Multifamily Feasibility and Market Studies

The cleanest public cost evidence for 2025 comes from a city. Detroit's housing department compared four recent multifamily projects and found costs between $278,000 and $331,000 per unit, with a composite 47-unit project at $14,310,000 ¹³. The study is useful because it is itemized and recent, and limited because it describes small urban infill in one Midwestern city, financed with public gap money, not the suburban garden community most agency files carry.

The real estate investment trusts fill in the Sun Belt. Camden reported three communities under construction at the end of 2025 totalling 1,162 homes, with $492.0 million of estimated cost and $278.2 million already incurred ²⁴. That gap between estimate and spend is the number a construction lender cares about, because cost to complete is where overruns live. Camden's own Charlotte estimate moved within one quarter, from $163.0 million at September 30, 2025 ²³ to $157.0 million at December 31 for the same 420 homes ²⁴. A sponsor's budget is a forecast, and a study should show its date, its contingency and the contractor's form of contract. Camden's other filings put a 393-home Nashville community at $184.0 million ²⁵ and Raleigh and Charlotte communities of 369 and 349 homes at $138.0 million and $154.0 million ²⁶, while a 498-home Nashville project of unstated type carries $300.0 million ²³. MAA's pipeline held 8 projects and 2,522 units at year-end 2025 ²⁷.

Coastal and urban product costs more, and the filings show it without saying why. AvalonBay's pipeline of 24 communities and 8,572 homes was budgeted at $3,307 million ²⁸; two of its communities in Maryland and Florida, with 612 homes and commercial space, cost $287.0 million combined ²⁸. Equity Residential started a 450-unit New York joint venture at $201.2 million in 2025 ³⁰. RAND's comparison of projects built between 2015 and 2024 found California costs more than twice those in Texas, with municipal fees of $12,000 a unit in Colorado and under $1,000 in Texas ³². A cost per unit borrowed from another state or another building form is not evidence for the subject.

What is not published

No public source gives the land share of total development cost, the share of furniture, fixtures and equipment, or a 2025 to 2026 split between hard, soft and land cost. Camden's balance sheet carries land at $1,784,349 thousand ²⁵, which describes a portfolio, not a project. The only published split is the 2022 developer survey that attributed 40.6% of multifamily development cost to regulation ³¹, an older figure that mixes fees, delay and design mandates. Equipment lead times were not found either. The study therefore takes land from the purchase contract or appraisal, hard cost from a dated bid, and soft cost from an itemized schedule.

Tariffs and the dated quote

Material prices moved after early 2025. Duties on Canadian softwood lumber rose from 14.5% to 35%, a further 10% Section 232 tariff applies to timber and lumber, and steel and aluminum carry a 50% tariff ¹⁵. NAHB puts the rise in building materials prices since December 2020 at 40% ¹⁵. Framing lumber, rebar, structural steel for podiums and aluminum storefront all sit inside an apartment budget, so a guaranteed maximum price signed before mid-2025, or an estimate built from square-foot rules, should be repriced before a lender relies on it.

Readers sometimes expect an SBA loss rate for apartments. There is no clean one, because the asset is ineligible, but the loan-level file is not empty either. SBA's 7(a) file lists 205 loans for $121.6 million under the residential lessor class over the ten fiscal years to 2025, and the 504 file lists 11 loans. The class is NAICS 531110, lessors of residential buildings and dwellings. It holds apartment owners, but also landlords of houses, duplexes and other residential buildings, and borrowers whose code was entered loosely, so these loans cannot be read as SBA financing for apartment buildings. Among seasoned 7(a) cohorts in the class, the charge-off rate by count is 3.9%, on a small base.

Sources. 7(a) loan count: approval FY2016 to FY2025 (ten full fiscal years). U.S. Small Business Administration, Office of Capital Access, 7(a) FOIA loan-level data, as of 2026-06-30. https://data.sba.gov/dataset/7a-504-foia
7(a) dollar volume: approval FY2016 to FY2025 (ten full fiscal years). U.S. Small Business Administration, Office of Capital Access, 7(a) FOIA loan-level data, as of 2026-06-30. https://data.sba.gov/dataset/7a-504-foia
504 loan count: approval FY2016 to FY2025 (ten full fiscal years). U.S. Small Business Administration, Office of Capital Access, 504 FOIA loan-level data, as of 2026-06-30. https://data.sba.gov/dataset/7a-504-foia
7(a) charge-off rate: Charge-off rate by count, seasoned cohorts FY2011 to FY2020. U.S. Small Business Administration, Office of Capital Access, 7(a) FOIA loan-level data, as of 2026-06-30. https://data.sba.gov/dataset/7a-504-foia

