But in the City of London, some actors moved before the headline arrived.
Did the Bank of England operate through an open market,
or through a closed access network?
The whole project turns on that one distinction. To answer it, the paper asks three smaller questions, each tied to a different period and a different test.
Did a stable set of intermediaries sit at the front of the crisis information chain in the Victorian period, closest to the Bank?
Did the 1866 transatlantic cable, which made prices travel faster, shrink that advantage and open access to the Bank?
Did the access network outlast the First World War, from Victorian bill brokers to later authorised money-market counterparties?
If markets were becoming modern and efficient, we might expect access to the Bank to grow more open over time. But if London finance stayed a closed access network, then modernity did not simply mean democratization. It meant older elite relationships were formalized into institutional rules.
Anyone could reach the Bank directly. Access widens and diffuses. Crisis support is a public facility, open on equal terms.
A recognised gatekeeper layer rings the Bank. The wider public sits outside it. Crisis support flows through known intermediaries.
The project enters a real disagreement. One tradition sees finance as shaped by elite networks and private access. Another sees information becoming faster and markets becoming efficient. The paper takes both seriously.
Finance was shaped by elite networks, private information, and privileged access.
"Gentlemanly capitalism": a London elite whose advantage was relational, built on geographical location and social ties.
Interlocking directorates worked as private information conduits, giving some directors a better informational advantage.
The theory that a crisis is fundamentally an information event, where who knows what, and when, matters.
Victorian crisis liquidity ran through the discount-house and bill-broker layer, with privileged access to the Bank.
Information became faster, prices became efficient, and simple insider-profit stories can be myths.
The 1866 cable cut the London to New York information lag "from ten days to zero" and compressed price gaps from 5–10% to 2–3%.
The Bank's 1866 lending was politically negotiated and deliberately ambiguous, not a transparent public rule.
The famous insider-profit legend is "more legend than fact", a caution against assuming early access became profit.
Measure dated visibility and access. Never assume foreknowledge or profit.
Maybe both are true. Prices became faster — but access to the Bank remained mediated.
Price information and institutional access are two different things. The cable settled the first. This project tests the second.
A crisis is easier to understand as a small cast of institutions than as a wall of numbers. The Bank of England sits at the centre. Around it are the firms that could reach the Bank, the banks that funded them, the press that made stress public, and the wider public that usually learned last.
They buy and hold short bills, funded by overnight money lent by the banks. The same role runs down the middle of the whole story: the bill broker the Bank gave a facility to in 1830 becomes the authorised counterparty of 1996.
Revealed by · BoE transaction ledgers + the 1967 discount-market historyThey matched lenders and borrowers of short IOUs, called bills of exchange. They are the ancestors of the discount houses and the first firms the Bank dealt with on a recognised footing.
Revealed by · ledgers + the 1906 bill-market reconstructionAlso called merchant banks. They guarantee, or "accept", a bill for a distant borrower, lending their name rather than their cash. They vet borrowers and form part of the recognised layer.
Revealed by · the 1914 ledger + the 1906 bill reconstructionLarge ordinary deposit banks that supply call money to the discount houses and later partner in rescues. They fund the network and help organise support.
Revealed by · Römer's 1931 study + the 1978 lifeboat accountThe Bankers' Magazine and similar titles reported money-market strain for professionals. They show stress was legible to insiders before it reached the public.
Revealed by · The Bankers' MagazineHMD newspapers and Hansard form the broad public record. This is the comparison point the whole argument turns on: the moment a crisis became visible to everyone.
Revealed by · HMD newspapers + HansardNo single record proves the argument. Each source answers one question and stays silent on others. The pattern appears only when they are linked in order, which is what makes this digital history.
Shows who reached the Bank's window, by name and amount.
Cannot show what anyone privately knew.
Dates the Bank's earliest defensive signal, day by day.
Cannot show the market rates firms actually paid.
Dates when stress became speakable, to insiders then the public.
