Roger Farmer re-litigates an old controversy about macroeconomic data, concluding
What do we learn from this? Much the same as we learn from the fact that unemployment has a unit root. Just as unemployment can remain persistently high, so GDP can remain persistently below trend. There is no evidence that the economy is self-correcting.
Pointer from Mark Thoma. My comments:
1. Although unemployment has a unit root, Ed Leamer finds that there is some persistence to changes in payroll employment. See his textbook. You might also Google Kling-Leamer-momentum-employment.
2. Unit roots in macro data cause a huge problem. If you don’t correct for them, you get spurious correlation. If you do correct for them, you tend to get noise. That is one reason I call macroeconometrics The Science of Hubris.
3. In stock market returns, econometricians have been able to identify long-term mean reversion even though the short run is a random walk. Can something similar be done with GDP data?
4. If real GDP truly is nonstationary, then how can we rescue the concept of potential GDP? If you think of the economy as ultimately self-correcting, then what it corrects to is potential GDP. If the economy is not self-correcting, then the concept of potential GDP can have no objective basis.
I come to this issue with a desire not to praise the concept of potential GDP, but to bury it. From a PSST perspective, there is no such thing as potential GDP. The patterns of specialization and trade that are sustainable now are the patterns that are sustainable now. There were other patterns that were sustainable in the past, and entrepreneurs will discover still other patterns sustainable in the future, but right now we cannot describe those alternative patterns as potential. Right now those alternative patterns are either not sustainable or yet to be discovered.
Macroeconomist and blogger Stephen Gordon argues that the unit root concept is a zombie that should be put to rest. I don’t think that’ll happen, because unit roots generate far too many papers for too many authors (a gift that keeps on giving), but I see his point. He concludes, “There’s little evidence that we should be operating under the unit-root hypothesis during recessions. Or during expansions, come to that. (And) The world would be a better place if we could all agree that that the whole unit root literature never existed.”
http://worthwhile.typepad.com/worthwhile_canadian_initi/2009/03/the-unitroot-zombie-dead-but-still-wreaking-havoc.html
Tanks for the link, I found his paper interesting.
What is a unit root?
Sounds plausibly phallic.
As if that weren’t bad enough it is often tested for using the augmented Dickey-Fuller test.
Basically it means the regression coefficient in an auto regression model is statistically equal to one. i.e. Yt = a*Yt-1 + sigma; a = 1. therefore Yt ~ N(Yt-1, sigma), or, your best guess for tomorrows value is todays value. Or, random shocks in previous time periods are preserved, they dont decay. Or Brownian motion/random walk. Or you cant really make meaningful structural predictions.
Thank you. That helps. I know more than I did before, which means my knowledge is not a unit root.
“Basically it means the regression coefficient in an auto regression model is statistically equal to one. i.e. Yt = a*Yt-1 + sigma; a = 1. therefore Yt ~ N(Yt-1, sigma)”
Can somebody translate this into English, maybe with some simple pictures, for those of us who wonder if an auto regression model is a Toyota or a Ford?