What the two audits show
Two federal rental-assistance audits published September 17 point to an oversight problem that a simple count of payment errors would obscure. The smaller multifamily sample contained fewer exceptions, yet its annualized lower-bound dollar estimate reached about 90 percent of the corresponding estimate for HUD’s larger tenant-based program. That is a comparison of the inspectors’ reported estimates, not a finding that 90 percent as much money was stolen or could be recovered.
For tenant-based rental assistance, HUD’s inspector general identified 74 exceptions among 300 sampled payments. The multifamily project-based review identified 42 among 230. Dividing exceptions by the respective samples produces 24.7 percent and 18.3 percent, a 6.4-percentage-point difference. Those are unweighted sample frequencies. Both audits used stratified designs, so these arithmetic percentages should not be presented as nationally estimated error rates or proof that one program operates more accurately.
The dollar comparison runs differently. The tenant-based report’s Appendix A puts estimated annual funds that could be better used at $1,430,175,998; the multifamily appendix gives $1,291,265,109. The second is 90.3 percent of the first, although its payment universe contains only 58.2 percent as many monthly records: 1,338,780 versus 2,300,130. This divergence is the central finding of our cross-report analysis. Counting affected transactions and estimating affected dollars answer different accountability questions.
Counts and dollars answer different questions
The sampling appendices explain the difference. The multifamily estimate uses a weighted average improper amount of $122.79 per monthly payment and a lower confidence bound of $80.38. The corresponding tenant-based figures are $77.80 and $51.82. The lower-bound amount per record is therefore about 55 percent higher in the multifamily calculation. A smaller universe can still produce a similar national dollar estimate when the measured amount per record is higher.
Another distinction matters for readers judging the public cost. Each audit includes technically improper payments: assistance paid in the correct amount to an eligible recipient despite a procedural failure. The exception tables assign $10,109 to that category in the multifamily sample and $7,541 in the tenant-based sample. Those represent 37.1 percent and 29.7 percent of the respective sample exception-dollar totals. They are sample compositions, not national estimates of harmless paperwork, and they should not be subtracted from the national projections without reproducing the weighting.
The two projections also should not be added together and described as a single estimate with the same 95-percent confidence level. Each interval belongs to its own sample design and month. Combining them would require a separate treatment of uncertainty, as well as a clear explanation of what the resulting annualized measure represents.
Responses, method and limitations
HUD disputed parts of the findings. Its multifamily office argued that correcting the sampled issues would reduce program costs by less than $300, which it extrapolated to $32.47 million. The inspector general rejected that reasoning, saying missing eligibility records mattered and HUD had not explained the alternative extrapolation. The tenant-based office generally accepted the findings and two recommendations but did not concur with the precise dollar estimate; auditors said the disputed utility detail did not materially change it.
The disagreement makes the next oversight test concrete: whether HUD can document eligibility and calculations, identify actual payment changes, and demonstrate that corrective controls work. Producing a smaller net-loss figure would not, by itself, answer whether payments satisfied program rules. Conversely, labeling a payment improper does not establish fraud, an ineligible household or a collectible debt. Both reports place the estimates under funds to be put to better use; the public Oversight.gov listings report zero questioned costs.
Our analysis compared the two complete reports, their exception schedules, statistical appendices and management responses. Tenant-based payments were sampled for March 2025; multifamily payments for June 2025. Each annual projection multiplies a one-month lower-bound estimate by twelve. The findings therefore do not measure actual transactions over a full year, establish a trend or support a statistical ranking between programs. We reviewed the public report listings on October 6; they do not establish that the recommended changes have been completed. No new agency response was requested.
Sources and further reading
HUD OIG 2026-FO-0006, September 17, 2026 — complete tenant-based audit, responses and appendices ↗
HUD OIG 2026-BO-0002, September 17, 2026 — complete multifamily audit, responses and appendices ↗
Oversight.gov tenant-based report listing, checked October 6, 2026 ↗
Oversight.gov multifamily report listing, checked October 6, 2026 ↗
