What risk management cannot promise
07 July 2025
In the sixth instalment of this ongoing series, Cyril Louchtchay de Fleurian, head of securities finance and balance sheet strategy at Capteo: Strategy & Management Consulting, talks risks, repo, and cost
Image: stock.adobe.com
Repo looks like a sleepy little business. That is what most risk departments still believe. In truth, it is a dense, unstable cluster of risks — and the fact that hedge funds have made it their favourite P&L tool is no accident; they have understood that its apparent simplicity is an optical illusion. 2026 has just confirmed it.
The wrong tool, by design
Splitting risks into ‘classic’ and ‘recent’ suggests the recent ones are new types of risk — only in part. What has changed comes down to three distinct moves. First, dormant risks that have been repriced. Sovereign and sanction risks have always existed, but they sat in the tail, treated as negligible, until the Russian asset freeze of February 2022. The risks did not just emerge; its expected value jumped by an order of magnitude, dragged up by geopolitics. Second, risks that have changed regime: intraday and the direction of reserves. The mechanics are the same as before, but what was once comfortable has become binding. Third, risks that are structurally new, created by new participants (NBFI), and by the evolution of the infrastructure itself: T+1, mandatory clearing, distributed ledger technology, SOFR reflexivity. Those are genuinely new.
All of these risks converge on a single point: the convergence of unavailabilities. Whether collateral is frozen by sanctions, locked by a central securities depository (CSD) outage or held hostage by ransomware, the balance sheet effect is identical — the bank can neither unwind nor mobilise. The distinction between political, operational, and cyber causes dissolves into a single liquidity risk. Repo stops being a homogeneous pricing risk and becomes a portfolio of circular threats, interconnected in their dynamics as much as in their mechanics. That is the central insight.
What really separates the two families is not their vintage but four criteria that together dictate how a desk should treat them — and, as we will see, drive their intensity directly.
Maturity of the framework. Classic risks rest on a settled doctrine: Global Master Repo Agreement, Basel, netting opinions, Financial Stability Board haircut floors. The risk is known, written, litigated. Recent risks operate within a framework still under construction — or absent altogether: the Digital Operational Resilience Act only just in force, the mandatory clearing timetable still in flux, the legal doctrine around tokenisation unsettled. One does not discover the existence of a classic risk; one can discover a recent one, precisely as it materialises.
Nature of the risk. Classic risks are largely idiosyncratic and modellable: they are measured, provisioned, stress-tested as routine. The second family gathers risks that are often systemic and non-stationary; their distribution shifts with the geopolitical or monetary regime, which makes history a poor guide to what comes next.
Failure mode. Classic risks deteriorate gradually, which buys time to act: margin call, netting, orderly liquidation of collateral. Recent risks materialise in a discontinuous and binary way: collateral is either accessible or it is not, settlement either clears or it freezes. There is little or no slope to claw back on between the two states.
Hedgeability. This is the decisive criterion. The first family hedges or collateralises: one hedges rates, margins a counterparty, nets a default. The second is largely unhedgeable, and that is precisely its signature. One does not hedge or net a sanctions freeze, a Fedwire outage or ransomware on one’s CSD. If recent risks are structurally unhedgeable, then the entire risk management toolkit — pricing, modelling, hedging, margining, provisioning — is the wrong tool. By design, not by deficiency.
The second family is not “more of the same risk”. It is a break in hedgeability. The answer does not lie in pricing or modelling but in infrastructure redundancy, custodian diversification, and pre-positioned collateral. That is a different game entirely. A desk that meets second-family risks with first-family instruments — model, margin, hedge — has picked the wrong tool, and will only find out when the incident hits.
Three speeds of loss of control
These four criteria do more than classify. They predict how fast a risk escapes you. The less mature its framework, the more non-stationary it is, the more binary its failure, and the less hedgeable its nature, the faster and more irreversible the loss of control. That is precisely what the three tiers below measure — no longer the nature of the risk, but its operational intensity.
Tier 1 — existential, exogenous, T+0/T+1. Three signatures define them: total externality (the bank has no grip on the initial shock), immediate kinetics (the effect lands at T+0/T+1), and irreversibility (once the shock fires, options close). Neither the Liquidity Coverage Ratio, nor the Contingency Funding Plan, nor extra buffers will respond: the bank stays liquid on the balance sheet while paralysed in execution. This is precisely where repo turns into a systemic weapon.
