EUROPEAN COMMISSION EUROSTAT Directorate C: National Accounts, Prices and Key Indicators Unit C2: National Accounts production Participants: Eurostat: Silke Stapel, Ani Todorova, Christine Gerstberger, Hans Wouters, Jenny Runesson, Håkan Linden (STS), Frank Espelage (LFS) Statistics Austria: Johannes Chalupa Statistics Denmark: Rasmus Rold Sorensen Statistics Hungary: Klara Anwar Statistics Lithuania: Jurga Rukšėnaitė Statistics Slovakia: Peter Čirka Statistics Italy: Maria Giovanna Piras Statistics Spain: Cristina Gonzalez Fidalgo Czech Statistical Office: Jaroslav Zbranek Federal Statistics Office, Germany: Cristoph-Martin Mai National Bank of Belgium: Bernadette Boudry ECB: Colm Bates DG ECFIN: Christoph Maier Final minutes Task Force Employment flash estimates, 28-29 March, Luxembourg 1. Opening and introduction of participants The Eurostat Director for National accounts, prices and key indicators, Ms Silke Stapel-Weber welcomed all task force members and thanked them for their participation in the employment flash project. The chair opened the meeting and also welcomed all participants. The participants introduced themselves in a short tour-de-table. European Commission, 2920 Luxembourg, LUXEMBOURG - Tel. +352 43011 http://ec.eurostat.ec.europa.eu
2. Presentation of the project Eurostat presented the content and background of the project 'employment flash estimates'. The reasons for starting the project (user needs) as well as the purpose (assessing the feasibility of producing quarterly employment flash estimates at t+45 days) were explained. The main elements of the project were listed, of which the preparation of test estimates is the most important one during the project implementation phase. It was mentioned that some decisions have to be taken already soon: at what timeliness should the flash be compiled, what variables should be covered, etc. Finally, a distribution of work between Eurostat and the TF members was proposed. ECB asked about the project duration. Eurostat explained that the life cycle of the project is two years and the goal is to have the first publication in the beginning of 2019. There were concerns that the methodological revisions planned in many Member States in 2019 could negatively influence the results of the project. However since the publication of the employment flash will most likely be growth rates only, the effect should be limited. DG ECFIN asked if there would be a possibility to also advance the regular employment estimation from t+75 days. Also, a flash estimate at t+45 days is too late for their forecasting schedule, since the forecasts are released at t+40 days. Eurostat answered that advancing the t+75 regular estimate is being considered and awaits improvements in the internal system for production of national accounts and the expiration of many existing derogations. Eurostat also clarified that regular progress reports on the project will be provided to the NAWG and DMES. 3. State of play employment flash in Member States Results questionnaire State of play employment flash in Member States; availability and timeliness (t+45 and/or earlier/t+30) Eurostat started the introduction by providing information about the current employment news release that is published at t+75 days. Subsequently the results of the questionnaire on the availability of employment flash data in the Member States that was prepared for the November 2016 NAWG were presented. An important conclusion is that some employment data are available at or before 45 days after the quarter-end in about 14 countries. In the final part of the presentation some information was shown about the expected available information on employment flash estimates in Member States that do not join the TF. Tour de table The members of the task force briefly presented their work on developing flash estimates on employment. The information is available in the annex of this document. Many Member States are already able to estimate employment at t+45 days or even earlier. It was noted that several 2
Member States are considering to estimate employment at t+30 days, in order to publish the GDP and employment flash at the same time. 4. Requests regarding employment data from key users DG ECFIN ECB DG ECFIN produces macroeconomic forecasts three times a year for the European aggregates and Member States data. They use the employment data as input for these forecasts, but also for briefings and economic analyses. The forecasts are published at t+40, for which a timeliness of t+45 is too late. The preferred coverage of an employment flash news release would be including the Member States growth rates and a split of employees and self-employed. DG ECFIN also expressed the need to share and document the countries' estimation methods, which is valuable information for the country desks. The ECB mentioned several reasons why it needs early employment data: as input for computing other statistical indicators, for forecasting purposes, for internal briefing and conjuncture analysis. The ECB also showed what employment data it publishes and encouraged the Task force to also work on advancing estimates of hours worked and income. It was concluded that it is important to have 'hard' employment flash data that are available at the same time as other important indicators and preferably have a long term goal to also publish employment data at t+30 days. DG EMPL was not represented in the meeting but has expressed interest in the project. There was a brief discussion on the ideal timing of estimates in order to fit user needs. The main policy users of the European agregates expressed preference for an earlier date with more limited content and were of opinion that there would be synergies in producing GDP and employment flash releases at the same time. Regarding the decision on the exact publication dates, Eurostat explained that a preliminary release calendar for the coordination of flash estimates is usually presented to the May NAWG for its approval. The first publication of the possible employment flashes will be decided upon the results of the project and coordinated at national and European levels. The target is a first release in May 2019. 5. Experiences from the GDP flash project Last year Eurostat successfully finalised the project 'EU/EA GDP t+30 estimates'. The employment flash project has many similarities with the GDP t+30 project and can probably benefit from the lessons learned in the GDP project. Therefore Eurostat presented the purpose of the GDP +30 project and the reasons to start it. Furthermore Eurostat mentioned what work was done (e.g. preparation of test estimates) and how it was organised. Last but not least the strong 3
