Correlation does not imply causation

Correlation does not imply causation is a phrase used in the sciences and statistics to emphasize that correlation between two variables does not imply there is a cause-and-effect relationship between the two. Its converse, correlation proves causation, is a logical fallacy by which two events that occur together are claimed to have a cause-and-effect relationship. It is also known as cum hoc ergo propter hoc (Latin for "with this, therefore because of this") and false cause. It is subtly different to the fallacy post hoc ergo propter hoc, which in requiring a chronological component may be considered a subtype of cum hoc.

Usage
In the strictest sense, it is always correct to say "Correlation does not imply causation". With casual use of the word "imply" the idea of a causal connection is in some sense true, but that is because the word "implies" can loosely mean suggests rather than requires. And correlation is certainly needed for causation to be proved. However, in logic, the technical use of the word "implies" means


 * to be a sufficient circumstance.

This is the meaning intended by statisticians when they say causation is not certain. Indeed, p implies q has the technical meaning of logical implication: if p then q symbolized as p ⇒ q. That is "if circumstance p is true, then q necessarily follows."

In contrast, the everyday English meaning of "imply" is


 * To indicate or suggest.

To say a "Correlation does not suggest causation" is false: A demonstrably consistent correlation often suggests or increases the probability of some causal relationship (or implies it, in the casual sense of the word).

What the correlation does not do is prove causation, as arguments that use the cum hoc ergo propter hoc logical fallacy as a pattern of reasoning assert.

While it is not the case that correlation is causation, simply stating their nonequivalence omits information about their relationship. Tufte suggests that the shortest true statement that can be made about causality and correlation must be at least expanded to either
 * Empirically observed covariation is a necessary but not sufficient condition for causality.

or
 * Correlation is not causation but it sure is a hint.

Literal logical meaning of material implication
"Correlation does not imply causation" does not mean ~(p ⇒ q), which is equivalent to p & ~q, as in material implication as logicians and statisticians who use the phrase are not trying to say that "we will always have correlation without causation", when considering this the statement should be modal, and not propositional, which is why many people prefer the phrase "Correlation does not necessarily imply causation."

General pattern
The cum hoc ergo propter hoc logical fallacy can be expressed as follows:
 * A occurs in correlation with B.
 * Therefore, A causes B.

In this type of logical fallacy, one makes a premature conclusion about causality after observing only a correlation between two or more factors. Generally, if one factor (A) is observed to only be correlated with another factor (B), it is sometimes taken for granted that A is causing B even when no evidence supports this. This is a logical fallacy because there are at least four other possibilities:


 * 1) B may be the cause of A, or
 * 2) some unknown third factor is actually the cause of the relationship between A and B, or
 * 3) the "relationship" is so complex it can be labelled coincidental (i.e., two events occurring at the same time that have no simple relationship to each other besides the fact that they are occurring at the same time).
 * 4) B may be the cause of A at the same time as A is the cause of B (contradicting that the only relationship between A and B is that A causes B). This describes a self-reinforcing system.

In other words, there can be no conclusion made regarding the existence or the direction of a cause and effect relationship only from the fact that A is correlated with B. Determining whether there is an actual cause and effect relationship requires further investigation, even when the relationship between A and B is statistically significant, a large effect size is observed, or a large part of the variance is explained.

Examples

 * Sleeping with one's shoes on is strongly correlated with waking up with a headache.
 * Therefore, sleeping with one's shoes on causes headache.

The above example commits the correlation-implies-causation fallacy, as it prematurely concludes that sleeping with one's shoes on causes headache. A more plausible explanation is that both are caused by a third factor, in this case alcohol intoxication, which thereby gives rise to a correlation. Thus, this is a case of possibility (2) above.


 * Ice cream sales correlate with the number of people who drown at sea.
 * Therefore, ice cream causes people to drown.

This fallacy concludes that as the number of ice creams sold increases at the same time that a higher number of people drown, there is a causal relationship. In fact, both are caused by a common third factor: Summer.

A recent scientific example:
 * Young children who sleep with the light on are much more likely to develop myopia in later life.

