviation between the BAM program
fraud rate and the U.S. Virgin Islands fraud rate would not be material to
our total fraud estimate. Our scope also includes PEUC, FPUC, Extended
Benefits, and MEUC payments, which are not reviewed by the BAM
program. PEUC and Extended Benefits programs provide additional
7These 14 states were selected based on different criteria compared with the six SWAs
that were selected for interviews to obtain information related to the assistance provided.
As previously mentioned, we selected the six SWAs—California, Florida, Kansas, Nevada,
New York, and Washington—based on a range of (1) the fraud risk level identified in our
first objective, (2) the amount of grant funding received, and (3) the acceptance of DOL’s
offer of financial and technical assistance.
8The three major UI programs covered by BAM are state UI, Unemployment
Compensation for Federal Employees, and Unemployment Compensation for Ex-Service
Members. We downloaded quarterly BAM data, including the estimated fraud rate, from
DOL’s website on September 13, 2022; November 1, 2022; and July 10, 2023.
https://www.dol.gov/agencies/eta/unemployment-insurance-payment-accuracy/data. The
BAM program estimated fraud rates are generated for all 50 states, the District of
Columbia, and Puerto Rico. According to DOL’s guidance, the U.S. Virgin Islands is
exempt from operating a BAM program.
Page 41 GAO-23-106696 Unemployment Insurance
Appendix I: Detailed Information on the
Methodology GAO Used to Estimate Fraud in
Unemployment Insurance (UI) Programs
during the Pandemic
weeks of benefits to claimants who were receiving regular UI benefits.
Given the close relationship between these programs and the regular UI
program, we assume that the BAM program fraud rate is a reasonable
approximation for the fraud rate in the PEUC and Extended Benefits
programs. We make the same assumption for the portion of the FPUC
payments that were not attributable to the PUA program. Payments for
the MEUC program make up less than 0.1 percent of UI program
expenditures and so, any differences between the fraud rate estimated
using the BAM program and the actual MEUC fraud rate would not have a
material impact on our overall results.
Another difference between our scope and available data is that the BAM
program was suspended from April 2020 through June 2020. Further, at
the time we performed our analysis, BAM data were not available beyond
December 2022. Therefore, we used a statistical model to impute the
fraud rate for the months where BAM was suspended. 9 We then
approximated the BAM program fraud rate from January 2023 to May
2023 using the overlapping annual fraud rates reported yearly by DOL
through December 2022.
We also reviewed how BAM program fraud rates were estimated and the
limitations in BAM program reporting. Since BAM estimates are derived
from statistical samples, the actual rate is expected to lie within 95
percent of the intervals constructed from repeated samples of the same
size and selected in the same manner as the BAM sample. In addition to
sampling uncertainty, the estimate may be impacted by nonsampling
error. One of the limitations is that the BAM program may not cover all
potential types of fraud. For example, one state reported that the BAM
program may be less effective at detecting employer and employee
collusion. Further, states did not always update their BAM program fraud
determinations, given subsequent conflicting final adjudications. 10 From
our discussions with SWAs about these sources of uncertainty, we
determined that the risk of the BAM program fraud rate being understated
due to being unable to detect certain fraud types was higher than the risk
of the BAM program fraud rate being overstated due to the differences
between BAM investigators’ findings and subsequent adjudications.
9We imputed the fraud rate for the quarter where the BAM program was suspended using
DOL’s improper payment reporting quarterly data from the first quarter of 2017 through the
second quarter of 2021. We developed a linear regression model using explanatory
variables including 52 state dummy variables, state population adjusted claim count at
each quarter, and whether a quarter was pre-pandemic or not.
10For example, one state reported that the appeals and resulting adjudications may not be
completed in time for BAM program reporting.
Page 42 GAO-23-106696 Unemployment Insurance
Appendix I: Detailed Information on the
Methodology GAO Used to Estimate Fraud in
Unemployment Insurance (UI) Programs
during the Pandemic
For the PUA program, we selected a generalizable sample of 260 PUA
payments to identify the presence of fraud indicators associated with
identity theft and eligibility fraud. 11 The PUA payments we selected were
a subsample of DOL’s sample of 2,540 PUA payments. 12 We selected the
sample of 260 PUA payments from a stratified sample of 14 states—
Alaska, California, Colorado, Georgia, Iowa, Illinois, Massachusetts,
Montana, New York, Oregon, Pennsylvania, South Carolina, Texas, and
Vermont. 13 More details on the sample specification are provided below:
• We stratified the top 10 states which consumed approximately 75
percent of PUA program outlays into two strata—three states with
the top outlays (approximately 50 percent of entire spending) from
which we selected all three (stratum 1) and a random sample of
three states from the rest of seven states (stratum 2). We also
randomly selected eight states from 16 remaining states (stratum
3).
• For states in stratum 1, we randomly selected eight out of 19
weeks and sampled five PUA payments per week. For states in
stratum 2, we randomly selected four out of 19 weeks and
11Fraud indicators are characteristics and flags that serve as warning signs suggesting
potential for fraudulent activity. Fraud indicators can be used to identify potential fraud and
assess fraud risk. They are not proof of fraud. Additional review, investigation, and
adjudication is needed to determine if fraud exists. To that end, we will refer claimants with
presence of fraud indicators we identified to the DOL OIG for further review and
investigation.
12DOL selected this sample of payments to review to estimate the PUA improper payment
rate. In August 2023, DOL released its estimate of improper payments made from March
2020 to September 2021 under the PUA program, concluding that the PUA program had a
total estimated improper payment rate of 35.9 percent. DOL noted that its analysis
focused on the broader universe of improper payments, does not isolate fraud, and should
not be considered a fraud estimate for the PUA program.
13We selected a stratified random sample of states based on PUA expenditure using
DOL’s original sampling scheme and allocated a random sample of 260 PUA payments
across the selected states. These 14 states were selected based on different criteria
compared with the 14 SWAs that were selected for interviews to obtain information about
the operation of the pandemic UI programs, among other things. As previously mentioned,
we selected the 14 SWAs—Arizona, California, Florida, Georgia, Illinois, Kansas,
Michigan, Nevada, New Jersey, New York, Pennsylvania, Rhode Island, Texas, and
Washington—based on population sizes and BAM-estimated regular UI program fraud
rates. Further, these 14 states were selected based on different criteria compared with the
six SWAs that were selected for interview to obtain information related to the assistance
provided. As previously mentioned, we selected the six SWAs—California, Florida,
Kansas, Nevada, New York, and Washington—based on a range of (1) the fraud risk level
identified in our first objective, (2) the amount of grant funding received, and (3) the
acceptance of DOL’s offer of financial and technical assistance.
