FRBSF Economic Letter
1997-11 | April 18, 1997
The Consumer Price Index, our most common measure of consumer inflation, has been the subject of controversy recently. Most of the headlines reflect the debate about whether the CPI overstates inflation, but there are other disagreements about this measure as well, especially in the context of monetary policymaking.
This Economic Letter focuses on the debate over whether policy should respond to inflation in all goods and services or whether it should pay particular attention to the measure that excludes changes in food and energy prices — often described as the “core” inflation rate. In addition, the Letter discusses an alternative method of estimating underlying inflation and how using this method might affect policy.
A common argument for ignoring changes in food and energy prices is that although these prices have substantial effects on the overall index, they often are quickly reversed and so do not require a monetary policy response. Food and energy prices comprise almost one-fourth of the weight in the CPI, and their price changes do tend to be volatile; for example, between 1983 and 1996, the standard deviation of monthly changes of the all-items CPI was 0.18 percentage point, whereas that of the core CPI was only 0.12 percentage point.
A second argument is that changes in energy and food prices often are due to supply shocks, such as cutbacks in oil exports by the OPEC cartel or unusual weather that affects harvests. Such supply shocks affect the levels of these prices, but not necessarily their long-run growth rates. Hence, if the goal of policy is to control inflation, monetary policy might forgive these one-time changes in the price level. Also, negative supply shocks typically show up as increases in average prices accompanied by declines in employment and output. Some might argue that it may be inappropriate to raise interest rates to subdue inflation when output and employment already are low.
If monetary policy were to follow a regime in which changes in short-term interest rates were guided by the core inflation rate, a food or energy supply shock would provoke little or no policy response. As a result, the shock would be passed through into a higher price level, since there is no mechanism to drive other prices down when food or energy costs rise. Thus, by choosing such a regime, monetary policymakers essentially would be deciding that shocks to food and energy prices could be routinely accommodated.
A general argument against a procedure that systematically excludes “volatile” prices is that prices that move quickly may provide useful information about future trends that is not reflected in prices that change more slowly. That is, a rise in aggregate demand that might set off a period of higher inflation may initially show up in increases in certain sensitive prices that are set in more competitive markets. If these prices are ignored because they are “volatile,” these early signals of inflation may be missed.
Moreover, the argument that volatile prices should be ignored does not necessarily imply that food and energy prices should be systematically excluded. Instead, it implies that unusually big (or small) price changes should be excluded wherever they arise, and not only when they are in the food or energy segment. In fact, not all food and energy prices are volatile, and not all volatile prices are in the food or energy sectors. In a study of price variability since 1983, I found that the prices of home and vehicle fuels were very volatile, but those of other energy items, such as gas and electricity, were less so. Similarly, most of the volatility of food prices came from variations in fruit and vegetable prices, with much less variation in other food categories. Several price series that are included in the core CPI were at least as volatile as the food and energy price series. These findings suggest that if the aim is to construct a CPI series that excludes volatile items, simply removing all energy and food items may not be the best way to proceed.
The argument that changes in food and energy prices should be accommodated because they reflect supply shocks is potentially suspect because not all food and energy price changes are the result of supply factors and because supply shocks do not occur only in the food and energy sectors. Although several major supply shocks in the past have affected food and energy items — and have been used to justify purging these items — we cannot tell that this will continue to be the case. In the future, there almost certainly will be supply shocks to other categories of prices. Nor can we be sure that most changes in food or energy prices will be due to supply factors that have only one-time effects on their prices. The basic problem is that economists do not have good models that enable them to separate the effects of supply shocks from those of demand factors that are likely to add to long-run inflation.
Bryan and Cecchetti (1993) and Cecchetti (1996) have examined methods of removing noise from monthly price data. They argue that monetary policy should seek to control the underlying trend rate of inflation and should ignore the effects of large outliers on measured inflation, regardless of where these outliers arise.
