Trends in Global Value Chains

6. About the GVC datasets presented in this report

The OECD PYP ICIO dataset

GVC analysis relies heavily on inter-country input-output (ICIO) tables, a statistical framework that traces flows of goods and services between industries and across countries, covering both intermediate consumption and final uses.

The OECD has routinely published ICIO tables in current prices for several years, with annual time series stretching back to 1995 (https://oe.cd/icio). Current price values, however, conflate changes in real economic activity with movements in relative prices, making it difficult to assess how production networks have evolved in volume terms over time. Separating these two components requires tables in which each year's flows are expressed in the previous year's prices (PYP).

This publication marks the first release of the OECD PYP ICIO tables. Unlike constant price tables, which remain anchored to base year prices that become increasingly unrepresentative over time (especially in fast-changing sectors such as ICT and electronics), PYP tables form the natural input for chain-linking, whereby volume indices for TiVA and related GVC indicators are constructed by linking together successive one-year comparisons. This release covers the period 1996‑2024 at the same country and industry resolution as the current price series.

Finally, ICIO compilation draws on a wide range of data sources that, due to the constraints of statistical production, are not available with the desired timeliness and industry detail in all countries. The PYP tables for 2023 and 2024 are therefore based on a projection procedure whereby aggregate trade and GDP data for 2023 and 2024 are incorporated and gaps are filled using econometric techniques (see Annex A).

The OECD PYP ICIO tables are available at https://oe.cd/pyp-icio.

Sources of deflator data

The construction of PYP tables requires industry-level deflators for gross output and value added, as well as aggregate deflators for each category of final demand. Deflators are assembled from multiple sources, with priority assigned according to data quality and country coverage. For OECD members and a number of non-member economies, the primary source is the OECD Structural Analysis Database (STAN) (https://oe.cd/stan), which provides output and value-added deflators at a detailed industry level. For economies not covered by STAN, deflators are drawn from the United Nations National Accounts databases, supplemented where necessary by national sources and KLEMS productivity accounts. Deflators for the “rest of the world” aggregate are constructed as trade-weighted averages of the ten largest economies in that residual grouping, using output shares from the UN National Accounts.

Where industry-level deflators are not available for a given economy, missing values are imputed from the nearest available aggregate in the ISIC industry classification hierarchy. For example, a missing deflator for the chemicals industry may be approximated using the deflator for the chemicals and pharmaceuticals aggregate. Output deflators are similarly imputed from value-added deflators when only the latter is available. The assembled deflator dataset is then subject to a series of consistency checks to identify implausible year-on-year changes and internal inconsistencies between industry-level and aggregate deflators.

Construction procedure

Once complete annual deflator series are available, the construction of the PYP tables builds on the methodology developed by the WIOD project (Los et al., 2014[9]). The WIOD project was a widely respected ICIO-style statistical effort that served as a key venue for methodological development in the field of input-output data compilation.

As a first step, industry gross output and value added, as well as final use aggregates, are extracted from the current-price tables and deflated using the appropriate deflators to obtain time series of the corresponding aggregates in previous years’ prices. In any given year, the deflated aggregates yield the row and column totals of the PYP ICIO table to be estimated. The full PYP table is then obtained by applying an algorithm of the RAS family to identify the deflated ICIO that deviates the least from the current-price table while satisfying those marginal totals. Implicitly, this balancing procedure assigns a cell-specific deflator to each flow, which may differ from the aggregate deflators used for the marginal totals. The specific variant of the RAS algorithm used is GRAS, which naturally accommodates the negative elements that arise in particular in the changes-in-inventories columns (Junius and Oosterhaven, 2003[10]; Lenzen, Wood and Gallego, 2007[11]; Temurshoev, Miller and Bouwmeester, 2013[12]).

The Analytical AMNE database

Multinational enterprises (MNEs) play an important role in international trade, yet the challenge of collecting empirical evidence on their activities has meant we have not always had the full picture of key issues in trade and industrial policy such as the role of MNEs in GVCs and how firms combine trade and investment in their cross-border activities. A better understanding of MNEs allows for more informed debates about the role of these companies in the economy and in society more broadly.

