Handbook on Extended Supply and Use Tables and Extended Input‑Output Tables

1. What are extended supply and use tables and why are they useful?

Abstract

This chapter starts with a brief description of regular supply and use tables before turning to the analytical issues they face. The chapter then presents the concept of extended supply and use tables and factors for identifying the most relevant dimension to break down in industries in such a table in practice. The chapter further provides various country examples explaining policy relevance and showing results.

Introduction

This chapter starts with a brief description of regular supply and use tables (SUTs) presented in the Handbook on Supply, Use and Input-Output Tables with Extensions and Applications (United Nations, 2018[1]). Although they are useful tools, SUTs do not account for enterprise heterogeneity within industries and might result in biased indicators on globalisation. Complementing existing statistics, extended SUTs (ESUTs) do incorporate enterprise heterogeneity into a consistent framework through a breakdown of industries by type of enterprises (e.g. based on size, ownership, trading status or formal/informal). As such, they also help to respond to the increasing demand for statistical information by type of enterprise. In addition to presenting ESUTs from a conceptual point of view, the chapter also discusses how to identify the most relevant dimension to break down in an ESUT in practice. Selected country examples highlight that there is no ideal set-up to produce an ESUT. Rather, the choice of the dimension to focus on in an ESUT will depend on the country-specific policy needs, macroeconomic environment (e.g. trade openness) and data availability.

Usefulness of regular supply and use tables

Regular SUTs provide a balanced framework for describing economic activity by industry and by product in a country, a region or worldwide. This framework links product-level supply and use as well as industry-level inputs, outputs and gross value-added components.

The supply table shows the supply of goods and services by type of product with a distinction by origin: supply from domestic production by type of activity and supply from non-residents through imports. The sum of the two yields total supply.

The use table shows information about the uses of the different products, by either industries (intermediate consumption or gross capital formation), households, government and non-profit institutions serving households (final consumption expenditure), or non-residents (exports). Moreover, the use table provides information about the components of gross value added by industry – namely, compensation of employees, other taxes less subsidies on production, consumption of fixed capital and net operating surplus.

Table 1.1 shows the two basic identities linking the supply table and the use table. To be balanced, total supply by product must be equal to total use by product and total output by industry must be identical in the supply table and the use table.

Gross domestic product (GDP) can be calculated using the three approaches: 1) the production approach; 2) the income approach; and 3) the expenditure approach within the SUT framework. SUTs allow more easily identifying the sources of divergences across the goods and services account, the production account (by industry and by institutional sector) and the generation of income account (by industry and by institutional sector). The detail in products and industries allows one to find the source of a possible imbalance at a detailed level as well. For further details, see Chapter 2 of United Nations (2018[1]).

Table 1.1. Supply and use table framework

		Products				Industries				Final uses
		Agriculture, forestry, etc.	Mining and quarrying	…	Services	Agriculture, forestry, etc.	Mining and quarrying	…	Services	Final consumption expenditure	Gross capital formation	Exports	Total
Products	Agriculture, forestry, etc.					Intermediate consumption by product and by industry				Final uses by product and by category			Total use by product
	Mining and quarrying
	…
	Services
Industries	Agriculture, forestry, etc.	Output by product by industry											Total output by industry
	Mining and quarrying
	…
	Services
Value added						Value added by component and by industry							Value added
Imports		Total imports by product											Total imports
Total		Total supply by product				Total output by industry				Total final uses by category

Note: Empty cells by definition.

Source: United Nations (2018[1]).

Much of the analysis done within an SUT framework pivots around the input-output tables (IOTs), which are analytical tables derived from the SUTs. An industry-by-industry IOT shows how industries supply their output as input for industries (intermediate use) and final use (final consumption expenditure, gross capital formation and exports). The table also shows imports and value added by industry.

Analytical issues of regular supply and use tables

Regular SUTs (and associated IOTs) might not always provide the right answers in analysis. As Michel et al. (2019[2]) point out, such tables assume producers (industries) produce similar goods and services, i.e. relative homogeneity of the production functions (technical coefficients) of units classified under the same activity. This assumption is not always realistic. For instance, economies of scale play a different role in smaller and larger enterprises belonging to the same industrial activity.

