WHAT THE SOFT INSURANCE MARKET DOESN'T GIVE BACK

THE CASE

In May 2024, North Carolina's Department of Insurance gave NC State a choice for the coming fiscal year. It could keep its $5,000 property deductible and pay a 20% higher rate, or raise the deductible to $25,000 and pay 10% more. NC State took the higher deductible. In the budget, that reads as a saving. In practice the university agreed to pay the first $25,000 of every loss itself, a cost that appears in no premium schedule, income statement or appraisal until something breaks.

The state fund that insures NC State had made the same trade four years earlier, when it raised its own reinsurance retention from $2.5 million to $10 million. Its reinsurers were moving in the same direction: attachment points rose across the market after 2022, and North Carolina's reinsurance renewal cost 23.5% more in 2023. At each level, the buyer of coverage accepted a larger share of the loss to hold the price down, and the risk ended up with whoever owned the building.

The market has since turned. Marsh reports that U.S. commercial property rates fell 13% in the second quarter of 2026, the eighth quarterly decline in a row, and Howden called the January 2026 reinsurance renewal the largest risk-adjusted price drop since 2014. Howden also noted that attachments "remain elevated by historical standards." Prices are coming down. For the most part, the retained layers are not.

We call this the Retention Ratchet. Hard markets move retention down the chain to the owner. Soft markets bring the rate back but leave the retention in place unless the owner negotiates it back.

THE INVESTING CALL

We expect the 2026 soft market to open a gap between how owners and their counterparties value insured assets. An owner that books this year's premium decline as NOI growth, while keeping 2023-era deductibles, will carry a higher value in its own model. Lenders and buyers will focus on the retained layer instead, and price it through liquidity requirements, reserves and deductions from the bid.

The gap should show up first in renewal structures, in lender deductible and liquidity tests, in buyer diligence on statements of values and loss runs, and in bids for catastrophe-exposed assets. Appraisals will mostly miss it, since retained loss has no line in the income statement until a storm hits.

Right now it is also cheap to fix. Carriers are offering lower deductibles to insureds with good data and clean loss histories. An owner that takes the entire soft market as premium savings keeps a liability priced at its own cost of capital, and its eventual buyer will reprice that liability at a higher one.

WHY THE EVIDENCE COMES FROM UNIVERSITIES

Commercial owners negotiate retention privately. Broker renewal decks, deductible buy-down quotes and loss runs rarely leave the risk manager's inbox. Public universities, by contrast, are required to publish much of this: budget memos go out to every department, state funds report to legislatures, and university systems post premium schedules that show the rate change and the total funding change side by side.

That makes them a useful laboratory. The buildings are ordinary institutional real estate (dormitories, hospitals, labs, central plants, parking structures), and the state funds and captives that insure them buy reinsurance in the same global market that sets commercial property prices. The records are dated and free to read.

We use them here the way earlier notes used SEC filings. They happen to show, in public, the three things that determine an owner's actual insurance cost:

Market reports track the rate. Budgets pay for the rate and the exposure base. The owner's balance sheet carries all three.

CASE A · NORTH CAROLINA: EVERY LAYER ON THE PUBLIC RECORD

North Carolina's State Property Fire Insurance Fund insures state-owned buildings, including the UNC System campuses. It keeps the first layer of every large loss and buys reinsurance above that. Between legislative reports, fire marshal presentations and campus memos, each layer's movement is documented.

The fund's economics are thin. From 2019 to 2024 it collected an average of $28.5 million a year in premiums and spent $27.2 million a year on reinsurance, so about 96 cents of each premium dollar went straight to reinsurers. Reinsurers paid back $9.9 million a year in recoveries, a ceded loss ratio of 36.5%. After $15.9 million a year of retained claims and $2.4 million of administration, the fund ran an average annual deficit of $6.7 million. It is less an independent price-setter than a conduit for what the reinsurance market charges, and that market raised both its price and its attachment point.

The retained layer has been used. The state fire marshal's January 2024 presentation lists recent large claims: an NC State substation ($5 million to $6 million, 2022), a UNC-Chapel Hill water loss ($8 million to $9 million, 2023), an NC A&T boiler explosion ($4 million to $5 million, 2023), and an open UNC Wilmington claim of about $10 million. The first three all fall inside the $10 million retention. At midpoint estimates they total $18.5 million, all of it kept by the fund. With the pre-2020 retention of $2.5 million, reinsurers would have carried about $11 million of that, roughly 60%.

