Home > Statistics > Quarterly national accounts > 2010 > 1st quarter > Changes in the compilation methods of quarterly national accounts

Changes in the compilation methods of quarterly national accounts

In May 2010 a new information system was introduced in quarterly national accounts (QNA), which enables utilisation of improved compilation methods from before. Data based on the new compiling methods will be released on 9 June 2010. The method changes cause small revisions to the time series data of quarterly national accounts starting from 1990.

In Finland quarterly national accounts are prepared using a so-called indirect compilation method. This means that data are not formed by summing from source data and by adding to the thus obtained raw data various corrections, as is done in most statistics. Instead, data are compiled from source data indicators and the levels of annual national accounts by means of mathematical and statistical calculation methods. The indirect compilation method is used for the compilation of quarterly accounts in most countries, because sufficiently exhaustive source data by quarter are not available on the whole national economy.

Changing methods

The main phases for the compilation process of quarterly national accounts are in chronological order: production of source data indicators, benchmarking, extrapolation, deflation, chain-linking and seasonal adjustment.

Of these, the methods for benchmarking, extrapolation and deflation are changed along with the new information system.

The key changes are as follows:

• benchmarking to the levels of annual national accounts is now done directly to the indicator time series of source data, when previously benchmarking was done to the quarterly accounts time series formed by extrapolation
• the extrapolation method is changed to the annual benchmark-to-indicator ratio, while the benchmark-to-indicator ratio of the corresponding quarter of the previous year was used
• industry-level deflators/price estimates of value added are now formed by weighting product-level price data, while before use was mainly made of one price index for each sector/industry combination.

In this context, benchmark-to-indicator (BI) refers to the ratio of the value of national accounts to the value of the source data indicator.

Aiming towards higher-quality time series

When the indicator series are benchmarked directly to the levels of annual national accounts and the extrapolation method is changed to be based on the benchmark-to-indicator ratio of the annual level, the time series of the quarterly national accounts will more closely follow the quarterly development of the source data indicators used in the compilation.

On account of the new benchmarking and extrapolation methods, the turning points in the economy can be located more accurately and reliably than before in the current year as well as in previous years.

The new method for forming deflators is uniform with annual national accounts and thus it produces price estimates that are closer to the prices of annual accounts. Consequently, the volume estimates of the value added of quarterly accounts and of gross domestic product are also more accurate and closer to the actual volume data of annual accounts.

Compilation of series at current prices

Benchmarking

An indicator series is produced from the source data for each QNA transaction/industry. Indicator is then benchmarked using the proportional Denton method ¹⁾ to the corresponding transaction/industry of annual accounts. Benchmarking can be made to the whole time series starting from 1990Q1 until the latest year of annual accounts. The proportional Denton method is based on BI (Benchmark-to-Indicator) ratios and aims to keep the quarterly development of the indicator series as intact as possible in the benchmarked time series.

During the planning stage of the new information system, benchmarking methods based on regression modelling were also examined and tested using Eurostat's ECOTRIM software, but the proportional Denton method was found to be better for the needs of Finnish national accounts. The proportional Denton method is actually a mathematically optimal benchmarking method if the aim is precisely to retain the development of the indicator as well as possible in the benchmarked time series.

Extrapolation

The latest QNA estimates for which there are not yet annual account data are now estimated by extrapolating with an indicator using the latest annual BI-ratio. The annual BI-ratio is then the latest annual account value divided by the sum of the indicator values of the same year. In extrapolation the quarterly value of the indicator is multiplied by the annual BI-ratio. At the maximum, five quarters have to be extrapolated (e.g. 2009Q1-2010Q1 in connection with the release for June 2010).

In the old compilation method, benchmarked quarterly accounts values were extrapolated onwards using the change of the indicator from one year back (I_Q/I_Q-4). The extrapolated quarterly accounts values were benchmarked with the Denton method to the annual accounts levels once the annual accounts were completed. A weakness of the old compilation method was that if extrapolation got "wrong" for even one quarter (indicator values may get revised after the extrapolation moment), or an extrapolated value was changed when balancing the accounts, it was possible that all corresponding quarters of the following years cumulated that same "error" or correction. By extrapolating a previously extrapolated estimate a chain-linked time series is actually constructed. With time the quarterly distribution of the time series (or seasonal variation) may thus get distorted and the time series can no longer be used for analysing quarterly development and the turning points of the economy.

