1

COMMENTS ON THE NBS PAPER

“Accounting Methods of China’s Annual Expenditure-Based GDP”

Ramesh Kolli[1]

Additional Director General

Ministry of Statistics and Programme Implementation, India

Introduction

The expenditure based GDP estimates are compiled in China both at national and regional levels. These follow largely the 1993 SNA concepts. A long-term data series on expenditure based GDP is apparently available from 1978 for China.

It is important that GDP is compiled by all the three approaches, namely, production, income and expenditure and preferably through supply-use framework. The expenditure GDP assumes more importance as it depicts the final uses of output in terms of consumption, capital formation and net exports. In this context, the compilation of expenditure GDP estimates by the NBS is in line with the best practices of compiling national accounts.

At regional level, it is generally difficult to compile expenditure based domestic product estimates, due to the problems in the availability of data on exports and imports, as also on the inter-region movement of goods and services within the country. It will be interesting to know the procedures adopted by the NBS to overcome these data issues in compiling regional level expenditure based GDP estimates.

In India, regional domestic product estimates (known as state domestic product or SDP), are compiled by following the income originating approach, which is similar to the production approach GDP at the national level. This method provides estimates of domestic product at regional level by economic activity. Consequently, since expenditure based GDP is not compiled at regional level in India due to the lack of data on imports and exports, estimates of consumption expenditure, capital formation and net exports are not available at regional level from the national/regional accounts.

I. The frame and classification of expenditure-based GDP accounting

The framework and classification adopted to compile the annual GDP by expenditure approach in the NBS is as per the standard procedures. However, under the final consumption expenditure, the NPISH part has not been shown. Apparently, the NBS considers that most of the NPIs in China are either serving government or businesses and the actual contribution of NPISHs is relatively insignificant. However, in most developing countries, the NPISH is an important institution. A small sample survey of few registered charity units could provide some base line estimates. India is presently conducting a census of all NPIs, which includes preparation of a register of all the NPIs in India and collecting financial data from the functioning NPIs. The overall objective of this census is to implement the UN Handbook on NPIs, and provide separate accounts of NPISHs in the national accounts. These objectives are expected to be achieved by December 2010.

The third-class classification adopted for household consumption expenditure by the NBS broadly conforms to COICOP classification. It is not clear from the paper whether the NBS is adopting the COFOG classification for the government final consumption expenditure or not. Presenting general government expenditures by COFOG is important for understanding the purposes for which the expenditures are incurred. In India, a two-way classification of government expenditures is presented, namely, by purpose and by economic.

The manner in which the estimates of GFCF are prepared by the NBS is quite elaborate and takes care of all conceptual issues. Few areas of improvement could be (i) the procedures to estimate construction appear to be a bit complicated; (ii) the repair and maintenance part of output of construction should be subtracted from the output to get to GFCF, the NBS might be doing this, but it is not clear from the paper, (iii) under Gross fixed capital formation, the number of assets appeared to be too broad, particularly for the machinery and equipment part. Perhaps, it is desirable to attempt to break this item into few important assets, such as transport equipment, and (iv) at some stage, the NBS could also contemplate on including R&D expenditures and military capital expenditures which can be used by civilians under GFCF and valuables in the GCF.

In respect of change in inventories, the methods followed by the NBS are apparently to inflate the current year’s opening stock to the level of current year’s closing stock and then to take the difference. The practice followed in India is to bring all inventories of various years to constant prices, then take the difference to get to the change in inventories at constant prices and then inflate this to obtain the estimated change in stocks at current prices. However, this procedure needs adjustment to the data on inventories provided by the businesses to the prevailing market values, as businesses normally provide data on inventories normally at costs or at realisable values, rather than at the prevailing market value.

The data on imports and exports appear to be single entries in the expenditure GDP. These could be considered for presentation in terms of goods and services separately.

The NBS may also plan to building up a capital stock data series from the available data on GFCF using the PIM. This would enable them to switch over from ‘depreciation’ to ‘consumption of fixed capital’ There are procedures available for building up opening stock figures prior to the year for which estimates of GFCF are available. Alternatively, one could attempt to estimate opening capital stock from the capital output ratios of the first year for which GFCF data is in place and apply the same on output of the previous year. However, this needs to be done carefully at very detailed level, by enterprise/economic activity.

