Environmental Goods and Services Sector: Eurostat EU estimates

Transcription

1 EUROPEAN COMMISSION EUROSTAT Directorate E: Sectoral and regional statistics Unit E-2: Environmental statistics and accounts; sustainable development ENV/EXP/WG/04.4(2016) Point 4.4 of the agenda (12 April) Environmental Goods and Services Sector: Eurostat EU estimates Eurostat Unit E2 Working Group Environmental Expenditure Statistics Meeting of 12 and 13 April 2016 BECH Building Room Quetelet

2 1. Introduction In 2013 and 2014 Eurostat developed practical methods for compiling EGSS statistics from existing sources publicly available on Eurostat s website (environmental protection expenditure statistics, national accounts, energy and agricultural statistical, SBS and PRODCOM data, Labour Force Survey data, trade statistics and other sources) and tested them at country level. In the February 2014 meeting of the Working Group Environmental Expenditure Statistics first estimates of the EGSS variables for EU-28 were presented covering the period (Doc. ENV/EXP/WG/3.2(2014)). Improved EU-28 estimates for the period were published beginning of 2015: the main improvements were explained in section 2 of the Doc. ENV/EXP/WG/3.2 (2015) presented to the Working Group in March This document presents new estimates for the year 2013 and revised estimates for (section 2). They were published in Eurobase beginning of Section 3 discusses plans for further work: comparison of Eurostat s estimates with data countries transmit via the questionnaires, revised method for compiling EU-28 estimates, early estimates. Working Group delegates are asked to: Comment on the revisions and update of the EU-28 estimates for EGSS (see section 2), Indicate if they are interested to get (restricted) access to the detailed test calculations for your country underlying Eurostat s EU-28 estimates (access via CIRCABC), Comment on the plan for a revised method of EU-28 estimates, Comment on the proposal for an early estimate at EU-28 level and to share their experience with early estimates at country level (if any). In particular, they are asked whether they propose other methods for the early estimates than those proposed in section 3.3? 2. Revised and updated EU-28 EGSS estimates In early 2016 Eurostat published new EU-28 EGSS accounts estimates for At the same time estimates for the period were revised taking into account the improved availability of ESA 2010 national accounts data. Revisions in EGSS were in part a consequence of more complete ESA 2010 national accounts becoming available during 2015 and revisions of some other data sources. Newly available data on the organic farming labour force were also included as well as any other revision in the source data underlying the estimates. The EU-28 estimates are available in Eurostat s on-line database, data collections group environmental goods and services sector (env_egs). The method to compile EU28 EGSS data for the period used some 40 different data collections from national accounts, 1

3 SBS, PRODCOM, energy statistics, trade statistics and environmental expenditure statistics and other data, most of them publicly available in Eurobase. Eurostat produces test calculations for individual countries, which are not published. They are used to calculate the EU-28 aggregates (only the EU-28 results are published). Any country data published in the Eurostat database are those transmitted by the countries via the EGSS questionnaire. The table below shows the new employment estimates for the period The estimates compiled last year are shown for comparison. December 2015 estimates: Table 1: Revision of EGSS-EU28 employment estimates (employment in million FTE) Market Non-market Ancillary Total December 2014 estimates: Market Non-market Ancillary Total As a consequence of revised source data EGSS employment estimates for EU-28 were revised downwards on average by ca to -0.3 million FTE each year. The revisions for Poland, Romania and United Kingdom explain much of the revision for the EU total: Poland ( FTE for 2012), Romania ( FTE for 2012) and the United Kingdom ( FTE for 2012): In Romania the downward revision was mainly due to revised employment figures in the national accounts for NACE E37-39 (from to employed person) and new census figures for agricultural employment which suggested that the share of organic farming employment in agricultural employment is much lower than in the December 2014 estimates (where the shares for employment were based on land use shares). For Poland the main revision impact comes also from new figures on organic farming employment. For the United Kingdom the main revision impact comes from the estimate for the production of EGSS goods used for wastewater and waste management. The December 2014 estimates still used ESA95 data for NACE E, which were replaced in the December 2015 estimate by ESA 2010 data, which showed a strong downward revisions for UK investments in NACE E ( -35%). 2

