test their theories about relations among specific variables, they must use
statistical techniques designed for situations in which other things cannot
be held constant. Fortunately, such techniques exist, although their
application is usually neither simple nor straightforward.
Later in this chapter we provide a discussion of some graphical
techniques for describing data and displaying some of the more obvious
relationships. Further examination of data involves techniques studied in
elementary statistics courses. More advanced courses in econometrics
deal with the array of techniques designed to test economic hypotheses
and to measure economic relations in the complex circumstances in
which economic evidence is often generated.
Correlation Versus Causation
Suppose you want to test your theory’s prediction that “If X increases,
will also increase.” You are looking for a causal relationship from X to
because a change in X is predicted to cause a change in Y. When you look
at the data, suppose you find that X and Y are positively correlated—that
is, when X rises, Y also tends to rise (and vice versa). Is your theory
supported? It might appear that way, but there is a potential problem.
A finding that X and Y are positively correlated means only that X and
tend to move together. This correlation is consistent with the theory that
causes Y, but it is not direct evidence of this causal relationship. The
causality may be in the opposite direction—from Y to X. Or X and Y may