mREIT Leverage: Economic vs At-Risk Definitions

NLY at 5.6× and AGNC at 7.2× Are Not the Same Ratio

Agency mortgage REITs run levered books, and the multiple on the earnings release is not comparable across filers unless you read the footnote. Annaly (NLY) reports 5.6× economic leverage. AGNC Investment (AGNC) reports 7.2× at-risk leverage. The numerators genuinely differ, and Annaly’s picks up Treasury financing that AGNC’s leaves out. The denominators barely do. So most of the 1.6 turns is real: AGNC runs a harder book against a pure agency portfolio, while Annaly also holds residential credit and servicing rights that cannot be levered the same way.

There is a second error underneath it. The ratio is not the risk. What destroys book value at these companies is the funding mechanism behind the ratio, and a lower number tells you nothing about that on its own.

Two Definitions, Side by Side

Item	NLY economic leverage	AGNC at-risk leverage
End-FY2025	5.6×	7.2×
FY2025 range	5.6–5.8× (5.7 Mar, 5.8 Jun, 5.7 Sep, 5.6 Dec)	7.2–7.6× (7.5 Mar, 7.6 Jun, 7.6 Sep, 7.2 Dec)
Numerator	Recourse debt + TBA cost + net forward purchases	Repo + consolidated vehicle debt + net TBA at cost
Denominator	Total equity, including preferred	Total stockholders’ equity less goodwill

The denominators are nearly the same, so they are not where the gap comes from. Annaly’s numerator is the broader of the two, because it picks up Treasury financing that AGNC’s measure leaves out, which if anything understates the difference. The gap is mostly what each company owns.

NLY runs roughly 1.6 turns lower than AGNC on these company lines (7.2 − 5.6), or about 22% less levered (5.6 ÷ 7.2). Take that at face value: rebuilding both on one basis would not reverse it.

Repo, Haircuts and the Margin Call

Repo, short for a repurchase agreement, is how the whole thing is funded. The REIT sells an MBS to a lender today and agrees to buy it back in 30 to 90 days at a fixed price. It is a secured loan with the bond as collateral, and the difference between the two prices is the interest.

Three features of that arrangement drive everything else on this page.

The lender advances less than the bond is worth. The shortfall is the haircut, a few per cent on agency collateral. That shortfall is the equity the REIT has to fund itself, which is why a book of high-quality, low-haircut collateral can be levered so many times over.

The loan has to be rolled. A book funded on 30-day repo refinances itself roughly twelve times a year, at whatever rate lenders charge on the day. Lenders can also decline to renew.

The collateral is marked daily. This is the part that matters. When MBS prices fall, the collateral no longer covers the loan, and the lender issues a margin call for cash or additional bonds, usually settling the same day. The REIT meets it from its liquidity buffer. If the move is large or the buffer is thin, it meets it by selling MBS into the market that has just fallen.

That last step is the loop. Selling turns a mark into a realised loss the book can never recover, and because every levered holder is being called on the same day, the selling itself pushes prices lower and sets off the next round of calls. Leverage sets how far a given price move travels into equity, but the forced sale is what makes the damage permanent. In March 2020 several mortgage REITs could not meet agency repo margin calls and liquidated most of their portfolios; some never rebuilt them.

What Enters the Numerator

Component	Role in the book	Effect on leverage and spread
Agency MBS	Interest-earning assets	Duration and spread driver
Repo	Short-term secured funding	Liability cost; the main leverage line
TBA positions	Forward contracts on generic MBS, used to hedge or to hold exposure	Enter both companies’ numerators
Interest rate swaps	Rate hedge	Cut rate-driven book moves; cost spread
Net forward purchases	Trades agreed but not yet settled	Part of NLY economic leverage

TBA stands for to-be-announced: a contract to buy or sell agency MBS at a future date, with the exact pools named later. It is a forward, so it sits in derivatives rather than in debt, and a plain balance-sheet debt-to-equity figure misses it entirely. A REIT holding most of its exposure through long TBAs would look barely levered on GAAP debt while carrying the same interest rate and spread risk as one holding the bonds outright. Economic and at-risk leverage both pull that exposure back in, which is why they are the lines to read and a debt-to-equity ratio you build yourself is not.

Hedging shows up in the spread rather than as its own line. AGNC’s FY2025 net interest spread was 1.92%, after TBA and swap costs and excluding catch-up premium amortisation. Annaly’s FY2025 spread averaged about 1.45% on the same exclusion basis (1.35%, 1.47%, 1.50%, 1.49% by quarter). Hedge drag is the price paid for a steadier book when rates and prepayment speeds move.

