The Statistics Professor Who Wasted 10 Years

In 1933, Horace Secrist -- a statistics professor at Northwestern University -- published The Triumph of Mediocrity in Business. He had spent a decade collecting mountains of data proving what he believed was a fundamental economic law: businesses inevitably trend toward average performance.

His evidence was meticulous. He grouped firms by initial performance levels and tracked them over time. The initially most profitable firms became less profitable. The initially least profitable firms improved. Mediocrity was triumphing.

Then Harold Hotelling reviewed the book in the Journal of the American Statistical Association and pointed out a devastating problem: Secrist had simply rediscovered regression to the mean.

Hotelling wrote that Secrist's work was equivalent to "proving that tall fathers have shorter sons, and concluding that the human race is converging to a single height."

The variability of profit rates hadn't changed at all. Secrist had selected extreme groups and observed the inevitable statistical regression. He'd spent 10 years documenting an artifact.

---

What Galton Actually Found

Sir Francis Galton first described regression to the mean in 1886 while studying hereditary stature. He observed that:

• Tall parents tend to have children who are shorter than them (closer to average)
• Short parents tend to have children who are taller than them (also closer to average)

He called this "regression towards mediocrity."

The key insight is that this is not a force or a cause. It is a mathematical inevitability that occurs whenever:

1. There is natural variability in measurements
2. A measurement is selected because it is extreme
3. A second measurement is taken

The regression occurs with no intervention, no change in underlying ability, and no causal mechanism. It is pure statistics.

Why it happens: Any extreme measurement is likely the result of the true underlying value plus some random variation in the favorable direction. On a second measurement, the random variation is unlikely to be equally favorable. The measurement moves closer to the true underlying value -- which is closer to the mean.

---

The Flight Instructor Fallacy

Daniel Kahneman describes a conversation with Israeli Air Force flight instructors that perfectly illustrates the attribution error:

The instructors observed that:

• When they praised a cadet after an exceptionally good landing, the next landing was usually worse
• When they punished a cadet after a bad landing, the next landing was usually better

Their conclusion: punishment works, praise is counterproductive.

The reality: Both observations were regression to the mean. An unusually good landing would naturally be followed by a more average one -- whether praised, punished, or ignored. An unusually bad landing would also be followed by a more average one.

The instructors had constructed an entire philosophy of training from a statistical artifact. And because the artifact consistently confirmed their belief (punishment always "worked"), they never questioned it.

This is the regression fallacy: attributing regression to the mean to whatever cause happens to be present.

---

How This Misleads Every Business Decision You Make

Employee performance reviews:

• Your star employee has a mediocre quarter after an exceptional one. You worry about attitude or complacency. It's probably regression.
• Your struggling employee improves after your "tough talk." You credit your leadership. It's probably regression.

Marketing metrics:

• A viral post gets 10x normal engagement. Next post gets normal engagement. "What did I do wrong?" Nothing. Regression.
• Your worst-performing campaign is followed by an average one. "My A/B testing paid off!" Probably regression.

Revenue fluctuations:

• Best month ever followed by an average month: "Something's wrong with the funnel."
• Worst month ever followed by an average month: "My new strategy is working!"
• Both are likely regression, not signal.

"Best practices" cargo culting:

• Company X had explosive growth and published their playbook
• You copy their playbook and get average results
• You blame your execution
• In reality, Company X was measured during a statistical extreme. Their documented practices were what they happened to be doing during an outlier period.

The Sports Illustrated "cover jinx" is the most famous example: athletes featured after exceptional performance tend to perform worse afterward. Fans attribute this to pressure or distraction. It's regression to the mean.

---

The Sophomore Slump Is Just Math

The "sophomore slump" -- where a remarkable first performance is followed by a more average second -- appears everywhere:

• Sports: Rookie of the Year athletes often have less impressive second seasons
• Publishing: First-time bestselling authors struggle with their second book
• Startups: Explosive first-year growth often plateaus in year two
• Products: Viral launch metrics settle into modest adoption curves

In each case, the initial extreme performance attracted attention (that's why we noticed it). The subsequent regression is treated as a mystery requiring explanation.

It isn't a mystery. It's what happens when you select for extreme values and then measure again.

Why the attribution errors persist:

1. Humans are pattern-seeking -- we need causal explanations for everything
2. Regression to the mean is invisible -- there's no "cause" to observe
3. Narrative reasoning (System 1) prefers stories over statistics
4. Confirmation bias -- we remember the cases that fit our causal theory

---

What This Actually Means for You

For evaluating your own performance:

• Don't over-interpret your best or worst periods. Look at longer-term trends.
• After an exceptional result, expect regression and don't panic when it happens.
• After a terrible result, wait before making drastic changes. The situation may regress naturally.

For evaluating others:

• Be skeptical of both praise and blame following extreme performances.
• Use larger sample sizes and longer time horizons before attributing changes to causes.
• When someone "improves" after your feedback, consider whether regression explains the change.

For evaluating tools and strategies:

• If you adopted a new tool during a bad period and things improved, the tool may not be the cause.
• If you changed strategies after a great period and things got worse, the strategy change may not be the cause.
• Controlled experiments with proper baselines are the only way to separate regression from real effects.

For building a business:

• Don't copy "best practices" from companies measured at their peak.
• Don't abandon strategies based on single-period underperformance.
• Understand that regression to the mean makes every business look like it's declining from its best and recovering from its worst -- even when nothing has changed.

That amazing quarter wasn't your genius. That terrible quarter wasn't your fault. The math explains both. And once you see it, you stop making expensive decisions based on statistical noise.

---

Research basis: Galton (1886) original regression to the mean discovery; Secrist (1933) "Triumph of Mediocrity in Business" and Hotelling (1933) review in JASA; Kahneman & Tversky (1973, Psychological Review) on prediction biases; Kahneman (2011) "Thinking, Fast and Slow" flight instructor example; Sturman, Cheramie, & Cashen (2005, J. Applied Psychology) temporal consistency of performance ratings. The mathematical basis of regression to the mean is established statistical theory, not an empirical finding subject to replication.