Federal homeowner protections follow the land title — 78 percent of new manufactured homes are titled as personal property

What is happening?

Of the new manufactured homes constructed in 2024, 78 percent were titled as personal property and 18 percent as real estate, according to the Census and HUD Manufactured Housing Survey figures cited by the Federal Housing Finance Agency in a proposed rule published 2026-06-24. That classification decides which federal homeowner protections reach the household. A loan secured only by the home, with no lien on the land, is not a federally related mortgage loan under Regulation X, so it sits outside the Real Estate Settlement Procedures Act, outside the mortgage servicing and loss mitigation regime built on that same definition, and outside foreclosure. On default it runs through repossession under Article 9 of the Uniform Commercial Code, which permits a creditor to take possession without judicial process where that can be done without a breach of the peace.

The credit terms differ by the same line. FHFA reports a denial rate of 65.6 percent on chattel loans against 8.8 percent on site-built mortgages, and interest rates of 9.24 percent against 6.63 percent, all from its own analysis of 2024 HMDA data. FHFA does not publish the denominators behind those two denial rates.

Most of the households on the personal-property side do not own the ground under the home. Among chattel borrowers recorded in 2019 HMDA data, 49 percent rented the land and a further 24 percent occupied it under an unpaid lease, so roughly 73 percent held no ownership interest in the land as of 2019. For those households the home and the ground answer to two different parties, and a default runs on two tracks that do not wait for each other — repossession by the lender, eviction by the landowner.

Two headline shares in this file measure different things and must not be read as a trend. The Manufactured Housing Survey counts new homes shipped in a year including cash sales, which is where the 78 percent for 2024 comes from. HMDA counts financed purchases of new and used homes, which is where the CFPB figure of 42 percent chattel for 2019 comes from. CFPB published a reconciliation table for exactly this reason.

Whose problem is this?

No single body owns the boundary. The classification is state law, the protection regime keyed to it is federal, and the substitute credit channels are federal but separate from the agency that wrote the definition.

Where does this problem end?

The boundary matters here because the protection attaches to the collateral rather than to the household.

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

Now

Needs a new measurementthe target state: no source opened here names a target. Nothing states what share of manufactured homes should be titled as real property, what approval-rate parity would look like, how many chattel loans the Enterprises should purchase, or how many Title I originations would count as the program working. The FHFA proposal removes the prescribed-activity structure without substituting a number, and GAO asks for timelines and milestones rather than levels. Those recommendations remain Open three years after they were made.

How big is it?

Between 3.87 million and 4.09 million occupied land-lease homesites, using a site count as of 2025 and an occupancy rate as of 2021. The unit is homesites. Homesites are not households, and the two must not be swapped.

The chain runs on a single instrument. An industry association, citing a commercial data provider, put the estimated number of homesites in manufactured housing communities at 4.3 million across 44,000 communities as of 2025. GAO, citing an industry report, states that occupancy in manufactured housing communities has been at or above 90 percent since 2009 and averaged 95 percent in 2021. Applying the floor gives 3,870,000 and applying the 2021 average gives 4,085,000.

Converting homesites to households is not derivable from anything opened in this round. The national figure that 77 percent of manufactured-home households own their home, from the 2021 American Housing Survey, is computed across all manufactured homes including those on land the household owns, and no source states that community sites share that tenure mix. The missing artifact is specific — an American Housing Survey or American Community Survey cross-tabulation of manufactured-home tenure by land tenure.

The band is narrow because the only uncertainty inside the chain is vacancy. The real uncertainty sits outside it, and runs in both directions at once:

• Not counted, and larger than the band — the 27 percent of 2019 chattel borrowers who own the land under the home. They are outside the mortgage protection regime too, but conversion is legally available to them, so they are a different population.
• Not counted — owner-occupants on rent-free family land outside any community, 24 percent of 2019 chattel borrowers, roughly 90 percent of them in twelve southern states. No site inventory covers them.
• Not counted — applicants who were denied. With a chattel denial rate reported at 65.6 percent for 2024, most people who sought this financing never enter a loan-based count at all. Cash purchasers are outside HMDA entirely.
• Counted but should not be — renters of manufactured homes occupy homesites too, and 23 percent of manufactured-home households rented the home as of 2021.