4 The December 2015 estimates show for the years that EU-28 employment related to the production of environmental goods and services has stabilised slightly above four million FTE. Whereas in the years before 2011 energy production from renewable sources acted as a main driving force for EGSS employment growth, the expansion of renewable energies slowed down after Data from Eurostat s energy statistics show that the absolute annual additions to the renewable energy production capacities have fallen for photovoltaic in Germany by about 50% (2013 compared to 2012, i.e. growth in absolute terms between 2012 and 2013 was 50% lower than between 2011 and 2012). This was not outweighed by the employment growth in wind power capacities. Energy statistics also indicate decreasing annual additions to the capacities producing energy from renewable sources in France, Italy and Spain. Furthermore, the German data on investment in renewable energies indicate that investments in hydro, wind and photovoltaic power dropped from a peak in 2010 to 2013 by more than -40%. PRODCOM data show that the value of production of Photosensitive semiconductor devices, solar cells, photo-diodes, photo-transistors, etc. fell in Germany from 4.6 billion EUR in 2010 to 1.1 billion EUR in On the other hand the production of Generating sets, wind-powered increased in Germany from 1.6 billion EUR in 2010 to 2.6 billion EUR in 2013 and in Denmark from 3.0 billion EUR in 2010 to 6.0 billion EUR The Table 2 shows the shares of EGSS gross value added to GDP in the Member States in These are the unpublished Eurostat country estimates that are needed to calculate the (published) EU-28 estimates. According to those figures, EGSS gross value added ranges from 0.9% of GDP in Ireland to 3.0% in Austria. The estimates published in 2014 are reported alongside the revisions published in

5 Table 2: EGSS gross value added by 27 EU Member States, year 2012 December 2015 estimates December 2014 estimates million EUR % of GDP million EUR % of GDP Belgium Bulgaria Czech Republic Denmark Germany Estonia Ireland Greece Spain France Italy Cyprus Latvia Lithuania Luxembourg Hungary Malta Netherlands Austria Poland Portugal Romania Slovenia Slovakia Finland Sweden United Kingdom It is also possible to analyse the size of revisions of Eurostat's estimates for countries between December 2014 and December The EGSS gross value added shares in GDP in most countries with biggest weight in the EU-28 economy were not significantly revised (Germany, France, United Kingdom, Italy). The biggest upwards revisions (+0.3 percentage points and more) were found for Bulgaria, Denmark, Estonia, Spain, Cyprus, the Netherlands and Finland and the biggest downward revisions (-0.3% percentage points or less) for Luxembourg, Malta, Romania, and Sweden. At first sight it is surprising that the overall level of GVA was revised upwards, whereas employment was revised downwards. As the number of sources used for the estimates is high 4

6 and many source data have been revised, it is non-trivial to disentangle the various revisions effects on GVA: It is to be noted that for the December 2014 estimates there were still many data gaps in the ESA 2010 national accounts data and even ESA 95 data were not yet complete at the necessary detailed level in some Member States (e.g. national accounts data at current prices for NACE E37-39 were only available up to year 2009 for Spain). Many data gaps were filled in the December 2014 estimates with simple estimates, which in the December 2015 estimates were replaced by data published on Eurobase. Moreover, not all revisions of source data impact on output, GVA and employment in equal measure. For example, the use of the new census data for agricultural employment had no direct revision impact on the estimates of output and only partially on the derived gross valued added for organic farming (as land use shares were used for the split by organic and conventional farming of output, consumption of fixed capital and net operating surplus of agriculture). The use of more detailed data on organic farm output (by crop and animal products) may be a possible improvement for the future, however, there is no Eurobase collection on the absolute price of organic farming products. 3. Further work 3.1. Comparing Eurostat estimates with data transmitted by countries The Eurostat's unpublished country estimates can be compared with the data transmitted by the countries via their EGSS questionnaires. Table 3 shows this comparison at an aggregate level for EU countries that transmit data in the 2015 data collection round 1. 1 Croatia also sent data in the 2015 data collection round. However, Croatia it is not included in this comparison as at the time of drafting this document Eurostat estimates for Croatia are being elaborated. 5