Page 43 GAO-23-106696 Unemployment Insurance
Appendix I: Detailed Information on the
Methodology GAO Used to Estimate Fraud in
Unemployment Insurance (UI) Programs
during the Pandemic
sampled five PUA payments per week. Finally, we randomly
selected 10 payments from each state in stratum 3.
Methods for identifying the presence of fraud indicators for PUA
payments included the following steps. 14
• The DOL Office of Inspector General (OIG) used data analytic
procedures to identify the presence of fraud indicators for the DOL
sample of 2,540 payments and provided them to us. The 18 indicators
included multistate claims and shared or suspicious emails, among
other indicators.
• To identify additional fraud indicators from the DOL sample of 2,540
PUA payments, we conducted data matching to the Death Master File
(DMF) to identify potentially deceased individuals. 15 We also
conducted data matching to the National Directory of New Hires
(NDNH) for quarter 1 of calendar year 2019 through quarter 3 of
calendar year 2020 and quarter 1 of calendar year 2021 through
quarter 3 of calendar year 2021 to identify claimants’ unreported
wages. 16
• For our sample of 260 payments, we then manually reviewed the
programmatically generated flags and updated these flags when
necessary. The manual review included a review of state case files
14In this report, we do not detail all fraud indicators we identified so that potential
perpetrators of fraud do not become aware of fraud risks or exploit potential weaknesses
in the program.
15The Social Security Administration (SSA) DMF identifies Social Security number (SSN)
holders who are deceased. SSA maintains death data, including names, Social Security
numbers, date of birth, and date of death. SSA shares a comprehensive file of this death
information, which includes state death data, with certain eligible entities, including SWAs.
We used this comprehensive file, which we will call the “full death master file,” for our
analysis. A subset of the full death master file that does not include state death data is
available to the public.
16We did not obtain unemployment data for quarter 4 of 2020 because at the time we
requested the information, the data for that period were no longer available. NDNH is a
national repository of new hire, quarterly wage, and unemployment insurance information
reported by employers, states, and federal agencies. NDNH is maintained and used by
the U.S. Department of Health and Human Services for the federal child support
enforcement program, which assists states in locating parents and enforcing child support
orders. DOL does not have access to NDNH wage data; however, SWAs have access to
NDNH wage data.
Page 44 GAO-23-106696 Unemployment Insurance
Appendix I: Detailed Information on the
Methodology GAO Used to Estimate Fraud in
Unemployment Insurance (UI) Programs
during the Pandemic
and relevant publicly available information. 17 We also performed a
cross-match of information in the case files against the Social Security
Administration’s Enumeration Verification System (EVS). 18 For a
subset of claims showing indicators of potential identity theft, we met
with the DOL OIG to understand the related investigative data
pertinent to those claims.
• We aggregated the manually adjusted fraud flags to generate
consolidated fraud risk scores for each of the 260 PUA payments in
our sample. 19 The consolidated risk score was an overall assessment
of fraud risk for the sampled payment.
The above steps resulted in manually adjusted fraud risk scores for the
sample of 260 PUA payments and programmatically generated flags
indicating the presence of fraud indicators for the DOL sample of 2,540.
We performed multiple imputation using a fully conditional specification
method that uses observed data for fraud indicators and manually scored
fraud risk to impute fraud risk for the remaining DOL sample. The model
was developed as part of a multiple imputation procedure to populate
predicted manual fraud risk scores for the remaining portion of the 2,540
PUA payments. The multiple imputation step produced multiple versions
of the data, where each version contains a different set of imputed values.
The multiple imputation step helped account for the uncertainty arising
from the modeling procedure. We then used the DOL sample design and
sampling weights to calculate a national PUA fraud rate estimate, given
each imputed dataset. The PUA range was calculated using Rubin’s Rule
for multiple imputation, which accounts for the variability of each
individual estimate and the variability across the imputed estimates. 20
17Fraud indicators may sometimes be explained by events other than fraud. An important
goal of the manual review was to help account for alternative explanations of the observed
fraud indicators. For example, an address may have a large number of claims because it
is a multiunit dwelling and so, when assessing fraud risk associated with individual
addresses, we examined the size of the dwelling and whether it was multiunit.
18EVS provides information on invalid (never issued) SSNs and instances where there are
mismatches between SSN, name, and date of birth. EVS flags SSNs in which the name or
date of birth (or both) do not match its records for the SSN, as well as SSNs that have
never been issued by the SSA.
19In this report, we do not detail all the steps of our fraud scoring process so that potential
perpetrators of fraud do not become aware of fraud risks or exploit potential weaknesses
in the program.
20Rubin’s Rule is an approach to create pooled estimates from the results of multiple
imputation procedures. Multiple imputation is a tool for replacing missing data with multiple
plausible values, with the goal of accounting for the uncertainty arising from the procedure
used to perform the replacement.
Page 45 GAO-23-106696 Unemployment Insurance
Appendix I: Detailed Information on the
Methodology GAO Used to Estimate Fraud in
Unemployment Insurance (UI) Programs
during the Pandemic
We obtained additional evidence regarding the extent of fraud in the PUA
program by conducting an econometric analysis of PUA benefit payments
over time from March 2020 to December 2021. 21 Additional details about
the econometric analysis follow:
• The model relied on the expectation that the PUA benefit payment
should be associated with certain states’ conditions and the fraud
prevention tools or processes that states implemented.
• To obtain basic information about the operation of the PUA program,
including the fraud prevention processes or tools implemented by the
states, we surveyed the SWAs in 50 states and the District of
Columbia. 22 We also interviewed 14 SWAs to obtain information about
states’ PUA application process, job search requirement, and
implementation of identity verification services. We also obtained data
for a broad range of state-level conditions, such as disease burden
during the pandemic, unemployment rates, demographic composition,
and industrial composition. 23
• We developed an econometric model to predict the level of PUA
benefit payment if all states were comprehensively implementing
fraud prevention tools or processes. We used this approach to
estimate potential fraud that would have been prevented or deterred if
all states at all times had implemented all fraud prevention tools and
processes that were ultimately in use across states. We used the
survey data we collected from 48 states as a proxy for the extent of
21This period provides sufficient coverage of the PUA program, given that the program
started in March 2020 and expired by September 2021. Some states opted out of the PUA
program prior to its expiration date. States were required to accept new PUA applications
30 days after the state termination or program expiration (whichever comes first). We
selected December 2021 to provide two additional months of coverage.
22Three SWAs did not respond to our survey by the time deadline. The econometric
modeling is limited by the 48 states we received survey responses from because the
model utilizes the fraud prevention tools and processes from the survey.
