Explaining the US Consumer Price Index

The purpose of this brief is to explain some of the methods used to construct the CPI-U and why, in some cases, the index's estimates of inflation may differ from consumers' perceptions of how much prices are rising.² The brief focuses on six aspects of the CPI-U's construction: averaging regional price indexes to create a nationwide index; estimating the expenditure weights that BLS assigns to the major categories of prices in the CPI-U to account for the categories' relative importance; allowing for shifts in relative prices, a phenomenon known as economic substitution; adjusting for changes in the quality of various goods and services; measuring prices for medical care; and measuring prices for shelter.

Two criticisms of the CPI-U are not discussed. One is the necessary limitations of the index: Despite its broad coverage of prices, the CPI-U does not track the cost of many factors that affect a person's well-being—for example, crime, traffic, pollution, the prevalence of disease, the quality of education, and civil liberties. Cost-of-living measures rarely include such factors because few market prices or consumer expenditures are associated with them, making it essentially impossible to define and measure changes in their price or value.

To develop an overall measure of the change in prices, BLS must assign a weight to each price category in the CPI-U to account for the categories' relative importance. The weights represent the average share of total consumer expenditures on a particular good or service and are calculated by using data from BLS's Consumer Expenditure Survey. If consumers, on average, spend a relatively large amount of their annual income on a particular good or service—food, for example—BLS assigns a large weight to the price of that good or service in the CPI-U.⁵

Because the expenditure weights are averages, people whose patterns of spending differ greatly from the average might face changes in prices that are larger (or smaller) than those the CPI-U indicates. Elderly people, for example, usually spend more of their annual income on medical care than the average consumer does. As a result, if the cost of medical care (that is, care not covered by insurance or paid for by the government) shot upward, the elderly might feel that inflation was higher than the CPI-U indicated.⁶

BLS's procedures implicitly allow for substitution within most basic categories of goods and services in the CPI-U—such as substituting round steak for sirloin—which helps keep the index from overstating a rise in the cost of living.⁷ But some parts of the CPI-U, such as rents, certain utilities, and medical services, do not incorporate the procedures because consumers cannot easily substitute one good for another within those categories.

When economic substitution occurs across basic categories of goods and services—when consumers can partially offset an increase in the price of beef by buying more chicken—the official CPI-U misses it and overestimates inflation. To address that issue, BLS in 2002 introduced a CPI measure, the chained consumer price index for all urban consumers (chained CPI-U), that can pick up substitution across categories. BLS cannot reliably calculate that index, however, until data on consumer expenditures for an entire year are available, so it publishes the final calculation of the chained CPI-U more than a year after it publishes the official CPI-U. From 2002, when the methods currently used for calculating the official CPI-U were put in place, to 2005, the last year for which the chained index is not subject to revision, the average growth of the chained CPI-U was about three-tenths of a percentage point slower than that of the official CPI-U.⁸

BLS uses a number of statistical procedures to adjust prices in the CPI-U for shifts in the quality of goods and services over time.⁹ By removing the effects of such changes, BLS can estimate pure price change—the change that occurs in a good's or service's price if its quality is held constant. Some of the prices in the CPI-U that are adjusted for improvements (or degradations) in quality include those for new and used vehicles, computers, apparel, various appliances and electronic equipment, shelter, and medical care.

Quality adjustment requires BLS to identify material changes in an item and then determine how much those changes contribute to the overall change in the item's price. In the case of medical care, for example, suppose a dentist raises the price of a standard office visit because she decides to include an additional service, such as a fluoride rinse, that before might have been optional and priced separately. BLS determines the value of the added service and subtracts it from the change in the price of the office visit that it has identified through its survey. In the case of costs for shelter, suppose that rents increase because the average unit that BLS uses for its survey is bigger than it was in the past. To calculate the pure price change, BLS identifies the value of the added square footage and subtracts it from the price change it has recorded.¹⁰

Estimating pure price change is a more accurate way to measure changes in the cost of living than simply observing changes in market prices. Take the case of a new appliance that is more expensive than an older model but uses less energy. Switching to the new model could increase the cost of living because the new model costs more to buy than the old one did. But the value of the energy that the new model saves might offset the higher purchase price—in which case the new model will not increase the consumer's cost of living. BLS thus factors in energy efficiency as well as appliances' purchase prices in estimating changes in living costs.

Because BLS adjusts prices for changes in quality, the price of a good or service in the CPI-U whose quality is improving declines more quickly or grows more slowly than it would if those adjustments were not being made. For example, BLS has estimated that prices of computers fell by about 9.8 percent (on an annual basis) between March and September 2004. Without BLS's adjustment for improvements in the quality of computers, the drop in prices is about 6.6 percent.¹¹

No completely satisfactory way has been found to measure health care prices in a cost-of-living index. Rapid changes in quality and the large role that insurers play hamper BLS's measurements, and its estimates of underlying prices are not without controversy. That uncertainty adds to the growing concerns of consumers and policymakers about the rising costs of medical care.

The CPI-U aims to measure the prices of all health care that consumers pay for, either out of their own pockets or through a third party (such as a private insurer or government program) if the consumer pays a premium or some kind of fee—for example, a copayment. BLS uses the results of its surveys to directly measure the prices that consumers pay for care they purchase out of pocket; however, it must measure prices for health insurance indirectly, because a direct approach—attempting to measure the cost of the premiums that insurers charge consumers—raises insuperable issues of quality adjustment. Insurance products may differ widely, and a single product may vary from year to year. Thus, the annual change in the premium that an individual pays will reflect both changes in the price of insurance and changes in the amount or quality of the insurance being purchased.

Disentangling changes in price from changes in quality requires information that insurers have not so far been able to provide, and BLS therefore surveys doctors and other health care providers to determine the prices that they charge insurers. Those prices are assigned the greatest weight in the health care price index. The CPI-U also includes an estimated price for health insurers' overhead (which includes the cost of administration and of maintaining reserves plus profits). That price is assigned a relatively small weight.¹³

Prices for goods and services that are not paid for by the consumer out of pocket or through premiums or fees are excluded from BLS's price indexes. Thus, prices for medical goods and services that are covered by Medicaid or Part A of Medicare (the Hospital Insurance program) are not part of the index because the government pays for all costs associated with those programs. In contrast, prices for goods and services covered by Parts B, C, and D of Medicare (the Medical Insurance, Medicare Advantage, and prescription drug programs) are included because consumers pay premiums or fees to participate in the programs.