hen the Consumer Price Index is announced each month, it gets a lot of attention. People might assume that its creators know exactly what consumers are doing: how much they spend on an average trip to Costco, for example, or on a delivery from the local Thai takeout; how many DVD's they rent or buy; or how much they pay each month for their high-speed Internet service.

They would be wrong. The index, the main gauge of inflation and bedrock statistic of the federal government, is a surprisingly frail number. The calculations behind it are as much art as they are science.

But that has begun to change. Starting next month, the Bureau of Labor Statistics, which computes the index, will use some of the freshest data it has ever collected. It will readjust the weight, or importance, of goods and services in the index marketbasket, and the stores where they are sampled, far more often than it did in the past.

And later in the year the bureau will start accounting for something that consumers do all the time: substitute one kind of product for another — chicken instead of steak, cotton instead of cashmere, used instead of new — to save money when prices rise.

Whether these changes will change the official inflation rate significantly is unclear. But it is hard to underestimate the potential impact. The Consumer Price Index is not just a measure of living costs. It is used to help determine the size of Social Security checks, veterans' benefits and other federal payments to about 80 million people. It affects how the Federal Reserve sets interest rates and how the Internal Revenue Service adjusts tax brackets. It can even change the size of alimony checks.

This is hardly the first time that the bureau has changed the index in response to new economic thinking. But the 2002 changes, with some others in the last few years, are the fruit of prolonged debate, inside and outside the bureau, as well as recommendations by two government commissions, one going back to the 1960's and, more recently, the Boskin Commission, which reported to the Senate in 1996.

The changes are meant to better reflect consumer behavior and, in turn, to give policy makers, businesses and consumers a better tool to gauge how social and technological changes affect inflation.

Two-income households, for example, mean more spending on takeout food and day care. New attitudes toward health may mean less spending on cigarettes and soft drinks. The shift toward work outside the office means more purchases of products like laptop computers and pagers. The rapid growth of grocery warehouses means that consumers are buying more food in bulk and at lower prices.

 
In the past, such trends took years to affect the index because the bureau's marketbasket was based on consumer buying patterns surveyed 5, 10 or even more years earlier. The current index, for example, is based on consumer behavior in 1993- 95.

When the bureau releases its January report on the index, it will use new "relative weights" for products and services based on consumer choices in 1999 and 2000.

More weight will be assigned to college tuition and nursing home care, for example, and less to full-service restaurant meals and men's suits. Those weightings will then be adjusted every two years — again in 2004, based on 2001-2002 surveys. Although there will still be a lag, it will be shorter.

"The B.L.S. is trying to get closer to a measure of inflation that people are experiencing," said Stephen G. Cecchetti, an economics professor at Ohio State University and a former Federal Reserve Bank officer.

In theory, the calculation of the index is simple. It is based on a marketbasket of 211 goods and services — medicine, education, entertainment and so on — bought by the average family. The prices are tracked over time.

But the task of calculation is daunting. As the bureau chooses among millions of products, it is constrained by budget limitations and saddled with old technology. Bureau agents roam stores, looking at price tags and writing prices on pieces of paper. They interview store executives, visit homeowners to determine housing prices and ask consumers to keep daily diaries of purchases.

All kinds of variables, including new products, mean that the bureau has a tough time keeping up.

Consider the bureau's prolonged statistical neglect of the cellphone. Cellphones were a staple in almost three million American households as early as 1992, according to Motorola (news/quote), the big cellphone maker. But cellphones were not reflected in the index until 1998, when Americans bought almost 35.5 million of them. That was largely because the marketbasket was still locked into an outdated formula that had counted telephone service in the costs of housing — and cellphones were used outside the home.

The inflationary effect of new features on old products also bedevils the index. Consider a shopper who bought a Chevrolet Blazer in 1998 for $23,703 and now wants to replace it. The comparable 2002 vehicle, the Blazer LS, is priced at $26,310 — an increase of $2,607, or almost 11 percent. But because it now comes standard with a Bose sound system and a theft-alarm system, among other upgrades, the bureau will slice hundreds of dollars off the price increase when calculating the price inflation.

The bureau is now making such adjustments to more goods and services in its marketbasket. It is also trying to figure how to reflect quality decreases in air travel, for example. It had tracked product changes that affect the environment, like emissions systems, but stopped recently to re-evaluate that process.

Taken together, the changes should make the index more timely. Patrick C. Jackman, a bureau economist who is considered its top authority on the index, offered one example: Viagra, the male impotence drug introduced in 1998, was moved into the index in 2000.

Starting next year, the bureau will measure a more subtle variable: what happens when consumers buy one product instead of another that has risen in price. The bureau began accounting for substitution in a rudimentary way in 1999, literally apples to apples or cheese to cheese. If the price of Golden Delicious apples goes up, for example, that would seem to bump up inflation. But if a shopper switches to cheaper Granny Smiths, the bureau records what the shopper actually spent — which means less inflation.

In 2002, the bureau will do this kind of calculation across categories — measuring what happens when a shopper buys chicken, for example, because the price of lamb has increased. This, too, will result in a lower inflation rate, although the bureau is planning, for now, to report this number separately, not in the basic index.

Such changes in bureau procedures have been welcomed by critics like Michael J. Boskin, the Stanford economics professor who led the commission, established by the Senate in 1995. If the index does not record how technology is improving consumer life, or how consumers save money by buying cheaper goods, he contends, it overstates inflation.

Economists continue to debate the issue. If consumers are forced to buy more expensive products, they end up spending more — and to some experts, that should mean higher inflation. Michael Goldstein, a finance professor at Babson College who has worked at the New York Stock Exchange and with federal agencies, said the consumer index should "measure how much it costs to be a regular person in this society, and to keep up with everybody."

Even small changes in the index can make an enormous difference in financial calculations. The Boskin Commission concluded that the index overstated inflation by an average of 1.1 percentage points a year.

Under current calculations, according to the Social Security Administration, a worker with an average wage rate who retired in 2002 at 65, then collected Social Security checks until she died at 82, would receive $318,000. If the index were cut by 1.1 percentage points, as the Boskin Commission recommended, that retiree would get $277,000, or $41,000 less. For the government, of course, such a reduction in the index would mean savings of billions of dollars.

But the index will always be flawed, said Joel Best, a professor of sociology and criminal justice at the University of Delaware and author of "Damned Lies and Statistics: Untangling Numbers From the Media, Politicians, and Activists." The index, like many other statistics, is "meant to distill complexity into some relatively simple measure," he said.

 
While the bureau may be moving faster to change its data, its methods for gathering the data remain stuck in the nondigital past. When Dr. Boskin's commission was formed in 1995, he recalled, he traipsed after a bureau pricing agent as she went through a Nordstrom department store in Northern California looking at the tags on dresses and writing down the prices on paper. Now, six years later, even though legions of FedEx (news/quote) and U.P.S. deliverers have long been using handheld computers, bureau workers are still out in the field with paper and pencil.

The bureau is phasing in handheld computers; all pricing agents will have them by 2003. But it is unclear if or when the bureau will begin using another common technology — the bar-code scanner — that has revolutionized the way retailers and manufacturers follow price movements.

This pace of change concerns Dr. Boskin. The bureau has "taken some steps," he said. But the inflation rate, he added, "is a moving target."