← Back to Home
July 2026
ATL · AUS · CLT · DFW · FTL · HOU · MIA · BNA · MCO · RDU · SAT · TPA
The
Reef
Report
Issue 04
← Back to Home
Issue 03 asked whether the lines on the chart were lying. July's data answers a narrower, sharper question: where in the capital stack is the pain actually landing.
It isn't in the rent roll. Multifamily CMBS delinquency rose 28 basis points in June to 7.23%, while overall CMBS delinquency, across every other property type combined, fell 20 basis points to 7.35%. Multifamily is now the outlier, and the reason is specific: 65% of the balance that went newly delinquent in June was non-performing matured balloons. Not missed rent. Not empty units. Loans that hit their maturity date and couldn't refinance.
That is a math problem, not an operating problem, and the distinction changes what you underwrite. An asset with 94% occupancy and a full rent roll can still fail, not because the property is broken, but because its debt was priced for a world where a 10-Year Treasury near 4.7% and a 6.2% cost of capital didn't exist yet.
The rest of July's data agrees with that read. Apartment prices are sitting roughly 20% below their mid-2022 peak. Deal volume is still falling, down 10.7% year-over-year through the first five months of the year, even as $3.7 trillion in global private-equity dry powder sits mostly on the sidelines. Concessions just hit their deepest level in 25 years. None of that is a story about buildings. It's a story about capital that was priced for a rate environment that no longer exists, meeting a rate environment that does.
The opportunity is not that the correction is over. The opportunity is that it has become legible: you can now point to the specific mechanism (maturity, not occupancy) and underwrite around it.
Twelve dated, sourced data points from July 2026, read individually or as a set.
In DFW right now, 2000s-vintage product is cratering and 1960s-1980s product is holding flat. That is the math gap resolving itself by age.
Here is the clearest single data point this issue found, and it comes from inside the market, not from a research desk. Tracking DFW multifamily loan maturities and distress through 2026, broker Buck Poderski's numbers show properties built between 2000 and 2009 posting the steepest price declines year-over-year, over $40,000 per unit, while older vintage deals, 1960s through 1980s, saw stable pricing over the same period, some even ticking up slightly. Same metro. Same rate environment. Opposite direction.
The mechanism is the maturity wall from earlier in this issue, playing out exactly where you'd expect: 95-120 multifamily loans are maturing in DFW every quarter through 2026. Foreclosures went from 8 in 2024 to 40 in 2025. Preferred-equity takeovers roughly 8x'd over the same window. 2000s-vintage product was underwritten and levered at the top of the last cycle; that paper is hitting the wall now. 1960s-80s vintage product was mostly bought with less leverage, at a lower basis, by owners who never needed a 2026 refinance to work.
The uncomfortable part: the market knows this, and isn't rewarding it yet. The same tracking shows roughly 45 active buyers circling a typical 2020s-vintage listing in DFW right now, versus about 2 buyers per 1960s-vintage listing. Capital is still chasing new product even as the data says the old product held its basis better. That gap between what happened and what the market is pricing is the opportunity, not because old assets are exciting, but because almost nobody is looking at them.
This isn't a call to buy anything built before 1990 indiscriminately. The same research turned up a real warning alongside the opportunity: underwriting on older C-class assets sometimes assumes an unrealistic 35% expense ratio when actual opex is closer to 50%, which breaks the debt-service math at normal leverage. The opportunity is real. So is the discipline required to actually underwrite it instead of assuming the discount alone is the thesis.
The divergence is a size story as much as a vintage one, and it runs across every market this issue covers. Below the institutional radar sits a second market: buildings of 5 to 49 units, too small for Yardi Matrix's survey universe, too small for most institutional mandates. Over the trailing twelve months, 1,063 of these properties traded across eleven of our twelve covered metros, at a median of $171,429 per unit, with a median actual transaction cap rate of 6.30% where reported, and 87% of the traded stock built before 1990 (CoStar sale comps, exported July 30, 2026). Set that against the institutional tape in the same metros: sub-50 product clears at roughly 60-75 cents on the institutional per-unit dollar, Atlanta $116,977 versus $186,401, DFW $100,625 versus $160,057, Houston $91,667 versus $127,421, and at cap rates 150-300 basis points wider than the 4-handle-to-low-5s institutional ranges brokers quote for the same cities (Northmarq metro insights, July 2026: Atlanta 4.25-4.75%, Tampa 4.2-5.7%). Same market. Two different assets. Two different prices for the same dollar of rent.
