How a DSCR-Constrained Mortgage Works
Loan Sizing When Cash Flow Limits

In 2022 and 2023 we had a version of the same conversation almost every week. A sponsor would call about refinancing a stabilized property, tell us what he expected the loan amount to be which would work out to the normal LTV percentages he had encountered his whole career. We would run the sizing and our loan would come back at around 65% LTV. The response was usually some form of "why are you underwriting this below the stated LTV limit?"

We weren't being too conservative — we had the same rents, same expenses, same occupancy the sponsor was seeing. What had changed was which loan sizing criteria determined the loan amount.

A mortgage is not sized by one calculation. It is sized by several calculations that run independently, each producing its own maximum loan amount, and the loan is the lowest result of these independent tests. On a HUD 223(f) there are five of them: the amount requested in the application, the amount supported by value, the statutory per-unit limit, the amount supported by debt service coverage, and the cost of acquisition or refinance. On a 221(d)(4) there are four, with replacement cost standing in for value. Every criterion gets calculated. Only one of them (the lowest) ends up controlling.

The lowest result governs because each criterion protects against a different failure. The LTV test protects the collateral position — if the loan goes bad and the property has to be sold, the lender wants enough equity ahead of it that a sale in a soft market still covers the balance. The DSCR test protects the payment stream — the lender wants income to exceed debt service by enough that a bad year, a tax reassessment, or an unplanned capital item does not put the loan into default immediately. Both are cushions against being overlevered; they just measure leverage against different things. A loan that clears the value test but fails the coverage test is still an overlevered loan, which is why passing four of five criteria does not get you the fifth criterion's number.

From roughly 2012 through early 2022, the criterion that controlled a stabilized refinance was almost always value. Rates were low enough that the debt service coverage test produced a bigger number than the LTV test on most properties, which meant the LTV test was the binding one and the debt service test sat silently above the constraining LTV amount. A whole generation of borrowers learned to think about leverage as a simply a percentage of value.

When the ten-year treasury moved, the debt service ceiling dropped below the LTV ceiling and stayed there. The LTV percentage did not change. HUD's maximum LTV on a market-rate 223(f) has actually gone up since then. What changed is that the higher interest rates meant that the property's cash flow stopped being able to carry a loan that large.

That is what "DSCR-constrained" means. The loan is not limited by what the property is worth. It is limited by how much debt service the property's net operating income can cover at the required coverage ratio, and everything else — value, cost, statutory limits — sits silently above that number and does not bind.

The practical consequence is that leverage expressed as a percentage of value becomes an output rather than an input. When a sponsor asks what our maximum LTV is, and the deal is DSCR-constrained, the honest answer is that the LTV maximum is 87% and it is not going to matter. What matters is NOI, the rate, and the amortization term.

The NOI is HUD's, not the operating statement's. HUD applies minimum vacancy factors by affordability type, requires specific treatment of management fees and replacement reserve deposits, and handles commercial income, ancillary income, and real estate taxes in ways that frequently differ from how an owner keeps his own books.

To turn net operating income into a loan amount, you need to know how much annual debt service a dollar of borrowed principal costs. That figure is the annual loan constant: total annual principal and interest divided by the original loan balance.

The constant has two components. The first is the interest rate and the second is the amortization component — the portion of the annual payment that goes to paying down the principal, expressed as a percentage of the original balance.

At 6.00% interest rate on a 35-year amortization, the annual constant is 6.842%. Of that, 6.000% is interest and 0.842% is amortization. Borrow ten million dollars and the annual principal and interest payment is $684,200.

The amortization component behaves in a way most people guess wrong. It gets smaller as the interest rate rises. At 4.00% on 35 years it is 1.313%; at 6.00% it is 0.842%; at 7.00% it is 0.666%. A level payment at a higher rate devotes more of each installment to interest, so less principal comes off in the early years. This is why the spread between two rates produces less difference in the constant than the rate difference alone would suggest — and it is the mechanical reason behind the MIP error covered in Section 05.

Annual loan constant by rate and amortization term. The amortization component is the annual constant less the interest rate.

