A 30-year affordability promise the tax code lets owners leave at year 15 — the statutory buyout price is a measure of what investors put in, not of what the building is worth as restricted housing

What is happening?

The federal Low-Income Housing Tax Credit buys a 30-year affordability commitment, and the same statute lets an owner out after year 15. The route is called a qualified contract, and it sits inside Section 42(h)(6) of the Internal Revenue Code, the very paragraph that imposes the 30-year minimum. About 10,000 rent-restricted units a year are released through it — an estimate produced by the National Council of State Housing Agencies, published in a letter to the Federal Housing Finance Agency as of 2023-07-12 and repeated by the Federal Reserve Bank of Chicago as of 2025-06-23. No federal agency publishes a count of its own.

The mechanism is short. After year 14 the owner files a written request with the state housing credit agency. The agency then has one year to produce a buyer who will pay a price the statute fixes and who will keep the units affordable for the balance of the extended use period. If no such buyer appears, the extended use period terminates by operation of law. There is no hearing, no public interest finding, and no discretion for the agency to refuse. Existing low-income tenants keep three years of protection from no-cause eviction and from rent increases, and after that the units may go to market rent or be sold unrestricted.

The price decides the outcome. The statute sets it at the applicable fraction of outstanding debt, plus investor equity indexed to the cost of living, plus other capital contributions, less cash distributions. That is a measure of what investors put in, not of what the building is worth to a buyer obliged to keep renting it cheaply. The fifteen organizations that wrote to the Federal Housing Finance Agency as of 2023-07-12 reported that the formula price nearly always exceeds the market value of the property as affordable housing, which is why the search rarely produces a buyer.

The stock this sits inside is large. About 2.46 million units were actively subject to LIHTC rent restrictions as of December 2023, out of more than 2.8 million financed since the credit was enacted in 1986, with roughly 86,000 rent-restricted units placed in service each year since 1990.

Whose problem is this?

The party best placed to object is the state housing credit agency, and that agency is the one the statute deputizes to carry out the exit. Once an extended use agreement has been signed without a waiver, the agency runs the search, records that it failed, and administers a termination it has no power to stop.

Where does this problem end?

The boundary matters because nobody here is breaking a rule and there is no enforcement gap to close. Foreclosure and the qualified contract are the only two exceptions the statute grants to the 30-year commitment, and this document is about the second one.

What is the state now, and what should it be?

Now

Needs a new measurementthe target state: no source opened here states how many units the program is willing to lose through this route, or by when the annual figure is meant to reach zero. The two federal bills that would repeal the route set a date after which new allocations lose the option, and neither they nor any other opened source sets a goal for the existing stock. There is no published benchmark against which twelve months of legislative silence counts as late.

How big is it?

Between 100,000 and 130,000 households, as of 2026-08-08. That is the count of households in rent-restricted units that have already been released through a qualified contract. The low end is the cumulative figure the NLIHC letter states as of 2023-07-12, which is a floor by its own wording. The high end adds three years at the only rate any source publishes, about 10,000 units a year, which is the NCSHA estimate. One rent-restricted unit is treated as one household, because every source opened counts units and none publishes household size for LIHTC properties.

The width of the band is three years of accumulation, not a disagreement about measurement. The band is one-directional in a way worth stating plainly. Nobody publishes a collection start date, a method, a per-agency response rate, or whether an agency that reports nothing is recorded as zero, so if reporting is incomplete the true figure sits above this band rather than below it.

Forward exposure is not derivable. How many households live in units that could still be released this way cannot be computed from any published source. It would require the count of units whose extended use agreements do not contain a waiver, and the state agencies hold every one of those agreements while no agency, association or researcher aggregates them. That is the number this problem is actually about, and it does not exist.

One context figure sits outside the range and must stay outside it. 845,000 units are scheduled to lose affordability requirements between 2024 and 2035, measured against the December 2023 stock. That total is all-causes — scheduled term expiration, foreclosure, casualty and qualified contract together — and no source decomposes it. It sizes the expiration wave inside which this route operates. It is not an estimate of this route, and treating it as one would overstate the quantity by an unknown but large factor.

Under what conditions does it arise?

1. The exit is a right, not a breach. An owner who files a qualified contract request is complying with the statute. There is no bad actor to find, no violation to cite, and no enforcement action anyone could bring. Every instrument built for catching noncompliance is the wrong shape for this.

2. The price formula drifted away from the thing it was meant to limit. The buyout price grows with inflation applied to the original equity contribution. That was thought in 1989 to be a significant limit on returns, in part because tax credit prices were then only about 43 cents for a dollar of credit. As credit prices rose, the equity share of project financing rose with them, and the formula began producing a price above what any buyer of restricted housing would pay. The safeguard became the mechanism.

