10 Price Indices for Inflation Targeting
- Carol Carson, Claudia Dziobek, and Charles Enoch
- Published Date:
- September 2002
In Recent Years, an increasing number of central banks have adopted inflation-targeting policies: the use of price statistics as control indicators for monetary policy.1 Put in simple terms, an inflation target is a policy-defined rate of change in a specified price indicator (or a set of such ranges and indicators). In inflation targeting, the ease or tightness of monetary policy is dictated by whether the indicators are increasing at a rate below or above their policy-determined targets. These inflation indicators also may be useful in policy-setting environments with a less formal or automatic relationship between monetary ease and the deviation of the indicator from the target.2
Credibility is of major importance in choosing the inflation indicator. Because the consumer price index (CPI) is the price indicator most widely known by the general public, because it is seen as being relatively free of manipulation, and because it generally is available on a timely basis, it is a natural first choice as an inflation indicator.
Once inflation targeting has become an established and credible policy, a central bank might consider a “second generation approach” using indicators that, while having distinct advantages for inflation-targeting purposes, have a lower profile than the CPI. As we will discuss below, as an indicator for inflation targeting the CPI has certain limitations that warrant consideration of other inflation indicators, at least in the more advanced stages of implementing inflation-targeting policies.
Monetary policy requires a broad set of economic statistics. This chapter focuses on one subset, namely, price statistics. First, monetary policy needs for real sector statistics (national accounts and price statistics) are discussed. Once the general outlines of the monetary policy needs for statistics on the prices and volumes of transactions and stocks have been set, attention is turned to whether the available economic data series—comprising generally the CPI or transactions-focused derivatives thereof for prices, GDP volume for output, and real estate and stock market indices for asset prices—are properly designed for policy use.
For this purpose, scope and valuation issues pertaining to the choice of inflation-targeting indices are discussed, and a set of criteria derived that such indices should meet. With these criteria, the adequacy of commonly used indicators (in particular, CPIs) and possible alternatives, such as total demand deflators and supply-based indices such as producer price indices (PPIs), is assessed.
There has been recent concern about whether interest rate policy should react to asset prices as well as transaction prices. In elaborating these issues, however, the problem of getting the best transaction price measure is principally considered. The national accounts framework, as described in the System of National Accounts 1993 (1993 SNA) is used as a reference. Although the 1993 SNA emphasizes the need to present integrated data on flows and stocks, this happy marriage has not yet resulted in widely available data on stocks. In fact, most countries do not yet compile balance sheets providing stock data, and this paucity of data makes the use of official statistics in incorporating wealth measures into interest rate policy less than viable in all but a few countries. Nevertheless, in a subsequent section of the chapter asset price indicators are considered.
Later discussion focuses on measures of core inflation and aspects of price index formulas that are relevant for the use of various price indicators in inflation targeting. Finally, some further areas for investigation and future prospects for inflation measurement will be discussed.
Monetary Policy Needs for Statistics
The traditional monetary policy instruments have either been to control bank reserves to target the money supply or to set central bank-controlled interest rates (such as the overnight bank rates) that then affect other interest rates. Inflation-targeting regimes are geared toward these latter interest rate instruments. There seems to be broad consensus that interest rate policy today can be characterized by so-called Taylor rules for setting targets for interest rates or other policy instruments. A Taylor rule determines the target interest rate generally as a linear function of the deviation of a selected inflation3 index from its target, and may also consider the deviation of the level or the growth in output from its potential, full-employment level or growth rate (Meyer, 2001). In determining this “reaction function” of the Taylor rule, the intensities or coefficients with which the central bank reacts to deviations of inflation and output from their target levels are set to minimize the variation of inflation and output from their targets over time. The Taylor rule framework dates from the early 1990s (Taylor, 1993) and permits a transparent quantitative calibration of a policy’s “relative intensity of reaction” to changes in target variables. The importance to monetary policy of inflation in goods and services transaction prices and in the volume of associated transactions, however, can be traced at least as far back as Irving Fisher (1911). Even within the modern Taylor rule framework, other variables may be integrated as additional arguments determining the policy rate of interest. Exchange rates and, particularly, asset prices are examples of such variables.
Central banks often distinguish between publication targets used in their communications about the inflation-targeting policies and operational targets used to forecast the future levels of published targets. The inflation measure typically used for publication targets is the CPI while a variety of other indicators such as the PPI are used in forecasting future targets. Other inflation measures with broader transaction coverage, however, could be used for publication targets. In this chapter a number of alternative measures for the inflation-targeting index will be explored. These include the GDP deflator and the domestic final purchases price index on the demand side, and the PPI, the value-added deflator, the implicit output price index, and the adjusted domestic supply price index on the supply side.
Scope and Valuation of Transaction Price Indices
The scope of the price statistics to be used for inflation targeting should be related to the policy objective. If the objective of inflation-targeting policy is household welfare, then the real income experience of households is most relevant, which argues for a focus on household consumption and the CPI. This conclusion would also hold if poverty alleviation is an issue. In these instances the expenditure aggregate of the typical CPI would generally be the appropriate scope for measuring inflation. The degree of its appropriateness could be seen as subject to the elimination of the components of CPI weights that represent internal transactions (transactions within one unit, such as a household) and to judgments about how broad coverage should be to properly capture household inflation. The imputed rent of homeowners in some national CPIs is the main example of internal transactions relevant for CPIs. The inclusion of net acquisitions of houses in other national CPIs is an example of how the scope of coverage can be broadened beyond consumption expenditure.4
However, the stabilization of the rate of inflation experienced by households is not a universally accepted policy objective, nor is it universally rejected. In the long term, welfare, income, and poverty are affected by transactions other than household consumption. Such transactions include production, investment, and savings. Furthermore, decisions concerning savings and investment are made not only by households but by corporations as well. Nonfinancial corporations make decisions on profit prospects rather than what households decide to save. Financial corporations, particularly insurance companies and pension funds, make important investment decisions on behalf of households without their involvement. In quite a few countries the use of household savings is mostly through these institutions. This fact advocates a broader coverage of the price indicators concerning both transactions and institutions.
