This chapter discusses the three most common extension criteria for constructing an extended supply and use table: exporter or trading status, ownership or group affiliation, and size. The chapter explains why each criterion is relevant, relating to academic literature. Subsequently, it explains in detail how to define the various categories of enterprises and how to classify the enterprises in practise, pointing out the various data sources. This is illustrated with several country examples. The chapter also discusses the different types of units that can be used in the compilation process.
Handbook on Extended Supply and Use Tables and Extended Input‑Output Tables
Abstract
The first step in constructing an extended supply and use table (ESUT) is to choose the source(s) of enterprise heterogeneity for breaking down industries (and product categories) in the supply and use framework. Any such source of heterogeneity is also referred to as an extension criterion. So far, the construction of an ESUT has mainly focused on the following three major enterprise characteristics as extension criteria: 1) exporter or trading status; 2) ownership status or group affiliation; and 3) enterprise size.
The industry breakdown in an ESUT may be based on a single extension criterion or a combination of several extension criteria. Most of the ESUTs that have been produced to date are for a single extension criterion, e.g. exporter status for the People’s Republic of China (hereafter “China”) (Koopman, Wang and Wei, 2012[1]), ownership status for the United States (Fetzer et al., 2023[2]), size for the Netherlands (Chong et al., 2019[3]). But there are also examples of ESUTs compiled combining these three criteria for all industries (for the Nordic countries; (Statistics Denmark and OECD, 2017[4])) or with the chosen criterion depending on industry characteristics (for Japan; (Hagino and Kim, 2021[5])).
This chapter provides a detailed description of how enterprise categories may be defined for each of these three criteria and which data can be used to classify enterprises in the defined categories. As the definition of categories is closely related to the available data sources, both are discussed together for each extension criterion. Regarding data sources, the focus here is only on those used for the classification of enterprises into the previously defined categories, while Chapter 3 will discuss those used in the construction of an ESUT. Naturally, some data sources will serve for both – enterprise classification and ESUT construction. For example, data on exports of goods are used for determining the exporter status of enterprises and for constructing an ESUT.
In this context, it must be noted that the classification of enterprises into heterogeneous categories generally concerns legal units (some national statistical offices [NSOs] do this for enterprises), while the construction of supply and use tables (SUTs) is based on statistical units as the producers of goods and services. Establishments (local kind-of-activity units) are the preferred statistical units for SUT construction to limit secondary output. When it comes to ESUT compilation, these establishments must be linked to the legal units (or enterprises) classified into heterogeneous categories according to the chosen extension criterion. Normally, such a link is included in the business register. But it is up to the compiler to decide whether all establishments that are part of a legal unit (or enterprise) should be classified in the same category. This is not an issue for countries where the legal unit (or enterprise) is the statistical unit for the construction of SUTs.
Finally, the discussion in this chapter focuses on the situation where compilers have access to the business register and the relevant enterprise-level data sources that allow classifying enterprises. Some compilers, in particular researchers outside NSOs, may not have access to such enterprise-level data sources. This does not rule out the construction of an ESUT, but it implies that these compilers have to rely on more aggregate data. Potential aggregate data sources are mentioned below for each extension criterion, and the consequences of their use – for both the definition of enterprise categories and the ESUT construction process – are briefly discussed.
The exporter status is a source of enterprise heterogeneity that has been investigated extensively in the academic literature. Exporters have been found to be more skill-intensive, more capital-intensive and more productive than other enterprises (see, for example, Melitz (2003[6]); Bernard et al. (2012[7])), and it has been shown that they rely proportionally more on foreign inputs (Bas, 2009[8]). This heterogeneity in production technologies is the motivation for breaking down industries according to the exporter status of enterprises. The resulting ESUTs reveal the specific input structures for the production of exports, which allows, among other things, to improve estimates of domestic and foreign value added in exports (see de Gortari (2019[9]) and Michel, Hambÿe and Hertveldt (2023[10])) .
Defining categories of enterprises
Although the breakdown of industries according to the exporter status of enterprises appears to be conceptually relatively straightforward, several issues need to be addressed in practice. Beyond special trade regimes (see below), the standard approach is to divide enterprises in each industry into exporters and non-exporters. However, the heterogeneity between exporters and non-exporters is not necessarily clear-cut. According to enterprise-level research, it is mostly the major exporters that are different in terms of production technologies (Bernard et al., 2012[7]), while minor exporters tend to be similar to the bulk of non-exporters. Therefore, a threshold may be imposed when defining exporter status to distinguish export-oriented enterprises from domestic-oriented ones – i.e. enterprises that mainly serve the domestic market – to increase the homogeneity of within-industry categories of enterprises in the ESUT. The threshold can either be relative – a minimum percentage share of exports in an enterprise’s output – or absolute – a minimum value of total enterprise-level exports. A double threshold in both relative and absolute terms is a powerful way of selecting big enterprises that are major exporters. The absolute threshold will discard small exporting enterprises, and the relative threshold will subsequently discard big enterprises that only export a minor share of their output.
For the optimal choice, compilers may want to test the sensitivity of the categorisation of enterprises (and potentially also the sensitivity of analytical results) with respect to different types of thresholds and different threshold levels (both in relative and in absolute terms). While the application of a threshold is likely to increase within-category homogeneity, it also implies that the category of export-oriented enterprises does not account for all exports, i.e. domestic-oriented enterprises account for at least a small share of exports.
Existing ESUTs for Mexico provide an example of a distinction between exporters and non-exporters (Ostolaza-Berman and Téran Vargas, 2019[11]). Examples of the application of a threshold can be found in work on ESUTs and EIOTs for Belgium (Michel, Hambÿe and Hertveldt, 2023[10]), relative threshold) and Denmark (Nilsson, Rørmose Jensen and Holst Jensen, 2018[12]), absolute threshold). In addition to these thresholds on the value of exports, compilers of ESUTs may also consider alternative export-related sources of heterogeneity, e.g. separating out enterprises that export to multiple destinations or enterprises that export to more distant destinations rather than only neighbouring countries.
