5. Liquidity Management
Liquidity is a bank's ability to meet its payment obligations when they fall due, without incurring unacceptable losses. Banks fail from a lack of liquidity far faster than from a lack of capital: a solvent bank that cannot pay on time is closed within days. This chapter explains where liquidity risk comes from, how treasury forecasts and positions cash across time horizons from intraday to months, and how buffers, collateral and contingency plans keep the bank able to pay under stress. Most of these activities rely on systems that engineers build and run: cash position services, maturity ladder engines, intraday monitors and collateral inventories.
<a id="51-liquidity-risk-concepts"></a>
5.1 Liquidity Risk Concepts
This subchapter defines liquidity risk, identifies where it comes from, explains why real cash flows differ from contractual ones, and draws lessons from historical bank runs.
Funding Liquidity vs Market Liquidity
Why the distinction matters. "Liquidity" is used for two different things that are easy to confuse. A bank can hold billions of assets and still be unable to pay a depositor this afternoon, because those assets cannot be turned into cash fast enough or without a large loss. To manage liquidity risk, treasury must look at both sides: whether the bank can obtain cash, and whether its assets can be converted into cash.
What they are.
- Funding liquidity: the ability of the bank to meet its cash obligations (deposit withdrawals, maturing borrowings, payments, margin calls) as they fall due, either from cash it holds or by raising new cash. It is a property of the institution.
- Market liquidity: the ability to buy or sell an asset quickly, in size, without moving its price significantly. It is a property of the asset and its market. It is usually described along four dimensions:
- Tightness: the bid-ask spread, i.e. the cost of a round trip.
- Depth: the volume that can be traded at or near the current price.
- Immediacy: how fast a trade of a given size can be executed and settled.
- Resilience: how quickly prices recover after a large trade.
| Aspect | Funding liquidity | Market liquidity |
|---|---|---|
| Question answered | Can the bank pay what it owes, on time? | Can the asset be sold for cash quickly at a fair price? |
| Belongs to | The bank (institution) | The asset and its market |
| Main indicators | Cash balances, maturity gaps, rollover rates, survival horizon | Bid-ask spread, traded volume, price impact, haircuts |
| Typical failure | Depositors withdraw, lenders refuse to roll funding | Fire sales at deep discounts, assets cannot be repoed |
How they interact. The two reinforce each other in a stress, a mechanism known as the liquidity spiral. When a bank loses funding, it sells assets; forced selling pushes prices down; lower prices and higher volatility lead repo lenders to raise haircuts; higher haircuts mean less cash per unit of collateral, so the bank must sell even more.
Funding stress (deposits leave, lenders refuse to roll)
|
v
Bank sells assets / needs more collateral <-----------+
| |
v |
Prices fall, spreads widen (market liquidity drops) |
| |
v |
Haircuts and margin calls increase |
| |
v |
Less cash per unit of collateral ----------------------+
Analogy. Funding liquidity is whether you have money in your wallet when the bill arrives. Market liquidity is how quickly you could sell your car to pay that bill, and how much below its value you would have to accept for a sale by tonight.
Example: raising EUR 500m by tomorrow.
| Option | What the bank gives up | Cost or loss | Speed |
|---|---|---|---|
| Overnight repo of Bunds, 2% haircut | EUR 510.2m of Bunds pledged (500 / 0.98) | Repo interest only; Bunds stay on balance sheet | Same day |
| Outright sale of Bunds | EUR 500m of Bunds | Bid-ask cost, a few cents per 100 | Standard settlement unless traded for same day |
| Outright sale of BBB corporate bonds at 97 (fair value 100) | EUR 515.5m nominal (500 / 0.97) | About EUR 15.5m loss versus fair value | Days, buyers must be found |
| Unsecured interbank borrowing | Nothing pledged | Credit spread; may be unavailable in stress | Same day, if lenders are willing |
The same funding need costs very different amounts depending on the market liquidity of the assets used to meet it.
Key takeaways
- Funding liquidity is the bank's ability to pay; market liquidity is the ability to convert an asset into cash without a large price impact.
- Market liquidity is measured with spreads, depth, immediacy and resilience.
- The two interact: funding stress triggers asset sales and higher haircuts, which worsen funding stress (liquidity spiral).
- Repo usually monetizes assets faster and with less loss than an outright sale.
🧪 Practice
- Beginner: Classify each indicator as funding or market liquidity: bid-ask spread on a covered bond; percentage of maturing interbank borrowing rolled over; daily traded volume in a government bond; overnight deposit outflows.
- Intermediate: A bank must raise EUR 200m from a portfolio of corporate bonds. In normal markets it can sell at 99.5; in a stress at 95. Compute the nominal amount to sell and the loss in each case, and explain why the bank might prefer repo.
- Interview: "How would you estimate the market liquidity of a bond portfolio in a treasury system?" Hint: bid-ask spreads and volumes from market data, days-to-liquidate using a maximum share of average daily volume, haircuts as a proxy, issue size and age, and stressed versions of each measure.
Sources of Liquidity Risk
Why it matters. Liquidity risk does not come from one place. A bank can have perfectly stable retail deposits and still fail because of a derivative margin call or a frozen wholesale market. Treasury must identify every channel through which cash can leave, or fail to arrive, so that each one is measured, limited and covered in stress scenarios.
The main sources.
| Source | What happens | Typical trigger |
|---|---|---|
| Retail deposit withdrawal | Customers move savings and current account balances | Rumors, rating downgrade, better rates elsewhere |
| Wholesale funding rollover | Interbank, CP, CD or repo lenders do not renew at maturity | Counterparty concern, market-wide stress |
| Off-balance-sheet drawdowns | Customers draw on committed credit and liquidity lines | Market stress, customers losing their own funding |
| Collateral and margin calls | More collateral must be posted on derivatives, repo or CCP positions | Market moves, rating downgrade triggers |
| Market liquidity of assets | Buffer assets can only be sold at deep discounts | Fire sales, flight to quality |
| Currency mismatch | Bank has liquidity in one currency but needs another | FX swap market dislocation |
| Intraday | Payments cannot be made on time during the day | Delayed incoming payments, system outage |
| Trapped liquidity | Cash exists in the group but cannot be moved to where it is needed | Local regulation, ring-fencing, capital controls |
| Concentration | A few depositors, lenders or maturity dates dominate | Loss of one large counterparty |
Two further distinctions are used in every stress framework:
- Idiosyncratic vs market-wide: idiosyncratic stress affects only this bank (fraud, large loss, downgrade); market-wide stress affects everyone (2008, March 2020). The worst case is a combined scenario.
- Contractual vs non-contractual: some outflows are legal obligations; others are actions the bank takes to protect its reputation, such as buying back its own debt or supporting a sponsored fund.
Analogy. A household can run out of money because salary stops (funding), because an unexpected bill arrives (contingent), because a lender demands early repayment (rollover), or because the house cannot be sold quickly (market liquidity). A good household budget considers all of them.
Example: mapping risk sources to data feeds.
Risk source Data needed by the liquidity engine
-------------------------- ---------------------------------------------------
Deposit withdrawal Deposit balances by product, customer type, insurance
Wholesale rollover Borrowings with maturity dates and counterparties
Off-balance-sheet lines Undrawn commitments by facility type and customer
Collateral calls CSA terms, rating triggers, margin history
Asset market liquidity Securities inventory, eligibility, haircuts, prices
Currency mismatch All of the above per currency, FX swap capacity
Intraday Payment flows with timestamps, credit lines
Key takeaways
- Liquidity risk comes from deposits, wholesale funding, contingent lines, collateral calls, asset market liquidity, currencies, intraday payments, trapped cash and concentration.
- Stress scenarios distinguish idiosyncratic, market-wide and combined stress.
- Non-contractual, reputation-driven outflows must also be considered.
- Each source requires specific data, which makes liquidity management a data-integration problem.
🧪 Practice
- Beginner: For each event, name the source of liquidity risk: a corporate client draws EUR 300m on its revolving credit facility; a CCP increases initial margin; a 3-month CD is not renewed.
- Intermediate: A bank has ample EUR liquidity but large USD loans funded with 1-month FX swaps. Describe what happens if the FX swap market dislocates, and which metrics would reveal the risk in advance.
- Interview: "Which sources of liquidity risk are hardest to capture in data, and why?" Hint: non-contractual and reputational outflows, behavioral drawdowns, rating triggers buried in legal documents, intragroup transfer restrictions, and intraday timing.
Behavioral Cash Flows
Why it matters. Contracts describe what customers may do; behavior describes what they actually do. A current account is contractually repayable today, yet most balances stay for years. A 25-year mortgage is often repaid early when the borrower moves or refinances. If treasury used contractual dates only, it would overstate short-term outflows from deposits and misstate the timing of loan inflows. Behavioral models replace contractual dates with realistic estimates.
What they are.
- Contractual cash flow: the amount and date set by the contract (maturity, coupon, notice period).
- Behavioral cash flow: the amount and date expected given observed customer behavior, under a specified scenario.
- Main products that need behavioral treatment:
- Non-maturity deposits (current and savings accounts): split into a volatile part, treated as short-term, and a core part, spread over a longer profile using a decay curve.