Such risks are not managed, they are anticipated. Active mapping of infrastructure dependencies, real — not declared — diversification of channels, the ability to reallocate collateral before conditions deteriorate. Without a transverse mandate empowered to arbitrate, those decisions remain scattered, late, and unenforceable — in other words, useless at the moment that matters.
Three risks belong here. First, central clearing as a choke point: cleared repo is no longer an open market where price arbitrates access to funding, it is a closed club from which one can be excluded with no price signal at all. Second, the asymmetry of speed: in a constrained market, velocity stops being an execution parameter and becomes an instrument of dominance — whoever mobilises first shuts the door on the rest. Third, repo as indirect sanction: faster and often more effective than any formal legal mechanism, because it strikes before any recourse is available.
Tier 2 — orchestration failures and self-inflicted risks. These risks do not fire alone; they turn a manageable stress into a crisis whenever decision making is scattered and unranked. The bank then penalises itself through three channels. Internal desynchronisation, which tips stress into crisis. Technical reputation, which hardens into an informal gating criterion on liquidity access and marginally degrades every condition. And pre-emptive anticipation, where haircuts and eligibility criteria turn forward-looking and exclude market participants before any visible crisis. These are amplifiers, not detonators: absorbable in an orchestrated bank, existential in a siloed one. Three levers neutralise them: a unified orchestration of collateral, cash, and margin flows; a standardised set of behaviours under stress; and an anticipatory read of the buffer’s vulnerabilities. Without them, these risks stay invisible until they combine — which is to say, too late.
Tier 3 — structural risks. These trends trigger no crisis. They settle in over years and reshape the funding environment continuously. The effect is not a shock but an erosion: structural balance sheet cost drifts upwards, business models lose viability, relative competitiveness shifts quietly between banks. Predictable, yet systematically underestimated because diluted in time — no incident makes them visible before the gap is already entrenched. These are not managed, they are positioned: transverse treatment with a consolidated view, ongoing adjustment of buffer composition, anticipation of fragmentation zones, integration of every funding channel’s constraints into long-horizon trade-offs. This is the only category where time works in favour of the bank that moved early — and against the one waiting for a signal that will never come. Three markers compose it.
First, control over collateral eligibility: a quiet piece of rule-setting that redraws the boundary between financeable and non-financeable players. Second, fragmentation of liquidity pools, which ends a long-held assumption that a high-quality asset would be mobilisable in roughly uniform fashion across venues. Third, central bank repo, now an instrument of sovereignty: a tool of financial industrial policy that, through its operational parameters alone, organises the relative competitiveness of bank balance sheets.
Two further mechanisms escape this grid because they cut across all three tiers.
Model monoculture. Synchronisation of haircuts and margins through converging methodologies (CCPs, triparty, vendors) and automatic feedback loops. Models — VaR, stress add-ons, concentration add-ons, rating triggers, liquidity scores — tend to converge because they share the same inputs, the same vendors, and very similar logics. Under stress, this produces a synchronised cascade of margin calls and haircut revisions across infrastructures and counterparties: a liquidity shock that is sharper and faster, not because the fundamentals have changed, but because the models recalibrate all at once in the same direction. It is a risk distinct from eligibility. The asset may remain eligible, yet the haircut and margin function can jump on cliff effects driven by automatic recalibration — and that shock propagates mechanically through the system.
Indirect governance through liquidity. Repo — and more broadly secured funding access — disciplines markets. It also disciplines the institutions themselves, turning liquidity into a lever of indirect governance. A bank structurally dependent on a handful of CCPs, a few dominant dealers, and an increasingly conditional access to central bank liquidity gives up part of its strategic freedom. It no longer arbitrates its business model on return and risk alone, but on what remains financeable through the existing pipes. It becomes less able to push back on instructions, contest norms, or defend non-aligned choices without paying an immediate liquidity cost. The effect is concrete and quick: some activities become de facto undesirable because they consume too much high-quality collateral or worsen the funding profile; some clients become too expensive in liquidity terms even when they are profitable on P&L; some strategies become impossible to deploy without any formal prohibition being issued, simply because the infrastructure and its gatekeepers reprice or ration access to refinancing.
