National accounts: revision of 1995-2021 time series

On 23 May 2024, the revised macroeconomic statistics for the base year 2021 were published. The revision resulted in adjustments to the interrelated figures for the balance of payments, the international investment position, government finance and the national accounts. Following this, the revision results for the 1995-2023 time series and the first quarter of 2024 were published on 24 June 2024. The adjusted figures include both annual and quarterly series, which comply with European delivery requirements under the ESA transmission programme¹⁾. This article looks at the time series for government finance and national accounts. Additional context for the balance of payments and international investment position time series is provided here.

CBS’s online database contains a comprehensive overview of the new time series for the three main national accounts systems: the Supply and Use Tables (SUT), Sector Accounts (SA) and Labour Accounts (LA). The revised government finance tables are also available here.

The revision was carried out in close cooperation with De Nederlandsche Bank (DNB). The revision for the base year 2021 and the adjoining time series covered both the national accounts (CBS) and the balance of payments (DNB). There is consistency between the national accounts and balance of payments data, and the publication dates were harmonised.

This article offers a brief explanation of how the time series for 1995-2021 were compiled. The period after the 2021 revision year was not included in the time series revision, but is part of the regular production cycle. The regular estimate for the first quarter of 2024, published on 24 June 2024, includes all previous series after the revision.

This article sets out the guiding principles for the time series project and the methodology used for each system. The government finance accounts have also been revised; as an integral part of the SA, these follow the same system and are therefore not described separately. In the final section, the adjustments for GDP and some other key indicators are presented. The revision also used a new method for seasonal adjustment of the new quarterly series, which is set out in Annex A.

The focus of the revision for the 2021 reporting year was on incorporating the latest insights from the various statistics and information sources used in the national accounts. Since 2015, the base year of the previous revision, new sources have become available, new statistical insights have been gained and new methods have been developed. The resulting changes have been incorporated into these new estimates for the 2021 reporting year.

For more information on the 2021 base year revision, see the previous articles published on 23 May 2024:

• News release 2021 revision
• Revision publication

The base year for this time series revision is the revised 2021 reporting year. The adjusted figures for 2021 were published on 23 May 2024. For users of national accounts statistics, rapid availability of consistent time series is of great importance. CBS is meeting this need by publishing the entire revised series on 24 June 2024, together with the regular figures for the first quarter of 2024. This early publication date is made possible in part by the use of mathematical algorithms and standardised methods to backcast time series in a largely automated way.

In estimating the time series, the following guiding principles were used:

• The time series covers the period from 1995 to 2021. Some time series are shorter, in line with the Netherlands’ European obligations.
• The SUT and LA time series are final and will, in principle, not be adjusted again until the next source revision. Parts of the SA’s non-financial and financial accounts and financial statements are reviewed annually.
• The length and granularity of the time series are in line with international obligations, and additional user requirements have been incorporated wherever possible.
• Algorithms were used as much as possible in backcasting the series, with gradual revisions over time. Only essential and substantial new information on the pre-revision years and quarters was included separately in the process (as so-called time series layers, see Section 3).

For parts of the Supply and Use Tables (SUT), Sector Accounts (SA) and Labour Accounts (LA), new methods and software were developed to construct the time series. These methods, which are in line with the previously established guiding principles, are explained below. There are many similarities between the methods used to produce the time series for each of the three systems. Section 3.1, on the methodology behind the SUT series, offers the most detailed explanation, as much of this also applies to the other systems. Sections 3.2 (Sector Accounts) and 3.3 (Labour Accounts) focus mainly on system-specific details.

3.1 Methodology for Supply and Use Tables (SUT)

A high level of aggregation was used in compiling the SUT annual series. This has resulted in a limited set of figures on industries, expenditure and product groups, in which the desired publication variables have been established. No revised Input-Output Tables (IOT) were created for the 1995-2020 time series. Moreover, product group-specific information is not available for all series. These tables are available for the 2021 revision year and beyond, however.