Read across a row and you can see how much amortization term is worth. At 6.00%, moving from a 30-year schedule to a 35-year schedule drops the constant by 35 basis points. Moving from 30 to 40 drops it by 59 basis points. Applied to a fixed debt service budget, a 59 basis point reduction in the constant produces roughly 9% more loan — with no change in rate, credit, or property performance.

That is one of the structural reasons HUD proceeds are larger than conventional proceeds on cash-flow-constrained deals, and it compounds with HUD's lower DSCR requirement. A 223(f) amortizes over 35 years and a 221(d)(4) over 40, against 30 year amortizations on most conventional and agency executions.

An interest-only structure does not change any of this on the sizing side, which is the part sponsors most often have backwards. A borrower looking at a full-term interest-only agency quote will sometimes assume the DSCR loan sizes off the interest-only payment — no amortization component, a lower constant, a bigger loan. The agencies do not underwrite it that way. Coverage is measured against the amortizing principal-and-interest payment even when the loan never amortizes a dollar, so the amortization component stays in the sizing constant regardless of what the borrower actually pays each month. Interest-only buys cash flow during the hold; it does not buy proceeds at closing. We ran the present-value comparison of the two structures in Interest-Only vs. Amortizing.

The LTV ceiling moves with appraised value and does not care about rates except indirectly, through cap rates. The DSCR ceiling moves with the loan constant, which moves with every rate change.

Take a market-rate 223(f) on a property appraised at $18,000,000 with $1,000,000 of net operating income underwritten for debt service. The LTV ceiling at 87% is a fixed $15,660,000 regardless of where rates go. The DSCR ceiling is not fixed:

35-year amortization, 87% of NOI, 0.25% annual MIP included in the constant. Statutory and cost criteria excluded for clarity.

The crossover on this deal sits at approximately 3.99%. Above that rate, debt service coverage controls the loan. Below it, value controls.

The 65% figure from the conversations in Section 01 is right there in the table at 6.50% — $11,600,229 against an $18,000,000 appraisal is 64.4% of value. The sponsor was not wrong about HUD's leverage. He was quoting a number that had been produced by the LTV test in a rate environment that no longer existed.

Two things follow from the shape of that column. Above the crossover, rate impacts proceeds. Going from 6.50% to 5.50% on this deal is worth $1,396,105 — nearly 12% more loan on identical property performance. This is why rate lock timing matters so much more on a DSCR-constrained deal than on an LTV-constrained one, and why we spend real time on rate assumptions during a sizing conversation rather than treating them as a placeholder.

Below the crossover, rate improvements stop impacting proceeds entirely. At 3.50% the debt service test supports $16.7 million and the LTV test caps the loan at $15.66 million; dropping the rate to 3.00% raises the DSCR number further and changes nothing about the loan. If your deal is LTV-constrained, a better rate lowers your payment and improves your coverage, but it does not increase the mortgage amount at all.

Where the crossover sits on your deal

Note that 3.99% being the crossover point is only applicable to that above example. The crossover moves with the cap rate the appraiser lands on and with which LTV limit applies to the transaction, and it is worth being able to locate without running a full sizing.

Set the two ceilings equal and solve. The DSCR ceiling is NOI times the coverage percentage divided by the loan constant. The LTV ceiling is value times maximum LTV, and value is NOI divided by the cap rate. NOI cancels off both sides:

On a market-rate 223(f) the coverage percentage is 87% and the maximum LTV is 87%. They cancel:

The binding test flips exactly where the all-in loan constant — rate plus MIP plus amortization — crosses the property's cap rate. Below the cap rate, LTV controls. Above it, DSCR controls. The same thing happens on an affordable deal, where the coverage percentage and the LTV limit are both 90% and cancel just as cleanly. That is not a coincidence, and Section 04 covers where it comes from.

Cash-out also breaks the symmetry. Criterion 10 holds a cash-out 223(f) to 80% LTV while the coverage requirement stays at 87% of NOI, so the terms no longer cancel and the crossover constant rises to 1.0875 times the cap rate. A tighter LTV cap means the LTV test binds across a wider band of rates — which is why a cash-out deal on the same property can be constrained by different criteria at the same interest rate.