The 7–8× Range, and What It Does Not Tell You

For screening, agency books typically sit around 7–8× on at-risk-style metrics. AGNC’s 7.2–7.6× fits. Annaly’s 5.6–5.8× sits below it because Annaly is genuinely running a lighter book against a mixed portfolio, not because the definition flatters it.

Leverage style	FY2025 range	Applies to
At-risk (AGNC-style)	7.2–7.6×	AGNC, and the 7–8× screening range
Economic (NLY-style)	5.6–5.8×	NLY, unless you rebuild peers on it
Debt-to-equity you build yourself	Not filed	Misses TBA exposure; do not substitute

The range only holds for agency paper. Agency MBS carry a guarantee from Fannie Mae or Freddie Mac, so the holder takes rate and prepayment risk but almost no credit risk, and lenders accept them at small haircuts. Nothing about 7× transfers to a book of commercial mortgage loans, where the borrower can default and the collateral has no guarantee behind it. A hybrid or CRE lender at 3× can be carrying more risk of permanent loss than an agency book at 8×. Lower leverage is only safer when the thing being levered is the same.

Rithm Capital, the listed hybrid, does not publish an economic leverage line comparable to Annaly’s or AGNC’s, so there is no like-for-like leverage rank across the sector to be had. Read its segments instead.

Book Value Sensitivity at Leverage

Leverage amplifies mark-to-market moves into book value per share. Wider MBS spreads cut asset values whatever Treasury yields do, and the loss lands in equity at the leverage multiple.

Spreads move enough to matter. AGNC reported the current-coupon MBS spread at 124 bps on 31 March 2026, well in from a stress peak near 160 bps in April 2025. That narrowing went the REITs’ way, and the same 36 bps in the other direction, at 7× leverage, would take a large bite out of book rather than trimming a quarter of carry.

The chart shows the book value hit as a percentage of equity across spread widenings and portfolio durations at 7× leverage. The marked point, a 25 bp widening on a five-year book, lands near 10% of equity.

Worked Mini-Example: 25 bp × 5 Years × 7× Leverage

Illustrative inputs, chosen for round arithmetic rather than to describe any filer:

Input	Value
Common equity	$10.0B (500M shares × $20.00 BVPS)
At-risk leverage	7.0×
MBS exposure	$80.0B (equity plus 7× borrowings, TBAs included)
Net interest spread	1.50%
Operating expense	$150M/year
Portfolio duration	5.0 years

Core earnings: $80.0B × 1.50% − $0.15B = $1.05B, or $2.10/sh on 500M shares.

Dividend: $1.90/sh, so coverage of about 1.10× on core earnings.

Book value mark from a 25 bp spread widening: 0.25% × 5.0 years × $80.0B = $1.0B, which is 10% of the $10.0B equity.

A spread move that small, over a few sessions, wipes out roughly a year of spread earnings. It also shrinks the collateral behind the repo book by $1.0B, which is when the margin calls arrive. Faster prepayment shortens effective duration and changes the arithmetic mid-quarter, usually in the direction of a smaller mark and a lower reinvestment yield.

What Hedging Fixes, and What It Does Not

Swaps hedge the wrong half of the problem for the example above. They protect against Treasury yields moving. They do nothing about the spread of mortgages over Treasuries, so if yields sit still and MBS cheapen by 25 bp, the full loss reaches book value. That gap between the exposure and the instrument hedging it is basis risk, and on an agency book it is the main uninsured exposure.

Which is why a hedge ratio near 100% does not mean the book is protected. It means the notional value of the hedges is close to the notional value of the funding. It says nothing about spread risk. It also assumes the hedge and the MBS move together in duration, and they do not: mortgages extend when rates rise and shorten when rates fall, always the unhelpful way round, so a hedge sized for today is short of the mark in a sell-off and too large in a rally.

Read leverage alongside the spread level and its trend, the gap between experienced and projected prepayment speed, and the book value path. Those tell you whether the hedge book actually delivered. For how the underlying businesses differ, see agency versus hybrid mortgage REITs.

When the Leverage Number Falls

A falling ratio is usually read as prudence, and sometimes it is. It can also be the aftermath: a REIT that sold assets to meet margin calls repays repo as it goes, and reports lower leverage next quarter on a smaller book. So read the ratio beside book value per share. Leverage down with book holding is a choice. Leverage down with book down is a consequence.