Public housing carries a USD 169.1 billion capital needs estimate as of 2025-10 against the FY2026 Capital Fund enacted at USD 3.2 billion

What is happening?

The United States has 899,047 public housing units. In October 2025 the Public and Affordable Housing Research Corporation published the first national estimate since 2010 of what it would cost to preserve them: USD 169.1 billion, or USD 188,090 per unit in 2025 dollars. About 267,000 of those homes, 30 percent, are in developments that failed their most recent Real Estate Assessment Center physical inspection — twice the number that failed in 2019.

Congress enacted USD 3.2 billion for the Public Housing Capital Fund in FY2026, in a bill signed on 2026-02-03. The same research puts newly accruing needs at USD 3,597 per unit each year, which across 899,047 units is about USD 3.23 billion. The year of funding is consumed by the year of new decay before the accumulated backlog is touched.

Meanwhile, units leave the inventory. HUD has estimated that about 10,000 public housing units are lost every year to demolition and disposition, a figure reported by NLIHC in 2024 that no more recent count replaces, and the route HUD had opened for rebuilding those units stopped accepting new requests on 2026-05-12.

Whose problem is this?

The body that sets the appropriation and the body that absorbs the consequence are not the same. Congress funds the repair account; the unit that goes unrepaired leaves the inventory through an approval at HUD, and the household that moves out of it lands in a different budget account.

Where does this problem end?

The boundary matters because the standard is not missing here. A dollar figure per unit exists and the appropriation is set below it every year.

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

Now

What it should be

Needs a new measurementno federal target for eliminating the backlog was found, and there is no current official estimate of the need either. The 2010 assessment is the last one the federal government produced, and the USD 169.1 billion figure belongs to a research organization rather than to HUD. The nearest thing to a stated target is the FY2026 request from the five industry organizations, USD 5.0 billion a year, which those organizations describe as enough to cover newly accruing needs and explicitly not enough to reduce the backlog.

How big is it?

The affected population is the people who live in public housing. The range runs from 475,170 to 1,600,000 as of 2025, and the population file carries the same two numbers.

The upper bound is the whole resident population, about 1.6 million people. The lower bound is the part of that population living in the roughly 267,000 homes in developments that failed their most recent physical inspection. Dividing 267,000 by 899,047 and applying the result to 1.6 million residents gives 475,170 people, and that share is 29.7 percent when rounded.

The range is not a confidence interval. It is the distance between two measured subsets of the same portfolio, and the choice between them is a judgment about whether a failing inspection score is the threshold of harm.

Under what conditions does it arise?

1. The bill and the payment are in different units. The need is a stock of USD 169.1 billion. The appropriation is a flow of about USD 3.2 billion a year, and the stock grows by roughly that same amount each year through new accrual. A flow that equals the growth rate holds the stock still. 2. Decay is not an event. A failing inspection score does not close a building on a fixed date and does not stop occupancy. The loss shows up as deterioration spread across years rather than as a dated failure that forces a response. 3. There is a legal exit and no legal entry. Section 18 demolition and disposition, and RAD conversion let a unit leave the public housing inventory. Congress ended the requirement to replace lost units one for one in 1995, and the Faircloth Amendment caps how many public housing units may exist at all. 4. No federal number is in circulation. The last official national assessment was produced in 2010. Every figure in use as of 2026-08 comes from advocacy or research organizations, and those figures range from USD 70 billion to USD 169.1 billion.

What has been tried?

Two directions are visible in this table and only one of them preserves housing. The measuring, the funding requests and the one bill point at repairing the stock; the repositioning notices point at shrinking it, and the shrinking route is the one that has moved most recently.

What was found?

Why is it still unsolved?

Cost structure — the repair bill is far larger than any appropriation on offer, and the cheapest response available to the party holding a failing building is to stop holding it.

The arithmetic is the first half. The accumulated need is USD 169.1 billion. The enacted Capital Fund is USD 3.2 billion a year. New needs accrue at USD 3,597 per unit, about USD 3.23 billion across the portfolio each year. The appropriation covers the new needs that accrue each year and leaves the accumulated backlog untouched. Even under the most generous reading, in which every dollar went to the backlog and nothing new accrued, USD 169.1 billion at USD 3.2 billion a year is a little under 53 years.

The second half is what happens to a building whose repair bill cannot be met. Section 18 demolition and disposition, and RAD conversion both remove the unit from the public housing inventory, and removal ends the repair obligation. Conversion does not add federal money. It reassigns the capital and operating amounts a development already receives, so that the property can borrow against a stream it is already getting, and the conversion cap has been raised from 60,000 units to as many as 455,000. Congress ended the one-for-one replacement requirement in 1995 and the Faircloth Amendment caps the number of units that may exist, so a unit that leaves is not merely unrepaired. It is gone. The route by which agencies could use unused Faircloth authority to rebuild lost units was wound down on 2026-05-12.

The books hide the result. Congress raised total HUD funding by more than USD 7.2 billion in FY2026, and in the same bill the public housing operating account fell by USD 477 million. A unit that is removed from the inventory stops generating a repair bill and cannot be replaced. The household moves onto a voucher in a different account, the capital need disappears from the portfolio total, and the backlog figure improves for a reason that has nothing to do with repair.

