The New CPI Is Better, More Precise Than the Old CPI

by: Mark J. Perry

In a comment Thomas Blair said: "The way CPI is calculated has changed, which makes comparisons between eras or even years irrelevant."

An anonymous commenter writes "Your reasoning is so specious. Firstly you have to recognize that inflation [CPI] is measured differently from the 70s so you're comparing apples to oranges."

The BLS provides CPI data here going back to 1978 using the new CPI-U-RS, which allows for a comparison of inflation calculated in the late 1970s using the old CPI (CPI-U) to inflation calculated using the new CPI (CPI-U-RS), see top chart above for a comparison of the two different CPIs between 1977 and 1998 and bottom chart for a comparison of inflation rates from December 1978 to December 2007 (blue line is inflation using the old CPI and red line is inflation using the new CPI).

Yes, it is true that the old CPI overstated inflation (see top chart above) and it is true that inflation was overstated in some years using the old CPI (see 1979, 1980 and 1981 in bottom chart, when the blue line is above the red line), but then actually understated inflation in the 1983-1984 period (blue line is below red line), and was about the same using either method for most of 1982 and in 1985-1986.

Bottom Line: The old CPI slightly overstated inflation, and revisions were made to correct for the slight upward bias and make the new CPI a more accurate measure of inflation. Some reports seem to suggest that the new CPI creates a downward bias for the way inflation is now reported, and that it would be higher today if it was calculated using the old, upwardly biased method. I think it's better to think of today's CPI and inflation rates as being better, more precise and more accurate measures than those in the past, and any adjustments should be made to previous years' inflation, not today's rates.

Further, I don't think the differences in the old CPI-U and the new CPI-U-RS are meaningful in any way that contributes to the discussion about comparing today's inflation to the 1970s inflation. It's far from a specious, irrelevant, apple to orange comparison, it's much more like comparing one variety of apples to a newer, better variety of apple. Inflation in the late 1970s was at double-digit level by either CPI, and inflation today, measured more accurately than ever before, is nowwhere near double-digits.

Some of the recent CD discussion has focused on the possible difficulties of comparing today's inflation rates to the inflationary 1970s period because of the changes that have been made to the way the BLS calculates the CPI. There was also some discussion about the timing of the changes to the CPI. Thanks to CD reader Spencer for the suggestion to check the BLS website for a comparison of the CPI-U-RS series (new) vs. the CPI-U series (old).

I was able to locate the BLS study "Consumer Price Index Research Series: Using Current Methods, 1978-1998," which finds that "the measured rate of inflation would have been lower since 1978 if methods currently used in calculating the Consumer Price Index for All Urban Consumers had been in place from that year to the present."

In other words, the old BLS method of calculating the CPI-U overstated inflation (which has been taught in Principles of Macro for decades), and various adjustments (14 in all, see Exhibit 1 in the study) made between 1983 ("changed homeowners’ component from cost of purchase to to value of rental services) and 1999 (the last 5 in 1998-1999 were maybe in response to the 1996 Boskin Commission Report?) attempted to make the CPI/inflation more accurate.

How big were the changes between the old CPI-U and the new CPI-U-RS after the 14 changes? Table 2 above tells the story:

  1. Although most changes to the CPI lowered inflation rates under the revisions, some of the changes to the CPI actually increased inflation rates (see the row above titled "Effects of all other changes" that adjusted for the fact that the old series "understated" inflation in some ways).
  2. Although the net effects in each period above were negative, meaning that inflation was overstated under the old methodology, the differences in inflation rates were relatively minor. For example, inflation from 1978 to 1982 averaged 9.46% annually under the old method and 8.46% under the new method, suggesting that the old method overstated inflation by about 1% per year.

For the other periods, the annual differences between old and new CPI inflation were relatively lower: -0.13% from 1983-1986; -0.35% from 1987-1997; and -0.23% in 1998, or an average of about 1/4 of 1% per year!

Bottom Line: With reasonable accuracy, we can fairly compare today's inflation rates calculated from the CPI-U-RS used by the BLS today to the inflation rates of the 1970s that used the old CPI-U, with a possible difference annually of only about 1% at the most. In other words, if you subtract 1% from the double-digit 10-14% annual core inflation in the mid- and late-1970s, you still get inflation way, way higher than the 1.5%-2.5% range of core inflation over the last ten years.

My position remains that unless, and until, we have core inflation approaching double-digits, inflation is not a problem. And I don't think there is near enough of a difference between the old CPI-U and the new CPI-U-RS to have that issue make any significant contribution to the inflation discussion.