Confounding Variables

Confounding variables are those that affect other variables in a way that produces spurious or distorted associations between two variables. They confound the "true" relationship between two variables.

For example, if we have an association between two variables (X and Y), and that association is due entirely to the fact that both X and Y are affected by a third variable (Z), then we would say that the association between X and Y is spurious and that it is a result of the effect of a confounding variable (Z). Of course, Z may be an confounding variable when it comes to this particular relationship, but it might not be for other relationships.

Confounding variables also can affect two variables that do have some causal connection. For example, if X and Y are associated and also causally related (for example, if X affects Y), the association between X and Y may reflect not only their causal connection but also the influence of a third variable (Z) that affects both of them. Thus, the association between X and Y may exaggerate the causal effect of X and Y because the association is inflated by the effect of Z on both X and Y. In this case, we could say that the relationship between X and Y is confounded by Z, even though it is not a purely spurious relationship.

Not all researchers use these terms or use them in exactly the way that we have defined them here. Some use the term "spurious variable" or "extraneous variable" to refer to a variable that produces a purely spurious association between two other variables. The term confounding variable sometimes is used more narrowly to refer only to the second example that we discuss above, where a causal connection between X and Y is distorted by the effects of a third variable. Regardless of the terminology used, the possible confounding effects of a third variable are recognized by everyone. Thus, we must always remember that the simple bivariate association between two variables can be quite unrepresentative of the true causal connection between the variables.