One simple method of reducing the effect of outliers is to average data over periods longer than one month. Policymakers already do this informally: policy rarely changes in response to a single monthly inflation number. Another procedure is to eliminate outliers in each month. Bryan and Cecchetti suggest measuring inflation in any month by the median of monthly changes in prices rather than by the average change. The median price change is the one such that half of all price changes are larger and half are smaller. An alternative to the median is to compute a “trimmed mean,” in which a chosen proportion of unusually large and small price changes is excluded before the average is computed. For example, using a 15% trimmed mean reduces the standard deviation of monthly changes in the CPI to .09 percentage point. Since such “limited influence” estimators of inflation are not affected by unusually high or low values, they may come closer than the standard measures to indicating the underlying trend in prices.
Cecchetti (1996) tested this possibility by performing a simulation experiment that assumes that trend inflation is measured by a three-year centered moving average of the all-items CPI. He found that the average deviation (measured by the root mean squared error) between monthly inflation and trend is reduced substantially by using a limited influence estimator of monthly inflation rather than either the all-items or the core CPI measure. This experiment suggests that a limited influence measure would more closely reflect the underlying trend of inflation than the conventional indexes. Also, a limited influence measure would have the important advantage of being agnostic with regard to the sources of volatility rather than attempting to remove only price changes associated with food and energy shocks.
It is possible to reduce volatility further by averaging the limited influence estimator over several months. Cecchetti finds that precision is most enhanced by focusing on a three-month moving average of the trimmed mean. This measure of inflation is substantially less volatile than the standard monthly indexes. Between 1983 and 1996, the standard deviation of monthly changes in this three-month average is .07 percentage point.
The accompanying figure shows the volatility of the all-items CPI, the core CPI and a 3-month moving average of the 15% trimmed mean (labeled Limited Influence); following Cecchetti, trend inflation is measured by the centered 36-month moving average of CPI inflation and shown by the thick black line. Clearly, the limited influence indicator is less volatile around the trend than the other measures; thus, it might be a more reliable monthly indicator of trend inflation than either the overall or core measures. However, the limited influence series diverges from the trend from time to time. On several occasions (mostly in the 1970s), the trimmed mean lags behind the trend at a turning point. This lag may be because the limited influence estimator leaves out useful information. This reminds us that an inflation measure that excludes some prices because they are volatile risks missing early signals of increasing inflation.
The volatility of the CPI presents a genuine dilemma for policymakers. On the one hand, some short-run movements in the index can contain useful information about incipient inflation; on the other hand, many short-run movements are soon reversed, and they can mask the underlying trend of inflation. Beyond the dilemma of whether to “smooth” the data is the dilemma of how to do it. Both the core measure and the various limited influence estimators have been less volatile than the all-items CPI. However, the differences in volatility between these alternative indexes are not large. Thus, the choice has to be made on conceptual rather than empirical grounds. In principle, it may be appropriate for policy to respond differently to one-time supply shocks than to demand shocks. This is the justification for focusing on a measure that excludes food and energy prices. Unfortunately, however, except in extreme cases, it is unlikely that we shall ever find ourselves in the happy position of being able to identify the effects of contemporaneous supply shocks with full confidence. We may be kidding ourselves if we think that focusing on the core CPI will remove most supply shocks while leaving the effects of demand shocks in the data. Thus, a limited influence estimator of inflation may be a superior smoothing device to a method that simply strips out all food and energy price changes, because it does not prejudge which types of shocks will be removed from the data.
Bryan, Michael F., and Stephen G. Cecchetti. 1993. “Measuring Core Inflation.” NBER Working Paper No. 4303.
Cecchetti, Stephen G. 1996. “Measuring Short-Run Inflation for Central Bankers.” Paper prepared for Economic Policy Conference, Federal Reserve Bank of St. Louis (October).
Opinions expressed in FRBSF Economic Letter do not necessarily reflect the views of the management of the Federal Reserve Bank of San Francisco or of the Board of Governors of the Federal Reserve System. This publication is edited by Sam Zuckerman and Anita Todd. Permission to reprint must be obtained in writing.
Please send editorial comments and requests for reprint permission to
Attn: Research publications, MS 1140
Federal Reserve Bank of San Francisco
P.O. Box 7702
San Francisco, CA 94120