Consider this classic example reported by psychologist Daniel Kahneman about a seasoned flight instructor observing how praise and criticism affected the performance of his flight cadets. The instructor noted that when he praised a cadet after a particularly clean execution of a flight manoeuvre, that cadet’s performance in the next run invariably worsened. On the other hand, when he shouted at a cadet after a particularly bad flight, that cadet’s performance in the next run improved. The instructor’s conclusion from these observations was that showering praise wasn’t really an effective way of improving performance.
Readers who are in supervisory positions can very easily relate to this phenomenon. After a salesperson wins a big bonus for a particularly stellar performance in one year, s/he generally produces mediocre results in the following year. Similarly, those who are put on “probation” for lousy performance in one year tend to recover in the following year. So, does this imply that praise has a negative impact on performance whereas reprimand affects performance positively?
Not necessarily. The decline after a spectacular performance and improvement after a particularly bad performance may simply be because of a statistical phenomenon called regression towards the mean. In simple terms, it implies that extreme outcomes will generally be followed by less extreme outcomes on subsequent trials, purely for statistical reasons.
Why? Because most outcomes depend on a combination of two factors: input (or effort) and luck (or chance). While one can keep the effort part fairly consistent over time, chance, by definition, varies from one instance to the next. An extremely good outcome needs both high effort and great luck. It is unlikely that one would have the same great luck on subsequent trials. Hence the drop in performance after a particularly impressive stint.
The statistical nature of this effect is often misunderstood. In the 1930s, a famous book titled The Triumph of Mediocrity in Business suggested that all businesses were heading towards mediocrity. This was based on a massive analysis of data which showed that businesses that did particularly well tended to do poorly the following year, while those that did particularly poorly tended to perform better the following year. Although the error in the analysis was later pointed out by several people, the failure to consider and understand this effect continues.
One can observe the phenomenon of regression towards the mean in a number of other fields. It is common for cricketers who have had a great series to fall back to more mediocre performance in the next one. In the US, there is an interesting belief among sportsmen called the Sports Illustrated jinx. It is believed that being featured on the cover page of the popular Sports Illustrated magazine is a kiss of death for sportsmen. Immediately after being featured on the cover, they fail to live up to expectations and their careers go into a decline.
From a statistical standpoint, this phenomenon is simple to explain—athletes are featured on the cover only after they have had some major accomplishments. These accomplishments have to be of an extraordinary nature in order for the athlete to be put on the cover of Sports Illustrated.
Hence, it is no surprise that these accomplishments are difficult to match or surpass in the following months or years as the athlete’s performance will revert back to the mean over time. Regression towards the mean also explains why sequels of blockbuster movies seldom live up to their audiences’ expectations or why mutual funds that outperform their competitors one year fail to repeat that feat in the following years. The phenomenon of regression towards the mean has the potential of manifesting itself in every place where the outcome is at least partially dependent on chance or random events. It could be a student’s scores on entrance tests, the quality of novels written by an author or the effectiveness of decisions made by a manager. After an unusually strong or glaringly weak outcome, things will tend to revert back to the mean.
A common mistake in performance evaluation is to reward teams for “improvement” over the previous year. What usually happens is that the team that performed the worst in a particular year becomes the “most improved team” the following year even if there was no great increase in effort on their part.
Also, the best performing team is likely to lose out on any reward because they end up performing worse even if this performance is still better than that of most other teams.
So, do not be disappointed if you are unable to match last year’s spectacular sales performance or repeat that hole-in-one feat—those outliers are difficult to replicate. Statistical forces are at work, just as they are when they pull you up after a particularly abysmal performance!
Praveen Aggarwal is an associate professor of marketing at the Labowitz School of Business & Economics at the University of Minnesota Duluth and Rajiv Vaidyanathan is a professor of marketing and director of MBA programmes at the University of Minnesota Duluth. Send your comments to firstname.lastname@example.org