Three Scenarios for the Dollar Swap Network When Crisis Drawdowns Stay Near Zero

The Federal Reserve's H.4.1 release dated 13 August 2026 recorded $132 million of central bank liquidity swaps outstanding for the week ended 12 August, against total Reserve Bank assets of $6,759,955 million. That is roughly two thousandths of one percent of the balance sheet — the residue of periodic small-value operations run to confirm that the plumbing still works. Read as an activity measure, the number says the offshore dollar funding system is calm. Read as a measure of what the swap network is worth, it says almost nothing.

The standing network is a contingent claim, not a flow. Its value is set by the terms on which it can be drawn and by who sits inside the perimeter, not by the balance outstanding on any given Wednesday. Forgetting that produces a specific analytical error: treating drawdown size as the crisis thermometer. The record of the three most recent episodes shows why that reading fails.

The Network as It Currently Stands

The architecture has three levels, and the differences between them are mechanical rather than rhetorical.

Tier one — the standing swap lines

The Federal Reserve maintains permanent, standing dollar liquidity swap arrangements with five foreign central banks: the Bank of Canada, the Bank of England, the Bank of Japan, the European Central Bank and the Swiss National Bank. These became open-ended on 31 October 2013, converting what had been a rolling series of temporary agreements into an indefinite facility. Each is reciprocal, so the Federal Reserve also holds the capacity to obtain Canadian dollars, sterling, yen, euro and Swiss francs.

Pricing was set at the US dollar overnight index swap rate plus 25 basis points with effect from the operations scheduled in the week of 16 March 2020, a reduction of 25 basis points from the prior spread. The same coordinated action added weekly 84-day maturity operations alongside the existing one-week tender. Days later, on 20 March 2020, the six central banks agreed to raise the frequency of 7-day operations from weekly to daily, commencing 23 March 2020.

Two features matter. There is no headline cap on amounts available to these five counterparties, and access requires no negotiation, board approval or programme: the line exists, the price is published, and the counterparty central bank decides.

Tier two — the collateralised facilities

Beneath the swap lines sits a different instrument. The Federal Reserve's FIMA Repo Facility, created as a temporary measure on 31 March 2020 and made standing on 28 July 2021, allows foreign and international monetary authorities with accounts at the Federal Reserve Bank of New York to exchange US Treasury securities for dollars overnight or for seven calendar days. The seven-day rate is the weekly overnight index swap rate plus 25 basis points; the overnight rate is the minimum bid rate for the Standing Repurchase Agreement Facility. The Federal Reserve states that the offering rate generally exceeds private repo rates when the Treasury market functions normally — by design, the facility is unattractive until it is needed.

The euro side of the system now has an expanded analogue. On 14 February 2026 the ECB announced changes to EUREP, the Eurosystem repo facility for central banks, introducing standing access in principle for all central banks outside the euro area, subject to exclusion on money laundering, terrorist financing or sanctions grounds. The stated rationale referenced "profound structural shifts related to geopolitics and changes in the international financial system", and the objective of helping counterparties address euro liquidity shortages swiftly. Implementation began on 24 July 2026. The published parameters set a maximum borrowable amount of EUR 50 billion and a minimum transaction size of EUR 10 million, maturities of one day to one week with a rollover option, pricing at the main refinancing operations rate plus a Governing Council spread, and eligible collateral limited to euro-denominated marketable assets other than covered bonds. Five national central banks — the Deutsche Bundesbank, Banco de España, Banque de France, Banca d'Italia and De Nederlandsche Bank — act as operational counterparts.

Tier three — everything else

Outside those two tiers, dollar and euro access is negotiated case by case. The North American Framework Agreement provides the Federal Reserve Bank of New York with a $2 billion arrangement with the Bank of Canada and a $3 billion arrangement with the Bank of Mexico, alongside a separate $9 billion Treasury-managed line with Mexico. The Eurosystem's own swap network is similarly stratified: unlimited standing reciprocal lines with the Federal Reserve, the Bank of Japan, the Bank of England, the Swiss National Bank and the Bank of Canada, but a capped line with the People's Bank of China at EUR 45 billion (or CNY 350 billion in the other direction) running to 8 October 2028, and non-reciprocal standing lines of EUR 24 billion with Danmarks Nationalbank and EUR 10 billion with Sveriges Riksbank.

