What is the difference between an agency and a hybrid mortgage REIT?

Agency mREITs (Annaly, AGNC) own mortgage bonds guaranteed by Fannie Mae, Freddie Mac or Ginnie Mae, funded with short-term repo borrowing and hedged with swaps. Credit risk is close to absent, so they run leverage of roughly five to seven times equity on their own reported measures, and what is left is interest rate and prepayment risk. Hybrid and CRE lenders (Rithm, Starwood, Blackstone Mortgage) take real credit risk on loans, servicing rights and origination, at far lower leverage. The same mREIT label covers three different risk books.

How do agency mortgage REITs make money?

They borrow short via repurchase agreements, buy longer-duration agency MBS, and hedge rate risk with swaps and TBAs. Profit is the net interest spread between asset yield and funding cost, amplified by leverage, minus hedge drag and operating expense. FY2025 spreads: NLY ~1.45% average (Q4 1.49%), AGNC 1.92% full year. A spread under two per cent only builds a business at leverage: assets near eight times equity turn 1.92% into roughly 15% on equity before costs, and turn a 1% fall in bond prices into 8% off book value.

Why do hybrid mREITs trade at different multiples than agency names?

Agency books are valued on price-to-book against BVPS or tangible net BVPS (NLY ~1.09× and AGNC ~1.23× as of 10 June 2026). Hybrid Rithm (~0.73×) and CRE lenders Starwood (~0.94×) and Blackstone Mortgage (~0.89×) carry credit, servicing and office-exposure risk that agency P/B bands do not capture. A discount is not cheapness: an mREIT marks its book to market every quarter, so a low multiple says the market expects those marks to fall further.

Should I screen agency and CRE mortgage REITs with the same metrics?

No. Agency screens centre on BVPS trajectory, economic return, net interest spread, leverage definition, and CPR (the speed at which the mortgages behind the bonds are paid off early). CRE lenders need distributable earnings coverage, non-accrual balances, office mix, and loan-book marks. A single dividend-yield rank mixes machines that share a tax status but not a risk factor.

Agency vs Hybrid Mortgage REITs: Two Different Machines

Educational analysis for professional use. This guide is not investment advice or a recommendation to buy or sell any security, and it is not personalised.

The Label Hides Three Different Machines

A mortgage REIT owns no buildings. It owns mortgage debt: either bonds backed by pools of American home loans, or loans it has written itself against commercial property. It shares a tax status with the rent-collecting REITs and almost nothing else, and the three kinds of mortgage REIT do not much resemble each other either.

Agency names such as Annaly and AGNC run a borrowed-money carry trade on government-backed mortgage bonds. Rithm Capital earns from writing loans, administering them, and the value of the servicing contracts it owns. Starwood Property Trust and Blackstone Mortgage Trust write commercial property loans and absorb the losses when a borrower fails. Screening them as one sector is how investors misread leverage, book value and dividend safety.

What moves the quarter differs in each case. Agency earnings turn on rates, the spread earned over funding, and whether the hedges worked. For the commercial lenders it is loans going bad, office collateral, and whether cash earnings covered the dividend. Rithm sits between the two: its servicing assets gain value when rates rise, which is the opposite of what a bond book does.

Agency Carry: A Guarantee Bought with Leverage

“Agency” means Fannie Mae, Freddie Mac or Ginnie Mae. Each guarantees the principal and interest on the mortgage bonds it issues, so when the homeowners behind a pool default, the bondholder is still paid. Credit risk, the thing that ruins most lenders, is close to absent.

Everything else remains. A bond paying a fixed coupon falls in price when yields rise. And borrowers refinance whenever it suits them, so the bonds pay down at a speed nobody controls, which is prepayment risk.

