SESSION 4 Extrapolation modelling econometric and sampling techniques

SESSION 4: Extrapolation, modelling, econometric and sampling techniques used in the preparation of rapid estimates Review of papers 1

4 papers • • Economic projection at NESDB – Thailand Monthly estimate of GDP in Peru Canadian Monthly GDP estimates Singapore Advance GDP estimates 2

Economic projection at NESDB – Thailand • Through 2 models – Current quarterly model (CQM) • Uses data on indicators to estimate GDP from both demand supply side • Short-term projections • In a range, progressively reduced to point estimate – Quarterly financial model (QFM) • CGE model • Used 29 endogenous and 28 exogenous variables • For medium and long-term projections of GDP 3

Comments • Monthly data is available on a number of indictors, both from supply and demand side – some indicators from VAT • Most countries use this data directly to extrapolate benchmark GDP estimates • GDP short-term estimates are given in a range, which at most times becomes not readily usable. • Long-term projections using CGE model – is it really the mandate of NSOs • Most of these models do not pick up abnormal or sudden events 4

Monthly estimate of GDP in Peru • Based on indicators – Monthly index of production • information provided by public and private agencies • 1994 base – Administrative registers – Surveys • Value of sales and personnel – Output as an indicator to extrapolate GDP • 45 days timeliness 5

Comments • Old IP base year – relative price movements are fixed – Output as proxy to GDP 6

Singapore Advance GDP estimates • Output based QGDP – 2 weeks (advance) • Broad 3 industry break-up – 8 weeks (preliminary) • More detailed industry break-up • Advance estimates use – Data on indicators (one month less data) – ARIMA models – Judgement • Quality – assessed from next set of QGDP release – Generally unbiased 7

Comments • Uses available data at the end of the quarter – For 3 key industry-groups – Generally 2 months data, given that most indicators have a timeliness of one month – Very good advance information, provided nothing drastic happens in the last month 8

Canadian Monthly GDP estimates • Since 1919, Industry output in the form of Index of Physical Volume of Business • Since 1971, monthly GDP by industry • Annual estimates prepared in parallel to benchmark monthly estimates • Since 1986, benchmarked to IO accounts • Reconciled to income-expenditure accounts • Approach – Projector or indicator, rather than econometric, of latest IO accounts – Focus on production in real terms – timeliness of 60 days – Compiled for 240 industries 9

Contd. . • Sources – Surveys (output and employment), government accounts, private firms, industry associations • Extrapolators – – Turnover and inventories (deflated) Physical indicators Employment Use of GST – being explored • Releases – Volume terms • Constant price Laspeyres • Chain volume (focus of dissemination) – closer to chained-Fisher due to integration with QNA • Revisions – Monthly, alongwith quarterly, annual, IO tables, rebasing and historical 10

Comments • • • Comprehensive compilation For how many industries, data is released Too many revisions – Can monthly revisions be avoided – Only alongwith QNA and ANA • Any plans for monthly GDP expenditures 11

General comments • Source data – Administrative sources (Peru) • Indicator approach – alternative data – Tax data (as in Thailand) – Companies quarterly accounts – Government expenditures • both extrapolation and econometric models – Output as proxy for value added • Periodicity – Monthly GDP for all countries within one month, may be difficult in view of the timeliness of data on indicators – should be within one month – Therefore, minimum quarterly – both production and expenditure • Timeliness – 45 days appears minimum for QNA – Singapore experience of advance estimates based on partial data can be considered for replication • Revisions – Can we avoid revisions till the annual national accounts release • Dissemination – Providing commentaries alongwith estimates 12

Thanks 13