OECD UNITED NATIONS JOINT OECD/ESCAP MEETING ON NATIONAL ACCOUNTS System of National Accounts: Five Years On. Bangkok, 4-8 May 1998

Transcription

1 OECD UNITED NATIONS ORGANISATION FOR ECONOMIC CO-OPERATION AND DEVELOPMENT ECONOMIC AND SOCIAL COMMISSION FOR ASIA AND THE PACIFIC JOINT OECD/ESCAP MEETING ON NATIONAL ACCOUNTS 1993 System of National Accounts: Five Years On Bangkok, 4-8 May 1998 Title: Price and Volume Measurement: Where do we stand? Agenda item: 9 Author: Economic Planning Agency, Japan

2 PRICE AND VOLUME MEASUREMENT: WHERE DO WE STAND? Mr Hiroki Owaki Department of National Accounts Economic Planning Agency, Japan May The purpose of this paper is to provide some information on discussion held in Japan relating to the introduction of chain indices in the national accounts. In conclusion, although merits of those indices are well understood, it is still too early for the Department to incorporate them in its system of national accounts owing to several reasons including their non-additive nature, their vulnerability to revision practices and cost for implementation. First, current practice in compiling deflators and main aggregates at constant prices are described. Second, current revision practice is provided with some revision history. Third, sample estimates of chain-type Paasche deflators for main aggregates of expenditures are introduced. Fourth, pros and cons of the introduction of chain indices are discussed with some concluding remarks. I. Current Practice in Price and Volume Measurement 2. All constant price estimates in the national accounts of Japan are made by the deflation method and therefore any physical volume or quantity estimates are not involved. 3. Estimation of deflators includes two steps. First, the basic unit deflators are constructed. They are composed of the basic unit deflators corresponding to the six-digit level of commodity classification in the commodity flow method, deflators for construction and deflators for services rendered by the general government and the non-profit private institutions serving households. The first one is constructed for each of around 400 commodity groups in accordance with the Paasche price index formula. 1 Price data are taken from many kinds of existing price indices such as the consumer price index and the wholesale price index. 2 Nominal estimates on relevant transactions of around 2,200 commodities, made by the commodity-flow method, are utilized for weight. Deflators for construction and services rendered by the general government and the non-profit private institutions serving 1 In the national accounts, constant price series should be the values obtained by multiplying quantities with respective prices for the base period, expressed as Q t (where P: price, Q: quantity, subject 0: base period and subject t: current period). Therefore, deflators, by which nominal prices ( P t Q t ) are divided and constant price series ( Q t ) are derived, should necessarily be constructed as in the Paasche-type, expressed as. In the current national accounts, attempts are made to construct Paasche-type deflators at as detailed a level as possible. 2 In case price indices are not available, unit value indices are prepared although changes in quality or in product composition are necessarily reflected by changes in the price indices. When unit value indices are not available, cost of inputs should be adopted. 2

3 households are estimated on the basis of cost for production of those services. And cost indices are also constructed in accordance with the Paasche index formula, using data including those from the commodity flow estimation. 4. Second, for major demand and production aggregates, commodity compositions are identified at the 400 commodities level. And series of values at constant prices are prepared by deflating nominal values for each commodity included in the aggregate concerned by the basic unit deflators and then dividing the total nominal values of the aggregate by the sum of the real values for each commodity. Accordingly, values at constant prices of a particular year are estimated implicitly by using Paasche-type deflators. Gross domestic product by industry at constant prices is estimated by the double deflation technique in which it is derived by subtracting intermediate consumption at constant prices from value of production at constant prices. II. Revision Practice 5. Japan is one of those countries who have adopted fixed-based constant price estimates as volume measurement and new benchmark year prices have been applied retroactively for a certain length of the estimation period. Currently, 1990 prices are used for constant price estimates on the expenditure accounts for 1985 onward prices are used for constant price estimates for the period of 1970 to prices are used for the rest of the estimation period (1955 to 1969). 6. With regard to the production accounts, two kinds of constant price estimates are provided. In one of them, the full length of the constant price series is expressed in the latest base year (1990) prices. In the other one, constant prices series are linked in three years (1970, 1980 and 1985). Consequently 1965, 1975, 1985 and 1990 prices are used for constant price estimates for the period of 1955 to 1969, 1970 to 1979, 1980 to 1984 and 1985 onward. 7. This practice of splicing earlier constant price estimates on to later ones is taken because it would be unreasonable to evaluate quantities at prices of a particular year for the period of more than forty years. At the same time a certain length of the period for which constant price estimates at prices of a particular year is identified in order to ensure that those constant price estimates are additively consistent. The length of ten to fifteen years has been considered to be appropriate The benchmark year revision takes place every five year, leading to revision to both nominal and real estimates. The latest experience shows that the benchmark year revision resulted mostly in upward revision for deflators, which had real growth implication in the opposite way. 5 The growth of 3 The latest benchmark year revision was held in Linking constant price estimates by different base years makes series of constant price estimates lose additive consistency. In the expenditure accounts, any gap arising from linking is closed by allocating the gap to components while the total is estimated by adding up components in the production accounts. In either case, since more emphasis is put on additive consistency, we don t have such practice as publishing any gap as it is and asking users to deal with it. 5 The reason for the upward revision of deflators can be illustrated in the following way. Growth of deflators from year 0 to year 5 with the base year of year 0: A= (ΣP 5 Q 5 )/(ΣP 0 Q 5 )/1=(ΣP 5 Q 5 )/(ΣP 0 Q 5 ) Growth of deflators from year 0 to year 5 with the base year of year 5: B = 1/ΣP 0 Q 0 )/(ΣP 5 Q 0 )=(ΣP 5 Q 0 )/(ΣP 0 Q 0 ) Then, B/A=[(ΣP 5 Q 0 )/(ΣP 0 Q 0 )]/[(ΣP 5 Q 5 )/(ΣP 0 Q 5 )] = Laspeyres prices index / Paasche price index > 1 This relationship holds for growth between the two benckmark years and does not necessarily hold for other periods. 3

