Data availability and problems

I-O tables: For some countries the I-O tables were not available for the selected base year (1995). In this case the I-O table had to be compiled from the MAKE and USE matrices (see the example of Greece in Revesz-Zalai (2003)).

The first usual problem is that the break-down of the I-O tables (and the nace 2-digit level break-down of the national accounts data for branches) are not sufficiently detailed in the case of the energy sectors. It can be seen from the branch classification table in section 5.1.

The second frequent problem of the I-O tables is that they are not consistent with the National Accounts. The NA data figures differ from the I-O table data mostly in the case of the foreign trade flows, but to a less extent it is so in the case of the gross outputs, wages, and stock accumulation. In the case of the foreign trade, the national accounts contain certain double accounting of the value of the materials to be processed, or of the goods to be repaired or reexported. Although in the I-O table (and output indicator) only the processing fee is accounted as export, in some of the I-O tables such reexport-like figures can be found, which is rather difficult to explain within the standard CGE modelling theoretical framework, so it is difficult to decide what to do with them (e.g. netting out, which however, for certain years may result in negative net export values due to the time lag between the import and export of such goods). This problem will be further discussed below, when we present the other problems of foreign trade data too.

The NA also showed some phenomena, which is difficult to take into account in CGE models. For example, in some of the national accounts the foreign sector had wage income (Hungary, Austria) (from which industry?) and SSC income, while it paid indirect taxes and received indirect subsidies (estimated tax content of the expenditures of the inbound tourists?).

Different I-O tables may account the imputed output of bank services (FISIM) differently.

Ideally they have to be allocated to actual users (estimating the interest margins by clients using reference interest rates for the domestic and foreign currencies), but this can not be done accurately using just the figures of the I-O table. Note, that in some countries (e.g. in Bulgaria) the Statistical Office just eliminated the FISIM by treating it as the own-consumption of the financial sector. However, it has certain problems, mainly the resulting apparent negative value added and the implicit assumption that the user-structure of the indirect services (interest margin) is the same as the users of the directly charged banking services.

In the GEM-E3 model the NPISHs are aggregated with the household sector (while in the I-O table they have separate columns in the final consumption). So in the model we had to treat the NPISHs as part of the household sector too. However, several coutries do not provide enough data for the NPISHs (Jellema et al [2004] ) so the missing data have to be estimated.

Foreign Trade Data: Apart from the above mentioned problem of value of the double-accounting of materials to be processed (or repaired or the like), the most common problem is the apparent inconsistency of the bilateral foreign trade data of the partner countries, which arises partly from the different valuation of the trade flows (the import is at c.i.f. parity while the exports are valuated at f.o.b. parity, so a so-called c.i.f.-f.o.b. correction is needed), but

also from different accounting methodology, classification and missing or false reporting. To see the resulting differences usually mirror-checks (i.e. one country’s reported import from an other country is the same as the other country’s reported export to the first country) are made by commodity groups (branches) and partner countries.

An other notable problem of these trade statistics, is that they usually contain the merchandise trade, but they do not include the services (for which data are rather unreliable and may be based on the foreign balance of payment statistics or specific surveys). A further general problem is that exports and imports with 'custom-free-zones' are not allocated to countries of origin or destination.

In the case of the printed publication of foreign trade statistics an additional special fundamental problem is that it contains only the large trade flows, which means that it is almost impossible to reconstruct the structure of the export and the import with small countries like Danemark or Portugal. The national account data for exports and imports by branch (in the new Cronos database) are too bad to be used (Eurostat is currently modifying them).

Consumption related data: The macrostatistical data for personal consumption by categories show the domestic consumption, while HBS data by definition contain the national consumption. Therefore, one has to estimate the tourists’ consumption by categories too.

Investment data: From the New Cronos database it was feasible to extract information regarding to gross fixed capital formation by branch, changes in inventories and acquisitions less disposals of valuables. In addition information on investments by product was available through the main aggregates of EUROSTAT. Combining this information with the investment structure derived from the projected tables the final investment by product transaction could be obtained.

The distinction of investments between the institutional sectors (Households, Firms, Government) was made by incorporating the information realised from the investment matrices as well as from the respective structure of Greece and United Kingdom.

Environmental data: Emission of air pollutants can be estimated quite accurately from the energy consumption data. However, in some cases uniform emission coefficients – derived from the EU15 practice - may result in rather unrealistic emission estimates. For example, in Hungary the extracted ‘coal’ is barely more than lignite, so its (per Joule or per ton) emission coefficients are far higher than those of any coal products within the EU15. Similarly the modellers have to bear in mind that in the Baltic countries fossil energy extraction practically means only pet and oil shale which are not typical product neither of the coal nor the oil industry.

Auxiliary data: When missing, the capital stocks have to be estimated from the capital incomes or amortization data (e.g in the case of Austria using exogenous type or industry-specific amortization rates)

Data for household groups: Inter-household transfers do not appear in the national accounts since by definition their aggregate value is zero. Therefore we can not adjust proportionately the HBS data for the macroeconomic total because it would set each

inter-household transfer figures to zero. Hence one has to find another way how to assess the representativity of the HBS’s figures for the inter-household transfers.

The usual problems in the HBSs are the need for reclassifying the special categories to scientific (national accounts) categories, the uneven representativity of the various income and expenditure categories, the underrepresentation of the poor, the rich, the overloaded, and the mobile (moving, commuting, immigrant, etc.) households, the annualization of the data (covering usually a period of one month or less), the matching of household incomes (and consumption) with members, the imputation of missing data (in-kind benefits, various capital incomes).

Apart these general data problems in chapter 8 we present the Hungarian example, showing how to cope with the missing data.