Black Swan Event

A black swan event, a term introduced by Nassim Nicholas Taleb in his 2007 book of that name, is an event with three attributes. In his words: "First, it is an outlier, as it lies outside the realm of regular expectations, because nothing in the past can convincingly point to its possibility. Second, it carries an extreme impact. Third, in spite of its outlier status, human nature makes us concoct explanations for its occurrence after the fact, making it explainable and predictable." Taleb summarizes the triplet as "rarity, extreme impact, and retrospective (though not prospective) predictability."

The name comes from the older philosophical example he opens with: Europeans held that all swans were white, an "unassailable belief as it seemed completely confirmed by empirical evidence", until Australia produced a black one. His point is that "one single observation can invalidate a general statement derived from millennia of confirmatory sightings", which is a claim about the limits of induction rather than about probability.

The third criterion is the one that does the work, and it is the one most summaries drop. A black swan is not merely a rare, large event. It is a rare, large event that becomes obvious in hindsight, so that the people who did not anticipate it end up convinced they nearly did. That retrospective clarity is why the concept keeps failing to change behavior: each time, the last one is filed as explicable and the next one is treated as unimaginable.

Because criterion one is written from a particular vantage point, black swan-ness is relative rather than absolute. An outcome that nothing in an investor's own experience pointed to may be entirely ordinary in a longer data series, in another country's history, or to a counterparty who can see the whole position. The same event is a black swan for one party and a scheduled cost of doing business for another, which is why the term describes a state of knowledge as much as it describes an event.

Two failure modes follow, and they run in opposite directions. The first is over-application. Almost anything sharp and unpleasant now gets called a black swan, including declines that arrive on a regular schedule and risks that were written down in advance. Taleb himself has rejected the label for the 2020 coronavirus pandemic in public statements, arguing that a global pandemic was foreseeable, had been foreseen and warned about, and was therefore not a black swan at all. Whatever one makes of that particular judgment, it illustrates the test: if the possibility was in the record, criterion one is not met.

The second failure mode is the more expensive one. Calling something a black swan can function as an excuse, converting a foreseeable exposure into an act of nature that no one could have planned for. That reading is the exact opposite of Taleb's argument, which is that most of the consequential outcomes in the world come from the tails and that plans built on the middle of a distribution are therefore fragile in a way their builders do not notice.

The modeling consequence is worth stating carefully. Standard risk models describe returns with a normal distribution and summarize them with a standard deviation, which is a well-behaved measure that assigns extremely small probabilities to extremely large moves. Real financial returns produce large moves more often than that model implies. Our page on standard deviation covers why, and the practical upshot here is only this: a probability quoted for a severe outcome is a property of the model, not a measurement of the world.

So what is left to do, if the events cannot be forecast? The concept points at structure rather than prediction. It argues for asking what would still be true after an outcome the plan did not contemplate: whether spending would have to be funded by selling at the bottom, whether any obligation forces a sale at a chosen moment, whether a single failure can take out more than one part of the plan at once, and whether the plan depends on being able to transact in a market that may not be functioning. None of that requires knowing what the event will be.

The mechanism the concept describes is that outcomes far out in the tail can dominate a long record, so an average computed over the ordinary years describes almost none of the result.

A hypothetical illustration, chosen because the arithmetic is checkable. A portfolio starts at $100,000 and gains 10 percent a year for nine consecutive years. Compounding, $100,000 multiplied by 1.10 nine times is $235,794.77.

In the tenth year it loses 60 percent. Sixty percent of $235,794.77 is $141,476.86, leaving $94,317.91.

Now compare two descriptions of that decade. The arithmetic average of the ten annual returns is nine years of positive 10 percent plus one year of negative 60, divided by ten, which is positive 3 percent a year. The actual result is that $100,000 became $94,317.91, a loss of $5,682.09, or about negative 0.58 percent a year compounded. The average return is positive and the money is gone. The single tail year, one observation in ten, decided the entire outcome, and no summary statistic computed from the other nine would have hinted at it.

What the concept is good for

• It names a real and specific failure of induction: a long run of confirming observations can leave a belief completely unprotected.
• It shifts attention from the probability of a bad outcome, which is usually unknowable, to the consequences of one, which are often estimable.
• It explains why plans that look robust on historical data can be fragile, since the data contains the ordinary years by construction.
• It has a built-in test for its own misuse, in the requirement that nothing in the past pointed to the possibility.

Where it goes wrong in practice

• The label is applied to foreseeable events, which converts a planning failure into bad luck.
• It is unfalsifiable in the moment: nothing can be identified as a black swan in advance, by definition, so it cannot inform a specific forecast.
• It can be used to justify doing nothing, on the grounds that the important risks are unknowable, which is not the argument Taleb makes.
• Preparing for extreme outcomes has a real ongoing cost, and a plan that carries too much of it may simply fail slowly instead of quickly.