Wednesday, June 2, 2010

The writers at the Economix blog should lose their Excel priveleges

... or at least learn the meaning of "post hoc ergo propter hoc".

Yesterday I remarked on how bad it was for Ed Glaeser to even post a plot of change in the unemployment rate against stimulus spending, particularly when he doesn't want to get in the weeds of endogeneity problems. The whole point is that you stimulate economies that are weak and you don't stimulate economies that are strong. Simply plotting stimulus against unemployment tells you exactly nothing about the impact of fiscal policy, and because of endogeneity and the risk of confusing correlation with causation, you risk people underestimating the impact.

Now this morning on the same blog Casey Mulligan blogs on Ricardian equivalence (the theory that people cancel out public borrowing by saving more in anticipation of higher taxes in the future) by plotting private savings against public borrowing. Again, though, this confuses correlation with causation. The government is borrowing now precisely because we're in a paradox of thrift economy. I'm not saying Ricardian equivalence doesn't hold to a certain extent (I guess I'd call it "Ricardian inequivalence", though) - I'm just saying that governments borrow precisely when there is too much saving for a healthy macroeconomy.

Why do dumb blog posts like this get thrown up? I think it's because of how easy it is to download a few data points and make a supposedly informative graph in Excel. Hence my insistence that the writers at the Economix blog should lose their Excel priveleges. These are good economists. They would never make these inferences in journal articles. But when you're explaining it to a wider public, it's easy to plot a few points and do an informal interpretation, glossing over all the fallacies you employ to get to your conclusion.

The fact is, a positive relationship between public borrowing and private savings could potentially be evidence for two different things:
1. On the one hand, it could illustrate Ricardian equivalence, which if it held strictly would largely eviscerate the prospect that fiscal policy is all that helpful, or

2. It could illustrate the reality of the paradox of thrift and a predictable government response to the paradox of thrift.

Well that really helps, doesn't it? (No, it doesn't at all).

If you make a strong theoretical case with lots additional evidence and reasons for believing your case, then simple plots can be helpful illustrations. But if you ever see simple plots like this presented as evidence for anything in economics, ask yourself "in what sense could the causality be reversed". Think that through, consider those alternatives, and always be skeptical of anyone that tries to present a simple graphic as an open and shut case. And finally - keep straight what you know based on the data and what you think you know based on the theory, and what is a mix of both. I don't claim to know that fiscal policy works based on the data. That's a very hard case to make because of the empirical obstacles I mentioned in the previous post. You get yourself in a heap of trouble when you claim to have empirical support for things that you actually don't (or that you have only weak support for).

No comments:

Post a Comment

All anonymous comments will be deleted. Consistent pseudonyms are fine.