- Mortgages and fixed-rate loans: prepayments, measured with the conditional prepayment rate (CPR), an annualized share of the outstanding balance repaid early.
- Credit lines and overdrafts: expected utilization changes.
- Term deposits: early withdrawal, where permitted, and rollover at maturity.
- Business-as-usual vs stress: behavior changes in a crisis. Deposits that are sticky in normal times leave faster under stress, and line drawdowns increase. Each scenario therefore has its own behavioral assumptions.
Behavioral assumptions are models: they must be documented, approved (typically by ALCO), validated and backtested (see Forecast Accuracy and Backtesting below). The same models also drive interest rate risk measurement in the banking book (Chapter 6).
Converting CPR to a monthly rate. Prepayment models usually work monthly, using the single monthly mortality (SMM):
SMM = 1 - (1 - CPR)^(1/12)
CPR = 8% -> SMM = 1 - 0.92^(1/12) = 0.69% of the balance repaid early each month
Analogy. A gym's members can all cancel at the end of the month (contractual), but the owner knows from experience that only about 3% actually do. The owner plans with the behavioral number, while keeping a reserve for a bad month.
Example: contractual vs behavioral profile of EUR 10bn of current accounts.
| Bucket | Contractual outflow | Behavioral outflow (BAU) |
|---|---|---|
| Overnight | 10,000m | 1,500m (volatile part) |
| 1 month - 1 year | 0 | 1,700m |
| 1 - 3 years | 0 | 2,800m |
| 3 - 5 years | 0 | 2,400m |
| Over 5 years | 0 | 1,600m |
| Total | 10,000m | 10,000m |
The total is the same; only the timing changes. Under a stress scenario, the volatile part would be larger and the core profile shorter.
Key takeaways
- Behavioral cash flows replace contractual dates with observed customer behavior.
- Non-maturity deposits are split into volatile and core parts; loans are adjusted for prepayments.
- Assumptions differ between business-as-usual and stress scenarios.
- Behavioral models need governance, validation and backtesting.
🧪 Practice
- Beginner: Compute the SMM for a CPR of 12%, and the expected early repayment in the first month on a EUR 2bn mortgage book.
- Intermediate: Explain why the same behavioral deposit profile should not be used for both a business-as-usual funding plan and a 30-day severe stress test.
- Interview: "How would you store and apply behavioral assumptions in a liquidity engine?" Hint: versioned model parameters by product and segment, effective dates, scenario as a dimension, separation of contractual cash flow generation from behavioral overlays, and reproducibility of past runs.
Deposit Stickiness and Run-Off
Why it matters. Deposits are the largest funding source for most banks, and they are formally repayable on demand. Whether a bank survives a stress depends on how much of that money stays. Stickiness describes how likely deposits are to remain; run-off rates quantify how much is assumed to leave in a given scenario and horizon.
What drives stickiness.
| Factor | Stickier | Less sticky |
|---|---|---|
| Deposit insurance | Fully insured (e.g. up to EUR 100,000 per depositor in the EU) | Uninsured balances above the limit |
| Relationship | Salary account, multiple products, long tenure | Single-product, rate-seeking customer |
| Purpose | Operational (payroll, payments, cash management) | Surplus cash looking for yield |
| Depositor type | Retail and small business | Financial institutions, large corporates |
| Concentration | Many small depositors | Few large depositors, connected communities |
| Channel | Branch-based relationship | Fully digital, money moved in seconds |
Regulatory run-off rates. The Liquidity Coverage Ratio (Chapter 9.2) assigns standard 30-day run-off rates to deposit categories. Selected Basel values:
| Category | 30-day run-off rate |
|---|---|
| Retail, stable (insured, established relationship) | 5% (3% in some jurisdictions with strong insurance) |
| Retail, less stable | 10% or higher |
| Operational deposits (clearing, custody, cash management) | 25% |
| Non-financial corporates, non-operational | 40% (20% if fully insured) |
| Financial institutions, non-operational | 100% |
Banks also calibrate internal run-off rates from their own data, which may be more conservative. The digital era has accelerated runs: what once took weeks of queues outside branches can now happen in hours through mobile apps, amplified by social media.
Analogy. A cinema with assigned seats and season ticket holders empties slowly when there is a small problem; a crowd standing near the exit leaves at the first rumor. Insured, operational, retail deposits are the season ticket holders.
Example: stressed 30-day outflows for a deposit book (EUR millions).
| Category | Balance | Run-off | Outflow |
|---|---|---|---|
| Retail stable | 20,000 | 5% | 1,000 |
| Retail less stable | 8,000 | 10% | 800 |
| Operational deposits | 6,000 | 25% | 1,500 |
| Non-financial corporates, non-operational | 5,000 | 40% | 2,000 |
| Financial institutions | 3,000 | 100% | 3,000 |
| Total | 42,000 | 8,300 (19.8%) |
Financial institutions are 7% of the balances but 36% of the stressed outflow.
Key takeaways
- Stickiness depends on insurance, relationship, purpose, depositor type, concentration and channel.
- Run-off rates convert deposit balances into stressed outflows over a horizon.
- Wholesale and uninsured deposits dominate stressed outflows even when they are a small share of balances.
- Digital channels and social media make modern runs much faster.
🧪 Practice
- Beginner: In the example, recompute the total outflow if EUR 2bn of financial institution deposits are replaced by EUR 2bn of retail stable deposits.
- Intermediate: A corporate client holds EUR 250,000 across three accounts at the same bank. Explain why insurance coverage must be determined per depositor rather than per account, and how this affects the classification.
- Interview: "Design a deposit classification engine for liquidity reporting." Hint: customer and counterparty type from reference data, aggregation across accounts per depositor for insurance limits, operational relationship flags, joint accounts, data quality fallbacks to the most conservative category, and an audit trail of classification decisions.
Contingent Liquidity Exposures
Why it matters. Some of the largest outflows in a crisis do not come from the balance sheet at all. They come from promises the bank has made (credit lines, guarantees) and from contract clauses that require more collateral when conditions worsen. These exposures are dormant in normal times and activate precisely when the bank is under pressure, which makes them dangerous.
What they are.
- Committed credit and liquidity facilities: customers can draw undrawn amounts at any time. In a market-wide stress, companies draw lines to hoard cash (seen in 2008 and March 2020). Liquidity facilities that back commercial paper programs are drawn in full when the CP market closes.
- Derivative collateral calls:
- Market value changes: when rates or FX move against the bank, variation margin must be posted under CSAs and to CCPs (Chapter 3.6).
- Rating downgrade triggers: many CSAs reduce thresholds or require additional collateral if the bank's rating falls. The LCR assumes a downgrade of up to three notches.
- Initial margin increases: CCPs raise margin when volatility rises.
- Guarantees and letters of credit: the bank pays if its client defaults on an obligation (Chapter 10.3).
- Securitization and conduit support: liquidity lines to vehicles the bank sponsors.
- Non-contractual obligations: buying back the bank's own debt to support market making, or supporting sponsored money market funds, to avoid reputational damage.
Analogy. Co-signing a friend's loan costs nothing today. The cost arrives only if the friend gets into trouble, often at the worst possible moment for you too.
Example: three-notch downgrade scenario (EUR millions).
CSA counterparty Current threshold Threshold after downgrade Their exposure Extra collateral due
(bank's rating) to the bank
Bank X 50 0 120 50
Bank Y 25 0 80 25
Insurer Z 100 10 160 90
----------------------------------------------------------------------------------------------
Total additional collateral call 165
(Collateral posted = exposure - threshold, so a lower threshold means more collateral.)
Plus: undrawn corporate liquidity facilities 2,000 x 30% assumed drawdown = 600
Total contingent outflow in scenario 765
Key takeaways
- Contingent exposures include committed lines, collateral calls, guarantees and non-contractual support.
- They activate in stress, often at the same time as other outflows.
- Rating triggers in CSAs create outflows linked directly to the bank's own creditworthiness.
- Measuring them requires legal terms (thresholds, triggers) to be captured as structured data.
🧪 Practice
- Beginner: Explain why a liquidity facility backing a client's commercial paper program is assumed to be drawn more heavily than a general credit line.
- Intermediate: Rates move by 100 bp and the bank's net derivative position loses EUR 400m of value against collateralized counterparties. What is the liquidity impact, and how does the timing of margin calls affect it?
- Interview: "How would you compute the collateral impact of a rating downgrade across hundreds of CSAs?" Hint: store thresholds and minimum transfer amounts per rating level, use current exposure per netting set, apply the post-downgrade terms, aggregate by currency and eligible collateral, and keep CSA data reconciled with legal documents.
Historical Bank Runs and Lessons
Why study them. Today's liquidity rules, metrics and stress scenarios are largely a response to past failures. Each run exposed a weakness that was not measured or not taken seriously. Knowing these cases explains why the systems engineers build ask for the data they do.
