Skip to content

TBA and the Dollar Roll

One-line definition

"TBA" (to-be-announced) is how the agency MBS market trades pass-throughs without specifying which exact pools will be delivered — a dollar roll is a short-term financing trade built on top of that convention: sell TBA for near-month settlement, buy back a "substantially similar" TBA for far-month settlement, usually at a lower ("dropped") price.

Vocabulary

  • TBA (to-be-announced): a forward contract on generic agency MBS — issuer (Fannie Mae/Freddie Mac/Ginnie Mae), coupon, and settlement month are specified; the exact pools are announced only shortly before settlement (per SIFMA's "48-hour rule"). This is what makes agency MBS as liquid as it is — a buyer doesn't need to pick specific pools to trade the market.
  • The roll: selling a TBA position for near-month settlement and simultaneously buying the same coupon/program back for far-month settlement — the seller doesn't actually need to own or deliver a specific pool at any point if they roll before settlement.
  • The drop: near_price - far_price — the amount the far-month price is below the near-month price. A positive drop is normal (the buyer of the far leg is compensated for the coupon income they won't collect in the interim); an unusually large drop signals collateral scarcity or strong dealer demand to be short the near-month settlement.
  • Roll specialness: how much cheaper (or more expensive) financing via the roll is compared to the prevailing GC repo rate for the same collateral — see repo and financing for the Treasury-market version of the same idea.

Formula

implied_financing_rate = (coupon_income - drop_income) / near_amount / horizon_years

  • coupon_income: the coupon interest given up over the roll period (not annualized) — approximately face * coupon_rate / 12 for a standard monthly roll
  • drop_income: the drop converted to dollars — (near_price - far_price) / 100 * face
  • near_amount: near-leg proceeds, near_price / 100 * face
  • horizon_years: the roll period in years (a standard monthly TBA roll is 1/12)

pm.fixed_income.mbs.dollar_roll_implied_financing_rate implements this. It is a simplified version of the real calculation, which also nets reinvestment income earned on the sale proceeds and any scheduled principal paydown given up during the roll period — see Limitations.

Why PMs care

The dollar roll is one of the cheapest, most liquid ways to finance an MBS position — for a levered MBS investor (a bank, a mortgage REIT, a relative-value fund), rolling instead of financing in cash-and-repo can be materially cheaper when the roll is "special." Reading a roll's implied financing rate against GC repo tells you whether the roll is cheap (special — attractive financing, but a signal of collateral scarcity) or expensive (better to hold and finance in repo instead). Because implied financing rate embeds a prepayment-speed assumption (via the coupon income and principal given up), it is also a read on the market's collateral-scarcity and prepayment expectations, not purely a funding-cost number.

Worked example

A $1,000,000 face, 5% coupon TBA pool: near price 101.00, far price 100.625 (a 0.375-point drop), rolled for one month (horizon_years = 1/12).

  • coupon_income = 1,000,000 * 0.05 / 12 = $4,166.67
  • drop_income = 0.375 / 100 * 1,000,000 = $3,750.00
  • near_amount = 101.00 / 100 * 1,000,000 = $1,010,000
  • implied_financing_rate = (4,166.67 - 3,750.00) / 1,010,000 / (1/12) ≈ 0.495%

If GC repo for agency MBS is trading around 5.30%, this roll is deeply "special" — financing at ~50bp is far cheaper than repo, telling a PM that this coupon/program is in unusually high demand to be short for delivery (tests/test_mbs.py::test_dollar_roll_implied_financing_rate_hand_example computes this exact figure).

Common mistakes

  • assuming a positive drop always means cheap financing — a large drop can still be expensive financing if the coupon given up is even larger (see the "zero-drop" and "drop exceeds coupon" test cases in tests/test_mbs.py); it's the ratio to coupon income that matters, not the drop in isolation
  • treating "the pool" as if it's the same pool across the roll — TBA delivery only guarantees substantially similar collateral (same issuer/coupon/program), not the identical pool, which is precisely what distinguishes a dollar roll from a repurchase agreement
  • ignoring that implied financing rate is a blended read on financing and prepayment expectations — a change in the roll's implied rate isn't purely a funding-market signal

Limitations

This repo's formula omits two components the real calculation includes: reinvestment income on the near-leg sale proceeds over the roll period, and the value of scheduled principal paydown given up (not just coupon interest) — both are typically small relative to coupon income and drop for a par-priced pool but can matter for premium/discount pools or longer rolls. See specified pools for how pool-level prepayment characteristics feed into this same calculation in practice.

Free resources