Causation and Correlation
Causation and Correlation
The Classic Trap
Ice cream sales and drowning deaths: Both increase together. Naive conclusion: ice cream causes drowning. Reality: both are caused by a third factor — warmer weather. Hot days → more ice cream AND more swimming → more drowning.
This is the confounding variable problem: two things moving together doesn't mean one causes the other.
Common Patterns
| Pattern | Looks like | Actually |
|---|---|---|
| Confounding | A correlates with B | Third variable C causes both |
| Reverse causation | A causes B | B actually causes A |
| Coincidence | A correlates with B | Random noise, no relationship |
| Mediated | A causes B directly | A causes C which causes B (indirect) |
Why It Matters
Business: "We ran ads and sales went up" — did ads cause sales, or was it seasonal demand? Wrong attribution wastes budgets.
Health: "People who eat X live longer" — do they live longer because of X, or because the type of person who eats X also exercises, sleeps well, and has better healthcare access?
Policy: "Countries with more police have more crime" — policing doesn't cause crime; high-crime areas get more police (reverse causation).
The Fix
Randomized controlled trials (RCTs) are the gold standard — randomly assign treatment to isolate causation. When RCTs aren't possible: look for natural experiments, control for confounders, and always ask "what's the mechanism?"
Most business and life decisions are made on correlations mistaken for causation. The mental habit of asking "what else could explain this?" catches 80% of bad reasoning before it turns into bad action.