The type of risks that put banks in the greatest jeopardy – and led to the recent crisis – are called tail risks. They have a small probability of turning out badly but are extremely costly if they do. Banks took on two kinds of tail risks prior to the crisis. One was economywide default risk, the risk that a broad portfolio of assets, such as mortgages, would suffer deep losses. The other was liquidity risk, essentially the way they financed the first kind of risk.

Some banks – such as Citibank, Lehman Brothers, and Royal Bank of Scotland – loaded up on both risks, holding enormous quantities of mortgage-backed securities on the asset side and paying for them with short maturity debt on the liability side. Why did they do it? The simple answer: It was very profitable, provided the tail events did not materialize. Think of insurers that write a lot of earthquake policies (another tail risk). If you didn’t know they were writing earthquake insurance and not setting aside reserves, you would think they were enormously profitable until there’s a quake. For banks, there was always the threat of a day of reckoning when liquidity dried up and defaults skyrocketed. But they set aside few reserves against that happening.
…
Particularly worrisome, as my colleague Douglas Diamond and I have argued, is that once banks are leveraged enough that they will be severely distressed if economywide liquidity dries up, they double down on risky bets.
…
Here’s the drill: To make it harder for tail-risk-taking banks to grow, all banks should be required to issue a minimum level of debt (say, 10% of assets) that is automatically impaired – either converted to equity or written down – if the bank suffers sufficient losses. This will quickly change debt holders’ views on risky expansion. Moreover, no financial institution should be allowed to hold this debt.