With regard to prior investigations of export-related enterprise heterogeneity in the academic literature, it appears as natural to take the exporter status as the source of heterogeneity. Nevertheless, focusing on exporters that also import, so-called two-way traders, may represent a further means of reducing within-industry heterogeneity (Ahmad, 2023[13]). In more general terms, compilers of ESUTs may categorise enterprises according to their trading status, i.e. non-traders, exporters, importers and two-way traders, then determine which categories to separate out and whether to impose thresholds on exports and imports at the level of the enterprise.
Data for classifying enterprises
Ideally, the classification of enterprises by exporter or trading status is based on an individual identification of exporters, importers or two-way traders through enterprise-level export and import data. Depending on data availability and quality, this may be restricted to exports and imports of goods – merchandise trade – or encompass exports and imports of both goods and services. Enterprise-level data on merchandise trade typically come from customs data (see United Nations (2011[14]) or an equivalent source (e.g. Intrastat and Extrastat declarations for European Union [EU] countries). Customs data usually contain monetary values of merchandise trade flows by enterprise, and also by product category and partner country. They may also comprise transaction codes through which enterprises operating under special trade regimes can be identified (see below). Enterprise-level data on exports and imports of services are mostly obtained through surveys for balance of payments statistics. Imposing thresholds requires additional enterprise-level data on total sales or turnover, which may be obtained from enterprises’ annual accounts, tax records or structural business surveys.
Such data on exports and imports are generally collected for legal units or enterprises. Unless the construction of SUTs is based on legal units or enterprises, a conversion is needed to attribute these exports and imports to establishments (local kind-of-activity units). In work for the United States, this conversion is based on establishment-level employment data (Fetzer et al., 2023[2]). The compiler must then decide whether all establishments that belong to a legal unit (or enterprise) should be classified in the same category, e.g. whether to consider all establishments of an exporting enterprise as exporters.
Moreover, for exports of goods, the identification of exporters or export-oriented enterprises is generally based on direct exports, i.e. cross-border transactions for which the enterprise is itself responsible. However, enterprises may also export goods indirectly through distributors or intermediaries (wholesalers), which do not transform the goods and only apply a margin on their sales of these goods. Such indirect exports can represent a substantial part of total trade, in particular in small open economies. It is, for example, estimated that approximately 30% of Danish exports of goods go through wholesalers (Nilsson, Rørmose Jensen and Holst Jensen, 2018[12]). In practice, attributing goods that are exported through wholesalers to the original producers is a challenge because it requires additional information on domestic transactions between producers and wholesalers. Value-added tax (VAT) transaction datasets that record all domestic transactions subject to VAT (see, for example, Dhyne, Magerman and Rubínová (2015[15])) appear to be the most promising data source for this purpose. However, even with such data, major methodological challenges remain to be solved, e.g. how to determine whether goods delivered by producers to wholesalers correspond to the goods exported by these wholesalers. This field has not yet been explored in depth and further work is needed.
The issue of trade through wholesalers also arises on the import side. But then the issue of imports distributed by domestic wholesalers is already a challenge when constructing regular SUTs and in particular for the compilation of the import use table. In theory, information on indirect imports through wholesalers used in the construction of regular SUTs should make it possible to estimate indirect import flows to identify all importers and two-way traders.
When ESUT compilers do not have enterprise-level data on exports and imports available to them, they must rely on other data sources. In some countries, structural business statistics contain information on the exporter status of enterprises, and this information can be merged into the business register for the classification of enterprises into the categories of exporters and non-exporters. Of course, this requires access to the business register. Moreover, with such data, it is not possible to apply thresholds.
Finally, for the construction of ESUTs by exporter or trading status, compilers may also use more aggregate data. In particular, the industry-level data on trade by enterprise characteristics (TEC; Box 2.1) published by Eurostat and the OECD provide a second-best alternative, at least for the disaggregation of industry totals. Compilers must be aware that the use of such data does not leave room for choosing a definition of exporter or trading status or for applying thresholds.
Trade by enterprise characteristics (TEC) data go beyond conventional international trade data. Rather than mapping trade flows between countries by types of goods or services, TEC data put the focus on the enterprises that export and import, reporting trade flows broken down by different categories of enterprises. Such data provide additional perspectives for the analysis of international trade, and they may be used in the construction of an extended supply and use table (ESUT).
The OECD publishes TEC data for its member countries and a number of non-member countries (see: www.oecd.org/en/data/datasets/trade-by-enterprise-characteristics-tec.html), and also Eurostat publishes data for EU member countries with a distinction between international trade in goods by enterprise characteristics (TEC) (Eurostat[16]) and service trade by enterprise characteristics (STEC) (Eurostat[17]). These are compilations of TEC data produced by NSOs. For the purposes of producing enterprise-level breakdowns of export and import data, NSOs combine enterprise-level trade data with other enterprise-level data sources that allow categorising enterprises.
In practice, the data on exports and imports of trading enterprises are cross-tabulated by the industry of the enterprises and by several enterprise characteristics including size, ownership/group affiliation and trading status. The data published by the OECD and Eurostat also provide numbers of trading enterprises by industry and enterprise characteristics. When using these TEC data in constructing an ESUT, compilers must be aware of a series of caveats among which the following deserve to be specifically mentioned: they have to take the definition of the enterprise characteristics as given (e.g. the enterprise size classes); the underlying coverage of enterprises may differ between the TEC data and the regular SUT that the compilers want to disaggregate; the TEC data may comprise transactions that are not included in the national accounts and vice versa; and the wholesale industry often accounts for a much larger share of total trade in TEC data than in regular SUTs.