Selected cases.
| Case | Year | Funding weakness | What happened | Lesson |
|---|---|---|---|---|
| Continental Illinois (US) | 1984 | Heavy reliance on large wholesale and foreign deposits | Loan losses triggered rapid electronic withdrawals by wholesale depositors; rescued by the FDIC | Wholesale funding is flighty; runs need no queues |
| Northern Rock (UK) | 2007 | Retail deposits funded only around a quarter of the balance sheet; rest from wholesale markets and securitization | Securitization and interbank markets froze in August 2007; Bank of England support triggered visible retail queues; nationalized in 2008 | Funding concentration and market-wide stress can sink a solvent bank |
| Bear Stearns, Lehman Brothers (US) | 2008 | Large overnight repo funding, prime brokerage balances | Repo lenders withdrew or raised haircuts, clients moved assets, clearing banks demanded collateral | Secured funding is not stable when collateral or counterparty quality is questioned |
| Dexia (Belgium/France) | 2011 | Short-term funding of long-term public sector assets, hedged with swaps | Falling rates triggered large collateral calls on swaps; required state rescue | Derivative collateral calls are a major liquidity drain |
| Silicon Valley Bank (US) | 2023 | Concentrated, largely uninsured deposits from a connected tech community; large unrealized losses on long-dated securities | About USD 42bn withdrawn in one day, with more requested for the next; closed within 48 hours | Uninsured, concentrated deposits run at digital speed; collateral must be ready at the central bank |
| Credit Suisse (Switzerland) | 2022-2023 | Persistent loss of confidence after scandals and losses | Net asset outflows of roughly CHF 110bn in Q4 2022; acquired by UBS in March 2023 | Confidence erodes over months, then collapses quickly; large buffers may not be enough |
Recurring patterns.
Hidden weakness Trigger Run Outcome
(concentration, mismatch, (loss, rumor, downgrade, (deposits, repo, (rescue, merger,
uninsured deposits) market freeze) collateral calls) resolution)
| | | |
+---- months / years --------+------- hours / days -----+------- days ---------+
Regulatory responses.
- 2008-2010: Basel Committee Principles for Sound Liquidity Risk Management and Supervision, then Basel III with the LCR and NSFR (Chapter 9.2).
- 2013: BCBS 248 intraday liquidity monitoring tools (Section 5.3).
- After 2023: increased supervisory focus on uninsured deposit concentration, unrealized losses on securities, readiness to borrow from central bank facilities, and faster run-off assumptions.
Key takeaways
- Banks with acceptable solvency have failed because funding disappeared faster than assets could be monetized.
- Recurring weaknesses: funding concentration, maturity mismatch, uninsured deposits, collateral calls and loss of confidence.
- Runs now happen in hours, so monitoring and contingency actions must be near real time.
- Modern liquidity regulation is a direct response to these cases.
🧪 Practice
- Beginner: For Northern Rock and SVB, identify the main funding source that failed and the trigger.
- Intermediate: Which early warning indicators (Section 5.4, Contingency Funding Plans) might have flagged SVB's risk months before the run?
- Interview: "If a run on your bank started at 10:00 today, what data would treasury need within the first hour, and could our systems produce it?" Hint: real-time deposit outflows by segment, central bank and nostro balances, unencumbered collateral and its eligibility, prepositioned collateral, maturing funding, and dependencies on overnight batches.
<a id="52-cash-flow-forecasting-and-positioning"></a>
5.2 Cash Flow Forecasting and Positioning
This subchapter moves from concepts to measurement: how treasury knows today's cash position, projects future flows into maturity ladders, analyzes gaps against available liquidity, and checks whether its forecasts can be trusted.
Cash Position Management
Why it exists. Every payment, loan disbursement, securities settlement and maturing trade changes the balances the bank holds at the central bank and at correspondent banks (nostros). If a balance goes negative, the bank pays overdraft charges, fails payments or breaches regulatory requirements; if it is too high, cash sits idle instead of earning interest. Cash position management answers a simple but operationally hard question: for each currency and account, what will the end-of-day balance be, and what must be done before the market cut-offs?
What it involves.
- Scope: one position per currency, per settlement account (central bank account, each nostro), per value date (Chapter 1.4).
- Flow sources: payment systems, treasury management system (deals), securities settlement, loan systems, card schemes, corporate actions and coupons, fees and taxes.
- Flow states: each flow moves through statuses that determine how much it can be trusted:
FORECAST --> EXPECTED --> CONFIRMED --> SETTLED
(statistical (known deal (instruction (seen on statement
estimate) or payment) released or or RTGS notification)
received)
- Reconciliation: projected balances are reconciled with intraday and end-of-day statements from the central bank and correspondents (camt.052 and camt.053, or MT942 and MT940; Chapter 12).
- Actions: fund short accounts, sweep surpluses, move liquidity between currencies with FX swaps, and borrow or lend in the money market (Chapter 4.2).
Analogy. A cash position is like an air traffic controller's view of a runway: every incoming and outgoing flight (payment) has a scheduled time and status, and the controller must make sure the runway (account) never becomes overloaded or empty.
Example: morning cash position across accounts (millions, local currency).
| Item | EUR central bank | USD nostro (New York) | GBP nostro (London) |
|---|---|---|---|
| Opening balance | 2,000 | 120 | 45 |
| Confirmed outflows | -1,350 | -310 | -10 |
| Confirmed inflows | +900 | +95 | +60 |
| Expected outflows | -400 | 0 | 0 |
| Expected inflows | +250 | +40 | 0 |
| Projected end-of-day | 1,400 | -55 | 95 |
| Target balance | 1,500 | 20 | 15 |
| Action | Borrow 100 overnight | Fund 75 via EUR/USD FX swap | Sweep 80 to money market placement |
Key takeaways
- Cash positions are managed per currency, account and value date.
- Flows progress from forecast to settled; the position must show how much of it is certain.
- Projected balances must be reconciled with statements throughout the day.
- Treasury acts on the position before market and payment system cut-offs.
🧪 Practice
- Beginner: In the example, an expected USD inflow of 40 does not arrive. What is the new USD projected balance and the required funding?
- Intermediate: Why is it risky to treat FORECAST and CONFIRMED flows the same way when deciding how much to borrow? Propose how the position screen should display them.
- Interview: "Design a cash position service that updates in near real time." Hint: event-driven ingestion of flows keyed by currency, account and value date; status transitions with idempotent updates; handling amendments and cancellations; snapshots for end-of-day; reconciliation against statements.
Contractual Maturity Ladders
Why it exists. A cash position covers today. Treasury also needs to see, for every future period, how much cash the bank is contractually due to receive and to pay. The maturity ladder organizes all contractual cash flows into time buckets, making mismatches between when money comes in and when it goes out visible at a glance.
What it is.
- Time buckets: typically overnight, 2-7 days, 8-30 days, 1-3 months, 3-6 months, 6-12 months, 1-2 years, and longer. Short buckets are narrow because near-term flows matter most for survival.
- Content: principal and interest cash flows from all assets, liabilities and off-balance-sheet items, placed in the bucket of their contractual payment date.
- Per row: inflows, outflows, net flow (gap) per bucket, and cumulative gap.
- Treatment of products without fixed maturity: in a purely contractual ladder, demand deposits sit in the overnight bucket, and undrawn commitments are callable immediately. This is deliberately conservative; behavioral and stressed ladders adjust it (Section 5.1).
- Regulatory use: supervisors require contractual maturity ladders, for example the EU Additional Liquidity Monitoring Metrics (Chapter 9.2).
Analogy. A maturity ladder is a household calendar listing every salary payment, rent, loan instalment and bill for the next year, grouped by week and month, so you can see which months will be tight.
Example: contractual maturity ladder (EUR millions).
| Bucket | Inflows | Outflows | Net gap | Cumulative gap |
|---|---|---|---|---|
| Overnight | 1,200 | 3,500 | -2,300 | -2,300 |
| 2-7 days | 800 | 1,200 | -400 | -2,700 |
| 8-30 days | 1,500 | 2,100 | -600 | -3,300 |
| 1-3 months | 2,000 | 2,600 | -600 | -3,900 |
| 3-6 months | 2,500 | 2,200 | +300 | -3,600 |
| 6-12 months | 3,000 | 2,800 | +200 | -3,400 |
| Over 1 year | 30,000 | 9,000 | +21,000 | +17,600 |
The pattern is typical of a bank: short-term outflows (deposits, short-term borrowing) exceed short-term inflows, and the mismatch is recovered only in the long buckets where loans repay. This is maturity transformation (Chapter 1.1) made visible.
Key takeaways
- The maturity ladder places all contractual cash flows into time buckets.
- Net and cumulative gaps per bucket show where the bank depends on rolling or raising funding.
- A contractual ladder is conservative for deposits and ignores behavior; it is the starting point for behavioral and stressed views.
- Bucket boundaries, business day adjustments and data cut-offs must be applied consistently.
🧪 Practice
- Beginner: In the example, which bucket has the largest net outflow, and what is the cumulative gap at 3 months?
- Intermediate: Where would you place, in a contractual ladder, a callable bond issued by the bank, an amortizing loan, and a term deposit maturing on a Sunday? Justify each choice.
- Interview: "Build a maturity ladder engine for 20 million contracts." Hint: cash flow generation per product, business day adjustment, consistent snapshot date, partitioned aggregation by bucket, currency and entity, drill-down from bucket to contract, and reconciliation of totals with the general ledger.
