Your bank: the contingency drill and the likelier cost Scenario estimate
1. Can it pay on the day? Hour by hour
Pledged collateral counts only if it is already lodged with the FHLB or the Federal Reserve and the borrowing lines have been tested; the filings show neither.
2. What it costs over a year: return on equity
Return on equity at June 30, 2026, what each part of the scenario takes away, and what is left after a year.
3. Peers at the same setting
Return on equity at June 30, 2026 (grey ring) and after a year at this setting (navy dot), for the bank and its seven named peers (the Know Your Peers group). Medians on the cards are of its wider Statum peer group.
4. Every hike and run size
Return on equity after, at the beta set above. Each square is one hike and one run size (on a phone the grid is turned, run sizes across); the ringed square is the current setting. Red squares are below zero; every square carries its value.
How the simulator calculates
- It is Scenario C of Deposit Runoff Scenarios with one beta: the rate hike reprices every interest-bearing deposit at the beta set here, and the agent run takes the set share of deposits within the hours set here. "Borrows it back" refills the run with wholesale money at the short rate plus the hike plus the extra cost set in Assumptions; "Sells securities" spends cash, then sells securities at the sale discount set in Assumptions.
- Separately, noninterest-bearing deposits are paid the rate set in "Rate paid on noninterest-bearing deposits"; the beta does not apply to them, and nothing leaves because of it.
- "A year's pre-tax income" is pre-tax income for the first half of 2026, doubled. Return on equity and on assets start from the published UBPR figures. Each change is a year's change in net income, after a 21% tax (none for an S corporation), over total equity or total assets at June 30.
- The hour-by-hour chart spreads the run evenly over the hours set. Cash is cash and federal funds sold above the minimum the bank keeps; securities free to sell are those not pledged; pledged collateral counts only if the bank is ready to borrow against it that day.
- Market value, for banks whose holding company is listed: the latest closing share price times the shares outstanding the holding company last reported to the SEC. The 2023 share-price moves are closing prices; they show how the market treated banks that ran, and are not a forecast of any bank's share price.
Before the clock moves again: five steps for the board
- Measure how much of a one-day outflow cash on hand covers at 5%, 10% and 25% of deposits, compare it with peers, and find your bank's ring of the blast radius.
- Make the borrowing real: confirm the Federal Home Loan Bank and discount window arrangements, lodge the collateral now, and run a small test transaction, as the July 2023 guidance expects.
- Know which balances could move together: the largest uninsured accounts, operating accounts tied to a few commercial relationships, and deposits gathered through a single channel.
- Write the agent scenario into the contingency funding plan, with an outflow speed measured in hours, and take it to ALCO and the board.
- Decide which deposits you would pay up to keep, and at what rate, so repricing is a choice rather than a scramble.
A yardstick bankers know: an agentic run against rate hikes Scenario estimate
Pre-tax income lost in a year
The lines are the run, from 0% to 10% of deposits in steps of 0.5%. The yellow dashed lines are rate hikes in steps of 25 bp. Where a line crosses a dashed line, the run costs as much as that hike.
Money to borrow the same day
A rate hike of any size needs none: its cost arrives over months as deposits reprice. A run needs it before the day is out, for whatever cash on hand cannot cover.
The run, step by step
The hikes, step by step
How the two are compared
- Each is computed on its own, through the same model as Deposit Runoff Scenarios. The run moves no rates: it is Scenario C with no rate rise. The hike moves no deposits: it raises the rate paid on interest-bearing deposits by the bank's deposit beta, with nothing leaving.
- The run uses the assumptions in the Assumptions panel: what wholesale money costs, cash kept on hand, the sale discount. Widening the wholesale spread makes the run cost more.
- "Costs as much as" compares a year's pre-tax income lost. Hikes are linear: a 50 bp hike costs twice a 25 bp hike.
- All banks: the totals add every bank's dollars; the percentage is of the industry's annualized pre-tax income.
Across the industry: a run at every bank, one bank at a time Scenario estimate
Every bank, hit by the same agentic run on the same day, with no rate change. The cost in earnings assumes each bank replaces the money at the short rate plus the wholesale spread in Assumptions; the realized loss assumes it sells securities instead, at the sale discount in Assumptions.
Banks that would borrow the same day, by run size (40% is beyond any run on record)
For scale: Silicon Valley Bank lost about $42 billion in one business day in March 2023, and requests for about $100 billion more were queued for the next. Preliminary findings of the independent review of Silicon Valley Bank's failure, presented by the Federal Reserve in September 2026, found that social media neither started nor sped up that run. It pointed instead to a deposit base 94% uninsured and concentrated in venture-backed technology firms, securities losses larger than its capital, and no readiness to borrow from the discount window. This page assumes agents could concentrate decisions the same way at any bank. That is an assumption, not a finding.