Special trade regimes
The earliest efforts to separate out groups of enterprises according to their trading status in SUTs and input-output tables were related to special trade regimes: processing traders for China (Koopman, Wang and Wei, 2012[1]), enterprises operating under global manufacturing programmes for Mexico (De La Cruz et al., 2011[18]) and enterprises operating in free trade zones in Costa Rica (Saborío, 2015[19]). In all these cases, the aim of the special trade regime is to grant a tariff exemption to provide an incentive for foreign enterprises to locate the assembly stage of their production process in the country, with inputs imported and output sold on foreign markets. This implies that there are differences in production processes and especially input structures between enterprises participating in the special trade regime and other enterprises, due to the design of the special trade regime where participating enterprises purchase (almost) no inputs locally and sell their production abroad. The non‑participating enterprises do purchase and sell much locally. The resulting within-industry enterprise heterogeneity provides a strong incentive for the construction of ESUTs. Work on disaggregating industries in SUTs to isolate enterprises participating in these regimes is focused on manufacturing industries and generally based on customs data or specific data collected as part of the operation of the regime. A potential issue is that trade flows under these special trade regimes mostly do not imply a change in ownership and are therefore not recorded in regular SUTs following the rules introduced in the System of National Accounts 2008.
Ownership links and group affiliation represent another major source of within-industry enterprise heterogeneity. A vast and long-standing empirical literature based on enterprise-level data has emphasised that production processes of foreign affiliates and enterprises that are part of a multinational enterprise (MNE) group tend to be different from those of other enterprises (Brainard, 1997[20]; Bernard et al., 2012[7]; Yeaple, 2013[21]; Antràs and Yeaple, 2014[22]). In addition, it has been estimated that the export activity of foreign affiliates accounts for two-thirds or more of global trade (UNCTAD, 2013[23]; Miroudot and Rigo, 2022[24]). This illustrates that MNE groups have largely contributed to global integration over the past decades and continue to do so today through exports and local sales of their foreign affiliates.
From the perspective of the national accounts, there is friction between the compilation of accounts for individual countries based on residency rules imposed by the System of National Accounts on the one hand, and global production arrangements of MNE groups, on the other (Moulton and van de Ven, 2023[25]). The most striking evidence for this friction is the Irish 2015 real gross domestic product growth rate of more than 25%, which was driven by the within enterprise group relocation of intellectual property rights (FitzGerald, 2023[26]).
ESUTs with an industry breakdown by group affiliation are an important means of getting a better and more detailed grasp of the nature and scope of MNE groups’ activities, within an individual country and in global value chains. Such an extension is based on a categorisation of enterprises within industries according to their affiliation to an enterprise group.
Enterprise groups
The starting point for this categorisation is the definition of an enterprise group, which is an association of enterprises bound together by legal and/or financial links and controlled by a group head. An enterprise group may be domestic – when all linked enterprises are residents of the same country – or multinational – with links to enterprises in other countries.
From a theoretical point of view, control determines which enterprises belong to a group. Control implies the power to set the general policy of an enterprise and/or to appoint the majority of its directors. It may be exerted by an individual, by an enterprise or by a public authority. For an ESUT, the focus is on enterprises controlling other enterprises, so-called corporate control, while control by individuals or public authorities is generally not considered. Enterprise A is said to control enterprise B if A holds, directly or indirectly, more than half of the shareholders’ voting power in B. Furthermore, enterprise A exerts indirect control over enterprise B if A controls one or more enterprises (C, D, etc.) that control B. Finally, in circumstances where voting rights are widely dispersed, enterprise A may control enterprise B through effective minority control, i.e. without holding more than half of the shareholders’ voting power.
In practice, delineating an enterprise group based on the concept of control proves challenging because information about the voting power that allows exerting control is mostly difficult to come by or unavailable. Therefore, NSOs generally rely on ownership as a proxy for control when it comes to delineating enterprise groups. They determine ownership based on information about participation rates, i.e. shares in an enterprise’s capital, since this information is easier to come by. This section first describes ownership-based definitions of enterprise groups and group affiliation before turning to a description of potential data sources.
As regards terminology, enterprises that hold a participation in another enterprise are referred to as parent enterprises, while enterprises whose capital is held by another enterprise are referred to as affiliates. It is common to consider only participation rates of 10% or more as the sign of a lasting interest (foreign direct investment [FDI] threshold). Among affiliates, a distinction is made between subsidiaries and associates. An enterprise is a subsidiary if another (parent) enterprise owns more than 50% of its capital; it is an associate if another (parent) enterprise holds between 10% and 50% of its capital. Participation rates above 50% are taken to reflect control, and, therefore, an enterprise group is made up of a parent enterprise that is not controlled by another enterprise and all the direct and indirect subsidiaries of that parent enterprise. Associates are not considered to be part of the enterprise group. Finally, a joint venture is a special case where two or more parent enterprises have a minority participation of equal size (50% for two parents, 33% for three parents, etc.) in an affiliate.
The global group head is the parent enterprise at the top of the enterprise group that is not controlled by another enterprise. This implies that the underlying group structure is assumed to be hierarchical. The notion of global group head is similar to the notion of global decision centre or ultimate controlling institutional unit in the Foreign Affiliates Statistics (FATS; (Eurostat, 2012[27])) or in the Activities of Multinational Enterprises (AMNE; (Cadestin et al., 2018[28])). This is the enterprise that takes global decisions for the enterprise group as a whole. Global group head and global decision centre may, however, differ for large or complex enterprise groups, in particular when a natural person, a family holding, an enterprise located in a tax haven, a holding enterprise or an equity fund is at the top of the enterprise group.
Defining categories of enterprises
Four categories of home country enterprises can be distinguished based on affiliation to an enterprise group:
1. domestic stand-alone enterprises
2. enterprises that are part of a domestic enterprise group
3. enterprises that are part of a domestically controlled MNE group
4. enterprises that are part of a foreign-controlled MNE group.