To help address the data deficit, the OECD developed a comprehensive database on MNE activities across countries and industries. Linking statistics on the activities of multinational enterprises (AMNE) with the OECD Inter-country Input-Output (ICIO) tables, the Analytical AMNE database allows for the analysis of MNE activities in value-added terms, providing insights into production functions and sales activities of MNEs across the global economy.

The dataset considers three different types of firms, which we can illustrate based on the French motor vehicle industry:

• Foreign affiliates: Firms with more than 50% foreign ownership. To determine the parent country or country of control, the chain of control is traced from the affiliate back to the country of residence of the affiliate's ultimate controlling institutional unit (UCI) as outlined in OECD and Eurostat handbooks. For example, a subsidiary of a German motor vehicle manufacturer in France, in which the German firm holds a controlling stake.

• Domestic-owned MNEs or parent companies: Firms which are majority domestic-owned and themselves have more than 50% ownership stakes in affiliates abroad. For example, a French car manufacturer with majority-owned subsidiaries abroad.

• Other domestic-owned firms that are not recipients of significant international investment nor hold controlling stakes in any firm abroad. For example, a majority French-owned automotive parts supplier with no foreign affiliates.

In its 2026 edition, the database covers 81 countries plus a rest of the world aggregate and 41 industries for the period of 2000‑2023. The release includes harmonised and complete analytical AMNE statistics for gross output, value added, exports and imports of intermediate inputs for the three firm types, as well as ICIO tables split by domestic- and foreign-ownership. The database also includes a bilateral output matrix, which contains information on gross output by host economy, industry and parent economy. The breakdown of domestic-owned firms into domestic MNEs and non-MNEs is only available for the period of 2008‑2023, due to the limited availability of data on the activities of parent companies. All Analytical AMNE database components are consistent with the 2025 release of the OECD ICIO tables, with the exception of the ICIO tables for 2023 which are based on the projections of the OECD PYP ICIO dataset (see Annex A).

The database is available at https://oe.cd/gvc-mne.

Sources and methods

The construction of the Analytical AMNE database draws on foreign affiliate statistics from the OECD AMNE database, the Eurostat FATS dataset, OECD-Eurostat Trade by Enterprise Characteristics (TEC) Database, as well as from national sources. These are supplemented by Foreign Direct Investment (FDI) statistics from the IMF Direct Investment by Counterpart Economy Database and the OECD Benchmark Definition Foreign Direct Investment dataset. This data is used to estimate an initial set of harmonised AMNE statistics. In a second step, the initial estimates are linked to the OECD ICIO tables and used as constraints in a quadratic optimisation approach to estimate production functions and sales activities by firm type, resulting in ownership-split ICIO tables. Details on data sources and the construction of the Analytical AMNE database are available in Cadestin et al. (2018[5]) and Cai et al. (2023[13]).

The Trade in Value Added Indicators for Services by Mode of Supply (TiVA-MoS) database

Services play a central role in GVCs, both as direct outputs and as key intermediate inputs embodied in goods and other services. However, their international delivery is not fully captured by conventional trade statistics. In particular, standard Balance of Payments (BOP) statistics record only transactions between residents and non-residents and therefore do not fully reflect several important channels through which services are supplied internationally, such as sales by foreign-owned firms operating in host economies.

These different channels of international service delivery are commonly referred to as modes of supply. The General Agreement on Trade in Services (GATS) distinguishes four modes: (1) cross-border supply, where services are delivered from one country to another without requiring the movement of suppliers or consumers, broadly comparable to conventional merchandise trade; (2) consumption abroad, which occurs when consumers move to another country to consume services, for example in the case of tourism, education or health services; (3) commercial presence, which corresponds to services supplied through a locally established foreign affiliate; and (4) presence of natural persons, whereby individual service suppliers temporarily move to the country of the consumer to deliver the service.

For policymakers and analysts, a quantitative understanding of these modes of supply is essential to properly assess how countries and their trading partners engage in services trade, the role of foreign direct investment in services markets and the integration of services into GVCs.

The TiVA-MoS database

The Trade in Value Added Indicators for Services by Mode of Supply (TiVA‑MoS) database addresses this measurement challenge by integrating the GATS modes of supply into the OECD Trade in Value Added (TiVA) framework (Cai et al., 2025[14]). By doing so, it provides a consistent value‑added perspective on international trade in services that goes beyond gross flows recorded in traditional statistics and traditional TiVA measures that only cover cross-border trade in services (i.e. modes 1, 2 and 4 without distinguishing them).

