Decomposition of the ESA2010 Eurostat net tax matrices and the transformation to GTAP format

The European Commission’s DG JRC IPTS undertook the task of updating/transforming the GTAP database's Input-Output tables for the 28 EU countries using the 2010 EUROSTAT Supply, Use and IO Tables (SUIOTs) and 'Taxes less subsidies on products' (TLS) matrices and auxiliary data, notably the MacMap for import duties, the 'Excise Revenue Tables' and 'National Tax List' data of DG TAXUD, the Eurostat's national accounts database and various publications of the national statistical offices. The first task was the decomposition of the TLS matrix to its VAT, excise tax, custom duty, 'other taxes on products' and subsidy components. This decomposition is not required for a GTAP data update, but a higher degree of detail makes the EU data more suitable for tax policy analysis. In the decomposition we exploited fully that the different taxes/subsidies usually affect only a limited number of users or products. The estimation process consists of the following main phases: 1. Estimating the row-totals (totals by products) for the component matrices (TLS layers), 2. Estimating satisfactory priors for each TLS layers, 3. Elaborating an entropy-model to simultaneously and consistently estimate the TLS layers. Since data availability, accounting methods and the tax system are heterogeneous across the EU countries, the general entropy model had to take into account country-specific constraints and adjustments. A partly different pilot-entropy model for Hungary was elaborated to take into account the very specific characteristics of its tax system. The entropy model seems to produce reasonable results for the decomposition. Then the TLS layers have been transformed to GTAP-sectors and split to domestic and import related parts. The transformed IOTs in GTAP format had to be consistent with the official statistics figures for value added, output and total imports too. We developed and used an other entropy model which represented all consistency criteria and other constraints.