The fund then passed the pressure to campuses. NC State's May 29, 2024 memo laid out two options for fiscal 2024-25, both including a 4% inflation increase in building values:

The higher deductible saves 10.4% of the prior year's premium, every year, and costs up to $20,000 more on every occurrence. Because the deductible applies once per occurrence across all buildings, a tornado that damages twenty buildings still triggers one deductible. The trade mostly transfers the small, frequent losses (water leaks, equipment failures, minor fires) and leaves catastrophe exposure largely where it was. The breakeven is:

If NC State has more than 5.2 losses of $25,000 or more a year for each $1 million of prior-year premium, the lower premium costs it money. NC State's premium and claim count are not public, but the University of California's claim data gives a reasonable calibration: 635 property claims in fiscal 2024-25 on $74.8 billion of insured value, or about 8.5 claims per $1 billion, with an average paid claim of $51.8 thousand. At that frequency, and assuming every claim is large enough to absorb the full extra $20,000, the NC State trade only pays if the all-risk premium is above roughly $1.63 million per $1 billion of insured value, about 16 cents per $100. Below that rate, NC State bought a 10-point premium reduction with more than 10 points of retained loss. Smaller claims would lower the threshold, but the direction holds.

The same memo committed NC State to a six-to-nine-month building valuation review with an outside firm, which addresses the other driver, the exposure base. The rate got the attention; the retention and the valuation will decide what NC State actually pays.

The pool is also still catching up. On September 25, 2026, after eight straight quarters of falling commercial property rates, UNC-Chapel Hill's finance office passed along the fire marshal's projection for fiscal 2027-28: a further increase of approximately 5% to 10%, citing statewide claims experience and "the need to maintain adequate funding." Final rates have not been set. The fund kept the first $10 million of every loss through the hard market, and it is now paying for that during the soft one.

CASE B · UNIVERSITY OF CALIFORNIA: LOWER RATE, HIGHER BILL

The University of California finances property risk through Fiat Lux, its captive insurer, which keeps a retained layer and buys reinsurance above it. Each year UC publishes both the property rate change and the change in total property funding. The last two schedules:

A broker would describe this as a rate decrease over two years. UC's budget paid 10% more. The difference is the exposure base, which UC also reports:

Insured value went from $54.8 billion to $74.8 billion, 8.1% a year compounded. The most recent year's increase, 7.8%, lines up closely with the 7.2% non-rate growth in the next year's funding, so the arithmetic is consistent: the rate hardly moved, and the growing base accounts for the bill.

The claims data shows what that base is exposed to. UC reported 650 property claims in fiscal 2023-24 and 635 in 2024-25, up from 576 four years earlier, with an average paid claim of $51.8 thousand and water damage as the leading cause. These are the routine, deductible-sized losses that a higher retention hands back to the owner.

For commercial owners the lesson carries over directly. Replacement-cost inflation, new construction and valuation reviews all raise the base, and carriers enforce it through insurance-to-value reviews and coinsurance clauses. A buyer underwriting from a trailing insurance line set before the last valuation reset is using the wrong number, even in a year when rates fall.

CASE C · UTMB: WHEN THE RETAINED LAYER IS ACTUALLY HIT

Retention is easy to ignore until a large event. The University of Texas Medical Branch in Galveston is a documented example of what happens when one arrives.

Hurricane Ike made landfall on September 13, 2008. More than a million square feet of first-floor space flooded with salt water, as deep as eight feet in the campus's oldest building. Damage was estimated at about $710 million. The University of Texas System had $100 million of disaster coverage in place for a physical plant reported at roughly $900 million.

Insurance paid its $100 million. The rest was handled through operating decisions: UTMB closed beds, cut staff and kept its trauma center shut for months. The $107 million cut in projected business losses came from running a smaller institution.

UTMB also had something no private owner can count on, which is FEMA Public Assistance. It expected about $500 million in future FEMA reimbursement for capital costs and had received $72.6 million by January 2009. Without that backstop, the same uninsured loss would have fallen on the owner's equity, the lender's collateral and the price a future buyer was willing to pay.

For any asset exposed to hurricanes, wildfire, hail or flood, the underlying question is the same. If a loss exceeds what insurance pays, who funds the difference, how quickly, and how much will the next lender or buyer deduct for that uncertainty?

HOW RETAINED LOSS BECOMES A CAPITAL TERM

Retained loss doesn't appear in NOI, but lenders and buyers still put a price on it, and some of them already write it into their documents.