Cannot prove absence; scanning shapes the record.
Tests whether early access showed up as visible profit.
Cannot see unlisted firms or private gains.
Maps the funnel and the long institutional lineage.
Institutional authorship; not an outside critique.
Shows the layer still in active use up to 1997.
Sector aggregate; limited firm-level detail.
Read together, these answer one question each source cannot answer alone: not what anyone knew, but who could reach the Bank, and how early stress was visible.
In the Victorian crises, the same kinds of intermediaries recur at the front of the chain, and the Bank's defensive signal appears before the public rupture in every case.
The 1866 transatlantic cable made prices travel faster. It did not decide who could stand at the Bank's discount window. The insider-to-public lead grew rather than shrank.
After 1914 the structure took a more formal shape. The Bank kept working through recognised intermediaries until 1997, when it widened to a larger defined list, not to the public.
A note on language. The project measures dated visibility, positioning, and institutional access. It does not claim that anyone knew a crisis in advance or profited from one.
It is a fair objection that crises are unusual. The reason to study them is that a crisis makes a hidden structure visible. In calm years the access network is quiet. A discount house rolls over its bills, a clearing bank lends spare cash overnight, a trusted firm gets credit because everyone knows its name. These relationships matter, but they rarely leave a dated public trace. A panic makes them loud.
Access runs on trust and habit. It is real, and it shapes who gets credit, but it produces little that you can date or count from the outside.
The Bank's ledgers, its rate moves, its rescue operations, the newspapers and the official record all turn dense and datable at the same moment. You can finally see who reaches the Bank, who moves early, who is supported, and who is left outside the recognised layer.
The claim is not that the network existed only during panics. It is that a panic forces private relationships into the record. So a crisis works as a stress test that exposes the wiring, and comparing several of them lets us ask whether the same structure keeps reappearing.
The cases were not chosen for drama or severity, and the paper does not claim they are the most severe. They were chosen to vary along the dimensions that matter for the question: period, crisis origin (home-grown or imported), the communication technology of the day, and above all what kind of evidence survives about who could reach the Bank. Each case has to clear three bars at once.
Each is named in the scholarly literature as a defining moment for the Bank's crisis role, so the comparison engages real cases other historians argue about, not convenient ones.
The cases sit at different points on the map: home-grown bill-market collapse, imported transatlantic shock, single-firm rescue, war and payment rupture. If one Bank rule ran uniformly, their shapes would converge. They do not.
Each appears in the daily Bank Rate and the public record, and most carry a deeper source — a transaction ledger, daily account books, or a recent archival reconstruction — so they can be set side by side.
| Crisis | Origin | What evidence it carries | Role in the argument |
|---|---|---|---|
| Part I · the press-and-telegraph era, 1847–1914 — who reached the Bank, and how early | |||
| 1847 | home-grown | Transaction ledger · firm-level | The earliest named window: who stood at the Bank before the watershed of 1866. |
| 1857 | imported · US | Transaction ledger · firm-level | A pre-cable transatlantic shock; the baseline for the telegraph test. |
| 1866 | home-grown | Transaction ledger · firm-level | The Overend panic, the case Bagehot was describing; the broadest window in the sample. |
| 1873 | imported | Bank Rate · public record | Included where the evidence allows, and flagged as a weaker case with thin coverage. |
| 1890 | home-grown | Daily account books · no ledger | The Baring rescue. It enters the lead-time comparison but not the firm-level recurrence, because no transaction ledger survives. |
| 1907 | imported · US | Bank Rate · public record | A post-cable transatlantic comparison to 1857: did faster news shrink the lead? It grew. |
| 1914 | external · war | Transaction ledger (reduced) | The latest ledger case, on the far side of the cable; tests whether the same firms recur. |
| Part II · did the structure survive? 1906–1997 — evidence shifts from ledgers to aggregates | |||
| 1906 | prewar baseline | Bill-market reconstruction | A calm-year snapshot of the access funnel just before 1914. |
| 1931 | home-grown | Archival reconstruction (Römer) | The closest major post-1914 British crisis with a recent reconstruction of the Bank's operations. |
| 1973–75 | home-grown | Official BoE account | The lifeboat: the major later British rescue, with named institutions and a vetted committee. |
| 1976–96 | structural | BoE statistical database | Shows whether the recognised intermediary layer was still in ordinary use before reform. |
| 1997 | the reform | Institutional records | The institutional break: the year the Bank stopped dealing exclusively with the discount houses. |
Choosing structurally distinct cases on purpose commits the argument to a particular shape. It cannot be a tidy ranking of more or less modern rescues. It has to be that the form of the crisis shapes who can reach the Bank — and where a single uniform rule had operated, the measures across these cases would converge rather than diverge. With a dozen cases the paper compares structures and traces continuity; it does not claim a statistical law over all crises or all central banks. Full selection logic in docs/case_selection.md and §4–5 of the paper.
