Treasury carry after the funding bill

Sep 2026 · 5 min read · carry & roll, repo, treasury bills

Independent research note. The repo chart uses 2024 observations. Bond and bill examples use hypothetical inputs.

A Treasury position can earn coupon income and still have negative carry after financing. It can also appear attractive under an unchanged-curve assumption while leaving little protection against a yield increase. Both effects belong in the trade calculation before forming a view on relative value.

Consider a near-par Treasury yielding 4.20%, funded at an annualized 4.60%, with modified duration 4.4. Over a quarter-year, the income-minus-funding approximation is −10 bp of starting market value. If the bond rolls to a point on the curve with a yield 10 bp lower, the first-order price benefit is +44 bp. Together, that is +34 bp before transaction costs.

The roll assumption provides the entire positive return in this example.

Keep income, roll and a market shock separate

For a near-par bullet bond, a useful first-pass estimate is:

R_{\mathrm{excess}}\approx(y-f)h -D_{\mathrm{mod}}\Delta y_{\mathrm{roll}} -D_{\mathrm{mod}}\Delta y_{\mathrm{shock}}.

Here, yields and yield changes use decimal units; $h$ is the horizon in years. Excess return means return after the assumed financing cost, measured against starting bond market value. It is not a return on the investor's haircut or equity capital.

The roll change is the yield at the bond's shorter remaining maturity on today's curve minus its initial yield. An upward-sloping local curve can create positive roll return. This assumes the remaining-maturity curve stays unchanged over the horizon. A forward-curve scenario is a different assumption and can produce a different result.

Hypothetical case	Net income	Roll benefit	Total excess return
Base assumptions	−10 bp	+44 bp	+34 bp
No roll benefit	−10 bp	0 bp	−10 bp
Funding 100 bp higher	−35 bp	+44 bp	+9 bp
Additional 25 bp yield rise	−10 bp	+44 bp	−76 bp

The final row includes a −110 bp price effect from the additional yield shock. Under these assumptions, a rise of about 7.7 bp beyond the roll scenario would consume the base case's 34 bp cushion. That is a more useful risk comparison than quoting the annual yield alone.

Yield times horizon is an income proxy, not exact coupon accounting. A production calculation should reprice the remaining cash flows, include coupon receipts and accrued-interest changes, and subtract financing on the actual cash balance using the relevant day count. It should also include haircuts, coupon reinvestment and execution costs. A full horizon repricing already captures roll; adding a separate roll estimate would double-count it.

Use SOFR to monitor funding conditions

The New York Fed defines SOFR as a broad measure of overnight cash borrowing costs secured by Treasury collateral. It calculates the rate as a volume-weighted median of eligible repo transactions, with filtering to reduce the influence of specials.

A useful public monitor compares SOFR with the Fed's interest rate on reserve balances, or IORB. IORB provides a policy reference for eligible institutions' reserve holdings; it is not a funding rate available to every investor.

Daily SOFR minus IORB during 2024. The chart shows variation around zero and marks the November 25 SOFR methodology change. Open chart at full size.

On September 30, 2024, SOFR was 4.96% and IORB was 4.90%, a +6 bp difference. On December 31, the corresponding rates were 4.49% and 4.40%, a +9 bp difference. These observations justify checking financing conditions near reporting dates. They do not establish the cause of the moves or the rate a particular desk could obtain.

The New York Fed changed SOFR's methodology on November 25, 2024, including the treatment of affiliated trades and the low-rate volume trim in the cleared bilateral segment. The chart marks that date because a time-series analysis should record benchmark-definition changes alongside economic events.

SOFR also describes an overnight market. It does not lock three months of financing. For a particular Treasury, the relevant repo rate depends on the security, term, counterparty and balance-sheet conditions. Special collateral can finance below general collateral rates; obtaining that security for a short position can become more costly.

Compare bill yields on the same basis

A bill's bank-discount quote uses face value as its denominator and a 360-day year. An investment return uses the amount paid. Mixing those conventions can create an apparent spread before any economic difference exists.

For a hypothetical 90-day bill quoted at a 4.50% bank-discount rate:

P=100\left(1-0.045\frac{90}{360}\right)=98.875.

The annualized investment rate on an ACT/360 basis is:

\left(\frac{100}{98.875}-1\right)\frac{360}{90} \approx4.5512\%.

On a 365-day basis, the corresponding simple annualized rate is about 4.6144%. Treasury's bill calculation examples explain these conventions; longer bills require additional care when calculating coupon-equivalent yields.

Even after putting a bill and repo rate on the same day-count basis, a 90-day bill yield and one day's SOFR still have different horizons. A financed-bill comparison needs either locked term funding or a scenario for the daily funding path through maturity.

Make the financing assumption visible

A useful research table should show gross income, roll, financing and a rate shock separately. For the stylized Treasury here, 44 bp of assumed roll offsets 10 bp of negative financed income. A modest adverse rate move can erase the residual. That result makes the next research question concrete: how credible are the unchanged-curve and funding assumptions over the intended holding period?

The Python script generates the carry scenarios, bill conversions and funding observations. The research package documents units, missing-data treatment and the distinction between observation dates and publication times.

Related: Agency and SSA relative value.