Most federal mortgage insurance works one way: the government carries the risk and does the underwriting. A program created in 1992 and made permanent in 2000 does it differently. State and local housing finance agencies underwrite affordable apartment loans themselves, and agree in advance to carry part of any loss.
"HFAs may elect to share from 10 percent to 90 percent of the loss on a loan with HUD," GAO writes. Agencies taking half the risk or more can use their own underwriting standards without HUD's approval.
What it has built
Over the ten years to 2025:
- more than $12 billion in insured loans, adjusted for inflation;
- 776 projects, expected to produce or preserve about 93,670 rental units;
- 22 agencies in 18 states and D.C. actually lent, out of 37 ever approved;
- 73 percent of the projects also used low-income housing tax credits.
The money is concentrated: agencies in New York, Massachusetts and Maryland account for about 67 percent of the lending. Most projects are between 50 and 199 units.
Against the three traditional FHA multifamily programs GAO compared it with, the risk-sharing program is small: 7 percent of units over the decade, ranging from 21 percent in 2024 to 4 percent in 2021. It is also not the same product. The traditional programs are about half market-rate housing, by HUD's estimate; this one can only be used for affordable housing.
The oversight gap
GAO found the written oversight requirements for the two lender types broadly similar. What happens in practice is not.
Private lenders in the traditional program get "a full MAP lender review and quality control plan audit once every 3 years," and annually for newer lenders. For the housing agencies:
The officials told us HUD has not conducted on-site reviews of HFAs because of resource constraints.
HUD oversees them instead by reading what they periodically submit. The handbook that governs the program was issued in June 1995; HUD said in 2020 it planned to update it, and told GAO again this year that it plans to.
What is not known
- The loss record cannot be checked. HUD officials say there have been no claims since 2014. GAO notes that HUD's public default data "does not provide default rates by specific program," so the program's performance cannot be compared with the traditional ones from published figures. GAO publishes no claim counts, loss totals or default rates.
- Costs and speed cannot be compared. "Finally, data needed to methodically analyze production times and costs are not publicly or readily available." GAO's rough figures put both programs near or above 1,000 days from commitment to final endorsement, with caveats it says make the comparison unreliable.
- Nothing about rents or tenants. The report describes the mechanism by which cheaper borrowing "can" reach tenants as lower rents. It contains no rent levels, no tenant incomes and no evidence that rents were in fact lower.
One difference that does reach tenants' buildings: under the risk-sharing rules, new construction and substantial rehabilitation can avoid federal prevailing-wage requirements if the agency does not insure the construction loan and no other federal money triggers them. Those requirements generally apply in the traditional programs.
No recommendations
GAO makes none: the report is descriptive, written for the appropriations subcommittees under a Senate report provision. "We provided a draft of this report to HUD and Treasury for their review and comment. HUD and Treasury did not provide comments on the report."
Separately, the federal bank that finances many of these loans is in question. HUD stopped taking applications under that arrangement in 2019, resumed in 2022, and projects no new commitments in 2027; a footnote records that the administration "is examining the FFB program and considering its necessity."