Things tend to even out over time. If something is extraordinarily high or low on first measurement, it will often be closer to the mean the next time it’s measured. Reversion (or regression) to the mean refers to the phenomenon where extreme values in a random variable tend to move closer to the average in subsequent measurements.
Reversion to the mean was described by Francis Galton, a pioneer in the field of statistical genetics, proponent of eugenics and half-cousin of Charles Darwin. Galton observed that the children of exceptionally tall parents would often be shorter than their parents and more similar in height to other people.
Recency bias is a related concept. If the stock market has crashed last week, I shouldn’t make investments thinking that it’s more likely to crash again anytime soon than I’d have assumed two weeks ago. We tend to rely too much on events that just happened compared to historical ones. Being aware of reversion to the mean can prevent over-correcting in response to recent events, especially those that were impactful.
Relying on reversion to the mean to explain extreme evets can be misleading where there has been a fundamental change, or in statistical terms, once we start sampling from a different underlying distribution. For example, if a company develops a best-selling new product, its increased growth in the subsequent year might not merely be an anomaly but a new norm.
After a high-impact event, it can be difficult to figure out whether we’re dealing with a new underlying distribution or if we should keep relying on the old and well-established distribution, discarding the event as an outlier. As always, the best strategy is to consider both possibilities.
This post is part of the Encyclopedia of Concepts.
Nehaveigur 