Wednesday, April 08, 2009

Correlation v Causation

by Andrew Gillen

The Chronicle of Higher Education did a nice little write up of our new report, Financial Aid in Theory and Practice, which seems to have set off a bit of a firestorm of criticism in the comments.

One recurring claim is that I am confusing correlation with causation. I seem to remember hearing of that before… – oh yes, I’d cover it in the first week of class back when I taught.

The report uses basic econometrics to show a couple of things about the interaction between various financial aid programs and tuition. The thing is, econometrics can virtually never prove causality by itself. Even fancy econometrics with causality in the title (such as granger causality), is probably better thought of as hinting at causality as opposed to proving it. To claim causality, you basically need a theory that provides a reason to think that the one thing causes the other. To sum up, econometrics gives you the correlation, the theory gives you the causation.

Fortunately for me, I cleverly included a DOZEN PAGES OF THEORY, to explain the correlations and back up my claims of causality. If my conclusions are wrong, it is because my theory is wrong, not because I've confused correlation with causation. But to claim that my theory is wrong, you'd probably have to actually read the study to see what the theory is. I guess it's easier for those that don't like my conclusions to just claim I made some basic mistake.

I have no problem with people commenting on the ideas in the study before reading it. I do have a problem with people trying to discredit my work by accusing me of making a mistake that I did not make, presumably because a) they did not actually read the study, or b) they don’t know what they are talking about.

No comments: