“Early” and “Late” Cycle Verdicts are Baseless
The current U.S. expansion, now in its eighth year, has lasted a year longer than the previous expansion and is substantially longer than the historical average. This has led to many pronouncements that we are in the “late cycle” phase of this economic expansion, which must therefore end soon. In other words, despite the truism that “expansions don’t die of old age,” there is a widespread belief that expansions have natural expiry dates.
But business cycle expansions not only vary greatly in duration, they also do not follow standard statistical distributions, such as a well-behaved bell curve, that could provide a simple basis for predicting the duration of future cycles. In fact, in order to attempt such a forecast, it is critical to know how they are distributed historically. Using ECRI’s chronologies of recession dates for 21 countries, we calculated the durations of 120 business cycle expansions across all of these economies.
The chart shows a histogram of the durations of these expansions, presented as percentages of the total. While the range of durations is quite wide, lasting from under a year to well over a decade, business cycle expansions are not clearly clustered around any particular duration.
This is evident from the distribution of expansions, which is not unimodal, and certainly not bell-shaped. Instead, since the empirical distribution looks rather flat and has no real peak, it resembles a uniform distribution. Indeed, a chi-square test comparing it with a uniform distribution showed no significant difference.
Unfortunately, with such a distribution — where the length of expansion looks just as likely to be a year or less as it is to be in the three- to four-year range or the eight- to nine-year range — it is illogical to assert that the duration of an ongoing expansion would be around six to ten years merely because the last three U.S. expansions lasted that long. Thus, the notion that economists can tell whether we are in the early, middle, or late stage of an expansion is unreasonable, and unsubstantiated.
Some years ago, the IMF concluded from a 63-country study that economists’ “record of failure to predict recessions is virtually unblemished.” Keeping in mind this near-perfect record of failure to predict recessions — even a few months in advance — how could economists have any clue as to whether a recession is one or five years away?
In the end, the mid-cycle / late-cycle talk is based on the implicit assumption that expansions last between six and ten years, but this is based on just three data points, i.e., just the last three U.S. expansions. Even disregarding the fact that trend growth is noticeably lower in the current expansion than it used to be, this is hardly a big enough sample to draw any statistical inferences from.
Rather, what matters most is whether an economy has entered a cyclical window of vulnerability, making it susceptible to recessionary shocks. Here ECRI’s leading indexes, which have demonstrated significant leads in real time at cycle turning points, continue to provide crucial insight.