Finance / WORLD

BIS Paper Presents a Liquidity Stress Tool for Nearly 13,000 Banks

A paper published on 11 November introduces a free tool that compares banks in more than 100 economies using annual balance sheet data and three shock scenarios. Its authors say it is neither an early warning system nor a supervisory test.

Reading sizeSources ↓Share story ↗
Conceptual illustration of financial data layers and information flows.Illustration
Symbolic illustration

The Bank for International Settlements published a working paper on 11 November 2026 introducing an interactive tool that compares banks’ ability to withstand liquidity shocks. The dataset covers nearly 13,000 commercial banks in more than 100 economies, with annual observations beginning in 2006. The Bank Liquidity Stress Test tool is hosted on the website of the ASEAN+3 Macroeconomic Research Office. The paper is by Christian Schmieder, Min Wei and Li Lian Ong. The authors describe the tool as a way to examine peer institutions and banking systems under common assumptions, not as a safety verdict on any individual bank.

The tool addresses gaps in existing public measures. Supervisory stress tests generally use confidential cash flow data and are rarely published. Public Liquidity Coverage Ratio disclosures are concentrated among large internationally active banks, begin mostly in the mid 2010s and provide a single regulatory ratio. The new platform applies scenarios to standardised balance sheet items and reports inflows, outflows and net liquidity at bank and economy level. AMRO’s page lists five interactive modules: coverage statistics, an economy level test, a bank level test, breakeven analysis and a user defined test. The platform also offers pre run results and bulk data downloads.

The framework compares liquidity that a bank could generate from its assets with potential funding outflows over a 30 day stress horizon. It uses three nested scenarios. Scenario 1 is described by the authors as an adverse event that is rare but manageable. Scenario 2 is at least as severe as the calibration used for the Basel III Liquidity Coverage Ratio. Scenario 3 is a highly severe assumption intended to resemble conditions faced by some of the hardest hit banks during the global financial crisis. Haircuts, encumbrance and runoff rates in Scenario 2 are twice their Scenario 1 levels, and those in Scenario 3 are four times higher. These are research assumptions, not an amendment to Basel rules or a new mandatory threshold for banks.

The paper’s central measure, the breakeven stress factor, estimates the minimum shock intensity at which a bank’s net liquidity reaches zero. In other words, it scales the rates in Scenario 1 by the same factor until projected inflows no longer cover outflows. A factor of 2.5 means that, under the model’s assumptions, the bank retains positive net liquidity through a combined shock 2.5 times as severe as Scenario 1. A value below one indicates a shortfall even under Scenario 1. The factor is not a forecast of realised losses. The authors designed it as a screening and comparison measure for resilience under defined shocks. In reality, stress types need not intensify together or in the same proportions, so the common scaling assumption simplifies how funding outflows and asset haircuts interact.

The database draws on annual bank financial statements in Moody’s Analytics BankFocus. It starts in 2006. Where consolidated figures are unavailable, the dataset uses unconsolidated statements. Some gaps are filled by linear interpolation, late current year reports can use the previous year’s value, and fiscal years are aligned to calendar years. A series is not extended when reporting has ceased for more than a year. Some bank years are removed if core indicators are missing or appear to be outliers. The authors report approximately USD 170 trillion in total assets for the banks covered in 2025, while noting that not all 2025 statements had been reported. They therefore use 2024 as the most complete reference year. The size of the sample does not mean that every bank has equally current or detailed data.

The paper finds that measured liquidity buffers have generally improved in most regions since Basel III was introduced. The authors report that the weakest tenth of banks in their sample could now withstand Scenario 2, the severe adverse case, while median banks in all regions were close to withstanding Scenario 3. This does not establish that every bank is prepared for a run or unlikely to fail. It is a result for a particular balance sheet dataset, filtering process and set of assumptions. Basic and granular data approaches also produce different results for some banks. The granular method applies category specific asset haircuts and funding runoff rates, while the basic method relies on broader totals. That difference makes it necessary to ask whether an apparent weakness reflects the bank’s position or the limits of public reporting detail.

Country rankings show some persistence even as liquidity profiles change. Between 2007 and 2024, 44 percent of economies in the weakest fifth remained in that group, while 40 percent in the strongest fifth retained their position. The Spearman correlation between country rankings was about 0.47. This does not mean that rankings remained fixed. The paper finds some improvement and convergence after the global financial crisis, alongside continuing structural differences among banking systems. For researchers and policymakers, a long time series is useful because it can help distinguish a system’s persistent position from a change in one year’s conditions.

