Banks: the cost of the December snapshot

Year-end is approaching. At our fictional bank, a desk lets a book of market financing mature, repays its own lenders and resumes the business in January. Fifty billion euros temporarily disappear from its exposures. The closing disclosure accurately describes that day. It says rather less about the preceding three hundred and sixty.

That gap between a day and a year has become a live issue for Basel. On 1 October 2026, following its 28–29 September meeting, the Committee announced that it had approved changes aimed at reducing window dressing by global systemically important banks. Details were scheduled for later in the month. Our search, closed on 10 October, did not locate a detailed final standard establishing the chosen observation frequency and implementation timetable. [1]

The public record already shows why the issue matters. A measurement date can influence how much capital a bank must maintain and how much financing it offers other market participants. A rule intended to protect the system also creates an incentive to rearrange business around its calendar. This investigation follows that incentive from the regulatory score to the investor trying to fund a bond portfolio over the holidays.

Five days in the frame

A global systemically important bank, or G-SIB, is an institution whose distress could cause substantial disruption elsewhere in the system. Basel’s score measures that relative importance through size, financial connections, cross-border activity, complexity and services that are difficult to replace. It is concerned with the consequences of failure, rather than an estimate of how likely the bank is to fail. [3]

Observable indicators make the framework workable. They also create a distinction between everyday activity and the amount recorded on the reporting date. Research presented by the Bundesbank in September 2024 documents stronger year-end exposure reductions at systemic banks, followed by a stronger rebound early in the next year. Banks close to a regulatory threshold face particularly sharp incentives. [10]

Consider a 365-day year. Our bank records €1 trillion of exposure for 360 days, falling to €950 billion for the last five. These are teaching assumptions, unrelated to the accounts of any real institution. The 31 December snapshot shows a 5% reduction. The average of all 365 observations is €999.315 billion, just 0.0685% below the ordinary level.

The measurement calendar explains the entire difference. An average of twelve month-end observations produces €995.833 billion. An average of four quarter-ends produces €987.5 billion. Identical transactions, taking place on identical dates, generate four different representations of the bank. These calculations illustrate options discussed around the March 2024 consultation. They do not establish the contents of the final changes approved in 2026. [2][11]

Window dressing can therefore involve transactions that genuinely take place. A loan has been repaid, funding has matured, or a contract has actually been terminated. The prudential problem is whether the selected observation represents the business adequately. A lasting reduction and a brief interruption can produce the same year-end number, although their effects over the rest of the year are very different.

That distinction matters when discussing responsibility. An allegation of false accounting would require a different evidentiary case. The issue examined here is an incentive embedded in the measurement method and its effect on behaviour.

A few score points and two billion euros of capital

In the international framework, each of the five broad categories accounts for 20% of the score. An institution’s indicator is divided by the aggregate for the assessment sample, converted into points and weighted. The indicators cover different stocks and flows: annual payment activity, for instance, is not a closing balance-sheet snapshot. The final score is rounded to the nearest whole basis point. [3]

The result helps determine an additional capital buffer. The first bucket, from 130 to 229 points, requires extra CET1 capital equal to 1% of risk-weighted assets. Between 230 and 329 points, the requirement rises to 1.5%. CET1 is the highest-quality layer of regulatory capital, principally ordinary shares and retained earnings after regulatory adjustments. [3][4]

Apply this mechanism to our institution. All the inputs are hypothetical. Other banks have a combined €19 trillion of exposure and remain unchanged. We fix the weighted contribution of the other categories at 133 points, with no binding cap or other adjustment. At €1 trillion of exposure, the bank contributes 100 points through size, producing an overall score of 233.

When exposure falls to €950 billion, the collective denominator also falls, to €19.95 trillion. The size contribution becomes 95.238 points. The total is 228.238, rounded to 228. In this illustration of the international schedule alone, the bank moves down one bucket. In practice, national authorities may delay the release of the associated buffer. With €400 billion of risk-weighted assets held constant, the associated surcharge falls from €6 billion to €4 billion.

