The Treasury Basis Trade and the Price of Balance Sheet

Primary dealers held about 42 billion dollars of net outright Treasury positions on an average week in 2015. In 2026 the average is close to 497 billion. Over the same period the repo financing dealers raise against Treasury collateral more than doubled, from about 1.3 trillion dollars to about 3.0 trillion. Both series come from the same weekly New York Fed collection, and together they describe the balance sheet that a cash-futures basis position ultimately rents.

The basis itself cannot be shown here. Constructing a clean cash-futures series needs licensed futures settlement data and CUSIP-level cash quotes, and a reconstruction assembled from what is free would look like a measurement without being one. Figure 1 shows the two quantities that are properly observable instead: how much balance sheet is committed to the Treasury market, and what that capacity costs when it becomes scarce.

Section 01What the trade actually is

In the Federal Reserve's description, the trade combines a short Treasury futures position, a long Treasury cash position, and repo borrowing that finances the cash leg and supplies the leverage. The asymmetry is the point. Only the cash leg needs funding: the fund buys the bond, pledges it in repo, and borrows most of the purchase price, while the futures leg absorbs margin rather than principal. An unlevered basis of a few basis points becomes a respectable return only after that borrowing. The thinness of the spread is what forces the leverage, and the leverage is what makes the position matter to anyone other than its owner.

Set against the size of the market being intermediated, the committed inventory stays small. Dealer net Treasury positions at the latest reading are about 1.4 per cent of the 31.5 trillion dollars of marketable Treasury debt outstanding. A record inventory in absolute terms is still a thin cushion relative to what has to clear through it.

Section 02The price of balance sheet

Every component of the structure is a charge for someone's balance sheet. Repo financing has to be rolled, often overnight, at a rate that can move. The pledged cash leg carries a haircut that can be raised. The futures leg carries initial and variation margin that rises when prices move. The Federal Reserve note is explicit that the trade is generally highly leveraged and exposed both to changes in futures margins and to changes in repo spreads. Those two exposures are the real position: the basis pays for warehousing a bond, financing it, and standing ready to meet margin.

Most of the time that rent is invisible, which is why the lower panel of Figure 1 uses a logarithmic scale. It plots the gap between the 99th and 1st percentiles of the transactions underlying SOFR, published daily by the New York Fed. When the market is functioning, identical overnight cash against identical collateral trades within a narrow band: the median reading is 16 basis points, and the 95th percentile is 29. When capacity binds, the same measure reaches 355 basis points at year-end 2018, 675 in the week of 17 September 2019, and 180 in mid-March 2020. A price that sits at 16 for years and then prints 675 is not a price that anyone was monitoring as a risk.

This is how a trade in the safest asset in the world turns from stabilising to destabilising with no change in the bonds. When funding tightens or margins jump, a levered holder must find cash it may not have, and the fastest route is to unwind: sell the cash Treasury, buy back the future. When many funds hold the same position they sell into one another. The Federal Reserve note records that hedge funds unwinding the cash-futures basis likely contributed to the March 2020 Treasury market instability, and that absent prompt central-bank intervention the situation may have been considerably worse.

net_basis_pnl  ~=  convergence + bond_carry
                   - repo_cost - transaction_costs

cash_needed_in_stress  ~=  variation_margin
                          + haircut_increase
                          + repo_rollover_gap

return_on_equity  =  net_basis_pnl / committed_equity

Section 03Clearing rewrites the plumbing

If the trade is a position in balance sheet and funding, clearing rules are part of its economics. The SEC's Treasury clearing standards require covered clearing agencies to bring eligible cash and repo transactions into central clearing through their direct participants. In Release 34-102487, effective 4 March 2025, the Commission extended the compliance dates by one year: to 31 December 2026 for eligible cash market transactions and to 30 June 2027 for eligible repo transactions.

Central clearing moves the risk rather than removing it. Netting and standardised margin can lower bilateral counterparty exposure and the balance-sheet cost of intermediation, which would make the market more resilient and the trade cheaper to run. Concentrating positions at one counterparty and hard-wiring margin rules also relocates the stress: a clearing house's margin model becomes a systemic variable, and a procyclical margin call becomes a single centralised event rather than a scattered set of bilateral ones. Which effect dominates is genuinely unsettled, and the compliance dates above mean the first full test is still ahead.

In modern markets the scarce input is rarely a view. It is a balance sheet, and the safest assets can be made fragile by the way they are financed.

What Figure 1 supports is a narrow claim: the balance sheet committed to intermediating Treasuries is at a record in absolute terms, it is financed through repo at a scale roughly seven times its own size, and the price of that financing is stable until it abruptly is not. What the figure cannot do is attribute any of it to basis positioning. Neither dealer statistics nor SOFR percentiles separate a basis trade from any other use of the same balance sheet, and no public daily series does. A complete account of the trade therefore ends not with the spread but with a cash-flow map: who finances the bond, who sets margin, when collateral can move, and how much depth remains if many holders reach for the exit in the same week.

This research is analysis and commentary for general information. It is not investment advice, an offer, or a solicitation, and it contains no price forecasts. Quantities in the text are the author's own calculations from the sources cited; regulatory descriptions summarise the releases named.