In the 2021 revision year, adjustments were made for almost all industries and expenditure categories. In areas where the adjustments were limited, the figures from the previous revision year, 2015, were assumed to be correct, and the adjustments were gradually backcast to this previous base year. Larger adjustments were backcast to the 2010 reporting year, the revision year before 2015. Thus, an explicit decision was made to make as few changes as possible to previous years, unless the previous estimate contained serious omissions that needed to be corrected. The same principles were applied to the SA and LA as well.

The larger the adjustment for the 2021 revision year, the greater the changes to the developments in the time series. For the SUT, a conscious choice was made to attribute the adjustment in value growth to volume growth, which means that the pre-revision prices determined the post-revision prices. This was done because pricing data is based on directly observed data sources, while volume figures (also in the regular estimates) are derived from value and price trends. Only for parts of the economy where volume developments were considered to be of higher quality than price developments (as in the case of some adjustments to the data on insurance companies) have price developments been adjusted.

Exactly how the adjustments for the 2021 revision year have been backcast in the time series varies by industry and final expenditure category. First, the adjustments for the 2021 revision year were looked at by industry or expenditure category. Subsequently, it was determined whether additional information was available on the size of the adjustments in the other years of the time series. Finally, the adjustments for the other years were determined, with the smaller revisions again tapering to zero in 2015 and the larger ones in 2010.

Updates to the national accounts time series are never made in isolation. For example, the adjustments to output and consumption of hospitality services must match. In addition, adjustments in one system, such as the SUT, need to be coordinated with adjustments to the other systems, such as the SA. In aligning these changes, the SA figures always follow the output and income generation data from the SUT (except for government estimates).

Ideally, adjustments in the revision year gradually become smaller further back in the time series, to ensure that the overall economic picture does not change too radically. The economic pattern before the revision is therefore very similar to the economic pattern after the revision. This is also included as one of the guiding principles in the methodology, and it is clearly visible in the graphical representations of the various economic aggregates before and after revision (see Section 4).

Below, the method for creating the SUT time series is set out in several steps.

3.1.1 Time series layers

The adjustment from the 2021 reporting year to the 2021 revision year can generally be divided into two parts: a part that can mostly be backcast using the old figures, and a part for which this is not the case, because the old figures are not representative of the phenomenon that has been adjusted. Adjustments of the latter type are also called time series layers.

The exact characteristics of a time series layer are determined in a revision project. For this revision, time series layers were created in response to a number of GNI action items (globalisation, spreads, health care costs), adjustments made for several large multinational companies, illegal activities and the transfer of social workshops and job pools to general government (see the revision publication for more information on the GNI action items and the other adjustments).

When using a time series layer, the impact of specific revision projects on the previous years is determined separately based on the available information. Some of the time series layers extend back to 1995, the beginning of the time series. This information was first incorporated into the time series before the rest of the time series was backcast using algorithms. Revision projects and time series layers were also used to create the time series for the SA and LA systems (see sections 3.2 and 3.3).

3.1.2 Backcasting adjustments

This section offers a detailed description of the backward adjustments to the production and expenditure GDP approaches.

Production approach to GDP:

1) The total output by branch of industry in current prices and the underlying commodity groups were backcast using the following formula:

A) If the recalibration for the 2021 revision year was less than or equal to 4 percent, the difference was backcast to the 2015 revision year.

B) If the recalibration was greater than 4 percent, the difference was backcast to the 2010 revision year.

The backcasting method used was relatively linear: the backcasting rate decreases each year by a fixed difference (linear) and is multiplied by the initial value (relative). The table below illustrates the backcasting process using sample figures. The pre-revision series is in the ‘initial’ column. For 2021, the revision adjustment (recalibration) is 100, which is less than 4 percent of the total. This means that the recalibration has been backcast to the previous revision in 2015. If the adjustment had been 300, the recalibration would have been backcast to the 2010 revision.

2) Output by product group was also backcast using the formula in step 1. Differences between total output and the sum of all product groups were adjusted in the product group with the highest output. These differences arose because the backcasting periods for the product groups and the total were not always the same.

3) The total value added was backcast using the same formula as in step 1. As a result, the backcasting period for total value added is always equal to that of total output.