Some shocks were home-grown in London. Others arrived from abroad, most often from New York. Pick a crisis to trace its origin, the recognised actors positioned closest to the Bank, and the gap between market visibility and public information.
A long market build-up, then a sudden public rupture when Overend suspended.
Read as visibility and access, not foreknowledge. The dates mark when stress was measurable to the market versus when it became public, not what any person knew or did.
Pick a crisis. The actors are laid out in the order trouble first reached them. For the three crises with surviving transaction ledgers (1857, 1866, 1914) that order is read straight from the Bank's own books: the date each kind of firm first appears at the discount window. Every link carries the gap in days. The same recognised layer keeps arriving first, and the wider public arrives last.
Each link is a measured gap in dated records, not a claim about who knew what. Ledger order for 1857, 1866 and 1914 (BoE transaction ledgers); Bank-Rate-signal to public rupture for the crises without surviving ledgers. Computed in src/v7_flow_network.py.
The recurring firms form a network. Each named institution is joined to the crisis periods it appears in, and two firms are linked when they sit at the Bank in the same period. Node size is eigenvector centrality, the standard graph measure of how embedded a node is in the well-connected core. The discount houses, in green, are the spine, and one firm sits at the centre of the whole 150-year structure.
A graph-theoretic read of the recurrence evidence. Memberships are the documented named firms from the v6 recurrence tables (Victorian ledgers; Accominotti et al. 2025; Römer 2025; the Bank's discount-market records). Centrality computed with NetworkX in src/v7_flow_network.py; it measures structural position across periods, not transaction volume.
For each crisis the paper lines up three clocks: the Bank or market signal, the specialist financial record, and the public rupture. A market-facing signal comes first in every case. The lead is longest when trouble grew inside London finance, and it almost vanishes in 1914, when war froze the system overnight.
If the City's advantage were only about faster news, the transatlantic cable should have weakened it. Before 1866, market information crossed the Atlantic slowly, by ship. After the cable, a price could travel in minutes.
Prices became faster. Access did not become open. The lead grew rather than shrank, so the telegraph and the club ran through different channels.
Of the ten intermediaries with the highest access scores, all ten reached the Bank's window early. They were mainly discount houses, bill brokers, and merchant banks, exactly the kinds of firms named in the City's own histories. Overend, Gurney recurs until it fails in 1866. Frühling & Göschen and Stern Bros reappear in 1914.
This is not a claim that the same firms controlled everything. Membership turns over. The durable feature is the recognised layer, not any single name.
If listed financial firms were being repriced before a public rupture, the Yale Investor's Monthly Manual index should fall. It does not. In 1873, 1890, and 1907 the sector was still rising into the rupture.
The index is monthly and sector-level, and several key discount houses were never listed. A negative result here means only that no broad listed-sector price footprint is visible. It cannot prove that no individual insider benefited.
The Bank stays fixed at the centre. The layer between the Bank and the wider public changes shape across 150 years, but it never opens. Each form is a recognised intermediary, not a public market.