4 GDE deflators for 1980 to 1985, 1985 to 1990 and 1990 to 1993 was revised up by 0.4 percent, 0.1 percent and 0.1 percent, respectively (Table 1). 6 Of particular note is that changes to growth of deflators of private fixed capital investment were relatively considerable, reflecting larger changes in relative prices for commodities included in this category of expenditure. The change ranged from 0.7 percent to 1.6 percent for 1990 through III. Sample Estimates for Chain-type Paasche Deflators 9. The Department of National Accounts has produced chain-type Paasche deflators for Gross Domestic Expenditure as supplement to official estimates for deflators (Table 2). 7 The estimates show that the growth of Paasche price indices is smaller than that of chain-type Paasche price indices. And the magnitude of changes in the growth of deflators is limited with the range of 0.0 to 0.5 for most of the expenditure aggregates for most of the estimation period although a number of significant changes (more than one percent) are observed for some aggregates for some years. With regard to GDP deflators, the changes were insignificant except for 1995 when they were compared with real GDP growth rates for the same period. 10. The Department needs more time and resources to work out the new type of deflators until they could potentially be used for official estimates. Issues to be addressed include compiling chain-type Paasche indices for all individual deflators to be integrated to a higher level of chain-type Paasche price indices. 11. Another issue is related to the current practice in estimating deflators for major aggregates implicitly because those deflators are not directly measurable. 8 This practice is based on the assumption that constant price estimates have the property of additive consistency. Unless this understanding is wrong, implicitly estimated chain indices are always misleading because constant price estimates by chain price indices do not have such property. The Department needs to have clear understanding as to whether implicitly estimated chain indices are identical to those which would be estimated if they were directly measurable. The issue is also related to the appropriateness of use of the double deflation technique in estimating gross domestic product at constant prices in the production accounts. IV. Discussion on Chain Indices 12. Although the Department has worked on chain Paasche price indices, it has not adopted chain indices in its official estimates yet not only because the work on the new type of indices is still under way as noted above, but also there are some other issues to be addressed. 6 The revision for the period of 1980 to 1985 was related to some technical issues other than the change of the benchmark year from 1985 to We call them chain-type Paasche indices because they are not recognized as pure chain indices mainly due to the occasional inclusion of simple Paasche price indices at lower level of commodity breakdown. The formula used was the following. D t =D t-1 *((ΣC t /(ΣC t *d t-1 /d t )) where D t : Integrated Deflator C t : Nominal Value D t : Individual Deflator And we should note that deflators for higher level aggregates including gross domestic expenditure were estimated using nominal values and their corresponding deflators for their components. 8 Examples are the following. Net exports of goods and services = Exports of goods and services Imports of them. Gross Domestic Expenditure = Domestic demand + Net exports of goods and services. 4

