Banks’ window dressing: relative scores and the choice to intermediate

What was decided, and what could change

On 16 October 2026, the Basel Committee on Banking Supervision (BCBS) finalised averaging for global systemically important bank (G-SIB) stock indicators.^[1]

The link to markets starts with the comparison group. If accepting another bank’s business changes the recipient’s assessment as well, another large bank may not take every displaced transaction on identical terms. A temporary year-end gap in service and the choice of who intermediates business throughout the year are different problems. The amendments invite a fresh examination of both choices.

The published standard alone cannot predict an individual bank’s response or a market price. Banks face constraints involving capital, funding, collateral and client revenues. Changing one assessment method will not create capacity if another constraint remains binding. This article follows three relationships—peer-relative assessment, the destination of transactions and the terms offered to clients—to distinguish conditions under which behaviour could change from those under which it might not.

A year-end photograph is not a year of business

A year-end stock records assets, liabilities or other amounts on one date. Ordinary business can move with seasons and clients’ payment schedules, so a year-end decline is not by itself proof of deliberate window dressing. The measurement problem arises when a bank temporarily contracts only around the reference date while performing much the same role before and afterwards. Does the assessment capture its usual role?

An annual average reduces the influence of a short-lived reduction. Cutting an amount briefly has a different effect on the average from reducing it for a sustained period. This could weaken the incentive to step back only at year-end. But a bank seeking a lower average could instead reduce business for longer. Better measurement and greater service supply are not equivalent outcomes.

Our discussion of MAS minimum cash balances focused mainly on one bank’s required cash and its allocation across days. Under G-SIB assessment, a bank can maintain the same average amount yet move in relative terms when the comparison group changes. The analysis therefore requires both a bank’s choices across time and its relationship with other banks.

A relative score includes the other banks

Relative scoring already exists. Individual indicators generally compare a bank’s amount with the sample aggregate; methodology-specific rules combine indicators into an overall assessment.^[3]

The denominator is not a budget controlled by the individual bank. Growth elsewhere in the sample can lower the share represented by an unchanged amount at that bank; contraction elsewhere can raise it. Other banks’ decisions therefore enter its assessment. Its own financial statements alone cannot explain the change: expansion at the bank must be separated from contraction among its peers.

The ratio does not directly measure the proportion of a bank’s assets that is dangerous, nor the share of borrowers whose ability to repay has deteriorated. The assessment concerns systemic importance. A larger relative role is different from immediate financial distress. Reading a higher score as evidence of a credit crisis would misunderstand the purpose of the framework.

When one bank shrinks, the others’ shares can rise

Consider one indicator and, purely for explanation, a sample containing only banks A and B. Let their amounts be a and b. A’s share is a divided by a+b. If both amounts are positive and A alone reduces its amount, both A’s numerator and the sample denominator fall. A’s share nevertheless falls, while B’s share rises even though B has changed nothing. This illustrates a mathematical relationship, not a forecast for real banks.

B need not take over any of A’s transactions for this to happen. Whether A’s activity disappears or moves outside the sample, a smaller sample total can raise B’s share. This is an easy place to confuse business transfers with assessment changes. A rising share at B does not establish that it won A’s clients. Its own amount and the denominator must be examined separately.

Even a reduction at A does not determine the direction of its overall assessment if other indicators grow or other adjustments apply. The calculation explains one indicator; it is not a model of capital requirements or bank rankings. An assessment of an actual bank requires the relevant reporting item, counterparty and consolidation boundary. The simple ratio cannot be substituted for its complete G-SIB score.

Peer contraction can change an unchanged bank’s assessment

Now keep A’s amount unchanged and reduce only B’s. A’s numerator is constant, but the combined denominator falls, so A’s share rises. Calling this aggressive expansion by A would be inaccurate. A framework measuring relative importance can show a bank taking a larger role in the whole even when it has done nothing. Assessment changes without a change in behaviour must be distinguished from changes in management strategy.