7 Table 3: Comparison between EGSS data transmitted by countries via the EGSS questionnaire and Eurostat s estimates, year 2013 Output GVA Employment (1000 FTE) BE BG CZ DK Quest. Eurostat Quest. Eurostat Quest. Eurostat Quest. Eurostat Output GVA Employment (1000 FTE) Output GVA Employment (1000 FTE) IE FR LV LT Quest. Eurostat Quest. Eurostat Quest. Eurostat Quest. Eurostat NL PL AT RO Quest. Eurostat Quest. Eurostat Quest. Eurostat Quest. Eurostat : Output GVA Employment (1000 FTE) Quest. SE Eurostat Quest. SI Eurostat EUR = BGN = CZK = DKK= PLN= RON = SEK= LTL Broadly, data are similar, but differences can also be surprisingly high. Detailed analyses may help to find out the reasons and improve Eurostat s estimates and/or (possibly) the methods applied by the countries. Such analysis was made in some detail for Germany in the preparation for and during an information visit to DESTATIS in spring 2014 (main results of this comparison have been presented to the 2015 meeting of the Working Group, Doc. ENV/EXP/WG/3.2 (2015)). Possible reasons found were various scope differences related to environmental activities and products and different estimates of factor productivities (e.g. for labour). Due to resource constraints it was not possible to further pursue a more detailed and systematic comparison in However, Eurostat makes available its calculations at country level to the NSIs on request. Eurostat s estimates are based on a standard approach used for all countries (as described in the EGSS practical guide, see Doc. ENV/EXP/WG/3.2 (2015)) 6

8 ensuring a high comparability of the results across countries and years. It may be interesting for the Working Group that countries compare their own approaches and results with Eurostat ones Revised method for EU-28 estimates Eurostat will continue releasing new EU-28 estimates every year and revising the estimates for the previous years. This is rather time-consuming. In 2015 it took about 3 months to extract new data, fill data gaps, and compile the EGSS data for EU-28. This process was particularly long in 2015 partly because there were many gaps in the new ESA 2010 national accounts data which replaced ESA 95. Correspondingly many technical changes in the estimation files were necessary. Unfortunately, the full changeover to ESA 2010 could not be completed in 2015 and this is still an ongoing task for A lengthy estimation process undermines the timeliness of the estimates. A speedier process would allow publishing earlier estimates and base them on the most recent underlying sources. To make this happen it had been proposed to reorganise the process by always using the previous years EGSS estimates available as benchmarks and to revise the calculations for only a fraction of the countries every year (see Doc. ENV/EXP/WG/3.2 (2015)) 2. However, this plan was not implemented in 2015 as it was deemed preferable to finalise the changeover to ESA 2010 data first. Hopefully, after the 2016 exercise this changeover can be finalised (with ESA 2010 data in national accounts more complete), so that the approach may be pursued in As the coverage and comparability of country data reported to Eurostat via the EGSS questionnaire improves, Eurostat s country estimates will be gradually replaced by the questionnaire data in the calculation of the EU-28 aggregates. Since countries are at different stages of developing EGSS accounts it may be necessary to integrate for a certain number of years the results from Eurostat s approach with the results from the questionnaires. An easy and flexible approach needs to be developed for this. Eurostat intends to work on a case by case basis where one country after the other is analysed in discussion with the country. Once country data are good they will be included in the EU estimate Early estimates Policymakers and other users of environmental accounts request early estimates to increase the relevance of the data and its analysis. The early estimates of economy-wide material flow accounts (at T + 9 months) are considered a big success. EU-28 EGSS currently become available at the end of second year after the reference year T (i.e. at T + 24 months). Eurostat aims to develop earlier estimates for EGSS. 2 The plan foresees the following: 5 countries revised and updated every year: the 5 countries in this group would be selected considering their economic weight in the EU-28, the quality of the basic sources available (in particular completeness and timeliness) and representativeness (geographical and socio-economic development). The share of these 5 countries in the EU-28 GDP would be approximately 60%. Other countries based on a benchmark approach: for the remaining countries the full calculations would not be repeated according to an annual schedule. The estimates for the benchmark years would be extrapolated using a simpler method based on indicators on their overall economic development.. 7

