Arguments,
with the numbers.

Articles

1. Repo’s moving day is in June. The dealers haven’t packed

   Treasury repo has 247 days to move into central clearing. On the New York Fed’s weekly numbers, the big dealers’ uncleared share has not budged in a year.

   Oct 26, 2026Repo marketsRepo series, part 2

   Anyone who has moved flat in New York knows the routine. The date goes in the calendar months ahead, everyone agrees it matters enormously, and three weeks out you are still eating dinner off the boxes.

   Treasury repo is at that stage, only with more weeks left and fewer boxes packed. On 30 June 2027, 247 days from now, the SEC’s clearing rule reaches the repo market. In the four weeks to 14 October, 55.4 per cent of the big dealers’ cash lending against Treasuries was still uncleared, by the New York Fed’s weekly data. In the same weeks of 2025 it was 54.2 per cent. In dollars the uncleared pile grew, from $1.44tn to $1.55tn. Lending through the clearinghouse’s sponsored service went from $0.59tn to $0.58tn.¹

   Box count: unchanged.

   

   A quick refresher for anyone with a life. In cleared repo, a central counterparty, the Fixed Income Clearing Corporation (FICC), stands between borrower and lender. Uncleared, or bilateral, repo is a private deal between the two, at whatever haircut they agree. The SEC’s rule requires clearing wherever at least one side is a direct member of a clearing agency. In practice that puts the obligation on the dealers.

   The lenders, at least, packed early. The share of money market funds’ repo lending that went through FICC sat between 2 and 6 per cent through 2022. It started climbing once the SEC proposed the rule that year, reaching 10.4 per cent by June 2023 and 35.9 per cent by January 2025. The rule was only adopted on 13 December 2023, so the funds were moving well before anything was final.²

   Then in February 2025 the SEC pushed both deadlines back a year, to 31 December 2026 for cash trades and 30 June 2027 for repo. Every fit I tried shows the funds’ move slowing after that, by 6 to 10 points a year, to about 2 points a year since. The share stood at 39.6 per cent in September. The extension may explain the slowdown. So may saturation, since the easy boxes get packed first. The data can’t tell the two apart.

   

   The van at the kerb

   The rest of the market got as far as the pavement. The Office of Financial Research, the Treasury’s financial-stability arm, tracks the cleared share of the repo the rule actually targets. It peaked at about 72 per cent at the end of 2025, then fell back to about 62 per cent in early 2026. Across all Treasury repo it went from a peak of 53 per cent to about 46 per cent.

   Part of that peak is a year-end effect tied to banks’ G-SIB surcharge, which is measured at year-end. Tidy for the inspection, then back to normal. One quarter is a short run, and the OFR’s own word for it is a plateau.

   Even the packing list is still being argued over. Industry groups asked the SEC to clarify and widen the exemption for trades between affiliates, one of the “critical issues” in the February 2025 extension.

   The boxes that won’t close

   Every move has a few boxes nobody wants to tape. In repo they belong to hedge funds. The OFR finds they clear their cash lending (ie reverse repo) at significantly lower rates than their borrowing. Those reverse repos often carry negative haircuts, and about 68 per cent are netted against other repos. The New York Fed reckons hedge funds’ repo with dealers was about 30 per cent cleared in late 2025.

   Price is roughly a wash. An earlier OFR brief finds funds pay about the same rate cleared or bilateral, and most funds active in Treasuries already use the cleared segment. The costs sit in margin and in tenor. Meanwhile, by the New York Fed’s count, 75 per cent of uncleared bilateral Treasury repo carries a zero haircut.

   No margin! Well, not until June.

   My reading is that a fund borrowing at zero haircut has every reason to stay bilateral until the day it can’t. That is an inference from the incentives, not something the data show fund by fund.

