Core Path · L6.6.07

Stressed outflow and drawdown models

14 minDeposit run-offDrawdownsConcentrationUninsured depositsDigital runsSVB

By the end you will be able to

  1. 1Estimate run-off rates from a bank's own deposit history (quantiles, the worst window, an episode) and say why calm history understates a run.
  2. 2Estimate stressed drawdowns of committed facilities from utilisation history.
  3. 3Measure depositor concentration and the uninsured share, and apply the Risk Engine's concentration rule to get a stressed run-off.
  4. 4Model the speed of a digital run and set the own-data rates beside the LCR's prescribed rates and the 2023 evidence.

Why it matters

On 9 March 2023 Silicon Valley Bank lost over $40bn of deposits, and it expected more than $100bn to go the next day: about 85% of its deposit base in two days (Federal Reserve review, p. 4). For retail and corporate deposits the LCR assumes 3% to 40% over 30 days. RiskAtoms Bank's own savings never fell more than 4.8% in a month. Which number should the bank plan for?