Saturday, May 4, 2013

Complete Crowding-Out or Near Identities: Cato Statistical Silliness

Via David Silby comes a really dumb argument by Cato’s Chris Edwards:
The chart below shows the Excel plot of the results. The downward slope of Excel’s fitted trend line means that higher government spending growth in a year corresponds to reduced private GDP growth that year. For example, if real government spending growth was zero, private GDP would be expected to grow at 4.2 percent. If real government spending growth was 5 percent, private GDP growth would be expected to fall to 2.8 percent.
Aha – statistical proof that fiscal stimulus crowds-out private spending. Oh wait, what did you say David:
What’s really happening, of course, is that during recessions, government spending goes up because of unemployment insurance and welfare programs in general get more (unfortunately) customers. It’s not that government spending knocks down private GDP, it’s that government spending tends to go up when GDP is shrinking.
I was alerted to this Cato craziness by a friend who must have posted this somewhere:
Chris Edwards (Cato) needs to be careful with regressions that only serve to validate identities. Think of the following Y (or GDP) = P (private spending) + G (government purchases. May I rewrite this? P = Y - G. OK if good economic policy keeps the variation in Y limited, then there has to be a negative correlation between P and G - by definition. This does not prove crowding-out all.
I know the folks at Cato what a smaller government but stupid statistics is just embarrassing.

1 comment:

Anonymous said...

"Dear colleague, you regressed an accounting identity. Anyway, that identity is wrong".