The asymmetry of error
Asymmetry first; the figure comes later. The first question is not what inaction costs, but which error one chooses. Treat a hedgeable risk as unhedgeable and the result is near-certain over-provisioning: the loss is bounded, visible, booked — a carry cost the balance sheet absorbs and a committee can arbitrate. Treat an unhedgeable risk as hedgeable and the cost is zero, until the incident. At that point, the loss is neither bounded nor provisioned: it represents the gap between an optionality that one believed one held and a frozen position one discovers the moment it stops unwinding.
The two errors are not symmetric. The first is paid continuously and can be steered. The second is paid only once, in full, with no warning. A system calibrated to minimise the visible cost optimises precisely the wrong error. Inaction is not a deferred expense; it is a short option sold at zero premium, with an unknown strike.
My desk had a short-term funding line, rolled every month for two years, at a spread that was genuinely unbeatable — from a single client. Every month, P&L and funding validated the choice: cheaper than any diversified alternative, and by a wide margin. The trader was, fairly enough, congratulated for the optimisation. In the desk committee, the question came up from time to time; the answer was always the same, and it was a good one — over two years, diversification would have cost a multiple of anything it could ever have protected. The roll fell on the last Thursday of the month; it had become a routine no one really looked at any more. The month the client needed his cash elsewhere, the line did not get more expensive: it vanished. It took us 20 minutes to realise it was not a delay. We had to refinance, in a hurry, a volume that had quietly swollen with time and habit. For two years, the optimisation had been booked as performance. It was a premium collected on a risk no one had ever named — and for good reason: the one who could have named it was precisely the one being congratulated.
The absence of transverse steering of executable liquidity does not produce only a single, identifiable shock. It also produces a continuous erosion, invisible in the usual performance indicators. These costs are neither a compliance failure nor a case of excess risk-taking: they come from trade-offs forced upon the bank — late, or over-cautious — imposed by operational uncertainty. The orders of magnitude that follow are not observed measurements. They are bounds modelled from published parameters, which any institution can rerun on its own data.
The only directly documented item is also the most telling: the extra cost of funding under stress. In September 2019, US repo went from two per cent to a peak around 9–10 per cent in a single session (BIS Quarterly Review, December 2019). The year-end reports of the International Capital Market Association’s European Repo and Collateral Council consistently document repo spikes of tens to hundreds of basis points on the European repo market at balance sheet dates. Assuming a differential of a few tens of bps between prepared mobilisation and forced mobilisation is therefore a prudent assumption — applied precisely to the volumes that cannot wait, at the moment they cannot wait.
The other items follow. Lacking fine-grained visibility on what is actually mobilisable, banks oversize their precautionary buffer. Two documented frictions feed that inertia: an asset encumbrance ratio of around 25 per cent on average across the EU (EBA) and an intraday settlement inefficiency of five to six per cent of volumes (ECB, TARGET Services 2024). Together they leave a fraction of the buffer inert at the very moment it is required.
For context, and outside the figures: that inertia is structurally worse in Europe than in the US, where cross-border collateral mobility costs around 65 per cent less in post-trade fees (AFME, October 2025). That is a market structure fragmentation point, not an input to the model. The immobilisation has a mechanical price: the trapped fraction multiplied by the cost of equity (around 10 per cent, EBA). The full calculation and its assumptions are set out in the box below.
One item is deliberately left unquantified. The preemptive or disorderly withdrawal from certain repo and securities financing transactions activities under stress erodes franchise and client relationships. No public data measures it, because it does not show up as a dated loss but as reduced capture of the recovery. Quantifying it precisely would be spurious precision; flagging it as a structural blind spot is the real thing. Excluding it from the figures makes them conservative — a lower bound.
