September 07, 2003

Measuring Economic Growth

At least three people have written emails saying:

Dear Professor De Long,

I am curious about your reaction to the following article published in the FT on Friday 9/5. The author makes the point that a big fraction of reported GDP growth is an accounting mirage due to using "hedonic pricing". Could you please discuss this. Thanks a lot.

The article is by Kurt Richebacher, former chief economist for the Dresdner Bank. His main point is:

Investment in computers soared by $38.4bn, or 12 per cent, from $319.1bn to $357.5bn. The trouble is that much of this boom-like increase in computer investment never occurred. The apparent surge is a consequence of the hedonic deflator that US government statisticians use when measuring computer output and investment... pricing produced $32.1bn of GDP in real terms, about 43.9 per cent of the reported second-quarter GDP increase of $73.1bn. In its absence, GDP would have grown a mere $41bn, implying a growth rate of 1.68 per cent. The important thing about hedonic pricing is that it measures dollars that nobody pays and nobody receives...

Briefly, I think Richebacher is mistaken: unless I am very confused, increased demand by businesses for computers accounted for only one-tenth of the 3.1% annual rate of growth in the American economy in the second quarter--not for half.

I think Richebacher is 100% right conceptually, but is wrong empirically: that is, his main criticism is a good point--but the Commerce Department's Bureau of Economc Analysis agrees that it is a good point, and half a decade ago shifted the way it computes real GDP growth in order to eliminate the problem Richebacher identifies. I think Richebacher is tripped up by the details of chain-weighted real national accounts. Richebacher's is a natural mistake to make: no mere human can follow all the technical changes in their national accounts that all OECD governments' statistical services are making all their time.

As I understand what's going on: Nominal computer investment jumped by 8% in the second quarter (i.e., at a 32% annual rate neglecting compounding), computer prices fell by 4% in the second quarter (which seems reasonable) (i.e., at a 16% annual rate), thus real computer investment jumped by 12% in the second quarter (i.e., at a 48% annual rate). Our current numbers are chain-weighted. That means that you calculate the aggregate growth rate by weighting the growth of each component by its *current* expenditure share. The current expenditure share of computer investment is about 0.7%. So computers are responsible for an 0.08% jump in the level of real GDP in the second quarter--responsible for 0.3% of the annualized real GDP 3.1% annual growth rate.

Richebacher, however, thinks that the weight computer investment gets in calculating the aggregate growth rate is the base year price expenditure share. When measured at 1996 prices, computer investment turns out to be not 0.7% but instead 3.2% of real GDP (that's the $357 billion number Richebacher talks about): the price of computers has fallen astronomically over the past decade. Multiplying that 3.2% expenditure share by the 12% jump and annualizing it, you get Richebacher's conclusion that computer investment is responsible for 1.6% of the annualized 3.1% second quarter real GDP growth rate.

It is this that Richebacher objects to: the hedonics and other procedures that have identified falling computer prices since 1996 lead, he thinks, to a vast overweighting of computer investment in today's national accounts.

Richebacher would be right if the BEA had not switched to chain-weighting. If the BEA were still using the old Laspeyres indices and base-year prices it used up until the second half of the 1990s, his critique would be correct. Indeed, his critique is the main reason the BEA switched to chain-weighting: that rapid changes in the prices of investment goods meant that the old Laspeyres-style base-year-price weighted national account measures were producing gross distortions of reality. The BEA is trying to do two things. First, it is trying to report the impact of the extraordinary fall in computer prices on the magnitude of the computer investment currently going on in America: American businesses are spending about $70 billion buying computers, but that is buying as much in the way of computer power as $350 billion would have bought back in 1996. Second, it is trying to calculate overall real GDP growth giving each component of production and demand an appropriate weight: although we are investing a huge and rapidly-growing amount in computers, that is not an overwhelmingly important source of economic growth: you see, the falling prices of computers mean that even a huge amount of computer investment is not that big a deal as far as its value relative to the other sources of value in the economy are concerned. It's a tricky task--to try to convey these two messages in one set of national accounts--and I think the BEA is walking its tightrope very well.

