SEC reopens fixed income cross trading
Reopening cross trading hands back the bid-offer on roughly 3.7% of fund fixed income volume, but only where marks sit at level 2. Also this week: a long end repriced to 2000 levels, and credit's scoreboard going dark.
SEC reopens fixed income cross trading, excluding level 3 books
The SEC proposed amendments to Rule 17a-7 on 9 October that would restore cross trading in most fixed income securities, which the 2020 fund valuation rule effectively shut off from its September 2022 compliance date. Eligibility is keyed to the fair value hierarchy: level 2 — priced off observable inputs that are not a live quote on the CUSIP itself, meaning dealer quotes, prints on comparable paper, a benchmark curve plus a spread, or a vendor's evaluated price — comes back in, while level 3, where the mark rests on the manager's own unobservable assumptions, and investments not categorised in the hierarchy stay out. Execution must be at either the value used in the fund's next NAV calculation that day, or a price the adviser determines reasonably represents the current market from one or more unaffiliated pricing sources, with non-firm indications and accommodation quotes explicitly excluded. The chief compliance officer takes quarterly review over from the board and must back-test at least annually against the prices realised when cross-traded bonds are later sold; comments run 60 days from Federal Register publication, and compliance is 12 months after adoption for complexes above $10bn in net assets, 18 months for the rest.
Explicit saving is the bid-offer on the crossed side, which the proposal's economic analysis puts near $173m a year across registered funds; the ICI's 2021 submission counted more than $200bn of fund fixed income cross trades in 2020, about 3.7% of their fixed income trading, of which 99.6% was level 2 — so for a conventional bond book almost all of the old volume is eligible again, while interval and tender-offer funds, holding roughly $242bn of level 3 or non-levelled assets out of $637bn gross as of September 2025, get very little, and the Commission said plainly the amendments may not materially expand what is available to BDCs. The larger number is the back-test: tying each crossed price to the realised price when that bond is eventually sold means joining execution records, a security master and a valuation history that in most firms sit in three systems with three owners, and the complexes that cannot do it will simply not cross, which is how a rule that costs nothing to use ends up available only to desks that already built the plumbing.
SEC · Atkins statement · AltsWire · Securities.io · ICI
Dealers take 2.5% as yields hit 24-year highs
Both the 10- and 30-year reached fresh 24-year highs this week in a global selloff, and the auctions that followed were among the best received of the year. Tuesday's $39bn 10-year reopening cleared at 5.300%, stopping through when-issued by 1.7 basis points, with a 2.77 bid-to-cover against a 2.51 twelve-month average and indirect bidders at 80.3% against about 71.5%; primary dealers were left with 2.5%, the lowest take on record for the tenor. Wednesday's 30-year cleared at 5.618%, the highest since August 2000, on a 2.54 cover against a 2.41 average, with indirects at 72.3% against 69.1% and dealers at 6.8% against 10.3%. The September FOMC minutes landed the same afternoon.
Indirect allocations at those levels invert the mid-September picture, when the problem at the long end was foreign bids going missing and a 20-year printing with a wider tail — distribution is plainly not the constraint now, the clearing level is. That makes the long end a price problem rather than a liquidity one, and the price problem is the harder one for construction: hedging and basis stay workable, but every discount rate in the process moves again, and carry on a 5.6% thirty-year starts competing directly with credit and private allocations that were sized when the long end paid four.
TFTC · Helious · CNBC · Federal Reserve
Credit's electronic scoreboard breaks as share changes hands
Trumid reported Q3 ADV of $9.6bn, up 38% year on year against roughly 16% growth in TRACE, and September ADV of $9.9bn, up 17% against about 8%, calling it a third consecutive quarter of record overall market share — without publishing the share. Tradeweb's September figures, out on 6 October, put fully electronic US credit ADV at $11.1bn, up 29.7%, with 21.9% of fully electronic high grade and 7.6% of high yield on TRACE, cash credit portfolio trading ADV up 56% and single-dealer non-comp portfolio trades up 125%; the same release notes that a FINRA TRACE change effective 31 August lowered reported high grade volumes and that prior periods were not restated, so the share series does not survive the break. MarketAxess, whose $167-a-share cash sale to ICE goes to a shareholder vote this month after the parties refiled their HSR notice and pushed the DOJ review to 29 October, has stopped publishing monthly volume statistics altogether. On price, Tradeweb disclosed preliminary Q3 variable fees of $2.091 per million across all products, Trumid publishes no fee per million, and MarketAxess's monthly credit fee-per-million disclosure stopped with the rest, so a like-for-like comparison of the three cannot currently be established from public sources.
