A treasurer I once worked with had a simple test for any liquidity number on a slide: "Show me the cell it came from." He had learned this the hard way — a bank where the internal LCR model showed a comfortable buffer, and the supervisor's SREP review found a gap because the internal team had counted undrawn committed facilities on both sides of the ratio. The bank was not lying. It had simply built its LCR calculation from memory instead of from the delegated act. This article is about that gap — between a ratio that looks fine in a spreadsheet and one that survives a supervisor's questions. We will take one concrete balance sheet, build both Basel III liquidity ratios from scratch, stress them, and see exactly which line items move the number. Along the way we will cover what the Banking Regulation — Basel III/IV & Prudential Supervision course covers in depth, but from the practitioner's angle: what actually happens when you sit down with a balance sheet and a blank worksheet.
Why the two liquidity ratios are harder than the capital ratios
Capital ratios are about the asset side: you classify exposures, apply risk weights, sum. Liquidity ratios are about time. The LCR asks one question — can this bank survive 30 days of a specific, prescribed stress? The NSFR asks another — is the funding structure stable enough over a one-year horizon? Neither question has a single right answer in the way a risk weight does. The answer depends on how you classify deposits, which assets you call liquid, and how you treat the things that sit in between.
That is why the same bank can produce three different LCRs from three teams. The regulation is precise — the Basel Committee's LCR framework and the EU's delegated regulation on the liquidity coverage ratio spell out the haircuts and run-off rates in detail — but the application requires judgment at dozens of points. Judgment is where people go wrong.
A worked example: building the LCR from a balance sheet
Take a simplified bank. Total assets of 1,000. On the liability side: 400 of retail current accounts, 200 of retail term deposits maturing in under 30 days, 150 of corporate operational deposits, 100 of unsecured wholesale funding from financial institutions maturing in under 30 days, and 150 of long-term debt. On the asset side: 100 of cash at central bank, 200 of Level 1 government bonds, 100 of Level 2A covered bonds, 150 of corporate loans maturing within 30 days, and the rest longer-dated loans. More on the topic: “Quant Finance and Structured Products”.
Step one: HQLA and haircuts
High-Quality Liquid Assets are not "whatever you can sell." They are a defined set with defined haircuts. Level 1 — cash, central bank reserves, high-quality sovereign debt — gets a 0% haircut. Level 2A — certain covered bonds and agency securities — gets a 15% haircut. Level 2B — lower-rated corporate bonds, some equities — gets 25% to 50%. There is also a cap: Level 2 assets together cannot exceed 40% of the HQLA pool after haircuts, and Level 2B cannot exceed 15%.
In our example: 100 cash × 100% = 100. 200 Level 1 bonds = 200. 100 Level 2A covered bonds × 85% = 85. Total HQLA = 385. The cap does not bind here because Level 2 is only 85 of 385, well under 40%.
That number is the easy part. The hard part is the next line.
Step two: the 30-day net cash outflow
This is where most errors live. The stress outflow is not "how much cash might leave." It is a prescribed set of run-off factors applied to specific liability categories. Retail stable deposits run off at 5% under the standard; less stable retail at 10%. Operational deposits from corporates get 25% if they qualify as operational; non-operational wholesale funding from financial institutions gets 100%. Undrawn committed facilities get a drawdown factor — typically 10% for corporates, 30% for financial institutions, 40% for some retail credit lines. More on the topic: “Profession”.
In our example, applying standard factors:
- Retail current accounts (stable): 400 × 5% = 20
- Retail term deposits under 30 days: 200 × 10% = 20
- Corporate operational deposits: 150 × 25% = 37.5
- Unsecured FI funding under 30 days: 100 × 100% = 100
- Total outflows: 177.5
Inflows are capped at 75% of outflows. Say 30 of the 150 corporate loans mature within 30 days and qualify as inflows. Inflows = 30, which is under the 75% cap of 133. Net outflow = 177.5 − 30 = 147.5. LCR = 385 / 147.5 = 261%. Comfortable.
Now change one assumption. Suppose the corporate operational deposits do not actually qualify as operational — the bank cannot demonstrate the 30-day operational purpose, or the deposits are held by a financial institution rather than a corporate. The run-off jumps from 25% to 100%. Outflows rise to 177.5 + 112.5 = 290. Net outflow = 260. LCR = 385 / 260 = 148%. Still above 100%, but the buffer has halved. That is the kind of reclassification that turns a comfortable ratio into a board-level conversation.
NSFR explained: why short-term funding of long assets is punished
The Net Stable Funding Ratio is the structural twin of the LCR. It compares Available Stable Funding (ASF) — the equity and liabilities weighted by how stable they are over a one-year horizon — to Required Stable Funding (RSF) — the assets weighted by how illiquid they are.
ASF weights: regulatory capital and long-term debt over one year get 100%. Stable retail deposits get 95%. Less stable retail get 90%. Wholesale funding with maturity under six months gets 0%. That last one is the punishment. If you fund a 10-year mortgage book with three-month interbank borrowing, the mortgage gets an RSF weight near 65% to 85%, and the short-term funding contributes nothing to ASF. The ratio falls.