These four ownership categories are discussed below, with examples for the latter three categories.
Domestic stand-alone enterprises
Domestic stand-alone enterprises are not part of an enterprise group, i.e. they are neither subsidiaries nor parent enterprises.
Enterprises that are part of a domestic enterprise group
Enterprises that are part of a domestic enterprise group are either the global group head or subsidiaries of an enterprise group which is made up exclusively of enterprises located in the home country (Figure 2.1).
As Figure 2.1 illustrates, enterprise A is the global group head and enterprises B, C, D and E are also part of the group as subsidiaries. They are all located in country 1, which is the home country. The perimeter of the group is indicated by the oval in the figure.
Enterprise E is an indirectly controlled subsidiary of enterprise A. In this example, A exerts its control through two subsidiaries, C and D, with minority participations in E that sum to more than 50%. Enterprise F is an associate since enterprise A and its subsidiaries hold less than 50% of its shares; it therefore does not belong to the enterprise group (it is outside the oval in the figure).
Enterprises that are part of a domestically controlled multinational enterprise group
Enterprises that are part of a domestically controlled MNE group are home country enterprises that belong to an enterprise group with a home country enterprise as global group head and at least one subsidiary in another country.
This is illustrated in Figure 2.2. Here, the perimeter of the group is again given by the oval. It is made up of enterprises located in three countries separated by dashed lines. Enterprises in each country have a different colour. Enterprise A is the global group head located in country 1 and all other enterprises shown in the figure (B to G) are subsidiaries of A. Of these, enterprises B, E and F are located in country 1 (domestic subsidiaries of A), while enterprises C and D are located in country 2 and enterprise G in country 3 (foreign subsidiaries).
From the perspective of country 1, this MNE group is domestically controlled since the global group head (enterprise A) is located in this country. Hence, enterprises A, B, E and F in country 1 are part of the category of enterprises that belong to a domestically controlled MNE group.
Enterprises that are part of a foreign-controlled multinational enterprise group
Enterprises that are part of a foreign-controlled MNE group are home country enterprises that belong to an enterprise group with a global group head located abroad. According to the terminology defined above, they are foreign subsidiaries, but in prior work on ESUTs, they are mostly referred to as foreign affiliates.
This is also illustrated in Figure 2.2. From the perspective of countries 2 and 3, this MNE group is foreign-controlled since its global group head, enterprise A, is located abroad, in country 1. Hence, enterprises C and D in country 2 are part of the category of enterprises that belong to a foreign-controlled MNE group. The same holds for enterprise G in country 3.
Joint ventures require specific attention. While a joint venture should not be considered as a domestic stand‑alone enterprise, the compiler must decide whether to classify the enterprise as part of a domestic enterprise group, part of a domestically controlled MNE group or part of a foreign-controlled MNE group. This decision should be based on the location of the parent enterprises and either on extra information on the joint venture collected by the NSO or on a pre-specified rule.
Data for classifying enterprises
Ideally, the classification of enterprises according to group affiliation is based on information on group structures from a groups register. Compilers may either use an existing groups register or build a groups register themselves. Since the construction of a groups register is a work-intensive undertaking, the use of an existing groups register reduces the workload for constructing an ESUT substantially.
In many countries, the NSO produces a groups register on a regular basis. In some cases, group structures are identified directly through specific surveys, but they are mostly derived from data on participations (ownership data). As mentioned earlier, the use of ownership data is generally motivated by the lack of data on control. In practice, group structures are derived from data on participations through an algorithm that identifies all enterprises that belong to a group. The algorithm is based on a 50% plus one share participation threshold so as to proxy for control. It must identify group heads and attribute all subsidiaries to their group head taking into account indirect control.
NSOs use several types of sources for data on participations in the construction of a groups register. They should ideally identify the parent, the affiliate, their respective countries of residence and the participation rate. These data sources include administrative data such as enterprise accounts and tax records, in which enterprises are required to report information about their shareholders and their affiliates, as well as surveys on shareholder structures and affiliates, mostly conducted by central banks or chambers of commerce. In France, for example, the NSO constructs its groups register LiFi, which stands for “Liaisons Financières”, mainly based on shareholder information from tax authorities and the Banque de France. This information is used as input into an algorithm that identifies enterprise groups.
A further example is provided by the groups register for Belgium, which relies heavily on information that enterprises have to report as part of their annual accounts (balance sheet and profit and loss account). In three separate sections of the appendix of these accounts, enterprises list their (domestic and foreign) affiliates with an indication of the rate of participation; provide information on their shareholder structure; and indicate whether they are part of a group that establishes consolidated accounts and reports information on the consolidating enterprise and/or their parent enterprise. This is complemented with information from a group structure survey, which is conducted by the Belgian National Bank and is the basis for selecting samples for FDI and FATS surveys. This group structure survey identifies foreign ownership links for large Belgian enterprises. Reporting enterprises are asked to provide the name and country of residence of all enterprises in the group, resident or non-resident, and the direct participation rate between any two enterprises in the group.
When there is no groups register or when compilers do not have access to the NSO’s groups register, they must determine the group affiliation of enterprises themselves. This is work-intensive, and the quality of the result largely depends on the underlying ownership data. With access to the type of administrative data sources described above, compilers can construct a groups register by applying the appropriate algorithm. Otherwise, they may use survey data from different sources, some of which is also used by NSOs for the construction of groups registers.
Dedicated surveys on enterprise groups and their affiliates are indeed conducted in most OECD member countries. For the United States, the Bureau of Economic Analysis collects data on participations of and in US enterprises through surveys on inward and outward FDI. Likewise, national central banks and NSOs in EU member states gather information on participations through surveys of FDI and through FATS. Data from these surveys are useful not only for identifying group structures and classifying enterprises but also for breaking down industries in ESUT construction.