1. Reference bank (assumed parameters)
2. Sourced parameters
3. Modelling assumptions (conservative)
4. Calculation - base case
Item A - buffer immobilisation
Inert buffer = €240bn × 8 per cent = €19.2bn
Annual cost = €19.2bn × 40 bps = €76.8m
ROE drag = €76.8m / €70bn ? 11 bps
Item B - extra funding cost under stress (annualised expectation)
Cost per episode = €50bn × 30 bps = €150m
Annual expectation = €150m × 0.5 = €75m
ROE drag = €75m / €70bn ? 11 bps
5. Result - recurrent structural cost
Reading: for a European G-SIB, the recurrent cost of not steering executable liquidity transversely sits, in the base case, at around €150m a year, or roughly 22 bps of ROE - borne every year, outside any crisis, and carried by no identified cost centre. What is robust is not the point estimate but the property: positive, recurrent, unattributed, and conservative (franchise erosion excluded).
Methodology note: the latest editions of EBA (encumbrance, cost of equity), ECB TARGET and AFME figures evolve over time. The balance sheet parameters of the reference bank are declared orders of magnitude, to be replaced by each institution’s own data for an institution-specific figure.
Taken on their own, each of these effects looks marginal. Taken together, they form a recurrent structural cost, borne year after year, regardless of any major crisis.
The cost no one carries
The price of inaction is not a loss. It is a rent. €152 million. 22 bps of ROE. Year after year, outside any crisis, for a bank that has neither taken too much risk nor breached compliance. It is the price of a massive blind spot, not of a fault. And a blind spot does not correct itself: it sits in no mandate, charges no cost centre, triggers no alert. It is simply paid — quietly and indefinitely.
The diagnosis ends there. What follows is no longer a question of measurement. If these risks are unhedgeable, then the answer is not a better model, a finer margin or a bigger buffer: those instruments treat the wrong error. The answer is organisational. It requires an organisation able to mobilise, transform, and move its liquidity under constraint, at market speed — not to measure it cold. Some banks have organised themselves accordingly. In most, collateral, cash, margins, and infrastructure dependencies are still managed in silos, optimised locally, and defended separately. The regime has changed. The organisation has not.
Those 22 bps are not a loss. They are a fee paid to inaction — recurrent, and entirely recoverable. The next articles in this series will show how an executable, cross-functional liquidity capability turns those silent, accepted costs into explicit and governed decisions — and why enterprise liquidity management will become as structural a function as asset-liability management or risk management. Not because it makes for a new org chart, but because €152 million a year tends to surface in the end.
The wrong tool, by design
Splitting risks into ‘classic’ and ‘recent’ suggests the recent ones are new types of risk — only in part. What has changed comes down to three distinct moves. First, dormant risks that have been repriced. Sovereign and sanction risks have always existed, but they sat in the tail, treated as negligible, until the Russian asset freeze of February 2022. The risks did not just emerge; its expected value jumped by an order of magnitude, dragged up by geopolitics. Second, risks that have changed regime: intraday and the direction of reserves. The mechanics are the same as before, but what was once comfortable has become binding. Third, risks that are structurally new, created by new participants (NBFI), and by the evolution of the infrastructure itself: T+1, mandatory clearing, distributed ledger technology, SOFR reflexivity. Those are genuinely new.
All of these risks converge on a single point: the convergence of unavailabilities. Whether collateral is frozen by sanctions, locked by a central securities depository (CSD) outage or held hostage by ransomware, the balance sheet effect is identical — the bank can neither unwind nor mobilise. The distinction between political, operational, and cyber causes dissolves into a single liquidity risk. Repo stops being a homogeneous pricing risk and becomes a portfolio of circular threats, interconnected in their dynamics as much as in their mechanics. That is the central insight.
What really separates the two families is not their vintage but four criteria that together dictate how a desk should treat them — and, as we will see, drive their intensity directly.
Maturity of the framework. Classic risks rest on a settled doctrine: Global Master Repo Agreement, Basel, netting opinions, Financial Stability Board haircut floors. The risk is known, written, litigated. Recent risks operate within a framework still under construction — or absent altogether: the Digital Operational Resilience Act only just in force, the mandatory clearing timetable still in flux, the legal doctrine around tokenisation unsettled. One does not discover the existence of a classic risk; one can discover a recent one, precisely as it materialises.
Nature of the risk. Classic risks are largely idiosyncratic and modellable: they are measured, provisioned, stress-tested as routine. The second family gathers risks that are often systemic and non-stationary; their distribution shifts with the geopolitical or monetary regime, which makes history a poor guide to what comes next.