9 The Eurostat EU-28 EGSS estimates are based on some 40 different data collections from national accounts, SBS, PRODCOM, energy statistics, trade statistics and environmental expenditure statistics and other data, most of them publicly available in Eurobase. Some of those data are available earlier than T+24 months. For example, most energy statistics would be available for a T+12 month estimate. For some countries also national accounts data are available earlier with fewer detail. However, by far not all data used are available before T+24 months. An early EGSS estimate based on the present Eurostat methods would use very incomplete data sources and would require more gap filling. Moreover, producing EU estimates based on country level estimates which won't be published are too costly for the benefits achieved. Therefore this section explores an alternative approach for early EU-28 EGSS estimates which would not require EGSS country level estimates by Eurostat. The approach follows a rate-of-change-model applied to EGSS EU-28 data and the application of specific assumptions about the development of partial productivities. The model for early estimates proposes combining default estimation approaches with indicator based approaches and expert guesses. Annex 1 proposes a more detailed concept for this approach. 4. Conclusions To ensure a high level of comparability across countries the EU-28 estimates use the same set of sources and apply the same set of methods top all countries. The country specific calculations can be made available to countries, and Eurostat can offer support to countries in methodological questions. The EU-28 estimates provide long time series that are comparable over time. However, the process to compile EU-28 estimates should be reorganised to speed it up using a benchmark approach. This is expected to produce timelier and less burdensome estimates. When country data transmissions become more complete respecting the future data delivery obligations established by Regulation 691/2011, the underlying Eurostat estimates at country level will be gradually replaced by the data reported by countries in their EGSS questionnaires (for estimating EU-28 EGSS). In this case the Eurostat estimates may still be useful for creating long time series and for validation purposes. Even once all countries transmit comparable EGSS data, the approach developed for the EU- 28 estimates may be useful to create longer timer series than those reported and used to produce early estimates. To produce early estimates (e.g. at T+12-months or T+18 months) at aggregate EU-28 level, a simplified approach based on rates-of-change of volume and price indicators and specific assumptions on partial productivities may be conceived. 8

10 Annex 1: Concept for the early EGSS estimates at EU-28 level The approach for early estimates follows a rate-of-change-model applied to EGSS EU-28 data and the application of specific assumptions about the development of partial productivities. The model for early estimates proposes combining default estimation approaches with indicator based approaches and expert guesses. Level of detail A simplified early estimate approach can in principle cover all characteristics of the EGSS accounts. Here the focus is on: Output Gross value added Employment broken down by the categories of the data map used for the EU-28 data compilation (see EGSS practical guide ), namely: Ancillary production of EGSS Non-market production other than wastewater, waste and water management services Market and non-market production of wastewater, waste and water management services Market production of EGSS other than wastewater, waste and water management services: non-capital goods and services Market production of EGSS: capital goods and services and further broken down by CEPA/CreMA and NACE categories. Some further categories: Non-market production of wastewater, waste and water management services Market production of wastewater, waste and water management services Non-market production Market production Total EGSS activities can be calculated by fixed ratios and simple aggregation, respectively. Output Output estimates can be based on rates-of-change for output volumes and output prices. Accordingly, the value change is then given by: Value change (%) = [(1 + Volume change (%)/100) * (1 + Price change (%)/100) -1] * 100 Using the data of the latest available year (e.g. 2013) an early estimate (e.g. for 2014) of the output value is calculated as follows: Output 2014 = Output 2013 * [Value change (%)/ ] 9

