Compound · Issue 02 · Free
Cap rate isn't yield
What recurring capex actually does to multifamily REIT yields — and why coastal names, not Sun Belt ones, carry the lowest percentage drag despite spending the most per door.
—
The same family-office CIO I wrote about two weeks ago called back. He’d read the piece on the post-tax tradeoff between REITs and private multifamily and had what he called “one nagging follow-up.” The 6.74 percent implied cap rate on MAA flows through to roughly a 4.0 percent dividend yield in the math I’d published, and he wanted to know where the missing 270 basis points went. “Debt service, capex, G&A — fine,” he said, “but those are three different problems with three different magnitudes, and I want to know which one is doing the actual work.”
It’s a fair question, and the answer is two things at once. First, the 270 basis-point figure was a stylized approximation; with actual prices from the May 4 close, the gap is closer to 160 basis points and it runs through four layers, not three. Second, and more useful: one of those four layers is the one this piece is about, because it’s the line that varies meaningfully across the multifamily REIT universe and it’s the one almost nobody backs out before ranking names.
Cap rate is the headline number. Capex-adjusted FCF yield — by which I mean cap rate net of recurring capex normalized by enterprise value — is the comparable one. The term isn’t standard sell-side language; sell-side closer-equivalents are “economic cap rate” or “implied AFFO cap rate.” The metric is what matters. Across the five multifamily REITs in this piece, published cap rates span 87 basis points — coastal names tighter, Sun Belt wider, in the conventional direction. Capex-adjusted FCF yields narrow that range to 74 basis points and pull two names — EQR and CPT — onto essentially the same number despite a 14 bp gap in headline cap rate. And the percentage capex haircut, which is what does the narrowing, is lowest at the coastal names — not the Sun Belt ones — despite coastal names spending the most per door in dollar terms. The cap rate dispersion overstates the real economic dispersion, and the rank-by-capex-burden moves against the dollar rank, not with it.
Why the gap matters
Every REIT disclosure splits capital expenditures into something like “recurring” and something else like “redevelopment” or “value-add” or “revenue-enhancing” — and no two agree on where the line goes. MAA, UDR, and CPT each pick a different threshold; the gap widens once you walk over to EQR or ESS. The discipline of reading the footnotes is older than this asset class. It’s also the discipline almost nobody applies when sizing a position.
Recurring capex is what’s pulled out of property NOI to produce AFFO — the number that supports the dividend. A REIT with a high cap rate and a high recurring capex burden has a smaller AFFO than a REIT with a lower cap rate and a lighter one. Once you net for capex, the two can converge depending on the spread. Brokers will quote cap rate to three decimal places. They almost never quote capex-adjusted FCF yield. Ranking REITs on cap rate alone is comparing a gross income stream to a portfolio you haven’t adjusted for what it costs to maintain.
The 270 basis points, decomposed (actually 160)
Start with MAA. At the May 4, 2026 close, MAA traded at roughly $135.83 per share, giving a market cap of approximately $15.8B against $5.66B of total debt — enterprise value $21.46B. Annualized same-store property NOI of $1,314.8M against that EV produces a 6.13 percent implied cap rate. That’s 61 basis points below Issue 01’s 6.74 percent figure — a real revision, not a rounding difference. Issue 01 used a stylized NOI extrapolation rather than the actual Q1 2026 supplemental same-store NOI run-rate; Issue 02 uses the latter. The gap is the cost of publishing rough numbers, and it’s why the methodology in this piece is tighter than what shipped two weeks ago.
MAA’s quarterly dividend of $1.53 per share annualizes to $6.12, which at the same share price gives a 4.51 percent dividend yield. The gap from cap rate to dividend yield is 162 basis points — not the 270 the CIO’s question implied. Issue 01 was using a rougher approximation than I should have published; this piece is using cleaner numbers.
Those 162 basis points run through four layers, not three. The fourth — easy to miss — is AFFO retention. Multifamily REITs distribute most of AFFO as dividends, typically 70 to 85 percent of it, and retain the balance to fund growth capex and balance-sheet flexibility. MAA’s implied payout ratio at these inputs is ~80 percent, near the top of that range. Retention is real but it’s a policy variable, not a structural one. Boards can and do move it.
The other three — debt service, recurring capex, corporate G&A — together account for the rest:
| Layer | bps | Cumulative |
|---|---|---|
| Implied cap rate (gross EV yield) | — | 6.13% |
| + Leverage uplift (1.36x to gross levered equity yield) | +220 | 8.33% |
| − Interest expense (3.9% wtd avg on $5.66B debt) | −140 | 6.93% |
| − Recurring capex ($134M TTM ÷ $15.8B equity) | −85 | 6.08% |
| − Corporate G&A ($64M TTM ÷ $15.8B equity) | −41 | 5.67% |
| = AFFO yield on equity market cap (structural number) | — | 5.67% |
| − AFFO retained (~80% payout ratio — policy, not structural) | −116 | 4.51% |
| = Dividend yield | — | 4.51% |
The four drags do not have equal weight. Interest cost is the biggest single piece (−140 bps) and is mostly a leverage-policy choice. Retention is the next (−116 bps) and is fully a board choice. Recurring capex (−85 bps) is in the middle by size. G&A (−41 bps) is overhead and scales gently across the portfolio. Recurring capex is the one that varies meaningfully across the universe, the one that’s hardest to change once a portfolio is built, and the one this piece is about.