An unaffiliated compiler of the same SBA data puts the cumulative 7(a) charge-off rate for the whole real estate sector at 2.9% across 21,152 loans, against 4.0% for all 7(a) lending ³⁸. That sector holds brokers, managers and lessors of every kind, and it is cumulative over a book that includes young loans. Neither number belongs in an apartment credit memorandum as a loss estimate. Our note on SBA default economics explains why the measurement basis changes the answer.

The stress sits where leverage met floating or resetting rates. In CMBS, the multifamily delinquency rate rose from 6.86% in August 2025, then a nine-year high, to 6.94% in January 2026 and 7.69% in July 2026 ³⁹⁴⁰⁴, and matured loans that could not refinance made up 66% of newly delinquent balances in July ⁴. Special servicing of multifamily CMBS had already reached 8.61% in August 2025 ³⁹. The agency books are far cleaner and moved in different directions. Fannie Mae's serious delinquency was 0.78% at March 31, 2026, and the company warned it could rise later in the year ⁴¹; instead it fell to 0.60% at June 30 after a portfolio modification and foreclosures, though the company still expects more delinquencies ⁴⁴. Freddie Mac's rate rose, reaching 0.64% in August 2026, against 0.48% a year earlier and a prior August peak of 0.35% in 2011 ²².

The bank evidence is concentrated in New York City. Flagstar reported $8.8 billion of loans on buildings at least half rent-regulated in the first quarter of 2026, of which $4.3 billion were criticized or classified ⁴², and in July it put about $11 billion of multifamily loans, at a weighted average coupon of about 3.90%, as resetting or maturing between June 30, 2026 and December 31, 2027 ⁴³. Low coupons written years ago are the problem: a building that covered its debt at the old rate may not cover a new loan at today's rate, whatever its occupancy. That is why a refinance study tests coverage at the takeout lender's current rate and sizing, as our note on coverage-constrained takeout works through.

No SBA Inspector General finding on residential rental loans was found, and no USDA multifamily finding either. HUD's own Inspector General found in 2024 that HUD did not issue notices within 15 days to 45% of sampled properties that failed to remedy physical deficiencies ⁴⁵, which is why inspection history belongs in the file for any HUD-assisted or HUD-insured acquisition.

RealPage's quarterly reading of MSCI transaction data put the average price per apartment unit at $211,356 in the first quarter of 2025, $213,629 in the second, $227,167 in the third and $220,684 in the fourth ^51525354. The comparison that matters is the 2015 to 2019 average of about $151,000 ⁵⁴: a building bought or appraised in that period and refinanced now carries a different basis from one bought at the 2021 and 2022 peak. MSCI's price index was down 18.9% over the three years to September 2025 and roughly flat over the last twelve months ⁵⁷, which is the arithmetic behind the CMBS maturities that cannot refinance.

Cap rates on trades sat between 5.41% and 5.63% through the middle quarters of 2025 ⁵²⁵³. Volume recovered to $51.0 billion in the fourth quarter of 2025 ⁵⁵ and then fell to $6.2 billion in April 2026, half the year-earlier level ⁵⁶. CBRE's survey respondents said the conflict between the United States and Iran had reduced the investment activity they expected for 2026 ⁵⁸. CBRE's first-half 2026 survey found the all-property cap rate essentially flat, with infill multifamily the most bearish segment, and a median view that the 10-year Treasury would need to fall to 3.75% to restart sales volume ⁵⁸.

For a study, valuation enters in two places. The appraisal sets loan to value at closing, and the exit cap rate sets the refinance or sale that repays the construction loan or the investor. National class-level cap rates from CBRE's survey are gated, and a widely repeated set of Class A, B and C national averages carries no stated period, so we do not use it. The study should take cap rates from closed local sales of comparable age and class, state the spread it adds for the exit, and show the loan at a higher exit cap as a sensitivity. A single quarter's price per unit is a check on the appraisal, not a substitute for it.