The TiVA‑MoS 2026 vintage provides bilateral trade values by mode of supply for modes 1 and 4 (combined), 2 and 3, covering 19 service industries across 82 economies, plus a “rest of the world”, over the period 2000‑2023. In addition to gross trade values, the database includes a range of TiVA‑based indicators that shed light on the domestic and foreign services value added embodied in trade, across the different modes of supply. The dataset allows for a detailed analysis of trade in services along GVCs.

TiVA‑MoS indicators and methodological documentation are available at: https://oe.cd/tivamos.

Sources and methods underlying TiVA-MoS

While modes of supply are analytically important for understanding services trade, their direct identification following the GATS classification requires information that is not typically reported in national statistical systems. As a result, TiVA‑MoS indicators are estimates that combine and reconcile information from several sources, notably national accounts, balance of payments statistics and the OECD Inter‑Country Input‑Output (ICIO) framework.

More specifically, TiVA‑MoS indicators are derived by applying standard techniques of value‑added decomposition to input‑output variables computed from the Analytical Activities of Multinational Enterprises (AMNE) ICIO tables, which are split by firm ownership (domestically-owned versus foreign‑owned). These ownership‑split input‑output tables are combined with information from the AMNE bilateral output matrix, which further decomposes the gross output of foreign‑owned firms according to the country of the ultimate controlling entity.

Standard ICIO tables provide information on non‑resident expenditures, which allows the identification and estimation of Mode 2 (consumption abroad). The Analytical AMNE data, in turn, enable the identification of Mode 3 (commercial presence), by isolating the production of services foreign‑owned firms in host economies. To obtain a complete bilateral breakdown of Mode 3 by parent country, the foreign‑owned components of the split ICIO tables are further disaggregated using the bilateral output matrix under a proportionality assumption. Specifically, for a given host country and industry, the exports of foreign‑owned firms are allocated across parent countries in proportion to each country’s share in the foreign‑owned gross output of that industry.

Interpreting input-output-based GVC indicators: Key caveats and limitations

ICIO tables are powerful tools for analysing production linkages across countries and sectors, but they are not designed to capture every detail of firm-level supply chains. They operate at relatively aggregated sectoral levels, which can mask important differences across products, firms and technologies within the same industry. This lack of granularity is particularly relevant when analysing critical goods, strategic inputs or highly specialised supply chains, where vulnerabilities may arise at a product, firm or facility level that is not visible in aggregate data.

Input-output analysis also relies on proportionality assumptions. In most applications, industries are assumed to use the same input structure for all outputs and destination markets, and imported inputs are allocated proportionally across using industries. These assumptions are necessary to construct consistent international tables, but they may not always reflect the behaviour of individual firms or the specific sourcing patterns of particular products. As a result, GVC indicators should be understood as economy-wide and sector-level estimates rather than direct measures of individual supply chains.

Timeliness is another important consideration. Building high-quality ICIO tables requires the integration, balancing and reconciliation of many national and international data sources. This inevitably creates time lags. While this publication includes projections that improve timeliness, it should be understood that input-output based indicators are well suited to analysing structural patterns and medium-term trends. They are less appropriate for real-time monitoring during fast-moving disruptions. For crisis management or short-term supply chain monitoring, they need to be complemented by more timely sources.

There are also measurement challenges related to services, multinational enterprises and intangible assets. Services are increasingly embedded in goods, supplied digitally, bundled with products, or provided through affiliates abroad. Multinational enterprises organise production through complex ownership structures, intra-firm transactions and intangible assets that are difficult to capture fully in national accounts and input-output systems. The new OECD datasets make significant progress in addressing these challenges, but they should still be interpreted as part of a broader evidence base rather than as a complete representation of all international business activity.

These limitations do not reduce the value of the GVC indicators presented in this publication. On the contrary, they clarify how the indicators should be used. Input-output data are especially valuable for identifying broad structural linkages, quantifying value-added relationships, comparing countries and sectors, and assessing how trade, investment and services are connected across economies. They are less suitable for drawing conclusions about specific products, individual firms or immediate operational risks without additional information. Used appropriately, they provide an indispensable foundation for evidence-based policymaking.