Correlation Does Not Equal Causation

psychpost | 5/06/2012 08:00:00 AM
Different types of correlation. [image source]
The dependence (or association) between two different variables is known as their correlation. Correlation is a very important tool for finding trends and as a starting point for research. A positive correlation is a relationship in which an increase in one variable accompanies an increase in another. A negative correlation is a relationship in which an increase in one variable is accompanied by a decrease in another (or vice-versa).

A very common mistake when looking at correlational data is that the reader may interpret the results to be causal (X causes Y). Although this conclusion can be very appealing, there are two problems which we encounter.

Problem 1, Direction-of-causation problem: A correlation between two variables does not carry information about which variable causes what effect. When looking at two correlated variables, X and Y, it is possible that X might cause Y or that Y might cause X.

Problem 2, Third-variable problem: The correlation of two variables may actually be the result of a third variable that is not taken into account. A popular example of this third-variable problem has been seen in research about children's IQ scores being positively correlated with mothers breastfeeding them as infants. The problem with this early research is that mothers of higher Socioconomic Status are more likely to breastfeed than women of lower SES. This third variable, Socioeconomic Status, is known to have positive effects on many variables including IQ score (for high SES) and also negative effects (for low SES).

1 response to "Correlation Does Not Equal Causation"

  1. I always have to be suspicious of a weak correlation. When you look at the difference between that and none, it seems like it could be very easily down to chance.

Post a Comment