No, Typewriters aren't still in the CPI. Here's how that works.

7/19/2013 02:44:00 PM
A lot of conservatives, hard-money types, and Ron Paul types believe that true inflation is way higher than the official Consumer Price Index (CPI), compiled by the Bureau of Labor Statistics (BLS), figures would suggest. One popular claim has been that if the BLS still measured CPI the way they used to, then inflation would actually be over 10 percent, not the paltry 1.7 percent the current methodology suggests. For example, here is Naill Ferguson (whom I've written about before):
"The way inflation is calculated by the Bureau of Labor Statistics has been “improved” 24 times since 1978. If the old methods were still used, the CPI would actually be 10 percent. Yes, folks, double-digit inflation is back. Pretty soon you’ll be able to figure out the real inflation rate just by moving the decimal point in the core CPI one place to the right…."
The specific 10 percent claim comes from a guy named John Williams, who appears to have misrepresented his own work since contrary to his original claims--and how his work was interpreted by Ferguson and others--he actually was "not going back and recalculating the CPI" (his words), and hence did not determine what the CPI would be under the old methodology. I think that misrepresentation is important to point out, because this is a guy who runs a website premised solely on the belief that the government is misrepresenting economic data. You really have to wonder why Ferguson, a Harvard professor, is relying on a crackpot like Williamson to lend credibility. And you have to wonder why Harvard doesn't seem to understand that Ferguson has become a walking embarrassment to their venerable reputation.

As this report proves, the 10-percenters are all wrong. Not only do they have no understanding of BLS methods, they dramatically overstate the impact of these methodological changes: BLS does have that information readily available, and we're talking about revisions amounting to three-tenths of a percentage point, not the 7 percentage points Williamson claimed. It's a great document to read, even if you aren't a 10-percenter, because it really goes into detail about how these data are collected, which is something that we should all know. Perhaps if everyone read this, there'd be less derp and more useful, well-informed critiques of BLS methodology. Look, no one is arguing that the current methods are perfect!

Anyway, a big part of the criticism of CPI, which became popular among many of my own conservative friends, was that CPI systematically understates true inflation because it assumes that people will substitute to cheaper goods. That is, they think that people switch to cheaper goods as the cost of living rises, and that CPI only reports the difference in the old good and the new different but cheaper good, ignoring the new higher price of the old good. That would be a huge oversight if that were the case. However, that's not what they do. I think misconception arises from conflating two aspects of what the BLS actually does.

First, the BLS does assume that people won't continue to buy the same quantity from firms posting the highest price increases for the same product that can be purchased nearby for less. BLS data is broken down by specific product--for example, apples--and by specific region--for example, Chicago. So suppose there are two apple sellers in Chicago, and one of them doubles the price it charges for an apple, but the other one does not change it's price at all. It would be extremely unreasonable to say that the price of apples in Chigaco doubled over this period. It would also be unreasonable to say that there was a 50% increase in the price of apples in Chicago which is what a simple arithmetic mean would say, because people can still go to the other seller and buy them at the original price. Instead the BLS uses a "geometric mean." In general, a geometric mean is always less than or equal to the arithmetic mean (if you pursue a math-related career, you'll probably be asked to prove this theorem on an exam at some point). The geometric mean will tend be "biased" towards the firms that post the smallest increase in price, which is what you want because consumers will be biased towards buying the cheapest ones anyway. This assumption is built upon a solid foundation of reams of empirical research. Though it may not be the absolute best method, it is clearly better than the arithmetic mean for same-product comparisons. And for the record, there's no pure mathematical reason for favoring the arithmetic mean over the geometric mean. They are alternative concepts of "average," and neither is more "correct" or "biased" than the other. The bottom line, however, is this: it is simply not true that the BLS figures are biased when people switch to cheaper goods as their standard of living falls--the BLS does assume that given the choice to buy the same good at a cheaper price from a different firm, some of the consumers will choose to do so. And for the record, if the price of all apples rise by the same amount, then the geometric and arithmetic means are identical.

Second, I think the above issue about substitution between cheaper and more expensive versions of a product gets conflated with the completely separate issue of how the BLS deals with the changes in types of goods firms sell. At the heart of the issue are cases like the typewriter. At one point in time, lots of people were buying typewriters, and so their prices were tracked and factored into the CPI. But no one buys them anymore, so it makes absolutely no sense to continue to keep them in the CPI. No one sells them either, so the BLS couldn't keep them in the CPI if they wanted. When possible, the BLS deals with obsolete products disappearing from the market by identifying the next closest substitute and decomposing the cost into a pure price change component--the portion of the change in cost due solely to a change in relative prices and inflation--versus a quality component, which is the value added premium consumers are willing to pay for a new and better product (or the cost decrease due to a loss in quality--for example, when food makers decrease the serving size). The point is, the BLS doesn't pretend we still buy typewriters, nor do they assume that a computer is an identical substitute for a typewriter. Rather, they statistically account for the impact the computer has had on the price of basic word processing, while recognizing that the other features of the computer represent a genuine increase in the standard of living since the typewriter days. And all of this is done with statistical models and hard data collected from samples of representative probability samples of real consumers--The BLS is not making any subjective assumptions to impute values for the CPI.

Now, if you want to claim that the government is cooking the books, engaged in an outright fraud and that I shouldn't trust anything these researchers say, you are wrong. We have alternative, independent measures of prices that corroborate the government data. I'm not sure what it is about some people that makes them want to believe such an extreme claim that is unsupportable by any evidence.

As a final point, some people I've debated with seem to think that the Federal Reserve targets "core-CPI" rather than the full CPI-U index, and that this somehow leads to inflation or instability or something. First off, the premise is false: the Fed actually targets the Personal Consumption Expenditure (PCE) deflator , which does use the same pricing data as the CPI but is weighted according to the PCE component of the Gross Domestic Product data. The PCE deflator is not necessarily representative of an individual consumer (there's a difference between the behavior of the average consumer and the average behavior across consumers), but certainly unbiased for the economy as a whole. However, the PCE deflator is not itself the best predictor of the future path of the PCE deflator, due to the large weight it gives to a few highly volatile prices. Internally, the Fed uses the so-called "trimmed-mean" statistic, which is much better at forcasting the PCE deflator. The focus on the "core-CPI"--which is just CPI without food and energy prices--really only exists in Fed communiques with the public, because the core-CPI gives similarly accurate forecasts of inflation, but is much more intuitive and easy to explain to the public. If you doubt my claim that the core is a better predictor of overall prices than the overall price index, and that the "trimmed-mean" is better than the core, here's one of many papers on the subject. The bottom line is that the Fed cares about all prices, economy wide, and does not selectively pick indices that understate inflation.

So no one is arguing that the government's inflation data is absolutely flawless. Moreover, from an individual's perspective, this flawlessness isn't really possible: any individual's spending habits will differ in various ways from the average consumer represented by the index. But if you're argument is that the government's price indices are way off base, let me be clear: you are wrong.