Under what conditions does it arise?

1. The default classification is personal property. In the majority of states a manufactured home is titled as personal property unless someone takes an affirmative step to change it, as GAO recorded in 2023. Nothing has to be decided against the household for the outcome to occur; the outcome is what happens when nobody acts.

2. Conversion is closed to the household that rents the ground. Forty states had a conversion statute as of 2015 on the count of a consumer-law nonprofit. Ten states permitted real-property title on leased land as of the CFPB citation in 2021, and then only under conditions on lease term and foundation type, frequently requiring the cooperation of the landowner. A right that the landowner can decline is not a right the household controls.

3. The home cannot be moved out of the situation. Moving an installed home costs roughly USD 5,000 to USD 15,000 by an estimate reported in 2025, and GAO notes that most owners do not relocate homes once placed, because moving is cost-prohibitive and risks damage. Community occupancy has been at or above 90 percent since 2009, so there is little slack to move into.

4. Three quarters of the product comes from five firms. The top five lenders wrote nearly 75 percent of chattel loans in 2019, and the four largest manufactured-housing originators are specialty lenders offering primarily chattel. A story in which competition closes the rate spread has few competitors to work with.

5. The loan cannot be refinanced out of. Fewer than 4 percent of chattel originations in 2019 were refinances, against about 31 percent of manufactured-home mortgages and 44 percent of site-built loans. Whatever rate the household took at origination is close to the rate it keeps.

What has been tried?

Every federal instrument aimed at this market operates on the credit side. None of them changes the title, and the title is what decides the default.

What was found?

Why is it still unsolved?

Institutional gap — the protection regime is keyed to a property classification that the affected households cannot obtain, and every institution built to serve them on the other side of that line originates approximately nothing.

The first movement is that the protection follows the title rather than the household. Regulation X defines a federally related mortgage loan as one secured by a first or subordinate lien on residential real property bearing a one-to-four family structure or a manufactured home. The manufactured home is named in that definition; the lien on real property is the gate. Section 1024.30 then applies the servicing subpart, including early intervention and loss mitigation and the restrictions on referral to foreclosure, to any mortgage loan as defined by the same anchor. So a single definitional clause removes settlement disclosure, servicing duties and loss mitigation together. This is not an exemption written against manufactured housing. Nobody wrote these households out. They are outside because the collateral is not land, and GAO describes it in exactly those terms, as two laws that do not always apply or fully apply rather than as a carve-out.

The second movement is that the household renting the ground cannot obtain the title. In most states the classification defaults to personal property. Forty states have a conversion procedure on the books. Ten permit conversion for a home on leased land, and then only under conditions on lease term and foundation type that frequently require the landowner to agree. For the roughly 73 percent of chattel borrowers who do not own the land, this is not a paperwork failure they could correct. A household that rents the ground cannot convert the title without the cooperation of the party it pays rent to. And because the home and the ground belong to two parties, default runs on two tracks at once — the lender repossesses under Article 9 without judicial process where that can be done without a breach of the peace, and the landowner evicts under landlord-tenant law. Neither track is foreclosure and neither has to wait for the other.

The third movement is that the substitutes were built and did not run. Title I is the only federal guarantee designed for personal-property home loans; its ceiling sat at USD 69,678 from 2008 while the average new home passed USD 108,000, and it originated zero personal-property loans in 2021 and in 2022. The Ginnie Mae channel for those loans reached zero in FY2022, because only Title I loans are eligible and there was nothing to pool. Duty to Serve has offered chattel extra credit every year since the program began and the Enterprises have never bought one. In June 2026 FHFA proposed removing the prescribed-activity structure so that chattel work would count more freely, and that proposal names why the pilots failed — insufficient industry data, no securitization infrastructure, no robust performance data. Which is to say the market lacks the data that would let it be underwritten, and lacks the data because the market never ran. Meanwhile the newest federal statute on manufactured housing, enacted 2026-07-11, raises the price of the guarantee and lengthens the permitted term without using the words chattel or personal property once. The 2014 HOEPA episode shows what happens when the boundary is approached through price instead — lenders priced just under the new thresholds, volume rose, and the title status did not move at all.

What observation would mean it is solved?