What observation would mean it is solved?

Candidates — (a) a repeat national assessment showing the backlog estimate falling in real terms (b) the share of units in developments failing the most recent physical inspection dropping below 30 percent (c) the yearly appropriation exceeding yearly accrual for several consecutive years.

(a) alone is the most dangerous. The backlog also falls when the worst units are demolished, so an estimate that drops after a decade of repositioning may be measuring subtraction rather than repair. Any use of (a) has to be read against the unit count in the same year.

(b) carries a measurement trap of its own — HUD is transitioning inspections to the National Standards for the Physical Inspection of Real Estate, and a change in the instrument moves the failure share independently of the condition it measures. (c) is the cleanest of the three but the accrual figure it depends on is an estimate from a single study, not a measured quantity, and no federal number exists to check it against.

What is it connected to?

Fills with researchthe relation to the wider affordable housing shortage, to the Housing Choice Voucher account that receives displaced households, and to local land values that shape which developments are proposed for disposition was not examined. Relation type and evidence grade were not established.

What these sources do not say

• Whether the Capital Fund was actually cut in FY2026. NLIHC reports a decrease of USD 210 million; the Bipartisan Policy Center shows USD 3,200 million in both FY2025 and FY2026 and describes the account as unchanged. Neither states the baseline behind the NLIHC figure, and we could not resolve it. The two do reconcile on the Operating Fund once shortfall funding is added to formula funding.
• What HUD itself now thinks the need is. The last official assessment is from 2010 and the USD 169.1 billion figure is not a federal number. We could not open the HUD budget justification for the Public Housing Fund to see whether a departmental estimate is published there.
• How many units were lost in 2024, 2025 or 2026. The figure in circulation is about 10,000 a year, attributed to HUD by NLIHC in 2024 without a date or a study behind it. No current count was found, and the 139,000 net loss figure was reported in 2016.
• The geography of the need. Neither the USD 169.1 billion nor the 267,000 failing homes were published with a state or agency breakdown in anything we opened, so the concentration of the problem is unknown.
• What share of the Capital Fund reaches capital work. No source we opened reports how much of the enacted USD 3.2 billion reaches physical improvements, so the arithmetic above treats the whole appropriation as if it reached repair.
• The cost of the alternative. No source compared what the federal government spends relocating a household onto a voucher after demolition against what it would have spent repairing the unit. That comparison is the one that would show whether repositioning saves money or moves it.

See the evidence

The PAHRC report was not read directly — the publisher blocked the fetch — so its findings are taken from three separate accounts that agree on the headline figures of USD 169.1 billion, USD 188,090 per unit and roughly 899,000 units. The USD 3,597 per unit yearly accrual appears in only one of those three, and the portfolio total of about USD 3.23 billion is arithmetic performed here on that figure and the unit count; it is not a published number. Two separate estimates of yearly need happen to land close together and were kept apart rather than blended: the HUD figure of USD 3.4 billion from 2010 and the PAHRC figure of about USD 3.23 billion from 2025. The resident count differs between two of the opened sources, at about 1.6 million in the NLIHC summary and over 1.5 million in the release, and the higher figure was taken as the upper bound of the affected population.

The backlog estimates in circulation diverge for reasons of scope, vintage and method rather than disagreement about the same quantity. USD 70 billion and USD 90 billion are NAHRO estimates of the repair backlog from 2024 and earlier; USD 169.1 billion is a 2025 estimate of full preservation cost built from observed hard construction costs at properties that actually converted under RAD, and it excludes efficiency, climate resilience and regulatory compliance work, which makes it a floor rather than a ceiling. On appropriations the two opened sources agree that FY2026 enacted USD 3.2 billion for capital, and they disagree about whether that is a cut; the Operating Fund reconciles exactly once shortfall funding is added, at USD 5,501 million for FY2025 against USD 5,024 million for FY2026, a difference of USD 477 million. This is a Path A output, so observation_refs is empty and provenance_mode: press-derived.

People affected

Estimated range 475,170–1,600,000 As of 2025

Sensitivity The range is not a confidence interval. It is the distance between two measured subsets of the same portfolio, and the choice between them is a judgment about whether a failed physical inspection is the threshold of harm. The lower bound rests on one conversion that the sources do not perform: residents per unit is taken as the portfolio average of about 1.78, so failing developments are assumed to house families of average size. The report notes that units with more bedrooms carry higher construction costs, which suggests failing stock may skew larger, in which case 475170 understates the count. In the other direction the upper bound counts residents of units in good condition, whose exposure to the backlog is financial and future rather than present. Neither bound is a count of people who moved out: about 10000 units are estimated to leave the inventory each year through demolition and disposition, and those households are no longer in either figure. The unit count and the resident count come from the same October 2025 research and were not cross-checked against a federal inventory file, because HUD sources could not be opened.

Regional breakdown Every figure available is a national aggregate. Neither the USD 169.1 billion estimate nor the 267000 failing homes were published with a state or agency breakdown in any source that could be opened, and splitting a national total across states by population would be proportional allocation with no basis, since public housing stock and its condition are concentrated in particular older cities rather than distributed with population