Liquidity Gap Analysis
Why it exists. A negative cumulative gap does not by itself mean danger: the bank may hold liquid assets that can be sold or repoed to close it. Liquidity gap analysis combines cash flow gaps with the bank's counterbalancing capacity to answer whether, in each time horizon, the bank has enough liquidity to cover its net outflows, and by how much.
What it is.
- Liquidity gap: inflows minus outflows per bucket; cumulative gap sums them up to each horizon.
- Counterbalancing capacity (CBC): liquidity the bank can generate in the horizon: unencumbered liquid assets after haircuts, central bank eligible collateral, and other reliable sources. It is not a cash flow but a stock that can be monetized (Section 5.4).
- Net liquidity position: CBC plus cumulative gap. A negative value indicates that the bank cannot cover its outflows in that horizon.
- Static vs dynamic gap: a static gap considers only the existing book; a dynamic gap adds planned new business (new loans, deposit growth, refinancing).
- Scenarios: the analysis is run under business-as-usual and several stress scenarios, each with its own run-off rates, drawdowns and haircuts.
- Limits: treasury sets limits, for example a minimum net liquidity position per bucket, or a maximum negative cumulative gap as a percentage of total assets.
Example: combined stress scenario (EUR millions).
| Bucket | Inflows | Stressed outflows | Net gap | Cumulative gap | CBC after stressed haircuts | Net liquidity position |
|---|---|---|---|---|---|---|
| Overnight | 900 | 2,600 | -1,700 | -1,700 | 6,000 | 4,300 |
| 2-7 days | 600 | 1,900 | -1,300 | -3,000 | 6,000 | 3,000 |
| 8-30 days | 1,100 | 2,800 | -1,700 | -4,700 | 5,800 | 1,100 |
| 1-3 months | 1,500 | 2,600 | -1,100 | -5,800 | 5,500 | -300 |
| 3-6 months | 2,000 | 1,800 | +200 | -5,600 | 5,500 | -100 |
The bank covers stressed outflows for the first month, but the position turns negative in the 1-3 month bucket. CBC declines over time because stressed haircuts rise and some assets are used. Treasury would respond by lengthening funding, increasing the buffer, or reducing short-term wholesale reliance.
Analogy. Gap analysis is like checking your bills against your salary and savings: the gap is salary minus bills, and the savings you could withdraw are your counterbalancing capacity.
Key takeaways
- Gap analysis compares cumulative cash flow gaps with counterbalancing capacity per horizon.
- Net liquidity position = CBC + cumulative gap; negative values show where the bank cannot cover outflows.
- Analyses are run for business-as-usual and stress scenarios, statically and dynamically.
- CBC must use stressed haircuts and only unencumbered, operationally available assets.
🧪 Practice
- Beginner: In the example, by how much would CBC need to increase so that the net liquidity position is non-negative in every bucket?
- Intermediate: The bank replaces EUR 1bn of 1-week wholesale borrowing with 6-month borrowing. Recompute the cumulative gaps up to 3 months and explain the effect.
- Interview: "Why is it a mistake to count the same bond in the CBC and as an inflow in the ladder?" Hint: a bond can be monetized once: either sold or repoed before maturity, or held to receive its redemption; double counting is a classic data-modeling error, avoided by consistent treatment rules per asset.
Survival Horizon
Why it exists. Ratios and bucket tables can be hard to interpret for decision makers. The survival horizon condenses liquidity risk into one intuitive number: under a given stress scenario, how many days can the bank continue to meet its outflows using its own resources, without new market funding or extraordinary central bank support?
What it is.
- Definition: the number of days until the net liquidity position (CBC minus cumulative net stressed outflows) first becomes negative.
- Granularity: usually computed daily for the first weeks, because the crossing point must be precise.
- Scenarios: idiosyncratic, market-wide and combined. The reported metric is typically the shortest.
- Risk appetite: the board sets a minimum survival horizon, for example 60 or 90 days under combined stress. Breaching it triggers escalation and contingency actions.
- Relation to the LCR: the LCR (Chapter 9.2) tests a standard 30-day stress; the survival horizon is an internal, bank-specific measure that can extend further and use the bank's own assumptions.
Example: daily survival horizon under combined stress (EUR millions).
Day: 1 5 10 20 30 40 50 51
Cumulative net outflow 900 2,400 3,300 4,300 5,000 5,500 5,980 6,020
CBC 6,000 6,000 6,000 6,000 6,000 6,000 6,000 6,000
Net liquidity position 5,100 3,600 2,700 1,700 1,000 500 20 -20
EUR m
6,000 |-------------------------------------------------x---- CBC
| ****** x
4,000 | ******** *****
| **********
2,000 | ****
| *** * = cumulative net outflow
0 +---------+---------+---------+---------+---------+----> days
0 10 20 30 40 50
Survival horizon: 50 days
| Scenario | Survival horizon | Risk appetite minimum | Status |
|---|---|---|---|
| Idiosyncratic | 74 days | 60 days | Within appetite |
| Market-wide | 95 days | 60 days | Within appetite |
| Combined | 50 days | 60 days | Breach: escalate to ALCO |
Analogy. A ship's captain asks how many days the crew can last on current food and water if no port can be reached. The survival horizon asks the same question about cash.
Key takeaways
- The survival horizon is the number of days until stressed net outflows exceed available liquidity.
- It is computed per scenario, daily in the short term, and reported at its worst value.
- Boards set a minimum survival horizon as part of the risk appetite framework (Chapter 4.1).
- It complements the LCR with a bank-specific, longer-horizon view.
🧪 Practice
- Beginner: In the example, what is the survival horizon if CBC were EUR 5,000m instead of 6,000m?
- Intermediate: CBC is constant in the example. Explain why, in practice, CBC should decline over a stress and how that affects the survival horizon.
- Interview: "Treasury wants to see the survival horizon for 50 scenarios recomputed every hour. How would you design it?" Hint: precompute contractual cash flows once, express scenarios as parameter sets applied as vectorized overlays, cache CBC by asset with scenario-specific haircuts, and recompute incrementally when positions change.
Forecast Accuracy and Backtesting
Why it matters. Every liquidity number in this subchapter depends on forecasts and assumptions: tomorrow's customer payments, deposit run-off rates, prepayment speeds, line drawdowns. If they are systematically wrong, treasury borrows too much or too little, buffers are mis-sized, and stress results are meaningless. Measuring forecast accuracy and backtesting assumptions keeps the framework honest and is expected by supervisors as part of the ILAAP (Chapter 9.2).
What it involves.
- Short-term forecast accuracy: compare forecasted with actual daily flows, by account and flow category.
- Error: actual minus forecast.
- Bias (mean error): shows systematic over- or under-forecasting.
- Mean absolute error (MAE): average size of errors regardless of sign.
- Hit rate: share of days within a tolerance band.
- Timing vs amount errors: a flow that arrives one day late is a different problem from a wrong amount.
- Behavioral model backtesting: compare model assumptions with observed behavior over history, especially in the worst observed periods. A stress assumption should be at least as severe as the worst historical experience, unless justified.
- Root cause analysis: missing or late feeds, wrong value dates, flows booked directly in the ledger, customer behavior changes, calendar effects (month-end, holidays, tax dates).
- Feedback loop: findings lead to fixing data feeds, recalibrating models, or adding new forecast components.
Example 1: daily forecast vs actual by category (EUR millions).
| Category | Forecast | Actual | Error |
|---|---|---|---|
| Customer payments out | -620 | -705 | -85 |
| Card scheme settlement | -110 | -110 | 0 |
| Securities settlement | -250 | -180 | +70 |
| Loan disbursements | -90 | -140 | -50 |
| Maturing interbank placements | +300 | +300 | 0 |
| Net | -770 | -835 | -65 |
The net error is only -65, but the sum of absolute errors is 205: errors in different categories partly offset each other. Accuracy must therefore be measured per category, not only on the net.
Example 2: weekly metrics.
Day Mon Tue Wed Thu Fri
Error -65 +40 -120 +15 -310
Bias = (-65 + 40 - 120 + 15 - 310) / 5 = -88 (actual outflows systematically higher)
MAE = ( 65 + 40 + 120 + 15 + 310) / 5 = 110
Pattern: Friday outlier -> investigate weekly flows missing from the forecast
Example 3: behavioral backtest.
| Deposit segment | Model 30-day stress run-off | Worst observed 30-day outflow (10 years) | Assessment |
|---|---|---|---|
| Retail, less stable | 10% | 6.2% | Conservative |
| Corporate, non-operational | 40% | 31% | Conservative |
| Financial institutions | 60% (internal) | 72% | Not conservative: recalibrate |
Key takeaways
- Forecasts must be measured with error, bias, MAE and hit rate, per category and account.
- Net errors can hide large offsetting errors in individual categories.
- Behavioral and stress assumptions are backtested against the worst observed history.
- Root causes are often data problems (missing feeds, wrong value dates) rather than modeling problems.
🧪 Practice
- Beginner: Compute the bias and MAE for the errors +20, -50, +10, -30, -150, and describe what they suggest.
