Murphy’s law is well-known in the form: “Whatever can go wrong, will go wrong” and similar variations on the theme. But the intellectually interesting substance of Murphy’s law is:  “Whatever can go wrong, will go wrong, given enough time.”

When a financial calamity has a very small probability of occurring—let’s say a 1 percent chance that it will and 99 percent that it won’t in any given year—we tend not, as a practical matter, to worry about it much. In most years, nothing will happen, and when it hasn’t happened for a long time, we may even start to treat the risk as essentially zero. Professors Jack Guttentag and Richard Herring authored a classic paper that gave this tendency the provocative name “disaster myopia.”

Banking and finance are full of events with a very small expected probability, but which are very costly when they do happen – e.g., a financial crisis.

Suppose the chance of a financial crisis is 1 percent annually. Suppose you optimistically start your banking career at the age of 23 and work to age 68, by which time you will be seasoned and cynical. That will be 45 years. Because you have given it enough time, the probability that you will experience at least one crisis during your career grows from that 1 percent in your trainee year to a pretty big number: 36 percent.

We observe in the real world that financial crises occur pretty frequently—every decade or two—and that there are a lot of different countries where a financial crisis can start. We also observe that virtually no one—not central bankers, regulators, bankers, economists, stock brokers or anybody else—is good at predicting the financial future successfully. Do we really believe the risk management and credit screens of banks, regulators and central banks are as efficient enough to screen down to a 1 percent probability?  I don’t.

Suppose instead that the probability of the banking crisis is 2 percent, with 98 percent probability that it won’t happen in a given year. Are banks even that good?  How about 5 percent, with a 95 percent probability of not happening?  That would still feel pretty safe. One more dubious of the risk-management skills of bankers, regulators and the rest might guess the probability, in reality, is more like 10 percent, rather than 1 percent. Even then, in most years, nothing will happen.

How does our banker fare over 45 years with these alternate probabilities?  At 2 percent chance per-year, over 45 years, there is a 60 percent probability he will experience at least one crisis. At 5 percent, the probability becomes 90 percent of at least one crisis, with a 67 percent chance to see two or more. If it’s 10 percent, then over 45 years, the probability of experiencing at least one crisis is 99 percent, and the probability of experiencing at least two is 95 percent. Since we learn from troubles and failures, banking looks like it furnishes the probability of an educational career.

In the last 45 years, there have been financial crises in the 1970s, 1980s, 1990s and 2000s. In the 2010s, we have so far had a big sovereign default in Greece, set the record for a municipal insolvency with the City of Detroit, and then broke that record with the insolvency of Puerto Rico. And the decade is not over. All of these crises by decade have been included in my own career around banking systems, of now close to 48 often-eventful years. The first one—the Penn Central Railroad bankruptcy and the ensuing panic in the commercial paper market—occurred when I was a trainee.

Since 1982, on average, a little less than 1 percent of U.S. financial institutions failed per year, but in the aggregate, there were 3,464 failures. Failures are lumped together in crisis periods, while some periods are calm. There were zero failures in the years 2005-2006, just as the housing bubble was at its peak and the risks were at their maximum, and very few failures in 2003-2004, as the bubble dangerously inflated. Of course, every failure in any period was a crisis from the point of view of the careers of then-active managers and employees.

A further consideration is that the probability of a crisis does not stay the same over long periods—especially if there has not been a crisis for some time. As Guttentag and Herring pointed out, risks may come to be treated as if they were zero, which makes them increase a lot. The behavior induced by the years in which nothing happens makes the chance that something bad will happen go up. In a more complex calculation than ours, the probability of the event would rise over each period it doesn’t occur, thanks to human behavior.

But we don’t need that further complexity to see that, even with quite small and unchanging odds of crises, given enough time across a career, the probability that our banker will have one or more intense learning experiences is very high, just as Mr. Murphy suggests.

Image by Ionut Catalin Parvu

Featured Publications