U.S. to Change Way It Measures Inflation By John M. Berry Washington Post Staff Writer Friday, April 17, 1998; Page A01 The government's main inflation gauge will be changed to account for the fact that consumers respond to the rising prices of many items by shifting to lower-cost substitutes, the Bureau of Labor Statistics announced yesterday. The announcement marked the end of a series of changes in the consumer price index that were begun in 1995 to correct for the CPI's tendency to overstate the actual rise in the cost of living. The changes affect the vast majority of Americans and the budgets of governments at all levels because the CPI is used to determine cost-of-living adjustments in benefits, such as Social Security, and for adjusting features of the income tax such as standard deductions and tax brackets. When the change announced yesterday is incorporated into the CPI next January, the cumulative effect of all the revisions will be to trim the annual increase in the index by about eight-tenths of a percentage point. The latest change, which will shave about two-tenths of a percentage point from the reported inflation rate, is designed to reflect shoppers' natural tendency to alter their purchasing patterns when the price of an item rises faster than an available substitute. At a press conference, BLS officials cited the example of what typically happens when the price of a particular brand of ice cream goes up: Many consumers will either switch to another brand, buy a different-sized package, go to another store or switch to a close alternative, such as frozen yogurt. The CPI now is based on changes in the cost of a fixed market basket of goods and services and assumes that consumers don't switch to a less expensive alternative when they have the opportunity. The new procedure for calculating the index is based both on economic theory and actual changes in consumer spending patterns in response to price changes as shown by checkout scanner data from supermarkets, BLS Commissioner Katharine G. Abraham said. "The evidence we looked at suggested that consumers are sensitive to price changes," Abraham said. The latest changes build on previous modifications in the CPI aimed at improving its accuracy in accounting for inflation. These include, among other things, updating the market basket to reflect more recent information about actual consumer spending patterns, improving the tracking of hospital charges, and measuring rents more accurately. The changes have been implemented amid a heated debate in recent years about whether the CPI was overstating inflation. The BLS began researching problems with the index in the early 1990s. In late 1996, a commission headed by Michael Boskin of Stanford University concluded that the CPI was overstating the actual increase in the cost of living by more than a percentage point a year. Although many analysts challenged the commission's findings, economists broadly agree that there has been an upward bias in the CPI. In effect, the changes mean that retirees whose benefits are linked to the CPI were routinely getting larger COLA'S than those to which the rising cost of living entitled them. Similarly, taxpayers were paying less in taxes than would have been the case with a more accurate measure. Thus, the overstatement of inflation worsened the government's budget problems by increasing spending and reducing revenue. The revisions in the index will improve the government's finances while causing benefits to rise more slowly than they otherwise would. In January, an average retiree who last year had a monthly Social Security benefit of $749 got a cost-of-living increase of about $16 a month, or $192 a year. That was the result of a 2.1 percent rise in the CPI from the third quarter of 1996 to the third quarter of last year. If the rise in the CPI over that period had been 2.4 percent, the likely figure had the BLS not begun changing its methods of calculating the CPI, the increase in the monthly benefit would have been about $19 instead of $16. For all of this year, that would have been $228 rather than the actual $192. Over time the differences grow. Assuming the CPI rises at 2 percent annually for the next three years, for example, by 2001 monthly benefits would be about $20 less than if the CPI were calculated using the old procedures. The change announced yesterday affects all but 15 of the 211 categories of goods and services in the CPI market basket. But the 15 items are so important that they account for almost 30 percent of the total index. These include housing, most utilities and some government fees such as automobile registration charges. The BLS decided not to use the new technique for these because consumers don't have ready access to a substitute, at least not without making a major change in their lives by, say, moving to another location. © Copyright 1998 The Washington Post Company Subject: Re: Boskin Report had it all wrong Date: Tue, 13 Apr 1999 07:04:02 GMT From: jfo@feri.de (John F. Opie) Organization: FERI GmbH Newsgroups: alt.politics.economics, alt.society.liberalism, alt.fan.rush-limbaugh References: 1 , 2 On Tue, 06 Apr 1999 12:24:28 -0500, Jim Blair wrote: >The basic form of consumer price indices have not changed since >1921, or since 1950. > >.... > The idea that the Consumer Price Index over states inflation, > and that COLA's based on it overcorrect for inflation has been > taught in basic economic classes since at least the 1950's when > I was in ECON 1. > ... >This from the Econ 100b course currently being taught at U Cal >Berkeley: > >QUOTE: > >Biases in measurement? Yes: > >The CPI overstates rates of price increase by 1/2 to 1.5 percent per year >(and the GDP deflator has similar biases). We are getting richer and more >productive faster than the official numbers report. > >UNQUOTE ... >Note that the bias is introduced in part because the "fixed basket of >goods" concept does not account for substitution, or for new goods >introduced since the base was established. Updating the contents of >the basket more frequently has the effect of reducing the magnitude >of the bias, but not its direction, or its existence. Hi Jim - I thougt i'd just toss this into the fray: The methodology of calculating the basket of goods has changed. Chain-weighting is now used: that means that the basket changes every June. There is no now longer a five-year revision every time the basket changes (which used to be every five years) but rather a one-year revision when the new basket is determined. A preliminary new basket is then used for a year, then checked against actual consumption, and revised for the year. Thereafter there is no further revision. Which means, if you really want to be pedantic, that you actually can't compare CPI from year to year, since it now changes in its content. What you can compare is what it costs for average consumption, but the contents of this consumption change every year. Which makes the CPI even less useful a tool for indexing increases, for instance, in social security, since the basket of consumption for retired folks is a different basket than general consumption. This, however, addresses only in part the criticism of the Boskin report. One of the major problems that I see is the way that data is collected: individual stores report their price changes over time. What is not reported is the shift in volume from one kind of retail outlet to another: if a small town receives a Costco or a Walmart on the town fringes, prices actually paid generally plummet for things like consumer electronics and the like. However, this will not show up in the CPI, since the Walmart prices don't change - they were already low - and the local store prices don't change - they have higher costs and can't afford to drop the price, even in the face of competition. What happens is that the consumer shifts his preference from local store to Walmart/Costco or whatever, and ends up with a price reduction which never shows up on the CPI. Add to this the entire problem of quality changes and the CPI remains flawed, flawed, flawed. But the BLS would have to hire about 20 statisticians and economists and it would probably take about 3-5 years to build up a set of "proper" consumer price deflators. John