Contagion: the blast radius of one failure Scenario estimate
The rings: banks that share the failed bank's weaknesses
2023: nine days from Silicon Valley Bank to Credit Suisse
- 10 MarSilicon Valley Bank fails.
- 12 MarSignature Bank is closed. Regulators say both banks' depositors will be made whole.
- 13 MarRegional bank shares slump, led by First Republic.
- 14 MarMoody's cuts its outlook on the US banking system to negative.
- 15 MarCredit Suisse borrows up to 50 billion Swiss francs from the Swiss National Bank.
- 16 MarEleven large banks deposit $30 billion at First Republic.
- 19 MarUBS agrees to buy Credit Suisse.
- 1 MayFirst Republic fails.
In 2023 the turmoil reached Credit Suisse within a week. Agents acting on the same signal for many depositors could put the first rings into the same day.
How the rings are drawn
- Ground zero is a hypothetical bank that fails the way Silicon Valley Bank did. No real bank is named as ground zero, and the page does not list which banks sit in which ring: it shows counts only, and only the bank you pick learns its own ring.
- The weaknesses follow the preliminary findings of the independent review of Silicon Valley Bank's failure: a deposit base 94% uninsured, securities losses larger than its capital, and no readiness to borrow from the discount window. The tests and thresholds are this page's own; readiness to borrow is not in the filings, so the third test uses cash on hand. Each is measured from the bank's June 30, 2026 filing: uninsured deposits (filed estimate, else the account-size estimate, plus foreign-office deposits) over total deposits; amortized cost less fair value of available-for-sale and held-to-maturity securities over Tier 1 capital; cash and federal funds sold above the cash the bank never spends, over deposits.
- Each ring's outflow runs through the same model as the run simulator, with no rate change. "Would borrow the same day" means cash on hand falls short; "short even after selling securities" means cash plus securities free to sell, at the sale discount in Assumptions, still falls short. The pre-tax cost assumes the money is borrowed back at the short rate plus the spread in Assumptions; the loss assumes securities are sold instead.
- The outflow for each ring is an assumption you set, not a finding. Credit Suisse was a foreign bank and is outside this data; it appears as history only.
What this dashboard is for
It is a contingency drill for AI agents, the planning that safety and soundness asks of every bank, alongside the likelier cost. The likelier case is paying up: the money stays, but the bank starts paying interest on noninterest-bearing deposits, so it earns less on them. The contingency is a run: agents move many depositors' money on the same signal, within hours, insured or not. Neither is a forecast. For the ordinary case, use Deposit Runoff Scenarios.
1. Pick a bank
Type a name, city, certificate or RSSD in Your bank. The page opens on JPMorgan Chase Bank as an example.
2. Your bank
Set how much agents move (steps of 0.5% of deposits), how fast (hours), the rate hike, and the beta the remaining deposits reprice at. The cards at the top give the answer in three numbers: how much leaves, how much must be borrowed or raised before the end of the business day, and what it costs over a year. Chart 1 shows hour by hour when cash, then securities free to sell, then pledged collateral would run out; chart 2 shows what each part of the scenario takes from return on equity. For a bank whose holding company is listed, a market-value panel sets the run against how the market treated the banks that ran in 2023. The first four buttons set the drill: a mild, a moderate, a stress and a severe run. The last two set the likelier cost: nothing leaves, and noninterest-bearing deposits are paid 1%, or 3% as a ceiling. The page opens on the moderate run.
3. Against rate hikes
Sets the run, from 0% to 10% in steps of 0.5%, against rate hikes in steps of 25 bp, and shows where a run of a given size costs as much as a given hike, and the same-day borrowing a hike never needs.
4. Across the industry
Every bank hit by the same run on the same day: how much leaves, how many banks would borrow the same day because cash on hand falls short, how much they would borrow, what it costs the industry in pre-tax income and in realized losses.
5. Contagion: the blast radius
One bank with Silicon Valley Bank's profile fails, and the run spreads to banks that share its weaknesses: heavily uninsured deposits, securities losses large against capital, thin cash on hand. Set how much each ring loses in a day and where each weakness starts. The page shows counts only; it never lists which banks sit in which ring, and only the bank you pick learns its own.
What it cannot tell you
- How much money AI agents actually manage, or how fast they would move it. Both are assumptions you set; no filing reports them.
- How much a bank can borrow on the day. FHLB and discount-window capacity are not in the filings; pledged collateral counts only if the bank is ready to use it.