The 13th OECD-NBS Workshop on National Accounts December, 2009
Accounting Methods of China’s Annual Expenditure-Based GDP (by Wu You, NBS)
Comments on NBS paper by Roger Jullion, Statistics Canada
China currently publishes annual estimates of expenditure-based GDP in current prices only. These data are published alongside the official production or industry-based GDP estimates which are released quarterly. The NBS paper discusses the methodology currently used to estimate the annual expenditure-based estimates along with the level of detail at which these estimates are prepared and the price deflators used to produce the constant price estimates. The methodologies used to compile these estimates have undergone continued improvement since they were first compiled on a trial basis in 1989. The paper discusses each of the five major expenditure-side categories, i.e. household consumption expenditure, government consumption expenditure, gross fixed capital formation, inventory change and net exports of goods and services. While the provinces have established annual expenditure-based accounting systems, the following comments are confined to the national estimates only.
Household Consumption Expenditure
The primary sources of data for estimating 8 of the 11 categories of household consumption are the Rural Household Expenditure and Urban Household Expenditure surveys. Estimates are made for goods produced and consumed by households and for payments in kind. While the current methodology categorizes household spending into 11 categories, more detailed information is available from the household expenditure surveys. Therefore one suggestion is to compile the data at a more detailed level and in particular, breakout some of the larger expenditure categories such as spending on motor vehicles, energy products and some selected services. The data would be more analytically useful and data problems could more easily be identified with more detailed information available. In addition, the constant price estimates would be more accurate if deflation was carried out at a more detailed level.
In some cases, NBS analysts have much better data available for selected spending components such as motor vehicle sales, and these alternative data sources could more easily replace or be compared to the household expenditure data if the estimates were compiled at a more detailed level. At Statistics Canada we use a combination of supply-side and demand-side information in compiling our household consumption estimates and NBS should continually confront the demand side data with survey data on the supply side, e.g. retail and wholesale sales, household services and transportation surveys.
Adequate coverage of high income households is recognized as a challenge for household expenditure surveys in many countries and as a result NBS analysts should pay particular attention to those infrequent purchases made by these households where the results can be volatile due to large sampling errors. The floating population also poses measurement challenges in determining the most appropriate average spending data to use.
NBS uses supplementary information to compile estimates of FISIM, insurance and owner occupied rents. For FISIM, aggregate FISIM should be allocated by sector rather than an explicit FISIM estimate constructed for the household sector. If household spending on FISIM is estimated independently from the total FISIM output then there is a risk of the two methodologies getting out of step with one another. Since household FISIM is attributable to both depositors and borrowers, it is not clear from the paper why the difference between the interest rate on loans less the interest rate on deposits is divided by two. Are the depreciation estimates for housing based on PIM (Perpetual Inventory Method) estimates of housing capital stock?
With respect to the price deflators, detailed consumer price indexes should be used to deflate the individual expenditure components. For financial and insurance services, volume indicators of activity or deflated financial assets and liabilities could be considered as alternative approaches to using the aggregate fixed asset investment price index.
Government Consumption Expenditure
The paper lists the industrial sectors included as part of government spending. Does government consumption expenditure, as compiled by NBS, include all non-profit institutional units, including those serving households? Included in this list are nursery and bury services to households and other services as well as culture and sports. Is only the government contribution to these organizations included? These government payments to non-profit organizations serving household may be more appropriately treated as transfers to the household sector and/or to NPISH where the expenditures would be reflected.
When current operating expenditures are defined in the paper, what does NBS include in wages and “welfare”? Are these employer contributions to employee benefit plans e.g. pensions? Why is the component (subsidies on individuals and families-pension costs-living allowance-relief costs-school aid costs-production subsidies) included? What does this residual represent? Payments to individuals and families should be treated as transfers and excluded from government final consumption. Depreciation should be at replacement cost rather than at original cost.
Government deflation could be improved by deflating the government wages and salaries portion of government current expenditure with an average government wage deflator or alternatively by projecting the base period wage and salaries level on government employment or hours worked. Other goods and services could be deflated with a combination of PPI’s and CPI’s. Government depreciation estimates in real terms is generally estimated as part of the PIM calculation.
Gross Fixed Capital Formation
In compiling residential construction, the paper indicates that to the value of residential investment NBS adds the value added of residential sales less corresponding fees in land acquisition, purchase and resettlement compensation. The value added from residential sales appears to refer to the mark-up or the difference between the sales value and the cost of the project. This component should be added only if the value of residential investment excludes this mark-up. It is not clear from the paper that this is the case. A similar comment can be made with respect to the non-residential construction estimates.