For management, this also means that the future assessment cannot be fixed entirely through the bank’s own decisions. A plan using almost all available room can be upset by peer contraction. Holding extra room against that uncertainty could leave some intermediation capacity unused. The margin each bank chooses will depend on its management targets and assumptions about its peers.

For clients, an unchanged bank balance does not guarantee unchanged terms in the following year. Revenue targets and funding costs can change too. When a bank changes its terms, however, a client can distinguish its own transaction needs from the bank’s wider allocation choices and peer effects. Asking which constraint changed which term is more useful than treating the word “regulation” as a complete explanation for a price increase.

Proportional contraction need not change relative positions

A different result follows if every bank reduces the same indicator by the same proportion. Each numerator and the sample denominator shrink proportionately, leaving simple relative shares unchanged. Total market balances can fall without changing each bank’s share. Stable scores therefore cannot establish that the supply of intermediation has remained stable.

This is not a one-line proof that relative assessment is defective. Comparing who plays the largest role in a system is a different task from measuring how much transaction capacity the system provides. A measure of relative importance cannot also serve as a complete test of aggregate market health. Analysis needs amounts measured consistently and services available to clients alongside ranks and shares.

These three comparisons show why movements in scores alone cannot identify who withdrew or how much capacity remains. Comparing available collateral at a common time requires consistent measurement conditions; this case also requires a consistent bank sample. Only the same indicator, period and comparison group allow changes at the bank to be separated from changes at its peers.

Transfers within the sample can preserve the denominator

Suppose B takes over the business that A gives up. If the same amount enters the same indicator and other effects can be set aside, B’s increase offsets A’s reduction and the sample total is unchanged. A’s share falls and B’s rises, while the client’s transaction continues. The central issue is who performs the function within the sample, rather than the disappearance of that function.

In real transactions, netting conditions, collateral treatment and consolidation boundaries can mean that the recipient records a different amount from the original bank. A reduction in one item can also accompany growth in another. The transfer of contracts with the same nominal amount does not prove that the sample denominator is constant. The comparison is valid only under its stated assumption of equal recognition in the same indicator.

The recipient may consider more than additional fee income. Accepting business can increase its relative role and use capacity that could have served another client. A transaction unwanted by the seller may therefore be difficult for the buyer to accept. The existence of another bank is different from immediate replacement at the same price and tenor. Clients’ alternatives depend on the recipient’s capacity.

Leaving the sample is not the same as eliminating risk

If the recipient lies outside the assessment sample, A’s reduction in an indicator can enter the sample calculation without the recipient’s increase entering its denominator. Under that assumption, the sample total falls although the client’s transaction continues in the economy. A smaller number within the measured banking group is different from business disappearing across the financial system. It does not establish that the outside recipient is safe.

Funding stability, redemption terms and collateral calls can differ depending on whether the recipient is a bank, a fund or another intermediary. The final amendments do not provide evidence of actual destinations or transfer amounts. Treating migration outside the sample as inevitable would invent unverified concentrations and liquidity risks. Such claims require separate evidence about recipient balances and contracts.

The distinction matters in evaluating supervision. A lower score at a sample bank does not prove that the policy reduced risk, but movement outside the sample is not automatically a failure either. Distribution can help if recipients have adequate capital, stable funding and an ability to continue service. Evaluation must add the contract, holding period and allocation of losses to the question of who holds the business.

If business disappears, examine the client’s alternatives

A third possibility is that a transaction does not move to another intermediary but ceases to occur. If a client voluntarily reduces unnecessary borrowing, the lower balance is not evidence of a supply shortage. If a client wants to continue but cannot find a counterparty, intermediation contraction can reach the real economy. The same fall in balances has different meanings when driven by demand choices and supply constraints.

Completed volumes alone cannot distinguish them. Analysis needs requested amounts, offered limits, reasons for rejection and the availability of replacement counterparties. Even an unchanged quoted price can accompany exclusion of smaller clients or shorter commitments. Examining quantities, tenors and the range of counterparties together can reveal adjustments absent from price statistics.

