Eurostat Expert meeting on Short Term Statistics UN ESCWA Amman, 16 February 2016 Arto Kokkinen - Eurostat On European experiences in Quarterly GDP flashes 1
Eurostat Content 1.Background: Flash estimates in Europe 2.QNA compilation practices in Europe 3.QNA data sources 4.Flash vs. regular Q GDP 5.GDP t+30 estimates: Why and how? 2
Eurostat 1. Background: Flash GDP estimates in Europe T+45 d flash estimates produced since 2003 gradually increased number of countries, now only 6 of EU28 MS are not compiling a t+45 flash T+30 d flash for Euro Area/EU begins 17 MS involved, seven publish at t+30 (UK, BE, ES, LT, LV, AT, FR) most give an internal estimate, country coverages > 90%, most countries have 2/3 months data (mainly monthly STS in Europe + X,M & defl. Ret trade) 3
Eurostat 2. QNA: compilation practices in Europe 1.Same principles used for compiling ANA can be used for QNA (even SUT framework) but at a less disaggregated level. 2.Methods that benchmark preliminary quarterly estimates of a QNA variable, or a quarterly indicator, to the corresponding ANA. (Direct GVA extrapolation by indicator in QNA, after ANA benchmark QNA to ANA figures) 3.Temporal Disaggregation Methods that impute values for a QNA variable by modeling the relationship between (annualized) preliminary quarterly estimates of a QNA variable, or a quarterly indicator, and the corresponding ANA. (Indirect GVA extrapolation by modelling quarterly indicator and annual variable relation) 4
Eurostat 3. QNA data sources for t+60d publication Output approach (O): good quarterly sources for many (Nace) branches. (STS) not necessarily for all (e.g. some services). Expenditure approach (E): Good sources in C, G, X-M(goods); issues in GFCF (I') and inventories. Sometimes output indicators used for expenditure items (C). Income approach (I): good sources for wages, taxes (possibly only late in the quarter). difficult for operating surplus (= often residual). Most reliable approach (O or E) in balancing depends on the data sources in the country this is then followed in the GDP Flash 5
Eurostat 4. Flash GDP vs. regular Q GDP Available earlier (in days) than the traditional estimates. Accuracy: there is a trade-off between timeliness and accuracy. The loss in terms of accuracy in flash is kept as small as possible. Limited set of information. For instance, often information from surveys concerning the whole quarter is not available. Estimation included: parts or areas that are estimated by statistical methods where -e.g. two months of short-term statistics and other available indicators (such as Economic Sentiment Indicator) are used typically in the framework of autoregressive distributed lag (ADL) models, ARIMAX models and time series regression techniques. 6
Eurostat 5. GDP t+30 Why?....(1) On the Euro Area/EU level t+30, 29/4/2016 User needs Key users – European Commission, European Central Banks o Policy making o Monitoring economic situation o Forecasting Other users Logical next step after GDP t+45 estimates Fits in release pattern PEEI indicators Releases at 30 – 60 – 90 days 7
Eurostat 5. GDP t+30 … data …(2) STS dataRequirement: To Eurostat Deflated retail tradet+1M IPIt+1M + 10d Construction (vol)t+1M + 15d Services (turnover)t+2M In t+30 Q GDP estimate: 3 rd month data problem: - third month data for defl. ret. Trade, IPI, construction, services missing, - Services TO; even second month may be missing, In t+30: third month of the quarter needs to be estimated In some countries STS gives internally an early estimate to QNA 8
Eurostat Sharing country experiences and best practices on the national GDP t+30 estimations in a Task Force Development of guidance document on (national) methods and estimation techniques Guidance for non-experienced Member States (contractor Cambridge Econometrics) STS internal 3rd month estimates in some countries 9 5. GDP t+30 …How country estimates ? (3)
Eurostat On European experiences in Quarterly GDP flashes 10 Expert meeting on Short Term Statistics UN ESCWA Amman, 16 February 2016 Arto Kokkinen - Eurostat