4) The intermediate consumption time series was calculated as total output minus total value added.

5) Intermediate consumption by product group was also backcast using the formula in step 1. Differences between total intermediate consumption and the sum of all product groups were adjusted in the product group with the highest intermediate consumption.

6) The outcomes in the previous year’s prices (the constant prices) were calculated as follows: the values per product group in current prices were deflated based on the existing pre-revision price indices. Total output and total intermediate consumption, both in constant prices, were calculated by adding up the product groups in constant prices. Value added in constant prices was calculated as output minus intermediate consumption.

Expenditure approach to GDP:

7) The imports of goods at current prices time series was calculated in the same way as the output time series (see step 1), meaning that total imports of goods was backcast in a relatively linear fashion. The same applies to imports by product group. The difference between total imports and the sum of the different product groups was then adjusted in the product group with the highest imports.

8) The trade balance of goods was backcast in a relatively linear fashion.

9) The total exports time series was calculated as a residual of the trade balance and total imports.

10) Exports by product group were calculated using the same method used for output (see step 1). Differences between total exports and the product groups were adjusted in the product group with the highest exports.

11) Steps 7 to 10 were also carried out to produce the service imports and exports time series.

12) The time series for consumption by government, households, and non-profit institutions serving households (NPISHs) and the time series for investment were calculated in the same way as the one for output (see step 1).

13) The vast majority of the inventory adjustments relate only to the 2021 reporting year and therefore have no impact on the time series. Adjustments were made to the time series only for a very limited proportion. The remaining absolute differences between pre- and post-revision changes in inventories and valuables were backcast linearly. The absolute differences were used because these series can be either positive or negative; using the absolute mutations yielded results closer to the original pattern.

14) All values in the previous year’s prices (constant prices) for expenditure were calculated in the same way as in step 6. The totals of expenditure in constant prices are always the sum of the product groups.

3.1.3 Automated integration

After going through the above steps, the remaining integration differences between supply of and demand for goods were generally found to be limited. In most cases, the time series layers were consistent (with supply equal to demand), and due to the high level of aggregation of the product groups, lower-level supply and demand differences often cancelled each other out. As a result, the remaining integration differences could easily be resolved using automated procedures, preserving existing mutations as much as possible. A number of preconditions had to be met for this. For example, supply of and demand for goods had to be identical, and data from the Tax and Customs Administration on product and non-product taxes and subsidies were used exogenously, as well as LA data on employee remuneration.

In addition to the annual time series, revised SUT quarterly time series have also been published. These were created using an automated calculation procedure. The old quarters have been reconciled to the new yearly totals, while the old quarter-on-quarter developments have been kept intact as much as possible. Where necessary, additional information on adjustments in the time series before 2021 has also been added (through the use of time series layers), such as for the time series on employee remuneration and taxes. The quarterly series have been used as input for the seasonally adjusted GDP and underlying series figures (see Annex A), and also as input for the SA (see Section 3.2) including the Government Accounts.

3.2 Methodology for Sector Accounts (SA)

The methodology for estimating the SA time series is essentially the same as that described above for the SUT. One important difference, however, is that parts of the non-financial accounts and the financial accounts and balance sheets are revised annually, including a full reconciliation of the time series.

The inputs to the time series are (a) the pre-revision figures from 1995 to 2021, including those revised annually after the 2015 benchmark revision, (b) the outcomes of the various revision projects and (c) the final results of the revision of the 2021 benchmark reporting year. The outcomes of the revision projects were largely added to the pre-revision figures as time series layers, which sometimes go as far back as 1995, as with the SUT. The remaining difference with the 2021 post-revision figures was, with some exceptions, linearly backcast to the previous 2015 revision. This yielded other developments and levels. The exact methodology differs for each activity/transaction, and the resulting time series were integrated after backcasting.

3.2.1 Non-financial transactions

For the transactions that also appear in the SUT, the figures from the revised SUT time series were used. The information by branch of industry has been attributed to sectors (the grouping of (parts of) industries by institutional sectors) using dual classification. For the other transactions, such as interest and dividend flows, the recalibration was backcast linearly.