Below that, the fallback is the multilateral safety net. The IMF's published lending capacity stood at roughly SDR 695 billion, including SDR 364 billion under the New Arrangements to Borrow and SDR 141 billion in bilateral borrowing agreements. Large in aggregate, slow in practice: drawing on it requires a programme, not a phone call.

What March 2023 Showed That 2008 and 2020 Did Not

The conventional monitoring instruction is to compare drawdown volumes against prior episodes. The historical series makes that instruction look sensible until the third data point arrives.

Swap balances peaked at $583,135 million on 17 December 2008. They peaked again at $448,946 million on 27 May 2020. Both episodes fit the intuitive model: acute stress, large drawdowns, a visible line item that any observer could track weekly.

March 2023 broke the model. On 19 March 2023, the same six central banks announced a coordinated increase in the frequency of 7-day operations from weekly to daily, starting 20 March and running through at least the end of April. That is the identical operational signal issued in March 2020. Yet the balance sheet response was negligible. The Wednesday level for 22 March 2023 was $587 million, with the weekly average at $502 million — under $600 million on either measure, and the high for the entire year. Against the 2020 peak, that is roughly one eight-hundredth of the volume, following the same announcement.

The announcement channel and the drawdown channel are separate mechanisms. In 2023 the first fired at full strength and the second barely registered. An observer watching only the balance sheet line would have concluded nothing happened.

The plausible reading is that 2023 was a confidence problem inside banks whose home central banks held ample dollar reserves, rather than a market-wide inability to roll dollar funding. The backstop was reaffirmed, the reaffirmation was priced, and the facility went almost unused. Whether that generalises is what the next episode tests.

Scenario A — The Announcement-Only Episode

Trigger. Stress concentrated in one or a small number of institutions, in a jurisdiction whose central bank holds substantial dollar reserves and whose banking system retains access to private funding at a wider but functioning spread. Credit differentiation rises; market-wide rollover capacity does not fail.

Path. A coordinated statement moves 7-day operations to daily frequency. Cross-currency basis widens for days. Drawdowns stay in the low hundreds of millions or single-digit billions. The 84-day tenor is either not reactivated or offered and lightly used.

Assessment. On the record of the three most recent episodes, this branch is the most likely single outcome for stress that originates as an institution-level credit event rather than a market-structure failure. The reason is mechanical: where the shortage is confidence in a borrower rather than scarcity of the currency, official dollars do not solve the binding problem, so the facility is not drawn.

What would falsify it. Drawdowns crossing roughly $20 billion within two weekly reports would indicate the shortage is broader than a credit event, moving the episode toward Scenario B.

Scenario B — The Core Drawdown Episode

Trigger. A simultaneous demand for dollars across many end users who fund dollar assets from non-dollar liability bases. This is a rollover failure, not a credit judgement: the constraint is that intermediary balance sheets cannot absorb the size at any spread within the horizon required.

Path. Daily 7-day operations plus reactivated 84-day tenders. Drawdowns build over three to six weekly reports rather than arriving at once, since foreign central banks must first allocate to their own domestic counterparties. In 2020 the ascent from negligible to the peak took roughly ten weeks, which is a useful sense of pace rather than a forecast.

Assessment. Materially less likely than Scenario A in any given episode, because the conditions required are more specific. It is also the branch with the largest consequences, which is why the tail deserves attention disproportionate to its frequency.

What would falsify it. Balances plateauing below prior peaks while basis spreads normalise would suggest the squeeze was absorbed and the episode is reverting to Scenario A.

Scenario C — The Perimeter Episode

Trigger. Stress concentrated outside the five standing counterparties. A banking system or corporate sector with heavy dollar or euro obligations faces a rollover wall while its central bank holds reserves that are adequate in aggregate but not in the maturity or the currency needed.

Path. This is the branch where the tier structure becomes decisive. Tier-one counterparties draw little or nothing, so the Federal Reserve's swap line stays near zero throughout — the same visual as Scenario A on the H.4.1, produced by a completely different underlying condition. Relief, if it arrives, runs through the collateralised facilities: FIMA repo for holders of US Treasuries, EUREP for holders of eligible euro-denominated paper. The binding constraint moves from price to eligibility. An authority with no qualifying collateral is outside both, regardless of how sound its balance sheet is.

