Estimated bilateral trade in services by industry (EBTSI) Methodology used to create a matrix of bilateral trade in services by industry for the OECD global I/O.
Download ReportTranscript Estimated bilateral trade in services by industry (EBTSI) Methodology used to create a matrix of bilateral trade in services by industry for the OECD global I/O.
Estimated bilateral trade in services by industry (EBTSI) Methodology used to create a matrix of bilateral trade in services by industry for the OECD global I/O table Constructing OECD MRIO tables • What is needed: A perfect matrix of bilateral trade flows (goods and services) decomposed by industry and by end-use • Issues: – Overlap between goods and services trade statistics – Missing data in the case of trade in services – Lack of adequate methodology to assess trade in intermediate services – Correspondence between EBOPS and ISIC classifications Allocated versus non-allocated trade in the OECD TISP database (2005) 100% = Exports reported for EBOPS 200 (total trade in services) with world as partner 120% 100% 80% Average = 69% 60% 40% 20% Share of total trade accounted for when adding over all partners and industries USA TUR SWE SVN SVK RUS PRT POL NZL NOR NLD MEX LUX KOR JPN ITA ISR IRL HUN HKG GRC GBR FRA FIN EST ESP DNK DEU CZE CHL CAN BEL AUT AUS 0% Methodology • • Step 1: bilateral matrix based on “hard” data – OECD TISP database – EUROSTAT database (for EU countries) – UN trade in services database (for non-OECD economies) – IMF (for world total and rest of the world) Step 2: reconciliation of bilateral data – • Exports preferred and then mirror imports Step 3: Estimation of missing bilateral flows – Prediction of zero trade flows (missing versus zero) – Extensive margin – Estimation of bilateral trade flows – Intensive margin • Step 4: Industry decomposition based on national accounts • Step 5: Collection of bilateral by industry data (conversion to ISIC) • Step 6: Optimisation to estimate bilateral trade flows by industry – First set of constraints: total bilateral = total by industry for a given country – Second set of constraints: existing data for specific bilateral relationships Data coverage • Three years: 1995, 2000 and 2005 (panel data for the estimation 1999-2009) • 58 countries + “rest of the world” • 15 services industries (ISIC Rev. 3) 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 Construction Wholesale & retail trade; repairs Hotels & restaurants Transport and storage Post & telecommunications Finance & insurance Real estate activities Renting of machinery & equipment Computer & related activities Research & development Other Business Activities Public admin. & defence; compulsory social security Education Health & social work Other community, social & personal services Gravity model and estimation • Structural gravity: Anderson and Yotov (2010) • Estimation technique: quasi-maximum likelihood Poisson – To take into account zeros – To address the “adding-up” issue in gravity estimation (the fact that the sum of estimated trade flows exceeds the total of observed trade flows) Optimisation Trade matrix normalised to national accounts figures Total bilateral C1 C1 C2 C3 ROW Total 81 30 8 119 Ind 1 C1 45 15 110 9 49 C1 C2 C3 ROW Total . 4 . 35 C2 50 C3 21 19 ROW 12 22 8 Total 83 122 83 32 320 Imports by industry C1 C2 C3 ROW Total Ind1 58 64 30 17 169 Ind2 46 40 19 18 123 Ind3 15 6 0 7 28 Total 119 110 49 42 320 42 Exports by industry C1 C2 C3 ROW Total Ind1 35 74 43 17 169 Ind2 37 40 35 11 123 Ind3 11 8 5 4 28 Total 83 122 83 32 320 Ind 2 C1 C2 C3 ROW Total C1 Ind3 C1 C2 C3 ROW Total C1 National accounts Predicted zero . 10 . 37 2 0 . 11 C2 C3 ROW Total . . . . . . 58 64 30 17 169 . . 74 C2 . . . 40 C2 . 0 . 8 . 43 C3 3 . . 35 C3 . . . 5 17 ROW . . . 11 ROW . . 0 4 Total 46 40 19 18 123 Total 15 6 0 7 28 Known figure coming from normalised trade statistics (used as lower bound) Robustness checks • Comparison between estimated and observed data (in particular for industries where EBOPS and ISIC categories are similar). • Full estimation (at the bilateral by industry level) versus estimation + optimisation How can you help? • We need more bilateral data broken down by industry – Some information on unallocated trade? • Identify zeros (as opposed to missing or confidential data) • Data for 1995