The club did not simply disappear. It became formalized, then widened in 1997.
The index scores six yes-or-no criteria per period. The first four separate an open market from a mediated one: an open market would not need named institutions, a formal list, restricted access, or a fixed intermediary route. The last two are about how far we can trust the reading. Toggle a criterion to see what each verdict rests on.
The pattern only appears when separate digitised sources are placed on one timeline. A contemporary saw one ledger, one newspaper, or one crisis. Click a card to turn it over and read where each source came from, and where it falls short.
Bank archives via Anson et al. 2017.
Working records of the Bank, not later memoirs.
Patchy coverage; shows access, never motive.
Naming the recurring inner circle and its timing.
BoE Millennium dataset.
Official, continuous, standardised series.
A rate rise can have several causes.
A common daily clock for the lead-time test.
British Library HMD, Hansard, Internet Archive.
Contemporary printed records.
Survival, scanning, and OCR shape the corpus.
The public and specialist visibility clocks.
Yale International Center for Finance.
A contemporary market record, not interpretation.
Monthly, sector-level; key houses unlisted.
Testing whether prices fell before ruptures.
Accominotti, Lucena-Piquero & Ugolini (arXiv).
A large, near-published scholarly reconstruction.
1906 is not a crisis year.
The prewar access funnel.
Römer 2025, Economic History Review.
Recent peer-reviewed archival study.
Raw data not released; the diagram is source-derived.
The interwar actor-flow network.
BoE Quarterly Bulletin, 1967.
The Bank describing a market it ran.
Institutional authorship, not an outside critique.
The long-run institutional lineage.
BoE Quarterly Bulletin, 1978 Q2.
From the organiser of the rescue.
A rescuer's account of its own rescue.
The selective rescue network.
BoE statistical database, series RPMATJD.
Official structured balance-sheet data.
Sector aggregate; exposed coverage from 1986.
Showing the houses still in active use before 1997.
The Bank of England's access network did not stay the same. But it also did not simply open.
From Victorian bill brokers to authorised counterparties, crisis access remained mediated through recognised financial actors until 1997.
This is what makes the project digital history: the argument appears only when ledgers, newspapers, rates, prices, and institutional records are linked into one timeline.
Crises reveal hidden institutional structures.
Faster information did not automatically democratize access.
Digital history makes the pattern visible by linking sources never designed to be read together.
The argument depends on linking sources that were never built to be read together. Each card opens to show where the data came from, what it holds, why the project needed it, how it was processed, what it showed, and where it falls short.
The Bank of England, transcribed from its own archival discount ledgers (the C28 series) by a named team and published alongside the 2017 working paper.
Loan-by-loan rows for the 1847, 1857, 1866 and 1914 crises: discounter name, amount, rate, collateral, and whether the bill was accepted or rejected.
It is the only source that names who actually reached the Bank's discount window during a crisis, which is the heart of the access question.
Reshaped into a long-form table, then names were resolved from 1,476 raw spellings to 1,314 entities and classified by type (discount house, bill broker, merchant bank, clearing bank).
The concentration and breadth of crisis lending: 1866 had the broadest distribution (top-five share 19%, HHI 154), while 1857 was far more concentrated (top-five 42%, HHI 503). It also names the recurring inner circle.
No 1890 ledger exists in the release. The 1914 ledger has fewer columns (11 versus 28). 86% of resolved names still carry a manual-review flag, so type shares are lower-bound estimates. The ledger records the what of lending, never the why.
The Bank of England's long-run macro compilation, which records the Bank's announced policy rate at daily frequency back to 1833.
One row per rate announcement: date and the end-of-day Bank Rate in percent per year.
It provides a single, uniform daily clock that every crisis can be measured against, and the rate level needed for the penalty-rate test.
The raw sheet holds 67,578 rows because of intra-day announcements; these were de-duplicated by date to give one settled daily rate.
