What Bank Failures Teach Us About Risk

Banks are strange businesses. They take your money, lend most of it to strangers, keep a thin slice as reserve, and then promise you can have it all back whenever you want. This works perfectly – until it does not. And when it does not, the results are spectacular in the worst possible way. In 2023 alone, Silicon Valley Bank, Signature Bank, and First Republic collapsed in a matter of days. Credit Suisse, a 167-year-old institution, was forced into a shotgun merger. These were not small-town banks run by amateurs. They had risk committees, chief risk officers, and thick binders of regulatory compliance. None of it mattered when the fundamental trust broke down. If you invest in bank stocks – or simply keep your money in a bank – understanding why banks fail is not optional. It is basic financial literacy.

Why Do Banks Actually Fail?

The mechanics of bank failure are almost boringly simple, which makes it even more embarrassing that it keeps happening.

Leverage is the original sin. A typical bank operates with 10:1 or even 15:1 leverage. For every dollar of equity, the bank has lent out ten or fifteen dollars. This means a 7-10% decline in asset values wipes out all the equity. Gone. Zero. The bank is insolvent. In normal times, this leverage is what makes banks profitable – you earn a spread on money that is not really yours. In bad times, this leverage is what kills them. Every single major bank failure in history traces back to this arithmetic. It is not complicated. A 10% loss on a 10:1 leveraged balance sheet means 100% of equity is destroyed.

Maturity mismatch is the structural weakness. Banks borrow short (your checking account, which you can withdraw tomorrow) and lend long (30-year mortgages, 10-year commercial loans, long-duration bonds). This is literally the business model. The spread between short-term borrowing costs and long-term lending rates is how banks make money. But it also means banks are perpetually vulnerable to changes in interest rates and depositor behavior. Silicon Valley Bank is the textbook example. They loaded up on long-duration government bonds when rates were near zero. When rates rose rapidly in 2022-2023, those bonds lost enormous value on a mark-to-market basis. The bonds were perfectly safe if held to maturity – but SVB’s depositors did not want to wait that long. They wanted their money now.

Bank runs are the kill shot. A bank can survive bad loans. A bank can survive mark-to-market losses. What no bank can survive is all its depositors showing up at once and demanding their money. Because the money is not there. It is lent out, invested, locked up in illiquid assets. This is not fraud – it is the fundamental structure of banking. SVB lost $42 billion in deposits in a single day. In 2023. With smartphones. No lines around the block, no newsreels of panicked crowds. Just a group chat, a few tweets, and $42 billion moved to other institutions in hours. The speed of modern bank runs is genuinely terrifying. What used to take weeks now takes a lunch break.

The combination of high leverage, maturity mismatch, and the possibility of a run means every bank is, in some sense, always one bad week away from trouble. The question is not whether banks can fail. The question is what separates the ones that do from the ones that do not.

Does “Too Big to Fail” Make Things Better or Worse?

When a large bank gets into trouble, governments face an ugly choice: let it fail and risk systemic contagion, or bail it out and create moral hazard. There is no good option. There is only a less bad option.

The moral hazard problem is real and unresolved. When banks believe they will be rescued, they take more risk. This is not cynicism – it is rational economic behavior given the incentive structure. If profits are private and losses are socialized, the optimal strategy is to maximize leverage and take on as much risk as possible. Heads you win, tails the taxpayer loses. The 2008 bailouts reinforced this lesson. The 2023 response to SVB reinforced it again, when regulators guaranteed all deposits – even those above the $250,000 FDIC insurance limit – to prevent contagion.

Credit Suisse showed that even a global institution can be forced into extinction. Swiss regulators essentially forced UBS to acquire Credit Suisse over a weekend. AT1 bondholders – who thought they held relatively safe senior debt – were wiped out entirely while equity holders received some compensation. This inverted the traditional creditor hierarchy and sent shockwaves through the banking sector globally. The lesson: when regulators are in crisis mode, the rules you thought you understood might not apply.

FTX was a quasi-bank failure without any banking regulation. This is worth examining because it shows what happens when bank-like activities operate outside the banking regulatory framework. FTX accepted customer deposits, lent them out (to Alameda Research), took enormous leveraged positions, and then could not meet withdrawal requests. It was a classic bank run on an institution that was not a bank. No deposit insurance. No central bank backstop. No orderly resolution framework. Customers lost everything. The parallel to traditional bank failures is almost exact – leverage, maturity mismatch, run dynamics – except with zero institutional safeguards.

The uncomfortable truth is that moral hazard and too-big-to-fail are features of the system, not bugs to be fixed. As long as banks are interconnected, as long as payments infrastructure depends on a handful of institutions, governments will intervene to prevent cascading failures. Knowing this changes how you should analyze bank stocks.

