How Reliable Is The Consumer Price Index?

The consumer price index is fast becoming the Rodney Dangerfield of economic statistics. It just doesn't get any respect these days. Indeed, taking potshots at the CPI, the most often used inflation measure, has become a favorite sport among forecasters. "The monthly CPI report should be like a cigarette package," says Lawrence A. Kudlow, chief economist at Bear, Stearns & Co. "Warning: Excess confidence in this index can be damaging to your investment health."

One reason economists are so critical may well be that the CPI, compiled by the Labor Dept.'s Bureau of Labor Statistics, hasn't behaved as expected during this recession. Even with March's small 0.1% gain in the CPI excluding food and energy, the "core" rate of inflation is rising at a rapid 5.5% from a year ago. But economic theory holds that the inflation rate should slow during a recession.

SHOPPING SPREE. Does the CPI deserve all this disrespect? To find out, this writer spent a day with the BLS with two questions in mind: How accurately does the government pick up price changes, and more important, does the data collection bear close resemblance to the real world where most of us spend our money? The answers seem to be that the BLS does a good job on the accuracy front. But improvement could be made to reflect how most consumers actually shop.

On Apr. 11, I paired up with Rena Brunner, a BLS field economist, who has spent the past 15 years checking prices. For three weeks of every month, Brunner and 350 other data collectors visit the nation's shopping malls, hospitals, and other suppliers of consumer goods and services to compare price changes from month to month. On any given day, a BLS worker can gather 30 to 70 different prices, but on this day, Brunner will collect prices for about a dozen items representing categories that account for about one-third of the CPI.

Collecting data for 95,000 items may seem overwhelming. But there is a method to this pricing madness. The BLS surveys consumers every year on what they buy and every four years on where they buy it. And because these two reports show that many people buy produce in the C-Town market in Jersey City, N. J., Brunner starts the day there, checking the fruit and vegetable prices.

Brunner inspects a display of limes where a sign says "3 for 99~." That's the same as in March--so there's no inflation, right? Not necessarily, Brunner explains. "We always weigh products that are sold as units just to see if they're smaller or bigger." And sure enough, three limes now weigh in at 5 ounces, compared with March, when they tipped the scale at 8 ounces. So instead of no price change, the cost per ounce of lime has skyrocketed by 60%.

NO SWITCHING. The visit to the produce section shows, however, a major shortcoming of the CPI. The index is a fixed-weight index. That is, each item carries the same importance from month to month, and none can be substituted for another product in the survey.

Because items cannot be switched, the CPI misses an important way that consumers deal with inflation: They simply don't buy a product that becomes too expensive. For instance, Brunner only checks the price of Macintosh apples. But a consumer might buy the Golden Delicious variety if it's cheaper. Same fruit, less cost. The CPI's rigid product mix is one reason why the price deflator used for consumer spending--which captures substitution--is up by just 4.5%, while the CPI has accelerated (chart).

Another criticism of the CPI is the use of list prices. By haggling or offering to pay cash instead of using a credit card, shoppers can often get a discount from the list price, and detractors say the CPI misses this. But a trip to a car dealer shows that the BLS does, indeed, try to capture the art of the deal.

At DiFeo Buick in Jersey City, Brunner checks the prices of three models. First she compares the March and April list prices for a basic car and selected option packages. Then Brunner asks Sales Manager Robert Caroselli how much of a discount buyers have been getting. Caroselli gives estimates for each model, and all seem more generous than the concessions made in March.

Brunner says that car dealers are a source of frustration for data collectors because there is no way to verify the accuracy of the discount estimate. Each customer may buy the same car at a different price. Caroselli admits that's true. "A regular, loyal customer or someone with a trade-in may do better," he says. And especially in today's sluggish market, he adds, "we'll do whatever we have to do to make a deal."

CHIC BOUTIQUES. After the car dealership, Brunner heads to a discount store to price infant apparel. Clothing items have strong seasonal patterns, and in April, prices are taken for spring/summer baby's sleepwear, which is on sale at 30% off the March price. The other survey item, a sleeper for autumn and winter, is already off the shelf, however. Brunner won't start looking for that item until August, when the store will roll out its colder-weather fashions.

Pricing baby clothes is, of course, necessary, but I'm a bit more interested in the inflation rate for women's apparel. Gathering these prices is trickier because of the wide range of retailers and the variety of styles. Brunner says the survey covers chic boutiques, department stores, and discount outlets. And since retailers change women's clothing styles frequently during the year, it can be difficult to track the price of a specific dress. Instead, a data collector may look for a dress with the same cut and in the same fabric. This ensures consistency between the monthly reports, but it doesn't mimic the behavior of shoppers who may opt for a cheaper wool-blend dress over a pricier pure-wool outfit.

Even with the problems of fashion styles or new-car discounts, the price of goods poses less of a challenge for the BLS than does the task of pinning down service prices. One reason, which economists have harped on, is that the BLS does not adjust the changes in prices that come from improvement in the quality of services, especially medical care, which is one of the fastest-growing components of the CPI.

At the Jersey City Medical Center, Brunner checks the prices for certain lab tests and emergency-room procedures, which haven't changed much since March. She explains that the numbers picked up by the BLS are costs set by the government's diagnostic-related group (DRG) plan. That program was started in order to rein in soaring medicaid and medicare expenses and is now used by private insurance companies. The problem is that DRG prices change only once or twice a year, while the hospital's labor and other costs may be rising. More important, these prices don't show if the patient is picking up a larger share of the tab in the form of higher deductibles. So the DRG prices don't reflect the costs that most consumers will pay or the costs to the hospitals.

In services, however, no part of the CPI has come under as much fire as the housing component, especially the segment called homeowners' equivalent rent, which accounts for 19% of the total index. Since about two-thirds of the population own the home they live in, determining changes in the costs of running a home is key to measuring rises in the overall cost of living. The BLS once used mortgage payments when calculating housing inflation. But in the 1970s, the acceleration in home prices and interest rates caused mortgage payments to skyrocket, sending the entire CPI upward. The problem was that only a small percentage of households were actually taking out new mortgages. Instead, the BLS adopted equivalent rent, which estimates what a homeowner would pay to lease a home.

To calculate the equivalent rent, a house occupied by its owner is matched up with a rental unit with similar characteristics, such as same neighborhood and equal number of rooms. Field operators such as Brunner then call the renters to find out what this month's rent is. The reported rent from the leased housing is then used to determine what rent an owner would pay for his or her home.

Because most leases are usually set once a year, only a small proportion of rents are adjusted in any given month. That means it takes time for changes in the housing market to show up in BLS numbers.

So how is the April CPI shaping up? The prices checked by Brunner are only a few of the thousands included in the CPI. And each item must be combined with prices from other regions and then seasonally adjusted by analysts in Washington. But the bigger discounts in new cars and the bargains to be had in infant wear suggest that the recession has finally slowed inflation. The April data, to be released on May 14, may post a small gain of about 0.2%. But I'm still a little worried about those soaring lime prices.

Before it's here, it's on the Bloomberg Terminal.