- Intermediate: Securities settlement shows +70 in Example 1. Give two plausible explanations, one about timing and one about data.
- Interview: "Design a forecast accuracy monitoring service." Hint: store each forecast version with timestamp (bitemporal), match flows to actuals by reference and category, compute metrics per dimension, alert on bias and outliers, and support drill-down to individual flows.
<a id="53-intraday-liquidity"></a>
5.3 Intraday Liquidity
This subchapter zooms into a single business day: how banks monitor and use liquidity minute by minute so that payments settle on time in real-time gross settlement systems.
Intraday Liquidity Monitoring
Why it exists. In an RTGS system (Chapter 2.2), each payment settles individually and only if the sender has enough funds or credit at that moment. A bank may end the day with a comfortable balance and still be unable to make a payment at 11:00 because the incoming payments it was counting on arrive at 15:00. Failing to pay on time can breach contractual deadlines, cause fails in securities and FX settlement, and, for large banks, delay payments for the whole system. Intraday liquidity monitoring makes sure the bank knows its available liquidity and its usage throughout the day.
What it involves.
- Sources of intraday liquidity: opening central bank balance, intraday credit from the central bank against collateral (see Intraday Credit Facilities below), balances and credit lines at correspondent banks, and incoming payments.
- Uses: outgoing payments, securities settlement, CLS pay-ins, margin payments to CCPs, ancillary system settlement.
- BCBS 248 monitoring tools (2013): the Basel Committee defined metrics that supervisors expect banks to monitor and report. The main ones:
- Daily maximum intraday liquidity usage: the largest negative net cumulative position (payments sent minus received) during the day.
- Available intraday liquidity at the start of the day.
- Total payments sent and received.
- Time-specific obligations (payments that must settle by a given time).
- For banks providing correspondent services: payments made on behalf of customers and intraday credit lines extended to them.
- Intraday throughput (see Payment Throughput and Queuing).
- Stress: usage is also assessed under scenarios such as own credit stress, a counterparty's failure to pay, or the loss of a credit line from a correspondent.
Example: one day on the EUR central bank account (EUR millions).
Opening balance 800, intraday credit line against collateral 1,200, available liquidity at start of day 2,000.
| Time | Sent | Received | Net | Cumulative net position | Balance |
|---|---|---|---|---|---|
| 08:00 | 300 | 100 | -200 | -200 | 600 |
| 09:00 | 650 | 200 | -450 | -650 | 150 |
| 10:00 | 500 | 350 | -150 | -800 | 0 |
| 11:00 | 400 | 700 | +300 | -500 | 300 |
| 12:00 | 900 | 300 | -600 | -1,100 | -300 (intraday credit used) |
| 14:00 | 300 | 800 | +500 | -600 | 200 |
| 16:00 | 200 | 650 | +450 | -150 | 650 |
| 17:00 | 100 | 350 | +250 | +100 | 900 |
Time Cumulative net position (each # = 100 of liquidity used)
08:00 -200 ##
09:00 -650 ######
10:00 -800 ########
11:00 -500 #####
12:00 -1,100 ########### <- daily maximum intraday liquidity usage
14:00 -600 ######
16:00 -150 #
17:00 +100 (net receiver for the day)
- Daily maximum intraday liquidity usage: 1,100, or 55% of the 2,000 available.
- Intraday credit used at peak: 300; headroom at peak: 900.
- The day ends with a positive balance, so no overnight borrowing is needed.
Analogy. Intraday liquidity is like a cashier's till: it is not enough to balance at closing time; there must be enough change in the drawer at every moment a customer pays.
Key takeaways
- In RTGS systems, liquidity must be available at the moment each payment settles, not only at end of day.
- BCBS 248 defines standard intraday metrics, including the daily maximum liquidity usage.
- Available intraday liquidity combines balances, collateralized credit and correspondent lines.
- Usage must be monitored in real time and analyzed under stress scenarios.
🧪 Practice
- Beginner: In the example, which hour causes the largest single-hour net outflow, and what is the balance after it?
- Intermediate: The 11:00 receipt of 700 arrives at 13:00 instead. Recompute the cumulative position and the peak use of intraday credit.
- Interview: "How would you compute the BCBS 248 metrics for 250 business days per currency and per account?" Hint: ordered settlement events with RTGS timestamps, running sums per account and day, minimum of the cumulative series, handling missing or out-of-order events, and reconciling totals with statements.
Payment Throughput and Queuing
Why it exists. Each bank would like to receive payments before sending its own, because incoming payments provide free liquidity. If every bank waits, nothing moves: the system gridlocks, and payments pile up late in the day, creating operational risk near the cut-off. Throughput guidelines and queuing mechanisms address this coordination problem.
What they are.
- Throughput: the cumulative share of a day's outgoing payment value that has settled by a given time. Some systems or supervisors set guidelines; UK CHAPS, for example, historically expected 50% by 12:00 and 75% by 14:30.
- Free-riding: delaying outgoing payments to use others' liquidity. Throughput guidelines limit it.
- Internal queuing: the bank's payment hub holds payments and releases them according to priority, available liquidity and counterparty limits.
- Central queuing in the RTGS: payments that cannot settle immediately wait in a queue managed by the system. T2, for instance, supports priorities (highly urgent, urgent, normal) and sender limits per counterparty.
- Liquidity-saving mechanisms (LSMs): algorithms in the RTGS that look for queued payments that offset each other bilaterally or multilaterally and settle them simultaneously, needing only the net amount.
Example 1: throughput against guidelines (total day's payments EUR 10bn).
Time Settled (cumulative) Throughput Guideline Status
10:00 2.5bn 25% - -
12:00 4.8bn 48% 50% Behind
14:30 7.9bn 79% 75% OK
17:00 10.0bn 100% - -
Example 2: gridlock resolved by multilateral offsetting.
Each bank has a balance of 0 and one queued payment of 100:
A ---100---> B
^ |
| 100
100 |
| v
+---- C <----+
Settling one at a time: impossible (each sender has 0).
LSM multilateral offsetting: net position of each bank = 0
-> all three payments settle simultaneously with zero liquidity.
Analogy. At a four-way stop where every driver waits for the others, nobody moves. A traffic officer (the LSM) who waves all cars through together solves the deadlock.
Key takeaways
- Throughput measures how early in the day payments are released.
- Free-riding by individual banks can gridlock the system; throughput guidelines discourage it.
- Banks queue internally by priority and liquidity; RTGS systems also queue centrally.
- Liquidity-saving mechanisms settle offsetting payments together, reducing liquidity needs.
🧪 Practice
- Beginner: A bank sends EUR 6bn per day. What value must have settled by 12:00 to meet a 50% guideline?
- Intermediate: In Example 2, A's payment to B is 150 instead of 100. With balances of 0, which payments can the LSM settle, and how much liquidity would A need to settle all three?
- Interview: "Design the release logic of an internal payment queue." Hint: priority classes, available liquidity and reserved buffers per account, bilateral counterparty limits, time-of-day rules for throughput, cut-off awareness, manual overrides with audit, and avoiding starvation of low-priority payments.
Intraday Credit Facilities
Why they exist. Even with careful timing, a bank's outgoing payments will at some moments exceed its balance. Rather than forcing banks to hold enough reserves to cover the intraday peak, central banks and correspondent banks provide intraday credit: funds that must be repaid by the end of the same day. This keeps payment systems flowing while limiting the cost of holding idle reserves.
What they are.
| Provider | Facility | Key features |
|---|---|---|
| Eurosystem | Intraday credit on T2 | Free of interest; fully collateralized with eligible assets; if not repaid by end of day, converted into overnight borrowing at the marginal lending rate |
| Federal Reserve | Daylight overdrafts on Fedwire | Collateralized overdrafts are free; uncollateralized overdrafts are charged a fee; capped by a net debit cap |
| Bank of England | Intraday liquidity via intraday repo | Against eligible collateral, repaid by end of day |
| Correspondent banks | Intraday credit lines on nostro accounts | Committed or uncommitted; priced; may be cut in stress |
Key considerations:
- Collateral: the main cost is the opportunity cost of collateral pledged. Assets pledged to the central bank but not used to obtain credit at the end of the day can generally still count as liquid assets, but the treatment must be checked against local rules.
- End-of-day risk: unrepaid intraday credit becomes overnight borrowing, often at a penalty rate, which is visible to the central bank.
- Correspondent dependency: uncommitted lines can be withdrawn without notice in a stress, exactly when they are needed most.
Example: intraday credit capacity in a collateral pool (EUR millions).
Collateral pledged to central bank pool (market value) 1,500
Central bank haircuts -120
------
Collateral value = maximum intraday credit 1,380
Already used for overnight/term central bank borrowing -180
------
Available intraday credit 1,200
Analogy. Intraday credit is like a shop's float lent each morning by the head office: the cashier can use it freely during the day, as long as it is back in the safe at closing.
Key takeaways
- Intraday credit bridges timing mismatches between outgoing and incoming payments.
- Central bank intraday credit is usually free but collateralized; unrepaid credit becomes overnight borrowing.
- Correspondent intraday lines are valuable but may be withdrawn in a stress.
