Liquidity Stress Testing Under LCR & NSFR

Since 2015 every bank has had to run a Liquidity Coverage Ratio (LCR) and, since 2018, a Net Stable Funding Ratio (NSFR). Both are calibrated to a specific supervisory stress and both are, on their own, extremely poor liquidity risk-management tools. They are compliance metrics. Real liquidity management sits in the internal stress testing that supports the ILAAP.

What LCR and NSFR actually measure

LCR asks a narrow question: does the bank hold enough High Quality Liquid Assets (HQLA) to survive a 30-day stress that combines a wholesale-funding freeze, a partial deposit outflow, and drawdowns on committed facilities, all with prescribed run-off rates? Above 100% and you pass. That's it.

NSFR asks a structural question: over a one-year horizon, does the amount of stable funding cover the assets that need stable funding? It penalises reliance on short wholesale funding for long-dated assets. It is a duration-of-funding rule dressed up as a ratio.

Both are important. Neither tells a treasurer anything they wouldn't already know from a decent ALM report.

Where internal stress testing has to do more

An LCR-only framework has three well-documented weaknesses:

• A single 30-day horizon. Real liquidity crises don't stop on day 30. Term-funding markets can be closed for months.
• Prescribed run-off rates. The Basel run-offs are meant to be conservative for a typical bank. They are not calibrated to your deposit base, your wholesale profile, or your geopolitical exposure.
• No management actions. LCR is a static point-in-time metric. It does not model what the bank will actually do as liquidity tightens.

Internal liquidity stress testing (ILST) — the analytical core of ILAAP — has to cover the ground LCR does not.

Designing scenarios that are plausible and severe

The classic mistake in liquidity stress design is scenarios that are severe but not plausible (a 90% overnight deposit outflow with no macro trigger) or plausible but not severe (a shallow recession with a 2% deposit outflow). Neither drives useful action.

A useful scenario has three properties:

1. A narrative trigger. Idiosyncratic (name-specific credit event, operational loss, cyber), market-wide (funding-market freeze, sovereign stress), or combined. Every ratio move should be traceable back to a plausible cause.
2. Differentiated assumptions. Retail deposit outflows should differ by segment (insured vs uninsured, digital vs branch, business vs personal). Wholesale roll-over rates should differ by tenor, counterparty type, and secured vs unsecured. Facility drawdowns should differ by borrower rating and utilisation.
3. A defined horizon and re-planning point. Typically 30 days survival, 90 days with contingency actions, and 12 months to full re-planning. A scenario that ends on day 30 with a shrug is not a scenario, it is an LCR calculation.

Behavioural assumptions are where the model lives

Deposit stickiness under stress is the single biggest driver of internal liquidity results — and the least defensible one, because banks by design almost never observe their own genuine deposit runs. Every behavioural assumption should be:

• Benchmarked against external references (2008, 2020, 2023 US regional bank episodes).
• Segmented (insured/uninsured, corporate/retail, digital/branch).
• Sensitivity-tested — run the ILST at a plausible base and at a plausible-severe run-off assumption, and report the range, not just the point.

Pre-committed management actions matter more than the number

An ILST result of "we survive to day 45" tells nobody what to do. The ILAAP has to document, in advance, what actions the bank will take, in what order, at what triggers:

• Repo capacity: what is pre-positioned, at what haircut, with which central bank.
• Deposit pricing: at what point do we widen retail rates, and by how much.
• Committed lines: what has been genuinely tested that we can draw.
• Asset sales: what is the actual liquidity of the securities portfolio net of the market conditions in the scenario itself.
• Business restrictions: at what point do we stop new lending, stop draw-downs on existing committed facilities where legally possible.

None of these actions add HQLA on the day of the crisis. All of them extend the survival horizon. The pre-agreement is the point — the moment a bank tries to negotiate them under stress, they don't happen fast enough to matter.

What a defensible liquidity framework looks like

• LCR and NSFR calculated cleanly, monitored daily, with buffers over the 100% floor that management can defend.
• Three to five internal stress scenarios spanning idiosyncratic, market-wide, and combined, each traceable to a plausible trigger.
• Behavioural assumptions segmented, benchmarked, and sensitivity-tested.
• A Contingency Funding Plan with pre-committed actions, tested drawdown capacity, and named owners for each action.
• An ILAAP document that ties all of the above together and refreshes at least annually, plus after any material market event.

Liquidity risk is the risk most likely to kill a bank in a week rather than a decade. Treating LCR and NSFR as anything more than the entry ticket has been, historically, the single most common cause of that week.