In addition to dedicated surveys, some NSOs have added specific questions to business surveys in other fields (e.g. structural business statistics, research and development, innovation, exports of services, information and communication technologies) to collect information on group affiliation. Although such information is not necessarily comprehensive, it may complement other data sources and allow classifying large enterprises. Besides, results from profiling carried out by NSOs are another source of information on group affiliation. In manual profiling, the global decision centre and the global group head as well as their location must be identified.
The identification of domestic affiliates or subsidiaries is often a blind spot when it comes to determining group structures and group affiliation, because data on domestic participation links tends to be scarce. This holds true in particular for the determination of group structures based on FDI surveys, and it may even be an issue in the construction of a groups register by the NSO. Therefore, some countries, for example France, have added a specific question on domestic subsidiaries to their FATS surveys to identify the full (domestic and cross-border) scope of enterprise groups.
Another limitation of groups registers and other data on ownership links for individual countries is that, in general, they only cover direct foreign ownership links. Data sources that report indirect foreign ownership links are the exception. In terms of the graphical example of a MNE group, data sources for country 1 are unlikely to cover the ownership link between enterprises C and D in country 2. Hence, the algorithm will not identify the full scope of the group. By the same token, data sources for country 2 will normally fail to identify the link between enterprise A and enterprise B in country 1. In that case, the global group head – enterprise A – is not identified by the algorithm. Such incomplete identification of group structures may lead to errors in the classification of enterprises according to their group affiliation.
The ideal solution for addressing this limitation in data on group structures for individual countries would be to construct a groups register at a multi-country or even global scale. Several international organisations have taken initiatives in this field.
Eurostat compiles the Euro Groups Register (EGR), which is a statistical business register of MNE groups operating in EU and European Free Trade Association countries. It is based on group structure information delivered by the NSOs of these countries, which is sometimes not publicly available. In the process of compiling the EGR, Eurostat validates and consolidates these data on group structures from individual countries; the EGR is then complemented by commercial data sources and other open data sources available online to depict the complete group structure of the MNE groups. Although it is not global but limited to the MNE groups with operations in the EU and European Free Trade Association countries in geographical scope, the EGR can provide valuable input for the classification of enterprises according to their group affiliation. Information from the EGR has been used together with national data for the classification of enterprises in the most recent compilation of ownership-extended SUTs for Belgium (Hambÿe, Michel and Trachez, 2023[29]).
At the global scale, the OECD and the United Nations Statistics Division have developed the Multinational Enterprise Information Platform (see Pilgrim and Ang (2024[30]). It contains a physical and a digital register of the world’s largest MNE groups with their affiliates and subsidiaries, aiming for a global coverage. They are built exclusively from publicly available data. The data used include, among others, information from companies’ annual reports and from the Global Legal Entity Identifier Foundation. To date, the register covers the 500 largest multinational groups. This limited coverage restricts the use of the register for ESUT compilation. It may, however, prove to be valuable input for confirming, correcting or complementing available national information on group affiliation.
Besides, there are also private initiatives on compiling group structure information at a multi-country or global scale. The most prominent example is ORBIS, a global database of company accounts, put together and published initially by the Bureau van Dijk, and recently taken over by Moody’s. The business information compiled by Dun & Bradstreet is another example of such data. These databases have been used in academic work on multinational groups and enterprise heterogeneity and represent a potential source for classifying enterprises in the process of ESUT construction. Data from ORBIS have been used in the construction of the OECD’s Analytical AMNE Database (Cadestin et al., 2018[28]).
For the construction of an ESUT with an industry breakdown by group affiliation, compilers must not necessarily strive for a full breakdown into the four categories of group affiliation shown above. Even with a less detailed categorisation, an ownership-extended SUT can provide valuable information on within-industry heterogeneity and enhance analytical possibilities. For example, categories 1, 2 and 3 may be merged into a single category of “domestic enterprises”, either due to data constraints or because the aim is limited to distinguishing between “domestic enterprises” and “foreign affiliates”, i.e. enterprises that belong to a foreign-controlled multinational group. Such a breakdown of industries into “domestic enterprises” and “foreign affiliates” can be found in the multi-country tables in the OECD’s Analytical AMNE Database (Cadestin et al., 2018[28]). In a similar vein, industries have been broken down into the three categories of “non-MNEs” or “domestic enterprises”, “domestic MNEs”, and “foreign MNEs” in the ESUT work for the United States (Fetzer et al., 2023[2]) and for Belgium (Hambÿe, Michel and Trachez, 2023[29]). This amounts to merging categories 1 (domestic stand-alone enterprises) and 2 (enterprises that are part of a domestic group) into a single category referred to as “non-MNEs” or “domestic enterprises”.
Furthermore, ESUT compilers often face data restrictions that either directly impose a definition of enterprise categories upon their work or make it necessary to adopt an alternative approach to enterprise classification. For example, the data available to the compiler may only record whether an enterprise has an FDI link, i.e. participation of more than 10% in a foreign enterprise, without mention of the exact participation rate. Such data are limited to direct ownership links and no distinction is made between subsidiaries and associates. Hence, it does not allow for identifying group structures. It may nevertheless be used for classifying enterprises in the process of ESUT construction, and tables with an industry breakdown based on these data will likely yield valuable insights on enterprises with an FDI link. Besides, it would be of interest to explore the impact of different participation rate thresholds. Instead of the 50% participation threshold to proxy for control, compilers may choose a lower participation rate threshold, e.g. the 10% FDI threshold. When applying the usual algorithm, this would lead to the identification of enterprise networks based on FDI links rather than groups where enterprises are bound together by control links. Comparing extended tables based on these two alternative thresholds could provide insights into whether it is investment or control links that account for differences in technology.
