Flash Estimates of Gross Domestic Product International Seminar

Flash Estimates of Gross Domestic Product International Seminar on Early Warning and Business Cycle Indicators 15 December 2009

Overview n Recap of contributions in Ottawa as background. n Diversity of approaches, different stages of deveopment, adequacy of data sources (statistical infrastructure), terminology (flash estimates, nowcasts, basic compilation). n Stocktake by UN ESCAP (not of just GDP estimates) but also of other short-term economic statistics confirms diversity of not only approaches but of readiness and capacity n Three country papers (Morocco, Belarus and China)

UNESCAP Assessment n Covers 58 ESCAP member and associate member countries, results by geographical region (ENEA, SSWA, NCA, Pacific) and income (high, medium, low) n Great diversity ranging from advanced, wide coverage, state-of -art methodologies (e. g. Australia) to less advanced (availability and methodologies) and under-resourced at early stage of development n Assessment covers 6 aspects – national accounts, selected economic indicators, timeliness, availability of data sources, business registers, classification, seasonal adjustment

UNESCAP Assessment n National accounts ¨ ¨ n Only half compile quarterly GDP; for SSWA and Pacific most economies do not have quarterly GDP Minimum Requirement Data Set: 50% able to produce 6 or more tables (26% 7 tables, 24% 6 tables). UNSD Milestones – 60% 2 nd milestone. Only more developed members able to go beyond. I-O (as well as Supply-Use Tables) not widespread; only 19 compile I-O tables. Basic economic statistics: CPI, Industrial Production Index, Employment, Trade, BOP, Wages, Earnings, Services Volume

UNESCAP Assessment n Data Sources - Censuses and surveys (generally available); administrative data (less developed) n Business Registers – 34 members n Classifications – ISIC, ISCO, COICOP, SITC. Less available or used: ISCED, CPC. n Seasonal adjustment – not widely carried out though importance recognised. Most common methodology: X 12 ARIMA.

UNESCAP Assessment n Some members relatively advanced but most are less developed, requiring assistance and support. n Sub-annual data not widely available; little or no expertise in seasonal adjustment. n Conclusion: Flash GDP extremely challenging for most; need to develop basics and build up statistical infrastructure (business registers, administrative data sources, expertise).

Belarus n Monthly GDP or flash (or nowcast) compiled using output or production approach n Quarterly and annual GDP compiled using all three approaches (output, expenditure, income) n Timeliness Monthly – T + 12 (based on state statistical reports and physical volume indices on output) Quarterly – T + 90 (supplemented by government finance, BOP data, etc) Annual – (further supplemented by even more data)

Belarus n Compilation Production data on major branches (ie industries) Cumulative rather than by discrete month n Question: Is quarterly the cumulation of the 3 months? Annual the cumulation of the 4 quarters? n If cumulation, will be difficult to do seasonal adjustment and assess recent trends. n If not cumulation, how to reconcile sub-quarter estimates with quarter estimates, sub-annual with annual estimates. Benchmarking carried out?

China n Flash quarterly GDP estimates using production approach n Compiled industry by industry n VA at current or constant prices is obtained by extrapolating by growth rates adjusted by an ‘adjusting coefficient’, then deflated or re-inflated by appropriate price indices n Essentially a nowcast (and therefore extremely timely, T + 10 to 15) obtained by extrapolation under various assumptions. Revised T + 45 (later T + 100). n Successful in providing early indications of economic activity. But problems and challenges remain.

China n Not sufficiently dis-aggregated (not an uncommon problem with flash or rapid estimates) for quarterly economic monitoring n Cumulative (and not discrete) quarterly estimates making it difficult to assess trends (same as Belarus’ flash monthly estimates). n No expenditure-based quarterly GDP estimates n Though highly timely, may need to be even more timely requiring forecast of flash quarterly estimates, i. e. forecast of nowcast.

China n Going forward, will move to discrete quarterly GDP, allowing for seasonal adjustment and better adjustment of trends n Compile also quarterly expenditure-based GDP estimates, which will improve economic analysis and monitoring but will also require reconciliation between the two n As with other sub-annual discrete estimates, will need to address how to benchmark quarterly estimates with more complete annual data.

Morocco n Morocco’s national accounts complete, but generally with annual periodicity, i. e. good but is of low rather than high frequency. n Need to supplement with conjuncture analysis, high frequency synthetic data, anticipatory (possibly soft) expectations data and analysis of business cycles n Quarterly GDP estimates, released by National Institute (T + 1 month) as 1 st estimate of the conjuncture analysis, is used as leading indicator. Also release at same time forecast of next quarter’s GDP.

Morocco n Flash quarterly GDP estimate (T + 45) not only of production, but also of income and expenditure crucial. n Use also high frequency indicators (like CPI, IPI, etc) prescribed by SDDS, consumer sentiments data and business cycle analysis. n High frequency indicators for some important industries (notably agriculture) not available. n More timely estimates (forecasts? ) of annual GDP obtained using econometric models with coincident indicators as explanatory variables.

Conclusion n Need for more timely assessment of trends (whether based on flash, 1 st estimate, forecast or nowcast) is clear. But will be challenging for most countries. n Primary priority must still be to build up on statistical infrastructure and capacity, compile regular (may refer to them as flash or rapid, as is the case with Belarus and China) quarterly GDP estimates on a more timely basis. n Sentiments (whether of consumers or businesses), coincident and leading indicators arrived at from business cycle analysis and studies are useful and valuable supplements.