It Overstates Inflation! Does Not! Does So!

Fueled by the issue of quality changes, the consumer price index debate rolls on

Are new apartments bigger than old ones? Does better heart surgery compensate for the end of the house call? And is life better because of cellular phones, Windows 95, and Apple Cinnamon Cheerios? Questions like these aren't just fodder for idle conversation. Figuring out changes in quality is critical to settling a trillion-dollar debate: the controversy over whether the federal government is exaggerating consumer-price inflation.

To date, the shrillness of the inflation debate has been directly proportional to the scantiness of data. Last December, a blue-ribbon panel of economists headed by Stanford University's Michael J. Boskin told Congress that the consumer price index overstates inflation by something like 1.1% a year. It said the mismeasure, if not corrected, would add $1 trillion to the national debt by 2008. The two costs: excess payments for inflation-indexed programs such as Social Security, and excessive shielding of taxpayers from the effect of inflationary "bracket creep."

ON TARGET. Now two researchers from the Bureau of Labor Statistics, keeper of the CPI, are firing back. In a document released this May, Brent R. Moulton and Karin E. Smedley tear holes in some of the commission's calculations. But they balk at producing their own estimates of the overstatement (or understatement) of inflation, and say the bureau is improving the CPI as quickly as possible. "We're in the business of collecting data using standard, verifiable techniques and not in the business of speculating," says Moulton, who is chief of the bureau's price and index-number research division.

On balance, the Boskin commission's estimate of mismeasurement in the CPI may be roughly right. A poll of 320 academic economists by The Wall Street Journal found that 79% thought that there was some degree of overstatement: 21% thought it was 1% or more, while another 35% thought it was 0.6% to 1%. Boskin himself now says his panel's estimate might have been one- or two-tenths low. But the only way to know for sure is to dive down into the details of everyday life--and answer questions like the ones posed above. BLS Commissioner Katharine G. Abraham calls it "the house-to-house combat of constructing price-change measures."

To be sure, quality questions aren't the only issues in the inflation debate. The Boskin commission calculates that half a percentage point of the overstatement by the BLS comes from failure to recognize how consumers change their purchasing patterns in response to changes in prices. They may substitute a cheaper brand of pickle, buy chicken instead of beef, or shop at Home Depot instead of the local hardware store. In April, the BLS introduced an experimental alternative CPI that allows for substitution within narrow categories such as pickles. This adjustment alone would have reduced measured inflation by 0.3% for the past year.

But correcting the so-called substitution bias in the CPI is simple compared to measuring quality. The Boskin commission attributed the remaining 0.6% of the inflation error to undercounting the benefits to society from improved quality and the introduction of new goods. Counting those benefits is a mind-boggling task.

Consider apartments. The Boskin commission argued that inflation in shelter was overstated because the BLS did not give adequate weight to growth in the size of apartments, which helped offset higher rents. It said that the size of the average apartment grew 20% from 1976 to 1993. But Moulton and Smedley demonstrated that the actual size increase was only about 6%. If anything, they say, the CPI might have understated inflation in rents. Likewise, in women's apparel, the Boskin panel said old Sears, Roebuck & Co. catalogs show prices rose more slowly than the government says. The BLS responds that the Boskin panel focused on items that were discounted because they had gone out of style.

For most products today, the BLS uses a crude method for estimating quality change. Say a TV set disappears from the shelves, replaced by a new model with a better picture costing 5% more. If the inflation rate of other TVs was 2%, then the BLS assumes that the rest of the increase, 3%, can be attributed to higher quality--namely, the better picture. But the true test of quality is how the new set sells. For instance, if it gains market share, the quality must have risen more than the BLS's 3%. Says Yale University economist William D. Nordhaus: "We actually don't know how much quality change exists in the BLS numbers."

FOOT-DRAGGING. Not that other quality-measurement techniques they use are much better. With cars, for example, the BLS sometimes calculates the value of a new feature by asking the manufacturer how much it cost to add--at best, a rough approximation of its value to the consumer. The BLS uses a different method for some other items, including computers and apparel, basing the change in quality on the improvement in quantifiable features--such as the processor speed of a PC. But a computer that's twice as fast might not be twice as valuable to the buyer.

Economists differ on whether the BLS is moving quickly enough to improve its quality measures. Yale's Nordhaus is sympathetic: "It's a gargantuan problem that they're tackling. They collect almost a million price quotations a year." Yet, Jerry A. Hausman of Massachusetts Institute of Technology has stated publicly that the entire bureau should be shipped off to Guam. "They're dragging their feet," Hausman says. "They're hurting the economy."

Hausman has invented an ingenious way to measure the consumer benefit from the introduction of new goods. Not only is the BLS slow to include new products with rapidly falling prices in the index--cellular phones still aren't in--but it neglects to measure the extra value of new goods vs. what was available before. In a 1994 paper, Hausman calculated that the introduction of Apple Cinnamon Cheerios by General Mills Inc. in 1989 improved the overall welfare of the American public by an average of 27 cents per person per year. As for cell phones and service, he calculates that leaving them out means that the telecommunications portion of the CPI is overstated by about 2.3 percentage points per year. Hausman argues that the Boskin commission's estimate is, if anything, "probably on the low side."

The BLS is taking some steps toward improvement. This year, it started measuring medical-care costs by pricing bundles of services instead of discrete items such as pints of blood. That's a step in the right direction: People value the outcome of a hospital stay, not how many injections and operations they got there. The bureau is also moving to speed up the introduction of all kinds of new products into the index.

NO. 1 DEFECT. But the BLS remains hamstrung by a data-collection methodology that makes it difficult to compare the quality and price of products over long periods. For one thing, the set of priced products differs from city to city, so there's no way to come up with a national price for, say, a particular model of microwave oven. Boskin panel member Robert Gordon, a Northwestern University economist, says he finds better data by going through old copies of Consumer Reports magazine, which list detailed prices and specifications of various products. "Consumer Reports is doing what the BLS should do," says Gordon. Hausman, among others, says that the BLS should make far more use of bar-code scanning data from supermarkets and other stores.

The BLS has been getting an earful from economists since 1961, when a panel headed by the late Nobel laureate George J. Stigler stated that underestimation of quality was "the most important defect" of the CPI. The sooner it gets fixed, the better. After all, a trillion here, a trillion there, and pretty soon you're talking real money.