The market study for an apartment project rests on four measures. The capture rate is the share of age, size and income-qualified renter households in the primary market area that the subject must lease to reach stabilized occupancy. The penetration rate is the share that all existing and proposed comparable units must capture together. The absorption period runs from the first certificate of occupancy to the signed lease on the last unit needed for stabilization, and the trade association's convention assumes three to six months of pre-marketing before it ⁷⁰. Market occupancy closes the loop. HUD's own research office notes the absence of a fully developed national practice standard for market analysis ²⁰, so thresholds are set program by program.

State housing finance agencies publish the most explicit thresholds. Illinois updated its standards on August 3, 2026, treating capture of 5% for family projects and 7% for elderly projects as historically acceptable and expecting overall affordable penetration under 25% ¹⁰. Those are one state's rules for tax-credit housing, not a national norm, but they show what a reviewer reads: capture by income band, not a single blended figure. Market-rate agency lenders apply the same logic less formally, through rent comparables and absorption evidence.

Building the count

Demand starts with households, not population. The 2024 American Community Survey counted 46,101,640 renter-occupied units nationally and 28,579,477 units in buildings of five or more ⁶⁶, out of 146,740,964 housing units ⁶⁶. A study repeats that count for the primary market area: renter households by income band and age, the share in multifamily buildings, household growth from the census estimates, and the turnover that frees units each year. It then subtracts the competing supply, every lease-up and under-construction project in the trade area, and checks the result against absorption the comparables actually achieved. Our note on why the trade area is the study covers the boundary, which moves the capture rate more than any other choice.

Absorption is the test

The Census Survey of Market Absorption is the public record of how fast new apartments lease. Of apartments completed in the fourth quarter of 2025, 49% were leased within three months and 90% within twelve ², against 61% and 96% for 2014 completions ¹¹, and the three-month figure fell to 41% for the first quarter of 2026 ¹². The survey covered 77,380 completed units in the fourth quarter of 2025 ². National demand is not the constraint: RealPage measured net demand of 271,300 units in the year to the second quarter of 2026 ⁶⁴, and 187,000 units were absorbed in that quarter alone, a strong peak season ⁶⁵. Seasonality matters for lease-up schedules, because the late-year quarters showed net move-outs in 2025 ⁶⁴. Our note on absorption forensics explains how to read a comparable's lease-up from its own history.

Unit mix and size set the rent per square foot that comparables are read against. Multifamily units started in the fourth quarter of 2025 averaged 1,068 square feet ⁶⁸, and those started in the second quarter of 2026 averaged 1,053 with a median of 1,008 ⁶⁹. A subject with larger units should show that comparables of that size lease at the rent it assumes.

An apartment study has to say which segment the subject belongs to before it can pick its comparables. Building form separates garden-style communities from mid-rise and high-rise buildings. Rent tier separates market-rate units from workforce housing with moderate income limits and tax-credit housing with deep ones. Specialty product separates student housing, leased by the bed on an academic cycle. All five are treated in full below, each with its own evidence.

Three neighbors have their own hubs. Single-family and townhome rental communities built for rent are financed by single-family rental lenders and studied against house rents, on the build-to-rent hub. Manufactured housing communities lease pads, not units, under tenant site lease protections, and are covered on the manufactured housing hub. Projects whose income comes from more than one use, and the income-share and occupancy tests that sort them, including SBA's owner-occupancy route, sit on the mixed-use hub; an apartment building whose commercial space stays within the agencies' limits on commercial income, which the mixed-use hub sets out, stays here. Age-restricted independent apartments use the elderly thresholds described under demand; licensed care belongs to the assisted living hub.

How it earns. An elevator building earns higher rent per square foot on smaller units in places where land is scarce and walkability is part of the product. Rents and margins for the format are not published separately; the national rent and absorption series cover all apartments. Buildings of 20 or more units held 10.8% of U.S. housing units in 2024 ⁶⁶, concentrated in the largest metros, so comparables are usually plentiful but heterogeneous in age and rent regulation.

What it costs. Essex's 2025 filing is the only one in our evidence that states a cost per unit: $573,000 for a 543-unit South San Francisco project budgeted at $311.0 million, started in the first quarter of 2025 ¹⁴. Every other per-unit figure in circulation is a reader's division of totals. Structure, podium and parking costs are not broken out in any source, and construction runs longer than for garden product, which adds interest carry and lease-up timing risk. A study should carry the general contractor's schedule and a carry budget, not a square-foot rate.

