Like most people, I suppose, I get my hair cut every four weeks. If, either by consulting the calendar or the mirror, I am “due” for a haircut, I head off and get one. The passage of time or the growth of my hair since my last visit is a very reliable predictor of the timing of my next one.
If, on the other hand, I spend an evening sitting at a roulette table, hypnotized by the spinning wheel, and stake a bet each time on lucky number seven, the length of time since it was last called out has no bearing on when it will be called again. I may feel that I am “due.” But the reality is that each spin is independent, with an exactly 1-in-38 shot of providing me with a win. Of course, given a 1-in-38 shot, seven will eventually come up. But the number of spins since the last seven has no bearing on that probability.
Since the short, sharp bear of February and March 2020, the stock market has moved very steadily higher, with the S&P500 logging a more than 100% increase by earlier this month. This has led many to argue that we are “due” for a correction. But does the increasing length or strength of a bull market actually make the next correction more imminent? In other words, is the stock market more like haircuts or roulette wheels?
This is actually a relatively easy question to investigate. First, to be precise, a stock market correction is defined as a decline of at least 10% from a recent peak. That peak does not have to be an all-time high. However, by convention, we only count a new correction as occurring if the market has rebounded at least 10% from the trough of the last one.
Using these rules, there have been 36 corrections in the S&P500 since 1950 and the average length of time between the start of corrections has been 748 days.
If corrections were as regular as haircuts then there would be no variation in this number. The expected length of time between corrections would be exactly 748 days, with a standard deviation around that expectation of 0 days. With 579 days elapsed since the last correction started, we could note on our calendars that we have 169 days to go until the next one.
If, on the other hand, the length of time since the last correction had no bearing at all on the timing of the next one, then the length of time until the next correction would technically follow what is known as a geometric distribution, with an expected wait of 748 days until the next correction and a standard deviation around that expectation that also, conveniently, equaled 748 days.
So which of these most closely resembles actual history?
Well, since 1950, while the average length of time between the start of corrections has been 748 days, the longest has been 2,640 days, between 1990 and 1997, and the shortest has been 131 days, in the winter of 1980.
Importantly, the standard deviation of these times has been 540 days. This is 72% of the standard deviation we would expect if the probability of correction was independent of the timing of the last one. So mostly roulette wheels, but partially haircuts.