Assessment. This branch has become more probable relative to the pre-2026 configuration, not less, and the ECB's February 2026 decision is the evidence. A central bank does not broaden standing access to a backstop facility while citing structural shifts in the international financial system unless it judges the perimeter problem to be live. The change reduces the severity of the branch without reducing its likelihood.

What would falsify it. EUREP or FIMA usage disclosed at material size while swap balances remain near zero would confirm this branch is running. Sustained silence across all three channels during visible currency pressure would suggest the stress is being absorbed by reserve sales rather than official liquidity.

Boundaries the Evidence Imposes

Four conditions break the framework above, and each has a precedent.

A zero balance is not evidence of calm. This is the 2023 lesson in reverse. Because the perimeter tier cannot draw on the Federal Reserve's swap line at all, an episode concentrated outside the five counterparties produces the same near-zero reading as a genuinely quiet market. The line item measures core-tier usage, and only that.

Reporting lag exceeds the useful horizon. The H.4.1 is weekly. A funding squeeze that resolves or escalates within three trading days will be visible in the balance sheet only after the decisive period has passed. Anyone using this series as a real-time input is using it wrong; it is a confirmation instrument.

Collateralised facilities are not substitutes for swap lines. FIMA repo and EUREP convert existing high-quality securities into cash. A swap line creates dollars against a foreign currency the counterparty issues. Those are different economic operations. Treating the 2026 EUREP broadening as an extension of the swap network overstates what changed — it extends the collateralised tier, which helps entities that already hold the right assets.

Backstop pricing distorts the signal. Both the FIMA facility and the swap lines are priced above normal market rates by design. Usage therefore starts only after private channels have already deteriorated. The absence of usage is consistent with private markets clearing at spreads that are wide, stressed and still cheaper than the official rate. There is a range of conditions in which nothing appears in the data and conditions are nonetheless worsening.

What to Watch Next Week

• The H.4.1 release, published Thursdays. The central bank liquidity swaps line, currently $132 million. Any move into the billions is a regime change; movement within the low hundreds of millions is operational testing.
• Announcements of a shift from weekly to daily 7-day tenders. This has been the leading official signal in both 2020 and 2023, and it precedes any balance sheet movement by at least a week.
• Reappearance of 84-day operations. A tenor extension communicates expected duration rather than expected size, and the two have not moved together.
• EUREP operational reporting following the 24 July 2026 implementation. First-year usage will indicate whether the broadened access is being treated as a live facility or as a dormant statement of intent.
• Foreign official reverse repo balances at the Federal Reserve, $349,641 million for the week ended 12 August 2026. A sharp decline would indicate foreign official accounts drawing down dollar cash before turning to any facility, which typically comes first.

Concrete Framework — The Order That Matters

1. Establish the baseline before the episode. Record the current swap balance, the current operation frequency, and whether 84-day tenders are being offered. As of 12 August 2026 those are $132 million, weekly, and no. Deviations are only meaningful against a written baseline.
2. Classify the stress by tier before interpreting the data. Determine whether the institutions under pressure sit within the five standing counterparties, within the collateralised tier, or outside both. The same H.4.1 reading means three different things depending on the answer.
3. Order the indicators by lead time, not by prominence. Frequency announcements lead. Tenor extensions follow. Counterparty identity comes next. Balance sheet volume is last and confirms rather than predicts.
4. Track both official channels, not one. Swap line balances alone are now an incomplete picture. FIMA repo activity and, from the third quarter of 2026, EUREP activity cover the collateralised tier where the swap series is structurally blind.
5. Set a falsification threshold in advance. Decide, before the episode, what reading would move the assessment from Scenario A to Scenario B — a specific dollar figure across a specific number of weekly reports. A threshold chosen during the episode will be chosen to fit the position.
6. Recheck the perimeter every quarter. Membership and terms change. The 31 October 2013 standing arrangements, the 28 July 2021 FIMA conversion and the 14 February 2026 EUREP broadening were each announced with little market reaction and each altered who can obtain official liquidity and on what terms.

The swap network's contribution during a crisis is mostly invisible in the series that purports to measure it. The 2008 and 2020 episodes made the instrument famous for its drawdowns; 2023 showed the drawdowns are not the mechanism, and the 2026 architecture change addresses a tier the drawdown series never covered. Whether that tier holds under a genuine squeeze is untested.

This article is analysis of policy mechanics and public data. It is not investment or financial advice.