Statistical Quality of the Trade in Value Added (TiVA) Database

It is important to stress that the indicators presented in the TiVA database are estimates. Official gross statistics on international trade produced by national statistics institutions result in inconsistent figures for total global exports and total global imports; inconsistencies which are magnified when bilateral partner country positions are considered. The global input-output tables from which TiVA indicators are derived, necessarily eliminate these inconsistencies, such as those that reflect different national treatments of re-exports and transit trade (e.g. through hubs such as the Netherlands and Hong Kong), to achieve a coherent picture of global trade. For the countries for which data is presented, total exports and imports are consistent with official national accounts estimates, however bilateral trade positions presented in the database (based on gross flows) and those published by national statistics institutions may differ. Work is on-going within the international statistics community to achieve coherence in international trade flows, particularly in the area of trade in services, where significant differences exist when comparing national statistics.

Some assumptions are necessarily used in creating global input-output tables and the Trade in Value-Added indicators.

•  Production assumption

Indicators created via input-output techniques are limited by the degree of industry disaggregation provided by the tables. The national input-output tables used by the OECD are based on a harmonised set of 37 industries.  Any given indicator therefore assumes that all consumers of a given industry’s output purchase exactly the same shares of products produced by all of the firms allocated to that industry. This boils down in practice, (but is not the same thing) to assuming that there exists only one single production technique for all of the firms (and all of the products) in the industry grouping. We know that this is not true and that different firms, even those producing the same products, will have different production techniques, and, so, technical coefficients, and we also know that different firms produce different products and that these products will be destined for different types of consumers and markets. Of chief concern in this respect is the evidence that points to exports having very different coefficients to goods and services produced for domestic markets, particularly when the exports (typically intermediate) are produced by foreign owned affiliates in a global value chain. Because exporting firms are generally more integrated into value-added chains they will typically have higher foreign content ratios, particularly when they are foreign owned, as such the estimates provided in this release should be considered as prudent. Generally they will point to lower shares of foreign content than might be recorded if more detailed input-output tables were available, with consequences for all other indicators presented. One important innovation in the indicators presented here is to use specially constructed input-output tables for China that differentiate between processing firms, other exporting firms, and those that produce goods and services only for domestic consumption. Because of China's importance to trade this significantly improves the quality of the results.  

•  Proportionality assumption

At the national level the quantity and quality of information available to allocate specific imports to using industries varies. Where information is not available, countries and indeed practitioners necessarily use the 'Proportionality assumption'. This generally means that that for a given product one assumes that the proportion of intermediates that an industry purchases from abroad is equal to the ratio of imports to total domestic demand in that product.  For those countries where it has been necessary to use this assumption (and indeed others) refinements have been introduced by using trade data that differentiates between those imported goods in a given product grouping that are intermediate and those that are final domestic demand. On its own, this assumption is not expected to have a significant impact on total economy estimates but it will affect the import content of various industries, and so, by extension, bilateral trade estimates of trade in value-added. But the results are not expected to be biased in any particular direction.

Moreover, in addition, there are challenges with the underlying official national statistics on international trade which require specific mention.

•  Dealing with internationally inconsistent official trade statistics

It is a well known fact that the international trade statistics produced by national authorities are not globally consistent: total global gross exports do not equal total global gross imports. This inconsistency is larger when bilateral trade flows are considered and larger still when those flows are looked at on a detailed product level. Even if total gross exports from country A equal those imported by country B, there may still be differences when these flows are looked at on a product by product level. The global input-output tables used to produce Trade in Value-Added indicators necessarily resolve all of these inconsistencies. Total exports and total imports of a given country will be consistent with totals recorded in their official National Accounts statistics but the balancing process will necessarily introduce coherence adjustments to bilateral trade flows that will lead to differences between official recorded bilateral gross trade flows and those reflected within the input-output table. The results of this balancing will form the basis of dialogue with national statistics institutions as part of on-going international efforts to reconcile international trade statistics; particularly in the area of trade in services where official statistics on bilateral trade data are notoriously weak. The balancing does not introduce any directional or structural bias but, clearly, the quality of TIVA results will be significantly improved as global inconsistencies reduce.  This is not expected to have a significant impact on overall foreign content estimates broken down by industry but bilateral trade in value-added estimates will be affected.

Detailed information on the methodology and statistical developments envisaged to improve the quality of results are provided in the accompanying paper issued with this release: Trade in Value Added: Concepts, Methodologies and Challenges (Joint OECD-WTO note).