Agency lenders cap it and test liquidity against it. Fannie Mae's Multifamily Guide, effective October 31, 2025, caps the all-other-perils deductible at $50,000 for collateral insured under $10 million, $100,000 at $10 million or more, and $250,000 on a true blanket policy. Wind and hail deductibles are capped at 5% of the collateral's total insurable value and named storm deductibles at 7.5%. Freddie Mac's requirements, updated October 30, 2025, use the same limits.

Fannie's rule for expanded deductibles is the clearest example of how a lender prices retention. A borrower can carry up to $100,000 or $150,000 only if it shows liquid assets of at least four times the deductible, the loan is rated pass with no delinquency in the past 12 months, and the property condition rating is 2 or better. The lender has to assess the borrower's ability to pay the deductible over the full policy term, measure exposure by total paid expenses rather than just the excess over the standard cap, and re-check eligibility every year. In effect, retention becomes a liquidity requirement, only well-maintained properties qualify for it, and a condition downgrade can take it away partway through the loan.

Buyers price it with a risk load. A reinsurer spreads catastrophe layers across thousands of assets, while the buyer of a single coastal property holds that retention on its own. A rational buyer will charge more than the expected loss for carrying it, and the bid is where it has the leverage to do so.

Carriers price the base. Ryan Specialty's May 2026 property review calls accurate statements of values and complete construction, occupancy, protection and exposure data "critical" underwriting inputs. An understated base leads to a coinsurance penalty when a claim is paid and a valuation reset at renewal; NC State's 4% inflation adjustment and outside valuation review are the public version of this.

The soft market gives owners a chance to buy retention back. Marsh reported on July 23, 2026 that insurers are competing through "broader coverage, expanded policy terms, and lower deductibles." IMA reported that catastrophe deductibles "have come down from the post‑2023 highs, especially for accounts with strong loss histories," including firmly priced accounts in Florida, California and the Gulf Coast. Ryan Specialty found earthquake deductibles "largely holding firm" and continued scrutiny of wind and hail deductibles.

So lower retention is available, but selectively, to owners whose data and loss history support it, and it has to be asked for. At the reinsurance level, where the ratchet starts, Howden describes January 2026 pricing as back near where it was about four years earlier, "albeit with comparatively higher attachments and tighter terms."

THE RETENTION SHADOW PRICE

Most underwriting models carry a single insurance number:

The owner's full cost of risk includes a second term:

Whoever buys the asset next will price that second term with a load, since it holds the retention without diversification and has to fund any loss out of equity:

Here λ is the buyer's load on undiversified retained loss (λ = 1 means pricing expected loss only), and the premium credit is the premium reduction the carrier gives the owner for keeping that layer. When the Retention Shadow Price is positive, the owner has kept a layer for less than its next buyer will charge for it.

We track three related measures:

1. Retention Coverage: unrestricted liquidity divided by the largest single-event retained loss. Fannie Mae's expanded-deductible rule implies a benchmark of 4.0x. Below 1.0x, a single event forces a capital call, a covenant discussion or a sale.

2. Rate-to-Bill Gap: the change in total insurance dollars minus the change in rate. For UC over two years it is +11.5 points (+10.0% in dollars against −1.5% in rate). A positive gap means cost is being driven by the exposure base or by retention rather than by the market.

3. Ratchet Gap: the owner's current retention minus the retention the market would offer today at the same premium. It is a measure of what the soft market would give back if the owner asked, and what is forgone by taking the rate cut alone.

A WORKED SOFT-MARKET RENEWAL

This example is illustrative. It uses the agency deductible caps described above and premium changes in the range broker reports describe for catastrophe-exposed accounts in 2026. All asset-level inputs are assumptions.

Take a coastal multifamily asset valued at $150 million, with $120 million of total insured value and a $1.8 million premium on a 5% named-storm deductible, a typical 2023 structure. At renewal, the carrier offers two options:

Assume a 6% annual probability that a named storm causes a loss of at least 5% of insured value, and a 5.5% cap rate.

In the owner's model, Path A adds $3.3 million of value. A buyer that loads the retained layer at 2.0x sees a $2.0 million reduction instead. The $5.2 million difference on a $150 million asset falls within the 3% to 8% single-asset exposure range we set out in Note 01.

The breakevens are worth stating. For a buyer, Path A only beats Path B if the load is below 1.25x at a 6% event probability, or if the event probability is below 3.75% at a 2.0x load. An owner that can't support one of those two numbers in diligence should expect the buyer to substitute its own.