Groups registers and alternative data sources on ownership links generally refer to legal units (or enterprises). For countries in which the legal unit (or enterprise) is used as the statistical unit for SUT construction, the classification by group affiliation is straightforward. With establishments (local kind-of-activity units) as the statistical unit, a link between legal units (or enterprises) and establishments is again required. Normally, all establishments that belong to a legal unit (or enterprise) should be classified in the same category of group affiliation.
Finally, when ESUT compilers do not have access to individual enterprise-level data on ownership or group affiliation they may rely on more aggregate industry-level data, e.g. from the TEC or FATS databases. However, in that case, they must take the definition of categories and enterprise classification as given. The prime example of ESUT construction based on industry-level data sources is the OECD’s Analytical AMNE Database, which combines the TEC database and several sources of FATS-type data for breaking down industries in the OECD’s Inter-Country Input-Output Tables (Cadestin et al., 2018[28]).
Enterprise size is the third major source of within-industry enterprise heterogeneity that has been considered as a criterion for the construction of an ESUT. This reflects the classical view that large enterprises are different from small and medium-sized enterprises (SMEs) in terms of production technology. In other words, the scale of operations is assumed to influence cost and input structures. Besides, SMEs are traditionally in the focus of policy makers given that they represent the vast majority of enterprises active in a country and account for a large share of a country’s total employment. Moreover, in many circumstances, enterprise size matters for state aid eligibility. There has been growing interest in recent years in the interaction between SMEs and large enterprises within domestic and global value chains (ADB, 2015[31]; OECD, 2023[32]).
ESUTs with an industry breakdown by enterprise size class can provide valuable insights into the economy-wide importance of SMEs and large enterprises, and these tables can serve as input for analyses of the links between these categories of enterprises. In the first instance, such a breakdown requires a definition of enterprise size classes. In general, these size classes are based on enterprise-level thresholds for one or more variables. Employment stands out as the most widely used variable for setting size class thresholds. However, size classes may also be defined based on other enterprise-level variables such as turnover, total assets or the balance sheet total.
Defining categories of enterprises
The decision on the number of enterprise size classes is up to the compilers. Traditionally, three enterprise size classes are defined: large, medium-sized and small enterprises. Depending on the aims of the exercise and data availability, the number of enterprise size classes may be adapted. In particular, a fourth class of micro enterprises can be split out from small enterprises. Alternatively, small and medium-sized enterprises may be grouped together in a single SME category.
The choice of levels for the thresholds for defining size classes may be very different depending on country size. According to the European Commission’s definition (European Commission, 2003[33]), enterprises with 250 employees or more and a turnover of EUR 50 million or more are considered as large, while all others are considered as SMEs (see Box 2.2). However, this EU-wide threshold may imply that the number of large enterprises is very limited in smaller member states. Due to the size of the country’s market, thresholds for defining large enterprises and SMEs as identified by the Bureau of Economic Analysis are different for the United States: enterprises with 500 employees or more are labelled as large, enterprises with 100-499 employees are considered as medium-sized, enterprises with 20-99 employees are categorised as small, and all others are taken to be micro enterprises.
The European Union has adopted a definition of micro, small and medium-sized enterprises in order to harmonise the categorisation of enterprises for all member countries (European Commission, 2003[33]). This responds to the aim of ensuring cross-country consistency in assessing whether an enterprise is entitled to financial support restricted to small and medium-sized enterprises (SMEs). The definition is based on three variables: 1) the number of employees; 2) turnover; and 3) the balance sheet total. In addition, the definition also takes into account whether the enterprise is part of a group.
The category of micro, small and medium-sized enterprises is made up of enterprises that employ fewer than 250 persons and have an annual turnover not exceeding EUR 50 million, and/or an annual balance sheet total not exceeding EUR 43 million (Figure 2.3).
Within the category of SMEs, small enterprises are defined as those that employ fewer than 50 persons and have an annual turnover and/or annual balance sheet total that does not exceed EUR 10 million. Among small enterprises, micro enterprises are defined as those that employ fewer than 10 persons and have an annual turnover and/or annual balance sheet total that does not exceed EUR 2 million.
It is recommended to apply these thresholds at the level of the enterprise for independent enterprises and at the level of the enterprise group for dependent enterprises. An enterprise is considered as an SME if the group to which it belongs employs fewer than 250 persons and has an annual turnover of less than EUR 50 million or an annual balance sheet total of less than EUR 43 million.
Data for classifying enterprises
The data on enterprise-level employment required for determining enterprise size typically come from a census or social security records, which are exhaustive administrative data, from an enterprise’s annual accounts or from various surveys such as structural business surveys. Ideally, the number of employees of an enterprise should be calculated in full-time equivalents but if the relevant information is not available, then absolute numbers may be used. Standard data sources on enterprises’ turnover are their annual accounts (balance sheet or profit and loss account), VAT records, and structural business surveys and other surveys.
When it comes to breaking down industries for ESUT construction, the relevant statistical units (local kind-of-activity or legal units) that make up each industry must be categorised by size class, while size is determined at the level of the firm/enterprise. Thus, in practice, the categorisation of statistical units into size classes is dependent on the firm/enterprise definition and information linking these statistical units to firm/enterprise. Statistical units that are part of the same firm/enterprise should be categorised in the same size class, i.e. that of the firm/enterprise as a whole.
To determine the size of a firm/enterprise, compilers can rely either on specific enterprise-level information or on data for the individual statistical units. With thresholds based on employment, the source of information should not make a difference, since a firm’s/enterprise’s total employment normally corresponds to the sum of employment in all the statistical units that belong to the firm/enterprise. The situation is different for turnover because of transactions between the statistical units that make up the firm/enterprise. These transactions are part of the turnover of individual statistical units but are eliminated from consolidated accounts for the firm/enterprise as a whole.
As previously mentioned for the other extension criteria, compilers may rely on more aggregate data in the ESUT construction process when they do not have access to detailed enterprise-level data. For example, TEC data by enterprise size class could be used as a second-best alternative for the disaggregation of industry totals in regular SUTs.