3. Market rent moved away from restricted rent. A HUD study published in 2012 examined what happens to these properties at year 15 and largely found that they were staying affordable, with no significant public policy concern. The market changed shortly afterward. As the gap between market and affordable rents widened, brokers began advising owners to check whether their documents left the route open. No federal study opened here has revisited the question since.

4. The only working instrument is held before the credits are awarded. A state agency can require a waiver in its qualified allocation plan, and 48 of 54 agencies required or incentivized one in the 9 percent program as of 2023-07-12. That instrument binds new allocations only. It does nothing at all for a building whose extended use agreement was signed years ago.

What has been tried?

What was found?

Why is it still unsolved?

Institutional exemption — the statute that imposes the 30-year requirement grants, in the same paragraph, a lawful way out of it at year 15, and the party positioned to object is the party required to execute the exit.

The first movement is that there is nothing to enforce. Compliance and the problem are the same act. An owner who files the request, waits out the year and takes the building to market rent has done exactly what the Internal Revenue Code permits, and any investigation would end by confirming it. The literature reaches for the word loophole, but a loophole implies drafting that went wrong, and this provision is doing what it was written to do. What changed is not the text but the arithmetic underneath it. The price was set in 1989 to limit investor returns at a moment when credits sold for roughly 43 cents on the dollar and the equity share was small. As credit pricing rose, so did the equity term, and the formula began generating prices that no purchaser of rent-restricted housing could justify. The provision that was supposed to make the exit expensive now makes it close to automatic.

The second movement is that the only instrument that works has a fifteen-year fuse. A state agency can require a waiver, and almost all of them now do — 48 of 54 in the 9 percent program as of 2023-07-12. But a waiver written into a 2020 allocation binds a property that will not reach year 15 until 2035. Every building that can exit today was allocated under a plan written before anyone was watching. Near-universal coverage on paper is therefore a statement about the 2040s, not about the stock currently at risk, and it is easy to read the coverage figure as a solution when it is a promise about a cohort that does not yet exist. Six agencies in the 9 percent program and thirteen in the 4 percent program did neither require nor incentivize a waiver as of that date, and incentivize is not require.

The third movement is that only Congress can reach the existing stock, and the federal action that has actually passed went the other way. The provision is tax law, so no agency can amend it. Three named bills have carried a repeal across three Congresses and none has been enacted; the current one, H.R. 4572, would leave the route in place for every building allocated before 2025-01-01, which is essentially the whole 2.46 million. Meanwhile the one tax law enacted in this window expanded credit production and left the paragraph alone. What agencies have instead is leverage over money that belongs to someone else — Treasury conditioning recovery funds, HUD conditioning mortgage insurance, advocates asking the housing finance regulator to condition enterprise financing. Every one of those levers is prospective, partial, and irrelevant to an owner who needs none of that money. And the failure produces no event. An agency markets a property for a year at a price nobody pays, files the paperwork recording a good-faith search that came up empty, and the restriction lapses on schedule. Nothing breaks and no alarm sounds, which is part of why so little of this is counted.

What observation would mean it is solved?

Candidates — (a) Congress enacts a repeal of the termination route or a repricing of the qualified contract at value as restricted housing, applying to the existing stock (b) the annual count of units released through this route falls toward zero (c) every allocating agency requires a waiver, and the share of restricted units covered by one rises toward the whole stock.

(a) alone is weaker than it looks. The bill now pending would repeal the route only for buildings allocated on or after 2025-01-01, so enactment as written would leave the exit on the books for nearly every unit currently at risk. Its repricing provision is the part that would bite, and it applies only to written requests submitted after enactment. A repeal can be real and still miss the population the problem is about.

(b) alone is unreadable. The annual figure is a single trade association estimate with no published method, no stated coverage of the 54 agencies, and no rule for how a silent agency is recorded. A fall in that number could mean fewer exits or fewer reports of exits, and no source opened here allows the two to be told apart. A count nobody can audit cannot certify its own decline.

(c) alone counts paper, and counts it in the wrong unit. The 48 of 54 figure describes agency policies, not units. What would have to be measured is the share of extended use agreements that contain a waiver, and that quantity is knowable in principle and published nowhere. Even at full coverage the instrument reaches only future allocations, so the observation would say nothing about the buildings reaching year 15 in any of the next fifteen years. The three have to be read together, and (a) has to be read against which cohort it covers.

What is it connected to?

Fills with researchproject-based Section 8 contract opt-outs and the wider expiring-use portfolio, the preservation finance market and the buyers who acquire year-15 properties, state and local right-of-first-refusal and tenant opportunity-to-purchase laws, and the treatment of expiring affordability restrictions in other countries. Relation type and evidence grade were not confirmed in this round.