- Collateral pledged for intraday purposes must be tracked to avoid double use.
🧪 Practice
- Beginner: In the example, how much intraday credit remains available if the bank pledges another EUR 200m of bonds with a 5% haircut?
- Intermediate: Compare the liquidity risk of relying on a central bank intraday facility versus an uncommitted correspondent line for USD payments.
- Interview: "The central bank collateral pool value drops by 10% at 11:00 due to a market move. What should systems detect and do?" Hint: real-time revaluation of pledged collateral, recomputed credit capacity, alerts when usage approaches the new limit, options to pledge additional collateral, and coordination with the payment queue.
Critical Payment Prioritization
Why it exists. Not all payments are equally urgent. Missing a CLS pay-in or a CCP margin deadline can have severe consequences: failed FX settlement, default procedures at the CCP, and reputational damage. A delayed low-value supplier payment is far less serious. When liquidity is tight, the bank must know which payments to release first, and its systems must enforce that order.
What it involves.
- Time-critical payments (BCBS 248 "time-specific obligations"): payments with a hard deadline, such as:
- CLS pay-ins (Chapter 2.2).
- Margin payments to CCPs.
- Settlement of ancillary systems (securities settlement, card schemes, retail clearing).
- Repayment of central bank operations and interbank borrowing.
- Customer payments with contractual deadlines (e.g. corporate tax or securities purchases).
- Priority classes: each payment type is mapped to a priority, which drives the release order in the internal queue and the priority flag in the RTGS.
- Liquidity reservation: part of the available liquidity can be reserved for critical payments so that normal payments cannot consume it.
- Stress mode: in a stress, non-critical payments are throttled while critical ones continue; the decision is governed by the contingency funding plan.
Example: priority scheme and queue order.
| Priority | Payment types | Deadline | Liquidity rule |
|---|---|---|---|
| 1 Critical | CLS pay-ins, CCP margin, central bank repayments | Hard, fixed times | Released immediately; reserved liquidity |
| 2 Urgent | Interbank deals, securities DvP cash legs, ancillary systems | Same-day cut-off | Released if available liquidity above reserve |
| 3 Normal | Customer payments | Scheme cut-off | Released subject to throughput targets |
| 4 Deferrable | Internal transfers, nostro sweeps | Flexible | Released last, may be delayed |
Available liquidity: 400 Reserved for priority 1: 250 Usable for others: 150
Queue (arrival order) Decision
P3 customer payment 120 -> release (150 -> 30)
P1 CCP margin call 200 -> release from reserve (250 -> 50)
P2 DvP cash leg 80 -> hold: only 30 free above reserve
P4 nostro sweep 50 -> hold
Analogy. A hospital emergency room triages patients by severity, not by arrival time. Payment prioritization is triage for money.
Key takeaways
- Time-critical payments have hard deadlines and severe consequences if missed.
- Payments are mapped to priority classes that drive internal release and RTGS priority.
- Reserving liquidity for critical payments prevents normal flows from consuming it.
- In stress, non-critical payments are throttled according to the contingency funding plan.
🧪 Practice
- Beginner: Assign a priority class from the table to: a margin call from a CCP; a salary payment file; a transfer between two of the bank's own nostros.
- Intermediate: In the queue example, an incoming payment of 100 arrives after the decisions shown. Which held payments should now be released, and in what order?
- Interview: "Design a payment scheduler that guarantees critical payments meet their deadlines." Hint: deadline-aware scheduling (e.g. earliest deadline first within priority), forecast of incoming liquidity, reservation of liquidity, simulation of the rest of the day, alerting when a deadline is at risk, and manual override with four-eyes approval.
Real-Time Balance Tracking
Why it exists. All intraday decisions (releasing payments, using credit, prioritizing) depend on knowing balances right now. End-of-day statements arrive too late, and the bank's own ledger may not yet reflect what the RTGS or a correspondent has settled. Real-time balance tracking maintains a continuously updated view of each settlement account, built from events, and reconciled with external sources.
What it involves.
- Event sources:
- Payment hub: payments released and acknowledged.
- RTGS and central bank: settlement notifications and intraday account reports (camt.054, camt.052).
- Correspondent banks: debit and credit confirmations (MT900, MT910, camt.054) and intraday statements (MT942, camt.052), with varying timeliness.
- Securities settlement and CLS: settlement confirmations.
- Shadow (mirror) account: the bank's own replica of each external account, updated from events, with separate figures for:
- Settled balance: confirmed by the account provider.
- Projected balance: settled plus released and expected flows.
- Reconciliation: the shadow balance is regularly compared with the provider's reported balance; differences are investigated as breaks.
- Data quality concerns: duplicate notifications, out-of-order events, missing confirmations from correspondents, and time zone differences.
Example: event stream updating a USD nostro shadow account (USD millions).
| Time (CET) | Event | Source | Settled balance | Projected balance |
|---|---|---|---|---|
| 07:00 | Opening balance | Correspondent statement | 120.0 | 120.0 |
| 07:05 | Expected inflow from FX swap near leg 40.0 | TMS | 120.0 | 160.0 |
| 09:12 | Outgoing payment 25.0 released | Payment hub | 120.0 | 135.0 |
| 09:13 | Debit confirmation 25.0 | Correspondent (camt.054) | 95.0 | 135.0 |
| 09:13 | Duplicate debit confirmation 25.0 | Correspondent | 95.0 | 135.0 (ignored, same reference) |
| 14:40 | Credit confirmation 40.0 (FX swap) | Correspondent | 135.0 | 135.0 |
| 15:00 | Intraday statement balance 135.0 | Correspondent (camt.052) | 135.0 | 135.0 (reconciled) |
Analogy. A real-time balance is like a live flight-tracking map built from many radar stations: each station reports positions with delays and occasional duplicates, and the system must combine them into one reliable picture.
Key takeaways
- Real-time balances are built from events from payment hubs, RTGS systems and correspondents.
- Shadow accounts distinguish settled from projected balances.
- Duplicates, ordering and missing confirmations must be handled explicitly.
- Continuous reconciliation with the account provider is essential.
🧪 Practice
- Beginner: In the example, why does the projected balance fall at 09:12 but the settled balance only at 09:13?
- Intermediate: The 14:40 credit confirmation never arrives, but the 15:00 statement shows 135.0. How should the shadow account treat the expected inflow, and what control should flag the gap?
- Interview: "Design a real-time liquidity dashboard for 40 currencies and 120 accounts." Hint: event sourcing with idempotent consumers keyed by message reference, per-account ordering, snapshots for fast recovery, latency monitoring per source, reconciliation jobs, and clear indication of data staleness on screen.
<a id="54-liquidity-buffers-and-collateral"></a>
5.4 Liquidity Buffers and Collateral
This subchapter covers the defenses a bank holds against liquidity stress: the buffer of liquid assets, the collateral rules that determine how much cash they can raise, the central bank as the ultimate backstop, and the plan for using all of these in a crisis.
High-Quality Liquid Assets (HQLA)
Why they exist. Forecasts and gap analysis show how much cash the bank might need in a stress; the bank must hold assets that can actually deliver that cash. Not every asset qualifies: in a crisis, many markets freeze and only the safest assets remain sellable or repoable. High-quality liquid assets are the assets that regulators consider reliable sources of liquidity in stress, and they form the numerator of the Liquidity Coverage Ratio (Chapter 9.2).
What they are.
- Characteristics of HQLA: low credit and market risk, ease and certainty of valuation, low correlation with risky assets, listed on a recognized exchange, active and large market, and a history of being sought in a flight to quality.
- Levels under the Basel LCR:
| Level | Typical assets | Haircut | Cap |
|---|---|---|---|
| Level 1 | Cash, central bank reserves (to the extent usable in stress), securities of sovereigns and central banks with 0% risk weight | 0% | None |
| Level 2A | Sovereign securities with 20% risk weight; covered bonds and corporate bonds rated AA- or better | 15% | Level 2 in total at most 40% of HQLA |
| Level 2B | Qualifying RMBS (25%); corporate bonds rated A+ to BBB- (50%); qualifying equities (50%) | 25-50% | At most 15% of HQLA |
Jurisdictions adapt the rules: the EU, for example, treats extremely high quality covered bonds as Level 1 with a 7% haircut.
- Operational requirements: HQLA must be unencumbered (not pledged), under the control of the function managing liquidity (usually treasury), and periodically monetized through test sales or repos to prove the bank can access the market. Assets that cannot be legally or operationally used in stress do not count.
- Cost: HQLA yield little, so holding them reduces net interest margin. Sizing the buffer is a trade-off between safety and profitability (Chapter 4.2, Investment Portfolio).
Example: computing the HQLA stock (EUR millions, simplified caps).
Asset Market value Level Haircut HQLA value
Cash and central bank reserves 4,000 1 0% 4,000
German and French government bonds 6,000 1 0% 6,000
Covered bonds, AA 3,000 2A 15% 2,550
Corporate bonds, BBB 2,000 2B 50% 1,000
-------
HQLA before caps 13,550
Cap checks (simplified):
Level 2 <= 40% of HQLA <=> L2 <= 2/3 x L1 3,550 <= 6,667 OK
Level 2B <= 15% of HQLA <=> L2B <= 15/85 x (L1+L2A) 1,000 <= 2,215 OK
HQLA stock 13,550
The full LCR formula also adjusts for short-term secured funding and collateral swaps that move assets between levels (Chapter 9.2).
