GDP growth estimates for Europe at 30 days NTTS Conference, Brussel 16 March 2017 Eurostat – Hans Wouters
GDP growth estimates for Europe at 30 days Why introduce GDP t+30 estimates for EU and EA? Development work Quality acceptance criteria Test estimates Published GDP t+30 estimates Communication and information Conclusions
1. Why introduce GDP t+30 estimates for EU and EA? (1) Purpose: Assess the feasibility of producing GDP flash estimates at t+30 days for the EA/EU User needs Key users – European Commission, ECB Policy making Monitoring economic situation Forecasting Other users
1. Why introduce GDP t+30 estimates for EU and EA? (2) Logical next step after introduction GDP t+45 estimates in 2003 Fits in release pattern Principal European Economic Indicators Releases at 30 – 60 – 90 days
2. Development work Development national estimates Exchange of estimation practices and problems How to estimate missing (3rd month) data Development methodology for EA/EU estimates Aggregation of national estimates Special attention for rounding Preparation of EA/EU test estimates Development of quality acceptance criteria Development of a communication plan
3. Quality acceptance criteria (1) Purpose Assessment of test estimates Decision to go 'live' or not Quality criteria Unbiased GDP t+30 estimate Average revision between -0.05 and +0.05 ppt Max 67% revision in the same direction
3. Quality acceptance criteria (2) Quality criteria (continued) Limited average absolute revision Max 0.10 ppt at t+45 Max 0.13 ppt at t+60 Sufficient coverage 70% GDP for 2015 test quarters Availability communication plan
4a. Test estimates, general Purpose To assess expected quality of future GDP t+30 estimates Test quarters Real time estimations for 2014Q1-2015Q4 Retrospective estimations for 2012Q1-2013Q4
4b. Test estimates, EU t+30 results (1)
4b. Test estimates, EU t+30 results (2)
4c. Test estimates, EA t+30 results (1)
4c. Test estimates, EA t+30 results
4d. Test estimates, conclusions Successfully passed all a priori quality criteria Better than market forecasts at t+30 (tested for real-time quarters 2014Q1 onwards) But: revisions t+30 t+45 a bit larger than revisions t+45 t+60 Eurostat (and ESS) decided to start publishing GDP t+30 from 2016Q1 onwards
5. Published GDP t+30 estimates
6. Communication and information (1) Introduction videos in April 2016 Methodological working papers (available on Eurostat website) Methods and techniques for national GDP t+30 estimations Eurostat estimation procedure
6. Communication and information (2) Article in Eurona journal 1/2016 (available on Eurostat website) Presentations in meetings and conferences ESCB Working Group on General Economic Statistics, Febr. '16 IARIW – International Association for Research in Income and Wealth, Aug. '16 CIRET – pre-Conference Workshop on Rapid Estimates, Sept. '16 CESS – Conference of European Statistics Stakeholders, Oct. '16 NTTS – New Techniques and Technologies for Statistics, March '17
7. Conclusions User needs for earlier GDP estimates Project 'GDP t+30' aimed to assess feasibility of producing GDP flash estimates for the EA/EU at 30 days The project developed: Guidance for national GDP flash estimates Methodology for EA/EU estimation procedure Test estimates for 2012Q1-2015Q4 Quality acceptance criteria Test results passed quality acceptance criteria Eurostat started official publication in April 2016 Published estimates confirm test results
GDP growth estimates for Europe at 30 days NTTS Conference, Brussel 16 March 2017 Eurostat – Hans Wouters