As we have seen over the last few years, cap rates and interest rates do not move in lockstep, so the crossover point drifts independently of where deals are actually pricing. At the cap rates and rates we are quoting into today, the crossover sits well below market on nearly every stabilized property, which is why almost every market-rate 223(f) we size is DSCR-constrained.

The full four-criteria walkthrough, including the statutory limit and the cost-of-refinance test, is in How HUD Sizes a 223(f) Multifamily Mortgage. The parallel piece for new construction is How HUD Sizes a 221(d)(4) Mortgage.

Almost everyone in this business, us included, says "1.15 DSCR" when describing the coverage requirement on a market-rate HUD loan. But in a unique test, the MAP Guide actually does not say that. Rather, on HUD's official underwriting form (HUD-92264-A) the DSCR criterion is expressed as a percentage of net operating income:

The ratio in the MAP Guide is presented parenthetically and is explicitly noted as rounded. 1 divided by 0.87 is 1.14943. The industry rounds it to "1.15" when talking about it.

The distinction is small in dollars but worth understanding anyway. Sizing to 87% of NOI produces about $5,800 more loan per million dollars of NOI at a 6.00% rate than sizing to a true 1.15x would. On a property with $3,000,000 of NOI that is roughly $17,500. It runs in the borrower's favor, it is the actual calculation HUD performs, and if you build your own sizing model against dividing your NOI by 1.15 you will come in slightly below what the loan committee eventually approves.

The percentage framing also makes the mechanics cleaner to follow, because the percentage is the first step of the calculation rather than a ratio you have to invert:

Every other coverage requirement in the HUD programs works the same way. The "1.11" figure is actually 90% of NOI. The "1.05" figure on certain 223(a)(7) transactions is 95% of NOI.

The coverage percentages are married to the LTV limits

Once the requirements are written as percentages, something shows up that the ratio form hides completely. HUD's coverage percentage and HUD's maximum LTV or LTC are the same number.

It held under the prior standards too. Market rate sat at 85% LTV against 85% of NOI, which the industry called 1.176. When HUD moved market-rate LTV to 87%, the coverage percentage moved with it.

That pairing is what produces the crossover result in Section 03. Because the two percentages are identical, they cancel out of the crossover equation and the binding test flips precisely where the loan constant crosses the cap rate — at every affordability tier, at every leverage level. Only a cash-out transaction breaks it, because Criterion 10 imposes an 80% LTV cap that the coverage percentage does not follow down.

This is the section where we lose people, but it is critical because it is the difference between a sizing you can rely on and one that's wrong.

HUD's annual mortgage insurance premium is not part of principal and interest. It is a separate charge calculated on the outstanding principal balance and paid alongside the mortgage payment. The note amortizes; the MIP does not have its own amortization schedule. It is simply the annual premium rate applied to whatever the balance happens to be. This has two consequences for sizing.

Month one is peak debt service. Principal and interest are level for the life of the loan. MIP declines as the balance comes down. Total debt service is therefore highest at the very start and falls from there. Since the DSCR test has to be satisfied at the highest debt service the loan will ever carry, the test is a month-one test.

MIP enters the loan constant undiluted. Because month-one MIP equals the full loan amount times the annual premium rate divided by twelve, the annualized month-one MIP is exactly the loan amount times the premium rate. Which means you can add the premium rate straight into the constant:

At $1,000,000 of NOI, 6.00%, 35-year amortization, 0.25% MIP:

The last two lines tie exactly to the income available for debt service. That is the check that proves the method.

The shortcut that produces the wrong number

The common shortcut is to add the annual MIP rate to the note rate and run a payment calculation at the blended figure — 6.25% instead of 6.00% — then divide the debt service budget by the resulting constant. We have seen this in borrower models, in broker sizings, and occasionally from HUD originators who have been doing this a long time.

It does not produce a conservative number. It produces a larger loan than the correct method:

Per $1,000,000 of NOI, 35-year amortization, 87% of NOI, 0.25% MIP.

The reason traces back to Section 02. Running the payment at 6.25% instead of 6.00% raises the interest component by 25 basis points but lowers the amortization component, because amortization shrinks as rates rise. The blended constant comes in below the true constant of rate plus MIP plus amortization-at-the-actual-rate. In effect the shortcut amortizes the MIP, and the real loan does not.