Moreover, heterogeneity within industries has increased with globalisation. Enterprises may have specialised in specific tasks of the production process, with both greater (upstream) import and export orientation. For instance, a processing enterprise will have significantly less (recorded) imports than a non-processing enterprise just because it does not own the imports it uses. In addition, factoryless goods producers and enterprises involved in goods for processing arrangements would have even less (recorded) intermediate consumption of goods compared to those enterprises that choose to buy the material goods to be processed.

The increased heterogeneity within the industrial activities has considerably reduced the quality of input-output based indicators such as the Trade in Value Added (TiVA) indicators. These measure, amongst other issues, the extent of integration into global value chains (GVCs). But regular SUTs and IOTs do not always provide an accurate picture (see Box 1.1 for an illustration). The reason is that empirical evidence suggests that enterprises more engaged in global production arrangements have higher import content and, sometimes, higher productivity than others. The first leads to an overestimation of domestic value added in exports, the second to an overestimation of domestic jobs in exports. This is problematic.

Yamano and Webb (2022[3]) compare indicators obtained from a regular multi-country input-output table (MCIOT) with indicators obtained from an MCIOT where industries in the People’s Republic of China (hereafter “China”) and Mexico are split into processing trade and non-processing trade. The authors used different globalisation indicators, e.g. offshoring (Feenstra and Hanson, 1996[4]), import contents of exports (Hummels, Ishii and Yi, 2001[5]) and domestic value added in foreign final use (Johnson and Noguera, 2012[6]; OECD, 2013[7]). China and Mexico were chosen because processing trade is large in these countries and sufficient information is available to incorporate both types of producers.

Yamano and Webb (2022[3]) find that there is substantial bias in domestic value added embodied in international trade in both China and Mexico. For China, the biases were much higher in the mid-2000s (approximately 15%) than in more recent years (approximately 7%). Domestic value added embodied in foreign final use as a share of total domestic value added for China and Mexico was 18% and 20%, respectively, in recent years. But without splitting the export-oriented sectors in the MCIOT, the results are overestimated by 2.1 percentage points for China and 1.0 percentage point for Mexico.

Box 1.1. Fictitious example: Bias in case of strong enterprise heterogeneity

This box provides a simple fictitious example of why it is important to consider heterogeneity in certain circumstances when deriving Trade in Value Added indicators.

If one does not have any information about the import-export pattern of the underlying types of enterprises, one will estimate that for USD 50 in exports, imports embodied in exports amount to 50/200 * 50 = USD 12.5 (Table 1.2). However, if extra information is available, one can see that exports are mainly from large enterprises which also import more. The estimate of imports embodied in exports becomes 10/100 * 10 + 40/100 * 40 = USD 17.

Therefore, the import content of exports is underestimated in an analysis using a regular table, and therefore the domestic value added content of exports is overestimated.

Table 1.2. An example of bias

Values, USD

	Imports	Exports	Domestic sales	Total sales	Imports embodied in exports
Total – assuming all enterprises behave the same	50	50	150	200	12.5
Small and medium-sized enterprises	10	10	90	100	1
Large enterprises	40	40	60	100	16
Total, accounting for heterogeneity in enterprise size	50	50	150	200	17

Increasing the granularity in the data by providing more industry and/or product detail has been one way to tackle the problem of heterogeneity in the past. It is a natural approach, since within a standard SUT framework, efforts to minimise aggregation bias typically centre on choosing industry definitions and industry aggregation levels that maximise homogeneity within the resulting industry groupings. Depending on the specific situation at hand, this approach or the ESUT approach may be the best to reduce heterogeneity within groupings. In the decision process, policy demands, data availability for the specific ESUT and related thematic accounts (e.g. related to employment or environment), and the best option to reduce data-processing burdens can play a part too.

