« Cowen and Sawicky on WSJ Econoblog | Main | ECB raises interest rate »
November 30, 2005
Real GDP growth: How did 2005Q3 rank?
Is 4.3% growth of real GDP just ok, or is it _______ (use whatever superlative you like)? One of my frequent commenters, spencer, writes in response to my earlier post that since WWII real GDP growth has exceeded 5% almost a third of the time. Against that record, today's news sounds quite average.
But of course the volatility of real GDP has declined significantly since WWII. Here's the picture:
The words "structural break" come to mind.
Chang-Jin Kim and Charles Nelson were thinking along those lines when they wrote their paper "Has the U.S. Economy Become More Stable? A Bayesian Approach Based on a Markov-Switching Model of the Business Cycle" in the 1999 Review of Economics and Statistics. Drawing on work done by fellow blogger James Hamilton (1989 Econometrica), they find a break at 1984Q1. (Links are to the appropriate papers on JSTOR for those who have access.)
Bottom line: If you want to grade the current economy on a curve, don't use the 1950s to construct the grading scale (or anything before 1984, for that matter).
Interestingly, David Tufte (VoluntaryXchange) constructs a grading scale for GDP based on post 1983 data. I would contend that "eyeball econometrics" should lead you to believe that the relevant data to use to construct your grading scale is somewhere between 1983 and 1985. Kim and Nelson's result happens to fall right smack in the middle of that range, so I'm ok with that.
So the next step is to look at a histogram of the data from 1984 onward. Here you go:
This does not include today's data release, but you can see where it would go. Grading on a curve, it's a solid "B". Tufte concurs. Tufte's scale using 1983 as a starting point indicates that the last quarter falls in at the 69th percentile. Using Kim and Nelson's 1984 break point makes things look a little better. The last quarter comes in at the 73rd percentile of GDP growth since 1984.
Of course, the structural break model is just that, a model. The variance of GDP may change more than once. We haven't even addressed what might be causing the change. In general, I'm reluctant to grade a time series on a "curve" like this if there is heteroskedasticity (changing variance) such as this.
That said, I am very sympathetic to the reason why you might want to make the comparison. Thus, I am quite happy to give this report a solid "B" as long as everyone understands that a nice long string of "B"s and "C"s without any "F"s is a pretty good achievement indeed. Stable, sustainable growth gets a "B" or a "C" on this scale, but stable and sustainable growth is exactly what we want.
UPDATE: See the comments for an addendum about per capita GDP.
UPDATE: David Tufte responds by calculating presidential GPAs. For fun, try to guess Clinton's and Bush's GPAs before clicking over!
Posted by William Polley at November 30, 2005 10:23 PM
Trackback Pings
TrackBack URL for this entry:
http://www.williampolley.com/cgi-bin/mt-tb.cgi/442
Comments
Good point. By only looking at the post 1984 period the frequency of the times growth is over 5% is roughly cut in half from 31% to 15%. The share of negtive growth also drops from some 15% to 5.7%.
I'm well aware of the fact that growth has been more stable since the early 1980s and should have considered that in my analysis.
But the average also drops from 3.5% to 3.26%.
Prior to 1984 growth average 3.63%.
Posted by: spencer at December 1, 2005 10:45 AM
I don't know if I am repeating, my first post appears to have failed.
Good point. If you only look at the post 1984 period the tails of the distribution are cut sharply. The share of over 5% drop from 31% to 15% and the share under zero drops from 15% to 5%.
I'm well aware of the shift to a more stable environment and should have considered it.
But the average growth rate does fall from 3.5%
to 3.26%. Prior to 1984 the average was 3.63%.
Posted by: spencer at December 1, 2005 10:51 AM
Sometimes you have to reload the page to see your comment.
On the average growth rate:
Population growth was higher in the 1950s. Per capital real GDP growth in the two periods is much closer. About 0.04% lower in the more recent period. (Per capita average growth of 2.14% vs 2.10%)
I used population data up to 2004 from the Economic Report of the President and averaged the quarterly growth rates before subtracting population growth.
I'm not sure I'd make too much out of that difference. After 2005 data gets added in, it will probably be about the same.
Posted by: William Polley at December 1, 2005 11:32 AM
How much of the current gdp growth accrues to 'financial' that is not apparent in those pre 80? numbers? Can we see this sector as continuing to gain share in this gdp growth? Is it possible to view this as pulling forward gdp growth should the 'financial' side take a well-deserved rest?
Posted by: calmo at December 1, 2005 2:55 PM
Not just the financial sector but services in general have seen their share of GDP increase. Yet the growth rate overall has stayed about the same. It's long run reallocation of resources that has been taking place slowly but surely. I'm not sure what you mean by your last question. Why should the financial side (or services in general, for that matter) take a rest?
Posted by: William Polley at December 2, 2005 3:28 PM
With regard to Spencer's point, picking up on William's reply:
There is the lower and more stable rates of labor force growth since 1981 or so, after the Boomers had gotten absorbed into the labor market. Growth in GDP/worker might be a better measure for time-varying growth rates. Even better might be TFP, if we can get a handle on measurement error. Then maybe an even more neutral measure of productivity growth including things like human capital and infrastructure.
Oddly enough, after taking these things into account, the time-varying heteroskedasticy seems to get more pronounced even as the time-varying means get smoothed out a bit more. Pretty strange. Anyhow, instead of grading presidents on things outside of their control, I'd ask what could be done to improve fiscal policy, trade policy, and regulatory policy, the things that the president has control over. After all, the BEA will be issuing regrades for years to come.
Posted by: Chris R at December 9, 2005 12:54 AM