Commentators both inside and outside central banks also have cited inflation in asset prices, such as real estate and financial assets (shares of stock traded on exchanges, for example), as worthy of consideration in determining and evaluating the effectiveness of monetary policy.
Another problem with CPIs is that they include elements that are not transactions between economic agents and could never involve a financial medium of exchange. Thus they arguably are not appropriate to include in the compilation of an inflation target index. On the other hand, it could be argued as well that the CPI’s coverage is too limited, because it does not cover institutional sectors other than households, nor does it cover intermediate consumption (goods and services used as inputs in production), capital, and financial transactions.
Concerning valuation, it should be noted that CPIs basically use prices paid by final consumers, either in shops or in other outlets. These prices include certain taxes and subsidies that would seem to be outside the scope of inflation targeting.
These scope and valuation issues are the main reasons why, as inflation-targeting policies are developed over time, more comprehensive measures might be considered.
Dimensions of Scope
Our preliminary discussion of the CPI has touched on two aspects of the scope of an inflation target price index: its coverage of economic agents resident in the economy, and its coverage of the kinds of transactions and other economic flows that involve those agents. To translate this general economic vocabulary into the more specialized language of economic statistics, we appeal to the 1993 SNA. We will consider the coverage of economic agents represented by a price index as its institutional sector coverage, and continue with earlier language by referring as well to the transaction coverage of the index.
Institutional sector coverage
The 1993 SNA groups economic agents or institutional units resident in an economy into five principal institutional sectors: nonfinancial corporations, financial corporations, general government, households, and nonprofit institutions serving households.
The coverage of institutional sectors is an important dimension of a price index. For example, the CPI is clearly and universally limited to households. On the other hand, the conceptual scope of the PPI is all resident institutional units, or all five types. (Although, in practice, the PPI usually focuses on economic activities largely within the nonfinancial corporations sector, there is no conceptual reason not to include all economic activities.)
Transaction coverage is critical to the interpretation of price indices and the assessment of their fitness for various uses. For example, for good reasons many CPIs include in their housing expenditure weight the implicit rent paid and received by homeowners on their own houses, even though no monetary transaction is involved. Though most available price indices normally cover current goods and services transactions, the coverage of income transactions is not self-evident, and the coverage of financial transactions is even less so. For these reasons, it seems useful to look at the distinction between monetary and nonmonetary transactions, and at the classification of transactions. Because—as noted in the previous section—in addition to goods and services flows, assets also may play a role in monetary policy reaction functions, we first discuss these fundamental concepts.
Transactions and assets
Traditionally, in the national accounts the emphasis has been on economic flows (transactions), with only a peripheral interest in stocks (assets) and with little concern for the linkage between the two. The 1993 SNA has changed this balance fundamentally, emphasizing the need for an integrated presentation, recording all transactions in accounts and assets on balance sheets, and providing explicit links between them. For the discussion here it suffices to emphasize that transactions are flows that represent phenomena that can be observed over time, while assets are stocks that can be observed at a point in time. This distinction can be illustrated by the notions of income (a flow concept) and wealth (a stock concept).
Monetary and Nonmonetary Transactions
The 1993 SNA includes both monetary and nonmonetary transactions. Monetary transactions are mutually agreed-upon economic flows between agents “… in which one party makes a payment (receives a payment) or incurs a liability (receives an asset) expressed in units of currency” (see 1993 SNA, paragraph 3.16). The SNA does not define nonmonetary transactions, but explains them through examples such as barter transactions (in which products are exchanged for other products), wages in kind, production for own consumption (such as housing services by owners to themselves, and food produced by farmers for own consumption), and own account fixed capital formation (for example, an enterprise building its own offices).
Because nonmonetary transactions do not involve the use of money and thus do not affect the demand for money, for purposes of inflation targeting it might be deemed useful to exclude such transactions from the target price indices, either as items or as weights. For instance, in the analytical framework provided by Irving Fisher, the equation of exchange formula equates the demand for money with the product of prices and volumes of goods and services in the economy. In such a system nonmonetary transactions would be out of scope. Excluding nonmonetary transactions may not be possible in all cases (for instance, barter transactions often are not specifically identified in published national accounts), but important items such as owner-occupied housing and other production for own final consumption may well be identifiable.
The 1993 SNA (paragraphs 6.204–6.207) considers three valuation principles as appropriate depending on whether a transaction is viewed from the buyer’s or seller’s point of view: purchasers’ prices, basic prices, and producers’ prices. Purchasers’ prices refer to the amount paid by the purchaser per unit of a good or service, including taxes on products (such as sales taxes and excise taxes on gasoline and tobacco) and charges for transportation, distribution, and insurance invoiced by other providers in the same transaction, and excluding subsidies on products. Basic prices refer to the amount received per unit of a good or service by the seller, excluding taxes on products and charges for transportation, distribution, and insurance invoiced by other providers in the same transaction, and including subsidies on products. Producers’ prices differ from basic prices in that they include taxes less subsidies on products.
The choice of valuation principle is critical regarding the use of a price index as an inflation target. In the view of the authors of this chapter, price indices founded on valuations at basic prices should be preferred to those indices using valuations at purchasers’ and producers’ prices. The reason is that the net product taxes paid are independent of the underlying basic price movements and beyond the control of the monetary authority. Thus, changes in the purchasers’ prices of goods and services arising directly from changes in the rates of net taxes on products are not relevant to triggering and assessing monetary policy actions.
