On the basis of the data series transmitted by the EU Member States, Eurostat calculates aggregated short-term business indices for the EU and the euro area. This article describes the procedure used by Eurostat for the compilation of European aggregates. It is part of a set of background articles treating various methodological aspects of short-term business statistics.
Basic concepts: Indices, activity aggregation and geographical aggregation
Eurostat presents information on the short-term development of production, turnover, prices etc. (i.e. the various short-term indicators) in the form of indices and in the form of growth rates. Information in the form of absolute values (e.g. in euros) are generally not published in short-term statistics. In this context, it should be noted that short-term business statistics (STS) are infra-annual statistics. Data are not provided only for years but for months (or at least for quarters). So for many indicators in the year of reference (e.g. Y) there will be 12 different monthly index values. Only the average of these values will be 100. Conceptually, the reference year is to be distinguished from the base year which is the year from which the weights for the index construction are chosen (see below).
STS provides indices for fairly specific economic activities (e.g. the manufacturing of parts for motor vehicles) and also for more general activities (e.g. the production of motor vehicles) or even for total industrial production. In the activity aggregation, indices for more specific activities are combined to higher level indices by multiplying the index values of the specific activities with their weight in the higher level activity and summing up the weighted indices. Table 1 provides a simplified example of this procedure. The index value for activity C10.1 is calculated as the sum of the weighted indices for the processing of meat, of poultry meat etc.
For the national level the activity aggregation is generally carried out by the National Statistical Institutes which send complete data sets including indices for specific and more general activities to Eurostat (see below). Apart from time and activity, STS indices have a geographical dimension. At the European level the indices from Member States are aggregated to an index for the whole EU and to an index for the euro area (geographical aggregation). In this aggregation the differences in the magnitude of the various economic activities in the European countries are taken into account. The production of aggregated indices for the EU and for the euro area is the main added value provided by Eurostat's short-term statistics.
Data reception, control and pre-treatment
Each month, Eurostat receives several thousand STS data series from Member States. Before these series are loaded into the public database, Eurostat runs a couple of tests, e.g. for unusually high revisions of individual data entries or for unusual growth rates. Checks are also carried out to exclude negative index values. Where unusual data are detected, Eurostat contacts the National Statistical Institute to verify, and if necessary, correct the data.
Some data series also need a pre-treatment before they can be used for the calculation of European aggregates. In some statistical areas Member States send absolute data to Eurostat (e.g. for building permits). These have to be converted to indices. Occasionally Member States transmit data series with different base years. In particular, this can happen around the time a new base year is introduced. While some Member States will have already switched their data series to the new base/reference (e.g. Y=100), others might still be using the old base/reference year. The data then have to be re-referenced to the same year.
Finally, missing series sometimes have to be produced or estimated.
Production of missing seasonally adjusted data
Two cases need to be distinguished, the production of seasonally adjusted series and the estimation of missing aggregates. Eurostat currently applies an indirect method of seasonal adjustment. According to this method, seasonally adjusted European aggregates are calculated as the weighted sum of seasonally adjusted country data. Where a Member State does not transmit a seasonally adjusted series for a certain index this seasonally adjusted series is produced by Eurostat on the basis of the corresponding calendar adjusted series. Member States are legally obliged to transmit the calendar adjusted and unadjusted data for all indicators except prices - and seasonally adjusted data for the Principal European Economic Indicators of volume and value indicators.
Estimation of missing aggregates and activity aggregation
Sometimes not only seasonally adjusted data are missing but also some unadjusted or calendar day adjusted data. In such cases Eurostat has to produce estimates. Generally, such estimates cannot be carried out if an index value is missing at the lowest level of aggregation. If for example, an industrial production index for country A and month t is available for NACE Rev. 2 classes C10.11 and C10.12 (meat and poultry meat) but not for C10.13 (poultry meat products) Eurostat has no information on which it could base an estimate for the missing index.
However, it is often possible to produce an estimate for the next higher level of aggregation (e.g. for NACE Rev. 2 group C10.1, meat and meat products). Such an estimate is produced by Eurostat if the sum of the weights of the available lower level indices amounts to at least 80% of the weight of the higher aggregate. The index for the higher aggregation level is calculated by distributing the weight of the missing index proportionally over the available indices (see Table 2 for a simplified example). The relatively high threshold of 80% guarantees the quality of the estimations.
When Eurostat publishes estimated data these are flagged in the dissemination database with an "i". Seasonally adjusted data are flagged as Eurostat's estimates if Eurostat is seasonally adjusting the calendar adjusted national data or if Eurostat is aggregating the national data to estimate seasonally adjusted European aggregates (indirect seasonal adjustment).
