remove quality bias from the CPI!
I spent today's evening in a rather unusual way: instead of reading 10K reports, HiddenGems message boards, Caps blogs or the rants of bloggers at housingpainc.com, I started reading the Boskin commission report and the Bureau of Labor Statistics (BLS)'s response to it. While I still have to finish reading them (yes, when you encounter terms like Laspeyres formula or superlative indexes, this slows down the reading process considerably), I can already make two observations: a) this is some of the most interesting stuff you may find on the internet, for all its technical mumbo jumbo, and b) I still don't understand the process by which the Bureau of Labor Statistics calculates its CPI index. Maybe I can improve my understanding if I read some more. The point that I want to share with you now, however, is that much of the discussion appears to be centered around the so-called "quality change bias". What is quality change bias? The idea is that if today's car selling for 20K runs better than the 10K car of yesteryear, then we should say (for CPI calculation purposes) that it sells at 10K. Simply stated, the essense of the debate boils down to this: the Boskin group (which is composed of politicians looking for ways to lower inflation on paper) is pressuring the BLS to report lower prices than the ones actually recorded by their data collectors, while BLS (which is not directly concerned with the Treasury's financial woes) is trying to be accomodative but still holds on to its last vestiges of honesty. As a result, Boskin & Co. wants BLS to generate an extra 10% price reduction by 2008 by creative quality accounting, while BLS replies that they're already doing just that, will try to do more, but cannot quite go all the way because then honest people will refuse to shake their hands. This exchange of opinion is truly fascinating, and I recommend reading both articles to get a better understanding of the CPI calculation process (you might even understand how the imminent bancruptcy of Social Security is getting postponed year after year after year without any tax hikes and without any visible reduction of benefits). But my point is that quality changes should never be tracked by statistical agencies at all. It is wrong both conceptually and statistically.
The first reason we should throw out all quality adjustments once and for all is that, despite all claims of Boskin & Co. to the contrary, there is simply no way to quantify them. For example, there was no internet 20 years ago, and now, lo and behold, every chimpanze has a blog. Is that a quality improvement? Yes, certainly. Now, what kind of discount to the CPI does it warrant? 1%? 10%? 0.1? Your guess is as good as mine. But wait a minute. Now, dial-up becomes a thing of the past, and everybody can get broadband at the same price. So, how do we account for that? One obvious way is to compare the cost of broadband 10 years ago and now. But wait, there is a problem. 10 years ago, the technology is just emerging, so its price is not indicative of consumer preferences. It's indicative of what it costs to produce the service at the time and of what you can pay for the service if you're filthy rich and like to be an early adopter. The majority of consumers are simply not ready to buy at this price point. There is no reason broadband should be a part of CPI at that time (analogy: there's no valid reason to include cosmic tours into the consumer basket now that Dennis Tito has shelled out sixty million bucks for a ticket to the MIR space station). Here's another idea: look at what people would pay for dial-up now if given a choice between dial-up and broadband. Sounds better? But this comparison is also meaningless, because internet content has changed so much that the old dial-up is now nearly useless. In 1998, dial-up was perfectly adequate for the internet that existed back then. But suppose we made some crude estimate. Next question: today's internet offers hundreds of times more information that it did in 1998. So how much better off are we now? Should we assume that hundred times the amount of information is hundred times more useful? That's pure nonsense. We all understand that we do get more useful info from the net than we did in the past, but any numerical estimate is bound to be purely arbitrary. In other words, the Boskins could report any CPI figure that suits their purposes.
But it's not only a quantitative issue. Accounting for progress is also wrong conceptually because in reality, what we call quality improvements should be considered in the new context. Let's take an example. The houses people built in 1860 would look inferior by today's standards. Boskin, for example, would immediately point out that this housing was something like 100 times worse that what we have now. But before we congratulate ourselves with our progress in construction quality, here are few things to consider. In 1860, building a house took 3 weeks. All you needed was to find a clearing in the woods and to bring some axes, sows, hammers and nails. You did not need to buy land (it was free under the Homestead act). You did not need to hire workers (your 5 sons would make your construction crew). You did not need to buy materials (the logs came from the surrounding forest). You did not need to apply for a mortgage (in those good old days you could achieve your American dream without a FICO score). You did not need to worry about job loss or relocation (your job was right here, at the farm). You did not need to pay for title insurance, property insurance, mortgage insurance, mortgage origination fees, etc. And now, quite frankly, how many struggling homebuyers would opt for the 1860 solution to their housing problem (3 weeks of physical labor with virtually zero cash outlays) if such an option were still available in 2008? So, how do you compare housing then and now? The answer is, you can't make a valid comparison because the economic context has changed so much that you'd be comparing apples and oranges. Besides, the real cost of "new-and-improved" goods often cannot be determined. To continue the house analogy, the quality improvement observed in a modern McMansion vs. the 1860 log cabin may justify its steeper price tag, and actually translate into price deflation (in your mind, of course, and only if you're Boskin). Consider, however, that unlike in 1860, in order to afford this house today, you will need to have a good job, and in order to get a good job, you will have pay for college education. In the process, you will have to "consume" a hundred textbooks, a hundred terabytes of broadband internet traffic, many calories of overpriced food in your college cafeteria, and many more thousands of dollars of "housing services" in your overpriced college dorm. So what is the real cost of that house?
There is only one sensible solution. Quality considerations should be excluded from GDP and CPI indexes. They could be included into a separate "quality of life" index that should be computed by a group of philosophers traveling around the world and cross-comparing different countries. But GDP and CPI should count only those quantities that are countable. In this way, we will have some piece of statistics that we can trust and that could be used in financial planning, as opposed to the arbitrary Boskin-style hedonic indexes that can't measure hedonics and can't measure price changes.
One additional argument in favor of this simplification is that hedonics is useless to you as a consumer, because you cannot pay your bills with it. When you retire on Social Securiry, and walk into a grocery store, you have to buy real cucumbers and real tomatos, and pay for them with real dollars. You can ask the cashier for a discount on the grounds that today's cars look much more sexy than similar-priced cars of the 2007 vintage, but the cashier will not be impressed. Also, if you lend your money to the government by purchasing TIPS, you will want to be paid back with cold, hard cash, not with hedonic pleasures of computer users. Or, if you are still convinced that quality changes should have a place in the CPI, you should at least insist that the old product must still be available to consumers and its use must still be a feasible choice in today's environment (Let me formulate it in simple terms: first, let people build log cabins in Manhattan, then adjust prices of modern Manhattan housing for hedonics).
Technological progress does not lead to improvements in the quality of life, it merely changes the way we live, and it should be taken for what it is: a natural phenomenon akin to rain, wind, or evening sunset. It may be a good thing that today's flat-screen TVs are even flatter, computers are faster, the sun is brighter, the grass is greener, the music is louder, and so on. But in the final run, this and two dollars will get you on the subway.