| Sub-50-Unit Tape (TTM) | Sales | Median $/Unit | Median Actual Cap |
|---|---|---|---|
| Miami | 219 | $222,917 | 5.90% |
| Fort Lauderdale | 169 | $196,345 | 6.10% |
| Houston | 160 | $91,667 | 8.00% |
| Tampa | 140 | $138,462 | 6.75% |
| Dallas-Fort Worth | 125 | $100,625 | 6.35% |
| Atlanta | 107 | $116,977 | 7.47% |
When a metro shows this kind of vintage divergence, don't just note it, quantify your own market's version of it before you underwrite. Pull comps split by decade-built, not just by class letter. A "Class B" label can hide a 2005-vintage asset with 2026-vintage debt problems sitting next to a 1978-vintage asset that never had them.
Multifamily CMBS delinquency is diverging from the rest of the market, and the reason why is a single, sharp number.
Overall CMBS delinquency fell 20 basis points in June, to 7.35%. Multifamily moved the opposite direction, rising 28 basis points to 7.23% (Trepp, June 2026 data, reported July 8, 2026). Multifamily is now the property type working against the broader trend, not with it.
65% of all newly-delinquent CMBS balance in June was non-performing matured balloons, loans that failed at their maturity date, not from an operating shortfall. That's the mechanism behind the divergence above, in one number.
Two of this issue's covered states sit in the worst tier nationally for securitized multifamily DSCR below 1.0: Georgia at 15.95% and Texas at 15.70%, meaning roughly one in six or seven securitized multifamily loans in those states isn't generating enough cash flow to cover its own debt service (Trepp state-level DSCR data). Florida sits lower but still meaningful at 8.82%. This is not evenly distributed geography. The concentration lands exactly on this issue's markets.
National repricing data, not metro-specific, the metro-level comps are Harrison's own pull this issue.
Agency liquidity is up. So is the cost of everything else.
FHFA set 2026 agency loan-purchase caps at $88 billion each for Fannie Mae and Freddie Mac, $176 billion combined, up 20.5% year-over-year. That's the largest standing pool of multifamily debt liquidity available this year, and it's meaningfully bigger than last year's.
The Fed, meanwhile, is not coming to the rescue. On July 29 the FOMC held the federal funds rate at 3.50–3.75%, but the vote was 9-3, and all three dissents (Hammack, Kashkari, and Logan) preferred a quarter-point hike, not a cut (Federal Reserve, July 29, 2026). When the internal pressure on the committee points up rather than down, every refinance conversation this year happens without a rate rescue penciled in.
| Source | Term / Leverage | All-In Rate |
|---|---|---|
| Agency (Fannie/Freddie/HUD) | 10-yr fixed, non-recourse | 5.25% – 6.75% |
| CMBS Conduit | 10-yr non-recourse, ≤75% LTV, 1.25–1.40x DSCR | 5.50% – 7.10% |
| Bridge / Debt Fund | 65–75% LTC, strongest sponsors | High-5% – mid-6% IO |
| Bridge / Debt Fund (broad market) | Full range of sponsor quality | 10% – 12% IO |
Rates via PeerSense (live-updated, checked July 2026) and the Crittenden Report / Stormfield Capital on bridge liquidity. 10-Year Treasury at time of pull: 4.69% (FRED, July 24, 2026), closing at 4.67% post-FOMC on July 29; spreads run approximately 175–275bps over it.
The assumable-debt play is on the market right now. A 240-unit, 1985-vintage Tampa community listed this cycle carries 2.97% assumable debt with roughly seven years remaining, against the 5.25-7.10% range in the table above. That's the exact mechanism: a buyer stepping into that loan locks in a rate 250-400bps below anything achievable with fresh financing today, on an asset old enough that it was never competing with this cycle's new Class A supply in the first place.
The largest federal housing package in decades, a new state front on algorithmic pricing, and one rent-control measure struck down.
The 21st Century ROAD to Housing Act was signed into law July 11, 2026, the largest federal housing package in decades, spanning 12 titles and 60 sections. Multifamily-relevant provisions: the RAD (Rental Assistance Demonstration) cap rises by 100,000 units; the HOME program is reauthorized with up to 20% of its allocation newly permitted for new construction; USDA's Rural Housing multifamily preservation program is made permanent; and, adjacent but consequential for the single-family-to-rental capital pool competing with multifamily, corporate single-family-home purchases are capped at 350 units going forward, a compromise between a 50-unit proposal and a 1,000-unit one.