Combining size and group affiliation
In addition, SMEs that are part of an enterprise group are likely to be different from those that are not part of a group. The former are generally referred to as dependent or linked SMEs or even pseudo-SMEs; the latter as independent, autonomous or genuine SMEs. The rationale behind this is that dependent SMEs, even though they meet the usual criteria for belonging to the SME size class, are different in terms of production process and input structure. Indeed, they have access to significant additional resources provided by their group (finance, skills, services, research and development, intellectual property, etc.) due to which they are also likely to face lower barriers to trade.
Therefore, group affiliation should, if possible, be taken into account as an additional source of heterogeneity when breaking down industries by enterprise size, so that SMEs that are part of an enterprise group (domestic or foreign) are separated from SMEs that are not part of a group. This has been done in prior work on ESUTs for the Nordic countries ( (Statistics Denmark and OECD, 2017[4]), the Netherlands (Chong et al., 2019[3]) and Belgium ( (Hambÿe, Michel and Trachez, 2023[29]).
A combined size-ownership breakdown leads to defining the following main categories of enterprises:
1. independent (autonomous, genuine) SMEs: small and medium-sized enterprises that are not part of an enterprise group (neither domestic nor multinational)
2. dependent (linked, pseudo-) SMEs: small and medium-sized enterprises that are part of an enterprise group (domestic or multinational)
3. large enterprises.
This adapted categorisation of enterprises is in line with the European SME definition (European Commission, 2003[33]) (see Box 2.2. The categories of dependent and independent SMEs can each be further subdivided into micro, small and medium-sized enterprises.
The size class categorisation of enterprises can be further refined by taking into account the size of the group to which an enterprise belongs. SMEs would only be considered as dependent if they are part of a large group, with the threshold for a large group set at the same level as the threshold for a large enterprise. However, in practice, determining the size of an enterprise group can prove difficult, mainly due to a lack of data. Compilers may find aggregated data at the group level for groups that draw up and publish consolidated accounts. Otherwise, compilers will have to deal with several issues. First, they have to delineate enterprise groups to identify all enterprises that belong to a group, which requires detailed information on group affiliation in the home country and abroad. Then, values for the threshold variables (employment, turnover and/or balance sheet total) must be determined for the entire group. In this respect, the non-additivity of the financial variables – turnover and balance sheet total – is again a major issue. For both, the value of the group as a whole does not necessarily correspond to the sum of the values for all enterprises that are part of the group due to within-group transactions. Compilers may therefore decide to rely only on employment, which is additive. However, compilers will mostly have access only to data for enterprises located in their country and therefore not be able to determine the size of multinational groups. A possible solution for this issue is to consider all foreign-controlled subsidiaries of multinational groups as part of a large group, regardless of group size; and all domestically controlled enterprises as SMEs if the group to which they belong has fewer than 250 employees in the home country, regardless of the number of employees in its foreign subsidiaries. This approach has been applied in prior work for the Netherlands (Chong et al., 2019[3]).
References
[31] ADB (2015), Integrating SMEs Into Global Value Chains, Asian Development Bank, Manila, https://www.adb.org/publications/integrating-smes-global-value-chains.
[13] Ahmad, N. (2023), “Accounting frameworks for global value chains: Extended supply-use tables”, in Ahmad, N. et al. (eds.), Challenges of Globalization in the Measurement of National Accounts, Studies in Income and Wealth, 81, University of Chicago Press, New York, NY, http://www.nber.org/books-and-chapters/challenges-globalization-measurement-national-accounts/accounting-frameworks-global-value-chains-extended-supply-use-tables.
[22] Antràs, P. and S. Yeaple (2014), “Multinational firms and the structure of international trade”, in Gopinath, G. (ed.), Handbook of International Economics, pp. 55-130, Elsevier, Amsterdam, Netherlands, https://doi.org/10.1016/B978-0-444-54314-1.00002-1.
[8] Bas, M. (2009), “Trade, foreign inputs and firms’ decisions: Theory and evidence”, Working Paper No. 2009-35, Centre d’Études Prospectives et d’Informations Internationales, https://www.international-economics.eu/PDF_PUB/wp/2009/wp2009-35.pdf.
[7] Bernard, A. et al. (2012), “The empirics of firm heterogeneity and international trade”, Annual Review of Economics, Vol. 4/1, pp. 283-313, https://doi.org/10.1146/annurev-economics-080511-110928.
[20] Brainard, S. (1997), “An empirical assessment of the proximity-concentration trade-off between multinational sales and trade”, American Economic Review, Vol. 87/4, pp. 520-544, https://www.jstor.org/stable/2951362.
[28] Cadestin, C. et al. (2018), “Multinational enterprises and global value chains: the OECD analytical AMNE database”, OECD Trade Policy Papers, No. 211, OECD Publishing, Paris, https://doi.org/10.1787/d9de288d-en.
[3] Chong, S. et al. (2019), “The role of small-and medium-sized enterprises in the Dutch economy: An analysis using an extended supply and use table”, Journal of Economic Structures, Vol. 8/1, pp. 1-24, https://doi.org/10.1186/s40008-019-0139-1.
[9] De Gortari, A. (2019), “Disentangling global value chains”, NBER Working Paper, No. 25868, National Bureau of Economic Research, Cambridge, MA, https://www.nber.org/papers/w25868.
[18] De La Cruz, J. et al. (2011), “Estimating foreign value-added in Mexico’s manufacturing exports”, Office of Economics Working Paper, No. 2011-04A, US International Trade Commission, Washington, DC, https://www.usitc.gov/publications/332/EC201104A.pdf.
[15] Dhyne, E., G. Magerman and S. Rubínová (2015), “The Belgian production network 2002-2012”, NBB Working Papers, No. 288, National Bank of Belgium, Brussels.