The earliest market-facing signal date for each crisis, and confirmation that the Bank charged a penalty rate in three of the four core crises.
It is the Bank's own policy rate, not the market discount rate faced by counterparties, so it cannot speak to the spread individual firms actually paid.
Three digitised text corpora: mass-market newspapers, the parliamentary record, and the specialist financial press for professionals.
Dated, searchable text in which a crisis is named or discussed, at three levels of audience from insider to public.
These supply the public and specialist visibility clocks: the dates on which a crisis became speakable to professionals, then to everyone.
Dated mentions of each crisis were located to fix the public rupture date used in the lead-time comparison.
That the specialist record and the public record sit later than the Bank's market-facing signal in every case studied, with the gap widest for home-grown crises.
Survival, scanning choices and OCR quality shape which pages exist and are findable, so the corpus is a sample of the historical record, not the whole of it.
A cleaned, digitised version of the Investor's Monthly Manual, a nineteenth- and twentieth-century publication reporting listed securities.
One row per listed security per month: company name, industry code, security type, and the late-month quoted price where available.
It gives an independent market view of ordinary financial firms, separate from the Bank's own records, to test whether prices moved before a rupture.
Filtered to common and preferred stocks, split into listed financial institutions and non-financial corporates, with monthly returns and the share of declining securities computed.
No clear pre-rupture price footprint. In November 1890 listed financial institutions returned about −0.9% against −1.2% for non-financial corporates, a mild and not distinctive move.
Monthly and sector-level; quotes are often stale; several key discount houses were never listed; and the panel has a severe quotation gap around 1914.
A large scholarly reconstruction of the prewar London bill market built from rediscounted sterling bills.
Counts and relationships across tens of thousands of bills, linking the firms that drew, accepted, and rediscounted them.
It maps the prewar access funnel: how a very wide base of drawers narrowed to a small set of names that reached the Bank.
Used as a published structure to characterise the funnel from many drawers down to a recognised discounting layer.
A steep narrowing: a broad population of drawers passed through acceptance houses to a small recognised group of discounting firms at the apex.
1906 is a normal year, not a crisis year, so it describes the standing structure rather than crisis behaviour directly.
A recent peer-reviewed archival study of the Bank's operations during the 1931 sterling crisis.
An account of how liquidity flowed between the Bank, the clearing banks, and the discount houses during the crisis.
It supplies the interwar actor-flow picture at a moment when the project has no transaction ledger of its own.
Read as a source-derived network: actors and flows are taken from the published study rather than re-computed from raw rows.
About £25m of Treasury bills bought through the discount houses, with the Big Five clearing banks and named market firms still central. The recognised market layer carried crisis liquidity.
The underlying raw data are not released, so the diagram is derived from the study's account rather than independently reconstructed.
The Bank's own published account of the London discount market and its institutions.
An institutional narrative of bill brokers, discount houses, and the London Discount Market Association.
It establishes the long-run institutional lineage that connects the Victorian bill broker to the later authorised counterparty.
Read as a primary institutional source for the continuity of roles across the twentieth century.
That the discount market was a recognised, named, and self-organising layer between the Bank and the wider market, with continuity of function over decades.
It is the Bank describing a market it operated, so it is an institutional account rather than an independent critique.
The Bank's published account of the 1973–75 secondary banking crisis, known as the lifeboat.
The structure of the rescue: the Control Committee, the named participating institutions, the support amounts, and the entry criteria.
It documents a selective rescue network in the modern era, the late equivalent of the Victorian inner circle.
Read for the membership and governance of the rescue, and the scale of support over time.
Support peaked at about £1,285m across roughly 26 banks, organised by the Bank with the clearing banks on a vetted committee. A closed, mediated rescue, not an open facility.
It is the organiser's own account of its own rescue, so the framing reflects the Bank's perspective.
The Bank's structured statistical record of the discount houses' monthly balance sheets.
Monthly balance-sheet items for the discount houses, including the sterling deposits the Bank itself placed with them.