Transaction Price or Inflation Indices for Inflation Targeting
Some General Characteristics of Inflation Indices
Clearly, for transaction prices, the CPI is widely credible as well as frequent and timely. Nevertheless, when looking at a second-generation approach to inflation targeting, it is a good idea to draw back from focusing on CPIs specifically and consider some broad properties that would be desirable for an inflation target price index. We group these factors according to the scope and valuation concepts of the indicator that were the subject of the previous section. Our objective is to lay out the options within the standard macroeconomic framework for economic statistics—the system of national accounts. Considering the additional options for inflation measurement implied by this approach will lead one to ask what new statistics may be required for measuring inflation.
One of the main characteristics of the national accounts is that, in principle, supply and demand must balance. This general principle pervades the SNA, but most relevant in this context is the balance in the supply of and demand for goods and services. This balance applies to individual goods and services, and also, on the aggregate level, to GDP derived from the expenditure side (the expenditure approach) and from the supply side (the production approach). Thus, the appropriateness of price indices can be considered from the use side, with the CPI as a prominent example, and from the supply side, with the PPI as a prominent example.
This section considers an additional property desirable for inflation indices under the general heading of valuation. Inflation indices should be independent of fluctuations in the prices of imports denominated in the unit of account of the supplying country, because, like domestic taxes net of subsidies on products, such prices are beyond the reach of monetary policy instruments. On the other hand, exchange rates often are policy variables or policy instruments themselves. Accordingly, one could look for a way to retain in the inflation measure their contribution to change in the import price index, where the prices in the latter are expressed in terms of the domestic unit of account.
The conclusion of the foregoing discussion of the scope and valuation dimensions of inflation indices is that such an index should satisfy several criteria. An inflation index should
include all types of monetary transactions;
exclude nonmonetary transactions;
include all types of institutional units resident in the country;
exclude taxes and subsidies on products in the valuation of market transactions; and
permit transparent analysis of the contributions to change in the inflation index arising from the prices of exported and imported goods and services. Analysis should aim at eliminating from the inflation target, if desired, (1) all effects on the target index of the prices of imports in domestic currency, or (2) that part of the contribution to change in import prices attributed to price developments in foreign currencies, while retaining the impact of movements in exchange rates.5
Now a range of candidates for inflation indices will be considered.
Consumer Price Index
Credibility is a key to public trust in a central bank’s policy and has to a large extent prompted countries practicing inflation targeting to choose the CPI (or a derivative of it) as the price indicator. As mentioned, the CPI is familiar to the general public and largely beyond the control of the central bank, which thus avoids any suspicion of data manipulation. In addition, CPIs are among the timeliest statistics and are usually not subject to revision.
The CPI published by a country’s statistical authorities is almost always linked to some concept of household expenditure, usually—but not always—relating to the concept of final consumption as defined by the 1993 SNA. National CPIs are compiled according to a variety of scope concepts that differ in the extent to which they align with the final consumption by households in the national accounts.6 Most of these exclude from households’ final consumption certain goods and services that households consume but do not directly pay for. Two broad categories of CPIs can be distinguished: consumption CPIs and transactions CPIs.
The consumption CPI is also known as the cost of living CPI and covers prices from a basket of goods and services that represents household consumption. This household consumption is largely monetary, in the sense that it comprises expenditures on goods and services that households directly pay for, but it also includes nonmonetary items, principally the value of homeowners’ imputed rent.7 A typical consumption CPI covers monetary consumption and the implicit rent of owner-occupants. It is often linked to the cost of living based on an assumption that households’ welfare at a point in time depends on what they actually consume, not on their accumulation of property.
Another reason for including nonmonetary items such as implicit rent is international and intertemporal comparability. For instance, an increase in homeownership would decrease the weight of housing in the CPI if no imputation were made for the nonmarket housing services owners provide to themselves. On the other hand, imputed rent is not itself the outcome of a transaction and its price is, by definition, implicitly determined. Thus, in several countries (such as Australia and the United Kingdom) it has not been considered desirable to include it in an inflation index.
The consumption CPI employs a purchaser’s price valuation basis, so changes in the rates of taxes and subsidies on products affect it as immediately and directly as prices do. As indicated at the beginning of this section, this feature is not desirable for an inflation index. Of course, the effects of tax and subsidy changes can be removed as a contribution to change in the CPI if sufficient data are available, but it would be preferable not to have such effects in an inflation index at the outset.8
The value aggregate of the transactions CPI, often called the inflation CPI, comprises the monetary consumption expenditure of households. Such expenditure excludes homeowners’ imputed rent and other imputed items, but includes household capital formation in residential structures (net acquisition of, and major improvements to, residential structures). The typical transactions CPI covers households’ monetary consumption and net acquisitions of houses. As its name implies, the transactions CPI focuses on the prices of items purchased by households in transactions with other institutional units, and proponents of this index have advanced it as an answer to central bank needs for inflation-targeting policies (as well as general inflation monitoring).
Like the consumption CPI, however, the transactions CPI focuses exclusively on the price experience of households in making final expenditures. Doing so leaves aside the inflation in the prices of other transactions, which also generate a demand for money. Among these other transactions are (1) the intermediate consumption (goods and services used in production) not only of households but also of other institutional sectors such as corporations and the government, (2) the capital acquisition expenditures of institutional sectors other than households, and (3) the exports expenditures of nonresidents. Because all these transactions generate a demand for domestic money, the scope of the transactions CPI leaves much to be desired in its coverage of transactions in goods and services.
Further, like the consumption CPI, the transactions CPI employs a purchasers’ price valuation basis, so changes in the rates of taxes and subsidies on products affect it as immediately and directly as prices do.
