Treasury's own data flagged 21,123 potential duplicate rent payments

Treasury's Emergency Rental Assistance program paid out $46.55 billion in COVID-era rent and utility aid to renters, landlords, and utility providers.▸ A May 2024 GAO report⧉ found Treasury still had not completed a detailed assessment of the program's risk of improper payments, and that grantees holding $787 million in funds had never reported a single household-level payment. A separate Treasury Inspector General field audit-508%20compliant%20locked.pdf), opening actual case files at two overlapping Texas grantees, found the concrete version of exactly the problem GAO described in the abstract: of just 50 beneficiaries sampled in detail, 21 had been overpaid, and in the largest single category, because the City of Houston and Harris County were both paying the same renters for the same months without either one noticing.

Congress created Emergency Rental Assistance in two waves: $25 billion in December 2020 (ERA1, part of the Consolidated Appropriations Act) and $21.55 billion in March 2021 (ERA2, part of the American Rescue Plan).▸ Treasury sent the money to states, territories, and local governments -- grantees -- who paid landlords, utility providers, and tenants directly. Treasury distributed $24.98 billion of the ERA1 appropriation to eligible grantees within about two months, by February 2021. As of the dates GAO reviewed, grantees had expended 94 percent of ERA1 funds and 80 percent of ERA2 funds -- a combined $39 billion of the $45 billion allocated to grantees, with 99.98 percent of expended ERA2 dollars going to direct financial assistance, housing stability services, and administration rather than new housing construction.

The speed came at a cost to the paper trail.▸ In a 2022 report, GAO found household-level payment data missing for 26 percent of the $17.1 billion in ERA payments grantees reported for 2021 -- enough, by itself, to place the program above the statutory threshold Congress set for risk of significant improper payments. Treasury agreed to complete a detailed improper-payment risk assessment in response. By this report's cutoff date, April 16, 2024, it still hadn't. Treasury had completed a narrower risk assessment covering its own disbursements to grantees, which concluded the program wasn't significantly at risk -- but that assessment never accounted for missing data or duplicate payments at the household level, the exact gap GAO had flagged two years earlier.

The gap hadn't closed by 2023.▸ GAO's original count found 96 of 378 ERA2 grantees, 25 percent, hadn't reported any household-level payment data as of mid-2023 -- representing about $4.1 billion Treasury couldn't screen for improper payments. Treasury later told GAO that 58 of those 96 had been dropped from GAO's data file by a Treasury file-transfer error the agency didn't catch until it read GAO's draft report. Corrected, the number was smaller but still real: 38 of 378 ERA2 grantees, 10 percent, holding $787 million, had reported no household payment data at all. A further 13 grantees, holding $581 million, hadn't even reported their total expenditures, and 48 grantees hadn't reported the demographic breakdowns Congress required. None of these gaps appeared in Treasury's public quarterly reports, which GAO found gave no indication that data were missing, that 103 local jurisdictions -- including New York City, which passed on $397 million, and Los Angeles County and city, which passed on $212 million and $193 million -- had declined their allocations and transferred roughly $2 billion combined to their states.

GAO's report never opened a single case file.▸ The Treasury Inspector General's did. Prompted by hotline complaints and a request from GAO, OIG ran a desk review of ERA1 payments made by the City of Houston and Harris County -- two grantees serving overlapping populations, since Houston sits inside Harris County. Cross-referencing Treasury's own ERA1 database as of December 31, 2021, OIG identified 21,123 potential duplicate assistance payments between the two grantees' 148,428 total transactions, worth $134.9 million. Matching by street address, OIG narrowed that to 1,797 address groupings with a combined $20 million in payments -- households where both grantees appeared to have paid assistance to the same address.

OIG did not review all $20 million.▸ It selected the 29 highest-dollar address groupings -- about $1.4 million, seven percent of the flagged total -- and, from those, non-statistically sampled 50 beneficiaries worth $456,150 combined for full document review: comparing approved and paid amounts, checking supporting documentation against payment-file addresses, and testing for duplicate payments and assistance exceeding the program's 15-month cap. Of those 50, 21 -- 42 percent -- had been overpaid, for a combined $110,928. Eight beneficiaries had received duplicate assistance from both Houston and Harris County at once, worth $54,709; four had exceeded the 15-month assistance cap across both grantees, worth $16,294; one had an ineligible application -- lacking proof of identification -- that both grantees approved anyway, worth $30,000. One of the eight duplicate-assistance cases turned out to be both problems at once: a beneficiary who received $20,000 total was both an ineligible applicant, for lack of proof-of-residency documentation, and a recipient of payments from both grantees for the same period.

OIG is pursuing recoupment of $98,242 of the $110,928 as questioned costs; the grantees recharacterized the remaining $12,686 to other eligible expenses, a resolution OIG's report explicitly says does not settle the underlying ineligibility.▸ OIG also flagged a structural gap with no dollar figure attached: one beneficiary's file showed an apparent familial relationship between landlord and tenant, and the grantee could not produce documentation the arrangement was an arm's-length rental -- because neither Treasury's Office of Capital Access nor the grantees' own program guidance required one. Houston's management disputed that OIG's targeted sample reflects the program's true error rate; OIG responded that its findings rest on documented, individually verified transactions regardless of what a broader statistical sample might show, and recommended OCA require both grantees to review the entire $20 million population of flagged address groupings, not just the sample.