This result of a study at University of Pennsylvania Medical Center was published in the May 13, 1999 issue of Nature and received much coverage at the time in the popular press. However a later study at Ohio State University did not find any link between infants sleeping with the light on and developing myopia but did find a strong link between parental myopia and the development of child myopia and also noted that myopic parents were more likely to leave a light on in their children's bedroom . This is a case of (2).

Another example:
 * Since the 1950s, both the atmospheric CO2 level and crime levels have increased sharply.
 * Hence, atmospheric CO2 causes crime.

The above example arguably makes the mistake of prematurely concluding a causal relationship where the relationship between the variables, if any, is so complex it may be labelled coincidental. The two events have no simple relationship to each other beside the fact that they are occurring at the same time. This is a case of possibility (3) above; another such example is the hoax Mierscheid Law.

A more complex example:
 * Scientific research finds that people who use cannabis (A) have a higher prevalence of psychiatric disorders compared to those who do not (B).

This particular correlation is sometimes used to support the theory that the use of cannabis causes a psychiatric disorder (A is the cause of B). Although this may be possible, we cannot automatically discern a cause and effect relationship from research that has only determined people who use cannabis are more likely to develop a psychiatric disorder. From the same research, it can also be the case that (1.) having the predisposition for a psychiatric disorder causes these individuals to use cannabis (B causes A), OR (2.) it may be the case that in the above study some unknown third factor (e.g., poverty) is the actual cause for there being found a higher number of people (compared to the general public) who both use cannabis and who have been diagnosed as having a psychiatric disorder. Alternatively, it may be that the effects of cannabis are found more pleasureable by persons with certain psychiatric disorders. To assume that A causes B is tempting, but further scientific investigation of the type that can isolate extraneous variables is needed when research has only determined a statistical correlation.

Examples are abundant in political debate surrounding legal issues. For example, there is a correlation between the use of pornography and sex crimes. Individuals who frequently view pornography are more likely to commit sexual offences than those that do not view pornography. Some people point to this as evidence that pornography causes individuals to commit sex crimes, and hence they argue that pornography should be made illegal. Although such arguments are based on a logical fallacy, they can be politically compelling, particularly in highly emotional situations. For example, the correlation between possession of child pornography and paedophilia may be seen as a legitimate rationale for the banning of child pornography. In such a case, it may be deemed appropriate to err on the side of caution. If there is even a chance that child pornography leads to paedophilia, then it may be in the social interest to make its possession illegal.

Pastafarianism, a parody religion founded in 2005, satirically states that there is a correlation between the number of pirates and many natural disasters. Bobby Henderson, the creator of this religion, put forth the argument that:
 * Global warming, earthquakes, hurricanes, and other natural disasters are a direct effect of the shrinking numbers of pirates since the 1800s.

This helps to show that things with statistically significant correlations are not necessarily related.

Determining causation
David Hume argued that causality cannot be perceived (and therefore cannot be known or proven), and instead we can only perceive correlation. However, he argued that we can use the scientific method to rule out false causes.

Intuitively, causation seems to require not just a correlation, but a counterfactual dependence. Suppose that a student performed poorly on a test and guesses that the cause was not studying. To prove this, we think of the counterfactual - the same student writing the same test under the same circumstances but having studied the night before. If we could rewind history, and change only one small thing (making the student study for the exam), then causation could be observed (by comparing version 1 to version 2). Because we cannot rewind history and replay events after making small controlled changes, causation can only be inferred, never exactly known. This is referred to as the Fundamental Problem of Causal Inference - it is impossible to directly observe causal effects.

Well designed statistical studies replace equality of individuals as in the previous example by equality of groups. This is achieved by randomization of the subjects to two or more groups. Although not a perfect system, placing the subjects randomly in the treatment/placebo groups ensures that it is highly likely that the groups are reasonably equal in all relevant aspects. If the treatment has a significantly different effect than the placebo, one can conclude that the treatment is likely to have a causal effect on the disease. This likeliness can be quantified in statistical terms by the P-value.