The error is largest in low-rate environments, which is exactly when a sponsor is most likely to be running quick numbers to decide whether to pursue a refinance. A borrower who sizes his own deal this way at 3.50% walks in expecting a quarter million dollars per million of NOI that we cannot deliver.

Coverage improves every year

Because P&I is level and MIP declines, actual debt service coverage on a HUD loan rises over the life of the mortgage even without any improvement in property performance. On the $12,266,865 loan above, holding NOI flat:

The loan is tightest on the day it closes, and every month after that is easier. Layer in even modest rent growth and the coverage curve steepens considerably.

The 2025 MIP reduction

In September 2025 HUD finalized a notice reducing annual MIP to 0.25% across every FHA multifamily insurance program and eliminating the different Green/Energy Efficient, Affordable, and Broadly Affordable categories created in 2016. The change applied to applications submitted or amended on or after October 1, 2025, provided the loan has not reached initial endorsement.

Because MIP sits in the constant undiluted, 35 basis points of premium reduction costs the same as 35 basis points of rate. On a market-rate 223(f) at 5.50% the move from 0.60% to 0.25% is worth about 5.2% more proceeds. On a 221(d)(4) at 40 years the same 40 basis point reduction is worth roughly 6.2%, because the lower constant makes each basis point a larger share of the whole.

Coverage minimums are set program by program and rent structure by rent structure, not as one number across the platform. The table below covers the four we size against most often — 223(f), 221(d)(4), 223(a)(7) and 241(a), or 223f, 221d4, 223a7 and 241a as they usually get typed in a term sheet.

Two notes on that table.

The 223(a)(7) ratios were not changed by the January 2025 Mortgagee Letter, which addressed 221(d)(4) and 223(f) only. A 223(a)(7) still runs at 1.11, or 1.05 where the property is more than 90% covered by project-based Section 8 or is a Section 213 cooperative.

The 241(a) coverage test is combined across the existing HUD first mortgage and the new supplemental loan rather than run standalone on the new debt. On a seasoned property performing well above 1.11 on its existing loan, the excess coverage carries over to the new phase, which is the single most aggressive feature in the HUD toolkit. We wrote that program up in detail in The 241(a): HUD's Most Powerful Program Most Sponsors Have Never Used.

Healthcare properties financed under Section 232 run on a different set of ratios and sit outside this discussion entirely.

Everything else

Conventional and agency executions generally underwrite to a higher minimum coverage on a shorter amortization schedule, and the two work together against proceeds. A 1.25x minimum on a 30-year schedule is common for stabilized market-rate multifamily; life company and bank quotes often run tighter still, and debt funds size to a stressed or forward-looking coverage figure rather than in-place NOI. Coverage minimums on the agency side move with LTV tier, market designation, and affordability, so any single number is a generalization — the relevant point for a HUD comparison is the direction of both variables at once.

At $1,000,000 of NOI and a 6.00% rate, a 1.25x minimum on 30-year amortization supports approximately $11,119,441. The HUD 223(f) figure at the same NOI and rate is $12,266,865 — about 10.3% more. Roughly half of that gap comes from the coverage requirement and half from the amortization term.

On January 8, 2025, HUD issued Mortgagee Letter 2025-03, revising both the loan ratios and the coverage requirements on 221(d)(4) and 223(f) transactions.

Read as percentages of NOI, market-rate deals improved from 85% to 87% and affordable deals improved from 87% to 90%. The Mortgagee Letter took effect immediately for any application that had not reached initial endorsement, and there was no change to the Criterion 10 loan-to-value ratio governing cash-out refinances or to the underwritten vacancy factors.

Stack the coverage change against the September 2025 MIP reduction and the combined effect on a market-rate 223(f) is substantial:

Per $1,000,000 of NOI, 35-year amortization.

A market-rate deal that was tested in early 2024 and did not size to where the sponsor needed it may well size today, on identical financials, at an identical rate. We have re-run a number of these and it is worth pulling the old file before assuming the answer is the same.