Thus the BEA doesn't do what Richebacher thinks it does when it calculates real GDP growth. Because the BEA uses chain weights, you can't add up the changes in the reported real values of the components and get the change in the real value of the total because the calculated change in the total uses not the base-year-price but the current-price expenditure share. Growth in computer investment is responsible for only one-tenth--not half--of second quarter growth.

So I believe he's just wrong: the theoretical point is a good and completely correct one, but the Commerce Department's BEA took account of it and changed its procedures five years ago.

America's recovery is not what it seems

By Kurt Richebacher

FT.com site; Sep 04, 2003

American economic recovery is the world's great hope. So there was widespread satisfaction when the US Commerce Department last week revised its estimate of annual growth in the second quarter to 3.1 per cent, up from an earlier estimate of 2.4 per cent. A closer look at the numbers, however, tells a somewhat gloomier story.

Much of the growth in gross domestic product is due to a big jump in defence spending, which provided $40.6bn of the reported GDP growth of $73.1bn. Yet while the commerce department and some media reports noted that defence spending had risen by 44 per cent at an annual rate, many commentators failed to mention its role in raising GDP growth. All too often nowadays, new economic data are examined with a single question in mind: are they better than expected? Close analysis with a longer-term perspective is in short supply.

Still, the commerce department report seemed to contain good news on capital investment. "Real non- residential fixed investment increased 6.9 per cent in the second quarter, in contrast to a decrease of 4.4 per cent in the first," it said. "Non-residential structures increased 4.8 per cent, in contrast to a decrease of 2.9 per cent. Equipment and software increased by 7.5 per cent, in contrast to a decrease of 4.8 per cent. Real residential fixed investment increased 6 per cent, compared with an increase of 10.1 per cent." These numbers suggest a vigorous turn in capital spending. However, their strength owes a lot to the singular US habit of annualising many figures. Quarterly data, in other words, are about four times the reality that would be reported in other countries.

So what exactly did happen in the private sector? Consumption increased by $62.5bn and business fixed investment - vital for a robust, self-sustaining recovery - by $22.4bn.

A single component of business fixed investment accounted for more than its overall increase. Investment in computers soared by $38.4bn, or 12 per cent, from $319.1bn to $357.5bn.

The trouble is that much of this boom-like increase in computer investment never occurred. The apparent surge is a consequence of the hedonic deflator that US government statisticians use when measuring computer output and investment. The aim is to capture quality improvements by calculating how much it would have cost in 1996 to buy a computer of equivalent power to today's machines.

Measured in current dollars, however, this spending rose a lacklustre $6.3bn, from $76.3bn in the previous quarter to $82.6bn - far below previous peak levels. In other words, hedonic pricing produced $32.1bn of GDP in real terms, about 43.9 per cent of the reported second-quarter GDP increase of $73.1bn. In its absence, GDP would have grown a mere $41bn, implying a growth rate of 1.68 per cent. The important thing about hedonic pricing is that it measures dollars that nobody pays and nobody receives. And it grossly distorts international comparisons.

Recessions and times of slow growth are when businesses and consumers normally retrench, correcting boom-time spending excesses. A full-scale recovery, in turn, requires the economy's return to a sustainable pattern of consumption, investment and saving.

None of these adjustments is happening in the US. Alan Greenspan, the Federal Reserve chairman, has been fighting the consequences of excessive monetary looseness with still more of the same. Inevitably, economic and financial fundamentals, such as profits, national savings, debt levels, balance sheets and the trade deficit, continue to deteriorate across the board.

Profits, as calculated in the GDP and income accounts, are down again. Total before-tax profits were $591.5bn in the second quarter, compared with $621.6bn in the first quarter. Non- financial profits were $359.2bn, after $391.3bn in the first quarter. As div- idends are rising while profits fall, credit-financed dividends have soared in the non-financial sector to $102.8bn.