In addition, extrapolation using the change of the indicator from one year back produces a time series departing from the development of the indicator for the latest quarters if the quarterly distribution of the benchmarked QNA estimates for the last benchmarked year differs from the quarterly distribution of the indicator. This makes it difficult to interpret the economic situation for the latest, that is, usually the most important quarters.

Compilation of volume series²⁾

Deflation

The time series at the average prices of the previous year are formed from the benchmarked and extrapolated time series at current prices by deflation. Deflation is made by dividing quarterly values at current prices by the change of the price index from the average price of the previous year³⁾. Industry-specific deflators are formed for deflation of value added by weighting product-specific price indices of output with the help of supply and use tables.

In the old compilation method a deflator was formed for output of each sector/industry combination using one price index. For example, in the manufacturing industries the deflator was the producer price index of the corresponding industry, which usually measures mainly the development of the price of the goods produced. In the last few years the manufacturing industries have increasingly started to produce services as well, so deflation based on one price index may produce poor price and volume estimates. In addition, the base years of the price indices may already be old, in which case the weight structure of the index is also behind time.

In the new deflator forming method, price data for several different products are used for forming a sector/industry deflator for each 138 value added series, Product-level deflators are then weighted with the most current product-structure data for output, obtained from the annual supply and use tables. The product-level deflator formation method is also used in National Accounts. The National Accounts supply and use tables contain nearly one thousand products for each of which are defined a price index or other price data best describing its price development. The new method for forming deflators for value added concerns the Quarterly National Accounts quarters starting from 2004Q1. Older deflators for the years 1990-2003 have not changed much. The deflators 2007Q1-2010Q1 for value added in the release for June 2010 were calculated using the product-weights of the 2007 supply and use tables.

In the same connection, the use of so-called double deflation was discontinued, which means that intermediate consumption is no longer estimated separately at current prices or the previous year's prices. There are no reliable quarterly value or volume indicators for intermediate consumption and estimation based merely on price data has proved to be ineffective as far as the reliability of value added estimates is concerned.

Volume series at the previous year's mean prices are benchmarked after deflation to annual accounts with the pro rata method, that is, each quarter of the same year is raised or lowered in the same proportion. The benchmarking system of volumes does not change.

Chain-linking

The benchmarked time series at the average prices of the previous year are chain-linked using the annual overlap method. The chain-linking method has not changed. Annual levels of the chain-linked (reference year 2000) volume series are equal to annual accounts when the time series at current prices and at the previous year's prices used in chain-linking have first been benchmarked.

Seasonal adjustment

Seasonal adjustment is done with the Demetra software using the TRAMO/SEATS method, as before.

¹⁾ The compilation formula for the proportional Denton method can be found in the IMF's QNA Manual, Chapter 6, page 87 (Formula 6.3). In the same chapter examples are also given of annual BI extrapolation (e.g. Example 6.1 p. 85 or Example 6.A2.1 p.110): http://www.imf.org/external/pubs/ft/qna/2000/Textbook/ch6.pdf

²⁾ Information on the quarterly accounts volume compilations can be found in Chapter 9 of the IMF QNA Manual: http://www.imf.org/external/pubs/ft/qna/2000/Textbook/ch9.pdf

³⁾ The formula for the deflator is: D_Q = P_Q/(P_Y-1), where P is the point figure of the price index or price data in euros. Then the volume is: FP_Q = CP_Q / D_Q, where CP is the (benchmarked) current price value of quarterly national accounts.

Further information about the methods of quarterly national accounts is available from:

• Bloem, A., R.J. Dippelsman and N.Mæhle: Quarterly National Accounts Manual - Concepts, data sources and compilation. IMF, 2001. http://www.imf.org/external/pubs/ft/qna/2000/Textbook
• Handbook on Quarterly National Accounts. Eurostat, 1999.

Inquiries: Pasi Koikkalainen, +358 9 1734 3332, Samu Hakala, +358 9 1734 3756

---

Last updated 9.6.2010