II. Household Consumption Expenditure

The household consumption expenditure is estimated separately for rural and urban households. The scope and coverage to estimate consumption expenditure appear to be as per the standard norms. One comment that can be made here is with reference to the manner in which the consumption expenditure is estimated in the NBS, which is to compile directly from the survey results. If the sources of data for consumption expenditure and output are different, there is a possibility of mis-match between the data of consumption expenditure and output in the national accounts, which could lead to a large discrepancy between production and expenditure GDP. The ideal procedure to obtain consistent estimates of consumption expenditure (as also other aggregates) is through the supply-use framework.

In India, private final consumption expenditure (covering both households and NPISHs) is estimated through commodity-flow approach, starting from the overall output of each commodity and services and deducting from this, the intermediate use and all other final uses. In this procedure the under-estimation in consumption expenditure in survey results is taken care of and the PFCE esitmates are consistent with overall output estimates.

  1. Calculation at current price

Expenditure on food, etc.: The method of compiling is in order. However, consistency between production and consumption data needs to be ensured, if they are coming from two different sources. NPISHs consumption expenditure could be attempted and included.

Expenditure on FISIM: The method of compilation appears to be correct, except that the forumula of (interest of loans – interest rate of deposits)/2 may be (interest of loans + interest rate of deposits)/2. A second observation is that FISIM is normally allocated to industries and final users. The method followed by NBS is different from this procedure and may give rise to discrepancy between FISIM of industries and final users and the FISIM shown under financial intermediation. In India, FISIM from different financial institutions is allocated on the basis of indicators of deposits and loans, separately in respect of each financial institution.

Expenditure on insurance services: The formula for estimating this component appears to provide reliable estimate. However, in general, life insurance and non-life insurance components need to be treated separately, since expenditures on these two types of insurance have different purposes. While, premiums on non-life insurance by households is in the form of consumption expenditure, it is only the commissions part of life insurance premiums and premium supplements that constitute consumption expenditure.

Expenditure of owner-occupied dwelling services: The procedure is in order. However, the standard procedure is to adopt user cost approach if less than 25% of all dwellings in the country are actually rented. The second observation is that local taxes, fees if any may also be included in the cost of housing estimates. Further, it is important to build up capital stock data to estimate consumption of fixed capital.

Expenditure on other goods and services: NBS estimates this using per capita expenditure. Although this procedure is appropriate, an alternative option could be the commodity flow approach.

2 .Calculation method at constant price

The deflators used here are appropriate. In respect of financial and insurance services, the implicit price deflator of the output of financial intermediation services could be an alternative. Another alternative could be the rates of interest index of deposits and loans.

III. Government consumption expenditure

The coverage of government final consumption expenditure includes central government, provincial governments, local bodies and autonomous government institutions. It is presumed that the NBS takes into account the final expenditures of these institutions of general government, otherwise all these organs of government should be covered. Most of the local bodies and autonomous government institutions also generate own resources, besides the current and capital transfers they receive from the higher level of government.

1. Calculation at current price

The compilation procedure appears to be in order. However, there are two comments on this:

(i)pension costs are excluded from the consumption expenditure, which is correct, as they are transfers. However, the expenditures on wages and welfare of the government should also include the imputed pension liabilities of present employees. In India, since this imputed pension liabilities are not estimated, the actual pension payments are taken to be equal to the imputed pension liabilities of present set of employees and added to consumption expenditure of general government.

(ii)Depreciation estimates need to be in terms of consumption of fixed capital.

  1. Calculation at constant price

The deflator for depreciation is taken to be the Producers’ Prices of Industrial Products. An alternative to this could be the implicit deflator of GFCF. Similarly, CPI is used to deflate the other component of GFCE. In India, different price indices (producer, consumer and implicit industry-wise GDP deflators) are applied on each component of the GFCE, namely, the compensation of employees and purchase of goods and services, as also in respect of different Ministries. The consumer price index is also the wage index in India and is, therefore, applied on the compensation of employees.

IV.Gross Fixed Capital Formation

The procedure followed for estimating GFCF for construction though conceptually correct, appears to be somewhat complicated. The GFCF of construction is the output of construction, minus, repairs and maintenance. The output of construction is the value of material inputs+value of factor inputs (the gross value added is the value of factor inputs). The NBS is estimating the value of output in terms of sales of units, while in India, it is estimated on the basis of value of available of basic construction materials, other construction materials and the factor inputs. It is not clear from the paper whether the NBS removes the repair and maintenance part of construction output to arrive at GFCF of construction.

