May 2000
Federal Reserve Bank of Dallas

Measuring and Detecting Inflation

Introduction
The past couple of years have been challenging times for monetary policy makers. Inflation and unemployment rates have fallen to levels not seen in thirty years. Output growth has exceeded all expectations, making the U.S. economy the envy of the world. There is an active debate about whether the economy has entered a new era, one in which the conventional wisdoms that have guided macroeconomic policy for so long cease to apply. Some argue that the Fed has been too slow to respond to incipient inflation pressures, contributing to imbalances that threaten to bring the current expansion to an end with a crash. Others argue that the computer revolution has raised the economy's growth potential, allowing the Fed to pursue a low interest rate policy for longer than might otherwise be the case.

In arguing for the latter view, President McTeer has asserted that the best place to look for inflation is in the inflation statistics. In my presentation this morning I want to give you some background on how exactly inflation comes to be manifested in the inflation statistics. I will review in some detail two of the major inflation measures used as inputs in the policy making process, the Consumer Price Index (CPI) and the Personal Consumption Expenditures Deflator (PCE) and explain the key differences between them. I will also review the concept of core inflation, and show you how headline and core measures have behaved recently.

Rates of Change of Selected Prices
In an economy as large and advanced as that of the United States, where almost all economic transactions take place on private markets, there are literally trillions of prices agreed upon by buyers and sellers each day in the course of doing business.

Figure 1 plots the price changes for a select group of goods to give you some sense of just how dramatic are the differences in the rates of price change experienced by different items. In some years the price of gasoline increases dramatically, in others it declines. These movements primarily reflect political developments in the Middle East and the ability of OPEC to restrain output.

Likewise coffee prices can show dramatic swings from year to year. Crop failures in Brazil in 1994-95 and again in 1997–98 caused prices to increase by nearly 60 percent.

However, computer prices are driven by the pace of productivity growth in the computer manufacturing sector, and in recent years productivity has advanced at a pace so rapid as to lead to average annual rates of decline in computer prices of about 15 percent. In not one single year over the past decade have computer prices increased.

By contrast, although it is difficult to see in this figure, the prices of prescription drugs have increased every year since 1991, and in some years at rates in excess of the overall rate of increase of the CPI.

What Is Inflation? The Fed is not so much interested in how any individual price is changing over time as it is in how the price level is changing. The price level is the average price of all goods and services purchased over some period of time, for example, a month or a quarter. The rate of inflation is the rate of change in the price level from one month to the next, or one quarter to the next, and is an average of all price changes occurring between the two months or quarters.

It would be prohibitively costly to try to measure every single price change that occurs each month. Instead, government statisticians produce a variety of price indexes that are designed to address specific questions and that serve to guide discussions of monetary policy.

The oldest of the price indexes produced by the government is the Producer Price Index, or PPI. This index has been constructed since 1902, and measures the average price received at the factory gate by U.S. producers for their output. Analysts study the PPI report because it is the first inflation statistic released each month. However, the PPI is oriented primarily towards the goods-producing sectors of the economy, limiting its usefulness as a gauge of overall inflation pressures.

The most closely followed of the major indexes is the Consumer Price Index, or CPI. The CPI measures the rate of price increase experienced by the average urban consumer in the U.S. A broader measure of inflation at the consumer level is the so-called Personal Consumption Expenditure Deflator, or PCE.

The broadest measures of inflation are the deflators for Gross Domestic Product and Gross Domestic Purchases. These measures provide the most comprehensive assessment of what is happening to inflation in the U.S. economy. However they are only available at a quarterly frequency, which limits their ability to reveal in a timely manner information about changing inflation trends.

In what follows I want to focus on the measures of inflation at the consumer level. I have two reasons for this. First despite some obvious limitations, the CPI is the most closely watched of the government's inflation statistics. It is used to adjust Social Security benefits each year and a host of other expenditures. Many private contracts are indexed to the CPI, and it gives us a good indication of how inflation is affecting the average consumer. Second, the PCE is probably the most accurate of the government's inflation measures, yet does not seem to be widely understood outside the community of economists and central bankers. The PCE is the preferred inflation measure of Chairman Greenspan, and if nothing else I would like you to leave this presentation with an understanding of why this is. But before discussing the competing merits of the CPI and the PCE, I would first like to lay out some of the basic principles of price measurement.

Principles of Price Measurement
Given the impossibility of recording every single price change, the choice of which prices to measure is usually guided by posing a question that the resulting inflation statistic will help us answer. For example, we might want to know how much more money the average worker needs this year to be able to buy the basket of goods he purchased last year. If this is the question we want our inflation measure to answer, then we can narrow the set of prices we look at to those that impact household's budgets the most. We will want to know how much the cost of food, housing, and apparel has increased, and we will not be interested in how much the prices of industrial machinery or iron ore have risen.

