The Development and Production of GDP Flash Estimates in a Newly - - PowerPoint PPT Presentation

1

The Development and Production of GDP Flash Estimates in a Newly Industrialised Country: The Case of South Africa

Nazeem Mustapha and George Djolov

Dr Nazeem Mustapha

2

Contents ^ Introduction ^ Flash estimate definition ^ SA specifications ^ Methodology description ^ Data sources ^ Trial/experimental results ^ Conclusion

3

^ The recent economic crisis has underscored the importance of more timely statistics that serve as early warning signals on the state of an economy. ^ In this context an early estimate of GDP provides valuable official information to policy- makers and decision makers on total economic activity. ^ Flash GDP is one such statistic commonly compiled by various statistical agencies employing varying methods. ^ The aim here is to report the method and results of the development and production of a flash estimate of GDP at Statistics South Africa (Stats SA). ^ The use of the flash estimate as an early indicator of business cycle slowdowns and upturns will also be discussed demonstrating one possible use of the flash estimate. ^ The Stats SA experience could inform national statistical organisations operating under similar statistical production constraints how a flash estimate could be constructed

Introduction

4

Flash estimate definition

^ A flash estimate of GDP is the earliest picture of the economy according to national accounts concepts, produced and published as soon as possible after the end of a reference quarter, using a more incomplete set of information than that used for traditional/regular quarterly figures. ^ This definition does not spell out how to construct a flash GDP estimate. ^ It is the purpose of this presentation to showcase one possible interpretation of this definition as a schema workable in practice. ^ The suggested approach is based on the premise that identified methods or techniques for flash GDP in the existing literature should not be treated as cast in stone, since there is no unique solution to producing a flash estimate of GDP. ^ Thus, the Stats SA experimental approach offers one possible way by which this may be done.

5

^ Based on organisational requirements, the Stats SA flash estimate has to satisfy: fit into the existing statistical production processes of the statistical office, which are limited by statistical capacity; in line with well established national accounts approaches to GDP; brief and easy to communicate in the sense that it depicts a single number; derivation methodology has robustness as its main statistical property, in the sense that under situations where violations occur to the underlying assumptions of its derivation, the estimate’s reliability is unaffected or at most minimally affected; and adheres to international standards, in the sense that the data inputs, methods of production and quality criteria conform to what peer statistical agencies have tried and tested.

SA specifications

6

^ The South African flash GDP measures the year-on-year quarterly volume development of total value added (TVA) at basic prices. For this reason it is named Flash TVA. ^ Flash TVA production is independent from the detailed estimates of total value added coming from quarterly national accounts, with production systems and processes located in a separate unit, i.e. Socio-Economic Integration. ^ The Stats SA flash GDP is released at 35 days after the end of a reference quarter. ^ This compares favourably with the times of other statistical agencies: e.g. Statistics Singapore releases its flash/advance GDP at 17 days after the end of a reference quarter, the UK Office of National Statistics at 27 days, the US Bureau for Economic Analysis at 30 days, and the overall EU and OECD average is at 45 days.

SA specifications (continued)

7

^ The Flash GDP gives the first preliminary picture of volume changes to domestic output in a quarter. It is independent from the detailed estimates of total value added of quarterly National Accounts. ^ Its production draws on a mixture of actual observations and imputations for the cases where information is missing, involving data sources from within and outside Stats SA. ^ The imputation is usually conducted for the last month of a quarter for some data sources, and is done in order to meet the 35 day deadline. ^ The missing data comprises one-ninth of the quarterly change of total value added. ^ The flash estimate is revised upon receipt of data previously imputed. These revisions are reported quarterly.

Methodology description

8

^ Total value added is valued at basic prices and compiled according to the 1993 System of National Accounts. ^ The volumes of source data for Flash TVA are obtained by deflation using the PPI where

• necessary. This is done with reference to constant prices in terms of a 2005 base year.

Rebasement is done once every 5 years. ^ The quarterly coincident indicator is calculated by summing up the monthly indicator for three months. ^ The construction of the Flash TVA estimate involves combining an annually quarterised estimate of total value added obtained from changes in the population growth rate and income per capita, with an estimate obtained from a composite indicator which monitors the monthly changes in total output. ^ The Flash TVA estimate is the average of these two estimates, subject to quality control procedures on each individual estimate.