The private sector's strength derived completely from bubble-driven consumer spending. But that bubble has been pricked by the sharp rise in long-term interest rates. Mortgage re- financing activity is falling sharply.

Mr Greenspan has turned bubble creation into a virtue. In days of yore, asset prices were viewed as a by- product of underlying economic conditions. Now, by contrast, the real economy is heavily dependent on asset inflation to fuel borrowing and spending. These conditions are more favourable to recession than to recovery.

The writer is a former chief economist of Dresdner Bank de

Posted by DeLong at September 7, 2003 08:44 AM | TrackBack

Comments

a related subject is the role of technology hardware in productivity growth over the last decade or so. in a pretty good overview, the CBO concluded, "the production of computer hardware has accounted for a major share of the acceleration in [productivity] growth during the latter half of the 1990s." their piece also has a discussion of price metrics and an appendix on hedonic techniques, not to mention lots of good links to other papers by practitioners. good reading for those interested in the effects of computer hardware on economic measures.

see http://www.cbo.gov/showdoc.cfm?index=3448

Posted by: wcw on September 7, 2003 10:23 AM

Stephen Roach has complained for years that as the service sector becomes more significant than manufacturing there is a upward bias to productivity growth. Service workers may be putting in far more hours than measured. We have a medical practice and fine technical equipment, and spend about 70 hours a week working. Pleasing work, but never 35 or 40 hours. We also would not record such hours for any survey. How then do you measure our productivity?

Posted by: jd on September 7, 2003 10:56 AM

Now why wouldn't you record your hours when the Bureau of Labor Statistics surveyors came around?

Posted by: Brad DeLong on September 7, 2003 11:41 AM

"Richebacher's is a natural mistake to make: no mere human can follow all the technical changes in their national accounts that all OECD governments' statistical services are making all their time."

Methinks you are being too charitable Brad. Richebacher could and should have checked all this out before going to press. What we have here is a silly, and widespread, prejudice against 'hedonic prices'. Hmmmmmm.

Also note this:"Alan Greenspan, the Federal Reserve chairman, has been fighting the consequences of excessive monetary looseness with still more of the same. Inevitably, economic and financial fundamentals, such as profits, national savings, debt levels, balance sheets and the trade deficit, continue to deteriorate across the board............ Mr Greenspan has turned bubble creation into a virtue"

I smell Austrianism. If only all the simplistic solutions worked.

Posted by: Edward Hugh on September 7, 2003 12:52 PM

When comparing different countries, how uniform are the application of hedonic factors? Isn't it here a rather large susceptibility to who does the application?

DSW

Posted by: Antoni Jaume on September 7, 2003 01:25 PM

When we are surveyed or asked about our practice other than by trusted friends, we answer in such a way as to "keep" what we hope to be a competitive edge. We have ample reason to believe others follow the same patttern. We learned early on how compeitive others were, though the competitiveness is masked. Women must be especially carefully to protect a practice, for even in our enlightened time women are not easily part of the "club." Yes, it is still so and was so from the beginning. No complaints, we simply decided to be as competitive as the guys while also never admitting a whit.

Again, no complaints. We try to be realists. Should we start playing golf?

Posted by: jd on September 7, 2003 01:53 PM

There is a steady drum roll from bond and gold merchants - former bankers - about excess liquidity being created by the Federal Reserve. "Inflation is all about if only we knew how to properly measure inflation." Since I have never been able to figure out what Austrianism is about, I will pass on the school. The comment really takes the point that a slowing of the economy that is caused by structural problems should not be solved by lowering interest rates but allowing a weak economy to gradually build a new competitive structure. "No pain, no gain." I think we deserve as little pain as possible and the Fed actions since January 2001 were all called for. I have often argued that Japan has used fiscal policy "fairly well" in trying to stimulate the economy. I also wish no pain on the Japanese.

