I read a really interesting book on my trip to Boston last week. It’s called Greenspan’s Bubbles: The Age of Ignorance at the Federal Reserve, by William Fleckenstein. I’ve read Bill Fleckenstein’s columns on-and-off since 1999, when I found him through Herb Greenberg. He’s definitely an intelligent guy, and while he presents like a perma-bear, the reality is that he’s really just a very strong, traditional, bottoms-up fundamentals-based valuation guy.
He has a real axe to grind in this book, but I’m going to do a book review in a separate post. However, one of the topics he raised was so interesting to me, I had to write a post about it.
Summary: I think we seriously messed up our monetary policy in the 1990s.
To be most specific, I think that in the 1990s, we made a fundamental change to the way we track inflation statistics for the United States that on the surface seems logical. But unfortunately, the realities about the economics of computers are so extreme, they may have completely distorted the inflation numbers for the entire country. And if you distort the inflation numbers for the entire country, you run the risk of distorting the monetary policy of this country. In fact, if you seriously mess up inflation calculations, you also affect fiscal policy, social benefit policy, and even global economic stability.
Yeah, it could be that big.
OK, here’s the information from the book that got me thinking. It starts on Page 39, in the chapter called, “The Bubble King”. Fleckenstein explains three changes that were made to the way the US calculates consumer price inflation (CPI) in 1995:
- Change 1: Move from Arithmetic to Geometric Rates. Ok, this one is perfectly legitimate. After all, inflation rates compound year to year, so calculating the rate as a geometric progression is fundamentally correct. I was actually shocked to find out we didn’t do this before, frankly. True, at low percentages, arithmatic and geometric calculations don’t always vary alot, but they do vary, and geometric is absolutely the right way to calculate the number.
For those of you asking what the difference is, let’s use this example. Say over 5 years, the price of milk goes up 50%. Arithmetic calculation would say 50%/5 = 10% per year. The problem, of course, is that if you actually raise the price by 10% per year, you get a lot more than 50% because the price increases compound each year. In Year 1, you’d go from $1.00 to $1.10, and in Year 2, you’d go to $1.21. By Year 5, you’d be at $1.61, not $1.50. It’s just like compounding interest in your savings account. Geometric calculations take this into account. Instead of 10%, they would say the inflation rate was 8.45%, which over 5 years compounds to 50%.
Doing this lowers the number reported, but it’s fundamentally the correct number to report on an annual basis. So far, so good.
- Change 2: Asset Substitution. This one is a little murkier. Basically, the way that economists calculate inflation for consumer goods is that they take a representative sample of products – hundreds of them. They then track the prices for these products each year. If you’ve ever seen those funny articles that track the “price index of the 12 days of Christmas” every year, you get the idea. 🙂
Asset substitution covers the case where similar goods might be substituted by people if one rises in price more than the other. Inflation is lower for the person, because instead of buying the high priced item, they buy the lower priced item. For example, let’s say the basket of goods included a 12-ounce can of soda. If the price of soda skyrocketed for some reason, most people would not actually spend the money, but would drink less soda and more water. The extent to which that substitution happens means that the inflation rate is actually lower for people, because they don’t feel the full impact of the rise in price of soda.
Fleckenstein argues that this change was “truly absurd.” Like a lot of the analysis in the book, that’s a significant exaggeration. The truth is, the fundamental need for substitution is sound. But like any of these economic techniques, if abused, this type of power could lead to incredibly huge errors in the calculation of inflation.
- Change 3: Hedonic Adjustments. OK, this is the one that has me worried. The CBO describes these as “quality adjustments”. Once again, the logic behind them is sound. It’s the execution that’s troubling. Hedonic adjustments account for the fact that if you improve the quality and features of one of the items in the basket of goods, the price might rise due to that increase in feature set, not inflation. For example, if in 2001 a Honda Civic has 145 horsepower, and in 2004 a Honda Civic has 160 horsepower, then the 2004 Honda Civic actually has 10% more horsepower than the 2001 version. To the extent that people pay for horsepower, the inflation numbers are adjusted to reflect that part of the price increase in the Honda Civic is due to increase in function, not just inflation.
Like asset substitution, this could easily be abused, since it involves a judgement call – how much has the product improved vs. how much has the price just risen due to inflation. It’s a hard line to draw, especially since in 2004 there are no new 145 horsepower Honda Civics around for an apples-to-apples comparison.