The €2 billion is a difference in required capital, not that year’s profit. No corresponding payment appears in the bank’s cash account. If the capital it actually holds remains unchanged, its headroom above this requirement increases. Whether it can use that headroom depends on other requirements, supervisory expectations, distribution policy and management’s own safety margins.

This distinction is central to any discussion of cost. Equity supplied by shareholders has an economic cost and absorbs losses. An extra requirement can change returns on equity, prices charged to customers or the mix of businesses a bank chooses to run. Treating €2 billion of capital as €2 billion of annual expenditure would nevertheless be wrong. Nor does the calculation produce an automatic amount of additional lending.

A bank far from a threshold may lower its score without changing buckets. An institution just above a boundary can obtain a discontinuous benefit. The timing of its transactions then becomes unusually valuable. Under the daily average in our example, the rounded score remains 233, and the associated buffer remains €6 billion. The five-day interruption has almost disappeared from the annual measurement.

The financing that is allowed to expire

To follow the effect into the market, consider a repo: secured financing structured as a sale of securities with an agreement to repurchase them. An investor obtains cash against bonds. A bank can stand between that investor and a supplier of cash, borrowing from one and financing the other. Short maturities make these transactions especially responsive to reporting dates. [7]

In our simplified diagram, a money-market fund supplies €50 billion to the bank, which finances a bond portfolio for the same amount. Securities move as collateral under the contracts. The two relationships involve different counterparties and are assumed not to qualify for netting that would remove this book from the balance sheet. The exposure measure is assumed to change by the same €50 billion, with other regulatory adjustments fixed.

The bank lets the book mature before the reporting date. The investor repays €50 billion. The bank uses that cash to repay its lender. A financing receivable disappears from assets, and a financing liability disappears from liabilities. Ignoring fees and trading results, equity is unchanged. This simultaneous contraction explains how the institution can temporarily become smaller without selling branches or calling in household mortgages.

The money-market fund has its cash back. The investor must find another way to keep financing its portfolio: another bank, direct funding, its own liquidity or a reduction in positions. Each option reallocates cost or risk. Cash available somewhere in the system and a bank’s willingness to put a particular transaction on its balance sheet are separate constraints.

For one institution, the decision may look rational. It compares revenue forgone for a few days with the expected effect on regulatory indicators. Across the market, similar decisions can remove intermediation capacity simultaneously. Any collective cost may then appear on counterparties’ books: funding arranged earlier, different terms, assets sold or business deferred.

Netting is important throughout. Two books with the same gross transaction volume can use very different amounts of balance sheet, depending on counterparties and legal terms. A serious analysis has to identify the exposure recognised by the rule. Adding every transaction’s face amount would produce a much larger and much less useful number.

Year-end contractions in the research data

A paper by Claudio Bassi, Markus Behn, Michael Grill and Martin Waibel, published by the ECB in February 2023, uses euro-denominated repo transactions from September 2016 to June 2021. Reconstructing balance-sheet exposures after eligible netting, the authors estimate an aggregate contraction of €66.3 billion at the first three quarter-ends and €131.7 billion at year-ends. Their event-study reference is 28 trading days before the reporting date. [7]

These estimates describe the paper’s sample and period. They are neither bank losses nor a measure of the entire European repo market, and they say nothing directly about the turn into 2027. The second number includes the year-end effect: the two bars in the figure must not be added together. We have not recreated the confidential underlying transaction dataset.