The guarantee is also what makes the leverage possible, and the leverage is the business. An agency book earns a net interest spread of one to two per cent: the yield on the bonds, less the cost of funding them, after hedges. Nobody builds a listed company on 1.92% of their own money. So the REIT borrows against the bonds through repo (repurchase agreements, short-term borrowing secured on the bonds and rolled every few weeks), and AGNC’s 7.2× at-risk leverage puts its assets at roughly eight times equity. That turns a 1.92% spread into something near 15% on equity before costs. Lenders will fund a book that size because the collateral is guaranteed paper they can sell in an afternoon.

Read it the other way and you have the risk. If assets are eight times equity, a 1% fall in the value of the bond book takes 8% off book value per share. Agency mortgage REITs are the names in this sector that come apart on rate moves, and the guarantee is the reason: because the credit risk is small, they run the leverage that makes rate risk large. The commercial lenders take real credit risk at a fraction of that leverage. Neither one is the safe one.

Annaly and AGNC are the reference machines. Both fund agency MBS with repo, hedge duration with swaps and TBAs (forward contracts on mortgage bonds, used to hedge and to hold exposure without owning the bonds), and file book value per share every quarter. The market prices them on price-to-book, not on property values.

Metric	Annaly (NLY)	AGNC Investment (AGNC)
Market cap (10 Jun 2026)	~$15.9B	~$11.8B
Book value basis	Book value per common share	Tangible net BVPS (excludes goodwill)
End-FY2025 BVPS	$20.21 (+5.5% from $19.15)	$8.88 (+5.6% from $8.41)
FY2025 economic return	20.2%	22.7% on tangible common equity
Net interest spread	~1.45% FY avg (Q4 1.49% ann.)	1.92% FY2025
Leverage (company-defined)	5.6× economic leverage	7.2× tangible net BV at-risk
P/B (10 Jun 2026, on 31 Mar 2026 book)	~1.09× ($21.66 ÷ $19.82)	~1.23× ($10.29 ÷ $8.38)

Agency investors watch three lines: the spread, the leverage ratio, and the path of book value per share. Annaly’s 5.6× and AGNC’s 7.2× are built slightly differently, but not differently enough to explain the gap. AGNC really is running the harder book, against pure agency paper, while Annaly also holds residential credit and servicing rights that cannot be levered the same way; the leverage definitions guide takes both apart. AGNC’s monthly dividend cadence makes it the clean monthly-payout screen; NLY is the scale benchmark.

The spread between mortgage bonds and Treasuries drives both halves of the result. AGNC put the current-coupon MBS spread at 124 bps over a 50/50 blend of 5- and 10-year Treasuries on 31 March 2026, against a peak near 160 bps in April 2025. Wider is better for money invested next quarter and worse for the book already owned, because the bonds already held mark down as the spread moves out.

Hybrid: Servicing Rights, Not a Bond Book

Rithm Capital is the listed hybrid contrast, and it is not an agency carry trade in different clothes. Much of its value sits in mortgage servicing rights: the right to collect a small annual fee for administering somebody else’s mortgage, for as long as that mortgage survives.

Metric	Rithm Capital (RITM)
Market cap (10 Jun 2026)	~$5.16B
Book value per share (31 Dec 2025)	$12.66
P/B (10 Jun 2026)	~0.73× ($9.28 ÷ $12.66)
FY2025 earnings available for distribution	$2.35/sh
FY2025 dividend	$1.00/sh (2.35× covered)
Servicing portfolio	$852B of loan balances, incl. $256B serviced for other owners
Servicing rights owned	~$596B of loan balances, carried at $10.4B fair value
Origination & Servicing assets	$27.5B of $53.1B total assets

A servicing right is a bet against refinancing. It pays a fee for every year the loan stays alive, so it gains value when rates rise and borrowers stop refinancing, and loses value when they start. That is the opposite direction from a long-duration bond book, which is why Rithm’s quarter often reads the other way from Annaly’s. Rithm does not publish an economic leverage line comparable to NLY’s or AGNC’s in its FY2025 materials, so do not force a leverage rank across the two.