It shows whether the recognised market layer was still in active use right up to the 1997 reform.
Read as a monthly series to track the Bank's continued placement of deposits with the houses across 1986–96.
The Bank still placed deposits with the discount houses in roughly 85% of months across 1986–96. The structure was not a relic; it was operating until the reform.
It is a sector aggregate, and the most exposed coverage runs only from 1986, so firm-level detail before then is limited.
Also drawn on in support: the JST Macrohistory panel (annual UK macro context from 1870, used for 1890 and 1914) and the Bank of England Daily Account Books C1/38 (1890 page images, transcribed for October to December via a vision model and validated against weekly anchors). Both are documented in full on the Sources tab.
Every dataset is a derivative of some original record, made by particular people for particular reasons. This desk lays out where each source came from, who made it, how it was digitised, why it can be trusted, and the specific claims it cannot support. The previews below are archive-style representations, not reproductions of the originals.
Published by the Bank of England with Anson, Bholat, Kang & Thomas (2017).
Bank clerks who recorded each discount; transcribed by a named team from the archival C28 series.
Hand-transcribed from page images into a spreadsheet, with raw transactions and pre-computed metrics.
Working records of the Bank itself, with an independent secondary cross-check on the 1866 portion.
Clerks' name spellings vary; no 1890 ledger; the 1914 ledger is thinner; classification still mostly flagged for review.
It cannot show motive or decision rationale, a counterparty's wider balance sheet, or any credit that did not pass through the Bank's window.
A Millennium of Macroeconomic Data, the Bank's long-run compilation.
The Bank of England, drawing the rate from its internal and published records.
Daily observations digitised into a single sheet, then de-duplicated by date for this project.
The authoritative record of the Bank's own rate, used to cross-check the 1890 image transcription.
It is a policy rate, not a market rate, and intra-day changes had to be settled to one daily value.
It cannot describe the open-market discount rates that counterparties actually faced.
British Library digitised newspapers, the Hansard parliamentary record, and Internet Archive runs of the specialist press.
Contemporary editors, reporters, and parliamentary recorders.
Scanned and OCR'd by their respective libraries and archives.
Contemporary printed records, dated at the point of publication.
Survival, scanning selection, and OCR quality all shape which pages are present and findable.
Absence of a mention is not proof of absence; the corpus is a sample, not the complete record.
Cleaned files published by the Yale School of Management.
The original publication recorded listed securities; Yale produced the cleaned digital panel.
Filtered to common and preferred stocks and split into financial and non-financial sectors.
A contemporary market record independent of the Bank, with hundreds of quoted securities around 1890.
Monthly and often stale; key discount houses unlisted; a severe quote gap around 1914.
It cannot show whether any individual firm benefited, nor support a 1914 ordinary-bank comparison.
The Bank of England Quarterly Bulletin, 1967 and 1978 Q2.
The Bank of England's own staff.
Published institutional accounts, read here as primary sources on structure and membership.
Written by the institution that ran the market and organised the rescue, with named participants and amounts.
Institutional authorship: the Bank narrating its own market and its own rescue.
They are not independent critiques and should be read alongside outside scholarship.
The Bank of England Archive, via its digital image service.
Bank clerks recording the daily Banking and Issue Department balances in 1890.
Eighteen page images for October to December 1890, transcribed by a vision model after OCR proved too weak.
The Bank's own daily record; the transcription validates against weekly anchors at 92 to 100% for the variables kept.
Only three months are transcribed, and only a few balance-sheet variables passed validation.
It carries no counterparty identities and supports no cross-crisis daily comparison.
Source absence is not historical absence. A gap in the data this project uses, such as the missing 1890 transaction ledger, is a limit of the project's coverage, not proof that the underlying record never existed. Where a claim cannot be traced through this chain, the project does not make it.
No finance background needed. Each method below names the question it answers, the data it uses, how it works in simple language, what it produces, and where it can mislead.
