As discussed in the previous installment in this series, adjustments are at the core of the debate about inflation, but another way to look at the problem is to consider exactly what’s being measured here. Substitution, of one type of good for another, is a clear case of something being different. What about when you’re talking about the same item, though—say a TV set? Is an HDTV with a flat screen and built-in Internet access the same as an old tube TV without a remote?
These adjustments—for quality, utility, or both—are known in the trade as hedonic adjustments. Again, Shadowstats leads the charge against the current adjustment method, using government-mandated gasoline additives as the poster child for fraudulent adjustments. In its response, the Bureau of Labor Statistics outlines its methodology and gives two examples: a candy bar, which is selling for the same price as before but now weighs half an ounce less, and a TV, which is now available only as an HD model at twice the price. Is an adjustment necessary? If yes, how should it be calculated?
Under the constant basket of goods approach, no adjustment would be made for the candy bar, as the price per candy bar wouldn’t change, and the price increase for the TV would be registered as 100 percent, because the price increased by that much, with no adjustment for quality. The BLS, however, chooses to adjust for quality in both cases, on the grounds that ignoring the significant quality differences wouldn’t accurately reflect the actual changes in the price levels.
So, given this determination, how is the adjustment calculated? Once again, the devil is in the details. The BLS method, using regression analysis, is typical and consistent with other econometric analytical methods. The real question here is, does an adjustment need to be made at all? Most would suggest yes—a 1-ounce candy bar is different from a 1.5-ounce candy bar, and an HDTV is different from a standard TV. The question then becomes how that adjustment is made, and the method used, multiple regression, is among the best available as a general case.
The final major perceived flaw in the CPI calculation—and the biggest issue of contention—involves how housing costs are calculated and included in the indices. Briefly, the way the CPI is now calculated is based on treating a homeowner as if he rented out the house to himself. By treating the house as a small business, with the imputed rent less any costs—real estate taxes, insurance, maintenance—subtracted, the CPI calculation is indifferent to whether the house is owned or rented. Think about it: if you suddenly decide to rent out your house and rent an equivalent house elsewhere, has your cost of living (other things being equal) changed? Conversely, to look at the issue from a GDP perspective, if you rent out your house to a third party, has the economy grown at all? From these perspectives, the treatment of housing appears reasonable.
The principal argument against imputing rent as a measure of housing costs is that it doesn’t reflect asset price changes. As house prices rise, the argument goes, that cost should be reflected in the CPI. There is a more general argument to be made here: as asset prices increase in general, the CPI should reflect this. And so, generally, it does in the form of increasing prices for items in the basket of goods. What makes housing unique, however, is that the investment aspect of it—the owner’s benefit from increasing value—is separate from the value of the actual housing services. That is, if housing prices climb, rents don’t necessarily climb as much. In fact, that is what actually happened in the mid-2000s, vindicating the BLS position. The living cost, then, is best reflected by the rental levels.
Another way of evaluating this is to look at the experience of 2008–2012, approximately, when housing prices were declining and rents were going up. The cost of living for anyone looking to rent a house was rising, even as homeowners’ net worths declined. Clearly, with rents going up and more people being forced to rent rather than buy, the overall cost of living wasn’t going down—even as asset prices declined.
Yet another way to test whether the CPI is understated by using owner-equivalent rent is to look at other data series from the real world, over an extended time period. The National Association of Realtors Housing Affordability Index increased by 79 percent from 1983 to 2007, less than the BLS rental index increase of 140 percent, suggesting that actual housing costs are not understated by the BLS methods.
All of these methods generally support the reasonableness of the existing CPI calculations while leaving open the possibility that something better should exist. The test of the CPI against the housing value index suggests that there should be some other real-world data set against which the CPI can be benchmarked.
The idea of testing the CPI against real-world data is appealing, but inherently difficult, since the CPI itself is based on a wide array of collected prices. What is needed is an independent, consistent, unadjusted (or minimally adjusted) series of data that would generate a series of price adjustments. Historically, for cost reasons, this has not been feasible. In the past couple of years, however, the Internet has made just such a series possible. As a test for the CPI prepared by the BLS, it is the best available, and we’ll take a look at it in the final post in the series.