Other Use-Based Price Indices
Implicit GDP deflator
In the 1993 SNA, GDP is defined as the sum of final expenditures on consumption, capital formation, and net exports. As a result, the implicit GDP deflator9 is a function of the price indices for consumption, capital formation, and net exports. As such, it is implicitly a function of detailed CPI components, in conjunction with other indicators. An advantage of the GDP deflator is that it covers all institutional units in the economy. Note, however, that the purchasers’ prices principle used to value the expenditures against GDP makes the GDP deflator directly sensitive to changes in the rates of tax and subsidy on products. The GDP deflator also is inversely sensitive to movements in import prices.10 Finally, the GDP deflator does not cover intermediate consumption transactions in goods and services.11
Domestic final purchases price index
Woolford (1999) argues that the scope of the CPI is too narrow for inflation monitoring and suggests the domestic final purchases (DFP) price index. The DFP index covers GDP plus imports less exports, which is equal to final consumption plus gross capital formation. Such a price index would, like GDP, exclude intermediate consumption transactions and include taxes less subsidies on products. The DFP index would be directly sensitive to import prices according to their importance in consumption and capital formation, and to taxes less subsidies on products. Although the effect of imports and taxes could in principle be identified using contribution-to-change decompositions, these often require difficult data collection and compilation operations.
Supply-Based Price Indices
Other alternative measures that can be used to gauge general inflationary pressures include supply-based indices. The most prominent of these are the PPI and the implicit value-added deflator, but other supply-based indices could also be considered.
Producer price index
At least one alternative to the CPI is readily available and timely in many countries: the PPI. As an inflation index, the PPI has definite appeal. The price concept used is basic prices, so the PPI does not react directly to changes in the rates of tax and subsidy on products. Also, it covers the output of resident producers, so it is not directly affected by import prices. A number of countries, including almost all advanced economies, produce timely monthly PPIs, so the availability of the PPI is, in these countries, on par with the CPI.
In practice, however, the PPI has one major disadvantage: its coverage of economic activities. In principle, the PPI covers all market output, but no country has achieved this. In most countries the PPI is an industrial PPI that covers goods production and the production of the “industrial services,” comprising transportation and utilities. Transportation includes passenger and freight transportation by air and other modes, and utilities include power generation and distribution, water supply, and the like. In addition, most countries have an output price index for agriculture; thus, a widened PPI including agriculture could be produced. The coverage in the PPI of construction and of nonindustrial services such as business services, data processing, legal services, and personal services is not comprehensive in any country, but is on the agenda for improving the PPI in most countries that produce it.12 Thus, the PPI would be a definite candidate as an indicator for inflation targeting, but only with more comprehensive output coverage.
Implicit value-added deflator
If GDP is viewed from the production side, it is the sum of value added—derived as output minus intermediate consumption—by industries, plus taxes less subsidies on both domestically produced and imported products. The implicit value-added deflator can be defined as the ratio of value added in current and constant prices. This deflator can also be seen as a function of the price index for total output and the price index for intermediate consumption. As an inflation index, the value-added deflator has the advantage of covering all institutional sectors, not just households, but it still has two drawbacks, based on the criteria used here. First, though the deflator depends on intermediate consumption and thus on the prices of intermediate consumption goods, it does so only in the sense of netting intermediate transactions from total output. By implication, its transactions coverage purposely excludes intermediate consumption. Second, because intermediate consumption is valued at purchasers’ prices, like the GDP deflator it is immediately, if inversely, sensitive to changes in the rates of tax and subsidy on products. Furthermore, although the 1993 SNA prefers output valuation at basic prices, it also allows valuation at producers’ prices.
Implicit output price index
The implicit output price index (IOPI) is the most comprehensive price indicator for domestic output. It can be seen as a weighted average of the prices of the output of all domestic activities, although, in practice, it usually is derived as an implicit deflator (that is, as the ratio of total output in current and constant prices).13 Compared with the PPI, the IOPI has the advantage of full coverage of the economy, and compared with the implicit GDP deflator it has the advantage of also covering intermediate consumption. For countries following the 1993 SNA’s preference for valuing output at basic prices, it also has the advantage of being unaffected by changes in product taxes and subsidies. Its only disadvantage as compared with the PPI would be its frequency and timeliness: like the implicit GDP deflator, it is usually available only quarterly, with a lag of three months.
Adjusted domestic supply price index
It could be argued that the demand for money in an economy is better tracked by total supply, comprising output at basic prices plus imports free-on-board (f.o.b.), than by output only, as in the IOPI. From this angle the domestic supply price index (DSPI), whose domain is total supply, is the broadest inflation indicator among the supply-based indices of the prices of goods and services produced in the current period. The DSPI can easily be constructed by combining the IOPI with price indices on imports from foreign trade statistics, which are readily available in many countries.
The DSPI uses the same valuation principle as the IOPI, although the inclusion of import prices makes it sensitive to exchange rates. This sensitivity may be of interest in a target indicator for inflation. It allows analysts to see the contribution to change in the inflation indicator for total supply of changes in the exchange rate, as well as changes in the basic prices of domestic production, and in the f.o.b. prices of imports in the currency of the nonresident supplier. Monetary policy analysts may be interested particularly in a version of the DSPI from which the contribution to change arising from movements in the foreign currency prices of imports is excluded, as this eliminates the component of the index that is presumptively beyond the reach of their policy instruments. This exclusion would leave the import component of the DSPI sensitive only to movements in the exchange rate, a currency price often included among the price indicators closely monitored by monetary policy analysts, particularly in open economies.
To reflect the demand for domestic money, a case can be made for including the prices of transactions in existing goods in an inflation target indicator. By the same logic, the DSPI might be broadened further to include transactions in existing nonfinancial assets, both produced and nonproduced. (Below is a consideration of how the prices of nonfinancial asset stocks, as opposed to transactions, may enter into an inflation-targeting scheme, and the statistical implications of such an option.) This index is referred to as the adjusted domestic supply price index (ADSPI).