How Should You Evaluate a Bank Before Investing?

Bank stock analysis is different from analyzing a software company or a retailer. The balance sheet dominates. Culture matters more than you think. And the things that will eventually blow up are almost never visible in the annual report.

Culture is the first thing to evaluate, and the hardest. The banks that survive crises are the ones where risk management is embedded in the institutional DNA – not outsourced to a risk committee that meets on Thursdays. A chief risk officer is, frankly, often just an employee who makes everyone feel comfortable while doing questionable things. What matters is whether the CEO personally understands risk, personally cares about it, and has the authority and willingness to say no to profitable-but-dangerous activities. You can sometimes assess this by reading earnings calls, watching how management discusses risk (do they minimize it or take it seriously?), and tracking whether the bank has avoided the fashionable-but-stupid trends of each cycle.

Look at the loan book composition. What types of loans does the bank make? Commercial real estate in 2025 is a known risk area – office vacancy rates remain elevated, and many commercial mortgages originated in the low-rate era are coming due with significantly higher refinancing costs. A bank with 40% of its loan book in commercial real estate has a very different risk profile than one focused on consumer lending or residential mortgages. Concentration kills. Diversification across loan types, geographies, and industries is a basic safeguard.

Watch the deposit base. SVB failed partly because its deposit base was extremely concentrated – tech startups and venture capital firms, many with deposits well above the FDIC insurance limit, all connected to each other, all likely to panic simultaneously. A bank with millions of small retail depositors has a much more stable funding base. Those depositors are insured, they are not on Twitter, and they do not coordinate withdrawals in a group chat. The stickiness of the deposit base is arguably the single most important factor in a bank’s crisis resilience.

Examine interest rate sensitivity. Banks report this in their filings, but you have to actually read it. How much does the bank’s net interest income change if rates move 100 basis points? 200 basis points? How much duration risk is embedded in the securities portfolio? SVB’s failure was entirely predictable to anyone who read their 10-K and understood the duration mismatch. The information was public. Almost nobody acted on it until it was too late.

The fintech angle adds a new dimension. Digital banks, neobanks, and fintech lenders operate with the same fundamental risks as traditional banks – leverage, maturity mismatch, run potential – but often with thinner capital buffers and less regulatory oversight. Some are excellent businesses. Some are traditional banking risk wrapped in a mobile app with better marketing. Do not confuse a modern user interface with modern risk management.

Here is a simple screen for bank stock risk assessment:

• Tier 1 capital ratio above 12% – provides cushion against losses
• Loan-to-deposit ratio below 90% – indicates adequate funding
• Non-performing loan ratio below 1.5% – signals loan quality
• Deposit concentration – no single depositor or group should represent more than 5% of total deposits
• Management tenure and track record through previous cycles – did they survive 2008? 2020? 2023?

Key Takeaways

• Banks fail because of leverage, maturity mismatch, and runs. This has not changed in 300 years. Only the speed has increased. Modern bank runs happen in hours, not weeks.
• A chief risk officer is not a substitute for a risk-aware culture. The banks that survive are the ones where the CEO understands and owns risk management personally. You cannot outsource this function.
• Too-big-to-fail creates moral hazard. Banks that believe they will be bailed out rationally take more risk. As an investor, factor this into your analysis: which banks benefit from implicit government guarantees, and which do not?
• Deposit base stability matters more than loan book quality. A bank can survive bad loans over time. It cannot survive a run. Concentrated, uninsured deposit bases are the highest-risk factor.
• Read the interest rate sensitivity disclosures. SVB’s collapse was visible in public filings months before it happened. Duration mismatch is not hidden – it is just boring, and most people skip those sections.
• Fintech and crypto platforms can exhibit identical failure modes to traditional banks. FTX was a bank run without deposit insurance. The underlying mechanics – leverage, mismatch, loss of confidence – are universal.
• Complexity is the enemy of understanding. When financial products require 750,000 pages of documentation to understand, nobody actually understands them. If you cannot explain a bank’s risk profile in simple terms, that itself is the risk.
• Bank failures will happen again. The specific trigger will be different, but the pattern is always the same: leverage, overconfidence, a belief that this time is different, and then a sudden loss of trust. Position your portfolio accordingly.

The history of banking is a history of periodic crises, each one generating new regulations that prevent the last crisis from recurring while doing nothing about the next one. As an investor, the most valuable thing you can do is understand the basic mechanics – leverage, maturity mismatch, run dynamics, moral hazard – and then evaluate every bank through that lens. The banks worth owning are the ones run by people who genuinely lose sleep over what could go wrong. There are fewer of those than you think. And in a world where $42 billion can leave a bank in a single day via smartphone, the margin for error is exactly zero.