23Other state-level conditions included in the econometric model include a COVID-19
stringency index (i.e., a measure of the strictness of states’ closure and containment
policies that primarily restrict people’s behavior), regular state UI features (e.g., insured
unemployment rate, percentage with insufficient wage credit, and percentage of single-
claimant denials), labor market conditions (e.g., part-time worker for economic reasons,
percentage of self-employed workers, and reason for unemployment), occupational
composition (i.e., 26 occupation groups), and lagged variables (e.g., lagged disease
burden during the pandemic, and a lagged stringency index). We estimated the
econometric model by pooled ordinary least squares with standard errors clustered by
state.
Page 46 GAO-23-106696 Unemployment Insurance
Appendix I: Detailed Information on the
Methodology GAO Used to Estimate Fraud in
Unemployment Insurance (UI) Programs
during the Pandemic
states’ fraud prevention efforts. We used other state-level conditions
as control variables in our econometric model.
• We took two approaches to estimate the potential fraud amount in the
PUA benefit payment. Specifically, we estimated the potential fraud
amount in benefit payment (1) by the total differences between actual
benefit payments and predicted benefit payments from the
econometric model and (2) from the average effect of comprehensive
fraud prevention tools or processes (i.e., the estimated coefficient
from the econometric model). The potential fraud rate is calculated as
the ratio of potential fraud amount to actual benefit payment.
• The resulting model did not directly estimate the amount of fraud in
the PUA program. Instead, it approximated the fraud total as it was
reduced by state controls. As a result, this model does not cover fraud
that would be undetected and undeterred by environments with a
higher level of fraud controls. Conversely, the model may identify as
fraud a reduction in legitimate claims activity due to increased state
controls. In addition, while our model took into account a variety of
factors that may be correlated with PUA benefit payment, such as the
COVID-19 excess death rate, the COVID-19 stringency index, the
unemployment rate, and states’ fraud prevention efforts, we may not
have taken into account all possible factors. To account for these
limitations, we used the econometric model in conjunction with the
results obtained from our statistical sampling of PUA payments. The
two approaches provided generally consistent results on the scale of
fraud in the PUA program.
We generated the upper and the lower range of estimated fraud across
the regular UI program, FPUC, PEUC, MEUC, Extended Benefits, and
PUA using the approaches described above. The range obtained from the
BAM program fraud estimate was applied to regular UI, PEUC, Extended
Benefits, MEUC, and the estimated portion of FPUC that was not
attributable to PUA. 24 The PUA range, which we obtained from the PUA
sample and checked against the econometric modeling, was applied to
PUA payments and the portion of FPUC payments that arose from the
PUA payments. We combined the component estimates by summing the
lower limit of each component to create the lower end of the overall range
24FPUC payments were made in addition to payments made on an existing claim. DOL
reported that it could not calculate the portion of FPUC payments that were made for PUA
claims. We asked the 14 SWAs we interviewed to provide the FPUC percentages, and we
received seven written responses that referred to expenditure data. We validated our
estimation procedure by comparing our state-level results with the percentage reported by
those states.
Page 47 GAO-23-106696 Unemployment Insurance
Appendix I: Detailed Information on the
Methodology GAO Used to Estimate Fraud in
Unemployment Insurance (UI) Programs
during the Pandemic
and summing the upper limit of each component to create the upper end
of the overall range.
Our estimate includes different fraud types, including claimants
intentionally misrepresenting themselves as other individuals and
claimants intentionally misstating facts related to their eligibility. We
attempted to use a diverse set of methods in order to cover a substantial
portion of fraud types identified by the 14 states we interviewed. However,
given both the uncertainty associated with the hidden nature of fraud and
the resource limitations on the entities that investigate and adjudicate
fraudulent claims, it was not possible to identify and cover all potential
fraud schemes and types. We also could not eliminate the possibility that
some of the sampled cases that were identified as potential fraud may
have involved nonfraudulent overpayments or have been properly paid.
Due to these limitations, we do not provide a statistical confidence level
when reporting our likely fraud range.
Further, we requested, identified, and reviewed relevant reports from
state entities as of March 2023 related to estimating the extent of fraud
and potential fraud in UI programs during the pandemic to gain
information about fraud estimates reported at the state level. 25
We assessed the reliability of the DOL Employment and Training
Administration data, BAM program fraud estimates, DOL’s sample of PUA
payments, and DOL OIG fraud indicators on PUA payments by (1)
reviewing information about the data and the system that produced them;
(2) interviewing officials knowledgeable about the data, when feasible; (3)
performing electronic testing, when feasible; and (4) tracing information to
source documents, when feasible. We assessed the reliability of EVS,
DMF, and NDNH data by reviewing relevant information about the data
and performing electronic testing, when feasible. We determined that the
data were sufficiently reliable for the purposes of responding to our
objective.
25Throughout this report, we refer to estimates as projections or inferences based on
measures, assumptions, or analytical techniques. Estimates are often used when direct
measures are unavailable, incomplete, or unreliable.
Page 48 GAO-23-106696 Unemployment Insurance
Appendix II: GAO Unemployment Insurance-
Appendix II: GAO Unemployment Insurance-
Related Recommendations to the Department
Related Recommendations to the
of Labor
Department of Labor
Table 6 lists GAO’s 26 recommendations made since 2018 to the
Department of Labor (DOL) to help improve the Unemployment Insurance
(UI) system and their implementation status.
Table 6: GAO’s 26 Recommendations to the Department of Labor (DOL) to improve the Unemployment Insurance (UI) System,
Status as of August 2023
No. Status Report number, date Recommendation to DOL
1 Closed – GAO-22-105051, (priority)c The Secretary of Labor should examine the suitability of existing
implementeda October 27, 2021 fraud controls in the UI program and prioritize residual fraud risks.
2 GAO-22-104251, The Secretary of Labor should ensure that the Office of Unemployment
June 7, 2022 Insurance review the customer service challenges that states faced during
the pandemic, identify comprehensive information on customer service best
practices, and provide states with this information to assist them in improving
service delivery.
3 GAO-21-387, March The Secretary of Labor should ensure that the Office of Unemployment
31, 2021 Insurance collects data from states on the amount of overpayments waived in
the PUA program, similar to the regular UI program.
4 GAO-22-105051, The Secretary of Labor should identify inherent fraud risks facing the UI
October 27, 2021 program.
5 GAO-22-105051, The Secretary of Labor should assess the likelihood and impact of inherent
October 27, 2021 fraud risks facing the UI program.
6 GAO-22-105051, The Secretary of Labor should document the fraud risk profile for the UI
October 27, 2021 program.
7 GAO-21-265, January The Secretary of Labor should ensure that the Office of Unemployment
28, 2021 Insurance collects data from states on the amount of overpayments
recovered in the Pandemic Unemployment Assistance (PUA) program,
similar to the regular UI program.