3.2.2 Financial accounts and balance sheets

A similar strategy was used for the financial accounts and balance sheets. The aim here was to achieve optimal alignment with the final pre-revision balance sheet for the 2015 reporting year (plus the layers added in step 1) by gradually reducing the difference in the balance sheet to zero in 2015 (or further back in time in some cases). The choice was made to channel the adjustments largely through financial transactions (and partly through price and exchange rate movements). After that, the integration software ensured an integrated and consistent system.

Because government finance is part of the SA, it also went through the above procedure for compiling the revised time series.

3.3 Methodology for Labour Accounts (LA)

The pre-revision LA time series also served as the basis for the revised time series for 1995-2021. To arrive at a revised time series, several layers were added to this series, as with the SUT and SA. Broadly speaking, the following procedure was followed:

• Time series revisions based on several revision projects were added to the pre-revision time series for wages and social security contributions as time series layers.
• Then, to fully align the time series for wages and social security contributions with the 2021 revision year estimates, the remaining differences were also gradually backcast using a recalibration.
• In addition, the new LA time series for wages and social security contributions were aligned with those for the public sector, the Sector Accounts (SA) and the Supply and Use Tables (SUT).
• In conjunction with the revision of the wages and social security contributions figures, labour volume data was also revised; this includes data on jobs, labour years, hours worked and employed persons.
• To compile quarterly series on wages, social security contributions and labour volumes, the pre-revision quarterly patterns were superimposed on the post-revision annual totals (after some changes).

3.3.1 Employee wages and social security contributions

As mentioned, the pre-revision time series served as the basis for the employee remuneration and social security charges borne by employers time series. First, time series layers from various revision projects were added to this. Examples of important revision projects include the reconciliation of severance and transition payments with the observation from policy records of the Employee Insurance Agency (UWV), the recalibration of wages in kind (wages in the form of goods or services, such as the private use of a company car), and the revision of pension contributions, whereby the totals of De Nederlandsche Bank (DNB) are now used for both employee and employer contributions. For each project, an assessment was made to determine how many years back corrections were needed, as well as the size of these corrections per year. For example, the revision project on severance and transition payments took into account recent changes in laws and regulations.

Then, to fully align the time series for wages and social security contributions with the new 2021 revision year estimates, the remaining differences were backcast using a recalibration. These remaining differences for the most part consisted of changes in the enterprise population by branch of industry, and of a smaller other residual difference. The changes in the enterprise population were partly due to different estimates for undeclared work/the illegal economy; the optimal backcasting method for these figures was determined by branch of industry on a year-by-year basis. The General Business Register’s (ABR) new classifications for business units also played a role in the changes in the enterprise population. These changes were the result of confrontations with other sources, such as De Nederlandsche Bank’s population of financial institutions. They were gradually backcast to 1995, with the compounded macro-level impact kept at zero for each year, as shifts between industries should not result in macro-level differences.

The final residual differences for wages and social security contributions necessary to achieve alignment with the 2021 post-revision estimate were also backcast to 1995. Here, value indices for wages and social security contributions were used for each branch of industry to index the calculated residual difference for 2021 back in time.

3.3.2 Labour volume

Labour volume includes jobs, labour years, hours worked and employed persons (employees and the self-employed). Analogous to the method for wages and social security contributions, the time series were calculated based on the pre-revision figures, to which layers were added. Apart from the new estimates for undeclared work/the illegal economy, the separate revision projects had no effect on labour volume. However, the same recalibration made for wages and social security contributions, due to population changes and the remaining residual difference, was also made for labour volumes. In doing so, the same changes per branch of industry and period were used as for wages and social security contributions, so as not to distort the relationships between labour volumes and remuneration for different groups.

For the self-employed, additional hours worked and labour years were revised due to the revision of the Labour Force Survey (EBB) in 2021. Changes in the EBB questions led to a slightly higher average number of hours worked per self-employed person. The fact that respondents were asked to fill in the EBB twice allowed the trend break in 2021 to be quantified and backcast to 1995 with fixed percentages for each subgroup.