In sum, the ADSPI would comprise a combination, or index, of the IOPI, the import price index (excluding the contribution to change from movements in the foreign currency prices of imports, but residually including exchange rate effects), and the price index for sales of nonfinancial assets. It is the most comprehensive price index of goods and services transactions among the alternatives we have considered.
Some possible empirical implications of different transaction price indices
Figure 10.1 charts the CPI, PPI, gross fixed capital formation implicit deflator, and GDP deflator for the United States (as a large continental economy) and Norway (as a small open economy) to convey an idea of the empirical differences that valuation and coverage make over time. It is clear that the profile of the target interest rate would differ across these price series, perhaps markedly depending on the numerical size of the inflation coefficient in the reaction function. Thus, different inflation indicators would indicate different monetary policies. The two countries display different time paths for all indices, but both have large divergences, particularly from 1995 to the present, between the CPI and PPI. The chart indicates that the CPI and GDP deflator are visually correlated in both countries, while the PPI exhibits a distinctly different profile. Although the ADSPI is not shown, it would combine the profiles of the PPI and the capital formation deflator.
Figure 10.1.United States and Norway: Major Price Indices1
Source: IMF International Financial Statistics (IFS).
1 Annual percentage change in the following variables: CPI, consumer price index; PPI, producer price index; GDP deflator, gross domestic product deflator; and GFCF, gross fixed capital formation deflator.
Asset Price Indices for Inflation Targeting
Asset Prices and Monetary Policy
Of perennial concern to business cycle and monetary policy analysts has been the phenomenon of “asset price bubbles.” These are periods during which the prices of assets are thought to have increased beyond a reasonable assessment of their net present value.14 Such overvaluations generally lead to sudden retrenchments in asset valuations that cause major financial and real dislocations in the economy. Consequently, attention has focused lately on whether to include measures of asset valuations in the central bank’s Taylor rule. Such a move would help smooth swings in inflation and output by allowing central bankers to respond to the emergence of asset price bubbles by proactively setting nominal interest rates.15 Implementing such a rule would require timely, credible series not only on goods and services inflation and output growth, but also on asset prices.
Alchian and Klein (1973) are among the first to have considered asset prices in the formulation of monetary policy within a framework of intertemporal optimization, but Samuelson (1961) considered the role of assets into the measurement of inflation even earlier. The argument in favor of incorporating asset prices in the determination of monetary policy is that, because the present value of future service streams from assets determines their price, asset prices can be seen as directly measurable indicators of traders’ forecasts of future inflation in goods and services transaction prices. Changes in asset prices thus can be seen as changes in expectations about future inflation. On the other hand, the volatility of asset prices, particularly share prices, suggests that such forecasts of future inflation could vary greatly from moment to moment. Such volatility has worked against including asset price indices explicitly as policy variables for monetary policy.16
Alchian and Klein, and others since, have considered principally the case in which asset prices are incorporated along with transaction prices into a single inflation index. An important observation to take from recent analyses in the Taylor rule framework, such as those by Bernanke and Gertler (1999) and Cecchetti and others (2000), is that it is not necessary in an inflation-targeting policy framework to have a policy or trigger price index that includes within its scope both goods and services transactions and asset stocks. Rather, for the formulation of monetary policy, distinct indices are needed for each that are credible, frequent, and timely. Thus, following Cecchetti and others, the modern Taylor rule approach can provide for distinct consideration of both transactions price indices in the tradition of Fisher and price indices for stocks following Alchian and Klein.
Real and Financial Asset Prices
Should property and equity prices be considered along with the prices of goods and services in formulating monetary policy? Recently, the amount of economic literature on this topic has been substantial (see, for example, Bernanke and Gertler, 1999; Cecchetti and others, 2000; Goodhart, 2001; and Goodhart and Hofmann, 2000; and the references therein). The conclusions are broadly negative about including share prices in inflation measures, because share prices are extremely volatile and thus need to be calibrated into the Taylor rule with a numerically small coefficient to successfully minimize deviations from the target rate of inflation. Second, the evidence indicates that such indices have rather low predictive power regarding future inflation in the “flow” prices of consumption and production, which is the reason they would be included in an inflation index to begin with. Both points argue against inclusion of such prices in the central bank’s Taylor rule.
On the other hand, the evidence generally favors the inclusion of real estate price indices in Taylor rules, principally because they have some explanatory power for the CPI (Cecchetti and others, 2000; Goodhart, 2001). Cecchetti and others provide simulations of macroeconomic models supporting the view that reaction functions incorporating real estate prices would better achieve the objective of reducing the variability of goods and services (CPI) inflation and of output growth around their target levels. Nevertheless, Cecchetti and others also argue for consideration of financial asset (equity) prices as an additional determinant of the interest rate policy “reaction function” of the central bank for setting interest rates.
In all of these most recent analyses, separate incorporation of asset prices into the reaction function effectively permits their weight to be set much lower relative to goods and services price inflation than would be implied by the Alchian and Klein theory. Although Cecchetti and others (2000) examine an overall price index including both goods and services price indices and asset price indices, with a special weighting determined by a statistical model, their most persuasive evidence derives from simulations incorporating separate asset price indices into the central bank interest rate policy function.
The second section suggested that, in the context of the scope of definition of money for policy purposes, the decision to include capital and financial transactions in the inflation price index might be seen within a Fisherian framework. If the monetary policy instruments operated by the central bank focus on narrow money (monetary gold and SDRs and currency and deposits) plus securities other than shares (such as treasury bills and bonds), then the prices of transactions in shares and other equity and financial derivatives might be eligible for inclusion in the inflation indicator along with goods and services prices. On the other hand, if the policy focus were on broad money as defined by the current statistical standard for financial statistics, the IMF’s Monetary and Financial Statistics Manual (2000a), we would include shares and derivatives in money and exclude them from the inflation index. Indeed, broad liquidity aggregates tend to perform better in macroeconometric studies of money demand, but they are under less direct control of the policy instruments available to central banks than the narrow aggregates.
