Abstract
The Q&A in this issue features seven questions on the role of precautionary savings in open economies (by Damiano Sandri); the research summaries are "The Macroeconomics of Aid (by Andrew Berg, Rafael Portillo, and Luis-Felipe Zanna) and "The Building Blocks to Measure Inflation" (by Mick Silver). The issue also lists the contents of the March 2011 issue of the IMF Economic Review, Volume 59 Number 1; visiting scholars at the IMF during January?March 2011; and recent IMF Working Papers and Staff Discussion Notes.
Price indices serve as measures of inflation, deflators for national accounts aggregates, the basis for escalation payments, terms of trade analysis, and much more. Good economic analysis requires a proper understanding of deficiencies in the practice of compiling these measures. Research by IMF economists has contributed to our understanding of what is good practice. An article in the IMF Research Bulletin in September 2006 outlined IMF research in this area. This article considers subsequent work, the focus of which is on the building blocks of price indices.
Consumer price indexes (CPIs) are compiled in two stages. The first stage is the measurement of (unweighted) average price changes of well-specified items—for example, a 500-gram, pre-packed and sliced loaf of white bread with soft crust—using prices collected from a sample of representative outlets. Then, as a second stage, the CPI is compiled as a weighted average of these elementary indices for bread, sugar, mangos, electricity, shelter, television sets, haircuts and, of course, much more. The first stage elementary price indices are the building blocks of a CPI.
Similar considerations apply to measuring purchasing power parity (PPP) across countries, with unweighted parity estimates for well-specified items forming the building blocks for the higher-level weighted aggregates. For export and import unit value indices (XMUVIs), the building blocks are the unit value changes for the applicable tariff item codes, such as “prepared mustard”—the 10-digit code under the Harmonized Commodity Description and Coding System (HS) is 2103.30.20.00. More than 20,000 codes are available. The second stage is to take a weighted average of the unit value changes for each detailed HS code. Producer price indices (PPIs) are compiled at the elementary level using the price changes of well-specified goods and services averaged over different producing establishments; the overall PPI is the (second stage) weighted average of these elementary indices.
For each of these indices—CPIs, PPPs, XMUVIs, and PPIs—the accuracy of the final index critically depends on the adequacy of the aggregation procedure used for their building blocks. This has been the subject of much research at the IMF over the last five years. A defining feature of these elementary indices is that they are unweighted.
For CPI compilation there are three main aggregation methods used at the elementary level: the arithmetic mean of price changes (the Carli index), the change in arithmetic mean prices (the Dutot index), and the geometric mean of price changes, equal to the change in geometric means (the Jevons index). All three formulas have some intuition. Axiomatic index number theory clearly shows Carli to be biased and chained Carli, substantially so—the annual change in the CPI for Kenya for September 2009 was 17.9 percent using the (biased) chained Carli index, but only 6.7 percent with its replacement by the Jevons index. Such bias is not trivial.
Given the bias in the Carli index, the two principal formulas that should be used to calculate elementary indices are Dutot and Jevons. Silver and Heravi (2007a) provide an analytical framework, based on sample estimators, to show the difference between the Dutot and Jevons indices is determined by changes in the variance of prices. A deficiency of Dutot is that it is not invariant to the units of measurement (commensurable) and, thus, the extent of quality differences in items sold. We further decomposed the difference between the two formulas into a difference owing to product heterogeneity and a difference owing to essentially different types of averages in the index formula. This provided a theoretical and measurement framework for identifying why Dutot and Jevons differ after allowing (and correcting for) Dutot’s shortcomings with regard to commensurability.
Unit value indices based on customs information are widely used as the building blocks and surrogates for export and import price indices. In empirical work, Silver (2009a) demonstrates that they can seriously misrepresent inflation in traded products. Moreover, the use of unit value indices leads to even more serious errors in terms of trade indices. Unit value indices should appropriately be used only for homogeneous items. It is argued that with increasing product differentiation, as well as the dwindling availability of national customs data due to customs unions and increasing trade in services, the use of unit value indices based on customs data is a disservice to price measurement. This position is in line with the recommendations for best practice given in ILO and others (2009), where it is suggested that countries using unit value indices move to establishment survey-based price indices and provides a strategy for doing so. This is an important departure from international recommendations from the United Nations (1981) that are nearly 30 years old.