8 GAO-21-191, The Secretary of Labor should ensure that the Office of Unemployment
November 30, 2020 Insurance revises its weekly news releases to clarify that in the current
unemployment environments, the numbers it reports for weeks of
unemployment claimed do not accurately estimate the number of unique
individuals claiming benefits.
9 GAO-18-633, The Secretary of Labor should systematically collect sufficient information on
September 4, 2018 state profiling systems, possibly through DOL’s new UI state self-assessment
process, to identify states at risk of poor profiling system performance. For
instance, DOL could collect information on challenges that states have
experienced using and maintaining their profiling systems, planned changes
to the systems, or state processes for assessing the systems’ performance.
10 GAO-18-633, The Secretary of Labor should develop a process to use information on state
September 4, 2018 risks of poor profiling system performance to provide technical assistance to
states that need to improve their systems. DOL may also wish to tailor its
technical assistance based on state service delivery goals and technical
capacity.
11 Open – partially GAO-21-191, (priority) The Secretary of Labor should ensure that the Office of
addressedb November 30, 2020 Unemployment Insurance pursues options to report the actual number of
distinct individuals claiming benefits, such as by collecting these already
available data from states, starting from January 2020 onward.
Page 49 GAO-23-106696 Unemployment Insurance
Appendix II: GAO Unemployment Insurance-
Related Recommendations to the Department
of Labor
12 GAO-23-105523, The Secretary of Labor should design and implement an antifraud strategy for
December 22, 2022 UI based on a fraud risk profile consistent with leading practices as provided
in the Fraud Risk Framework.
13 GAO-22-105051, The Secretary of Labor should designate a dedicated entity and document its
October 27, 2021 responsibilities for managing the process of assessing fraud risks to the UI
program, consistent with leading practices as provided in our Fraud Risk
Framework. This entity should have, among other things, clearly defined and
documented responsibilities and authority for managing fraud risk
assessments and for facilitating communication among stakeholders
regarding fraud-related issues.
14 Open – not GAO-22-104438, (priority) The Secretary of Labor should ensure that the Office of
addressedd June 7, 2022 Unemployment Insurance examines and publicly reports on the extent and
potential causes of racial and ethnic inequities in the receipt of PUA benefits,
as part of the agency’s efforts to modernize UI and improve equity in the
system. The report should also address whether there is a need to examine
racial, ethnic, or other inequities in regular UI benefit receipt, based on the
PUA findings.
15 GAO-18-486, August (priority) The Assistant Secretary of DOL’s Employment and Training
22, 2018 Administration should provide states with information about its determination
that the use of state formal warning policies is no longer permissible under
federal law.
16 GAO-18-486, August (priority) The Assistant Secretary of DOL’s Employment and Training
22, 2018 Administration should clarify information on work search verification
requirements in its revised Benefit Accuracy Measurement procedures. The
revised procedures should include an explanation of what DOL considers to
be sufficient verification of claimants’ work search activities.
17 GAO-23-105478, July The Secretary of the Department of Labor should direct the Office of
2023 Unemployment Insurance Modernization and the Office of the Chief
Information Officer to update their processes for UI pilots to reflect leading
practices for pilot design and implement the leading pilot design practices that
address the weaknesses that we identified on its future pilots.
18 GAO-23-105478, July The Secretary of the Department of Labor should direct the Office of
2023 Unemployment Insurance to define UI IT modernization standards for states.
19 GAO-23-105478, July The Secretary of the Department of Labor should direct the Office of
2023 Unemployment Insurance to measure states’ UI IT performance against
established standards.
20 GAO-22-105162, The Secretary of Labor should develop and execute a transformation plan
June 7, 2022 that meets GAO’s high-risk criteria for transformation; the plan should outline
coordinated and sustained actions to address known issues related to
providing effective service and mitigating financial risk, including ways to
demonstrate improvements. Planned actions may include addressing audit
recommendations and determining whether legislative changes are needed,
as appropriate. Planned actions may also include achieving quantifiable
results in reducing improper payment rates, including those related to fraud;
improving efficiency in claims processing and restoring prepandemic payment
timeliness levels; better reaching current worker populations; and enhancing
equity in benefit distribution.
Page 50 GAO-23-106696 Unemployment Insurance
Appendix II: GAO Unemployment Insurance-
Related Recommendations to the Department
of Labor
21 GAO-22-104438, The Secretary of Labor should study and advise the Congress and other
June 7, 2022 policymakers on the costs, benefits, and risks of various options to
systematically support self-employed and contingent workers during periods
of involuntary unemployment outside of declared disasters, including
considering options’ feasibility and approach to fraud prevention.
22 GAO-22-104251, The Secretary of Labor should ensure that the Office of Unemployment
June 7, 2022 Insurance assesses lessons learned from the pandemic to inform its future
disaster responses efforts and support the Congress on ways to address
future emergencies.
23 GAO-22-105051, The Secretary of Labor should determine fraud risk tolerance for the UI
October 27, 2021 program.
24 GAO-18-633, The Secretary of Labor should update agency guidelines to ensure that it
September 4, 2018 clearly informs states about the range of allowable profiling approaches.
25 GAO-18-486, August The Assistant Secretary of DOL’s Employment and Training Administration
22, 2018 should monitor states’ efforts to discontinue the use of formal warning
policies.
26 GAO-18-486, August The Assistant Secretary of DOL’s Employment and Training Administration
22, 2018 should monitor states’ compliance with the clarified work search verification
requirements.
Source: GAO analysis of open recommendations to DOL. | GAO-23-106696
aRecommendations that have been closed as implemented are those where the agency has
completed all action(s) to implement the recommendation or the intent of the recommendation.
bRecommendations that have been partially addressed are those where the agency has completed
action(s) that contribute to the full implementation of the recommendation, but some actions remain
outstanding.
cPriority recommendations are those that GAO believes warrant priority attention from heads of key
departments or agencies. They are highlighted because, upon implementation, they may significantly
improve government operations, for example, by realizing large dollar savings; eliminating
mismanagement, fraud, and abuse; or making progress toward addressing a high risk or
fragmentation, overlap, or duplication issue.
dRecommendations that have not been addressed are those where the agency has yet to take any
action(s) to implement the recommendation.
