SVB 2023: How a Bank Run Showed Up in Market Data

Silicon Valley Bank went from announcing a capital raise to FDIC receivership in less than two days. Over the next eight weeks Signature Bank and First Republic Bank also failed, Credit Suisse was folded into UBS over a weekend, and discount-window borrowing at the Federal Reserve climbed above its 2008 peak. Yet SPY, the S&P 500 ETF, never lost 5% over any five trading days in the episode, and high-yield spreads widened only to around 500 basis points.

SVB was a duration and funding crisis, not a credit crisis, and it surfaced in a specific order: regional-bank stocks and short-dated Treasuries first and hardest, equity volatility second, corporate credit barely. A monitor built on credit spreads and the VIX would have under-read it. One built on bank equities and flight-to-safety flows would have caught it, then forgotten it within days.

Timeline: eight weeks, three US bank failures

Why SVB was fragile

The Fed's review names three structural weaknesses on top of poor management: a concentrated business model, interest-rate risk and heavy reliance on uninsured deposits.

• Uninsured funding. About 94% of SVB's deposits were uninsured at the end of 2022, more than double the large-bank average. Uninsured depositors can lose money in a failure, so they have every reason to leave at the first doubt.
• Deposit growth parked in duration. SVB grew from $71 billion to over $211 billion in assets between 2019 and 2021. Much of the deposit inflow went into long-dated agency mortgage-backed securities held to maturity (HTM), a book with a weighted-average duration of 6.2 years at end-2022.
• Losses outside the capital ratio. HTM accounting carries securities at amortized cost. At end-2022 the roughly $91 billion HTM portfolio was worth about $15 billion less than its carrying value, against about $16 billion of equity (ICAEW, from SVB's 10-K). Its reported CET1 ratio of 12% was 2 points above the large-bank average.
• A depositor base that moved as a herd. Clients were venture-backed companies and their investors, linked through the same networks. Barr's cover letter concludes that "social media, a highly networked and concentrated depositor base, and technology may have fundamentally changed the speed of bank runs."

Speed broke the old assumptions. Per the Fed's May 2023 Financial Stability Report, Washington Mutual's worst day in 2008 cost it just over 2% of deposits; SVB and Signature each lost more than 20% in a day. SVB's $42 billion on March 9 plus the $100 billion queued for March 10 came to roughly 85% of its deposit base.

What markets did, and in what order

1. Bank stocks moved first, partly before SVB spoke

On March 7, as Powell told the Senate rates would probably peak higher, the SPDR S&P Regional Banking ETF (KRE) fell 3.2% and the two-year yield reached 5%. The market was already pricing the channel that would sink SVB: higher rates eroding the value of banks' fixed-rate assets. After the March 8 announcement KRE fell 8.1%, 4.4% and 12.3% on the next three sessions, 23% in total. SVB's own shares fell 60% on March 9 alone, a reminder of how much a basket dilutes a single-name run.

2. Short-dated Treasuries gave the loudest signal

Treasury's constant-maturity two-year yield went from 5.05% on March 8 to 4.03% on March 13, about a full percentage point in three sessions, as investors bought safety and pared back expectations of further Fed hikes. The Fed's Financial Stability Report called the March 13 drop, more than 60 basis points, the largest one-day decline since 1987. New York Fed researchers found two-year note volatility reached more than twice its March 2020 level, and the ICE BofA MOVE index of Treasury option volatility hit its highest since late 2008. Bills were bid too: volume in BIL, a 1–3 month T-bill ETF, ran at 2.7 and 4.0 times its 20-day average on March 13 and 14 while its price edged up.

3. Equity volatility: elevated, not extreme

The VIX rose from 19.11 on March 8 to a closing high of 26.52 on March 13, touching 30.81 intraday. SPY's worst five-session loss was 4.7%. The VIX prices expected volatility in the S&P 500, an index dominated by companies that do not fund themselves with uninsured deposits. Our guide to reading the VIX covers what it does and does not measure.

4. Credit widened but did not break

The ICE BofA US High Yield option-adjusted spread widened 42 basis points to 503bp on March 13, the biggest daily widening since June 2020 and the first move above 500bp since October 2022, Reuters reported. In price terms it was small: HYG's worst five-day loss was 2.3%, and the SRLN leveraged-loan ETF's was 2.1%. SVB's losses sat on government-backed securities and were created by interest rates, which is why the credit-spread gauges most monitors lean on stayed comparatively calm.

5. The Fed's balance sheet showed the scale within a week

The Fed's weekly H.4.1 release for Wednesday, March 15 showed primary credit (the discount window) at $152.9 billion, above its 2008 peak. The new BTFP had lent $11.9 billion, and $142.8 billion of "other credit extensions" had gone to the bridge banks the FDIC set up for SVB and Signature. Total Fed assets rose $297 billion in a week, undoing nearly half of the shrinkage since the April 2022 peak. Federal Home Loan Banks added about $250 billion of debt in the week ending March 17.