How lenders underwrite it. Large owners fund towers from the balance sheet, so named construction lenders on 2025 to 2026 urban deals were not found. HUD 221(d)(4) is the long-dated alternative ¹⁶, and agency takeouts apply as for any stabilized building. Where for-sale condominiums are a fallback exit, the absorption survey shows that market is thin: 3,879 condominiums were completed in the first quarter of 2026 ¹².

Where it fails review. In New York City, rent regulation is the risk a study must quantify. The Rent Guidelines Board voted on June 25, 2026 for a 0% increase on one and two-year stabilized renewals, while its own index showed building operating costs up 5.3% ⁷⁴. Flagstar told investors that a three-year freeze has no NOI impact on buildings up to 70% regulated but cuts NOI by about 7% to 8% above that share ⁴², and it held reserve coverage of 2.87% on its regulated book in the second quarter of 2026 ⁴³. A study on a regulated building has to separate stabilized and free-market units and project each under the guideline orders. Outside New York, urban files fail on cost: a budget without a dated bid, a schedule without carry, and a rent premium over suburban comparables that is asserted rather than shown.

Which cases show it. The New York case and the North Carolina case each test a new market-rate mid-rise, and the Georgia case a community whose form sits between garden and mid-rise.

How it earns. A tax-credit property trades rent for capital. Investors buy the credits, the equity replaces debt, and in return the units carry income and rent limits for decades. The program has financed more than 4.13 million homes since 1986, housing about 9.62 million households, and the Joint Committee on Taxation scores it at $80.3 billion of federal tax expenditure for fiscal 2025 to 2029 ⁸⁴. Rents are capped by income band, so revenue cannot grow faster than area incomes, and the operating risk is on the expense side.

What it costs. The only national median is old: GAO found $218,000 per unit for new construction in 2011 to 2015, reaching $326,000 in California ⁷⁸. Recent evidence is higher and local. California awards from 2020 to 2023 cost $515,000 per unit without prevailing wage and more with it ⁷⁹, a New York City first phase delivers 576 units ⁸⁰, and a 40-home single-family tax-credit community in Georgia closed at $18.5 million in December 2025 ⁸⁰. A study should not lean on the GAO median.

How lenders underwrite it. The 2025 law raised 9% allocations permanently by 12% from 2026 ⁷⁷, to $3.36 per capita, with a small-state minimum that was $3,455,000 on the 2025 base ⁷⁶⁷⁷. Buildings placed in service after 2025 need only 25% bond financing, down from 50%, to earn 4% credits ⁸⁶, which opens more 4% deals but adds taxable construction debt. FHFA doubled each agency's annual tax-credit equity cap from $1 billion to $2 billion ⁸⁷, and the July 2026 housing law lifted the bank public welfare investment cap from 15% to 20% ⁸⁸. State agencies judge the market study: Illinois looks for capture of 5% for family projects and penetration below 25% ¹⁰.

Where it fails review. Expenses are the failure mechanism. Per-unit operating expense in Novogradac's data rose from $5,721 in 2021 to $7,657 in 2024, in a sample of about 194,000 units ⁸¹, after $6,932 in 2023 ⁸². Repairs and maintenance rose $242 per unit in 2024 and insurance $124, and insurance is up 287.1% since 2016 ⁸³. With rents capped, a pro forma that grows expenses at general inflation will show coverage the property cannot hold. The other failures are capture rates above the agency's threshold in a single income band, and a bond structure that assumes the old 50% test.

Which cases show it. None of the worked cases is a tax-credit project; the case list shows the market-rate method the tax-credit study adapts by income band.

All six multifamily cases use the same structure, a bank construction loan taken out by an agency permanent loan, because that is how most new apartment debt is arranged; they differ in size, building form and market. The Arizona case is the largest, a new market-rate community of about 220 units in a high-growth Phoenix-area suburb. The Florida case is a garden community of about 200 units in a high-growth Florida suburb, where the competitive pipeline is the main test, and the Georgia case is a garden or mid-rise community of similar size in a metro-Atlanta suburb. The North Carolina case is a new 200-unit market-rate mid-rise, and the New York case is the smallest, a new-build 150-unit mid-rise in a growing suburban submarket. The Texas case is a new market-rate apartment community of about 200 units.