Though in the recent monetary policy literature the distinction between the prices of transaction flows and asset stocks is a crucial one, this is not formally equivalent to saying that the price of a given type of good, such as a house, might not appear in both the transaction and asset price measures for use in interest rate policy functions. The good’s price would be weighted by sales transactions in the first instance, and by the value of stocks held in the second. For example, one could conceivably have house prices weighted by net acquisitions in the inflation indicator, and weighted by the outstanding value of single-family residential real estate in the asset price indicator.17 Finally, as noted earlier, the state of official statistics around the world for the prices of real and financial stocks generally is not as good as for transactions. If asset prices were to be formally and transparently included in determining interest rate policy, better stock price series would have to be developed to ensure the credibility of the policy rule.
Further Areas of Inflation Measurement for Policy Implementation and Assessment
Income Transactions in Inflation Indices
In addition to goods and services, another class of transaction that we have not touched on arguably affects the demand for money: income transactions. Taken across the total economy, gross national income—the aggregate of income transactions—differs from GDP by labor and property income receivable by residents from nonresidents less such income payable by residents to nonresidents (1993 SNA, paragraphs 2.183, 7.17).
One could identify a price index for some income transactions, such as a labor services price index for compensation of employees. Other income transactions, however, would be more problematic. For example, taxes do not have an intrinsic price component, and property incomes from services of capital and land have a price that is difficult to measure at the current state of the art. On the other hand, it often is argued that all income is eventually to be used, and from this perspective one could envision a general price indicator for total final demand (the implicit GDP deflator) or value added (implicit value-added deflator) that would also cover income transactions.
The Cost of Production Concept of Inflation
There is yet another approach to measuring goods and services inflation. To represent it, the total value of transactions in goods and services is decomposed into the components of the cost of production. This approach would begin with intermediate consumption, but would then show the payments to primary production factors that make up value added. These are compensation of employees, taxes less subsidies on production (other than levied on products), and operating surplus. Then would be added acquisitions of capital assets and taxes less subsidies on products to round out total transactions in goods and services in purchasers’ prices. In certain respects, the income approach to aggregate transactions intersects with the cost of production approach to goods and services. Compensation of employees is included under both approaches, and operating surplus in the cost of production approach can be seen as comprising the property income items from the income approach adjusted for realized holding gains. This approach is based on an interpretation of income-generation transactions that is not widely considered by national accountants. Nonetheless, it organizes the price information often monitored by macroeconomists on goods, produced services, and labor services into cost-push indicators for product price inflation. It would cover the purchaser’s value of total goods and services transactions less net taxes on products and operating surplus.18
A great deal of attention has focused in recent years on the impact of the index formula on the accuracy of price indices as inflation indicators. This subject includes both the aggregation of samples of elementary item price relatives into item indices, and the aggregation of the item indices into higher-level aggregates. The context of the formula discussion has been the CPI. However, formula issues extend as well to the broad coverage inflation indices discussed earlier: the PPI and the ADSPI.
Given the scope of any inflation index, the principal formula issue is to eliminate the substitution bias of fixed basket indices such as the Laspeyres and Paasche. Results from the economic theory of index numbers establish that so-called superlative price index formulas, such as the Fisher Ideal and Törnqvist, account for the behavior of economic agents in changing their pattern of transactions in goods and services in response to changes in relative prices. Considering price indices for the uses of resources such as the CPI, the substitution bias of Laspeyres indices is upward relative to an economic cost of living index for households from a Laspeyres perspective.19 On the other hand, the Paasche price index is downward biased relative to an economic index from the Paasche perspective.20 In the context of price indices for the supply of resources, such as the PPI and ADSPI, the biases run in opposite directions.
The scope of inflation indices in covering institutional sectors and types of transactions is, in the authors’ view, a more important issue than the formula-related bias of the CPI in medium-term planning for inflation indicators. Nevertheless, it is easy to see the legitimacy of central bankers’ concern. If the CPI were the central bank’s inflation target, the upward bias of the Laspeyres CPI that is almost universally produced could lead to an overly tight monetary stance.21 Similarly, for supply-based target indices for which the bias is presumptively the other way, the monetary stance could be overly lax. There is thus a strong argument for the use of superlative formulas such as the Fisher Ideal and Törnqvist, regardless of the inflation indicator used. It should be noted that the implementation of timely superlative indices requires that the series be revisable, because weighting information becomes available with a substantial lag compared with the more timely information on prices for almost all price indices. The sole exceptions to this observation are the export and import price indices. The weights of these indices are based on customs data, which are usually as timely as the prices of exported and imported goods. Consequently, in most cases the actual implementation of a superlative formula is in the form of a superlative series up to the latest period for which weights are available. After this period, the index is extrapolated by the change in a Laspeyres or weighted geometric index for the most recent periods. Superlative accuracy thus comes at the price of tolerating an explicit revision variance.
