How Banks Optimise HQLA Liquidity Buffers: Level 1, 2A and 2B Assets, LCR Haircuts, Yield, Duration, Encumbrance and Monetisation

Quick answer: a high-quality liquid asset (HQLA) buffer is not simply a pile of the safest securities. A bank needs enough eligible, unencumbered and operationally usable assets to survive stressed net cash outflows, while also managing yield, market-value sensitivity, currency needs, collateral uses and concentration. Level 1 assets receive the most favourable regulatory treatment; Level 2A and Level 2B assets receive haircuts and are subject to composition limits. The optimisation problem is therefore: build a portfolio that remains monetisable under stress and satisfies the liquidity requirement without carrying more low-yield or rate-sensitive inventory than the bank needs.

A liquidity buffer is valuable only if the bank can actually turn it into usable cash in the market, currency, entity and time window where the stress occurs.

Ownership boundary: buffer construction is not the same job as liquidity stress testing

How Banks Stress-Test Liquidity asks how fast stressed outflows consume available resources. How Banks Optimise Collateral asks which asset should be pledged to which obligation.

This article owns a different job:

Before the stress begins, what mix of liquid assets should the bank hold as its reserve portfolio?

1. LCR turns a reserve portfolio into a ratio

The Basel Liquidity Coverage Ratio is designed to ensure that a bank has an adequate stock of unencumbered HQLA to survive a significant 30-calendar-day liquidity stress scenario.

At a high level:

LCR = stock of HQLA / total net cash outflows over 30 days.

The regulatory calculation contains detailed definitions, caps and runoff/inflow rules. The mathematical intuition is simple: the numerator is a reserve of highly liquid assets; the denominator is a stressed funding need.

See the Basel LCR framework.

2. Level 1, Level 2A and Level 2B are not just labels

HQLA are divided into categories reflecting liquidity quality and stress behaviour. The precise eligible instruments depend on the applicable jurisdiction, but the architecture is widely recognisable:

HQLA category	General role	US LCR amount treatment
Level 1	Highest-quality assets such as qualifying central-bank reserves and US Treasuries under US rules	100% of eligible fair value, subject to rule details
Level 2A	High-quality but somewhat less liquid eligible assets	85% of eligible fair value
Level 2B	Eligible assets with more market/liquidity risk	50% of eligible fair value

US Regulation WW sets the qualifying criteria in section 249.20 and the HQLA amount in section 249.21.

The category changes the effective liquidity value. Two securities worth S$100 million on a market screen can contribute very different amounts to the regulatory buffer.

3. Haircuts convert market value into regulatory liquidity value

Suppose the bank owns S$100 million each of three eligible assets under the US rule:

• Level 1 contribution: about S$100m;
• Level 2A contribution: about S$85m;
• Level 2B contribution: about S$50m;

before considering composition caps, secured-financing unwind adjustments and other detailed requirements.

The haircut does not mean the Level 2B security is expected to lose 50% in price. It is a regulatory liquidity adjustment that recognises weaker stress liquidity and price performance relative to Level 1 assets.

4. Composition caps stop the bank from building the entire buffer from weaker assets

Under the Basel structure, Level 2 assets are limited relative to the total HQLA stock, and Level 2B assets face an additional sub-limit. US rules implement this through the excess-HQLA calculation.

Federal Reserve supervisory material summarises the US treatment: Level 2A and Level 2B assets together generally cannot exceed 40% of the HQLA amount, and Level 2B assets generally cannot exceed 15%, after the relevant calculations. See the Bank Holding Company Supervision Manual, liquidity section.

This makes the optimisation non-trivial. A higher-yielding Level 2 asset may be attractive until the bank hits a regulatory composition cap and the next unit contributes little or nothing to usable HQLA.

5. Yield creates the opportunity-cost problem

Level 1 assets can be extremely liquid but may yield less than loans, corporate bonds or other assets the bank could hold. Carrying a large liquidity buffer therefore has an opportunity cost.

A simplified portfolio objective might be:

Maximise expected income − liquidity shortfall penalty − market-risk penalty − funding/collateral opportunity cost

subject to:

• LCR ≥ required level;
• internal liquidity-buffer minimums;
• Level 2/2B composition constraints;
• currency and legal-entity constraints;
• concentration limits;
• unencumbered-asset requirements;
• duration or interest-rate-risk limits.