Page 51 GAO-23-106696 Unemployment Insurance
Appendix III: Financial Assistance Allocated
Appendix III: Financial Assistance Allocated
and Awarded to States
and Awarded to States
Table 7 lists the total amount of grants that the Department of Labor
(DOL) allocated and awarded to states for initiatives including fraud
prevention, detection, investigation, and recovery activities in the
unemployment insurance (UI) programs as of May 2023. These totals
include both Coronavirus Aid, Relief, and Economic Security (CARES)
Act and American Rescue Plan Act (ARPA) funding that was used to
address fraud in all UI programs. 1 The total amount allocated represents
the maximum amount of funds available for states to apply for in each of
these grants. The total amount awarded represents the amount of funds
that DOL approved for states to receive after the application process.
In June 2023, the Fiscal Responsibility Act of 2023 (FRA) was signed into
law. This law rescinded total ARPA funding for UI programs that had not
been awarded and reduced the total ARPA funding for UI programs from
$2 billion to $1 billion. In July 2023, DOL announced updated financial
assistance amounts for states, which reflected the FRA rescission.
Table 7: Coronavirus Aid, Relief, and Economic Security Act and American Rescue Plan Act Financial Assistance Dollar
Amounts Allocated (as of July 2023) and Awarded (as of May 2023) by the Department of Labor to States and U.S. Territories
State Total amount allocated Total amount awarded
(in dollars) (in dollars)
Alabama 19,761,600 16,390,322
Alaska 11,038,750 6,238,991
American Samoa 500,000 300,000
Arizona 33,023,350 29,182,281
Arkansas 16,663,800 10,226,684
California 51,533,850 40,911,465
Colorado 30,091,350 26,127,555
Commonwealth of the Northern Mariana Islands 1,050,000 1,026,060
Connecticut 19,628,600 15,794,849
Delaware 11,274,000 10,319,842
District of Columbia 9,593,000 8,634,624
Federated States of Micronesia 500,000 200,000
1Pub. L. No. 116-136, 134 Stat. 281 (2020); Pub. L. No. 117-2, 135 Stat. 4 (2021).
Page 52 GAO-23-106696 Unemployment Insurance
Appendix III: Financial Assistance Allocated
and Awarded to States
Florida 32,494,550 23,795,789
Georgia 29,929,350 15,726,189
Guam 1,050,000 1,026,060
Hawaii 11,407,750 8,845,328
Idaho 11,130,750 8,608,747
Illinois 31,208,350 25,094,049
Indiana 29,621,350 22,206,376
Iowa 19,353,600 13,327,338
Kansas 20,202,800 18,308,817
Kentucky 20,689,600 16,947,071
Louisiana 17,591,800 11,084,688
Maine 14,279,987 9,535,575
Maryland 30,197,350 22,818,913
Massachusetts 27,283,800 11,083,800
Michigan 33,037,350 27,823,810
Minnesota 24,148,800 4,860,000
Mississippi 19,960,600 15,068,629
Missouri 28,891,350 17,923,800
Montana 9,564,000 7,155,992
Nebraska 11,052,750 9,523,254
Nevada 21,941,600 19,908,013
New Hampshire 11,382,750 9,718,695
New Jersey 27,489,800 24,517,615
New Mexico 14,333,980 11,833,778
New York 38,573,550 34,315,180
North Carolina 29,969,350 17,923,800
North Dakota 9,544,000 3,862,417
Page 53 GAO-23-106696 Unemployment Insurance
Appendix III: Financial Assistance Allocated
and Awarded to States
Ohio 31,892,350 24,417,298
Oklahoma 19,598,639 14,740,241
Oregon 22,841,600 19,411,467
Palau 500,000 194,300
Pennsylvania 37,496,350 31,206,763
Puerto Rico 10,939,000 9,905,996
Republic of Marshall Islands 500,000 100,000
Rhode Island 11,879,750 9,255,707
South Carolina 18,867,800 16,103,076
South Dakota 9,408,000 8,460,169
Tennessee 24,552,800 14,874,660
Texas 34,001,550 29,994,524
Utah 16,136,800 11,550,670
Vermont 9,566,000 4,803,745
Virgin Islands 10,894,750 2,430,000
Virginia 31,524,350 26,496,795
Washington 32,921,350 25,991,442
West Virginia 11,385,750 8,849,212
Wisconsin 31,551,350 28,332,104
Wyoming 10,899,750 7,156,000
Total 1,158,347,156 872,470,565
Source: GAO analysis of Department of Labor data. | GAO-23-106696
Note: Total amounts in table 7 differ from those in table 1. Table 7 figures reflect only funds that were
allocated and awarded specifically to states and territories and does not include the $46 million
available for Tiger Team consultations nor the $200 million for unemployment insurance IT
modernization initiatives, as these funds were made available in lump sums and awarded based on
state applications.
Page 54 GAO-23-106696 Unemployment Insurance
Appendix IV: Fraudulent Overpayments Appendix IV: Fraudulent Overpayments
Recovered and Written Off in Unemployment
Recovered and Written Off in Unemployment
Insurance (UI) Programs
Insurance (UI) Programs
The tables below provide amounts that states reported to the Department
of Labor (DOL) for established fraudulent overpayments, recoveries, and
write-offs for the regular UI program (table 8) and pandemic UI programs
(table 9). With the exception of the Pandemic Unemployment Assistance
(PUA) program, DOL requires states to report fraudulent overpayments,
recoveries, and write-offs on a quarterly basis. For PUA, DOL requires
monthly reporting of overpayments and recoveries; DOL does not require
states to include PUA amounts written off.
It may take states many years to recover UI overpayments. Since regular
UI programs have been in existence longer than the pandemic UI
programs, states have had more time to recover overpayments that
occurred many years prior to the recovery. As a result, states might report
instances where the amount recovered for the regular UI program
appears to exceed the reported overpayments for a given reporting
period.