Separating a regulatory constraint from a seasonal fall in demand takes more than a chart with a December dip. The authors compare transactions of different maturities around reporting dates, including loans that cross the date and loans that expire beforehand. Their tests link behaviour to leverage-ratio and G-SIB incentives. That is a research finding within a particular identification design. [7]

The broader evidence is not equally strong across markets. A study released by the Basel Committee in March 2024, using 70 banks in 16 jurisdictions from 2010 to 2022, finds particularly strong evidence for derivatives notionals. Its repo results, using less granular data, are weaker and only marginally significant at the 10% level. The authors state that limitation explicitly. [8]

This qualification matters when interpreting December moves. The systemic surcharge interacts with the leverage ratio, client demand, other requirements and market conditions. The leverage ratio compares capital with a broad exposure measure. The G-SIB surcharge depends on systemic classification and then applies to risk-weighted assets. The same transaction can affect both calculations through different mechanisms. [3][4][7]

Fewer contracts can be a genuine improvement

The strongest counterargument concerns operations that actually reduce complexity. Derivatives compression terminates offsetting contracts and replaces them, where their terms permit, with a smaller set. A Federal Reserve research note published in January 2020 examines this practice and its relationship to US systemic indicators. Its historical description is not a comprehensive account of American rules in 2026. [9]

Take two hypothetical interest-rate swaps between the same bank and counterparty. Under the first, the bank pays fixed and receives floating on €100 million. Under the second, it takes the opposite position on €90 million. To make the comparison exact, assume identical dates, maturity, fixed rate and floating index, with no other contractual differences.

Following agreed compression, a single €10 million swap remains in the first direction. Gross notional falls from €190 million to €10 million, a 94.7% reduction. In our simplified setup, net interest-rate exposure remains that of the €10 million swap. These notionals are bases for calculating interest, not €190 million of principal exchanged between the parties.

Such a change can durably reduce the number of contracts that must be processed and some associated operational complexity. Proximity to December does not, by itself, make it artificial. A prudential benefit and an economic improvement can coexist. Establishing a merely temporary rearrangement would require examining subsequent rebuilding of positions and the risks remaining after the transaction.

GFMA, IIF and ISDA emphasise this distinction and seasonal client demand in their consultation response. They represent financial-industry interests. That interest is explicit, but the objection still forces an analyst to identify the relevant measure: contracts, gross notional, net exposure, balance-sheet usage or default risk. Research on threshold incentives adds another part of the explanation. It does not make those measures interchangeable. [12]

An average of which days?

Replacing one observation with an average reduces the influence of a brief interruption. The frequency of that average still creates different incentives. In our initial model, a five-day contraction has a substantial effect on four quarter-end observations and a tiny effect on 365 daily observations.

Now change the behaviour. The bank repeats the five-day contraction at every quarter-end. All four quarterly observations are now €950 billion. The daily average, comprising twenty days at €950 billion and 345 days at €1 trillion, remains €997.260 billion. One prepared annual snapshot has become four prepared snapshots. This demonstrates a sampling limitation, rather than a prediction of how banks will respond to the final reform.

Frequency also has a cost. In its 7 June 2024 response, the French Banking Federation describes the difficulties of collecting and validating high-frequency consolidated data, and favours an average of quarter-end amounts. A banking group must align entities, definitions and controls. That is a different undertaking from downloading daily market prices. The argument identifies an implementation constraint; it does not by itself determine the best collective trade-off. [11]

The denominator introduces another complication. A bank’s relative score also depends on what peers report. Return to bank A, whose exposure falls from €1 trillion to €950 billion. Bank B remains at exactly €900 billion. Its share of the aggregate nevertheless increases from 4.5000% to 4.5113%, because the sample total has declined. Its weighted size contribution rises from 90 to about 90.226 points. Holding everything else constant, A has changed the measurement of its peers as well as its own.

If every bank reduces the indicator by the same percentage, relative shares remain unchanged. A collective reduction in the indicator and a change in relative classification are different outcomes. This arithmetic complicates estimates of capital savings. A simulation that holds the aggregate denominator constant while banks reduce their exposures can overstate the benefit.