Posted by: anne on September 7, 2003 02:11 PM

Edward,

Richebacher _is_ an Austrian. Why the anti-Austrian sentiment?

Posted by: Kyle Markley on September 7, 2003 04:07 PM

"I smell Austrianism" is pretty funny out of context.

What is the Austrian school, exactly. What makes them so heterodox?

Posted by: Walt Pohl on September 7, 2003 04:20 PM

"...the price of computers has fallen astronomically over the past decade."

Er, have you been reading too much G W Bush?

Posted by: Keith M Ellis on September 7, 2003 09:37 PM

Walt,

I'm not an economist by profession, but I've read a fair amount of Austrian economics, and from what I can tell most of the difference between Austrians and everyone else is disagreement over methodology. In particular, Austrians favor economic principles and deductive logic over empirical methods. They tend to be much more laissez-faire than others.

I've seen comments to the effect that Austrians and other economists seldom understand each other -- I wish I knew why, but it's definitely true. For instance, Brad's comments on the gold standard (http://www.j-bradford-delong.net/Politics/whynotthegoldstandard.html) are totally unconvincing to an Austrian, they don't even recognize the Austrian pro-gold arguments.

Posted by: Kyle Markley on September 7, 2003 10:34 PM

Walt,

I'm not an economist by profession, but I've read a fair amount of Austrian economics, and from what I can tell most of the difference between Austrians and everyone else is disagreement over methodology. In particular, Austrians favor economic principles and deductive logic over empirical methods. They tend to be much more laissez-faire than others.

I've seen comments to the effect that Austrians and other economists seldom understand each other -- I wish I knew why, but it's definitely true. For instance, Brad's comments on the gold standard (http://www.j-bradford-delong.net/Politics/whynotthegoldstandard.html) are totally unconvincing to an Austrian, they don't even recognize the Austrian pro-gold arguments.

Posted by: Kyle Markley on September 7, 2003 11:12 PM

Walt,

I'm not an economist by profession, but I've read a fair amount of Austrian economics, and from what I can tell most of the difference between Austrians and everyone else is disagreement over methodology. In particular, Austrians favor economic principles and deductive logic over empirical methods. They tend to be much more laissez-faire than others.

I've seen comments to the effect that Austrians and other economists seldom understand each other -- I wish I knew why, but it's definitely true. For instance, Brad's comments on the gold standard (http://www.j-bradford-delong.net/Politics/whynotthegoldstandard.html) are totally unconvincing to an Austrian, they don't even recognize the Austrian pro-gold arguments.

Posted by: Kyle Markley on September 7, 2003 11:13 PM

Walt,

I'm not an economist by profession, but I've read a fair amount of Austrian economics, and from what I can tell most of the difference between Austrians and everyone else is disagreement over methodology. In particular, Austrians favor economic principles and deductive logic over empirical methods. They tend to be much more laissez-faire than others.

I've seen comments to the effect that Austrians and other economists seldom understand each other -- I wish I knew why, but it's definitely true. For instance, Brad's comments on the gold standard (http://www.j-bradford-delong.net/Politics/whynotthegoldstandard.html) are totally unconvincing to an Austrian, they don't even recognize the Austrian pro-gold arguments.

Posted by: Kyle Markley on September 7, 2003 11:16 PM

The switch from base-year to chain weighting obviously reduces growth rates when there is rapid technical progress and basket items are correctly priced hedonically. As a matter of interest, has the correction been made to old series? How much of the high growth rates of the past (say Europe in the 1960s)is a statistical artefact?

Posted by: James Wimberley on September 8, 2003 01:44 AM

Must agree with Ed Hugh (as usual) on the point about the reporter. (This is clearly an advantageous time to duck the Austrian issue.) Last week, Brad was complaining about a reporter who couldn't get his head around the notion that a deficit could be good at some times, bad at others. Richebacher is obviously more sophisticated than that, but still manages to screw up, in a way his readers are likely to take to heart, on the core issue of his article. Bad journalist! Bad!