What these three share is not venue rivalry but the quiet removal of the public arithmetic underneath it: one platform withholds the percentage, one withholds the volumes, and the denominator moved mid-quarter without restatement, all in the quarter when share is actually moving. That reaches a desk as a best-execution documentation problem before it reaches it as a venue-selection one — any policy citing platform market share as evidence of where liquidity sits is citing a ruler that changed length in August, and the only credit liquidity series defensible in a review from here is the one built out of a firm's own fills.
Tradeweb · Trumid · Minichart · FINRA
LSEG hands the invoice reconciliation to the agent
LSEG said on 8 October that it is extending its partnership with DataBP to give London Stock Exchange market data customers programmatic access to their own licensing and billing records through DataBP's Licensing Ledger, over a REST API and a Model Context Protocol server. The ledger holds licensing inventory, entitlements, permissions, usage terms, invoices and billing history, and access is authenticated and entitlement-scoped through the customer portal so a firm sees only what it is licensed to see. The stated use cases are an agent comparing licensed entitlements against invoice line items and flagging discrepancies, and checking whether a particular use is permitted; LSEG's framing of the problem is that licensing rights still sit in static PDFs and invoices are reconciled line by line. DataBP co-founder Brandon Baker said an AI agent "can do in minutes what teams currently spend days on", and Geoff Hammond, head of LSE market data, news and identifiers, described it as giving clients "structured access to the licensing data that underpins our own workflows". No pricing was disclosed.
Bloomberg's Enterprise MCP server, out on 29 September, is the prevailing direction — put the feed behind a protocol an agent can call, which is where the licence-as-API thread of the past month has run — and this diverges from it by putting the licence itself behind the protocol, with no published equivalent from Bloomberg, CME or ICE and therefore no competing terms to price it against. The value sits in the reconciliation rather than the lookup: CME's repricing of end-of-day and non-display data, back-billed to June 2025, produced a line-item dispute its user group threatened to litigate, with a worked example of $273,600 a year for one bank across five exchange products, three datacentres and two redistributors, and the binding constraint there was never the headline fee but the inability to prove from internal records what had been licensed in the first place.
LSEG · WatersTechnology · WatersTechnology on CME · Bloomberg
Citadel publishes its rankings to set the capital argument
Citadel Securities published a whitepaper on 8 October defending non-bank liquidity provision and, unusually, quoting its own non-public Q2 dealer ranking reports: first for US Treasuries by risk executed on both Tradeweb and Bloomberg, and first on hit rate for French and German government bonds on Tradeweb, second on Bloomberg. Its own figures put 2025 US Treasury volume at about $23trn, or $92bn a day, and a 14.2% share of investment grade credit block risk transfer on $406bn traded year to date; of the March Iran-conflict volatility it says it "maintained industry-leading low 'no-quote-rates' while trading a record $1.36 trillion in a day". The paper proposes five metrics for liquidity resilience across dealer-to-customer markets — non-quote rates, market share, execution cost savings, hit rates and risk executed — and argues bank capital rules do not fit non-bank market makers, aiming at AFME's February response to the FCA's engagement paper on market risk capital for investment firms, which held that newer entrants are untested in a market-wide fixed income stress, and at ESMA's proposal that systemic non-bank liquidity providers face requirements comparable to banks in similar activities. Tradeweb and Bloomberg declined to comment.
14.2% of investment grade block risk transfer is the figure that matters, because that is the share of large-size credit execution now priced off a balance sheet not capitalised like a bank's, and on ESMA's approach the capital charge lands in the bid-offer on exactly those blocks. Which five metrics are the right ones is secondary to who is proposing them — a liquidity provider rather than a venue or a regulator — and a desk that can already compute non-quote rates and hit rates by counterparty from its own RFQ logs will argue that file from its own data, while one that cannot will be arguing from Citadel's.