New Jersey signed the FAIR Act on July 20, 2026, a first-of-its-kind state law regulating algorithmic rent-setting software, following New York's Donnelly Act amendment from October 2025. It arrives as the RealPage/Willow Bridge thread keeps widening: Willow Bridge settled the DOJ's price-fixing suit on July 8 (no financial penalty, but it must stop using competitively-sensitive pricing algorithms and accept monitoring, following similar settlements by RealPage itself, Cortland, Greystar, and LivCor), and a new private class action was filed in Philadelphia on July 16 under that city's local algorithmic-pricing ban, seeking treble/statutory damages. What started as a DOJ antitrust matter is now a live front in state legislatures and city ordinances simultaneously.
Not every 2026 rent-control push landed. Massachusetts's Supreme Judicial Court struck a statewide rent-control ballot measure (capped at 5% or CPI, whichever lower) from the November 2026 ballot, not on the merits, but because a religious-facility exemption violated the state constitution. Advocates have already signaled a rewritten resubmission is coming.
If you're still using an algorithmic pricing tool anywhere in your revenue-management stack, treat this as a multi-front legal exposure now, federal, state, and city, not a single DOJ matter that settles and closes. And underwrite the ROAD Act's SFR cap as a genuine capital-reallocation event: capital that was chasing single-family rentals at scale now has a hard ceiling, and some of it is a plausible source of incremental multifamily demand.
Insurance, concessions, and property taxes, three lines moving against the operator at once.
None of these three pressures are new. What's notable in July's data is how far each one has moved from its historical baseline, simultaneously.
The same concession data, read from the renter's side of the table.
A 16.5% concession rate and an 11.1% average discount aren't just an owner-side cost, they're the clearest available signal of renter leverage in the market right now. Renters are negotiating, and operators are conceding at the deepest rate in 25 years rather than holding face rent and losing the lease entirely. That's a different posture than raising rent and accepting higher turnover, and it implies renewal economics are quietly becoming the more important number to track than new-lease asking rent.
The practical read: a rent roll that looks flat on face rent can still be eroding on effective rent, month over month, if concessions are climbing underneath it. Track the gap, not just the headline number.
Ask for trailing concession data by month, not just a current snapshot, a rent roll that looks stable on face rent can be quietly eroding on effective rent if concession depth is climbing underneath it.
One structural headwind, and one open question that's no longer safe to assume away.
Adoption is nearly universal. Returns are not.
| Category | 2026 Reality | Diligence Question |
|---|---|---|
| Leasing Automation | AI leasing chatbots are in wide use for screening and initial contact | Has the tool been tested for disparate-impact / fair-housing exposure? A $2.275M settlement already exists for algorithmic screening bias against voucher holders. |
| Institutional Context | The weak point is the data and context feeding the model, not the model itself | What proprietary data is the AI actually trained or grounded on, versus operating on generic assumptions? |
| Employee vs. Leadership Gap | Line staff use generative AI roughly 3x more than leadership assumes | Is there an actual usage policy, or is adoption happening informally and ungoverned? |
CRE finance sentiment steadied in Q2 2026, but only because expectations reset downward first.
At the June FOMC meeting, the dot plot flipped from a median projection of one rate cut by year-end to essentially none, 9 policymakers now see rates higher, 9 see them flat-to-lower. The CREFC Board of Governors Sentiment Index, surveyed June 25–July 6 and released July 13, rose to 101.0 from 100.1 in Q1, after Q1's sharp 20.2% shock-driven drop. That reads as stabilizing. It's really re-anchoring: interest rates were the weakest single reading in the survey for a second consecutive quarter, 53% of respondents expect a negative impact from elevated rates, 37% neutral, and only 11% positive.
The index improved on 5 of 9 core questions and softened on the other 4, led by borrower and investor demand. Sentiment isn't collapsing. It's adjusting to a rate environment that isn't going to rescue anyone in 2026, and pricing itself accordingly.
The July 29 FOMC decision confirmed the read: a 9-3 hold at 3.50–3.75% in which every dissent argued for a hike, the committee's internal pressure now points away from the cuts this market spent two years waiting on (Federal Reserve, July 29, 2026).