Table 8: Fraudulent Overpayments Recovered and Written Off in the Regular Unemployment Insurance Program, March 2020
– March 2023 (as of May 1, 2023)
State Total fraudulent Total fraudulent overpayments Total fraudulent overpayments
overpayments established recovered written off
(in dollars) (in dollars) (in dollars)
Alabama 5,334,028 5,438,234 86,621
Alaska 5,527,276 5,128,286 66,433
Arizona 22,646,270 28,971,386 2,347,835
Arkansas 7,337,139 8,512,982 593,003
California 367,107,069 208,812,336 721,819
Colorado 3,616,513 4,002,416 3,648,607
Connecticut 17,418,467 16,716,891 136,170
Delaware 2,020,605 1,250,144 0
District of Columbia 5,892,639 8,644,624 90,970
Florida 650,451 3,778,956 3,357
Georgia 34,095,676 11,777,959 2,805,866
Hawaii 4,777,313 1,410,768 0
Idaho 11,916,269 8,447,389 67,521
Illinois 41,460,515 32,144,067 42,449
Indiana 11,822,170 12,277,119 557,416
Iowa 9,680,894 12,877,258 252,374
Kansas 3,703,406 8,772,021 1,247
Kentucky 58,254,321 11,695,453 315,679
Page 55 GAO-23-106696 Unemployment Insurance
Appendix IV: Fraudulent Overpayments
Recovered and Written Off in Unemployment
Insurance (UI) Programs
Louisiana 12,580,087 11,815,248 95,432
Maine 1,577,516 2,275,281 235,984
Maryland 13,613,286 17,945,521 21,644
Massachusetts 29,818,650 24,567,457 209,736
Michigan 5,931,958 6,700,814 485,339
Minnesota 35,064,088 13,277,802 6,861,380
Mississippi 42,046,701 25,280,239 774,434
Missouri 18,018,078 11,880,754 0
Montana 4,992,667 3,067,642 103,562
Nebraska 596,004 2,015,814 2,367
Nevada 12,366,896 6,972,926 2,738,173
New Hampshire 2,245,432 1,655,128 19,291
New Jersey 48,755,967 56,429,439 0
New Mexico 5,884,274 6,370,515 0
New York 249,948,288 154,980,897 24,643,163
North Carolina 45,627,622 12,292,631 7,100,114
North Dakota 1,700,953 1,096,628 6,548
Ohio 89,502,081 20,849,677 1,310,304
Oklahoma 3,645,341 9,431,169 34,288
Oregon 30,389,656 17,070,197 212,176
Pennsylvania 16,790,032 55,540,248 2,949,926
Puerto Rico 4,593,412 1,761,063 0
Rhode Island 3,599,475 5,246,965 7,184
South Carolina 24,509,942 15,415,168 49,739
South Dakota 2,676,030 1,136,568 142,313
Tennessee 8,335,837 12,174,757 1,131,356
Texas 11,039,164 18,549,040 339,391
Utah 9,823,242 7,685,465 185,813
Vermont 3,073,722 1,540,421 137,029
Virgin Islands (U.S.) 286,554 86,168 0
Virginia 15,787,037 5,772,926 3,791,360
Washington 8,691,628 12,388,639 2,658,019
West Virginia 1,412,992 2,105,401 1,207,131
Wisconsin 16,852,283 14,946,811 88,601
Wyoming 2,186,092 2,472,432 101,195
Total dollars 1,397,224,008 953,456,140 69,380,359
Source: GAO analysis of Department of Labor data. | GAO-23-106696
Page 56 GAO-23-106696 Unemployment Insurance
Appendix IV: Fraudulent Overpayments
Recovered and Written Off in Unemployment
Insurance (UI) Programs
Table 9: Fraudulent Overpayments Recovered and Written Off in the Pandemic Unemployment Insurance Programs, March
2020 – March 2023 (as of May 1, 2023)
State Total fraudulent overpayments Total fraudulent overpayments Total fraudulent overpayments
established recovered written off
(in dollars) (in dollars) (in dollars)
Alabama 22,247,643 1,033,806 204,342
Alaska 5,342,585 423,050 4,018
Arizona 130,601,698 5,234,319 275,602
Arkansas 12,481,190 635,684 176,135
California 2,664,903 1,256,073 0
Colorado 382,048,674 960,449 211,868
Connecticut 6,755,530 1,940,377 5,695
Delaware 3,626,308 253,936 0
District of Columbia 6,244,260 577,386 1,380
Florida 204,916 23,578 0
Georgia 27,780,819 164,955 0
Hawaii 3,897,553 700,434 1,200
Idaho 10,849,661 1,968,828 13,023
Illinois 66,759,768 2,489,214 0
Indiana 39,084,328 5,130,582 182,229
Iowa 20,933,644 835,234 6,540
Kansas 2,994,687 126,665 0
Kentucky 52,503,988 333,067 68,565
Louisiana 18,410,377 224,916 47,468
Maine 3,583,045 55,083 9
Maryland 22,384,273 1,118,071 20,550
Massachusetts 74,012,489 1,101,390 1,249
Michigan 3,917,799 47,604 2,880
Minnesota 26,298,145 2,923,538 136,217
Mississippi 80,836,538 15,224,310 72,557
Missouri 38,924,684 2,128,148 0
Montana 5,424,176 1,084,775 93,474
Nebraska 1,054,461 115,505 60
Nevada 7,706,796 402,063 119,004
Page 57 GAO-23-106696 Unemployment Insurance
Appendix IV: Fraudulent Overpayments
Recovered and Written Off in Unemployment
Insurance (UI) Programs
New Hampshire 2,089,265 266,887 0
New Jersey 20,677 0 0
New Mexico 13,407,119 715,144 0
New York 668,496,890 76,996,784 2,420,554
North Carolina 129,529,137 8,579,215 517,805
North Dakota 2,586,402 588,327 34,502
Ohio 1,098,002,231 18,392,359 0
Oklahoma 4,396,159 99,922 1,500
Oregon 56,374,096 2,102,250 0
Pennsylvania 119,206,185 1,286,636 947,338
Puerto Rico 8,541,355 14,462 0
Rhode Island 2,139,227 753,555 125
South Carolina 71,462,001 19,806,306 109,552
South Dakota 4,617,004 937,022 54,565
Tennessee 32,503,303 4,437,687 563,783
Texas 387,441,479 14,749,133 2,605
Utah 19,075,483 2,292,424 65,705
Vermont 3,141,000 1,452,194 28,486
Virgin Islands (U.S.) 1,611,751 125,362 1,574
Virginia 69,352,768 4,778,565 33,914,465
Washington 14,335,162 2,958,616 22,031
West Virginia 56,921,173 0 810
Wisconsin 39,349,745 4,316,893 188,712
Wyoming 408,603 127,299 0
Total dollars 3,884,583,153 214,290,082 40,518,177
Source: GAO analysis of Department of Labor data. | GAO-23-106696
Note: Some states did not report totals for the pandemic UI programs during our review time frame.
Specifically, three states did not report information for Pandemic Unemployment Assistance (PUA),
three states did not report information for Pandemic Emergency Unemployment Compensation
(PEUC), two states did not report information for Federal Pandemic Unemployment Compensation
(FPUC), and 19 states did not report information for Mixed Earner Unemployment Compensation
(MEUC).
Page 58 GAO-23-106696 Unemployment Insurance
Appendix V: Nonfraudulent Overpayments Appendix V: Nonfraudulent Overpayments
Recovered, Written Off, and Waived in
Recovered, Written Off, and Waived in
Unemployment Insurance (UI) Programs
Unemployment Insurance (UI) Programs
The tables below provide amounts that states reported to the Department
of Labor (DOL) for established nonfraudulent overpayments, recoveries
write-offs, and waivers for the regular UI program (table 10) and
pandemic UI programs (table 11). With the exception of the Pandemic
Unemployment Assistance (PUA) program, DOL requires states to report
nonfraudulent overpayments, recoveries, write-offs, and waivers on a
quarterly basis. For PUA, DOL requires monthly reporting of
overpayments, recoveries, and waivers; DOL does not require states to
include PUA amounts written off.