Candidates — (a) a published national count of manufactured homes by title status and land tenure for the occupied stock (b) real-property titling available on leased land in all fifty states (c) a functioning federal channel for personal-property loans, measured as Title I originations and Enterprise chattel purchases above zero for consecutive years (d) servicing and loss-mitigation duties extended to personal-property home loans by rule rather than by title status.

(a) alone measures rather than changes. A stock count would say how many households are on the personal-property side and how many of them rent the ground, which is the number nobody currently has. It would not move a single household across the line. Its absence is nonetheless why none of the other three could be evaluated if it were achieved.

(b) alone can be declined in practice. Conversion on leased land still needs the cooperation of the landowner and a lease term the landowner sets. A statute that makes the procedure legally available in all fifty states leaves the household whose community owner refuses exactly where it was. Ten states already permit it under conditions and the share of new homes titled as personal property was still 78 percent in 2024.

(c) alone improves the price and not the consequence. Cheaper credit secured by a home that still goes through repossession is a better loan on the same collateral. The household would pay less and would still face a default process without judicial supervision, and would still face the eviction track separately.

(d) alone reaches one of the two tracks. Extending servicing and loss mitigation by rule would give the household the protections that follow from the lien side. The landowner track runs on state landlord-tenant law and would be untouched. None of the four closes it alone, because the outcome is produced by the joint of two legal regimes and each candidate reaches only one of them.

What is it connected to?

Fills with researchmanufactured housing community ownership consolidation and lot-rent increases, with an average site rent increase of 6.4 percent in 2024 and community sales estimated at USD 8 billion to USD 9.4 billion in 2021; resident-owned communities and state opportunity-to-purchase laws, adopted in 18 states by a 2021 count; HUD Code preemption of state and local building codes; and the chassis-requirement change in Title III of Public Law 119-101, which gives states one year, or two for biennial legislatures, to certify equal treatment of chassis-free homes in financing, titling, insurance, taxes, transportation and installation. Relation type and evidence grade were not confirmed in this round.

What these sources do not say

• How many manufactured homes in the existing stock are titled as personal property. Every published share is a flow. The Manufactured Housing Survey counts new homes shipped in a year, which gives the 78 percent for 2024. HMDA counts loans originated in a year, which gives the 42 percent for 2019. Nothing opened here counts the stock, and the stock is where the affected households live.
• How many manufactured homes are repossessed. CFPB states the limit of its own instrument directly — HMDA data cover applications and originations and carry no information about loan performance or default risk. Foreclosures are counted in several federal places. Repossessions under Article 9 are counted in none, so the event this entire problem turns on has no national series.
• What the denominator is behind the FHFA denial rates. FHFA publishes 65.6 percent for chattel and 8.8 percent for site-built from 2024 HMDA data and states neither denominator. Pulling the same year from the public HMDA Data Browser gives site-built single-family originations of 5,984,121 and denials of 1,827,215, and denials divided by the sum of denials and originations is 23.4 percent rather than 8.8 percent. Both figures can be correct under different denominators, and no source opened here says which one FHFA used.
• What share of chattel loans the reported share misses. Under the 2018 regulatory relief act, certain depositories and insured credit unions are exempt from reporting HMDA data points including the two manufactured-housing points specifically — secured property type and land property interest. In 2019 roughly 6,000 loans, about 5 percent of manufactured-housing loans, were exempt from both. The indicator that defines this problem is optional for some of the institutions that would report it.
• What any agency is aiming at. No source opened here states a target share of homes titled as real property, a target approval-rate gap, a target volume of Enterprise chattel purchases, or a level of Title I originations that would count as the program working. The FHFA proposal removes prescribed activities without substituting a number, and the GAO request for timelines and milestones has been Open since 2023.
• Whether the Title I loan-limit increases changed anything. The ceiling moved from USD 69,678 to about USD 105,532 in March 2024 and into statute at USD 106,405 in July 2026, with terms now permitted up to 30 years. No FHA production report broken out by the Title I manufactured home program for FY2024 through FY2026 was reachable this round, so the one experiment with a clean before-and-after has no published after.
• Which of the two population figures is mislabelled. The 2026 rulemaking states that manufactured housing serves over 20 million American families and cites nothing for it. The industry fact sheet, citing the American Community Survey, reports 16.6 million Americans residing in manufactured homes as of 2023. Families cannot outnumber individuals, so one of the two carries the wrong unit, and neither source says which.
• Who owns the land. GAO records that determining the extent of purchases of manufactured housing communities by investors is difficult because information is limited on who or what company owns them. The counterparty on the eviction track appears in no register.