Index number theory advocates the use of superlative price index number formulas, including Fisher and Tőrnqvist price indices, as target indices for heterogeneous goods and services. However, for homogeneous products and services, it is well recognized that superlative price indices can be misleading and unit value indices are the appropriate target index. Silver (2009b) provides a formal mathematical decomposition and understanding of why unit value and Fisher’s price indices differ, identifying both a levels and substitution effect. The paper draws attention to a continuum between homogeneous and heterogeneous products and a need to examine the issue of which index number formulas are appropriate for products within this continuum. There are many products that are only slightly differentiated to meet niche markets or to attain competitive advantage. The paper discusses the case of broadly comparable items and proposes a hedonically-based aggregator (see also Silver 2010a).
The building blocks for the measurement of PPPs by the World Bank’s International Comparison Program are similar to those of the CPI above—average prices across outlets for well-specified items, such as bread—but within and across countries in a region, as opposed to over time and covering all expenditure components of GDP and not just household consumption expenditure. Generally, country product dummy regressions are used to estimate elementary parity indices. The coefficients on the country dummy variables are estimates of the price parity for a product group between countries. These parity estimates are then aggregated to PPPs using weighted methods outlined in the World Bank’s (2007) International Comparison Program Handbook and Diewert (2008). Silver (2009c) considers deficiencies in the regression-based aggregation procedure using a panel-data framework and extends the analysis to panel estimators that incorporate quality adjustments for noncomparable products, primarily through the use of a hedonic country product dummy framework. Work has also been undertaken on IMF uses of PPP estimates (Silver, 2010b).
There are specific measurement issues for particular products, that is, for hard-to-measure goods and services. Zieschang (2010) tackles the difficult area of indirectly measuring the price change of financial intermediation services by banks through the spread between the rate earned on assets and the rate paid on liabilities. Silver and Heravi (2007b) consider alternative hedonic-based methods for the price measurement of goods and services, such as personal computers, that have a rapid turnover in quality characteristics. Silver and Heravi (2007c) and Diewert, Heravi, and Silver (2009) provide mathematical decompositions to better understand the difference between two seemingly equally plausible, yet quite different, hedonic approaches. The difficult area of new outlet (Wal-Mart) bias that occurs when the effect of consumer substitution to new supercenters is improperly incorporated is considered by Hausman and Liebtag (2009) and commented upon by Silver (2009d). Armknecht, Diewert, and Nakamura (2007) consider seasonal goods and services for which month-on-month price change measurement is problematic.
The IMF’s research usefully finds its way into practical recommendation contained in price index manuals (ILO and others, 2004a, 2004b, 2009; UN ECE and others, 2009).
These manuals set out internationally-accepted standards for measuring inflation that are promulgated in the IMF Statistics Department’s training, technical assistance, and report on standards and codes data missions. Similarly, a newly developed training course by the IMF’s Statistics Department and Monetary and Capital Markets Department on core inflation measurement was also built on research work including Roger (2000, 2010) and Silver (2007).
References
Armknecht, Paul A., W. Erwin Diewert, and Alice O. Nakamura, 2007, “Dealing with Seasonal Products in Price Indexes,” in Price and Productivity Measurement. Vol. 2 – Seasonality, ed. by W. Erwin Diewert, B.M. Balk, D. Fixler, K.J. Fox, and A.O. Nakamura (Vancouver, Canada: Trafford Press).
Diewert, W. Erwin, 2008, “New Methodology for Linking Regional PPPS,” ICP Bulletin, Vol. 5, No. 2, pp. 10–21.
Diewert, W. Erwin, Erwin, Saeed Heravi, and Mick Silver, 2007, “Hedonic Imputation Indexes Versus Time Dummy Hedonic Indexes,” in Price Index Concepts and Measurement, ed. by W. Erwin Diewert, John Greenlees, and Charles R. Hulten. (Chicago: University of Chicago Press for the NBER).
