Causality vs correlation....

When we look at economic outcomes, we observe correlations: event "x" happened along with event "y". There are several possible interpretations of this. First, "x" may have caused "y". Alternatively, "y" may have caused "x". Finally, some other variable "z" may have caused both "x" and "y". Without knowing the causality involved, we cannot answer most relevant policy questions available data.

One specific example from my own research: we observe that members of currency unions trade a lot more than other countries. That's the correlation. If being a member of a currency union induces people to trade more, this might make a compelling case for countries to form currency unions. But if countries trading more tend to form currency unions, then the correlation is reversed, and countries may not see an increase in trade from forming a currency union.

Economists spend a lot of time trying to find "natural experiments," where we can be sure of the direction of any possible causality. Often this is based on an unexpected change in "x" which was clearly exogenous- i.e. was not caused by "y".

One interesting application is to the effects of female leadership rather than male leadership. Read this slate article for more. (Thanks to Marginal Revolution for the link).