The debate ended July 29, in the long bond rather than the press release. A 9-3 hold with all three dissents pointing up, a chair telling markets the bond market is doing his tightening for him, and a 30-year at 5.21% for the first time since 2007. That is a regime, not a pause. Note what the short end did: it rallied. The market's argument is no longer about the next 25 basis points; it is about whether anyone is coming to rescue the long end, and nobody is. The 2026 maturity wall ($162 billion of multifamily paper due this year, nearly 40% of the hard CMBS maturities stacked in Q4) now reprices against a 19-year high in long rates with no rate rescue on the calendar. We underwrite at today's cost of capital, on today's rents, and book any future cut as unbanked upside. The sellers who understood this in July will transact in August. The ones who didn't will meet their lender's price ladder in the winter, and the buyer at the bottom of that ladder gets asked one question: "When can you close?"
What actually happened in July, metro by metro.
All twelve coral reef markets, June 2026 data. Source: Yardi Matrix MarketPoint (Jun-2026 metro reports) and Yardi Matrix survey series (occupancy). Vacancy = 100 minus Yardi survey occupancy. Deliveries column is Yardi's forward projection for the 12 months ending May 2027, the supply that's coming, not the supply that came.
| Market | Rent Growth (YoY) | Vacancy | Employment Growth | Price Per Unit | Pipeline | 12-Mo Proj. Deliveries | Trans. Vol (TTM) |
|---|---|---|---|---|---|---|---|
| Atlanta | −0.6% | 7.1%* | +0.2% | $186,401 | 27,088 | 20,364 | $4.5B |
| Austin | −4.8% | 8.0% | +0.6% | $234,740 | 22,593 | 16,756 | $2.3B |
| Charlotte | −2.0% | 6.5% | +1.1% | $206,515 | 26,456 | 19,959 | $2.0B |
| Dallas-Fort Worth | −2.0% | 7.5–8.2%‡ | +0.5% | $160,057 | 52,823 #1 US | 32,336 | $3.7B |
| Fort Lauderdale | −0.5% | 5.6% | −0.7% | $257,727 | 9,264 | 6,724 | $1.2B |
| Houston | −1.7% | 8.2–9.0%‡ | +0.2% | $127,421 | 32,646 | 18,015 | $3.6B |
| Miami | −0.5% | 4.6% | +0.1% | $299,005 | 18,279 | 7,871 | $1.1B |
| Nashville | −1.6% | 6.5% | +0.3% | $201,584 | 16,535 | 10,169 | $1.3B |
| Orlando | −1.7% | 6.3% | +0.9% | $208,591 | 21,948 | 15,338 | $2.5B |
| Raleigh | −1.0% | 6.6% | +1.0% | $202,417 | 12,809 | 9,233 | $1.2B |
| San Antonio | −3.2% | 10.4% | 0.0% | $121,455 | 13,837 | 10,589 | $0.8B |
| Tampa | −2.0%† | 6.7% | +0.3% | $187,391 | 19,616 | 13,379 | $1.3B |
* Atlanta vacancy reflects Yardi's Atlanta–Suburban survey market. ‡ Range across Yardi's component markets (DFW: Dallas–North / Dallas–Suburban / Fort Worth; Houston: East / West). † Tampa's MarketPoint report prints a national rank but no metro percentage, figure derived from Yardi's rent-per-unit survey series (Jun-2025 → Jun-2026). Transaction volume is trailing 12 months. All figures: Yardi Matrix, June 2026.
Twelve metros, one pattern with twelve local variations: supply is setting rents, employment is deciding who absorbs it, and the buyers who are still transacting are underwriting through the cycle, not around it. All data: Yardi Matrix MarketPoint, June 2026.
* Tampa's MarketPoint report prints a national rank (134/141) but no metro rent-growth percentage; the −2.0% figure is derived from Yardi Matrix's rent-per-unit survey series (Jun-2025 → Jun-2026).
Every date below was independently verified against a primary source, not assumed from a recurring pattern.
Issue 04 should leave the reader with one distinction, not one number.
The multifamily correction of 2026 is not primarily an operating story. It is a refinancing story. Rents are soft, concessions are deep, and insurance is expensive, all real, all worth underwriting carefully. But the number that actually explains why delinquency is rising specifically in multifamily, while falling everywhere else in CMBS, is that 65% of the newly-delinquent balance failed at maturity, not in operation.
That distinction matters because it points at what to underwrite. A performing asset with a bad refinance date is not the same risk as a badly-run asset. The capital sitting on the sidelines, $3.7 trillion of it globally, knows this too, which is exactly why deal volume hasn't recovered even as prices have fallen 20% from peak. Everyone is waiting for the same repricing to actually clear.
The opportunity is not that the correction is over.
The opportunity is in the gap between assets that are broken and assets that are simply mispriced against a rate environment that changed underneath them.