The machinery and equipment part of GFCF is estimated by NBS as sum of fixed-asset investment (units above 500 thousand RMB Yuan) and investment in fixed asset below this limit, besides the associated costs. Conceptually and in terms of coverage, the procedure followed by the NBS is correct. Data is also available from the sources directly (from enterprises). In India, due to the limitation of industry-wise data on net acquisition of fixed assets (quality of available data is poor), the commodity-flow approach is adopted whereby the machinery and equipment produced and imported (exports and other uses, particularly consumer durables, subtracted) is taken into account to estimate the GFCF of machinery and equipment.

In the NBS procedure, two improvements could be suggested, (i) the segregation of machinery and equipment to few more assets such as transport equipment and (ii) inclusion of valuables in the GCF. It appears that NBS does not take into account any part of military capital expenditures (used by civilians) as GFCF, which could be considered for inclusion in future, though 2008 SNA suggests capitalising weapons expenditures also. Also, the 2008 SNA treatment of R&D expenditures as GFCF could also be considered for implementation.

For the constant price estimates, different deflators are used by the NBS, which seem to be appropriate. In India, a number of specific price indices are compiled at industry/asset level and used as deflators.

V. Increase in inventories

The title of this section could be change rather than increase in inventories, as it is change in stocks (could be positive or negative) which is a component of expenditure GDP.

The definition provided in the paper is conceptually correct, except that when output is estimated as production*price, finished goods do not form part of inventories. They become part of inventories when output is estimated as sales+change in inventories.

NBS updates the inventories at the beginning of the year with relevant price indexes to bring them to the price levels of the reference year and when this estimate is subtracted from the inventories at the close of the year, the change in inventories data is obtained. This is done by industries (and by broad three groups of institutions, state-owned enterprises, large enterprises and others; it is presumed that inventories with government and government commercial undertakings are also taken into account). In India, the inventories of different years are first brought to the base year prices using relevant price indexes (generally the producer price indices at detailed level) and the difference between two successive years’ inventories is estimated as the change in stocks at constant prices. These estimates are then revalued to the current prices using the same price indexes. This procedure is followed at each industry/institution level.

In most developing countries, estimating inventories is a problem area, particularly for the household/informal segment of the economy. Generally, benchmark estimates are prepared based on benchmark enterprise surveys or economic census and proxy indicators are used to extrapolate base year inventories to other years for the informal sector.

The NBS paper mentions that economic census data is used to estimate the inventories for smaller enterprises. Apparently the extrapolator for the benchmark estimates is the retail price indices. In India, the data on short-term credits to the household industries is used in some cases.

Another problem area in the estimation of inventories is that the businesses report inventories at ‘cost’ or at realizable value. It is very difficult to estimate these inventories at current market prices. Only after this, one should apply relevant deflators to get to the inventories at constant prices

The NBS uses different price indexes to deflate the current price estimates of change in inventories in respect of different industry groups. The inventories in some industries (such as industry and trade) are further segmented broadly under three commodity groups, so that more appropriate deflators could be applied at commodity/industry level. The procedure is elaborate and in order, but NBS may consider using deflators on each year’s inventories (at industry/institution/commodity group) rather than applying deflators directly on change in inventories.

VI. Net Export of Goods and Services

Current prices: The method of compilation is as per the standards. It appears that foreign trade data is available only in terms of US dollars, which is converted into local currency with annual average exchange rate. It is better if the data from source is available in terms of CNY, rather than in terms of US dollars. Secondly, NBS compiles import data on FOB basis, which is the standard procedure. Most countries could present import data on CIF basis only. The sources for goods and services are respectively the Customs and BOP statistics. There could be discrepancy in the two sources of data in respect of goods.

Constant prices: The way NBS compiles net exports of goods and services at constant prices is quite interesting. All the four components of imports/exports and goods/services are deflated by the respective price indices (presumably the unit value indices for goods), and then the additions and subtractions are carried out to obtain net exports of goods and services. These are in terms of US dollars. The constant price estimates in terms of US dollars are then converted to CNY using base year annual average exchange rate. The price indices for services are the services part of CPI for exports and export of services price index for select countries for import of services. The procedure appears to be in order.