The paper’s retrospective analysis of Silicon Valley Bank, Signature Bank, First Republic Bank and Credit Suisse shows why the tool should not be treated as an early warning system. Their stress factors rose in 2020 and 2021, appearing consistent with the ability to withstand the most severe scenario. The authors link this to pandemic era deposit inflows invested in government and agency mortgage backed securities. Higher interest rates later reduced the market value of longer duration assets, while deposit outflows accelerated. The measured liquidity profiles deteriorated in 2022, but annual balance sheet data did not clearly signal the impending failures. In the authors’ examples, deposits fell by 20 to 25 percent within days, beyond the Scenario 3 runoff assumptions. An annual structural measure cannot alert users in real time to a rapidly developing run.

The user defined module has narrower coverage than the full dataset. It includes data from 2019 onward and, for bank level calculations, a maximum of the 100 largest banks in each banking system by 2024 total assets. Not every bank included in the other modules appears in this feature. The paper also says the framework does not model day by day or intraday payment timing, deposit concentration, whether collateral is operationally ready for central bank use, or the effect on prices when several institutions sell assets at once. A sufficiently prolonged liquidity problem can become a solvency problem, but the tool focuses on liquidity. Supervisory cash flow data and scenarios that incorporate banks’ behaviour therefore remain necessary.

The platform should make it easier for analysts to repeat comparisons among economies and institutions, but its results should not be presented as a credit rating, formal supervisory decision or assurance to depositors. The data year, basic or granular approach, scenario, weighting and aggregation method can affect an output. The paper also states that its views do not necessarily reflect those of the BIS or its member central banks. The useful conclusion is narrower: the tool offers a public starting point for studying structural liquidity differences across a broad sample. Determining whether a particular bank can meet payments during a fast moving deposit run requires more current, detailed and supervisory cash flow information.

Data

Table

Coverage and method of the BIS liquidity stress tool

Sample, measures and limitations described in BIS Working Paper 1394 and the AMRO tool.

Commercial banks coveredNearly 13,000Annual observations from 2006
Economies coveredMore than 100Full tool sample
Covered bank assetsApproximately USD 170 trillion2025, statements incomplete
Most complete reference year2024At time of paper preparation
Stress horizon30 daysEnd of horizon liquidity measure
Scenario 1 severity multiplier1 times the baseline calibrationAdverse
Scenario 2 severity multiplier2 times the baseline calibrationSevere adverse
Scenario 3 severity multiplier4 times the baseline calibrationHighly severe adverse
Weakest tenth of banksWithstands Scenario 2 in the authors’ analysisLatest comparison in paper
Country ranking correlationApproximately 0.472007 compared with 2024
Weakest country quintile retained44 percent2007 to 2024
Strongest country quintile retained40 percent2007 to 2024
Deposit outflows at selected failed banks20 to 25 percent of deposits within daysSelected 2023 failures
User defined bank test sampleUp to 100 largest banks per economy2024 assets; data from 2019 onward
Methodology

Indicators were transcribed from BIS Working Paper 1394 and its appendices on 12 November 2026. Coverage refers to the research sample based on annual BankFocus statements. It should not be treated as every bank in each country, current supervisory data or a complete global census. The authors use linear interpolation for some missing years, carry forward prior year values for late current period reporting, align fiscal years to calendar years and remove outliers. The breakeven factor is computed by multiplying Scenario 1 haircuts, encumbrance and funding runoff rates by a common factor. Scenario multipliers are model inputs, not observed shocks. The paper identifies 2024 as the most complete reference year and notes that 2025 reporting is incomplete.

Updated:

Structured coverage

Subjects and places

Subject proposals

Taxonomy version: IPTC Media Topics 2026-Q2

Continue reading

Finance

Mexico's reserves fall as monetary base also contracts in the week

Banco de México's preliminary weekly account for Oct. 23 showed international reserves down $61 million to $257.016 billion. The monetary base fell 12.103 billion pesos to 3.509 trillion pesos, while its annual growth rate was 7.1%.

Banco de México3 min read Data included
Finance

ECB paper estimates regulation favours secured bank funding

An ECB Occasional Paper estimates materially higher regulatory costs for unsecured interbank lending than for repo transactions backed by high-quality collateral, which may influence how banks redistribute reserves.

European Central Bank1 min read Data included