Consulting by RPM || Free Advice Blog

So Which Are the Predictions That Matter Again…?

Long-time readers know that my day of infamy occurred when I bet David R. Henderson $500 that year/year CPI inflation would break double-digits. (I actually bet Bryan Caplan first, but David’s bet was for a shorter time horizon, so that’s why my bet with David got more attention. I summarize everything here.) When that bet blew up in my face, Brad DeLong and Paul Krugman thought it was the most glorious day since Nixon took us off the gold standard. They were both quite convinced that because I had been wrong in my forecast of (price) inflation, that my underlying model of the economy was wrong, and therefore I needed to reconsider my entire worldview and in particular, my free-market policy recommendations. (All links are provided in the one above.)

So in that context, I tongue-in-cheek asked about the infamous Romer/Bernstein chart from early 2009, which made projections about the unemployment rate with and without the Obama stimulus package. As it turned out, the unemployment rate *with* stimulus turned out to be worse than what they were warning would happen *without* stimulus:

So I asked (again, tongue-in-cheek) why I had to be taken out to the woodshed for *my* bad prediction, but Romer/Bernstein’s bad unemployment prediction didn’t (apparently) affect the case for deficit spending. Here’s what Krugman said at the time:

Robert Murphy replies to Brad DeLong, and DeLong is not happy — for good reason. But I think there’s also a broader point.

Brad’s ire reflects Murphy’s apparent belief that his failed inflation forecast is OK because we just so happen to have faced a huge deflationary downdraft that offset the inflationary impact of Fed expansion. As Brad says…

My broader point, however, involves Murphy’s main argument, which is “Well, some Keynesians got their unemployment predictions wrong, so there.”

What’s wrong with this line of attack? Two things, actually.

First, it’s really important to distinguish between fundamental predictions of a model and predictions that an economist happens to make that don’t really come from the model. The prediction that huge increases in the monetary base will cause large increases in the price level, and that big government deficits will cause big increases in interest rates, are more or less inescapable if your model of the economy is one in which recessions are supply-side problems, not the result of inadequate demand. Conversely, the prediction that neither of these things will happen if the economy is in a liquidity trap is a fundamental prediction of Keynesian models. On the other hand, the unfortunate Romer-Bernstein prediction of a fairly rapid bounceback from recession reflected judgements about future private spending that had nothing much to do with Keynesian fundamentals, and therefore sheds no light on whether those fundamentals are correct.

In short, some predictions matter more than others. [Bold added.]

OK, so at the time I thought that seemed very convenient for Krugman & Co., namely that unemployment forecasts didn’t matter as much as (price) inflation forecasts, but fair enough. Let’s take Krugman at his word, that it was fatal to my worldview that my model predicted big inflation and it didn’t happen.

Now, EPJ links to a recent symposium on rethinking macro, and Paul Krugman’s contribution has this abstract:

This paper argues that when the financial crisis came policy-makers relied on some version of the Hicksian sticky-price IS-LM as their default model; these models were ‘good enough for government work’. While there have been many incremental changes suggested to the DSGE model, there has been no single ‘big new idea’ because the even simpler IS-LM type models were what worked well. In particular, the policy responses based on IS-LM were appropriate. Specifically, these models generated the insights that large budget deficits would not drive up interest rates and, while the economy remained at the zero lower bound, that very large increases in monetary base wouldn’t be inflationary, and that the multiplier on government spending was greater than 1. The one big exception to this satisfactory understanding was in price behaviour. A large output gap was expected to lead to a large fall in inflation, but did not. If new research is necessary, it is on pricing behaviour. While there was a failure to forecast the crisis, it did not come down to a lack of understanding of possible mechanisms, or of a lack of data, but rather through a lack of attention to the right data.

Well isn’t that special. Here, Krugman the academic says his IS/LM model performed great *except* on the one area of inflation forecasts. Specifically, with unemployment so high and real output below potential output, his model going into the crisis predicted accelerating price deflation, yet we actually observed modest price inflation.

Guy reminds me of Nurse Ratched.

(NOTE: I have previously pointed out Krugman’s inconsistency on these matters, but it’s still fun to comment when he writes a paragraph like the above.)