Who actually borrowed from the Bank during a crisis?
What kind of institution was each borrower?
Do the same firms or firm types keep reappearing?
How early was stress visible to the market versus the public?
When did each crisis become speakable, and to whom?
How narrow was access to the Bank?
How did liquidity move between institutions?
How did the Bank's balance sheet move through a crisis?
Was the access structure open or mediated, and how sure can we be?
Did markets reprice financial firms before a public rupture?
The findings, each with what was measured, what it shows, why it matters to the argument, and one line on its limitation. Charts are the project's own outputs; the diagrams for 1931 and 1973–75 are source-derived.
A market-facing signal appears before the public rupture in every crisis studied. The lead is longest for home-grown crises and almost vanishes when war froze the system in 1914.
Stress was measurable inside London finance well before it became public, which is what an access advantage looks like.
Comparing two US-origin shocks, the insider-to-public lead grew from 25 days in 1857 to 68 days in 1907, after the 1866 transatlantic cable made prices travel in minutes.
If the advantage were only about fast news, the lead should have shrunk. It widened, so price speed and access ran on different channels.
The firms closest to the Bank are mainly discount houses, bill brokers, and merchant banks. Names like Overend Gurney, Frith Sands, and National Discount sit at the top of the 1866 window.
The exact firms change across crises, but the recognised actor type at the front of the chain persists.
Turning the recurrence evidence into a graph and measuring it, Union Discount Co. has the highest eigenvector centrality (0.49) and the highest betweenness. It is the single node bridging the Victorian acceptance cluster to the 1996 authorised list, and half of the eighteen named firms are discount houses.
The word "network" in the title is now a measured object, not a metaphor. The connected spine of the graph is the discount houses; the rescued secondary banks of the 1970s sit on its outer edge.
Lined up at their trigger week, the crises share a shape: lending surges, reserves dip then rebuild, and Bank Rate spikes. 1914 dwarfs the others in scale.
A shared event-time frame makes very different crises directly comparable and shows the Bank's response was systematic.
Listed financial firms did not visibly fall before ruptures. Around the 1890 crisis their move was mild and no more negative than the wider market.
The project finds an access advantage, not a proven profit advantage. The two are not the same.
Crisis borrowing concentrates in a small group. 1857 was highly concentrated (top-five 42%, index 503), 1866 the broadest (top-five 19%, index 154). The prewar bill market funnelled many drawers down to a recognised discounting layer.
Access narrowed structurally toward the Bank, even when many firms drew bills lower down the chain.
In the 1931 sterling crisis the Bank bought about £25m of Treasury bills through the discount houses, with the Big Five clearing banks and named market firms central.
Even without a transaction ledger, the interwar Bank still worked through recognised intermediaries.
The 1973–75 rescue was organised by the Bank with the clearing banks on a vetted Control Committee, with entry criteria and named members.
The modern equivalent of the Victorian inner circle was still a closed, mediated network.
Across 1986–96 the Bank still placed sterling deposits with the discount houses in roughly 85% of months, right up to the 1997 reform.
The recognised market layer was not a relic by the 1990s; it was a working part of how the Bank operated.
Scored against the lender-of-last-resort criteria, the crises cluster high (1857 and 1914 score 5 of 6, 1866 scores 6), confirming a consistent, mediated structure.
The structure stayed recognisably the same for 150 years, then changed form at 1997 rather than simply opening.
| Firm | Periods |
| Union Discount | 5 |
| Overend, Gurney & Co. | 3 |
| Seccombe Marshall & Campion | 2 |
| Frith Sands & Co. | 2 |
| R. Raphael & Sons | 2 |
A small set of recognised firms appears across multiple crises, with Union Discount in five periods and Overend Gurney recurring until it fails in 1866.
It puts named evidence behind the claim that the inner circle was durable, not random.
Top counterparties at the Bank's window · 1866 ledgerContinuity score by crisis · a consistent mediated structure