December 30, 2001

Research Says CPI Improving, Needs Work

WASHINGTON (Reuters) -

Five years after a high-profile commission said the Consumer Price Index substantially overstated the rate of inflation, new research says improvements have been made although work still remains.

In December 1996, the so-called Boskin Commission, headed by influential Stanford University economist Michael Boskin, said the widely watched CPI overstated the rate at which prices rose by about 1.1 percent annually.

The report came at a pivotal time, as Congress was looking for ways to save money and balance the federal budget and smaller cost-of-living adjustments for retirees looked attractive. The report noted several different ways the index could be changed to remove its alleged upward bias.

Five years, several methodological changes and a spate of budget surpluses later, new research from the Federal Reserve and the National Academy of Sciences gives some direction for other possible changes but otherwise advocates a ``go-slow'' approach.

The Federal Reserve study, ``Measurement Error in the Consumer Price Index: Where Do We Stand?'', provides a smaller estimate of the CPI's alleged upward bias, at about 0.6 percent a year.

'LOW-HANGING' STATISTICAL FRUIT PICKED

``One reason our bias estimate is smaller than in several earlier studies is that the (Bureau of Labor Statistics) has recently made a variety of improvements to its procedures. Although much of the low-hanging fruit has already been picked, further progress remains possible, and our study has highlighted several potential areas for additional improvement,'' wrote the Fed study's authors, David Lebow and Jeremy Rudd. The study was posted on the Fed's World Wide Web site earlier this month.

While the National Academy of Sciences study, ``At What Price? Conceptualizing and Measuring Cost-of-Living and Price Indexes,'' provided several specific recommendations for further study and potential methodological changes, it did not contain an estimate of bias. But it also urged the BLS to move slowly on a major front in the debate over the CPI's accuracy -- how to account for improvements in the quality of products, even as their prices remain the same or fall.

``Relative to our view on BLS research, we recommend a more cautious integration of hedonically adjusted price change estimates into the CPI,'' the study said.

Patrick Jackman, a BLS economist, noted the agency has made several changes in recent years to improve the CPI and will be unveiling others soon.

For example, in the CPI report for January, to be released in February, the index will incorporate the most up-to-date ''weights'' for consumer purchasing patterns yet. The CPI is built on a ``basket of goods'' approach, based on periodic surveys of how consumers spend. In the past, the various weights given to goods in the basket have been updated with a lag of several years. That means the BLS has been slow to include new products, such as cellular phones, into its price index, often as they have fallen sharply in prices.

But starting with the January data, the report will reflect spending patterns seen in 1999 and 2000 and will be updated every two years.

Also, around mid-year, Jackman said the BLS plans to release an alternative CPI that will attempt to measure the ''substitution bias,'' the willingness of a consumer to buy more apples if their price should fall compared with other fruits. The Boskin Commission cited ``substitution bias'' as source of upward bias in the CPI.

QUALITY, SURVEY ERROR SEEN

But where the Boskin Commission saw the question of adjusting for quality changes as an almost exclusive source of upward bias, the National Academy report was more cautious. Jackman said quality change can work two ways -- improvements mean lower prices but deterioration can mean higher prices.

While many goods and services have improved in quality in recent years, others, such as air travel, may have declined, he said. Jackman said air travel quality may have fallen in light of skimpier in-flight meals, increased delays and more cramped seating than in the past.

Both reports, however, raised questions about the accuracy of the consumer spending survey around which the CPI is built. The Fed study said the survey may not reflect actual expenditures, and would account for 0.5 percentage point of the estimated 0.6 percentage point upward bias. Similarly, the National Academy study proposed an experimental CPI be created using ``weights'' culled from a separate government survey of spending.

Jackman said it's likely some ``sin items,'' such as alcohol and tobacco use, are underreported, but he said it was not obvious that one set of spending data was better than another.

``The question really is, is their data that much better than the CPI data or are both in error, but maybe in different directions,'' Jackman said.