I think the recurring capex line is the most under-discussed number in multifamily REIT analysis. Sell-side notes quote cap rate, NOI margin, and AFFO. They don’t typically reverse out the capex line and ask whether the AFFO they’re using is comparable across the cohort. It usually isn’t, and the differences hide in plain sight — in the supplemental, where the methodology footnote tells you exactly how each REIT decided what counts.
What capex per unit looks like across five names
Five multifamily REITs at Q1 2026, computed on a consistent methodology: recurring capex extracted directly from each company’s supplemental disclosure on a trailing-four-quarter basis, divided by each company’s own per-unit denominator. Cap rates are recomputed on May 4 closing prices for consistency across the cohort.
| Implied cap rate | Recurring capex / unit (TTM) | EV / unit | Capex haircut (bps) | Capex-adjusted FCF yield (≈ implied AFFO cap rate) | |
|---|---|---|---|---|---|
| MAA | 6.13% | $1,304 | $208,742 | 62 | 5.51% |
| UDR | 5.87% | $1,969 | $304,062 | 65 | 5.22% |
| CPT | 6.10% | $1,857 | $261,793 | 71 | 5.39% |
| EQR | 5.96% | $2,312 | $392,074 | 59 | 5.37% |
| ESS | 5.26% | $2,109 | $428,072 | 49 | 4.77% |
The dollar capex per unit varies by roughly 2x across the cohort — MAA at $1,304, EQR at $2,312. That’s the spread the discussion usually stops at, and it supports the intuitive story: older coastal portfolios cost more to maintain than newer Sun Belt ones.
The yield-relevant number is the capex haircut — recurring capex per unit divided by enterprise value per unit. The haircut compresses the dollar spread substantially: the range across the five names is 49 to 71 basis points, a 22 bp band. Higher per-unit capex tracks higher per-unit property values; the coastal names spend more per door but also own doors worth more, so the ratio narrows.
Two findings sit on top of the table. First, rank order is preserved, but CPT and EQR converge to within 2 basis points on capex-adjusted FCF yield despite a 14 bp gap in headline cap rate. Cap rate dispersion overstates the real economic dispersion.
Second — and more useful — the rank by percentage capex haircut moves against the rank by dollar capex, not with it. The two coastal names spend the most per door but carry the lowest haircut: ESS at 49 bps, EQR at 59. The three non-coastal names cluster higher, with CPT — a Sun Belt name — running the highest at 71 bps. MAA, the purest Sun Belt name in the cohort, sits in the middle at 62 — its lower per-unit value is offset by the cohort’s lowest per-unit recurring capex. The intuitive “older portfolios drag more” story doesn’t survive the percentage translation. Dollar capex tracks vintage and structure type. Yield drag tracks per-unit value, which moves with the cap rate the market assigns. Cheaper per-unit values amplify the same-dollar maintenance line. Recurring capex is a duration drag on the equity, but the drag is bigger when the per-unit base is smaller. Cheaper buildings carry their maintenance line less efficiently.
What drives the dispersion
Two layers, different drivers.
Dollar layer — vintage and structure. Per-unit capex tracks portfolio age and building type. Roofs are thirty-year goods, HVAC is fifteen to twenty, structural systems run decades but their failures are expensive when they come. High-rise coastal buildings have elevators, structured parking, zoned HVAC, and amenity spaces that cost more per unit to maintain than a garden-style Sun Belt walk-up. Dollar capex per unit is a portfolio-composition number.
Percentage layer — per-unit value. What turns the dollar dispersion into a yield haircut is the denominator. Coastal buildings cost more per door to maintain and also more to own; the higher per-unit value absorbs the higher capex spend. CPT’s $1,857 against $262K per-unit EV is a 71 bp haircut; ESS’s $2,109 against $428K is 49 bps. Percentage drag converges in a range that dollar comparison doesn’t predict.
Disclosure conventions cut across both. Each REIT decides where to draw the line between recurring capex (pulled from AFFO) and growth or redevelopment capex (excluded from AFFO denominator). UDR discloses “Total Recurring Cap Ex” and “NOI Enhancing Cap Ex” as separate lines (Q1 2026 supplemental, Attachment 12). CPT runs a “Reposition Expenditures” program — $21.5M in Q1 2026, against $16.2M of recurring capex over the same period (CPT 1Q26 supplemental, p. 20). Reposition is currently 1.33x the recurring line; reclassifying any portion across the divide moves the AFFO denominator meaningfully. EQR’s AFFO bridge has historically carried a larger redevelopment / value-add bucket relative to recurring than the trio (MAA, UDR, CPT — the three names I’ve been tracking since Issue 01), though the specific classification line moves quarter to quarter. The footnote tells you which way each name bends.
