have no direct causal connection; their movements may instead be jointly
caused by movements in some third variable, Z.
Here is an example. Suppose your theory predicts that individuals who
get more education will earn higher incomes as a result—the causality in
this theory runs from education to income. In the data, suppose we find
that education and income are positively correlated (as they are). This
should not, however, be taken as direct evidence for the causal prediction
that more education causes higher income. The data are certainly
consistent with that theory, but they are also consistent with others. For
example, individuals who grow up in higher-income households may
“buy” more education, just as they buy more clothes or entertainment. In
this case, income causes education, rather than the other way around.
Another possibility is that education and income are positively correlated
because the personal characteristics that lead people to become more
educated—ability and motivation—are the same characteristics that lead
to high incomes. In this case, the causal relationship runs from personal
characteristics to both income and education.
Most economic predictions involve causality. Economists must take care when testing
predictions to distinguish between correlation and causation. Correlation can establish that
the data are consistent with the theory; establishing causation usually requires more advanced
statistical techniques.