So, now that you’ve gotten your fill of Macroeconomics for the day, here’s the part where we may have wrecked our monetary policy.
Moore’s Law
Well, it’s not just Moore’s Law. It’s the pace of product improvement in the high tech industry, specifically hardware. It’s huge. It’s unbelievable. There has never been a manufactured good like it. There has never been a manufactured product, like the computer, that doubles in capability every 18 months. Hard drives double in size. I bought a 40MB external hard drive in 1993 for $200. I just bought a 1TB drive for the same price last month. That’s a 24,900% increase in storage for the same price in 15 years.
Try feeding that through “Hedonic Adjustment” and see what you get. A huge deflationary element.
Now, that wouldn’t matter, except for one thing: computers have become a decently large chunk of the US economy. Not huge mind you. The US economy is now over $13 Trillion. Computers are lucky to make up 2-3% of that. But 2-3% is actually a big number when you start feeding through it ridiculous improvements in “quality/features per dollar”.
Let me jump to page 101 of the book, in the chapter called “The Stock Bubble Bursts”:
James Grant, editor of the always insightful Grant’s Interest Rate Observer, was one skeptic who took the trouble to dissect the complicated subject that Greenspan seemed to accept at face value. In the spring of 2000, Grant published a study by Robert J. Gordon, a Northwestern University economics professor, who had prepared for the Congressional Budget Office a paper with a shocking revelation:
There has been no productivity growth acceleration in the 99% of the economy located outside the sector which manufactures computer hardware… Indeed, far from exhibiting a productivity acceleration, the productivity slowdown in manufacturing has gotten worse: when computers are stripped out of the durable manufacturing sector, there has been a further productivity slowdown in durable manufacturing in 1995-99 as compared to 1972-95, and no acceleration at all in nondurable manufacturing.
Grant backed that thunderbolt up with another study conducted by two economists, James Medoff and Andrew Harless. Their contention was that the use of a hedonic price index grossly misrepresented the actual data.
This is bad news. Bad bad news.
In case you are wondering, the fundamental question that our Federal Reserve and other governmental agencies concerned with the US economy ask themselves is how much of the growth in the economy is due to three factors:
- Population growth
- Productivity growth
- Inflation
If our calculation of inflation is off, it drastically changes our calculation for productivity. Productivity is the measure of how much economic value is generated from one time-unit of work. The 1990s were largely heralded as a decade of re-invigorated productivity growth. It’s why some people think Robert Rubin (or Bill Clinton) were great. It’s why people believed in a new economy driven by technological progress.
The data above is disturbing. Yes, it confirms that high tech might have had phenomenal impact on our aggregate numbers. But it’s totally misleading if it turns out that 98% of the economy was not, in fact, seeing productivity growth. Worse, it’s possible computers were actually masking continued weakness in every other area.
Look, I’m fairly sure that the people responsible for collecting this data are intelligent, and that this issue has likely been raised already. It’s also possible that this book and its citations are already known and discredited.
Still, I’m left with the following thoughts:
- Is the above data true? If so, does this mean the 1990s were not, in fact, a real productivity boom for the economy overall?
- If these issues are true and known, is the Federal Reserve, Treasury, Congress, et al taking this into account when they make monetary and fiscal policy decisions? If inflation is understated, then interest rate cuts, fiscal stimulus, and whole host of other policy decisions could be disasterous. We could end up with HUGE inflation in everything except computers to make the numbers balance. (I feel like this is like that line from “The Matrix Reloaded” – the system is desperately trying to balance the equation)
- When they make hedonic adjustments for computers, do they take actual utility into account? Sure, today’s Windows PC is 3x faster than one from five years ago, but the latest versions of Windows & Office are much more resource intensive than five years ago too. My Mac Plus booted faster than my PowerMac G5. How do they measure the hedonic adjustment for computers? Are they grotesquely over-estimating the increased value from hardware improvements, without discounting the resource requirements of software to provide equivalent “utility”?
Feel free to comment if you have pointers to information either confirming or refuting the above issues. This hits home for me as an issue that ties together two of my strongest personal interests – computers & economics.
Also, feel free to post this blog URL to other boards or forums where experts might be able to answer some of the above questions.
Psychohistory 