Sometimes, dividing an industry into several subindustries does not solve the heterogeneity problem related to IOTs. The reason is the following: input-output (IO)-multipliers use a matrix of technical coefficients. This matrix can be interpreted as the collection of production functions of different industries in the economy (e.g. the production function of the plastic manufacturing industry). Each production function in the matrix of technical coefficients is the weighted average of the production functions of individual establishments within the corresponding industry grouping. A key assumption that underpins the calculation of a matrix of IO-multipliers is that each industry column captures a set of establishments with similar, or homogeneous, input structures. Aggregation bias results when establishments with dissimilar, or heterogeneous, input structures are grouped together. Much research has been devoted to understanding and minimising the impact of aggregation bias on IO analysis. The heterogeneity within an industry, e.g. between small and medium-sized enterprises (SMEs) and large enterprises, generally does not depend on the different subindustries within an industry. This SME-large enterprise heterogeneity will be present in each subindustry that is not too small.

Extended supply and use tables accounting for enterprise heterogeneity

The ESUT approach has its focus on decomposition of industries by enterprise characteristics. Industries are split into heterogeneous groups, where each group itself is homogeneous. This leads to greater homogeneity of production functions within the resulting industry-by-enterprise-type groupings.

There are many ways to reconfigure the information in SUTs to show new details, but only a reconfiguration that decomposes industries by enterprise characteristics is considered an ESUT. Other common “extensions” of the SUT framework, such as digital SUT and thematic SUTs are not ESUTs and are described in Annex 1.A.

In practical terms, the extension typically consists of splitting columns (industries) further and possibly rows (products) in the SUTs by using (micro)data linked to related different official statistics (see the Handbook on Integrating Business and Trade Statistics (United Nations, forthcoming[8]) on micro-data linking). The dimensions used to break down industries and products in the SUTs define the type of ESUTs. For example, in the case of a breakdown by size, each industry could be split into two parts: 1) the SME part; and 2) the large enterprise part (see Table 1.3 for an example).

Similar to ESUTs, an extended input-output table (EIOT) is an IOT where industries are split by enterprise characteristics; see Table 1.3.C for an example. Note that other extensions of IOTs also exist, for example with information on investment, capital and labour. Additional information on energy, emissions, natural resources, waste, sewage and water could be added to the tables as well (United Nations, 2018[1]). However, such extensions are not EIOTs.

Table 1.3. Simplified examples of extended supply, use and input-output tables, by size

A. Extended supply table
Products	Industries													Imports					Total
	Manufacturing						Services
	SMEs		Large enterprises				SMEs		Large enterprises
Product 1	Output by product by industry by enterprise type													Imports by product					Total supply by product
Product 2
…
Total	Total output by industry by enterprise type													Total imports					Total supply
B. Extended use table
Products		Industries										Final uses								Total
		Manufacturing						Services				Final consumption expenditure			GCF		X
		SMEs		Large enterprises				SMEs		Large enterprises
Product 1		Intermediate consumption by product and by industry by enterprise type										Final uses by product and by category								Total use by product
Product 2
…
Value added		Value added by component and by industry by enterprise type
Total		Total output by industry by enterprise type										Total final uses by category
C. Extended input-output table
					Industries								Final uses							Total
					Manufacturing				Services				Final consumption expenditure			GCF		X
					SMEs	Large enterprises			SMEs		Large enterprises
Manufacturing		SMEs			Intermediate consumption of domestic production								Final uses of domestic production							Total use
		Large enterprises
Services		SMEs
		Large enterprises
		Imports			Intermediate consumption of imports								Final uses of imported products
		Value added			Value added by component
		Total			Total supply								Total final uses by category

Note: SMEs: small and medium-sized enterprises; GCF: gross capital formation; X: exports. Empty cells by definition.