Hausman, Jerry, and Ephraim Leibtag , 2009 “CPI Bias from Supercentres,” in Price Index Concepts and Measurement, ed. by W. Erwin Diewert, John Greenlees, and Charles R. Hulten. Chicago: (University of Chicago Press for the NBER).
International Labor Organization (ILO), OECD, UN Economic Commission for Europe, and World Bank, 2004a, Consumer Price Index Manual: Theory and Practice (Geneva: ILO).
International Labor Organization (ILO), OECD, UN Economic Commission for Europe, and World Bank, 2004b, Producer Price Index Manual: Theory and Practice (Washington: IMF).
International Labor Organization (ILO), OECD, UN Economic Commission for Europe, and World Bank, 2009, Export and Import Price Index Manual: Theory and Practice (Washington: International Monetary Fund).
Roger, Scott, 2000, “Relative Prices, Inflation and Core Inflation,” IMF Working Paper 00/58 (Washington: IMF).
Roger, Scott, 2010, “Inflation Targeting at Twenty: Achievements and Challenges,” in Twenty Years of Inflation Targeting: Lessons Learned and Future Prospects,David Cobham, ed. by Øyvind Eitrheim, Stefan Gerlach, and Jan F. Qvigstad (Cambridge, UK).
Silver, Mick, 2007, “Core Inflation: Measures and Statistical Issues in Choosing Among Them,” IMF Staff Papers, Vol. 54, No. 1.
Silver, Mick, 2009a, “Do Unit Value Export, Import, and Terms of Trade Indices Represent or Misrepresent Price Indices?” IMF Staff Papers, Vol. 56, No. 2, 297–322.
Silver, Mick, 2009b, “An Index Number Formula Problem: the Aggregation of Broadly Comparable Items,” IMF Working Paper 09/19 (Washington: IMF).
Silver, Mick, 2009c, “The Hedonic Country Product Dummy Method and Quality Adjustments for Purchasing Power Parity Calculations,” IMF Working Paper 09/271 (Washington: International Monetary Fund).
Silver, Mick, 2009d, “Comment on CPI Bias from Supercentres” by Jerry Hausman and Ephraim Leibtag,” in Price Index Concepts and Measurement, ed. by W. Erwin Diewert, John Greenlees, and Charles R. Hulten (University of Chicago Press for the NBER).
Silver, Mick, 2010a, “The Rights and Wrongs of Unit Value Indices,” Review of Income and Wealth, Vol. 56, No. 777 (June), 206–23.
Silver, Mick, 2010b, “IMF Applications of Purchasing Power Parity Estimates,” IMF Working Paper 10/253 (Washington: IMF).
Silver, Mick, and Saeed Heravi, 2007a, “Why Elementary Price Index Number Formulas Differ: Price Dispersion and Product Heterogeneity,” Journal of Econometrics, Vol. 140, No. 2, 874–83.
Silver, Mick, 2007b, “Hedonic Indexes: A Study of Alternative Methods,” in Hard-to-Measure Goods and Services: Essays in Honour of Zvi Griliches, ed. by Ernst R. Berndt and Charles R. Hulten. (Chicago: University of Chicago Press for the NBER.
Silver, Mick, 2007c, “The Difference between Hedonic Imputation Indexes and Time Dummy Hedonic Indexes,” Journal of Business and Economic Statistics, Vol. 25, No. 2, pp. 239–46.
United Nations, 1981, Strategies for Price and Quantity Measurement in External Trade, Series M, No. 69.
UN Economic Commission for Europe (UN ECE), ILO, IMF, OECD, Eurostat, World Bank, and UK Office for National Statistics, 2009, Practical Guide to Producing Consumer Price Indices (United Nations)
World Bank, 2007, International Comparisons Program Handbook, 2003–2006 (Washington: World Bank).
Zieschang, K., 2010, “FISIM Developments in the National Accounts and Implications for the CPI,” presentation at the ECE-ILO CPI Workshop on Financial Services in the CPI, Geneva (May 10).
Seven Questions on the Role of Precautionary Savings in Open Economies