What this means in practice
Ranking REITs against each other: the comparable number is capex-adjusted FCF yield, not cap rate. EQR and CPT end up within 2 bps of each other despite a 14 bp gap in headline cap rate. Two REITs with identical cap rates aren’t pricing equivalent economics if their per-unit capex profiles differ. Strip the maintenance line out, normalize by per-unit value, and the comparison gets honest.
Comparing public to private: anchor against your underwriting’s per-unit capex assumption, not the NOI margin. The comparison only works if you know which line you’re comparing. LP underwriting typically separates replacement reserves ($250 to $400 per unit per year — what lenders require) from full recurring capex ($500 to $900 per unit when underwritten honestly). The public REIT supplemental number is closer to the second. Comparing a private deal’s reserve line to a public REIT’s recurring line is the apples-vs-oranges this piece warns about. Run capex-adjusted FCF yield on both sides — but check that what you’re calling “recurring” matches what the REIT calls recurring before reading the gap. Some of the gap is categorization. Some of it is genuine under-allocation that catches up at refinance or reversion. For the comparison to a public REIT to mean anything, anchor against the $500-$900 line, not the reserve.
Modeling forward dividend yields: the question isn’t “how old is the portfolio” but “how much per-unit value is each capex dollar spread across.” Higher per-unit values absorb maintenance more efficiently. Lower per-unit values feel every capex dollar more sharply. The duration dynamic is about denominator, not age. Watch the per-unit value, not just the buildings.
What would change my mind
Two specific claims: (1) capex-adjusted FCF yield is the right comparable across the multifamily REIT universe and the rank-spread compresses meaningfully when you compute it; (2) the percentage haircut clusters in a 22 bp band because per-unit value and per-unit capex move together. What would flip either:
- Standardized capex disclosure. If NAREIT or the SEC defines “recurring capex” with teeth — same threshold, same categorization — the haircut dispersion either tightens further or reveals itself as something other than what I’m calling it. Right now the data are partly real and partly artifact of choice.
- Per-unit value diverges from capex. The convergence depends on per-unit value tracking per-unit capex across the cohort. If coastal per-unit values fall (cap rates widen at ESS/EQR specifically) without a commensurate move in capex, their haircuts widen and the order shifts. A 50+ bp cap rate widening at EQR or ESS without a capex disclosure change is the signal.
- The categorization itself moves. If CPT reclassifies its Reposition program out of growth capex, or EQR shifts NOI-Enhancing into recurring, reported numbers move without underlying economics changing. Q2 supplementals — late July — are the first chance to see it.
What we’re watching next
Q2 2026 supplementals drop in late July. Three lines to read:
- Per-unit recurring capex trajectory. The trio should stay clustered in dollar terms; ESS and EQR likely keep printing higher. The harder question is the percentage haircut. If Sun Belt per-unit values compress relative to coastal (trio cap rates widen more than the coastal pair), the inversion sharpens. If the gap narrows, it softens.
- EQR’s redevelopment categorization. EQR has historically carried a larger redevelopment / value-add bucket relative to recurring than the trio. A visible reclassification in Q2 — or a shift in the AFFO bridge — moves the comparison.
- Sell-side pressure on capex-adjusted yields. Almost no analyst frames dividend coverage this way. If one starts, the disclosure equilibrium moves. Slow process; worth tracking.
—
Methodology + sources
Implied cap rate: annualized same-store property NOI ÷ enterprise value (shares × 2026-05-04 close + total debt from most recent 10-Q, cash excluded). Trio cap rates here (MAA 6.13%, UDR 5.87%, CPT 6.10%) are recomputed on refreshed inputs and differ materially from Issue 01’s published numbers; same methodology, actual Q1 2026 same-store NOI run-rate instead of the stylized extrapolation used in Issue 01. Issue 02 numbers are the standing reference going forward.
Recurring capex per unit: extracted from each REIT’s Q1 2026 supplemental on a trailing-four-quarter basis. Each REIT’s own definition of “recurring” is preserved — definition variance is part of the thesis, not a bug. Per-unit denominator follows each company’s own disclosure convention.
| REIT | Q1 2026 8-K | Accession |
|---|---|---|
| MAA | filed 2026-04-29 | 0001193125-26-191620 |
| UDR | filed 2026-04-29 | 0000074208-26-000045 |
| CPT | filed 2026-04-30 | 0001628280-26-028942 |
| EQR | filed 2026-04-28 | 0001193125-26-187339 |
| ESS | filed 2026-04-28 | 0001140361-26-017477 |
Trailing four-quarter primary because Q1 capex is seasonally low; TTM smooths the run-rate.
The structured dataset, extended this issue with EQR and ESS coverage, is maintained internally with full per-record source provenance — URL, filing accession, page or section, raw quoted text, sha256, extractor version. Trustworthiness rated 8.5/10 in independent redteam audit. Every number cited traces to its source supplemental page in under 60 seconds.
What this is not: Investment, tax, or legal advice. A recommendation to buy or sell any specific REIT. A complete analysis of multifamily REIT capital allocation — subsequent issues will continue the thread.