Analogy. HQLA are like an emergency fund held in a current account and government savings bonds, rather than in a house or a stamp collection: they are worth less on paper but can be turned into cash immediately, even in a panic.
Key takeaways
- HQLA are assets that remain liquid in stress and can be monetized quickly with little loss.
- Level 1 assets count fully; Level 2A and 2B assets receive haircuts and are capped.
- HQLA must be unencumbered, controlled by treasury and periodically tested in the market.
- Holding HQLA has a cost in lower yield, so buffer size is a risk appetite decision.
🧪 Practice
- Beginner: Classify into HQLA levels: central bank reserves; a AA-rated corporate bond; a 0% risk-weight sovereign bond; a BBB corporate bond; an unlisted equity.
- Intermediate: In the example, the corporate BBB holding grows to EUR 6,000m market value. Apply the haircut and the simplified caps and compute the new HQLA stock.
- Interview: "How would you build an HQLA classification service for a securities inventory?" Hint: rules based on issuer type, risk weight, rating, asset class and listing from reference data; jurisdiction-specific rule sets; versioned rules with effective dates; explanation of each classification; and exception handling when data is missing.
Haircuts and Eligibility
Why they matter. When an asset is used to raise cash, the lender does not give its full market value. Prices can fall during the time needed to sell the collateral if the borrower defaults, and some assets are hard to value or sell. The lender protects itself with a haircut, and decides beforehand which assets it accepts at all (eligibility). Together, haircuts and eligibility determine how much cash a given portfolio can actually raise, in each venue.
What they are.
- Haircut: a percentage deduction from market value.
Collateral value = Market value x (1 - haircut)
Market value needed to borrow X = X / (1 - haircut)
- What drives haircuts: price volatility over the expected liquidation period, residual maturity, credit quality, market liquidity, currency mismatch between collateral and loan, and wrong-way risk (collateral whose value falls when the borrower is in trouble, e.g. own-group securities).
- Different haircut regimes: the same asset has different haircuts depending on the use:
- Regulatory (LCR) haircuts: for counting HQLA.
- Central bank haircuts: published schedules by asset category, maturity, coupon type and rating.
- Market repo haircuts: bilateral or tri-party, negotiated and changing with market conditions.
- CCP haircuts: for initial margin collateral.
- Eligibility criteria: asset type, issuer type and domicile, currency, minimum rating, listing, settlement location (CSD), and exclusions such as close links (central banks generally do not accept securities issued by the borrower itself or a closely linked entity). Concentration limits restrict how much of one issuer can be used.
Example: one bond, four venues (illustrative values).
Corporate bond, rated BBB+, nominal EUR 100m, price 98, market value EUR 98m.
| Venue | Eligible? | Haircut | Cash / value obtainable |
|---|---|---|---|
| LCR buffer (Level 2B) | Yes | 50% | 49.0 counts as HQLA |
| Central bank credit operations | Yes | 10% | 88.2 |
| Bilateral repo with a dealer | Yes | 8% | 90.2 |
| CCP initial margin | No (only government bonds and cash) | - | 0 |
To borrow EUR 50m in the bilateral repo, the bank must deliver 50 / 0.92 = EUR 54.3m market value of this bond.
Example: eligibility rule check.
rule CentralBankEligible(asset):
asset.type in {GOVT, SUPRA, COVERED, CORPORATE, ABS}
and asset.currency == "EUR"
and asset.rating >= "BBB-"
and asset.settlementCsd in ACCEPTED_CSDS
and not closeLink(asset.issuer, bank.group)
and asset.issuer not in EXCLUDED_ISSUERS
Analogy. A pawnshop lends you 60% of a gold watch's value and 20% of a used phone's value, and refuses items it cannot resell. Each pawnshop (venue) has its own list and its own percentages.
Key takeaways
- Haircuts protect the lender against price falls during liquidation; collateral value = market value x (1 - haircut).
- The same asset has different haircuts and eligibility in the LCR, at central banks, in repo and at CCPs.
- Eligibility rules depend on asset type, currency, rating, settlement location and close links.
- Liquidity systems must store haircuts and eligibility per venue and recompute them as data changes.
🧪 Practice
- Beginner: Compute the cash obtainable from EUR 40m market value of government bonds with a 2% haircut, and the market value needed to borrow EUR 40m.
- Intermediate: The bond in the example is downgraded to BB+. Which venues would it remain eligible for, and what happens to positions where it is already pledged?
- Interview: "Design a data model for collateral eligibility and haircut schedules across venues." Hint: venue-specific rule sets and haircut grids (asset class, maturity bucket, rating), effective dating, links to security master data, evaluation results cached per security and venue, and explainability for each decision.
Collateral Optimization
Why it exists. A treasury has many collateral obligations at once: repo borrowing, CCP margin, central bank credit, derivative CSAs, tri-party programs. It also has an inventory of assets with different eligibility, haircuts and value to the bank. If collateral is allocated carelessly (for example, delivering top-quality government bonds where lower-quality assets would be accepted), the bank uses up its best HQLA and funding capacity unnecessarily. Collateral optimization allocates assets to obligations so that all requirements are met at the lowest total cost.
What it involves.
- Inputs: asset inventory (available, unencumbered quantities); obligations (amount, venue, eligibility, haircuts, concentration limits, deadlines); cost of each asset.
- Cost of using an asset: opportunity cost such as loss of HQLA value, its value in the repo market (bonds in demand, "specials", are expensive to give away), funding cost, and operational cost of moving it between custodians or settlement systems.
- Approaches:
- Rules-based waterfall: for each obligation, allocate the cheapest eligible asset first. Simple and explainable.
- Mathematical optimization: formulated as a linear or mixed-integer program that minimizes total cost subject to eligibility, haircuts, quantities and limits. Better results across many obligations.
- Tri-party allocation: tri-party agents (Chapter 2.4) automatically allocate and substitute collateral from the bank's pool according to the lender's eligibility schedule.
- Substitutions: replacing an asset already delivered with another one, to free a needed asset.
Example: naive vs optimized allocation (EUR millions, haircuts ignored for simplicity).
Inventory Obligations and accepted collateral
A Government bonds 500 1 CCP initial margin 200 accepts: A
B Covered bonds 400 2 Tri-party repo 300 accepts: A, B
C Credit claims 600 3 Central bank credit 400 accepts: A, B, C
(loans pledged)
Naive (first obligation first, best asset first)
1 <- A 200 2 <- A 300 3 <- B 400
Remaining: A 0, B 0, C 600 -> unencumbered HQLA left: 0
Optimized (least valuable eligible asset first)
1 <- A 200 2 <- B 300 3 <- C 400
Remaining: A 300, B 100, C 200 -> unencumbered HQLA left: 400
Both allocations cover all obligations, but the optimized one preserves EUR 400m of HQLA for the liquidity buffer.
Analogy. When paying at a shop that accepts both vouchers and cash, you use the vouchers first, which are only good there, and keep the cash, which works anywhere.
Key takeaways
- Collateral optimization allocates assets to obligations at the lowest total cost.
- Cost includes lost HQLA value, repo value, funding cost and operational movement costs.
- Waterfalls are simple and explainable; optimization models handle many obligations better.
- Good allocation preserves the most liquid assets for the buffer.
🧪 Practice
- Beginner: In the example, a new CCP margin call of 250 arrives after the optimized allocation. Can it be covered, and with which assets?
- Intermediate: Redo the optimized allocation assuming haircuts of 0% for A, 10% for B and 30% for C. How much HQLA remains?
- Interview: "Formulate collateral allocation as an optimization problem." Hint: decision variables for quantity of asset i allocated to obligation j; constraints for eligibility, quantities, post-haircut coverage of each obligation and concentration limits; objective as sum of asset-obligation costs; discuss solver choice, run time and explainability for operations.
Encumbrance Monitoring
Why it exists. An asset that is pledged as collateral, or otherwise restricted, cannot be sold or repoed again. Such assets are encumbered. If treasury counts encumbered assets in its liquidity buffer, the buffer is overstated and will fail in a stress. High encumbrance also matters to unsecured creditors and depositors, because in a failure the encumbered assets go first to the secured creditors. Encumbrance monitoring tracks, for every asset, whether it is free to use.
What it involves.
- Sources of encumbrance: repo and securities lending, covered bond cover pools, securitizations, central bank borrowing, CCP and CSA margin, collateral pledged for payment systems, and assets reserved for legal requirements.
- Collateral received: securities received in reverse repo or as margin can sometimes be reused (rehypothecated); when reused, they are encumbered too.
- Pre-positioned collateral: assets placed in a central bank pool but not backing any borrowing are generally treated as unencumbered for reporting and buffer purposes, with local rules to be checked.
- Asset encumbrance ratio (as defined in EU reporting):
AE ratio = (Encumbered own assets + Encumbered collateral received)
/ (Total assets + Total collateral received available for encumbrance)
- Monitoring at position level: each holding (security and quantity, per custodian account) has a status such as free, pledged, lent, in cover pool, in transit. Status changes intraday as trades settle.