Methodology description (continued)

9

^ This approach of deriving the Flash TVA yields a robust Flash TVA estimate by producing a reliable measure of changes to domestic output. ^ This measure comes with the capacity to depict the pace of production in the economy. ^ Graphically, this pace is captured by the Flash TVA snapshot (to be presented under results). ^ The Flash TVA snapshot classifies the pace of production as normal, warm, hot, cool, or cold as represented by the regions between yellow and light-blue, between yellow and red, above red, between light-blue and blue, and below blue lines, respectively. ^ These reference lines are recalculated once every 5 years as part of good indexing practice. ^ The flash estimate is a statistic derived by a measurement model.

Methodology description (continued)

10

^ Construction of Flash TVA is done with data from 1998 to date taking account of changes/upgrades in National Accounts practices at Stats SA . ^ Value added and population data for quarterised estimate come from quarterly National Accounts and mid-year Population estimates as compiled by Stats SA. ^ Data for monthly indicators come from Stats SA, complemented by records from the South African Futures Exchange (SAFEX), South Africa’s Richards Bay Coal Terminal and Transnet. ^ 26/33 data points (79%) available each quarter to compute composite/coincident indicator. This ratio may change with the addition of new potential candidates in the construction of the composite indicator as part of beefing up coverage on services activity. ^ Missing data is imputed. ^ Table 1 clarifies the situation.

Data sources

11

Table 1: Availability of monthly variables

Data sources (continued)

m 1 m 2 m 3 Agriculture X X Impute Mining X X Impute Manufacturing X X Impute Electricity X X X Construction X X Impute Retail X X Impute Motor Trade X X Impute Wholesale X X Impute Transport X X X Services X X X Exports (commodity intensive) X X X

12

^ The results reported are from simulated production of Flash TVA as if it were an official statistic under realistic operational and data constraints as set out above. ^ The first estimate was produced at the beginning of January 2010 for the last quarter of

• 2009. Table 2 gives summary of the results to date.

Table 2: Key estimates

Trial results

Period GDP results Flash TVA results Total value added (b.p.) Y-o-y change in TVA Total value added (b.p.) Y-o-y change in TVA Total value added (b.p.) Y-o-y change in TVA Preliminary Revised 09Q4 409

• 1.1

411

• 0.7

410

• 0.8

10Q1 396 1.9 389 393 0.9 10Q2 408 3.1 403 1.8

13

^ The revised flash results differ from the standard GDP results by: 0.24% and 0.76% in level estimates 0.3 p.p. and 1.0 p.p. in change estimates ^ The flash estimates have been revised by: 0.24% and 1.54% in levels 0.1 p.p. and 0.9 p.p. in changes ^ As compared with the best possible forecast the results are as follows: 09/Q1 09/Q2 09/Q3 09/Q4 MAPE Flash TVA / MAPE ARIMA: 0.58 0.56 0.56 0.56 ^ Examined historically the flash estimate tracks the pace of production in tandem with the standard GDP as captured in the Flash TVA Snapshot, Figure 1.

Trial results (continued)

14

Figure 1: Flash TVA Snapshot (Historical review) ^ The flash estimates are consistently in the same regions as the GDP estimates, and also trace the latter’s turning points

Trial results (continued)

15

^ Subject to an independent and reliable statistical production the Flash TVA can be an official statistic with the advantage of reaching policy-makers and decision-makers quicker than the regular main aggregate. ^ This is not to argue that the latter is without value. The flash and regular estimates are complementary to each other, and a full fledged Flash TVA production programme can be an indispensable companion to its regular counterpart by giving a parallel view of the state of production in the economy as a quality tool. ^ The Stats SA experimental approach of a Flash TVA released 35 days after the end of a reference quarter strikes a healthy balance between being fast and giving a reliable reading

• f the main total value added aggregate that can also shed light on emerging conditions

as depicted in the snapshot. ^ The suggested approach offers a straightforward step-by-step methodology that can be applied in practice by an average qualified national accounts

• practitioner. As a result the South African method may also prove valuable to other countries
• n the path of developing their own flash estimates.

Conclusion

16

Thank you

Conclusion