# Aggregation methods

To provide an appropriate summary and global overview, the Economic Outlook projections for individual countries and regions, both OECD and non-OECD, are commonly aggregated into and reported separately for regional and economic groupings and also at the total OECD and world levels.  For example regional and global aggregates are commonly reported in the summary tables and thematic tables included in the General Assessment of the Macroeconomic Situation chapter and its Statistical Annex.  This section discusses the various methods used in constructing such aggregates and some of their implications.

The specific country and regional groupings and the composition of the respective aggregates are listed in the Economic Outlook database inventory. Aggregate series are typically constructed for the most important variables, including gross domestic product (GDP) and its demand components, inflation and labour market indicators, the government accounts and international trade related variables. The complete list of variables is available in the Economic Outlook database inventory. With the exception of the ratios, aggregates are only computed in quarterly frequency, with annual data derived from the quarters.

Weighting methods

The general class of weighting method used in the Economic Outlook, broadly consistent with current National Accounts practices, is a chain-linked method (ref) where the weights typically applied to rates of change or levels, depending on the concept being aggregated, move over time (year or quarters). The broad justification for such an approach, as discussed by Whelan (2000), (see also the OECD Statistical Glossary, is that moving weights take better account of composition changes over time, especially those related to changes in relative prices or technologies. Depending on the quantity under consideration, one of two specific weighting schemes is adopted:

Method 1 (especially relevant to varaiables expressed as ratios or shares or rates of interest):

Weights are applied to the variable in level terms:

where w is the relevant share of country i for period t in the region under consideration, and X is the variable to be aggregated.

Method 2 (relevant for series in absolute terms, e. g. real GDP, prices and exchange rates):

Aggregation proceeds in three steps. First, a weighted average, pct(.) of individual growth rates of the relevant quantity is computed:

This is converted into an index that takes on the value 1 in the base year.

Secondly, the aggregate level of the relevant quantity is computed for the base year by summing up the countries’ individual base-year levels of the relevant quantity:

For example, aggregate real GDP in the base year is obtained by adding up individual base-year GDP’s expressed in a common currency.
Finally, the resulting index series for aggregated growth is rescaled by the value of the aggregated quantity in the base year.

The choice of weights

With either method, the choice of weights used depends on the particular nature of the variables to be aggregated, and represent a country’s share in a certain quantity or concept for the relevant region.
For ratios, the denominator of the variable to be aggregated is used to compute the country’s weight in a region. For example, for the current account balance expressed as a per cent share of GDP (CBGDPR), the weights are based on GDP in the reference years, while for the unemployment rate (UNR, the share of unemployed in the total labour force), the weights are based on the size of the labour force (LF) in each country in the region.

For GDP and expenditures, the weights are computed using the series in nominal i.e. current value terms expressed in a common currency. For example, private consumption (CP) and real private consumption (CPV) are aggregated using weights based on CP (that is a country’s share in the region’s consumption).
For variables related to numbers of persons, the weights are based on the shares of the variable to be aggregated. For example, total employment (ET) is aggregated using weights based the share of the country in the region’s total employment.

For a selection of variables, for example interest rates, where there is no absolute regional concept, regional averages are computed using GDP weights (that is a country’s share in the region’s nominal GDP). Where variables are related to constant or current price values, these variables are first expressed in a common currency (e.g. euro for the euro area) or in Purchasing Power Parities (PPP’s).

Coverage issues: actual data versus OECD estimates

In the event of incomplete information, for example due to delays in reporting of recent historical values, the reliability of coverage for estimates for a specific economic aggregate is usually judged via a threshold value. Thus reported history for a specific aggregate is judged to stop as soon as the available country coverage falls below 2/3 for the zone in terms of base-year GDP. Aggregated values thereafter should be treated as “OECD estimates” rather than actual data.

Last updated: December 2011

Sources & Methods of the OECD Economic Outlook

Related Documents