3.3.3 Quarterly series

The pre-revision quarterly patterns served as a basis to compile quarterly series on wages, social security contributions and labour volumes. After some changes to the patterns for wages and social security contributions due to new insights, these quarterly patterns were superimposed on the post-revision annual totals by branch of industry.

3.3.4 Aligning time series with other systems

Total social security contributions for the Netherlands and wages and labour volumes for the public sector were aligned with government statistics. Wages, social security contributions and labour volumes were also aligned with the SA. In the LA, pension contributions, which are part of wages and social security contributions, were specifically adjusted in line with the SA to fully match DNB data. Finally, as usual, the data for employee remuneration from the LA was used for the SUT, after a confrontation and some adjustments to the figures.

This section presents pre- and post-revision figures for GDP and some other key indicators for the period 1995-2021. The largest adjustments are due to realignment with current source levels. As described earlier, the differences were backcast gradually, to the previous 2015 revision or, for larger differences, to the 2010 revision. The time series layers can lead to a more variable pattern, but these layers are usually small and barely visible in the development of the key indicators.

Upward revisions to GDP and most expenditure components have resulted in post-revision levels that are higher than pre-revision levels, but for the most part the pattern has remained virtually unchanged over time. The economic dip in 2009 is evident, as are the sharp downturn as a result of the pandemic in 2020 and the rapid recovery in 2021.

The way in which the quarterly time series are adjusted to account for normal seasonal influences has been modified in a number of ways in this revision. This annex first provides a general explanation of the seasonal adjustment method before zooming in on the differences between the seasonal adjustment method before and after the revision. The main difference is the method used to deal with discrepancies between the sum of various seasonally adjusted series and the corresponding seasonally adjusted aggregates. This is also known as the aggregation difference.

Introduction

Seasonally adjusted time series play an important role in the interpretation of quarter-on-quarter economic developments. Seasonal adjustment is the process of removing all usual seasonal influences and calendar effects from a time series. Usual seasonal influences are all developments of a similar magnitude that occur every year during the same period, and that are expected to continue. Calendar effects are the effects of year-on-year changes in the calendar. These are mainly the number of working days per period and the effect of leap days.

Seasonally adjusted quarterly series make it possible to identify real short-term economic developments, as many economic time series show similar trends between certain quarters each year. The holiday season, for example, always leads to a spike in toy sales in the fourth quarter. When analysing toy sales in the fourth quarter of any year, it is therefore not useful to look at the unadjusted sales trend compared to the third quarter, because this will always show an increase. Adjusting for the usual seasonal influences and calendar effects gives a better picture of the state and development of the economy in any given quarter.

Seasonal adjustment method

In order to achieve an optimal seasonal adjustment of a time series, first all calendar effects are removed from the data and outliers are temporarily taken out of the series. An outlier is an observation in the dataset that does not fit statistically with the other data points, such as the purchase of a new fleet of aircraft by an airline; the value for the quarter in which this purchase takes place will be much higher than the value for a typical quarter. The quarter in question is therefore identified as an outlier, so that its value is not included in the determination of the seasonal component of the series. After preprocessing the series to correct for calendar effects and outliers, the seasonal component of the series can be determined and removed using the seasonal adjustment software X-13ARIMA.

Within the national accounts, series are published at both detail and aggregate levels. When both individual (detail) series and their aggregates are adjusted for seasonal influences separately using theaforementioned process – also called direct seasonal adjustment – aggregation differences arise: the seasonally adjusted detail series do not add up to the separately adjusted aggregate.

The reason such aggregation differences arise is that each series has its own set of outliers, ARIMA model and seasonal filters. In most cases, this is enough for differences between the detail series andthe aggregate to arise. In addition, the seasonal component of most economic series is broken down non-linearly. This means that the size of the component depends on the size of the series. As a result, the seasonal components of the subsets do not add up to the seasonal component of the aggregate.