11 Methods must then be defined for the volume and price changes. Which method can be applied to a certain variable depends very much on when the early estimate is to be made. It is clear that in general the more specific indicators are available the longer the time-lag is: for an early estimate of T+6-months normally less specific indicators are available than for an early estimate of T+12-months, and for an estimate T+12-months less indicators are available than for an estimate T+18-months. For the price and volume changes a hierarchical overlay of methods should therefore be defined that ensures that at any time between T+1 and T+23 months an early estimate can be carried out. Of course, it has to be understood that an estimate at T+23 months can be based on a much richer quality and level of detail of indicators than an estimate at T+1 months, which may to a great extent have to rely on no-change assumptions or other simplistic approaches. The estimation must be designed in a way that allows specifying the methods of setting rates of change and partial productivities at upper aggregation levels (presented in the section 'level of detail' above) and taking them over to lower levels of breakdowns. If, however, specific methods are defined at lower aggregation levels the methods specification at higher aggregation levels should not work through to the lower levels. For example, if it is defined that the method for the category Market production of EGSS other than wastewater, waste and water management services, CEPA 13A uses physical data on renewable energy production as an indicator and no specific indicator is defined at the NACE breakdown for this category, the same indicator shall apply to all NACEs for this category. This can save time in elaborating the early estimates, but it is flexible enough to accommodate specific information (if available and with significant impact). In the following some simple methods are shown that can be used to estimate the characteristics. Methods for volume rates of changes may, for example, look as follows: 1 assumption of no-change from the previous year: 0% 2 same rates of change as the GDP volume change (from Eurobase) 3 same output volume rates of change as at the different NACE categories (from Eurobase) 4 rates of change following the rates of change of specific physical/volume indicator I (from Eurobase) 5 rates of change following the rates of change of specific physical/volume indicator II (from Eurobase) 6 rates of change following the rates of change of specific physical/volume indicator III (from Eurobase) 7 rates of change set by an expert Method 1 (no-change assumption) ensures that in any case a rate of change (=0%) is set irrespectively of whether data for the other methods are available. However, if data on GDP volume growth is available method 1 will be overruled by method 2, i.e. by the assumption that the volume changes in the EGSS follow those of GDP. This is 10

12 not a very specific assumption, but it acknowledges that changes in the economy wide production may have an impact on the overall level of production in the EGSS. Method 3 uses output volume changes for different NACE categories. If data on the volume changes are available this method overrules the previous two methods. The method basically assumes that the production in the NACEs is representative for the EGSS production in these NACEs. This is again not a very specific assumption, but it acknowledges that for some EGSS activities the developments in the NACE can be quite representative for the EGSS (e.g. NACE 37 and 38 for wastewater and waste management). Methods 4 to 6 use specific physical/volume indicators. For example, the variable primary production of wind power at EU-28 level from Eurobase energy statistics may be considered as a sufficiently good indicator I for volume changes in the market production of EGSS capital goods and services for CReMA 13A (production of renewable energies). But a better indicator may be the variable primary production of renewable energies at EU-28 level (indicator II) as it covers not only a part of the renewable energy but the totality of renewable energy. The rate of change from indicator II may, however, be overwritten with a rate of change from indicator III, which may for example be built as a composite variable representing changes in the production capacities of renewable energies. However, such composite variables are often not readily available in Eurobase and would have to be compiled from various Eurobase selections. The overlay structure of the indicators may also use only one variable but at different regional scope. For example, a sum of three big MS may represent an indicator I, a sum of 5 MS indicator II and the EU-28 aggregate indicator III. Indicator III would be used if EU-28 data are available, if EU-28 data are not available the sum of 5 or 3 MS would be used to indicate the rates of change. Method 7 for an estimate of the volume change should be an expert estimate. Field experts often have earlier access to information than the statistical system, but also the results of specific studies (e.g. projections to the future) may be assessed and used. This method can overrule all previous methods if an expert estimate is set. Methods for price rates of changes may, for example, be as follows: 1 assumption of no-change from the previous year: 0% 2 same rates of change as the general price level: [Index GDP current price/index GDP volume -1] *100 3 same rates of change as NACE specific price levels: [Index output current price/index output volume -1] *100 4 rates of change following the rates of change of specific price indicator I (from Eurobase) 5 rates of change following the rates of change of specific price indicator II (from Eurobase) 6 rates of change following the rates of change of specific price indicator III (from Eurobase) 7 rates of change set by an expert 11