Failure mode. Classic risks deteriorate gradually, which buys time to act: margin call, netting, orderly liquidation of collateral. Recent risks materialise in a discontinuous and binary way: collateral is either accessible or it is not, settlement either clears or it freezes. There is little or no slope to claw back on between the two states.
Hedgeability. This is the decisive criterion. The first family hedges or collateralises: one hedges rates, margins a counterparty, nets a default. The second is largely unhedgeable, and that is precisely its signature. One does not hedge or net a sanctions freeze, a Fedwire outage or ransomware on one’s CSD. If recent risks are structurally unhedgeable, then the entire risk management toolkit — pricing, modelling, hedging, margining, provisioning — is the wrong tool. By design, not by deficiency.
The second family is not “more of the same risk”. It is a break in hedgeability. The answer does not lie in pricing or modelling but in infrastructure redundancy, custodian diversification, and pre-positioned collateral. That is a different game entirely. A desk that meets second-family risks with first-family instruments — model, margin, hedge — has picked the wrong tool, and will only find out when the incident hits.
Three speeds of loss of control
These four criteria do more than classify. They predict how fast a risk escapes you. The less mature its framework, the more non-stationary it is, the more binary its failure, and the less hedgeable its nature, the faster and more irreversible the loss of control. That is precisely what the three tiers below measure — no longer the nature of the risk, but its operational intensity.
Tier 1 — existential, exogenous, T+0/T+1. Three signatures define them: total externality (the bank has no grip on the initial shock), immediate kinetics (the effect lands at T+0/T+1), and irreversibility (once the shock fires, options close). Neither the Liquidity Coverage Ratio, nor the Contingency Funding Plan, nor extra buffers will respond: the bank stays liquid on the balance sheet while paralysed in execution. This is precisely where repo turns into a systemic weapon.
Such risks are not managed, they are anticipated. Active mapping of infrastructure dependencies, real — not declared — diversification of channels, the ability to reallocate collateral before conditions deteriorate. Without a transverse mandate empowered to arbitrate, those decisions remain scattered, late, and unenforceable — in other words, useless at the moment that matters.
Three risks belong here. First, central clearing as a choke point: cleared repo is no longer an open market where price arbitrates access to funding, it is a closed club from which one can be excluded with no price signal at all. Second, the asymmetry of speed: in a constrained market, velocity stops being an execution parameter and becomes an instrument of dominance — whoever mobilises first shuts the door on the rest. Third, repo as indirect sanction: faster and often more effective than any formal legal mechanism, because it strikes before any recourse is available.
Tier 2 — orchestration failures and self-inflicted risks. These risks do not fire alone; they turn a manageable stress into a crisis whenever decision making is scattered and unranked. The bank then penalises itself through three channels. Internal desynchronisation, which tips stress into crisis. Technical reputation, which hardens into an informal gating criterion on liquidity access and marginally degrades every condition. And pre-emptive anticipation, where haircuts and eligibility criteria turn forward-looking and exclude market participants before any visible crisis. These are amplifiers, not detonators: absorbable in an orchestrated bank, existential in a siloed one. Three levers neutralise them: a unified orchestration of collateral, cash, and margin flows; a standardised set of behaviours under stress; and an anticipatory read of the buffer’s vulnerabilities. Without them, these risks stay invisible until they combine — which is to say, too late.
Tier 3 — structural risks. These trends trigger no crisis. They settle in over years and reshape the funding environment continuously. The effect is not a shock but an erosion: structural balance sheet cost drifts upwards, business models lose viability, relative competitiveness shifts quietly between banks. Predictable, yet systematically underestimated because diluted in time — no incident makes them visible before the gap is already entrenched. These are not managed, they are positioned: transverse treatment with a consolidated view, ongoing adjustment of buffer composition, anticipation of fragmentation zones, integration of every funding channel’s constraints into long-horizon trade-offs. This is the only category where time works in favour of the bank that moved early — and against the one waiting for a signal that will never come. Three markers compose it.