The fact that credit supply depends on more than policy rates connects with our guide to bank lending and the credit cycle. But these amendments do not imply a uniform reduction in ordinary corporate lending. The first task is to connect a potentially affected business to the assessment item, then establish clients’ reliance on that business for funding or investment. A broad reference to bank regulation cannot justify direct claims about every loan or household.

Does a smaller stock mean less business?

A stock is an amount outstanding at a point in time; it differs from the volume of business conducted over a period. Repeated short transactions can raise turnover without greatly increasing average balances. A few long commitments can produce a substantial stock. This relationship matters when considering whether banks might change contract lengths after averaging is introduced. A reduction in stock is not necessarily an equal reduction in service volume.

Unchanged turnover does not guarantee unchanged usefulness to clients. Replacing a long commitment with repeated short renewals makes the client depend on a new decision each time. Business may continue in calm markets, yet funding becomes less dependable if renewal can be refused during stress. Efficient turnover and the assurance of a long commitment should not be counted as identical services.

Frequencies vary; flow indicators remain distinct.^[2] Knowing a business’s year-end balance is therefore insufficient to calculate its change under the new measurement. The relevant throughout-the-year data are needed; turnover or a closing-date number cannot substitute for an average stock.

How banks choose where to use scarce capacity

A bank’s decision to continue a transaction depends on more than its revenue: it also depends on the value of business forgone elsewhere. If the same capital or funding could serve several clients, a standalone profit calculation is not enough. Long-term relationships, revenues from other services and the timing of balance-sheet use can enter the bank-wide allocation decision. Relative assessment is one input to that choice.

Under the existing G-SIB capital framework, additional loss absorbency is met with Common Equity Tier 1 capital, and requirements differ across buckets. It is not equivalent to setting aside cash in a separate box; the capital requirement is expressed against risk-weighted assets.^[4] Confusing that denominator with the systemic-score denominator produces the false idea that an assessment change immediately creates an equal cash shortage. Overall assessment and the applicable bucket connect the two.

An indicator’s small movement therefore need not mechanically produce a proportional change in required capital. The current bucket, distance from its boundary and the strength of other regulatory or management constraints matter. Avoiding a bucket change may be important for one bank while another constraint binds first elsewhere. Without bank-level disclosures, a uniform estimate of additional equity or reduced funding supply cannot be justified.

In securities financing, separate price from available quantity

Securities financing provides a concrete route into markets. In a repo, securities and cash are exchanged alongside an agreement to reverse the transaction later. The cash provider’s reverse-repo perspective likewise combines a purchase with an agreed subsequent resale.^[7] If a bank uses its balance sheet to intermediate this business, quantity, tenor and counterparty become allocation choices.

If one bank steps back only at year-end and another has ample room, business may move with little interruption. If recipients also constrain capacity at that time, clients may face different limits or tenors. Average rates on completed trades then omit the inconvenience of trades that never occurred. Whether clients refused quantity found another counterparty is part of an assessment of market functioning.

Our guide to dollar liquidity and funding flows helps connect funding markets to international asset prices. But these amendments alone do not establish a dollar shortage or falling bond prices. Demand, central-bank supply, collateral composition and individual banks’ funding conditions interact. The distinctive channel here is that relative assessment can enter the recipient intermediary’s choice within that wider setting.

For businesses, dependable commitments matter

Businesses operating internationally plan not only payments and investments but also their ability to obtain funds when needed. A change in banks’ offered quantities or tenors can lead them to consider other banks, prepare collateral or stagger renewals. The useful question is which business connects to which contract term, rather than how prominently regulation appears in the explanation. Services with the same name can differ in their availability at the required time.

