Monday, February 11, 2013

Fun with Paul and Brad

David Cushman gives DeLong and d'Okun of it to Krugman's waffling on his famous tiff with Greg Mankiw over how quickly economic growth would be restored after the 2008 unpleasantness;
I guess I should take it as a compliment that people who want to attack my record feel the need to make stuff up. Greg Mankiw posts the abstract of a piece claiming that Brad DeLong and I endorsed the early, optimistic Obama administration forecast for unemployment and growth. What? I was famously pessimistic at the time....
Brad tells me that the author of the piece also doesn’t know whatOkun’s Law is.
Better trolls, please.
Cushman doesn't take that lying down;
Had Professor Krugman read just the first two pages of my paper, he would have learned that I was dealing with not the January CEA forecast and January blog entries by DeLong and Krugman, but rather the February 2009 CEA forecast and March blog entries. 
Then, now that they're on the same page;
Okun’s Law was central to Krugman’s (2009b) discussion of the likelihood of recovery, which followed his approving reference to DeLong (2009). In that entry, DeLong had used Okun’s Law (without explicitly naming it) as the basis of a forecasting model deployed to further his chastisement of Mankiw. But, as I will conclude by showing, the DeLong model’s specification and empirical fit were inferior to, and the model has yielded poorer forecasts than, the vector-autoregression (VAR) model of the hypothetical econometrician in Cushman (2012)—a model that is also related to Okun’s Law. 
Okun's 'Law' (really an estimate of the relationship of changes in GDP to changes in unemployment) would seem to be in competent hands with Cushman;
I derived a VAR that, after consolidating various terms and generalizing to account for the accompanying discussion of serial correlation (Cushman 2012, 318), becomes:
yt − yt−1 = a + (3) kΣi = 1 bi(yt−i − yt−i−1) + kΣi = 1 ciunt−i + ε1, t 
unt = d + (4)kΣi = 1 ei(yt−i − yt−i−1) + kΣi= 1 fiunt−i + ε2, t
This VAR is a model of the relationship between economic growth and unemployment that allows various adjustment patterns through time. It is thus a
dynamic Okun’s Law model. 
What comes out in the wash being;
...the DeLong model forecasts show immediate and complete rebounds from the recession conditions at the end of 2008. For 2013 (the bet year proposed by Mankiw), the “DeLong ’86 start” forecast is in very close agreement with the CEA forecast. And the “DeLong ’50 start” forecast from the full, postwar estimation period that DeLong used is even more optimistic, reflecting its higher long-run growth rate.
This is certainly consistent with DeLong’s belief that the CEA forecast was “certainly the way to bet” (2009). Finally, comparison with the now-known, post-2008 real GDP values shows that the hypothetical econometrician’s forecasts in Cushman (2012), while insufficiently pessimistic, would have proven more accurate than the DeLong model forecasts.
Or, Mankiw was closer to the truth back in the day.  And, maybe Obamanomics wasn't such a great idea after all, if you wanted a robust economy.

(Thanks to Garrett Jones at Econlog)

No comments:

Post a Comment