RSF weights: cash and central bank reserves get 0%. Level 1 HQLA gets 5%. Loans to non-financial corporates with residual maturity under one year get 50%; over one year, 85%. Mortgages get 65%. Undrawn committed facilities get 5% of the undrawn amount.
There is a cap on the ASF side that catches people out: certain liabilities, including some forms of funding from financial institutions and certain short-term wholesale funding, are capped or excluded. And on the RSF side, encumbered assets — assets already pledged as collateral — are generally treated as required stable funding even if they are HQLA, because you cannot use them twice.
The mistakes that turn a passing ratio into a finding
Three errors show up repeatedly when teams build these ratios for the first time.
Double-counting undrawn facilities. A committed credit line is a liability-side item (it can be drawn, so it creates an outflow) and simultaneously an asset-side item (it is an undrawn commitment, so it attracts an RSF charge). Teams sometimes count the same facility twice in the LCR outflow — once as a wholesale liability and once as a drawdown — or forget it entirely on the NSFR side. The result is a ratio that is wrong in both directions.
Ignoring currency mismatches. The LCR is calculated in each significant currency and in the aggregate. A bank with a large USD book funded in EUR can show a fine aggregate LCR and a failing USD LCR. The supervisor looks at both. If your spreadsheet has one currency column, you are not done.
Forgetting the inflow cap. Inflows are capped at 75% of outflows. A bank that expects large loan repayments within 30 days and counts them fully will overstate the numerator of the net outflow calculation. The cap is not a detail — it is the difference between a ratio that survives stress and one that does not.
Stress-testing liquidity: what the standard ratios do not show
The LCR and NSFR are point-in-time snapshots under a prescribed stress. They do not tell you what happens if the stress is worse than the prescription, or if it lasts longer than 30 days. That is what stress testing liquidity is for — and it is where the internal liquidity adequacy assessment process (ILAAP) earns its keep.
A practical approach: take the LCR spreadsheet and run three scenarios.
- Idiosyncratic run. Double the retail run-off factors and move all wholesale funding to 100%..
- Market-wide stress. Apply haircuts to HQLA beyond the regulatory ones — a 10% haircut on Level 1, 25% on Level 2A — and assume no inflows at all.
- Combined and prolonged. Extend the horizon to 60 and 90 days, and assume no new wholesale funding can be raised.
The output is not a single number. It is a curve: how the ratio decays as the stress deepens and lengthens. That curve is what the supervisor wants to see in the ILAAP, and it is what tells the ALCO committee where the real cliff edge is. For a deeper look at how stress testing fits into the broader risk framework, the FRM — Financial Risk Manager (Part I + Part II): What VaR, Expected Shortfall and Stress Testing Basel Actually Hide walks through the model assumptions behind these numbers. More on the topic: “FRM”.
Who this work suits — and who it does not
Building LCR and NSFR in a spreadsheet is not glamorous work. It is classification, reconciliation, and arguing with the treasury team about whether a deposit is operational. If you enjoy that — if you like the moment when a line item moves and the ratio shifts by 40 points — this field rewards you. If you want to be in the markets, this is not that job.
The people who do it well tend to come from three backgrounds: regulatory reporting, treasury, or risk analytics. What they share is comfort with the fact that the regulation is precise but the application is judgment. There is no answer key. There is only the delegated act, the supervisor's template, and the number that survives both.
Learning paths and their trade-offs
You can learn this from the primary sources. The Basel Committee's LCR framework and the EU delegated regulation on the liquidity coverage ratio are public, detailed, and free. Reading them carefully and building the spreadsheet yourself is the most rigorous path — and the slowest. It takes weeks, and you will not know whether your classification decisions are right until someone checks them.
A second path is on-the-job learning inside a bank. You get feedback, but you also inherit the team's existing errors. If the team has been double-counting facilities for three years, you will learn to double-count.
A third path is a structured course. The Banking Regulation — Basel III/IV & Prudential Supervision course on asibiont walks through exactly this calculation with examples — HQLA haircuts, run-off factors, NSFR weights, and the caps — and lets you check your spreadsheet against a worked answer. That is the barrier self-study cannot close on its own: you can read the delegated act, but you cannot easily find out whether your reading is the one the supervisor will accept.
The bottom line
The LCR and NSFR are not mysterious. They are arithmetic with judgment layered on top. If you can classify a balance sheet, apply the prescribed weights, and respect the caps, you can compute both ratios in a spreadsheet in an afternoon. The skill is not in the formula — it is in knowing which line items move the ratio, and in stress-testing the result before the supervisor's numbers arrive. Start with one balance sheet, build the LCR cell by cell, then change one assumption and watch what happens. That is how the mechanism becomes intuitive. The course walks through these calculations with examples and gives feedback on your decisions — details and enrolment are on the Banking Regulation — Basel III/IV & Prudential Supervision page.