Central banks are loath to undertake a policy action on the basis of a price fluctuation that is quickly reversed, as the impacts of monetary policy register with a lag and may stretch over a number of quarters. Developing credible methods for extracting useful price signals from short-term noise in the target price index thus is a major concern. This inflation signal is known more commonly as “core inflation.” Thus, in terms of its scope and valuation concepts, the definition of the methodology for extracting a core inflation index or inflation signal from an underlying series is distinct from the definition of the underlying series per se. Signal extraction techniques, although most often applied to the CPI, may be applied to any of the potential inflation indices we have discussed in the previous sections of this paper. We note three main approaches to signal extraction for price indices, which we term adjusted coverage, statistical smoothing, and robust estimation methods.22
Adjusted Coverage Methods
Several countries construct core inflation measures by deleting from an inflation index components found or judged to be volatile. For example, in the United States “core CPI” excludes food and energy items. Though this approach has the advantage of transparency and easy communication to outside observers, it is essentially arbitrary for any given short-term movement of the overall index and often involves a fairly serious compromise of the carefully constructed coverage of the original index. Adjusted coverage is, at best, a crude signal extraction technique for producing a core inflation index.
Statistical Smoothing Methods
Another approach to signal extraction is to apply one or more of a variety of available statistical smoothing techniques. Most techniques of statistical time series filtering derive directly from engineering principles for signal extraction in electronic circuits, such as the widely used X-11 and X-12 seasonal adjustment filters designed by the U.S. Census Bureau. On the other hand, the recently developed “state space” methods are more closely related to linear modeling methods used in econometrics and offer some advantages in the incorporation of explanatory variables and handling unusual events in the time series to which they are applied. With regard to credibility, employing such smoothing methods to obtain a core inflation indicator can be seen as a technical operation that does not involve the judgment of monetary policymakers. However, it is easy to overly smooth a price series with these methods, eliminating movements that represent valid price signals for triggering policy action. Also, substantial analyst judgment and expertise is always involved in the model specification and interpretation process that complicates their explanation to outside observers.
Robust Estimation Methods
Robust estimation methods for obtaining core inflation signals from a target series delete or trim the price changes of component series in an index for individual periods. They do not, however, eliminate once and for all those component series of an index that are judged a priori to be “volatile,” as adjusted coverage methods do. So-called “trimmed mean” methods fall into this class, and have been implemented by Bryan and Cecchetti (1994, 1999) for the United States and Japan, and Roger (1997) for New Zealand. These methods have the advantages of being easy to communicate to outside observers and capable of being objectively applied without hint of intervention by the analysts using them as signals. They thus have significant credibility advantages over the other two methods for developing a core inflation measure from a given inflation index.
Clearly, wide public acceptance of the price index used for inflation targeting is crucial to the credibility of a monetary policy that is, in part, based on movements in the index. This chapter has, however, addressed other factors important to the effectiveness and focus of such policies. To frame the problem of selecting a concept for an inflation index suitable for implementing monetary policy, this chapter has proposed five scope and valuation criteria and identified additional formula and signal extraction considerations. The taxonomy used embraces and rationalizes the broad factors discussed in the literature on inflation indicators within the principal international system for economic statistics, the System of National Accounts 1993.
The current practice of using the CPI or its derivatives as an inflation index has the advantages of timeliness, availability, transparency to the public, and credibility, because the CPI is produced by an independent agency, usually the statistical office. The CPI is lacking in scope however. It covers only the inflation experience of households, whether the monetary and nonmonetary consumption components of a consumption CPI, or the monetary consumption and household capital formation components of a transactions CPI. Further, it was noted that the CPI’s purchasers’ price valuation basis makes it sensitive to tax and subsidy changes that do not belong in an inflation indicator for monetary policy.
Alternatively, if one went back to first principles and attempted to design an index that could be considered most appropriate for inflation targeting, one would want an index that has broader coverage of monetary transactions, is available on a timely basis, and could be produced by most countries’ statistical systems. Such an index would cover the inflation experience of all resident institutional units, not just households, and would incorporate valuation of transactions at basic prices that would free it from direct sensitivity to tax and subsidy changes. Further, the index would not be directly sensitive to exogenous price shocks from imports that are unrelated to exchange rates. On the basis of these criteria, the PPI and ADSPI have, in principle, desirable inflation measurement properties. However, in practice, the PPIs that countries produce have limited coverage of economic activities (mainly manufacturing industries). The industrial PPI does not meet the criteria set out above; for a PPI to be useful in inflation targeting, extension of its coverage would be needed. The ADSPI would answer all the criteria but timeliness, as some of its components are available only on a quarterly basis.
Prospects for Inflation Measurement
There are immediate, medium-term, and long-term options for constructing inflation indices. In the near term, the choice of inflation target indices is limited to existing series of monthly frequency and timeliness. Most countries are thus limited to the CPI, and countries with highly developed statistical systems are limited to the CPI and PPI. Though, in the authors’ view, the PPI is preferable to the CPI in principle, its lack of coverage of the service industries is a serious drawback. However, a number of statistical offices around the world are addressing this problem, so the PPI could represent a viable medium-term option.23 Compared with the CPI, use of the PPI in inflation targeting could make a qualitative difference, because it would include the prices of intermediate and capital goods not included in the CPI. Moreover, its trend has clearly differed from that of the CPI.
Still better than the PPI would be the ADSPI, which would add to the PPI the prices of transactions in the existing stocks of nonfinancial assets and reflect exchange rate movements without sensitivity to changes in the foreign currency prices of imports. This, too, could be a medium- to long-term option, because timely data on the sales and prices of nonfinancial assets are not available across the full range of such assets, hindering the development of timely, comprehensive price indices for such transactions. On the other hand, the ADSPI could be expected to move differently from the PPI, because its capital goods and intangible assets price component could display decidedly different movements than its price component for current production. We need further study of the relevance of the choice of inflation indicators for inflation-targeting policy, and the feasibility of expansion of the PPI as discussed above.
For a recent review of central banks in developing countries that maintain inflation-targeting policy regimes, see Schaechter, Stone, and Zelmer (2000). Meyer (2001) notes that among advanced economies, explicit inflation-targeting policies have been adopted in New Zealand in 1990, Canada in 1991, the United Kingdom in 1992, Australia in 1993, and Sweden in 1993. In addition, he makes a case for the European Central Bank having in effect adopted a more or less explicit inflation-targeting policy, and notes that “the Taylor rule [setting the interest rate policy target as a function of target inflation and potential output] is a useful characterization of U.S. monetary policy.” He also describes the various modalities by which these policies have been implemented around the world.