This turns the buffer into a constrained portfolio-optimisation problem, not an instruction to “hold more Treasuries.”

6. Duration creates market-value risk inside the liquidity buffer

A long-duration government bond can be Level 1 HQLA and still fall sharply in market value when interest rates rise. Regulatory liquidity quality and interest-rate risk are different properties.

If the bank must sell or repo the security during stress, lower market value can reduce the cash it raises. If accounting treatment keeps unrealised losses outside current earnings, those economic losses can still become relevant when the asset is monetised.

This is one lesson from the 2023 US banking turmoil: highly rated securities can remain creditworthy while their market values move enough to matter for liquidity and capital if the bank must realise or collateralise them.

For the duration mathematics, see How Banks Measure Interest-Rate Risk.

7. Unencumbered means available—not merely owned

An asset can be legally owned by the bank yet unavailable because it has been pledged, segregated, trapped in another entity or otherwise encumbered.

Federal Reserve liquidity-buffer rules require qualifying assets to be unencumbered and operationally available. See section 252.35 — Liquidity Stress Testing and Buffer Requirements.

The optimiser therefore needs a real-time distinction:

Total holding ≠ unencumbered holding ≠ HQLA-eligible holding ≠ operationally monetisable holding.

8. Monetisation is the world-return test

An asset qualifies conceptually as liquid because it should be saleable or repo-able in meaningful size during stress without an excessive loss of value. A bank should therefore demonstrate that it has operational capacity to monetise its buffer rather than merely list securities in a report.

A monetisation test can ask:

• Can the security be sold or repoed through an active route?
• Does the treasury function control it?
• Can settlement occur quickly?
• Are legal documents and accounts ready?
• How much market depth exists for the required size?
• What haircut or bid-ask widening appears under stress?
• Is the asset in the currency needed?

This is the liquidity equivalent of a scientific falsifier: if a supposedly liquid asset cannot be converted into usable cash when tested, its place in the buffer deserves challenge.

9. Currency mismatch can make a strong global buffer weak locally

A bank may hold abundant HQLA in US dollars while facing a severe shortfall in another currency. FX markets can normally transform one currency into another, but stress can widen basis spreads, reduce liquidity or create settlement timing constraints.

The buffer model should therefore map assets and outflows by significant currency. The question is not only total HQLA; it is whether the right liquidity can reach the right legal entity and currency before the deadline.

See How Foreign-Exchange Settlement Algorithms Reduce Principal Risk.

10. HQLA and central-bank collateral are overlapping, not identical, sets

An asset can be eligible at a central bank but not count as HQLA under a particular liquidity rule, or vice versa depending on jurisdiction and facility. The central-bank haircut can also differ from the LCR haircut.

This matters because an asset can have multiple liquidity values:

• sale value;
• repo value in private markets;
• central-bank borrowing value;
• regulatory HQLA value;
• collateral value for derivatives or CCP margin.

The buffer optimiser should not confuse those uses or double-count the same asset across them.

11. Concentration can make a liquid portfolio collectively illiquid

A security can trade actively in normal times and still become difficult to sell in enormous size if many institutions liquidate the same position simultaneously.

Useful concentration dimensions include:

• issuer;
• asset class;
• maturity bucket;
• currency;
• market venue;
• repo counterparty;
• central-bank facility reliance.

A diversified buffer preserves more routes to liquidity when one market becomes impaired.

12. Buffer size and buffer composition are separate decisions

Liquidity stress testing determines how much usable liquidity the bank wants under its scenarios. Buffer optimisation determines how to hold that liquidity.

A bank can have:

• a large but poorly monetisable buffer;
• a small but high-quality buffer that is still insufficient in size;
• a compliant LCR buffer but inadequate internal currency or intraday liquidity;
• an expensive buffer that exceeds risk needs and depresses profitability unnecessarily.

“More HQLA” is therefore not a complete strategy.

13. Creative-work lens: Dunkirk and usable reserve capacity

Dunkirk is not a banking source, but it makes a useful resource principle visible: reserve capacity matters only if it can reach the bottleneck in time. A vessel sitting safely elsewhere is not the same thing as evacuation capacity at the beach. A security held somewhere in a banking group is not automatically usable liquidity at the entity and currency where stress occurs.