Liquidity makes the case stronger. At 55% leverage the equity is $67.5 million, so Path A's $6.0 million retention is 8.9% of equity in a single storm. Applying Fannie's 4.0x multiple would call for $24 million of liquidity behind Path A and $14.4 million behind Path B. Neither agency applies that multiple to named-storm deductibles, but it shows how a lender reasons about retention when it chooses to.

THE CALLS · DATED, FALSIFIABLE, GRADED IN PUBLIC

We are putting three predictions on record. Each can be checked against public documents within about twelve months, and each has a stated falsifier. We will publish the results.

Call 1 asks whether a state pool passes market prices through with a delay. Call 2 asks whether retention stays put after the rate turns. Call 3 asks whether the exposure base, more than the rate, drives the total bill.

Call 1 starts from the fire marshal's preliminary projection of 5% to 10% and tests whether it holds as the commercial market softens. Call 3 carries the most risk: UC's improving loss projections narrowed its gap from 7.3 points to 4.1 points in a single year.

The underlying prediction applies to any insured owner: the soft market returns rate first, and returns retention only to owners who negotiate for it. If all three calls miss, the mechanism is weaker than we argue here and the worked example overstates the spread.

THE ACTION WINDOW

The policy renewal date is the formal deadline, but the useful work starts well before it, because carriers in this market reward owners who come in with good data.

This window will close. The conditions that are bringing deductibles down for well-documented accounts will reverse after the next heavy loss year, and none of the retentions documented in this note have come down on their own.

THE FIVE QUESTIONS

THE MENDRIA EDGE

Insurance is one of the few operating costs where the owner decides how much of the risk to keep. Most owners make that decision once a year from a broker's premium comparison and never put a number on the portion they keep.

Mendria does. The repricing graph connects each asset's insured values, deductibles, sublimits and loss history to its hazard exposure, the liquidity behind it, and the carrier, lender and buyer who will test it. From that we can show which assets carry retention the market would now take back, which carry retention a buyer will charge a multiple for, and which renewal choices move exit value by more than they move NOI.

North Carolina's state fund, the University of California and UTMB put each step of this in the public record. Commercial owners go through the same steps privately, and their next lender or buyer will price the result.

WHAT THE PUBLIC RECORD PROVES

The public record leaves several values private: NC State's premium, claim count and building-level values; the reinsurance limit above North Carolina's $10M retention; UC's retention and reinsurance structure by layer; UT System's 2008 program terms beyond the $100M limit; and every commercial owner's retention curve. Those gaps limit what this note can show, and they are the same inputs Mendria collects when it runs this analysis for an owner.

SOURCES

• • NC State, Insurance Premium Increase for FY 2024-2025, memo of May 29, 2024 · PDF
• • NC State Insurance and Risk Management, Property Insurance
• • UNC-Chapel Hill, All Risk Special Form Property Insurance
• • UNC-Chapel Hill Finance, Projected Rate Increase for 2027-2028, Sept. 25, 2026
• • NC Office of State Fire Marshal, State Property Fire Insurance Fund and Public School Fund, Jan. 30, 2024
• • NC General Assembly Fiscal Research Division, State Property Fire Insurance Fund, Mar. 4, 2025
• • University of California, FY 2025-26 Insurance Program Premium Schedules
• • University of California, FY 2026-27 Insurance Premium Schedules
• • UC Office of the President, Risk Services Biennial Report
• • UC Office of the President, Captive Insurance Financing Programs
• • David Callender, UTMB, Senate testimony, Sept. 25, 2009
• • Texas Observer, Storm Over UTMB
• • Texas Monthly, Emergency!
• • Fannie Mae Multifamily Guide, Property and Liability Insurance, effective Oct. 31, 2025
• • Freddie Mac Multifamily Seller/Servicer Guide, Chapter 31: Insurance Requirements — all-other-perils deductible caps per 31.5 ($50,000 / $100,000 / $250,000 blanket), wind/hail 5% of TIV per 31.7(b), named storm 7.5% of TIV per 31.7(d), updated Dec. 16, 2025
• • Marsh, Global Insurance Market Index Q2 2026, July 23, 2026
• • Reinsurance News, Howden on the January 2026 renewal
• • IMA Financial Group, Property & Casualty Markets in Focus Q2 2026
• • Ryan Specialty, May 2026 US Property Insurance Review