False Causality: Why Correlation Is the World's Most Dangerous Two-Card Trick

Countries with higher chocolate consumption per capita have more Nobel Prize winners per million inhabitants. The correlation is real — the data is in the charts. So... does chocolate make you smarter? This is the delightfully absurd end of a very serious reasoning error: false causality, the assumption that because two things are correlated, one must cause the other. It's the cognitive trap behind some of history's most consequential mistakes.

Post Hoc Ergo Propter Hoc

The Latin phrase post hoc ergo propter hoc — "after this, therefore because of this" — names the oldest and most intuitive form of false causality. I took vitamin C tablets and my cold got better; therefore the vitamin C cured my cold. I wore red socks on the day we won the match; my red socks bring luck.

Temporal sequence is powerful. It's actually the foundation of rational causal inference — causes do precede effects. But temporal sequence alone doesn't establish causation. The cold would probably have gotten better anyway. The team might have won regardless of the socks.

Aristotle listed this as a classic logical fallacy. David Hume made it the centerpiece of his analysis of causation in the Treatise of Human Nature (1739): we never observe causation directly, only constant conjunction and temporal sequence. We infer causation; we don't perceive it. This inferential gap is where false causality lives.

Correlation Is Not Causation — But Why Not?

When two variables, A and B, are correlated, there are several possible explanations:

1. A causes B (the assumed explanation)
2. B causes A (reverse causation)
3. C causes both A and B (confounding variable)
4. Pure chance (especially with small samples or data dredging)

The chocolate-Nobel-Prize correlation almost certainly reflects a confounding variable: national wealth. Rich countries both consume more chocolate and invest more in the education and institutions that produce Nobel laureates. Chocolate didn't cause the Nobels; GDP caused both the chocolate consumption and the research infrastructure.

Franz Messerli's 2012 paper in the New England Journal of Medicine presented this correlation with deadpan irony — and yet countless readers shared it seriously. The lesson: even obviously spurious correlations find credulous audiences when they're dressed in data.

Famous Cases of False Causality

Vaccines and autism: A 1998 paper by Andrew Wakefield (later retracted and stripped of credibility) suggested a link between the MMR vaccine and autism. Parents noticed their children were diagnosed with autism around the same age they received vaccines. The temporal sequence seemed compelling. But large-scale epidemiological studies involving millions of children found no causal link. The apparent connection was confounded by the fact that autism symptoms become noticeable around the same developmental age as vaccination schedules — not because of vaccines.

The stork-birth rate correlation: Across European countries, the number of storks and the birth rate are positively correlated. Obviously, storks don't bring babies — but both correlate with rural population density and traditional agricultural communities, which have both more storks (open farmland) and higher birth rates.

Ice cream and drowning: Ice cream sales and drowning rates peak simultaneously — in summer. The confound is heat and water. Neither causes the other.

Nicolas Cage films and pool drownings: Tyler Vigen's satirical "Spurious Correlations" website found that Nicolas Cage film releases in a given year strongly correlate with pool drowning deaths that year. Both trend with population size and economic cycles. This is a statistical artifact — the kind that emerges readily when large datasets are mined for correlations without prior hypotheses.

Smoking and lung cancer (delayed): In the 1950s, the tobacco industry actively exploited the difficulty of proving causation to deny the link between smoking and lung cancer. The correlation was clear; establishing causation required decades of epidemiological work, animal studies, and mechanistic research. False causality arguments ("it could just be correlation") were used as shields against genuinely causal evidence.

The Science of Establishing Causation

Epidemiologist Austin Bradford Hill (1965) proposed a set of criteria for moving from correlation to causation that became foundational in public health research. The Bradford Hill criteria include:

• Strength: Stronger associations are more likely causal
• Consistency: Replicated across different populations and contexts
• Specificity: The cause produces a specific effect, not everything
• Temporality: Cause precedes effect (necessary but not sufficient)
• Biological gradient: Dose-response relationship (more exposure = more effect)
• Plausibility: Consistent with known biological mechanisms
• Coherence: Doesn't contradict established knowledge
• Experiment: Evidence from controlled experiments

No single criterion is definitive; causation is a judgment call based on the totality of evidence. This is why establishing causal claims is genuinely hard — and why false causality is so persistent.

Confounding Variables and Why They're So Hard to Spot

A confounding variable is one that correlates with both the proposed cause and the proposed effect, creating the illusion of a direct relationship. Confounders are the invisible third parties in spurious correlations.

Socioeconomic status is one of the most pervasive confounders in social science research. Studies correlating any two socially patterned behaviors — income, education, diet, health, crime — have to control for SES or risk finding spurious links. Many headline-grabbing findings that "X causes Y" collapse when SES is properly controlled.

False Causality in Policy

Government policy is perhaps the highest-stakes arena for this fallacy. Programs implemented because of correlations — rather than established causal relationships — have wasted vast resources and sometimes caused harm. The "broken windows" policing theory (Wilson & Kelling, 1982) held that visible disorder causes serious crime. Decades of aggressive policing followed. Later research suggested the relationship was far more complex and the causal claims overstated.

Distinguishing correlation from causation is not just an academic exercise. It determines where we spend money, what we treat, who we imprison, and what we teach our children.

How to Think More Clearly About Causation

• Always ask: could a third variable explain both?
• Could the causation run the other way?
• Is there a plausible mechanism — not just a pattern?
• Has the finding been replicated in independent samples?
• Beware of data-mining: correlations multiply when you look at enough variables

And remember: temporal sequence is necessary but not sufficient. Just because B followed A doesn't mean A caused B. Millions of things happen every day before every other thing.

References

• Hume, D. (1739). A Treatise of Human Nature. (Analysis of causation as habit of mind.)
• Bradford Hill, A. (1965). "The Environment and Disease: Association or Causation?" Proceedings of the Royal Society of Medicine, 58(5), 295–300.
• Messerli, F. H. (2012). "Chocolate Consumption, Cognitive Function, and Nobel Laureates." New England Journal of Medicine, 367, 1562–1564.
• Wakefield, A. J. et al. (1998, retracted). "Ileal-lymphoid-nodular hyperplasia, non-specific colitis, and pervasive developmental disorder in children." The Lancet.
• Vigen, T. (2015). Spurious Correlations. Hachette Books.
• Wilson, J. Q., & Kelling, G. L. (1982). "Broken Windows." The Atlantic Monthly, March 1982.