Weights and branch lists
The above estimation example is valid for all levels of activity: If the available country data at one level of aggregation represent more than 80% of the weight of the next higher level of aggregation, the missing data are estimated with the help of so-called “branch lists”. A branch list is a complete and detailed list of weights for each country and each indicator of the NACE activities for which data have to be transmitted by the country according to the EBS Regulation along with the respective weights. Depending on the indicator different weights are used. Table 3 provides an overview of the STS indicators and their respective weights. The reporting countries use data for the STS weights in most cases from structural business statistics, but apply adjustments to take into account the different statistical units (enterprise in structural business statistics and kind-of-activity-unit in STS). However, other statistics such as international trade statistics are also used.
The weights used for the production of aggregated European STS indicators are taken from a single base year (as of March 2024 the year 2021 is used as base year). Since the weights remain unchanged over a number of years, aggregated European STS indices are fixed base indices. The weights are generally updated every five years.
For each series Eurostat uses specific weights representing the size of the respective economic activity in a national economy in relation to the whole EU or euro area. At a very detailed level this could lead to the identification of individual companies since in some special activities there might only be few companies in a country. Therefore the detailed weights used by Eurostat in STS are confidential. For higher level aggregates the weights can be found here.
Geographical aggregation
Eurostat only publishes European aggregates if the total available country data represent at least 60% of the total European weights for the EU or the euro area. The weights used for the geographical aggregation are the same as for the activity aggregation (see above Table 1). The threshold of 60% is applied to each individual series. Thus it might be the case that a production index for manufacturing of paper and paper products (NACE Rev. 2, division 17) can be published but not a production index for wallpaper (Nace Rev. 2, class 17.24). Data availability of European aggregates is better for the aggregated levels of the activity or product classifications, because more Member States are obliged to report higher level aggregates. For similar reasons it can happen that an indicator can be published for the EU but not for the euro area.
Updating of STS weights and indices
The economic structure of European economies changes over time. Some national economies grow faster than others and some activities may expand or shrink at a different tempo in the different countries. Therefore the fixed weights, used to calculate European indices, tend to reflect the economic reality less well if these weights are old. For this reason the EBS regulation foresees that the weights used for the aggregation of STS indicators (by Eurostat and also by the National Statistical Institutes) are generally updated every five years.
Theoretically, base years should be years that are "normal" or at least not greatly influenced by special events (like a financial crisis). Mainly for reasons of comparability and simplicity however, the business statistics regulation prescribes that the selected base years are generally years ending with 0 or 5.
The update of the weights should take place not later than three years after the end of the new base year.
The process of updating weights and indices is described with a simple example in Table 4 and Figure 1 using two price indices (Index 1 and 2) for a community of only two countries. It is assumed that in 'Y - 5' (the original base year) the weights for the selected economic activity were equal in the two economies. The reference year was also 'Y - 5' so in 'Y - 5' the indices for the two economies were 100 and the aggregated index ('Y - 5') was also equal to 100 (0.5 * 100 + 0.5 * 100).
Over time the price index increased strongly in economy 2 but went up and down again in economy 1. The aggregated index (AG N - 5) shows a relatively strong price increase as a result of the increase in economy 2 which is taken into account with a weight of 50%. In 'Y + 3' new weights data become available for the year 'Y' which show that the relative importance of the activities in economy 1 and 2 have greatly changed (0.95/0.05 compared with 0.5/0.5). The new aggregated index calculated in 'Y + 3' using the weights from 'Y' (AG Y) is now much lower since the price increase in economy 2 is taken into account with a weight of only 5%.
Obviously, statistics cannot simply switch from one index (AG N - 5) to a re-weighted index (AG N) since this would result in a break of the time series from 140 to 100 between 'Y - 1' and 'Y'. To combine the series with the different base years two methods are possible.
The first is to entirely recalculate the old index values with the new weights from 'Y' (AG N). This method also changes the rates of annual changes for older years (e.g. the rate of change between 'Y - 3' and 'Y - 2' becomes -13% instead of 0%). This method is used by Eurostat for the calculation of European aggregates. It is worth noting that the weights rarely change as much as in this example.
Under the second method, currently not used by Eurostat, the index levels are calculated backwards from 'Y' using the old rates of changes (AG linking). Arithmetically, this is equal to applying the ratio between the aggregated indices with the weights from 'Y - 5' and 'Y' for the year of change 'Y' (i.e. 100/150=0.67) to the old indices. As a result the rates of change which were observed between 'Y - 5' and 'Y' are maintained in the new re-weighted index, only the index level was shifted downwards (see Figure 1). In this method, it is important that both index series are scaled to 100 for the entire reference year.
The second method requires that the weights of price indices are volumes (quantities), not values, where the prices affect the weighting variables.
Statistical Institutes use different procedures to change base years. Generally, however, the second method is recommended for vertical aggregation of national time series.
It can be noted that the re-basing implicitly results in a re-referencing, i.e. the value for the new base year equals 100. (The opposite is not the case; a time series can be re-referenced without changes in the weights.)
