Regression Fallacy

The error of attributing regression to the mean to some causal intervention rather than recognizing it as a statistical inevitability. After an extreme performance (very good or very bad), subsequent performance tends to move toward the average—not because of any intervention but because extreme values are statistically unlikely to repeat. A CEO hired after the worst quarter will likely preside over improvement (regression to the mean), not because of their leadership but because extreme lows are followed by less extreme values. The regression fallacy leads to: praising interventions that coincide with natural regression upward and blaming interventions that coincide with natural regression downward.

When to use it

When evaluating interventions that follow extreme events; when 'crackdowns' seem to work (they coincide with natural regression); when performance improvement after a low point is being attributed to a specific cause; when any action taken at an extreme point seems to produce results.

How it can help

Whenever an intervention follows an extreme event and performance changes: ask 'would this change have happened without the intervention?' If someone was selected for remedial training after their worst performance, improvement is expected statistically—the training gets undeserved credit. If a team was restructured after an unusually bad quarter, the next quarter would likely improve anyway. The correction: compare the intervention group's regression with a control group's regression. Without a control, you cannot attribute change to intervention rather than regression.