   The inventory

   Then there’s the paperwork. Amendments to Form PF, the private-fund filing behind the OFR’s hedge-fund data, were adopted in February 2024. Their compliance date has been pushed back four times. It now falls on 1 July 2027, the morning after the repo deadline. The SEC’s rule on who counts as a dealer fared worse: a Texas court vacated it on 21 November 2024.

   None of this means the uncleared trades are hidden. The OFR collects data on uncleared bilateral repo from the large dealers. Supervisors can see them. The public can’t.

   Why bother

   A flat line eight months out leaves two ways to arrive. One is a scramble next spring, as funds that post no margin today start posting it, all at roughly the same time. That would land on the corner of the market that borrows most: hedge funds’ repo borrowing was $3.38tn in June, as I noted last time. The other is a second extension. After the first one the move slowed, for reasons the data can’t pin down.

   Neither outcome is a crisis. Both are better planned for in October than discovered in June. The cash deadline comes first, on 31 December, the same date the G-SIB surcharge is measured. Happy new year.

   Next time

   The move needs cash to run on. Twice in eight years, repo got tight when banks’ reserves at the Fed fell to about a tenth of the federal debt held by the public, at very different dollar amounts. That could be a rule of thumb worth having, or a coincidence. The next piece weighs the two.

   On current form the dealers will still be carrying boxes down the stairs at midnight on 30 June. The inventory clerk arrives at nine the next morning.

   1. Primary dealers only. The series is broader than the rule: it includes trades between affiliates and general-collateral trades the rule may not reach, so it does not measure the rule’s progress. It dips every year-end, so the comparison is always with the same weeks a year earlier; over 8-, 13- and 26-week windows the gap stays within 1.5 points. Weeks with suppressed figures are dropped.
   2. The 2022 starting point was chosen after looking at the data. It fits better than the December 2023 adoption date.

   Links

   The repo-market model behind this series →Data and code for the charts ↗Part 1: Has the Repo-Coaster finally been retired? →

2. Has the Repo-Coaster finally been retired?

   The standing repo facility has given the repo rate a ceiling. The priciest corner of the market missed the memo.

   Oct 26, 2026Repo marketsRepo series, part 1

   If you can’t tell from the greying hair of your fixed-income friends, the Treasury market has spent three weeks on a rollercoaster. The 10-year yield touched 5.31 per cent on 5 October and was back at 4.96 per cent by 22 October.

   Repo, historically, would be in the front car for this sort of ride: harness undone, arms in the air, whooping. This time it barely looked up from its phone.

   A quick refresher for anyone who has had better things to do. Repo is where the Treasury market borrows its cash. By the New York Fed’s count, some $13.5tn of collateralised loans are outstanding on any given day. The usual gauge of strain is the gap between SOFR, the benchmark repo rate, and the interest the Fed pays banks on their reserves (IORB). Through October that gap stayed between 3 basis points below and 2bp above. Even the priciest 1 per cent of trades peaked at 10bp over IORB, roughly where it averaged in January, a month in which nothing much happened at all.

   

   Repo has form, which is what makes this a little weird. On 2 January 2019 SOFR closed 65bp above the top of the Fed’s target range. On 17 March 2020 it was 44bp over the Fed’s rate.

   April 2025 was the first hint that the act had changed. The 10-year yield jumped 14bp on 7 April. Repo rose 7bp the next day, which only got it back up to the Fed’s own rate, and by 10 April it was below it again. The pressure came out somewhere else. Between 1 and 10 April the gap between 10-year swap rates and Treasury yields slid from −45bp to −61bp, and the Fed reckons hedge funds unwound about $60bn of swap-spread trades that month, plus another $40bn in May.

   

   (For the more conscientious repo nerds: I also checked whether repo’s jumps leak into other markets. Across eight years of month-ends and the seven unscheduled spikes of 2025–26, nothing measurable reached 10-year yields, swap spreads or the OFR’s stress index.¹)

   The ceiling

   What keeps repo in its seat these days is the Fed’s standing repo facility. Since July 2021 it has lent cash against Treasuries at the top of the Fed’s target range, 10bp above IORB. Let’s call that the line.

