South African National Accounts benchmarking experience Gerhardt Bouwer 1 Background Shared responsibility in compilation of South African national accounts Stats SA responsible for production, income SARB responsible for expenditure Close liaison between two organizations on reconciliation of their estimates Stats SA and SARB jointly benchmarks South Africa’s estimates of national accounts 2 South African benchmarking experience Previous benchmark saw a lot of changes in data, data sources and methodology These changes resulted in large level changes in data These changes influenced the national accounts time series data 3 Business Sampling Frame (BSF) Stats SA developed a new BSF, based on Income Tax and VAT database Survey data used for benchmark based on new BSF, while survey data before benchmark was based on previous used Business Address Register (BAR) The new BSF showed that SA economy was on a much higher level 4 Business Sampling Frame (BSF) cont. Year 2000 Output at basic prices PreBenchmark level (R’million) After Benchmark level (R’million) % change in level 1 619 481 1 893 686 16,9 5 Business Sampling Frame (BSF) cont. Value added (current prices) - Year 2000 Pre-Benchmark level (R’million) After Benchmark level (R’million) % change in level 54 951 63 391 15,4 Manufacturing 150 198 159 107 5,9 Trade, hotels, restaurants 107 299 122 702 14,4 Other personal services 48 979 51 382 4,9 All Industries 808 461 838 218 3,7 Mining 6 Business Sampling Frame (BSF) cont. 1998 1999 2000 2001 2002 2003 PreBenchmark: GDP (R’billion) 739 801 888 983 1 121 1 209 After Benchmark: GDP (R’billion) 742 814 922 1 020 1 165 1 251 % change in level 0,5 1,6 3,8 3,7 3,9 3,5 7 Economic Activity Survey (EAS) Stats SA started to use EAS with benchmark EAS formed cornerstone of the calculation of national accounts estimates Confrontation of data in SUT framework suggested that the distribution of economic activity between different industrial groups was not correct 8 Economic Activity Survey (EAS) cont. Two main explanation for this: Incorrect classifications Enterprise vs establishment 9 Supply and use tables (SUT) Started to use SUT to calculate annual GDP estimates SUT balanced on 95-industry, 48-commodity Confrontation between approaches on commodity level, previously only on big totals 10 Double-deflation Annual sets of SUT’s were developed for each year, which made it possible to introduce double-deflation Specific price indices could be linked to corresponding commodity groups Derivation of a weighted intermediate consumption and output for each industry Previously deflation by single price index 11 Quarterly estimates – Denton Method Quarterly estimates based on annual nominal and real estimates Preserve as much as possible of the short term movements in the new series Quarterly ratio average to annual ratios for each year 12 Conclusion Large level changes in data should be worked in over longer period than only two years More frequent annual data than periodical data might influence time series Establishment vs enterprise will always create problems Changing of methodology might also influence time series 13 End 14