Example: encumbrance status and ratio (EUR billions).
| Item | Amount | Encumbered |
|---|---|---|
| Total assets | 100 | |
| Collateral received available for reuse | 10 | |
| Assets delivered in repo | 8 | |
| Mortgages in covered bond cover pool | 12 | |
| Central bank pool: 5 pre-positioned, 3 backing drawn credit | 3 | |
| Margin posted to CCPs and under CSAs | 2 | |
| Collateral received and reused | 4 | |
| Total encumbered | 29 |
AE ratio = 29 / (100 + 10) = 26.4%
Analogy. A house with a mortgage still belongs to its owner, but it cannot be used to secure another loan until the first one is repaid. Encumbrance monitoring is the land registry that records which houses are already mortgaged.
Key takeaways
- Encumbered assets are pledged or restricted and cannot be used to raise new liquidity.
- Encumbrance arises from repo, covered bonds, securitizations, margin and central bank borrowing.
- Only unencumbered assets may count in the liquidity buffer and counterbalancing capacity.
- Encumbrance must be tracked per position and in real time, including reused collateral received.
🧪 Practice
- Beginner: In the example, the bank draws a further EUR 2bn from the central bank using its pre-positioned pool. Recompute the AE ratio.
- Intermediate: Why does high encumbrance worry unsecured creditors and rating agencies, even when the bank is profitable?
- Interview: "Design a service that answers, at any moment, which securities are free to use." Hint: positions per security and safekeeping account from custody and settlement events, encumbrance links to the trades or pledges causing them, pending settlements, reconciliation with custodian and tri-party agent statements, and point-in-time queries.
Central Bank Facilities and Lender of Last Resort
Why they exist. Even a well-managed bank can face outflows larger than its own buffer, especially in market-wide stress when private lenders disappear. Central banks can create money without limit, so they act as the ultimate source of liquidity. Their facilities serve two purposes: routine monetary policy implementation and, in emergencies, the lender of last resort (LOLR) role that prevents liquidity problems of solvent banks from turning into failures.
What they are.
| Facility | Central bank | Tenor | Purpose and features |
|---|---|---|---|
| Main refinancing and longer-term operations (MRO, LTRO) | Eurosystem | 1 week; 3 months and longer | Regular open market operations against eligible collateral (Chapter 1.3) |
| Marginal lending facility | Eurosystem | Overnight | Standing facility at a penalty rate above the policy rate; the ceiling of the rate corridor |
| Discount window, primary credit | Federal Reserve | Overnight to 90 days | Lending to sound banks against pledged collateral |
| Standing Repo Facility | Federal Reserve | Overnight | Repo against Treasuries, agency debt and agency MBS |
| Operational standing facilities, Indexed Long-Term Repo | Bank of England | Overnight; 6 months | Bilateral and auction-based lending against eligible collateral |
| Emergency Liquidity Assistance (ELA) | National central banks in the euro area | Varies | Emergency lending to solvent banks with temporary liquidity problems, often against wider collateral |
| Temporary crisis programs (e.g. Bank Term Funding Program, 2023) | Federal Reserve | Up to 1 year | Loans against securities valued at par, introduced after the SVB failure |
Lender of last resort principles. Walter Bagehot's 19th-century rule still guides central banks: in a panic, lend freely, at a penalty rate, against good collateral, to solvent institutions. The penalty rate discourages routine use; good collateral protects the central bank; restricting lending to solvent banks avoids keeping failed institutions alive.
Stigma and readiness.
- Stigma: banks have historically avoided the discount window, fearing that borrowing signals weakness. This can delay use until it is too late.
- Operational readiness: central bank lending is only usable if the legal agreements are signed, collateral is pre-positioned and valued, operational procedures are tested, and cut-off times are known. SVB's difficulties in quickly moving collateral to the discount window in March 2023 led supervisors to stress regular testing.
Readiness chain for central bank borrowing
Legal docs -> Collateral -> Collateral -> Test -> Ability to borrow
signed identified and pre-positioned transactions within hours
eligible and valued performed
Analogy. The central bank is like a fire department: it does not prevent fires, but it must be reachable instantly. Pre-positioning collateral is like keeping the hydrant connected and tested before the fire starts.
Key takeaways
- Central banks provide regular operations, standing facilities and emergency lending.
- The LOLR role follows Bagehot's principles: lend freely, at a penalty rate, against good collateral, to solvent banks.
- Stigma can delay use of facilities; supervisors now encourage regular testing.
- Access depends on operational readiness: legal agreements, pre-positioned collateral and tested procedures.
🧪 Practice
- Beginner: Explain why the marginal lending facility rate is set above the main policy rate.
- Intermediate: A bank holds EUR 3bn of eligible loans (credit claims) not yet pre-positioned at the central bank. What steps and data are needed before they can support borrowing, and why does timing matter in a run?
- Interview: "Treasury asks for a daily report of how much the bank could borrow from each central bank within one hour, one day and one week. How would you build it?" Hint: inventory of eligible assets per central bank, pre-positioned versus not, haircuts, operational lead times per asset type, existing usage, and cut-off times.
Contingency Funding Plans
Why they exist. In a liquidity crisis, decisions must be made within hours, under uncertainty and pressure. A bank that only starts planning when the run begins will act too late. The contingency funding plan (CFP) is a documented, tested plan that defines how stress is detected, who decides, what actions are available, and how the bank communicates. Supervisors require it, and it is closely linked to the recovery plan (Chapter 7.3).
What it contains.
- Early warning indicators (EWIs): market-wide and bank-specific signals with thresholds, for example own CDS spread, share price, rating outlook, deposit outflows, wholesale rollover rates, margin calls, and the survival horizon.
- Stages and escalation: typically normal, heightened (alert) and crisis, with clear triggers for moving between them.
- Governance: a liquidity crisis committee, its members and deputies, decision rights, and reporting frequency to the board and regulators.
- Action menu: each funding action with its estimated capacity, time to execute, cost, side effects and stigma.
- Communication plan: messages for staff, customers, investors, rating agencies and supervisors.
- Testing: regular simulations, test sales and repos of buffer assets, and test borrowings at the central bank, so that the plan works in practice.
Example 1: early warning indicator dashboard.
| Indicator | Current | Amber threshold | Red threshold | Status |
|---|---|---|---|---|
| Own 5-year CDS spread | 145 bp | 120 bp | 200 bp | Amber |
| Share price, 1-month change | -18% | -15% | -30% | Amber |
| Retail deposit outflow, 5 days | 1.1% | 2% | 4% | Green |
| Wholesale funding rollover rate | 72% | below 80% | below 60% | Amber |
| Rating outlook | Negative | Negative outlook | Downgrade | Amber |
| Survival horizon, combined stress | 55 days | below 75 days | below 60 days | Red |
Escalation rule example: three or more amber indicators, or any red indicator, moves the bank to the heightened stage and convenes the liquidity crisis committee.
Example 2: action menu (EUR millions).
| Action | Capacity | Time to execute | Cost and side effects |
|---|---|---|---|
| Repo of Level 1 assets in the market | 4,000 | Same day | Low cost |
| Borrow from central bank against pre-positioned collateral | 3,000 | Same day | Low cost; possible stigma |
| Sell Level 2 securities | 1,500 | 2-3 days | Price losses in stress |
| Intragroup funding from parent | 1,000 | About 1 week | Requires regulatory approval if cross-border |
| Deposit campaign with higher rates | 800 | 2-4 weeks | Margin cost; may signal distress |
| Reduce new lending and undrawn commitments | 1,000 | 1-3 months | Franchise and client damage |
| Retained securitization of mortgages for central bank use | 2,000 | 2-3 months | Operational effort; cover pool and rating work |
Stage: NORMAL ------------> HEIGHTENED --------------> CRISIS
Trigger: EWIs amber/red outflows accelerate, markets closed
Actions: monitor daily committee, execute action menu in order of
EWIs pre-position collateral, speed and cost, inform supervisor,
test repos, raise term activate communication plan
funding early
Analogy. A CFP is like a building's fire evacuation plan: alarms (EWIs), floor wardens (governance), marked exits (actions), and regular drills (testing), all decided before there is any smoke.
Key takeaways
- The CFP defines detection, escalation, governance, actions and communication for liquidity stress.
- Early warning indicators with thresholds trigger movement between stages.
- Actions are documented with capacity, speed and cost, and ordered accordingly.
- A CFP is only credible if it is tested regularly, including real transactions.
🧪 Practice
- Beginner: Using the action menu, how much liquidity can the bank raise within one day?
- Intermediate: Explain why a deposit campaign with higher rates may be counterproductive at the crisis stage, and in which stage it is more appropriate.
- Interview: "Design an early warning indicator monitoring service." Hint: ingestion of market data and internal metrics at different frequencies, configurable thresholds with versioning, aggregation rules for escalation, alert routing and acknowledgement, history for backtesting thresholds, and resilience so that it keeps working during a crisis.
<a id="6-asset-liability-management-and-interest-rate-risk"></a>