[33] European Commission (2003), Commission Recommendation of 6 May 2003 Concerning the Definition of Micro, Small and Medium-sized Enterprises, European Commission, Brussels, https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=CELEX:32003H0361.
[27] Eurostat (2012), Foreign AffiliaTes Statistics (FATS) Recommendations Manual, Publications Office of the European Union, Luxembourg, https://ec.europa.eu/eurostat/web/products-manuals-and-guidelines/-/ks-ra-12-016.
[16] Eurostat (n.d.), , https://ec.europa.eu/eurostat/databrowser/view/ext_tec06/default/table?lang=en.
[17] Eurostat (n.d.), , https://ec.europa.eu/eurostat/databrowser/view/ext_stec01/default/table?lang=en.
[2] Fetzer, J. et al. (2023), “Accounting for firm heterogeneity within US industries: Extended supply-use tables and trade in value added using enterprise and establishment level data”, in Ahmad, N. et al. (eds.), Challenges of Globalization in the Measurement of National Accounts, University of Chicago Press, https://www.nber.org/books-and-chapters/challenges-globalization-measurement-national-accounts/accounting-firm-heterogeneity-within-us-industries-extended-supply-use-tables-and-trade-value-added.
[26] FitzGerald, J. (2023), “National accounts for a global economy: The case of Ireland”, in Ahmad, N. et al. (eds.), Challenges of Globalization in the Measurement of National Accounts, University of Chicago Press, https://www.nber.org/books-and-chapters/challenges-globalization-measurement-national-accounts/national-accounts-global-economy-case-ireland.
[5] Hagino, S. and J. Kim (2021), “The usefulness of Japan’s extended input-output tables incorporating firm heterogeneity”, Journal of Economic Structures, Vol. 10/25, https://doi.org/10.1186/s40008-021-00255-3.
[29] Hambÿe, C., B. Michel and G. Trachez (2023), Ownership and Size Extended Supply and Use and Input-output Tables for Belgium, report under Eurostat Grant Agreement 101055571-2021-BE-NA, Belgian Federal Planning Bureau.
[1] Koopman, R., Z. Wang and S. Wei (2012), “Estimating domestic content in exports when processing trade is pervasive”, Journal of Development Economics, Vol. 99/1, pp. 178-189, https://doi.org/10.1016/j.jdeveco.2011.12.004.
[6] Melitz, M. (2003), “The impact of trade on intra‐industry reallocations and aggregate industry productivity”, Econometrica, Vol. 71/6, pp. 1695-1725, https://www.jstor.org/stable/1555536.
[10] Michel, B., C. Hambÿe and B. Hertveldt (2023), “The role of exporters and domestic producers in GVCs: Evidence for Belgium based on extended national supply-and-use tables integrated into a global multiregional input-output table”, in Ahmad, N. et al. (eds.), Challenges of Globalization in the Measurement of National Accounts, University of Chicago Press, https://www.nber.org/books-and-chapters/challenges-globalization-measurement-national-accounts/role-exporters-and-domestic-producers-gvcs-evidence-belgium-based-extended-national-supply-and-use.
[24] Miroudot, S. and D. Rigo (2022), “Multinational production and investment provisions in preferential trade agreements”, Journal of Economic Geography, Vol. 22/6, pp. 1275-1308, https://doi.org/10.1093/jeg/lbab031.
[25] Moulton, B. and P. van de Ven (2023), “Addressing the challenges of globalization in national accounts”, in Ahmad, N. et al. (eds.), Challenges of Globalization in the Measurement of National Accounts, University of Chicago Press, https://www.nber.org/books-and-chapters/challenges-globalization-measurement-national-accounts/addressing-challenges-globalization-national-accounts.
[12] Nilsson, M., P. Rørmose Jensen and J. Holst Jensen (2018), Compilation of Extended Supply and Use Tables in Denmark and Possible Applications in Input-Output Analyses, Paper presented at the 2018 IARIW conference, http://old.iariw.org/copenhagen/jensen.pdf.
[32] OECD (2023), OECD SME and Entrepreneurship Outlook 2023, OECD Publishing, Paris, https://doi.org/10.1787/342b8564-en.
[11] Ostolaza-Berman, R. and J. Téran Vargas (2019), The Mexican Single-country E-SUT 2013: Methodology and Results, Paper presented at the 27th IIOA conference, Glasgow, http://www.iioa.org/conferences/27th/papers/files/3744_20190501091_ESUTmexico.pdf.
[30] Pilgrim, G. and S. Ang (2024), “The OECD-UNSD Multinational Enterprise Information Platform”, OECD Statistics Working Papers, No. 2024/01, OECD Publishing, Paris, https://doi.org/10.1787/b7d90a06-en.
[19] Saborío, G. (2015), Costa Rica: An Extended Supply-Use Table, Paper presented at the 23rd IIOA conference, Mexico City, https://www.iioa.org/conferences/23rd/papers/files/2236_20150602101_EXTENDED-SUT.pdf.
[4] Statistics Denmark and OECD (2017), Nordic Countries in Global Value Chains, Statistics Denmark, Copenhagen, https://www.dst.dk/Site/Dst/Udgivelser/GetPubFile.aspx?id=28140&sid=nordglobchains.
[23] UNCTAD (2013), World Investment Report 2013: Global Value Chains: Investment and Trade for Development, United Nations, New York, NY, https://unctad.org/system/files/official-document/wir2013_en.pdf.
[14] United Nations (2011), International Merchandise Trade Statistics: Concepts and Definitions 2010, United Nations, New York, NY, https://unstats.un.org/unsd/trade/eg-imts/IMTS%202010%20(English).pdf.
[21] Yeaple, S. (2013), “The multinational firm”, Annual Review of Economics, Vol. 5, pp. 193-217, https://doi.org/10.1146/annurev-economics-081612-071350.