CRE Lending: Credit Spread and Collateral Risk

Starwood and Blackstone Mortgage write loans against offices, hotels, flats and warehouses. They earn a credit spread and they take the losses. Their reported earnings measure is distributable earnings, a cash-basis figure that stands in for the agency spread, and the question each quarter is whether it covered the dividend.

Metric	Starwood (STWD)	Blackstone Mortgage (BXMT)
Market cap (10 Jun 2026)	~$6.38B	~$3.10B
GAAP BVPS (31 Dec 2025)	$18.34	$20.75
P/B (10 Jun 2026)	~0.94×	~0.89×
FY2025 distributable metric	$1.69/sh DE	$(1.43)/sh distributable EPS
FY2025 dividend	$1.92/sh (0.88× covered)	$1.88/sh (uncovered after charge-offs)
Commercial lending book	$16.6B	$17.8B loans receivable, net
Office exposure basis	8% of $30.7B undepreciated assets (US office)	27% of net loan exposure

Office percentages are not comparable without reading the denominator. Starwood’s 8% is measured across the whole company’s undepreciated assets; Blackstone Mortgage’s 27% is on its net loan exposure alone. And BXMT’s FY2025 distributable EPS was $1.86 before charge-offs (loans written off as uncollectable) but $(1.43) after, against a $1.88 dividend. The distance between those two figures is the entire credit story.

CRE book value still uses price-to-book; see book value per share for the scoreboard frame on GAAP BVPS.

Worked Mini-Example: Same Sector, Different Screens

An investor ranks five mREITs by dividend yield alone and buys the top three. Using FY2025 filed data:

Company	Dividend/sh	Price (10 Jun 2026)	Indicative yield	Coverage metric
NLY	$2.80	$21.66	12.9%	Economic return 20.2% (BVPS + dividends)
AGNC	$1.44	$10.29	~14.0%	Economic return 22.7%
RITM	$1.00	$9.28	10.8%	EAD coverage 2.35×
STWD	$1.92	$17.22	11.2%	DE coverage 0.88×
BXMT	$1.88	$18.41	10.2%	Uncovered ($(1.43) DEPS)

The top three by yield are AGNC, NLY and STWD. Two of those are in good shape and one is not, and the yield column cannot tell you which. Starwood paid $1.92 against $1.69 of distributable earnings. Blackstone Mortgage, which the yield screen happens to reject, is in worse shape still: $1.88 paid against negative distributable earnings after charge-offs.

The deeper problem is that a dividend is not a return. The return is the dividend plus the change in book value, which is why these companies report economic return and why the agency names’ 20%-plus figures matter more than their yields: they paid double-digit dividends and grew book at the same time. A 14% dividend paid while book value falls 10% is a 4% year, and mREIT dividends are regularly funded out of book. Run the screen the other way and Rithm’s 0.73× book is not automatically cheap either. An mREIT marks its book to market every quarter, so a discount is the market saying it expects those marks to fall, not that it has overlooked something.

How to Split Your Workflow

Agency names (NLY, AGNC): start with book value per share, then leverage definitions, because the two filers do not mean the same thing by leverage. Set economic return history against the price-to-book you are being asked to pay.

Rithm needs a segment read: origination and servicing against the fair value of the servicing rights, with EAD coverage in place of an agency spread.

CRE lenders (STWD, BXMT): distributable earnings coverage, office mix on the stated basis, and non-accrual balances (loans where the borrower has stopped paying and the lender has stopped booking interest), each on the denominator the filing actually uses.

Equity REIT screens (REIT screening checklist) run on FFO, AFFO and cap rates. None of those exist here. Mortgage REITs need the spread-and-book framework in the Mortgage REIT Sector Primer.

One Label, Three Answers

Work out which of the three machines you are looking at before you compare it to anything. Dividend yield is the one number all five publish, and the one that tells you least.