There are several steps before an effect reaches business decisions. Changed measurement must alter a bank’s policy; that policy must produce terms worse than the firm’s alternatives; and the difference must affect investment or inventory management. Another institution or internal funds may absorb the adjustment along the way. A straight line from one regulation to lower capital spending overlooks substitution and absorption.

Useful checks extend beyond the offered rate: distinguish the promised amount from the amount actually available, maturity from the next renewal date, and the conditions under which the counterparty may reconsider support. Several bank relationships may provide less diversification than expected if the banks face the same constraint and contract together. Conversely, differing funding sources and contract terms may absorb one bank’s withdrawal effectively.

A higher reporting threshold is not a capital relaxation

The reporting threshold in Figure 4 is not a ratio specifying how much capital a bank must hold. Mixing a reporting threshold with a capital-adequacy ratio can falsely suggest a large relaxation in prudential requirements for the same bank.

Size alone does not determine sample membership.^[3]Treating size alone as proof of exclusion would distort an analysis of transferred business. What matters is actual membership of that year’s sample, not the counterparty’s name or apparent size.

A change in the boundary also complicates time comparisons. If entry or exit changes the total, movements in shares cannot be explained entirely by business expansion or contraction. Holding the old sample constant and using each year’s official sample answer different questions. A chart combining years must identify changes in bank coverage rather than presenting them as uninterrupted changes in business activity.

Separate the data year, assessment year and capital date

Figure 5 distinguishes the stages.^[2]A single year should not imply that data collection, assessment and capital application all change together.

Banks may need consistent definitions across departments and systems before collecting the assessment data. Accurate year-end disclosures do not prove that daily or monthly records can be assembled on the same basis. If records cannot be reconstructed later, preparation immediately before application is difficult. A future implementation date therefore does not establish that management consideration will wait until that year.

Preparation is different from contraction. One bank may change data collection while another revisits contracts or business allocation. Early developments need to be separated into systems investment, disclosure changes and actual commercial policy. Preparation costs alone do not prove reduced market capacity; the evidence must extend to terms offered to clients and completed business.

Reliable averages have operational value

Comparing balances through a year assumes that each observation measures the same thing. Different cut-off times, inconsistent classification of the same contract or incomplete treatment of consolidation changes can undermine comparison even with more observations. Higher frequency and better quality are related but different tasks. An average does not automatically repair inconsistent underlying records.

Building those records costs money, but it could also help management understand ordinary balance-sheet use. Knowing which clients and businesses use capacity at which times may create scope to adjust maturities or departmental allocation instead of mechanically cutting transactions. This is a possible operational benefit if better records inform decisions, not an outcome already demonstrated by these amendments.

Do not attribute every year-end contraction to this framework

Year-end contraction need not have a single cause. Client demand, bank funding conditions, internal risk controls and other regulatory or accounting considerations can affect behaviour at the same time. A remaining year-end dip after the amendments would not alone prove the policy useless. Nor would its disappearance prove better intermediation if weaker demand merely removed the seasonal difference.

The official abstract of a 2024 Committee working paper reports causal evidence that responses to the G-SIB framework contribute to window dressing. This is research by its authors, not an evaluation of the final amendments now announced.^[6] Evidence about past incentives can inform the rationale for reform; it is not a number measuring future success. Research on a different period and policy cannot be substituted for a post-implementation forecast.

Evaluation requires comparisons tied to the particular business and measurement change. Do otherwise similar activities respond differently, and was the trend already present before reform? Interest-rate levels and curves are separate market influences. Our guide to Treasury yields and the yield curve can help distinguish changes in the wider financial environment from bank-specific behaviour.

SG Group View: measure assessment quality and continuity separately

SG Group places particular weight on recipient economics when assessing continuity. The reason for accepting new business is easily overlooked beside the withdrawing bank’s balance-sheet reduction. If a recipient considers fees, future client business and the value of capacity elsewhere, competition for the transaction will involve more than price. A bank with spare capacity could win a client by offering a longer commitment.