Reference dates and reporting conventions also vary between jurisdictions. December provides a useful entry point, but the FSB list documents other financial-year reporting dates. A global assessment must align the calendar, data definitions and sample before applying the chosen methodology to the relevant institutions. [5]

The cost turns up on somebody else’s invoice

The value of a reporting date may be small for one trade and material for a large portfolio. Suppose a borrower must refinance €500 million for five days. An additional annualised spread of 100 basis points, or one percentage point, adds approximately €69,444 of interest using an ACT/360 day count. At 25 basis points the additional bill is €17,361; at 300 basis points it is €208,333.

These rates are assumptions, not year-end market quotations. The calculation covers only incremental interest over five days. It excludes fees, collateral changes, market moves and the cost of a forced asset sale. Its purpose is to translate an annualised price into money actually paid over a short period.

Availability may matter more than price. A counterparty offering to pay extra does not automatically obtain a line. The intermediary may reserve capacity for other clients, face its own balance-sheet constraint or decline a particular maturity. The borrower may instead arrange longer funding earlier and pay for certainty that it will remain available through the reporting date.

The direction of rates also depends on what is scarce. When market participants urgently need a specific security, its holder may obtain exceptionally cheap cash financing in exchange for supplying it. Scarce collateral therefore has different pricing implications from excess demand for cash. ICMA discusses this distinction in its review of the turn from 2025 into 2026. [13]

That recent experience offers a useful counterweight. An anticipated year-end premium eroded as the date approached, and final volatility remained relatively contained. Prefunding, available liquidity, market positioning and central-bank facilities formed part of the environment. ICMA is a market association, and its account draws partly on participant input. Most importantly, the episode gives no reason to treat every December as an impending crisis. [13]

The test comes after the snapshot

The regulatory calendar itself has several layers. The G-SIB list released by the FSB in November 2025 relied mainly on end-2024 data; associated increases in buffers apply from January 2027. A measurement, an annual announcement and the resulting requirement can therefore fall in three different years. Reductions follow different timing provisions, including supervisory discretion. [5][4]

The October 2026 approval begins another sequence. The final technical choices must be read and placed within each jurisdiction’s rules. Basel has no supranational legislative authority of its own. The consolidated framework already includes an SCO40 version, updated in November 2023 and effective in 2027. It predates the 2026 decision; its timetable belongs to an earlier change. [1][6][14]

Financing over the whole year offers a concrete test of the reform. Do reporting-date troughs become smaller? Do average exposures increase durably? Do banks move between buckets? Does the activity migrate to other intermediaries? Smoother seasonal patterns could coexist with less average financing capacity. Equally, pricing spread more evenly through the year could make the service more predictable without reducing it.

The comparison must also involve genuinely comparable transactions. Currency, collateral quality, tenor, counterparty, netting and contractual protection all affect repo economics. An isolated rate spike, a falling balance sheet or a changed G-SIB bucket cannot establish a bank’s motivation. The research examined here identifies behaviour in historical samples. It does not support a list of banks accused of misconduct in 2026.

The economic issue is clear. A bank can optimise a measure of its activity and shift part of the cost of the reporting calendar onto institutions using its balance sheet. A better rule would make the observation more representative, preserve genuine reductions in complexity and keep the cost of verification proportionate. The ultimate question concerns the service delivered over 365 days: who will finance the positions, on what terms, and with how much capital available to absorb losses?

Scope and method

Documentary research closed on 10 October 2026, with no interviews or access to banks’ confidential positions. Simulations use a fictional institution, exposures in billions of euros and fixed parameters to isolate mechanisms. The 365-day calendar is a teaching convention, not a statement about any prescribed regulatory averaging method. Score calculations apply the international schedule alone, without recomputing national requirements, supervisory adjustments or other ratios.

The empirical figure uses estimates published on page 14 of Working Paper 2771, which was visually inspected. Other quantitative figures show explicitly identified l0g calculations. The data and assumptions below each figure set out the calculations. The sources listed below identify document dates, research scopes and the limits of the conclusions.

Further reading

The collateral and financing chain is examined in Repo, the liquidity factory. The investigation into European banks and sovereign debt follows a different mechanism: the transmission of a rate shock through bank balance sheets.