After it has been determined which prices we will look at, the greatest challenge faced by statisticians is to ensure that they compare like with like. If the collection of goods and services consumed by the average worker stayed fixed over time this would not be a problem. However, in a dynamic economy, old goods are constantly disappearing, new goods are always being introduced and existing goods are always being improved.

At the most basic level, this requires that we measure the price per unit of what the consumer is interested in and not be fooled by differences in package size. Paying $10 for a 5 ounce martini as opposed to $5 for a 2½ ounce martini does not constitute a price increase.

A more difficult problem is controlling for differences in quality. When new model cars are introduced each year, the statisticians need to figure out how much of the price difference between the new and old models reflects quality improvements, and how much reflects inflation.

The final problem is to find some way of combining the detailed price information into a single inflation statistic. If we are interested in knowing how much more income the average worker would need this year to maintain last year's living standard, we will have gathered the prices of the various goods and services that the average worker buys. To summarize how these prices have changed, an obvious strategy would be to look at the average price change across all of these goods and services. An alternative is to assign an importance or weight to each individual price change based on the importance of that item in the household's budget.

The Consumer Price Index
The CPI is the most closely watched of the major price indexes published by the federal government. The CPI measures the average change over time in the prices paid by urban consumers for a specific basket of goods and services.

Each month Bureau of Labor Statistics (BLS) field agents collect about 80,000 individual price quotes from different retail outlets, doctors' offices, etc., around the country. The choice of cities in which to collect price quotes is determined by the distribution of the population; the choice of outlets from which to collect prices is determined by surveys of where households shop.

Because the focus of the CPI is on the inflation experience of consumers, it only includes the prices of business to consumer ("B2C") sales. The prices of business to business ("B2B") sales are not included.

The aggregate CPI number is calculated by weighting the individual price series by the share of the associated item in the expenditures of the average urban household in 1993-1995. Because these weights are held constant for long periods of time they become less representative of household 4 spending patterns. Goods and services that have experienced less rapid inflation and have become more important in households' budgets will be underrepresented in the index, while goods and services that have experienced more rapid inflation and have become less important in households' budgets will be overrepresented in the index. The net effect is to cause the CPI to overstate the true rate of inflation. Four years ago the Boskin Commission, appointed to assess the accuracy of the CPI, estimated that the use of fixed weights caused the CPI to overstate inflation by about half a percentage point a year.

Figure 2 shows the behavior of the CPI over the course of the current expansion. We see the pickup and decline in CPI inflation around the time of the Gulf War, and the subsequent stabilization through the middle of the decade. In 1997 and 1998 inflation declined further to less than 2 percent, but has since reversed course. Three weeks ago the BLS reported that in March of this year the CPI was up 3.7 percent, the highest it's been since the summer of 1991.

The PCE Deflator
The other principal measure of inflation at the household level is the PCE deflator. The PCE deflator is not a completely independent measure of inflation as it is built up from components of both the CPI and the PPI.

There are two major differences between the CPI and the PCE deflator. First, in terms of coverage. The CPI is only supposed to be representative of the price paid by urban consumers in the U.S. Thus it does not capture prices paid by rural consumers. The PCE deflator, on the other hand, is supposed to be representative of the prices paid by all consumers, urban and rural.

However the crucial difference between the CPI and the PCE has to do with how the individual prices series are combined in the PCE deflator. Specifically, the PCE deflator takes account of shifting spending patterns, as opposed to the CPI which relies on spending patterns from several years ago.

Recall that the CPI uses historical information on spending to add up individual prices. This means that as time goes by, goods that have become cheaper will tend to be underrepresented in the aggregate numbers, causing inflation to be overstated.

An alternative approach to combining the individual price changes is to use current spending patterns to add up individual prices. But this too has its problems. By using weights based on current spending patterns, goods and services that have become cheaper over time will be overrepresented in the index, while those that have become more expensive over time will be underrepresented. The net effect will be to cause the index to understate the true rate of inflation.

The obvious solution is to somehow use information on both this year's and last year's spending patterns. By averaging measures that overstate and understate inflation we ought to get closer to the true measure. This is exactly what the PCE deflator does, and accounts in part for its popularity with Chairman Greenspan as a measure of inflation.

Here we see the behavior of the PCE deflator over the course of the current expansion, along with the CPI (figure 3). Note that both measures show roughly the same trends in inflation over the past decade. They both capture the acceleration prior to the Gulf War, the subsequent decline and stabilization through the middle of the decade, the further decline in 1997-98 and the acceleration since late 1998. Note also that the PCE measure of inflation is almost always less than the CPI.