Bernanke and Gertler (1999) note that inflation-targeting policies can be implicit or explicit. Their empirical analysis of central bank “reaction functions” for the United States and Japan from the mid-1980s to the late 1990s suggest that monetary policy has behaved as if it had an inflation target, even though neither country had an explicit inflation-targeting policy over the period analyzed. This supports, for the United States, the Meyer remarks cited in footnote 1.
Broadly speaking, the monetary policy literature uses the unqualified term “inflation” to refer to the rate of change in the prices of goods and services transactions. The term “asset prices” or “asset price inflation” refers to the level or rate of change of the price index (indices) for asset stocks. This terminology will be followed throughout
In this chapter, the volatility of the prices of certain commodities or other subdomains of a price index are not considered to be a determinant of the scope of the index. Deleting high-volatility subdomains from the overall index belongs to one of three categories of methods of signal extraction considered in a later section of this chapter. That discussion focuses on the problem of extracting the “core inflation” signal from the short-term movements in an inflation index, whatever its designed scope. Consequently, the discussion of scope will not consider so-called core inflation indices constructed as, for example, a CPI less food and energy.
“Contribution-to-change” measures the effect on the aggregate price index arising from the change in one or more of its components. The contribution to change does not measure the full general equilibrium effect of a change in the relative prices of exports and imports as compared with those for goods and services originating from domestic sources and used by residents.
Inflation measures based on national CPIs are not strictly comparable across countries because of differences in coverage. To obtain comparable measures, a standard CPI definition would have to be determined to ensure that the same scope of transactions and other flows in goods and services is being measured. For example, the harmonized index of consumer prices (HICP), developed by the European Central Bank and Eurostat, is an attempt to construct such a standardized measure for making inflation comparisons.
The standard approach in CPI compilation and in the 1993 SNA is to treat durable goods as consumed at the time of purchase rather than to measure the flow of services provided by such goods. The one exception to this in the 1993 SNA, as well as in the CPIs of a number of countries, is to include price measures for the imputed rents of homeowners.
See Australian Bureau of Statistics (2000, pp. 9–24) for a recent assessment of the impact on the CPI of imposing a general sales tax.
The implicit GDP deflator can be defined as the ratio of GDP in current and constant prices.
Although the GDP deflator is an inverse function of the import price index, its other price components also are direct functions of import prices, depending on the import composition of consumption and capital formation. Thus, movements in the before-tariff prices of imports that are part of final consumption effectively cancel out in the GDP deflator. However, price movements of imported intermediate consumption items do affect GDP and the GDP deflator. In addition, movements in tariffs on imports, for both intermediate and final consumption, and taxes on exports that comprise part of the change in taxes on products affect the GDP deflator as a use-based price index.
The GDP deflator is, however, inversely sensitive to the prices of imports used in intermediate consumption, as previously noted.
The Organization for Economic Cooperation and Development (OECD) conducts an annual survey of producer price indices compiled by national statistical authorities, the results of which are available on the OECD website at www.oecd.org/std/indserv/index.htm.
In general, the volume index used to calculate the IOPI is itself partly composed of detailed volume indicators that are obtained by deflation of detailed outputs at current prices by detailed PPI components. Other components of the volume index are direct indicators of output. Working through the algebra, the IOPI is thus a hybrid of directly measured price indices from the PPI and indirect price indices. The latter are typically for the outputs of both market and nonmarket services.
Clearly, sustained low inflation implies less uncertainty about the future, and lower risk premiums imply higher prices of stocks and other earning assets. We can see that in the inverse relationship exhibited by price/earnings ratios and the rate of inflation in the past. But how do we know when irrational exuberance has unduly escalated asset values, which then become subject to unexpected and prolonged contractions as they have in Japan over the past decade? And how do we factor that assessment into monetary policy? (Greenspan, 1996, italics added.)
See Cecchetti and others (2000) for a comprehensive overview of the asset bubble issue in setting monetary policy, and Goodhart (2001) and Goodhart and Hofmann (2000) for a specific consideration of inflation and asset price measurement.
See Goodhart (2001), Bernanke and Gertler (1999), and Shibuya (1992), among others. Goodhart and Hofmann (2000) note, however, that real estate prices do seem to have some explanatory power for future inflation in consumption goods and services prices, and accounting for them in the reaction function thus seems to improve the performance of interest rate policy in stabilizing goods and services inflation and output growth. See also Cecchetti and others (2000) for a similar assessment of the role of real estate prices and a negative assessment of the usefulness of share prices in determining interest rate policy.
However, this weighting could increase the collinearity between the inflation and asset price variables, which may be econometrically problematic when calibrating Taylor rules with empirical macroeconomic models.
Operating surplus can be seen as comprising factor payments to the suppliers of capital services. The 1993 SNA does not take a view on direct measurement of the income components comprising it.
“Laspeyres perspective” denotes an economic price index oriented from the point of view of the preferences of households in an earlier reference period. See Bureau of Labor Statistics (1993) and Boskin and others (1996) on formula-related and other biases in the CPI.
“Paasche perspective” denotes an economic price index oriented from the point of view of the preferences of households in the current period.
See Clark (2001) for a recent review of core inflation measures.
Services account for more than half of GDP in advanced economies. The impetus for broadening the PPI to cover services is driven principally by interest in better measuring the volume of total output and value added, and in the contribution service industries and products make to growth. Such measures would lead to more accurate series on GDP from the production approach, and more accurate labor and multifactor productivity indicators.