See the evidence

Five primary documents were read directly and in full. The FHFA proposed rule was read from the GPO text after the Federal Register site itself returned a block page, and its footnote 30 was confirmed against the Federal Register full-text API. The CFPB report of May 2021 and GAO-23-105615 were both fetched as PDFs and their text extracted locally. Public Law 119-101 was fetched as a PDF and the three term counts were run mechanically on the extracted text, which is why the statement that the statute never uses the word chattel is a count rather than an impression. Two sections of Regulation X were read on the CFPB site because the eCFR redirected to a block page. Three figures rest on secondary reporting because the primary source refused the request — the 2024 Title I loan limits, after HUD returned HTTP 403; the survey of manufactured-home residents, after pew.org returned 403 on every path attempted; and the household count that the ACS would have supplied, which was not obtained at all because the Census API requires a key. One figure comes from an industry source with no federal equivalent — the count of 4.3 million homesites in land-lease communities, published by an industry association citing a commercial data provider, on which the population estimate in block 5 depends. Two disagreements are left visible rather than resolved. FHFA states that manufactured housing serves over 20 million American families with no citation, while the industry fact sheet citing the American Community Survey reports 16.6 million Americans residing in manufactured homes as of 2023, and families cannot outnumber individuals. And the FHFA site-built denial rate of 8.8 percent for 2024 does not reconcile with the 23.4 percent obtained by pulling the same year from the public HMDA Data Browser, because FHFA does not state its denominator. This is a Path A output (research-based definition), so observation_refs is empty and provenance_mode: press-derived.

People affected

Estimated range 3,870,000–4,085,000 As of 2025 homesites and 2021 occupancy

Sensitivity The width of the band is the vacancy assumption and nothing else. It spans 215,000 sites, about 5 percent of the total, and it is not a confidence interval — it is the distance between the floor and the average of one industry occupancy series. The real uncertainty sits outside the chain and runs in both directions at once, and it is far larger than the band. Downward, the count includes renters of manufactured homes, who occupy community homesites but do not own a home and so face the eviction track without the repossession track; 23 percent of manufactured-home households rented the home as of the 2021 American Housing Survey, and no source opened here states the owner-occupancy rate for community sites specifically. Upward, the count excludes two groups that face the same protection gap: the 27 percent of chattel borrowers who owned the land under the home as of 2019, for whom conversion to real-property title is at least legally available and who are therefore a different population, and the 24 percent who held an unpaid leasehold as of 2019, usually family land, roughly 90 percent of them in twelve southern states, for whom no site inventory exists at all. The count also excludes everyone denied credit, and with a chattel denial rate reported at 65.6 percent for 2024 that is most of the people who sought this financing, as well as all cash purchasers, whom HMDA does not observe. Converting homesites to households is not derivable from any source opened in this round. The national figure that 77 percent of manufactured-home households own their home is computed across all manufactured homes including those on owned land, and applying it to community sites would multiply across instruments measuring different populations, which is the error the CFPB reconciliation table exists to warn against. The missing artifact is specific: an American Housing Survey or American Community Survey cross-tabulation of manufactured-home tenure by land tenure. The trailing digits of both bounds are artifacts of the multiplication and carry no precision of their own, since the site count itself is published as 4.3 million.

Regional breakdown No source opened in this round gives a state-by-state count of land-lease homesites, of manufactured homes titled as personal property, or of chattel loans outstanding. The two state-level facts that were found are legal rather than numerical — that personal property is the default classification in the majority of states, and that ten named states permit real-property title on leased land under conditions. Splitting the national homesite figure by state population would be proportional allocation, and it would be wrong in a known direction, because manufactured housing is concentrated in the South and in rural areas rather than distributed with population. GAO records separately that even the ownership of manufactured housing communities is not systematically recorded, so no register exists from which a regional count could be assembled.