Measures of Core Inflation
So far I have talked about the headline measures of inflation. There are also a number of measures of what is termed "core" inflation constructed by the federal statistical agencies (and others) and used as inputs to the policy making process. Core measures strip out price developments that are believed to contain too much noise to be of any use in assessing underlying trends. A dramatic run up in the price of some commodity that looms large in households' budgets, but is likely to be reversed in the near future, probably does not tell us a lot about whether the underlying inflation picture has changed.

The best known of the core measures are the so-called Ex. Food and Energy measures, which are identical in all respects to the headline measures except that they exclude the prices of food and energy commodities. That is, the core measure is constructed by assigning zero weight or importance to price developments in the food and energy categories. This practice began in the 1970s when the industrial economies experienced large oil and agricultural commodity price shocks.

Another measure of core inflation is the so-called median CPI. The motivation for looking at this measure is that it may be misleading to always assign zero importance to food and energy prices. Some months it may be that other prices experience dramatic movements that bear no relationship to the underlying trend. So rather than delete food and energy prices every month, the median CPI deletes the biggest and the smallest price changes. The intuition behind this measure is not unlike that behind the scoring in ice skating competitions: the points awarded to each competitor are an average of the points awarded by each judge, excluding the highest and lowest.

Figure 4 shows the behavior of both headline and core CPI inflation over the course of the current expansion. Note that the measure of core inflation that is based on excluding food and energy prices is a lot less volatile than the headline measure. The deceleration in the core measure after the Gulf War is more gradual than the deceleration in the headline measure. In recent years the core measure is more stable than the headline measure: it does not decline as much in 1997-98, nor does it increase as much in 1998–2000.

The median measure is somewhat more volatile than either the headline CPI or the CPI excluding food and energy. Despite this, there is research showing that the median may sometimes do a better job than the traditional measure at detecting changes in the direction of inflation.

The government also produces a core measure of the PCE deflator, based on the exclusion of food and energy prices. However this measure is only available on a quarterly basis. Figure 5 shows how this quarterly measure of core PCE inflation compares with the monthly core CPI inflation rate. Note that as with the headline measures, the core CPI and PCE measures move in tandem over time. The core PCE rate has been below the core CPI rate for the past seven years, in some years by a significant amount.

Recent Developments
Recall that both the CPI and PCE headline inflation rates show an increase in inflation in recent years. This increase in inflation is not peculiar to these particular measures of inflation: if we were to look at narrower measures such as the PPI, or broader measures such as the Gross Domestic Product deflator or the Gross Domestic Purchases deflator, we would see the same thing. Some have argued that this development should be discounted as it primarily reflects the recent run up in oil prices and that the best gauge of inflation for monetary policy purposes is the core rate.

If we look at how the core measures have evolved over the past two years we do see a deteriorating situation, albeit much less dramatic than in the headline numbers. The CPI excluding food and energy and the median CPI have been surprisingly stable over most of this period, but both do show noticeable upticks in recent months. The deterioration in the core PCE is less evident, although it too has been drifting upwards gradually since the beginning of 1998 (from 1.2 percent in 1998:II to 1.6 percent in 2000:I). If we were to look at the broadest measures of core, the picture would be slightly worse, with both the Gross Domestic Product and Gross Domestic Purchases deflators excluding food and energy increasing from rates of increase at or below 1.0 percent at the beginning of 1998 to rates of increase in excess of 2.0 percent recently.

Conclusions
The CPI and the PCE deflator are probably the two most important gauges of inflation for monetary policy purposes. Neither measure provides a complete picture of what is happening with inflation, so we always need to look to the other statistics to confirm our assessment of what is going on. At a minimum, we need to take account of what the core measures are telling us, but it is also important to look to both the narrower measures (such as the PPI) and the broader measures (such as the Gross Domestic Product and Gross Domestic Purchases deflators) to complete the picture.

All measures of inflation have shown an increase in the inflation rate since 1998, in some cases by as much as one or two percentage points. To an important extent, these swings in inflation reflect the fact that oil prices were declining in 1997 and 1998, but over the past year have been rising. If we net out the effects of the oil price increases, the pick up in inflation is less dramatic and more recent. When headline inflation, whatever its proximate cause, begins to seep through to core inflation and more importantly, inflation expectations, the risks become very real that the inflationary process will take on a life of its own. Once the inflation genie is out of the bottle, it can be very difficult to get it back in. The current trajectory of interest rate policy is designed to prevent just such an occurrence.

—Mark A. Wynne

About In Depth This article is based on a presentation by Mark A. Wynne, policy advisor and research officer, Research Department, Federal Reserve Bank of Dallas. The views expressed are those of the authors and do not necessarily reflect the positions of the Federal Reserve Bank of Dallas or the Federal Reserve System.