# Liquidity and funding: surviving a run with LCR and NSFR
On March 9, 2023, customers of Silicon Valley Bank tried to withdraw an estimated $42 billion in a single day. That is roughly a quarter of the bank's deposits, gone in the hours between morning coffee and market close. By the next morning, regulators had seized the bank. SVB was not insolvent in the traditional sense when the run began: it had assets on paper. It simply could not turn those assets into cash fast enough to hand money back to fleeing depositors.
This is the core lesson of bank liquidity. A bank can be solvent (assets exceed liabilities) and still die in 48 hours because it is illiquid (cannot produce cash on demand). Two regulatory ratios exist to prevent exactly this: the Liquidity Coverage Ratio (LCR) and the Net Stable Funding Ratio (NSFR).
A bank borrows short and lends long. Deposits can leave tomorrow. Loans and bonds the bank holds may not mature for years. This mismatch, called maturity transformation, is how banks make money, and it is also their permanent vulnerability.
Normally this works because depositors do not all leave at once. A bank run happens when they do. Fear becomes self-fulfilling: if you think others will withdraw, you withdraw first, which makes the bank weaker, which makes everyone else withdraw.
In 2023, that fear traveled at the speed of a group chat. SVB's depositors were concentrated in tech and venture capital, a tight, connected community. Word spread in hours, and money moved by app.
The LCR asks a simple question: if the bank suffered 30 days of severe stress, could it survive on its own liquid resources?
The formula:
LCR = High-Quality Liquid Assets (HQLA) / Total net cash outflows over 30 days
Regulators require this to be at least 100 percent. In plain terms, a bank must hold enough easy-to-sell assets to cover a month of modeled withdrawals.
High-Quality Liquid Assets (HQLA) are assets you can convert to cash quickly without a big loss. Cash and central bank reserves are the gold standard. Government bonds count too, though with small "haircuts" (a discount applied to their value to reflect selling risk).
Here is the trap SVB fell into. It held large amounts of long-dated US Treasury and mortgage bonds. Those are high quality. But the bank had booked many of them as held-to-maturity (HTM), an accounting category that lets a bank carry a bond at its original price rather than its current market price, on the promise it will hold the bond until it matures.
When interest rates rose sharply in 2022, the market value of those old, low-rate bonds fell hard. Selling them meant crystallizing a loss. SVB did sell a chunk, reported the loss, and that disclosure was the spark that lit the run.
The deeper problem: those "liquid" assets were only liquid at a painful price. Liquidity and solvency collided.
The denominator matters just as much. LCR rules assign run-off rates to different deposit types, an estimate of how much of each will flee in stress.
SVB's deposit base was extreme. A large majority of its deposits were above the FDIC insurance limit (250,000 dollars per depositor in the US). Uninsured depositors have every reason to run at the first sign of trouble, because they can lose real money. A standard LCR model would flag this base as dangerously flighty.
Worth knowing: due to a rule change, banks of SVB's size were not subject to the full LCR requirement at the time. A key regulatory guardrail simply did not apply to it. You can read the official post-mortem in the Federal Reserve's review of SVB, which is candid about supervisory gaps.
If the LCR is about surviving a 30-day storm, the NSFR is about the structure of the ship. It looks at a one-year horizon and asks: is the bank funded with stable money, or with money that could vanish?
The formula:
NSFR = Available Stable Funding / Required Stable Funding
This must also be at least 100 percent.
The NSFR pushes banks to match long-term assets with long-term, dependable funding. A bank funding 10-year loans with hot overnight money would fail this test.
SVB's funding was heavily tilted toward volatile, uninsured, concentrated deposits, exactly the kind that an NSFR calculation would treat as low-quality funding. Its assets included long-duration bonds that had become hard to sell without loss.
That is a structurally fragile shape: unstable funding on one side, hard-to-move assets on the other. The ratios were designed to make precisely this mismatch visible and costly before a crisis, not during one.
Both LCR and NSFR use averages and standard assumptions. They can miss concentration risk, the danger that comes from having similar customers who behave the same way at the same time.
SVB's depositors were not a diversified crowd. They were interconnected founders and funds who shared advisors, investors, and Slack channels. When a few large venture firms told portfolio companies to pull their cash, the withdrawal was correlated and instant. Standard models assume depositors act somewhat independently. Here they acted as one herd.
This is the modern twist. A 1930s run required a physical line outside a branch. A 2023 run required a wire transfer and a rumor. Speed compresses the survival window from weeks to hours, which is why the 30-day LCR assumption itself may understate the risk for socially networked, digitally banked customer bases.
Vérification des acquis
1. A bank has assets that exceed its liabilities on paper, yet it collapses within 48 hours during a deposit run. What best explains this outcome?
2. Why is maturity transformation described as both a bank's source of profit and its permanent vulnerability?
3. A bank reports an LCR of 85 percent. What does this indicate under the regulatory standard?
4. Select ALL correct answers about why a bank run can become self-fulfilling.
Sélectionnez toutes les réponses correctes.
5. Select ALL correct answers describing the purpose and logic of the Liquidity Coverage Ratio (LCR).
Sélectionnez toutes les réponses correctes.
Ratios are the scoreboard. The real work is operational.
Diversify funding. Do not rely on one customer type. A mix of insured retail deposits, corporate accounts, and term borrowing is harder to spook all at once.
Hold genuine HQLA. Keep a buffer of cash and short-dated government securities that can be sold or pledged instantly, even in a bad market.
Use the central bank backstop. Banks can pledge assets to a central bank to borrow cash immediately. After 2023, the Federal Reserve created the Bank Term Funding Program, which let banks borrow against bonds valued at par (full face value) rather than depressed market prices. That program has since closed, but the lesson stands: access to a lender of last resort is part of a liquidity plan.
Run internal stress tests. Do not just meet the regulatory LCR. Model your own worst case, including a fast, concentrated run, and hold buffers above the minimum.
Manage interest rate risk on the asset side. SVB's real original sin was letting a huge bond portfolio sit exposed to rising rates without hedging. Liquidity risk and interest rate risk are linked: a rate shock destroyed the value of the assets that were supposed to be the liquidity cushion.