points of the project were emphasised; the very good cooperation between Member States and Eurostat was one of them. 6. Sources for employment flash estimates Availability of employment data in LFS Availability of early employment data in STS and inventory of sources National accounts based employment flash figures can only be compiled with help of sufficient source data. Labour force survey data (LFS) and short term statistics data (STS) are possible data sources. Eurostat elaborated the concept, results and data availability of the EU-LFS. Furthermore, the data requirements and data availability regarding the compilation of the monthly unemployment rate (MUR) were explored, including the data availability at national level. The MUR is currently compiled based on a gentleman's agreement, but a legal change is in progress, which would potentially require data transmission at t+25 days. Monthly employment estimates are transmitted by 13 Member States at t+30 days, and the publication of LFS unemployment rates mean that employment estimates could be derived in priciple. The availability of final quarterly LFS data at t+30 days is very limited. Regarding STS employment data, Eurostat presented the reporting requirements, concepts and coverage of the labour indicators. Eurostat also presented overviews of the STS monthly and quarterly data availability and their timeliness. The available index for the number of persons employed was discussed in more detail. This index is to be transmitted quarterly at t+60 days (or t+75 days for smaller Member States). However, nine Member States are on a voluntary basis transmitting the index monthly with average transmission delays of t+43 and t+48 depending on the NACE activities. The members of the task force asked to receive methodological documentation on the LFS flash calculation. Questions on the quality of the monthly indicator and the economic sector breakdown in LFS were also raised. A tour de table followed, where the task force members described the use of sources and timeliness needs. The extent to which the different sources are used differ widely in the countries, however LFS tends to be more used than STS, due to the later timeliness. There are many developments under way, both legally and methodologically in LFS and STS respectively, and some task force members considered to use more timely estimates in national accounts. Often the task force members refered to difficulties to obtain information on self-employed persons. The task force members asked to be kept aware about possible improvements of LFS and STS as well as the risk of breaks in the time series resulting from the adoption of the regulations integrating social and business statistics respectively. 4
7. Guidance document for national estimation techniques The availability of national employment estimates is crucial for the estimation of employment estimates for the euro area and the EU. Some countries are already quite experienced in estimating early employment figures, but for some other countries this might not yet be the case. Therefore Eurostat addressed the question whether it would be necessary to prepare a 'guidance document for the preparation of early national employment estimates'. In the presentation Eurostat suggested possible topics to be included in such a guide and a proposal how to organise the drafting process. It was also emphasised that this part of the project, if considered necessary, should be carried out by some task force members. The members of the task force confirmed the need for documenting national estimation techniques and best practices, which would be especially useful for the Member States that do not yet publish an employment flash. Data users and other data producers would also benefit from such a document. The existing paper on GDP flash techniques could be a starting point. From a tour de table it was noted that only few members currently have official documentation available of the national estimation methods of employment flash. Johannes Chalupa from Statistics Austria accepted to be the coordinator in preparing the methodological documentation. It is planned that an outline of the document could be circulated soon so that a first draft could aleady be made available before the next meeting in September-October 2017. Priority will be given to the estimation method at t+30 since it would represent something new for most of the countries. 8. National accounts employment flash estimates for the euro area and the EU Variables to be estimated Test estimates and reporting template Task Force's participants contributions to test estimates tour-de-table The preparation of test estimates is considered to be the core of the project as the test estimates should provide an impression of the expected quality of future euro area and EU employment flash estimates. Eurostat proposed to prepare test estimates (quarter-on-quarter and year-on-year growth rates) for the variable 'total employment in persons'. For the transmission of the national estimates Eurostat developed a draft reporting template. The participating Member States are requested to introduce their national estimates for 8 real time quarters and 8 back quarters in this excel template and send it to Eurostat via the secure edamis transmission channel. The retrospective quarters should be transmitted by end of 2017. Following discussions during earlier agenda items it was clear that an employment flash estimate at t+45 days is feasible and t+30 days is also a possibility. 5