First, control over collateral eligibility: a quiet piece of rule-setting that redraws the boundary between financeable and non-financeable players. Second, fragmentation of liquidity pools, which ends a long-held assumption that a high-quality asset would be mobilisable in roughly uniform fashion across venues. Third, central bank repo, now an instrument of sovereignty: a tool of financial industrial policy that, through its operational parameters alone, organises the relative competitiveness of bank balance sheets.
Two further mechanisms escape this grid because they cut across all three tiers.
Model monoculture. Synchronisation of haircuts and margins through converging methodologies (CCPs, triparty, vendors) and automatic feedback loops. Models — VaR, stress add-ons, concentration add-ons, rating triggers, liquidity scores — tend to converge because they share the same inputs, the same vendors, and very similar logics. Under stress, this produces a synchronised cascade of margin calls and haircut revisions across infrastructures and counterparties: a liquidity shock that is sharper and faster, not because the fundamentals have changed, but because the models recalibrate all at once in the same direction. It is a risk distinct from eligibility. The asset may remain eligible, yet the haircut and margin function can jump on cliff effects driven by automatic recalibration — and that shock propagates mechanically through the system.
Indirect governance through liquidity. Repo — and more broadly secured funding access — disciplines markets. It also disciplines the institutions themselves, turning liquidity into a lever of indirect governance. A bank structurally dependent on a handful of CCPs, a few dominant dealers, and an increasingly conditional access to central bank liquidity gives up part of its strategic freedom. It no longer arbitrates its business model on return and risk alone, but on what remains financeable through the existing pipes. It becomes less able to push back on instructions, contest norms, or defend non-aligned choices without paying an immediate liquidity cost. The effect is concrete and quick: some activities become de facto undesirable because they consume too much high-quality collateral or worsen the funding profile; some clients become too expensive in liquidity terms even when they are profitable on P&L; some strategies become impossible to deploy without any formal prohibition being issued, simply because the infrastructure and its gatekeepers reprice or ration access to refinancing.
The asymmetry of error
Asymmetry first; the figure comes later. The first question is not what inaction costs, but which error one chooses. Treat a hedgeable risk as unhedgeable and the result is near-certain over-provisioning: the loss is bounded, visible, booked — a carry cost the balance sheet absorbs and a committee can arbitrate. Treat an unhedgeable risk as hedgeable and the cost is zero, until the incident. At that point, the loss is neither bounded nor provisioned: it represents the gap between an optionality that one believed one held and a frozen position one discovers the moment it stops unwinding.
The two errors are not symmetric. The first is paid continuously and can be steered. The second is paid only once, in full, with no warning. A system calibrated to minimise the visible cost optimises precisely the wrong error. Inaction is not a deferred expense; it is a short option sold at zero premium, with an unknown strike.
My desk had a short-term funding line, rolled every month for two years, at a spread that was genuinely unbeatable — from a single client. Every month, P&L and funding validated the choice: cheaper than any diversified alternative, and by a wide margin. The trader was, fairly enough, congratulated for the optimisation. In the desk committee, the question came up from time to time; the answer was always the same, and it was a good one — over two years, diversification would have cost a multiple of anything it could ever have protected. The roll fell on the last Thursday of the month; it had become a routine no one really looked at any more. The month the client needed his cash elsewhere, the line did not get more expensive: it vanished. It took us 20 minutes to realise it was not a delay. We had to refinance, in a hurry, a volume that had quietly swollen with time and habit. For two years, the optimisation had been booked as performance. It was a premium collected on a risk no one had ever named — and for good reason: the one who could have named it was precisely the one being congratulated.
The absence of transverse steering of executable liquidity does not produce only a single, identifiable shock. It also produces a continuous erosion, invisible in the usual performance indicators. These costs are neither a compliance failure nor a case of excess risk-taking: they come from trade-offs forced upon the bank — late, or over-cautious — imposed by operational uncertainty. The orders of magnitude that follow are not observed measurements. They are bounds modelled from published parameters, which any institution can rerun on its own data.
The only directly documented item is also the most telling: the extra cost of funding under stress. In September 2019, US repo went from two per cent to a peak around 9–10 per cent in a single session (BIS Quarterly Review, December 2019). The year-end reports of the International Capital Market Association’s European Repo and Collateral Council consistently document repo spikes of tens to hundreds of basis points on the European repo market at balance sheet dates. Assuming a differential of a few tens of bps between prepared mobilisation and forced mobilisation is therefore a prudent assumption — applied precisely to the volumes that cannot wait, at the moment they cannot wait.
