Mervyn King delivered a provocative and intriguing 2017 Martin Feldstein Lecture at the National Bureau of Economic Research on the subject of “Uncertainty and Large Swings in Activity”(July 19, 2017). A written version of the presentation is available in the NBER Reporter (2017: 3, pp. 1-10), or you can watch the lecture and download the slides here. 

King’s argument has both a broad conceptual message for the study of macroeconomics, which is that it is literally impossible to demonstrate with statistics that a certain macroeconomic model is “true.” After all, drawing statistical conclusions requires a decent sample size. But to get a sample size of, say, 20 or 30 recessions in a given economy would take a long time–perhaps several centuries–and it is not plausible that any macroeconomic model remains “true” over that length of time. As King puts it (footnotes omitted):

“Let me give a simple example. It relates to my own experience when, as deputy governor of the Bank of England, I was asked to give evidence before the House of Commons Select Committee on Education and Employment on whether Britain should join the European Monetary Union. I was asked how we might know when the business cycle in the U.K. had converged with that on the Continent. I responded that given the typical length of the business cycle, and the need to have a minimum of 20 or 30 observations before one could draw statistically significant conclusions, it would be 200 years or more before we would know. And of course it would be absurd to claim that the stochastic process generating the relevant shocks had been stationary since the beginning of the Industrial Revolution. There was no basis for pretending that we could construct a probability distribution. As I concluded, `You will never be at a point where you can be confident that the cycles have genuinely converged; it is always going to be a matter of judgment.'”