The members of the task force were asked, in a tour de table, if they would be ready to provide estimates at t+44 and t+30 days during 8 real time quarters, back estimates for 8 quarters (t+45 days) for total employment, unadjusted and seasonally adjusted (growth rate or levels). All members were ready to provide real time estimates at around t+45 days. They were also willing to try to carry out estimations at t+30 days; either starting immediately or after the t+45 estimate is in place (during the autumn). Some expressed doubts of the availability of back estimates, but would try to reconstruct the information. 9. Estimation procedure for EA/EU flash estimates and development of assessment criteria for test estimates For the compilation of the euro area and EU estimates, the national contributions should be aggregated. Subsequently, the results of the euro area and EU test estimates should be assessed against some quality criteria. Regarding the compilation of the euro area and EU employment flash aggregates, Eurostat presented a possible method, based on the method that is currently in use for the GDP flash estimates. Concerning the definition of quality criteria, Eurostat showed the criteria that were used for the GDP t+30 project and that can probably serve as a starting point for the development of similar criteria for the employment flash project. In addition, the results of a first revision analysis (comparison between t+75 and t+165 employment data) were presented. Eurostat will prepare a proposal for the quality acceptance criteria that may be used to control the test estimates and present it for discussion at the coming meeting. 10. Presentation of time schedule and to do list Eurostat presented the proposed time schedule for the coming two years. The time schedule included the most important actions/deliverables, such as the preparation of the test estimates, the planning of the TF meetings and several other key events of the project. The next meeting is planned for September/October and if possible, taking place in a Member State. The following tasks were distributed: Eurostat: Propose release calendar. Create Circa folder where all relevant documents should be stored (presentations, flash documentation (LFS) etc.). Provide edamis identifier for data transmissions. Set up estimation procedure for test estimates. Will send reminders one week before test estimates are due. 6
Members of the task force: The Austrian delegate will prepare a draft outline of the methodological document, which will be circulated end April/beginning of May. The German delegate will contribute on the estimation methodology at t+30. Members should prepare back estimates 2015Q1-2016Q4 for t+45 days and transmit to Eurostat before end of 2017. Members should also prepare real time test estimates and transmit to Eurostat around t+45 days, and preferably also t+30 estimates if possible. 11. Conclusions and closure of the meeting The chair thanked all participants for their valuable contributions to this fruitful meeting. The next meeting will take place in September-October 2017. Eurostat invited the countries to consider offering a venue of this meeting. 7
Annex Czech Republic: An employment flash is already available at t+45 days. The unadjusted and seasonally adjusted growth rates are published. The sources are LFS and social security registers and at the time of publication, information for the first two months of the quarter is available. Denmark: Data sources are available at t+50 and the estimate is published at t+60 days, in both levels and growth rates. The data source is working time account from the tax authority. In order to produce a flash estimate at t+45 days, it would be necessary to either advance the source data or, since the source data is monthly, it could be possible to model the last month. Italy: At the time of publication at t+60 days, 100% of source data is available. At t+45 days, 85% of information would be available. There is also a plan to produce a t+30 estimate and analyse revisions against later estimations. The OROS survey is used for industry information, but is only available at t+55 days after the quarter. Lithuania: Total employment and A10 breakdown in persons and hours worked is published at t+45 days. LFS data is available at t+40 days for total employment and Arima models are used to nowcast for the breakdowns. There are very small revisions. Advancing the estimate to t+40 days could be realistic. Hungary: LFS is used as a data source to estimate self-employment and business surveys for employees. To produce a flash estimate for employment it could be possible to use the flash modelling approach from the GDP flash and model the 3 rd month out of 2 available months. Germany: The total employment is published monthly at t+30 days. At t+45 days, the A10 breakdown and hours worked data are published. Data sources are social security registers and LFS, however the published quarter is mainly based on expert calculations and forecasts. Austria: Number of persons employed is published at t+45 days. The main data source is social security registers available at t+26 days, complemented by LFS to estimate unpaid family workers. The A10 breakdown is derived from business registers. T+30 days could be a possibility. Belgium: Data is currently published at t+60 days, using econometric methods and based on available information on the number of bankruptcies, the value added for the manufacturing industry and the number of employees for previous quarters. Social security data is only available at t+55 days so it is not used in the estimation. The same method could be used for t+45 days and probably possible also for t+30 days The output of the model is a quarter-on-quarter calendar and seasonally adjusted growth rate for the considered quarter from which we derive a year-on-year growth rate for the number of employees. The growth for total employment can then be derived by making the assumption that the growth of the number of self-employed is similar to the growth observed for the previous quarter. Slovakia: Seasonally adjusted employment in persons is published at t+45 days. The estimate is based on different surveys, depending on the sector. The sources are available at t+37 days. T+30 estimate could be possible, if using monthly surveys. There are no revisions between t+45 days and t+60 days because there is no additional information available in that time span. Spain: The employment is estimated at the same time as the GDP flash estimate at t+30 days, but in full time equivalents. The employment data is based on LFS as a main source, but also social 8
security records, retail trade and services data from STS. Spain is planning to estimate at t+30 days, since there is no additional information available between t+30 days and t+45 days. 9