An easy overstatement is that curbing brief window dressing will resolve year-end funding-market instability. Other constraints can preserve the pattern. Measuring assessment quality separately from services shaped by competition and alternatives makes a gap between policy objectives and outcomes easier to identify.

Potential beneficiaries include clients troubled by temporary year-end service gaps and banks whose assessment becomes better aligned with their usual role. Potential costs fall on banks building reporting systems and clients whose terms change with business allocation. But gains and costs cannot be divided uniformly between banks and clients. Relationships, alternatives and the ability to pass costs into prices differ even within the same bank.

Three possible paths require different evidence

The first path is less brief year-end contraction with little change in the volume and tenor of business through the year. If clients retain similar quantities and commitments while the year-end gap narrows, that would support improved continuity. A higher year-end balance should not automatically be read as deterioration: activity previously absent only on the measurement date may be appearing in a more representative figure.

In the second path, other constraints remain binding and neither year-end behaviour nor client terms changes much. Assessment could improve while the supply effect remains limited. In the third, banks reduce outstanding amounts or long commitments throughout the year. A smaller year-end dip would not be an unambiguous improvement if available quantities and tenors also fell. Recipient capacity would need to be examined at the same time.

The announcement does not provide the bank-level data needed to attach probabilities to these paths. As in our guide to building macro scenarios, pairing conditions with evidence is more useful than an unsupported ranking of forecasts. Relevant evidence includes not only bank explanations but quantities and tenors available to the same clients, expansion at recipients and the sample aggregate. The framework should permit revision when different evidence emerges.

International standards still require national implementation

Basel standards are intended to be incorporated into jurisdictions’ legal frameworks. The Committee’s Charter describes implementation through national rulemaking.^[5] Publication does not automatically rewrite bank contracts worldwide on the same day. Tracking application requires separating the international timetable from measures adopted by the authorities supervising the banks concerned.

For a bank operating across borders, client location need not match the perimeter of supervisory consolidation. Business serving a firm in one country is therefore not necessarily a one-country issue. Reallocation by a large group serving international clients can transmit changed terms to several markets. The standard alone, however, provides no basis for ranking countries by their eventual exposure.

International effects depend on which banking groups perform which functions and who can replace them. Numerous stable local providers may absorb an adjustment. Several nominal counterparties within the same group may offer less choice than they appear to. Consolidated providers and actual contracts are more useful units than country names for tracing the international channel.

Read scores alongside amounts in the next disclosures

For future disclosures, first establish the official sample and indicator definition, then place the bank’s amount, the sample aggregate and its relative share side by side. Distinguishing a share-only movement from a change in amounts makes peer effects and sample turnover easier to identify. Currency conversion and acquisitions or disposals can also affect measurement; a changed number need not have a single cause.

Next compare the year-end observation with the distribution through the year. The same average can describe a bank with large temporary swings or one with a flat profile, with potentially different client experiences. Unpublished daily data should not be filled in by guesswork. Separate observable points from the bank’s explanation and preserve gaps as unknown. Replacing missing periods with zeros or constants can manufacture a before-and-after difference.

Finally, match client terms. Rate comparisons need the same transaction type, tenor and collateral conditions to distinguish regulatory effects from differences in the business itself. When bank explanations and market data disagree, examine scope and timing before declaring either wrong. Keeping bank assessment separate from aggregate service supply also allows improved measurement and reduced supply to be recorded together if both occur.

Final assessment: whose number falls, and who continues the work?

The question is whether the amendments move assessment away from the ability to present a small year-end number and toward an account of a bank’s role throughout the year. Understanding that direction requires peer-relative assessment as well as averaging over time. A single bank’s contraction, peer contraction and proportional contraction convey different information through the same ratio. The relationship between the measured number and the service also differs depending on whether the recipient is inside or outside the sample.

The next assessment should be read alongside evidence of which banks continued to offer clients commitments, and for how long. Connecting relative assessment to the work performed in markets requires evidence linking those two observations.