Why it stayed contained

Three US banks with roughly $550 billion of combined assets failed, and the damage stopped at the edge of the regional-bank sector. Four differences from 2008 explain most of it.

1. The assets were sound; the timing was bad. SVB held Treasuries and agency mortgage bonds that would pay in full at maturity. BTFP valued that collateral at par, so banks could borrow its face value instead of selling at a loss, the move that killed SVB. In 2008 the problem assets were mortgage credit whose eventual value nobody could pin down.
2. Deposits were guaranteed fast. The systemic risk exception, announced Sunday, March 12, covered every SVB and Signature depositor in full, and SVB customers could reach all their money on Monday. The question shifted from "will I get my money" to "which bank do I prefer."
3. Money moved within the system, not out of it. The Fed's Financial Stability Report notes the largest banks saw significant deposit inflows while smaller banks lost deposits.
4. Few counterparties were exposed. Barr warns that distress can spread "even if the firm is not extremely large, highly connected to other financial counterparties, or involved in critical financial services," a fair description of SVB. That same lack of connection kept losses bounded once deposits were covered, and the Fed judged US spillovers from Credit Suisse "muted." Our contagion map follows how stress travels between regions.

The bill landed on the Deposit Insurance Fund: FDIC estimates at the time of each sale were about $20 billion for SVB, $2.5 billion for Signature and $13 billion for First Republic.

What the official post-mortems concluded

Three official reviews were published on April 28, 2023.

• Federal Reserve (Barr review). SVB's board and management failed to manage their risks; supervisors did not fully appreciate the vulnerabilities as the bank grew; when they found problems, they did not push for fixes fast enough; and the Board's tailoring approach after the 2018 EGRRCPA, plus a less assertive supervisory stance, impeded supervision. Barr added that "the proximate cause of SVB's failure was a liquidity run," but "the underlying issue was concern about its solvency."
• FDIC, on Signature Bank. The root cause was poor management: rapid, unrestrained growth funded by uninsured deposits without sound liquidity risk practices. The FDIC said it could have escalated sooner and that examiner resource shortages hurt the timeliness and quality of its exams.
• GAO. Risky strategies plus weak liquidity and risk management drove both failures. Assets grew 198% at SVB and 134% at Signature from 2019 to 2021, against a 33% peer median. The San Francisco Fed downgraded SVB in June 2022 and began an enforcement action that August but had not finalized it when the bank failed; the FDIC did not substantially downgrade Signature until the day before it failed.

The common thread: growth rate and funding mix were the warnings, and both sat in public filings long before March 2023. Reported capital ratios were not.

How CrisisBoard's US rules read these moves

CrisisBoard launched in 2026, so it was not watching in 2023. Its rules are published on the methodology page and are mechanical, so the 2023 daily closes can be replayed through them. The US region tracks KRE, HYG, SRLN, BIL, SPY, S&P futures, IGV, BIZD and the VIX, and enters MONITOR at 8 points, PHASE 1 at 10 and PHASE 2 at 12.

What stayed silent matters as much. The broad-market drawdown rule missed by 0.3 percentage points, and HYG never left the "warning" band on CrisisBoard's reference scale. The funding-stress rule looks for T-bills being sold to raise cash, as in March 2020; in 2023 investors were buying them. The central-bank layer is the exception. Its CB_fed_emergency_qe rule (5 points) fires when Fed total assets grow more than 2% in a week, and the week to March 15 showed 3.6%, from $8.34 trillion to $8.64 trillion. The next week's 1.1% would not have qualified. That signal scores in the separate central-bank layer, not in the US totals above.

Practical lessons for reading bank-stress signals

1. Pair bank equities with the two-year yield. Bank stocks falling while the two-year yield plunges means the market is pricing funding stress and a policy pivot at once, not just a growth scare.
2. Separate duration stress from credit stress. Ask where the losses come from. If rates are eroding safe assets, watch bank equities, the MOVE index and the front end of the curve. If defaults are the threat, watch spreads.
3. Read flows, not only prices. T-bill ETF volume, the Fed's Thursday H.4.1 and Federal Home Loan Bank issuance showed the size of the run within a week.
4. Check funding mix and growth before the shock. Uninsured-deposit share and asset growth are public and were flagged by every post-mortem; capital ratios were not the warning.
5. Expect contagion along business models, not size. The banks that followed SVB shared its mix of uninsured deposits and rate exposure.
6. Do not let a falling score end the story. Short-window rules reset after the shock. The level against the pre-shock baseline tells you whether the damage has healed.