The emergence of aggregation differences in direct seasonal adjustment is thus unavoidable, which is why international guidelines do not require seasonally adjusted aggregates to match seasonally adjusted subcomponents. However, such inconsistencies do pose challenges to users of CBS’s macroeconomic statistics. For example, a separate calculation of GDP growth through the underlying seasonally adjusted expenditure components may not match the published GDP growth figure, making it difficult to identify which spending components have driven seasonally adjusted growth. Within CBS, the choice was made to help users by minimising aggregation differences.

Old method

Before the 2021 revision, aggregation differences were minimised by adding consistency between series at several points in the seasonal adjustment process:

• GDP model settings used for other series

  The settings used for the seasonal adjustment of GDP were also used for most of the detail series. Different settings were only used in exceptional cases, when the GDP model did not fit a detail series.

• Consistent inclusion of outliers

  If an outlier was identified within a series, it was also included in the higher aggregates, with exactly the same magnitude (in absolute values). The reverse was true as well: outliers in aggregates were distributed among the underlying series. However, determining the correct distribution of these outliers and their exact value is a complex process, as the seasonal adjustment software may produce different outlier estimates for each series.

• Consistent estimation of working day effects

  A structural time series model was used to estimate a consistent working day effect across all series. ²⁾

A major challenge with this approach is that the process of determining seasonal components as effectively as possible is strongly intertwined with the process of minimising aggregation differences. It is therefore difficult to distinguish the effects of these two processes on the seasonally adjusted results. To get a better understanding of this and increase transparency, a new method is being used from this benchmark revision onwards.

New method

The premise of the new method is that seasonal adjustment and the minimisation of aggregation differences are two separate processes.

Seasonal adjustment

In seasonal adjustment, the seasonal pattern of each series is optimally removed. This means that each series (both detail series and aggregate series) gets its own settings, and that outliers and working day effects are identified separately for each series, as effectively as possible. Outliers are included here only if there is a valid economic explanation for them, and if they are statistically significant in the series. If an outlier is of low statistical significance, a valid economic explanation may still lead to its inclusion.

Sub-periodisation of the working day effect

The effect of an extra working day in a quarter has changed considerably over the years: it used to have a fairly significant impact on many economic indicators, but in today’s economy a large number of sectors operate 24/7. To account for this, the data was divided into periods of roughly nine years each, after which an assessment was made for each period as to whether there was a significant working day effect. This effect was then estimated and removed from the series. Many series still have a significant working day effect at the beginning, but almost no series still has a significant working day effect at the end.

Splitting up of the “Value added by culture, recreation and other services” series

For the value added by this branch of industry, a completely different seasonal pattern has been observed since 2007. The series has therefore been split into two parts: one from 1995 to 2007, and one from 2007 to the present. Both series have their own settings to allow for the optimal removal of the seasonal pattern, after which the two parts are merged to form one series.

Minimising aggregation differences

After individual seasonal adjustment has been applied to all series, the aggregation differences are calculated. The aggregation differences are then minimised using an optimisation model that is also used for other processes within national accounts. This is called integration. The aim of the model is to minimise the development of aggregation differences. This is important because it reduces the effect of aggregation differences on GDP trends, making it easier to determine the contribution of underlying component indicators to seasonally adjusted GDP growth.

Thus, two constraints are imposed on the optimisation method used to compile seasonally adjusted series:

1. The underlying subseries plus the aggregation difference must add up to the aggregate in each quarter.
2. Aggregation gaps between two consecutive quarters must be minimised.

Due to the second constraint, the aggregation differences for each quarter are adjusted. Because of the first constraint, this affects the seasonally adjusted subsets and aggregates. The degree of adjustment to the series is then determined by applying two criteria:

1. Proportionality: the larger the value of the series, the more the series may be adjusted.
2. Quality: the worse the quality of a series’ seasonal adjustment, the more the series may be adjusted. This quality is assessed by the seasonal adjustment software.

The extent to which aggregation differences between two consecutive quarters must be reduced is determined in consultation with relevant experts. An assessment is also made to determine the extent to which both criteria (proportionality and quality) for distributing aggregation differences among the series must be taken into account.