Posted by: K Harris on September 8, 2003 05:39 AM

Professor DeLong is exactly right: what hedonic pricing giveth, chain weighting taketh away.

Two comments. First, several studies have found that carefully constructed conventional price indices for computers would also show very large declines.

Second, several OECD countries employ hedonic pricing for computers and measure real GDP growth with reference to a base year rather than by chain weighting. It is in those countries where GDP growth is likely to be overstated.

Posted by: Matt on September 8, 2003 09:30 AM

"In particular, Austrians favor economic principles and deductive logic over empirical methods. They tend to be much more laissez-faire than others."

They also have a theory about what causes recessions that's kind of crazy.

Posted by: Jason McCullough on September 8, 2003 09:43 AM

I think that discussion would be promoted if Mr. Richebacher were apprized of Mr. De Long's criticism to the effect that Richebacher is simply uninformed as to the statistical m.o. today applied. Richebacher's response could prove enlightening . . . . Since Mr. De Long uttered the criticism, it is presumably "his call" to apprize Mr. Richebacher of the criticism at issue.

Posted by: JD on September 8, 2003 10:38 AM

anne, recognizing that there are structural problems does not mean that short-term stimulus won't be effective. It depends very much on degree (i.e., percentage of economy affected).

Do you recognize that the command economy causes severe structural inhibitions to growth? What makes you think the Japanese economy is more similar to U.S. than the USSR? For instance, what are the upsides and downsides of lifetime employment? Why does it have positives in an ascending industry and yet can be so bad in a declining industry? If these same things apply to the U.S. and the USSR why not to the Japanese?

I may be wrong, but you seem to have a conclusion which you are trying to support.

Posted by: Stan on September 8, 2003 01:56 PM

The switch to chain-weighting apparently took place in 1995.

Posted by: Pooh on September 8, 2003 02:52 PM

You are correct to point out the flaw in Mr. Kurt Richebacher's article. Table 8.2 of the GDP tables at the BEA website shows clearly that the contribution of 'Computers and Peripheral Equipment' was 0.34 to the overall real GDP growth of 3.1% (both annualised) - about 11% as you point out.

Footnote 1 in Table 5.9 clearly warns against drawing the kind of conclusions that Mr. Richebacher has drawn:

"1. Includes new computers and peripheral equipment only. The chained-dollar estimates for computers can be used to accurately measure the real growth rate of this component. However, for computers, or for other components that exhibit rapid changes in prices relative to other prices in the economy, the chained-dollar estimates should not be used to measure the component's relative importance or its contribution to the growth rate of more aggregate series"

You have correctly said that the increase in nominal computers and peripheral equipment investment was about 8% and the deflation rate was about 4% in IIQ and hence a growth of 12% in real investment in C&PE.

This is correct as far as it goes.

But, if you look at the last three years, American real GDP has grown by a cumulative 4.6% (Q2, 2000 to Q2, 2003).

During this period, nominal investment spending on C&PE has dropped by 13% whereas in real terms, it has increased by about 43%.

During this same period, the average contribution of real C&PE to real GDP growth is around 11%. Hence, the entire cumulative real GDP growth rate over this period is covered by the hedonic price adjustment for the real C&PE investment alone.

If my analysis is correct (I do not know if it is correct), then the productivity acceleration in recent years is owed entirely to the hedonic price adjustment.

It will be useful if the BEA could provide a decomposition of the increase in real investment growth rate of C&PE of 43% over the last three years - how much comes from actual price deflation and how much comes from their quality adjustment, given that in nominal terms, we have seen a decline of around 13%.

Further, to the best of my knowledge, no other country uses hedonic price adjustment.

Interestingly, if we compare the nominal GDP growth rate of the Eurozone (12 countries in the single currency) and the US since the beginning of 1999, we find that the Eurozone nominal GDP is slightly ahead by the end of Q2 this year.

Over to you.

-Ananth

Posted by: Anantha Nageswaran on September 16, 2003 03:31 AM
Post a comment