Table 10: Nonfraudulent Overpayments Recovered, Written Off, and Waived in the Regular Unemployment Insurance
Program, March 2020 – March 2023 (as of May 1, 2023)
State Total nonfraudulent Total nonfraudulent Total nonfraudulent Total nonfraudulent
overpayments established overpayments recovered overpayments overpayments
(in dollars) (in dollars) written off waived
(in dollars) (in dollars)
Alabama 47,380,799 11,529,229 112,776 0
Alaska 15,476,061 10,194,698 1,364,678 49,153
Arizona 43,928,975 71,833,863 4,438,116 33,428,140
Arkansas 25,120,424 9,067,035 931,287 565,162
California 383,074,458 74,448,432 615,008 26,820,594
Colorado 424,452,848 73,660,486 178,554,852 16,360,216
Connecticut 40,496,978 16,660,804 125,799 14,914,773
Delaware 7,116,945 4,486,898 1,223 5,799
District of Columbia 20,673,007 12,461,994 187,347 782,328
Florida 551,418,239 117,439,534 1,010,897 83,429,035
Georgia 78,678,118 37,131,482 11,122,250 6,387,412
Hawaii 13,367,332 8,892,625 26,421 509,956
Idaho 11,286,626 6,118,128 87,220 1,340,533
Illinois 505,820,802 71,604,940 39,998 3,720,388
Indiana 130,310,856 42,274,396 1,705,318 413,075
Iowa 77,703,878 26,251,589 7,306,761 21,756
Kansas 27,616,883 15,079,029 79,114 1,302,812
Kentucky 99,458,610 14,200,321 192,193 18,859,966
Louisiana 48,495,490 14,147,877 3,293,869 1,287,455
Maine 15,365,554 6,821,988 1,337,325 425,533
Maryland 349,444,474 42,601,196 2,053,978 1,619,114
Page 59 GAO-23-106696 Unemployment Insurance
Appendix V: Nonfraudulent Overpayments
Recovered, Written Off, and Waived in
Unemployment Insurance (UI) Programs
Massachusetts 1,367,200,988 57,019,342 1,194,487 144,798,865
Michigan 666,096,100 103,120,133 4,211,053 24,735,474
Minnesota 91,354,905 48,231,327 7,725,462 0
Mississippi 38,426,052 20,667,653 885,411 0
Missouri 129,030,797 36,006,129 0 0
Montana 15,426,065 7,890,421 908,312 329,551
Nebraska 12,143,881 5,092,067 15,060 0
Nevada 590,253,434 100,425,550 89,758,756 123,422
New Hampshire 137,672,639 15,463,378 113,202 40,412,590
New Jersey 947,271,163 578,315,344 0 3,888,913
New Mexico 100,459,840 75,320,942 0 0
New York 154,716,407 26,982,627 4,068,249 0
North Carolina 139,698,272 39,977,173 3,337,465 489,645
North Dakota 28,262,811 9,795,120 4,680 209,987
Ohio 305,309,216 75,245,559 4,055,498 51,990,274
Oklahoma 49,782,801 12,001,587 23,888,778 0
Oregon 87,676,075 29,108,896 347,561 10,977,041
Pennsylvania 302,308,905 80,083,777 20,624,847 205,738
Puerto Rico 25,844,210 10,906,680 2,217 0
Rhode Island 16,229,099 6,529,288 2,279 3,559,709
South Carolina 73,148,544 19,912,613 47,507 461,678
South Dakota 7,453,683 3,772,849 376,342 215,637
Tennessee 25,969,352 15,060,447 2,892,466 1,160,183
Texas 1,409,516,992 327,603,965 2,717,743 250,609
Utah 21,381,955 10,551,492 1,515,540 456,137
Vermont 17,908,846 3,433,850 548,190 8,286,397
Virgin Islands (U.S.) 1,314,107 604,691 0 3,852
Virginia 492,758,561 35,002,667 38,372,310 79,320,701
Washington 979,462,447 144,110,439 15,890,685 26,213,615
West Virginia 21,217,090 5,132,688 4,117,132 0
Wisconsin 84,976,837 39,082,384 6,386,787 7,244,025
Wyoming 7,355,033 4,345,819 171,066 398,774
Total dollars 11,264,314,464 2,633,703,441 448,765,515 617,976,017
Source: GAO analysis of Department of Labor data. | GAO-23-106696
Page 60 GAO-23-106696 Unemployment Insurance
Appendix V: Nonfraudulent Overpayments
Recovered, Written Off, and Waived in
Unemployment Insurance (UI) Programs
Table 11: Nonfraudulent Overpayments Recovered, Written Off, and Waived in the Pandemic Unemployment Insurance (UI)
Programs, March 2020 – March 2023 (as of May 1, 2023)
State Total nonfraudulent Total nonfraudulent Total nonfraudulent Total nonfraudulent
overpayments overpayments recovered overpayments overpayments
established (in dollars) written offa waived
(in dollars) (in dollars) (in dollars)
Alabama $ 236,188,851 $ 13,519,425 $ 614,574 $ 194,044
Alaska 75,725,260 30,913,923 22,713 5,091,569
Arizona 89,071,647 17,828,558 12,245,327 44,353,167
Arkansas 142,445,184 3,720,193 1,298,600 858,622
California 39,043,749 331,151 0 0
Colorado 2,233,971,821 225,323,367 213,400,958 176,834,237
Connecticut 18,241,163 2,792,622 0 6,379,041
Delaware 12,084,126 2,852,438 0 600
District of Columbia 63,871,929 32,839,807 47,836 276,217
Florida 3,447,451,453 91,840,956 431,746 522,756,035
Georgia 47,262,434 2,874,432 0 3,471
Hawaii 17,657,259 12,859,371 0 492,500
Idaho 26,886,191 7,942,124 141,169 9,028,929
Illinois 2,702,495,829 82,414,901 0 86,727,948
Indiana 1,117,033,830 84,722,393 7,348,814 110,202,363
Iowa 106,714,289 11,122,778 199,375 20,260,328
Kansas 24,366,595 1,924,783 0 27,873
Kentucky 24,435,631 1,245,960 131,197 8,699,697
Louisiana 205,592,949 9,619,651 33,960 18,189,644
Maine 83,739,781 3,885,614 11 285,360
Maryland 3,971,797,918 67,376,375 10,979,855 261,603,083
Massachusetts 3,020,190,957 158,919,275 10,571,879 1,218,566,906
Michigan 1,949,204,677 39,983,007 1,290,096 10,274,767
Minnesota 55,309,611 13,304,736 112,128 0
Mississippi 349,628,955 29,671,998 192,409 0
Missouri 494,865,173 23,851,909 0 81,379,730
Montana 76,300,556 9,695,301 469,039 713,899
Nebraska 56,582,099 8,567,330 77,447 1,332,878
Nevada 986,460,095 52,029,201 25,964,582 3,236,471
Page 61 GAO-23-106696 Unemployment Insurance
Appendix V: Nonfraudulent Overpayments
Recovered, Written Off, and Waived in
Unemployment Insurance (UI) Programs