The creative work supplies the intuition. HQLA eligibility, encumbrance, legal transferability, market depth and settlement data provide the evidence.

14. The HQLA optimisation pipeline

1. Forecast stressed net cash outflows by relevant horizon and currency.
2. Set regulatory and internal liquidity-buffer requirements.
3. Inventory eligible Level 1, 2A and 2B assets.
4. Apply regulatory haircuts and composition caps.
5. Remove encumbered or operationally unavailable assets.
6. Estimate yield and funding carry.
7. Measure duration and market-value sensitivity.
8. Map sale, repo and central-bank monetisation routes.
9. Apply currency and legal-entity constraints.
10. Add concentration and market-depth penalties.
11. Optimise portfolio composition subject to LCR/internal constraints.
12. Run monetisation drills and stress market haircuts.
13. Reconcile buffer inventory with collateral usage.
14. Re-optimise as outflows, rates, prices and encumbrance change.

15. Failure modes

• Eligibility=liquidity assumption. Asset meets the rule but cannot be monetised operationally in size.
• Yield-only optimisation. Higher-return Level 2 assets crowd out resilient Level 1 liquidity.
• Duration blindness. Buffer market value falls sharply before it is needed.
• Encumbrance double count. The same security supports LCR, repo and margin calculations simultaneously.
• Currency aggregation. Global buffer is strong while one major currency is short.
• Composition-cap blindness. Extra Level 2 assets add little to recognised HQLA.
• Central-bank assumption. Facility eligibility is treated as guaranteed same-day usable funding without operational testing.
• Normal-market depth. Sale/repo assumptions ignore everyone else liquidating the same assets.

16. Diagnostics and falsifiers

• How much HQLA remains after haircuts and composition caps?
• How much of that stock is genuinely unencumbered?
• Which assets have never been monetised operationally?
• What happens to buffer value under a 100–200bp rate shock?
• Which currency has the weakest local buffer relative to stressed outflow?
• What share of liquidity depends on one repo market or one central-bank route?
• Does the bank’s optimiser preserve enough Level 1 capacity after yield optimisation?
• Which observation would show that a “liquid” asset is not liquid enough for the assumed horizon?

Suppose someone claims, “We hold S$10 billion of government bonds, so liquidity risk is covered.” A falsifier is evidence that much of the portfolio is encumbered, long-duration, trapped in another entity, denominated in the wrong currency or operationally unprepared for sale/repo. Gross holdings are not the same as usable stress liquidity.

17. Verification and update triggers

• reconcile HQLA classification with current rules;
• reconcile encumbrance across treasury and collateral systems;
• conduct periodic monetisation tests;
• stress duration and market-depth assumptions;
• review buffer currency composition after funding changes;
• re-optimise after major margin/collateral calls;
• track realised sale/repo haircuts against assumptions;
• update internal buffer targets when stress-test outflows change.

Connections across the finance-and-banking algorithms lane

• Liquidity stress testing — estimates how much usable liquidity the bank needs.
• Collateral optimisation — allocates specific assets once the buffer exists.
• Interest-rate risk — duration can change the market value of the liquidity reserve.
• Intraday liquidity — a shorter-horizon liquidity constraint that HQLA alone may not solve.

Research anchors

• Basel Framework — Liquidity Coverage Ratio.
• Federal Reserve — HQLA eligibility criteria.
• Federal Reserve — HQLA haircuts and composition calculations.
• Federal Reserve — Liquidity stress-testing and buffer requirements.
• Federal Reserve — Bank Holding Company Supervision Manual liquidity guidance.

The deeper lesson

HQLA optimisation is the mathematics of holding options before you know which one you will need. Level 1/2A/2B rules define regulatory liquidity value. Yield determines opportunity cost. Duration changes market value. Encumbrance removes freedom. Currency and entity boundaries determine where the reserve can be used. Monetisation testing returns the model to the world. A strong buffer therefore is not merely large—it is deployable.

Educational note: This article explains public banking and liquidity mathematics. It is not treasury advice, investment advice or a liquidity-buffer design for any specific institution.