Investment in construction which is currently based on the Fixed Asset Investment Survey should be compared or confronted with data from the Survey of the Output of Construction Enterprises and also with data on the supply of labour and materials used in construction. Significant differences in growth rates among these three sources should be investigated. This would ensure better coherence with the production-based GDP data where the output of construction enterprises is used as the basis of the value-added estimates.
A commodity flow model for machinery and equipment should be developed to ensure coherence between the fixed asset investment data and a commodity flow model where production, inventory, export and import data are used to derive the domestic supply of M&E. Any significant differences in the data generated from these two approaches should be investigated.
For computer software, own account software development should be included in addition to custom software development and ‘off the shelf’ software purchases of government and businesses.
Real output of construction is a major contributor to economic growth in China. Rather than using the aggregate construction and installation price index within the fixed assets investment price index as the price deflator for construction, would it be possible to distinguish between residential building prices, non-residential building prices and non-residential engineering prices in collecting individual prices from respondents?
Using balance sheet information to estimate gross fixed capital formation poses some challenges. Purchases of land, which are excluded from the national accounts definition of capital spending, would need to be separately identified on the balance sheet as would revaluations and other adjustments. Assets may not be entered on the balance sheet until they are ‘put in service’ making the estimation of capital spending on large multi year engineering projects difficult. While a survey of capital spending is the preferred data source since definitions can be specified, balance sheet information could be used for selected industries or portions of industries where survey coverage is incomplete.
Inventory Change
The estimation of inventory holding gains and losses is an important element in the measurement of inventory change. The paper correctly states that once the change in the volume of inventory is estimated this volume should be re-valued using average prices of the period. However the paper is not specific in describing which price (i.e. annual average, end of period) is used to inflate the beginning inventory of the period and how nominal and constant price inventory change is estimated. Perhaps a numerical example would be helpful to illustrate the steps in the computation.
In Canada, where inventory accounting is generally on a “first-in first-out (FIFO) basis, the ending or closing inventories in each period are deflated using an average of monthly price indexes that reflect the turnover period of the inventory. For example, if we determine based on sales and stocks that there is the equivalent of two months of purchases in inventory, we would average the purchase prices of the last two months of the period and use this as a price deflator for the closing inventory book values. The first difference of the deflated book values gives the constant price change in inventory. This volume change is re-valued to nominal values using average prices for the period. The calculation is done by industry and the prices used include CPI’s, PPI’s and import prices.
For the industry sector, the NBS paper mentions that the data are “estimated by census data in non-census years”. Which data source is used for the most recent time periods? Are inventory data collected for manufacturing and mining on monthly and/or on annual surveys? How would inventories of imported crude oil and mining commodities be tracked by NBS industry surveys?
For retail and wholesale inventories above the designated size, it is not clear what “computed by related financial term tables” refers. Source of the information is Department of Trade and External Economic Relations of NBS.
The assumption that inventories of unsold buildings is 5% of the inventories of the real estate industry appears low. What would the other 95% represent? Only new buildings are relevant for aggregate inventory change and/or aggregate gross fixed capital formation.
Net Exports of Goods and Services
The customs data for goods (for both values and prices) are compiled and published in $US. These customs data are converted to a BOP basis. Both the nominal values and the trade data at constant prices, as suggested in the paper, need to be converted to RBM before including them in expenditure-based GDP. The same principle holds for trade in services where data are compiled by BOP.
Deflation of goods uses unit value deflators available from Customs. For commodities where quality change may be important, e.g. electronic goods, NBS could consider using a detailed producer price index for exports if this price index accounts for quality change. For imports, the paper states that some imports prices should be adjusted according to consumer price indexes of the main import country or district. Is not clear which commodities this alternative method would apply to. Along those lines, NBS could use counterparty price indexes available from other countries as alternative import price indexes for those goods where quality change may be important. NBS should make the appropriate exchange rate adjustments to the $US-based prices in order to convert them to RMB.
For services, NBS uses CPI’s to deflate services exports and a basket of services export price indexes from China’s main trading partners to deflate imports of services. Alternatives to using domestic CPI’s would be to use average earnings indexes if available. Travel payments and receipts and transportation services should be deflated separately.
5