New Hampshire 157,710,695 10,463,106 2,090 7,898,054
New Jersey 107,166,709 2,192,073 0 0
New Mexico 551,502,937 32,095,142 0 32,555,980
New York 215,817,440 36,174,954 9,338,213 0
North Carolina 797,588,272 64,952,495 6,776,154 25,099,654
North Dakota 80,108,256 7,739,293 3,129 556,467
Ohio 5,403,592,587 141,796,582 170,389 415,455,971
Oklahoma 56,006,144 2,916,638 14,596,311 0
Oregon 111,384,170 5,742,835 0 5,380,564
Pennsylvania 2,925,184,909 377,007,276 2,372,584 1,886,589
Puerto Rico 183,737,814 25,850,537 0 0
Rhode Island 45,959,642 4,186,991 1,191 7,954,797
South Carolina 159,582,791 33,538,483 114,860 1,215,328
South Dakota 21,412,579 5,964,033 120,524 3,405,848
Tennessee 69,206,237 6,271,803 853,714 2,278,809
Texas 3,679,413,674 917,316,677 250,394 1,515,975,221
Utah 43,896,825 6,405,364 56,749 550,474
Vermont 5,655,900 3,464,664 10,518 156,469
Virgin Islands (U.S.) 3,541,245 460,419 0 3,852
Virginia 905,413,053 54,650,533 74,450,774 217,498,861
Washington 1,833,884,292 154,606,336 3,035,097 25,961,729
West Virginia 50,053,928 1,794,904 143,150 328,926
Wisconsin 175,610,406 48,153,324 2,173,020 18,814,471
Wyoming 24,531,788 3,025,447 100,083 2,003,431
Total dollars 39,282,582,305 2,990,743,488 400,142,659 4,872,750,874
Source: GAO analysis of Department of Labor data. | GAO-23-106696
Note: Some states did not report totals for the pandemic UI programs during our review time frame.
Specifically, three states did not report information for Pandemic Unemployment Assistance (PUA),
three states did not report information for Pandemic Emergency Unemployment Compensation
(PEUC), two states did not report information for Federal Pandemic Unemployment Compensation
(FPUC), and 19 states did not report information for Mixed Earner Unemployment Compensation
(MEUC).
Page 62 GAO-23-106696 Unemployment Insurance
Appendix VI: Comments from the
Appendix VI: Comments from the Department
of Labor
Department of Labor
Page 63 GAO-23-106696 Unemployment Insurance
Appendix VI: Comments from the Department
of Labor
Page 64 GAO-23-106696 Unemployment Insurance
Appendix VI: Comments from the Department
of Labor
Page 65 GAO-23-106696 Unemployment Insurance
Appendix VII: GAO Contact and Staff
Appendix VII: GAO Contact and Staff
Acknowledgments
Acknowledgments
Seto J. Bagdoyan, (202) 512-6722, BagdoyanS@gao.gov
GAO Contacts
Jared B. Smith, (202) 512-2700, SmithJB@gao.gov
In addition to the contacts named above, Gabrielle Fagan (Assistant
Staff Director), Daniel Flavin (Assistant Director), Dae Park (Assistant
Acknowledgments Director), Erica Varner (Assistant Director), Erin Barry, Ranya Elias, Cole
Haase, Daniel Harris, Lauren Kirkpatrick, Won Lee, Sophia Liu, Maria
McMullen, Isaac Pavkovic, Gloria Proa, Steven Putansu, Sabrina
Streagle, and April VanCleef made key contributions to this report.
Page 66 GAO-23-106696 Unemployment Insurance
The Government Accountability Office, the audit, evaluation, and investigative
GAO’s Mission arm of Congress, exists to support Congress in meeting its constitutional
responsibilities and to help improve the performance and accountability of the
federal government for the American people. GAO examines the use of public
funds; evaluates federal programs and policies; and provides analyses,
recommendations, and other assistance to help Congress make informed
oversight, policy, and funding decisions. GAO’s commitment to good government
is reflected in its core values of accountability, integrity, and reliability.
The fastest and easiest way to obtain copies of GAO documents at no cost is
Obtaining Copies of through our website. Each weekday afternoon, GAO posts on its website newly
GAO Reports and released reports, testimony, and correspondence. You can also subscribe to
GAO’s email updates to receive notification of newly posted products.
Testimony
Order by Phone The price of each GAO publication reflects GAO’s actual cost of production and
distribution and depends on the number of pages in the publication and whether
the publication is printed in color or black and white. Pricing and ordering
information is posted on GAO’s website, https://www.gao.gov/ordering.htm.
Place orders by calling (202) 512-6000, toll free (866) 801-7077, or
TDD (202) 512-2537.
Orders may be paid for using American Express, Discover Card, MasterCard,
Visa, check, or money order. Call for additional information.
Connect with GAO on Facebook, Flickr, Twitter, and YouTube.
Connect with GAO Subscribe to our RSS Feeds or Email Updates. Listen to our Podcasts.
Visit GAO on the web at https://www.gao.gov.
Contact FraudNet:
To Report Fraud,
Website: https://www.gao.gov/about/what-gao-does/fraudnet
Waste, and Abuse in
Automated answering system: (800) 424-5454 or (202) 512-7700
Federal Programs
A. Nicole Clowers, Managing Director, ClowersA@gao.gov, (202) 512-4400, U.S.
Congressional Government Accountability Office, 441 G Street NW, Room 7125, Washington,
Relations DC 20548
Chuck Young, Managing Director, youngc1@gao.gov, (202) 512-4800
Public Affairs U.S. Government Accountability Office, 441 G Street NW, Room 7149
Washington, DC 20548
Stephen J. Sanford, Managing Director, spel@gao.gov, (202) 512-4707
Strategic Planning and U.S. Government Accountability Office, 441 G Street NW, Room 7814,
External Liaison Washington, DC 20548
Please Print on Recycled Paper.