5 13. First, the non-additivity of chain indices, which has been widely recognized and discussed, is a serious drawback. Users, in particular, business cycle analysts including government officials in charge of such a project, are very sensitive to the non-additive nature of the indices because they pay strong attention to contribution to growth of an individual demand item to derive policy implication. Given that growth rates have become low recently with a range of 0-2 percent, the statistical discrepancy can not be ignored. Use of data at constant prices for macro-econometric models, which have been widely developed in the country in both private and public institutions, also requires constant price estimates to be additively consistent at least for the estimation period of the model. 14. Second, at the time of benchmark year revision, not only deflators but also nominal values are revised retroactively for around ten years owing to revision and/or release of basic statistics including input and output tables. For example, at the time of the latest benchmark year revision in 1990, growth rates of nominal GDP were revised up by 0.2 percent on average during 1985 to 1993, of which around 0.1 percent was due to changes in GDP deflators (Table 3). Therefore, real GDP growth rate would have been revised even if chain indices had been adopted in the past. More critically, chain indices themselves would have been revised because of revisions to commodity flow data used for weight. These points would lessen the theoretical robustness of chain indices against periodical revision. 15. Third, chain indices should not be used for quarterly estimates as described at paragraph in SNA1993. Using different kinds of indices for annual and quarterly estimates would make confusion to users and risk undermining credibility of official statistics on such a critical economic indicators as GDP. 16. Fourth, the implementation of chain indices is very costly. In the current practice, nominal values and prices for more than 2000 commodities are used in estimating basic unit deflators, which amount to around 400 commodity groups. A considerable amount of resources has to be available to the project. 17. Having discussed this way does not imply that we don t understand the usefulness of chain indices in analyzing long term economic growth. It s a matter of striking a right balance between merit and demerit for users and implementation cost accruing to compilers of statistics. As we discussed in section II, it might be recalled that the Department has linked constant price estimates at different base year prices with ten to fifteen year intervals. This practice could have been considered to be a compromise, taking into account the users conflicting interest, which are the additive consistency and the quality required as indicators for long-term economic growth. 9 Replacing the fixed-weight indices with the chain ones will lead to ignoring one of the two interests of users, which should still be deemed critical in the country. 18. At this stage the Department does not intend to change the current practice of price and volume measurement. However, the Department will keep studying and working on the chain-type Paasche price indices aiming at utilizing them in estimating main aggregates at constant prices and their growth rates. For the time being, these results will be published separately from official annual estimates for the system of national accounts as they were done so in the past It could be argued what the best length of period for which prices of a particular year are applied is. 10 The preliminary estimates on chain-type Paasche indices were published in the Quarterly Journal of the System of National Accounts (in Japanese) by the Department of National Accounts of the Economic Planning Agency. 5

6 REFERENCES EC, IMF, OECD, UN and World Bank, 1993, System of National Accounts 1993 Department of National Accounts of the EPA (Economic Planning Agency), 1980, A System of National Accounts in Japan: Definitions, Concepts, Sources and Methods Department of National Accounts of the EPA, 1994, A Report of the Special Research Group on the SNA1993 (in Japanese) Department of National Accounts of the EPA, 1996, Report on Revised National Accounts on the basis of 1990 (in Japanese) Department of National Accounts of the EPA, 1997, The Minute of the Third Meeting of the sub- Committee on Production and Expenditure of the Advisory Council on the SNA (in Japanese) Landfeld, J.S. and Parker, R.P., 1997, BEA s Chain Indexes, Time Series and Measures of Long-Term Economic Growth, Survey of Current Business, Vol.77, no.5. Young, A.H, 1992, Alternative Measures of Change in Real Output and Prices, Survey of Current Business, Vol. 73. Any comment on the paper will be appreciated. Please send them to hiroki.owaki@epa.go.jp. 6

7 Table 1. Japan: Revision in Growth of Deflators (annual average, in percentage) 11 Gross Domestic Expenditure Private Final Consumption Expenditure Please note that revision for 1980 to 1985 was related to some technical issues other than the change of the benchmark year from 1985 to

8 Private Housing Investment Private Fixed Capital Investment Table 1. (Continued) Japan: Revision in Growth of Deflators (annual average, in percentage) Please note that revision for 1980 to 1985 was related to some technical issues other than the change of the benchmark year from 1985 to

9 Government Final Consumption Expenditure Government Fixed Capital Formation Export of Goods and Services

10 Import of Goods and Services

11 Table 2. Japan: Paasche vs. Chain-type Paasche Deflators (growth rates, in percentage) By Paasche indices (A) 1/ Private Final Consumption Government Final Consumption Gross Capital Formation Fixed Capital Formation Private Public Exports of goods and services Imports of goods and services Gross Domestic Expenditure Domestic Demand By Chain-type Paasche indices (B) Private Final Consumption Government Final Consumption Gross Capital Formation Fixed Capital Formation Private Public Exports of goods and services Imports of goods and services Gross Domestic Expenditure Domestic Demand

12 Difference: (A)-(B) Private Final Consumption Government Final Consumption Gross Capital Formation Fixed Capital Formation Private Public Exports of goods and services Imports of goods and services Gross Domestic Expenditure Domestic Demand / 1990 is the base year. 12

13 Table 3. Japan: Revision in Growth of GDE (annual rates, in percentage) 13 Gross Domestic Expenditure at Current Prices Gross Domestic Expenditure at Constant Prices Please note that revision for 1980 to 1985 was related to some technical issues other than the change of the benchmark year from 1985 to

14 GDE Deflators