The Bureau of Labor Statistics' (BLS) consumer price index is among the most commonly used economic indicators in daily life. Annual increases for social security recipients are indexed to the CPI. Many wage contracts include a CPI escalation clause as do rental and lease agreements. Yet, the CPI is often maligned as an inaccurate cost-of-living measure. In fact, the CPI is not a true cost-of-living measure the way it is currently calculated. It tends to overstate the rate of inflation truly faced by consumers because it doesn't allow for substitutions when prices change. For instance, consumers might increase their purchases of chicken when beef prices rise, and may be just as well off in the bargain.

Fed officials have long followed other price measures, in addition to the CPI, in order to get a more inclusive estimate of inflation. Fed policymakers and economists closely monitor the personal consumption expenditure (PCE) deflator because it reveals shifts in expenditure patterns as well as changes in prices. This index reflects the inflation rate that consumers are facing once they have already made their monthly purchases of various goods and services.

BLS heavy on R&D
The Bureau of Labor Statistics is not insensitive to the need for accurate price accounting. The BLS has a long history of improving all of its inflation measures. Some adjustments are minor and are incorporated within current indexes bit by bit. The BLS's central index is the CPI-U (Consumer Price Index for All Urban Consumers). The BLS also has created new price indexes. The most recent addition to the CPI stable is the Chained Consumer Price Index for All Urban Consumers (shortened to C-CPI-U). This index is more than just an adjustment to the current CPI methodology. The premise behind the chained index is new and different. The chained CPI is not even a fixed basket of goods and services, but a variable bundle of consumer goods and services.

Chained index is superlative
Unfortunately, there is no way of describing the new methodology without defining key index terminology. We promise to keep this short and to the point.

The regular CPI is a Laspeyres index, which measures the current cost of a fixed basket of goods and services relative to its past cost. This type of index tends to overstate inflation because it doesn't account for an individual's willingness to reduce consumption of items that are rapidly rising in price.

A Paasche index measures the present cost of the currently consumed basket of goods and services relative to the past cost. This method tends to understate the inflation rate because it measures prices of goods and services that are recording slower price gains -- that's why individuals are choosing to purchase the items in question.

In order to correct for the problems of overstatement (Laspeyres) and understatement (Paasche), superlative indexes were designed about 25 years ago. In 1996, the Bureau of Economic Analysis abandoned its old methodology of calculating fixed price indexes in order to compute real (inflation-adjusted) GDP and shifted to superlative indexes for more accurate real GDP estimates. Now, GDP and PCE price deflators are Fisher-ideal indexes. Essentially, the Fisher-ideal index takes the geometric average of the corresponding Laspeyres and Paasche index for each period. (A geometric average is the square root of the multiplication of the Laspeyres and Paasche indexes.)

The chained CPI is a Tornqvist index. Its calculations are very similar to the Fisher-ideal index and it is also a superlative index. According to statistics experts, superlative indexes generate very good approximations of true inflation. A major practical difficulty exists, though, in the compilation. The superlative indexes require the calculation of a Paasche index for the current period. However, this index can't be calculated until the period is over and the pattern of expenditures can be observed. Consequently, superlative indexes must be estimated for the near-term period, and then subsequently revised as soon as actual consumption figures for the period become available. Moreover, these estimates are generally revised as frequently as the consumption expenditure data are revised.

Practical matters
The BLS released July CPI data on August 16; the old CPI will only be revised next year insofar as the seasonal adjustment figures are updated, but the unadjusted data is final. In contrast, the chained CPI data for July is considered "initial" and won't be final for another two years. Currently, year 2000 data are final, but figures for 2001 are considered "interim" while data for January through July 2002 are "initial" values. In February 2003 with the release of the January 2003 index, the data for 2001 will be revised and become "final" while data for 2002 will become "interim". The 2002 figures won't be "final" until the February 2004 release. The BLS expects that monthly index revisions will be small. But this is a change from the current CPI-U that is only revised for seasonal adjustment changes. The unadjusted data does not get revised, which is one of its major advantages. Typically, less credibility is accorded to oft-revised data.

The chart below was compiled using the Bureau of Labor Statistics data reported on August 16. Keeping in mind that the data still will go through some iterations of revisions, it is interesting (but not surprising) to note that the chained CPI (C-CPI-U) is more similar to the PCE deflator than the regular CPI. The chart below shows year-over-year changes in each of these indexes. The differences between the regular CPI and the chained CPI are larger in 2001 than they are for preliminary 2002 figures. Nevertheless, the chained CPI is more in tune with the PCE deflator than with the regular CPI, since the former two are measured in the same fashion. July data for the PCE deflator is not yet available. Incidentally, the chained CPI showed a lower year-over-year inflation rate than the PCE deflator in every month since July 2001, except September 2001.

BOTTOM LINE
When new data series are created and released to the public, they are given a glimpse at best or ignored at worst. It is probably a good thing that economists are cautious in using new series because they don't have an established history. Yet, the BLS certainly would not have released these figures if they were not ready for public consumption. It may take a while before market players start looking at these new inflation measures. After all, who knows if Fed policymakers find them useful? If Fed officials were going to ignore the data, then it would be pointless for equity and bond investors to pay attention. Most likely, though, Alan Greenspan & Company are indeed analyzing and monitoring the new chained CPI. It would be foolish to ignore the inflation trends posted by this measure which is technically superior to the old CPI.

The July data show that the chained CPI rose 0.4 percentage points less on a year-over-year basis than the regular CPI. At 1.5 percent, there is no question that the rate of inflation is moderate. But a 1.1 percent rate of inflation is even more moderate. Keep in mind that the real federal funds rate target (nominal rate less inflation rate) reveals whether the Fed's underlying monetary stance is accommodative or tight. If the CPI-U measures the true inflation rate at 1.5 percent as it was in July, then the real federal funds rate was 0.25 percent - indicating that the current policy stance is reasonably accommodative. However, if the true inflation rate were 1.1 percent as measured by the C-CPI-U, the real federal funds rate target was 0.65 percent, not quite as accommodative. Across the board, the real federal funds rate is lower when utilizing the regular CPI (FF-CPI) than the chained CPI (FF-CCPI). This means that Fed policy appears more accommodative with the old CPI than with the new chained version.

Evelina M. Tainer, Chief Economist, Econoday