What these sources do not say

• How the annual figure was produced. The estimate of about 10,000 units a year is attributed to the trade association of the very state agencies that process these requests. Neither publication that repeats it states the years covered, how many of the 54 agencies reported, or whether an agency reporting nothing is counted as zero. Two independent outlets carry the number and neither footnotes a method. The background paper that would presumably describe it returned HTTP 403 on both published paths.
• How many units still hold an unwaived right. This is the quantity that would size the remaining exposure. It is knowable in principle, since every extended use agreement either contains a waiver or does not and the agencies hold them all. No source opened here aggregates it, and the widely cited 48 of 54 and 41 of 54 figures describe policies rather than units.
• How often a search actually fails. The central mechanical claim is that the agency search predictably comes up empty. The support for it is the phrase nearly all cases, and nothing more. No source opened here reports how many requests were filed, how many ended in termination, how many were withdrawn, or how many found a buyer. The one statistic that would prove or falsify the mechanism is the one nobody keeps.
• What the cumulative figure covers. The letter says more than 100,000 since data collection began, and gives no start date. The summary page for that same letter, published by the same organization twelve days later, says since 1990. At the stated annual rate, 1990 through 2023 would be roughly 330,000, not 100,000. Neither version names who collects the data.
• What share of the 845,000 belongs to this route. The expiration total for 2024 through 2035 is published as one number covering all channels, and the authors state that their estimates possibly undercount actual exits without bounding the undercount. No source opened here separates this route from scheduled term expiration.
• What happened to the households. The statute gives existing tenants three years. No source opened here reports how many were still in place when those three years ended, what they paid afterward, or where they went. Units are counted at the moment of release and then leave the record entirely.
• How many people live in a released unit. No source opened here publishes household size for LIHTC properties, so the conversion from units to persons rests on an assumption none of them supply. The HUD tenant characteristics pages returned empty content on every attempt.
• What became of two federal executive actions. The 2023 request to the housing finance regulator has no published response — silence, not refusal. The HUD mortgage insurance requirement was announced with a comment deadline and a proposed applicability date, and no source opened here gives a final notice number, an issuance date, or its status in 2026.

See the evidence

Four documents were read in full and directly. The statutory text at Cornell, the bill text at the Government Publishing Office, the 2023 letter to the housing finance regulator, and the Illinois allocation plan carry every structural claim in this document. Everything quantitative reaches the reader secondhand, and that is the central weakness of the evidence here rather than an incidental one — the trade association that is the origin of the only recurring unit figures could not be opened, so every number about how many units have been lost is a citation of a citation. Where sources disagree the disagreement is left visible rather than resolved. The cumulative figure is given as more than 100,000 since data collection began in the letter and as more than 100,000 since 1990 on the page summarizing that letter, and at the stated annual rate those two statements cannot both hold. The one-year search window is stated by two secondary sources and could not be confirmed against the statute, because the fetch truncated before the relevant subparagraph on every attempt. The status of the HUD mortgage insurance requirement is reported here as announced and unconfirmed, because the document in circulation is a draft with no notice number and no opened source states that a final notice issued. Bill status rests on a single commercial tracker after the official sources refused, so it is written as no further action recorded rather than as a statement about what Congress has done. This is a Path A output (research-based definition), so observation_refs is empty and provenance_mode: press-derived.

People affected

Estimated range 100,000–130,000 As of 2026-08-08

Sensitivity The width of the band is three years of accumulation at the only rate any source publishes, not a disagreement between competing measurements. The band is one-directional. No source states when the underlying data collection began, how many of the 54 allocating agencies report, or whether an agency that reports nothing is recorded as zero, so incomplete reporting would place the true figure above 130,000 and not below 100,000. What this number fails to count is anyone still exposed rather than already released: forward exposure would require the count of extended use agreements that contain no waiver, and no agency, association or researcher publishes it. It also fails to count households that moved during the one-year agency search or before the three-year statutory protection ended, households that never arrived because a unit left the program before their turn, and rent burden after the protection lapsed. In the opposite direction, the limit is that the low end rests on a single sentence in a single letter whose method is not described anywhere, and the annual rate rests on a trade association estimate that two publications repeat and neither footnotes. The context figure of 845,000 units scheduled to lose affordability requirements between 2024 and 2035 is deliberately excluded, because it is an all-causes total covering scheduled term expiration, foreclosure and casualty alongside qualified contracts, and no source decomposes it.

Regional breakdown The only per-state material located is an NLIHC resource page dated 2024-12-09 whose figures exist solely inside a chart image sourced to a preservation dataset; the page text carries no numbers and none could be extracted. No agency or association publishes state-level counts of units released through this route, and apportioning the national figure by state population or by state share of the credit stock would manufacture values the sources do not support.