Correlation is not causation

"Correlation is not causation" is one of the unofficial mottos of modern science. When we observe that two things generally or always happen together, it is natural to wonder whether one may be causing the other. A correlation between two variables does not always mean that one variable has caused a change in the other. Often a third variable is causing both, such as shark attacks and ice cream sales. These two factors may correlate, but they are not directly related. Rather, both go up each summer as people go swimming at ocean beaches.

It is a logical fallacy to assume that correlation implies causation. Correlation does not prove causality, but non-correlation proves non-causality.[1]

The following is a flawed argument:

1. Event A occurs in concurrence with Event B
2. Therefore, Event A causes Event B

This is flawed because:

1. There may be a third, confounding variable C which causes A and B
2. B may cause A
3. The relationship may be a complete coincidence

Examples

1.

Hypothesis: As sales of ice cream rise, so do reports of shark attacks.

Conclusion: High sales of ice cream cause shark attacks.

Flaw: Shark attacks and ice cream sales follow a seasonal pattern. Both rise in the summer months. The seasons are a confounding variable.

2.

Hypothesis: The more people convert to Christianity, the more powerful God becomes.

Conclusion: God gains power from his followers.

Flaw: God is already all-powerful. His magnificence is what draws people to conversion. B causes A.

3.

Hypothesis: The construction of a stadium began in Alaska, and the value of your home in Texas has risen.

Conclusion: Stadium construction in Alaska causes higher home values in Texas.

Flaw: This is a coincidence between unrelated variables.

Correlation vs. Causation in Science

The difference between science and superstition lies in the ways an idea is tested. Scientists test a theory by drawing conclusions from it, hypotheses which can be confirmed or contradicted by observation.