The other items follow. Lacking fine-grained visibility on what is actually mobilisable, banks oversize their precautionary buffer. Two documented frictions feed that inertia: an asset encumbrance ratio of around 25 per cent on average across the EU (EBA) and an intraday settlement inefficiency of five to six per cent of volumes (ECB, TARGET Services 2024). Together they leave a fraction of the buffer inert at the very moment it is required.
For context, and outside the figures: that inertia is structurally worse in Europe than in the US, where cross-border collateral mobility costs around 65 per cent less in post-trade fees (AFME, October 2025). That is a market structure fragmentation point, not an input to the model. The immobilisation has a mechanical price: the trapped fraction multiplied by the cost of equity (around 10 per cent, EBA). The full calculation and its assumptions are set out in the box below.
One item is deliberately left unquantified. The preemptive or disorderly withdrawal from certain repo and securities financing transactions activities under stress erodes franchise and client relationships. No public data measures it, because it does not show up as a dated loss but as reduced capture of the recovery. Quantifying it precisely would be spurious precision; flagging it as a structural blind spot is the real thing. Excluding it from the figures makes them conservative — a lower bound.
1. Reference bank (assumed parameters)
2. Sourced parameters
3. Modelling assumptions (conservative)
4. Calculation - base case
Item A - buffer immobilisation
Inert buffer = €240bn × 8 per cent = €19.2bn
Annual cost = €19.2bn × 40 bps = €76.8m
ROE drag = €76.8m / €70bn ? 11 bps
Item B - extra funding cost under stress (annualised expectation)
Cost per episode = €50bn × 30 bps = €150m
Annual expectation = €150m × 0.5 = €75m
ROE drag = €75m / €70bn ? 11 bps
5. Result - recurrent structural cost
Reading: for a European G-SIB, the recurrent cost of not steering executable liquidity transversely sits, in the base case, at around €150m a year, or roughly 22 bps of ROE - borne every year, outside any crisis, and carried by no identified cost centre. What is robust is not the point estimate but the property: positive, recurrent, unattributed, and conservative (franchise erosion excluded).
Methodology note: the latest editions of EBA (encumbrance, cost of equity), ECB TARGET and AFME figures evolve over time. The balance sheet parameters of the reference bank are declared orders of magnitude, to be replaced by each institution’s own data for an institution-specific figure.
Taken on their own, each of these effects looks marginal. Taken together, they form a recurrent structural cost, borne year after year, regardless of any major crisis.
The cost no one carries
The price of inaction is not a loss. It is a rent. €152 million. 22 bps of ROE. Year after year, outside any crisis, for a bank that has neither taken too much risk nor breached compliance. It is the price of a massive blind spot, not of a fault. And a blind spot does not correct itself: it sits in no mandate, charges no cost centre, triggers no alert. It is simply paid — quietly and indefinitely.
The diagnosis ends there. What follows is no longer a question of measurement. If these risks are unhedgeable, then the answer is not a better model, a finer margin or a bigger buffer: those instruments treat the wrong error. The answer is organisational. It requires an organisation able to mobilise, transform, and move its liquidity under constraint, at market speed — not to measure it cold. Some banks have organised themselves accordingly. In most, collateral, cash, margins, and infrastructure dependencies are still managed in silos, optimised locally, and defended separately. The regime has changed. The organisation has not.
Those 22 bps are not a loss. They are a fee paid to inaction — recurrent, and entirely recoverable. The next articles in this series will show how an executable, cross-functional liquidity capability turns those silent, accepted costs into explicit and governed decisions — and why enterprise liquidity management will become as structural a function as asset-liability management or risk management. Not because it makes for a new org chart, but because €152 million a year tends to surface in the end.
NO FEE, NO RISK
100% ON RETURNS If you invest in only one securities finance news source this year, make sure it is your free subscription to Securities Finance Times
100% ON RETURNS If you invest in only one securities finance news source this year, make sure it is your free subscription to Securities Finance Times
.jpg)