13 Method 1 (no-change assumption) ensures that in any case a rate of change is set irrespectively of whether data for the other methods are available. However, if data on the GDP general price level is available (method 2) the no-change assumption will be overruled with the assumption that the price changes in the EGSS follow those of GDP. This is not a very specific assumption, but it acknowledges that changes in economy wide price levels should have some impacts on the level of prices in the EGSS. Method 3 uses price changes in output by different NACEs. Unfortunately up to now these data are unavailable for many MS and hence the indicator is not available in Eurobase at EU- 28 level. Indicators for single or groups of MS may be used instead. Methods 5 to 6 use specific price indicators. For example electricity, consumer prices at EU28 level from the Eurobase energy statistics may be considered as sufficiently good indicator I for price changes for the production of renewable energies in NACE D35. However, as D35 may also produce biogas a composite price indicator taking into account electricity and gas prices may be more adequate (indicator II). However, such an indicator is not readily available from Eurobase. A better indicate (indicator III) would present the energy generation prices (not price at consumer level). Such a price would have to be compiled by a decomposition of the consumer price using more detailed data on the various elements that constitute the consumer prices. Method 7 for an estimate of the price changes can be an expert estimate. This method can overrule all previous methods if an expert estimate is set. Limitations: The more sophisticated the construction of an indicator is, the more burdensome can be the simplified approach. When such composite indicators are not readily available in Eurobase but shall be used, they must be constructed as part of the early estimate approach. If in addition, some of the data are not available at EU-28 level (due to some MS failing to deliver the data) even more burdensome prior work is necessary to fill the gap. The advantage of the simplified approach in terms of resource savings over the full approach (that has been used to compile the data) may then shrink considerably. The simplified approach can only be better than the full approach if it is really kept simple. In contrast, using the full approach for early estimates has the advantage that it directly uses available data at MS level to the maximum possible (while missing data are filled up with simple assumption such as no-change, using trend values, ratios etc.). In fact, a very elaborated simplified approach may end up in replicating the same burdensome work that needs to be done if the full approach would be used directly. This does not in principle speak against the simplified approach as long it is consciously kept simple conceding that the available information for an early estimate is possibly not exploited to its maximum extent. Gross value added Gross valued estimates can be based on rates-of-change for intermediate consumption volumes and prices. Intermediate consumption is not a reporting variable in EGSS, but it can be derived from output and gross value added: Intermediate consumption 2013 = Output Gross value added

14 Value change (%) = [(1 + Volume change (%)/100) * (1 + Price change (%)/100) -1] * 100 Using the data of the latest available year (e.g. 2013) an early estimate (e.g. 2014) of IC and GVA is calculated: Intermediate consumption 2014 = Intermediate consumtion 2013 * [Value change (%)/ ] Output Gross value added 2014 = Output Intermediate consumption 2014 Methods for volume rates of changes of IC may be: 1 assumption of no-change from the previous year: 0% 2 set by expert: rates of change same as or x percent higher/lower than then output volume changes (implicitly assuming constant or decreasing/increasing partial productivities of IC products). Possible methods for price rates of changes are, for example: 1 assumption of no-change from the previous year: 0% 2 same rates of change as the general price level: [Index GDP current price/index GDP volume -1] *100 3 rates of change for NACE specific intermediate consumption price levels: [Index IC current price/index IC volume -1] *100 4 rates of change set by an expert Employment Employment estimates can be based on rates-of-change for employment. Employment 2014 = Employment 2013 * [change (%)/ ] Methods for the rates of changes may be: 1 assumption of no-change from the previous year: 0% 2 set by expert: rates of change same as or x percent higher/lower than output volume changes (implicitly assumes constant or decreasing/increasing partial productivities of labour). 13