   The line mostly held, with the odd embarrassing leak. SOFR closed above it on 16 of 1,090 days, the worst on Halloween 2025, when it finished 22bp over. Spooky.

   Then on 10 December the Fed announced it was taking the limit off the facility itself, from the next day. Eligible firms can now bid for up to $40bn per type of collateral at each of two daily operations, in what the New York Fed calls a “full allotment format” (ie they get everything they ask for). They promptly helped themselves to a record $74.6bn on New Year’s Eve.

   Since 2 January SOFR has closed above the line on precisely none of 203 days, October included. Before anyone declares the repo-coaster closed for good, a confession on the Fed’s behalf: in the same announcement, it also said it would start buying Treasury bills to top up reserves from 12 December, about $40bn in the first month. So the facility has to share the credit, and ten months is a short streak.²

   The penthouse

   The ceiling caps the benchmark. It doesn’t cap everyone.

   

   The priciest 1 per cent of SOFR trades has gone through the line on 25 of those 203 days, about one day in eight, by as much as 14bp. Before full allotment it was 7.3 per cent of days. Count the year-end scramble from 11 December as well and it’s 35 days, topping out at 25bp on New Year’s Eve, the day of that record take-up. These trades sit in the cleared bilateral bit of the market, where dealers lend to their clients.

   The odd bit is the timing. Every breach since 15 April has come on a quiet day. Take-up at the facility was about $10mn, the priciest tri-party trades sat 4.8bp below the line, and SOFR’s top 1 per cent sat 3.1bp above it. Calm downstairs, surcharge upstairs.

   My reading is that only dealers and other eligible firms get into the facility. It caps what they pay for cash, but not what they charge the clients who can’t get in themselves. Hedge funds are the usual suspects, though the data don’t say who paid, so file this under informed hunch rather than measurement.

   Nor does the ceiling do anything about the calendar. An ordinary month-end still lifts repo by about 7bp, the same as in 2018–19. Some traditions survive any amount of central-bank engineering.

   The fuel tank

   Why care about a few basis points in the penthouse? Because of what’s been built on top of repo.

   The Fed put hedge funds’ Treasury basis trade at about $830bn in September 2025, plus another $305bn in swap-spread trades. Their short futures positions have since shrunk, by about a sixth in risk terms from a peak in November 2025.

   Deleveraging! Well, no. Check the other side of the book. Hedge funds’ repo borrowing was $3.38tn in June, near its record, and futures now cover 39 per cent of their long Treasury holdings, down from 55 per cent in September 2024.

   So where did the hedge go? Economists at the Dallas Fed find that swap-spread trades now pay close to what the basis trade does, which fits a migration into swaps. Their explanation is refreshingly dull: with yields near the 6 per cent reference rate baked into Treasury futures, an option embedded in the contracts is worth less. Plumbing, in other words. Either way, more of the leverage is now hedged in a market where public data don’t show who sits on which side.

   Repo rarely starts the fire any more; the Fed has seen to that. It’s still where the fuel is stored. If something else gives Treasuries a proper shove, $3.38tn of borrowing could turn falling prices into forced selling, as in March 2020, whatever the benchmark happens to be doing.

   And the plumbing is being refitted while all of this sloshes through it. In 247 days the SEC’s clearing rule reaches Treasury repo. The New York Fed’s weekly data show 55.4 per cent of the big dealers’ cash lending against Treasuries still uncleared, against 54.2 per cent a year ago (not all of it falls under the rule, but still). More on that another day.

   So repo has finally been strapped in. A few of the passengers in the back still have their harnesses undone.

   1. Daily closes only, so anything that happened intraday could still be hiding in there.
   2. For the statistically inclined: zero breaches in 203 days is still consistent with a true breach rate of about 1.5 per cent of days.

   Links

   The repo-market model behind this series →Data and code for the charts ↗