Wednesday, October 16, 2013

Handy Concept #3: Regression to the Mean

     Intellectual Toolbox Week continues with the engine behind many an otherwise inexplicable disappointment in life.

Regression to the Mean

     Let's say you do something excellent: write a best-selling novel, bowl a perfect 300 game, pitch a no-hitter. Congratulations. What do we know about how your next effort might fare? Because you've already proven yourself as capable of something great, will you probably do even better next time? Or will you likely do worse?
     Sorry to report, the tendency is for you to do worse, because of a nagging little reality known as "regression to the mean." It tells us that when there are averages of something -- most books sell crap, most bowlers bowl under 200, most pitchers give up hits during a game—that any exceptional achievement is by definition a rarity and the next attempt will, probably, bend toward the average, as opposed to away from it. A study of more than 4,000 major league baseball players who had batting averages above .300 in a season found that 80 percent did worse the next season.
    The classic example involves military pilots. When trainees would do well in flying exercises, their superiors noticed that if they were praised for their performance, they would do worse on the next outing. So they started criticizing good performances, until they realized that outstanding flying was followed by lesser results whether the pilot was praised or criticized.  
    There is a comforting corollary to this -- just as superior performance will, on average, be followed by something worse, so a really bad job should be followed by improvement, again on average.
    Bearing this in mind helps us grasp specific examples of falling short, which should not be surprises, but expected.
    “Regression effects are all about us," Amos Tversky and Daneil Kahneman wrote in their influential 1973 paper on the subject.  "Most outstanding fathers have somewhat disappointing sons, brilliant wives have duller husbands, the ill-adjusted tend to adjust and the fortunate are eventually stricken by ill luck. In spite of these encounters, people do not acquire a proper notion of regression. First, they do not expect regression in many situations where it is bound to occur. Second, as any teacher of statistics will attest, a proper notion of regression is extremely difficult to acquire. Third, when people observe regression, they typically invent spurious dynamic explanations for it.” 
     That has to be of some comfort to guys like Frank Sinatra Jr. 


  1. Which is why it's probably not a very good idea for a famous dad to name his son after himself. I always felt it was somewhat narcissistic to do that. While it may open some doors down the line, is the pressure of higher expectations worth it?

  2. How do you explain Rep. Dan Lapinski then, he gets better and better every day?

  3. Easy. Because it wasn't humanly possibly for Lipinsky to get any worse than he was when he showed up. Plus it's up for debate as to whether he has surpassed his dad at what his dad was best at.