It is not necessary, nor generally feasible, to break down each single industry or provide each extension for every industry. As national statistical offices or researchers decide which dimensions to focus on, the opportunity to highlight important aspects of the economy as well as the level of interest by data users in a particular topic are important considerations, as is the availability of the necessary data. Extending SUTs means breaking down product and industry categories (that are fairly aggregated), and several criteria can be applied to this end. Breakdowns by size class would be advisable to analyse the role of SMEs within GVCs, and if the investment agenda is a high-priority policy, a breakdown by ownership will be essential. The decision-making process should consider not only the benefits achieved through mitigation of aggregation bias and practical considerations such as data availability, but also the potential insights that these tables will provide in their own right. Annex 1.B describes an econometric approach that provides guidance on choosing the most technically appropriate breakdown by industry. Another factor is the existence of thematic accounts along the same dimensions, such as factor inputs or environmental stressors. These are crucial for researchers for many types of analysis. This aspect should also play a role in deciding on the disaggregation of the industries.

The enterprise-type extensions that are used might vary from one economy to another. The most common extensions are splits by ownership, size class or exporter status (Table 1.4). These are further elaborated in Chapter 2. Because ESUTs are often used in the calculation of TiVA statistics, enterprise-type categorisations that simultaneously capture differences in production functions and differences in foreign trade patterns are of special interest.

Other types of heterogeneity can be relevant as well. For example, one can consider groups of enterprises related to special trade regimes: processing traders for China (Koopman, Wang and Wei, 2012[9]; Yang et al., 2015[10]; Jiang et al., 2016[11]; Chen et al., 2019[12]), processing traders for regions in China (Duan et al., 2023[13]), enterprises operating under global manufacturing programmes for Mexico (De La Cruz et al., 2011[14]; INEGI, 2018[15]; Yamano et al., 2022[16]) and enterprises operating in free trade zones in Costa Rica (Saborío, 2015[17]). INEGI (2018[15]) describes the methodology of a split of the regular SUT of Mexico by all dimensions at the same time, capitalising on their economic census data.

In many developing countries, completely different EIOTs could be most relevant. In those countries, a large part of GDP and employment is generated by the informal sector. There, policy makers must understand the contributions of the informal sector and its interdependence with the formal sector in developing policies. An informal sector EIOT for India played a critical role in estimating the informal sector’s contribution to environmental and social impacts. Mitoma (2023[18]) revealed that the informal sector significantly contributed to the carbon footprint of the top three supply chains with the highest CO₂ emissions in India. Mitoma and Yamano (2024[19]) studied the bias that could arise in IOTs that aggregated the formal and informal sectors. Their analysis of employment in vulnerable groups embodied in Indian exports showed that ordinary IOTs can overestimate or underestimate the impact of exports by 10-40% at the industry level.

More detailed examples are provided throughout this handbook.

Table 1.4. Examples of extended input-output table and extended supply and use tables

Authors	Regional scope		Dimensions
OECD (2015[20])	OECD Member countries	EIOT	Enterprise size
Fortanier et al. (2020[21])	OECD Member countries	EIOT	Ownership
Statistics Denmark and OECD (2017[22])	European Nordic countries (Denmark, Finland, Iceland, Norway and Sweden)	EIOT	Size, ownership, exporter status
Fetzer et al. (2023[23])	United States	ESUT	Size, ownership, exporter status
Chong et al. (2019[24])	Netherlands	ESUT – EIOT	Size
Michel and Hambÿe (2022[25])	Belgium	ESUT and EIOT	Exporter status
Hambÿe et al. (2022[26])	Belgium	ESUT and EIOT	Type of multinational
Sallusti and Cuicchio (2023[27])	Italy	ESUT	Size, ownership, exporter status
INEGI (2018[15]; 2023[28])	Mexico	ESUT	Size, ownership, exporter status

Note: EIOT: extended input-output table; ESUT: extended supply and use table.

Purposes of extended supply and use tables

The creation of an ESUT results in new groupings of enterprises within the economy. It provides an extra layer to existing statistics that is well-integrated into the core accounting framework of National Accounts. In particular, the complete, integrated view of enterprises by size class, ownership or trading status is consistent with macroeconomic indicators such as GDP.

The integrated framework of ESUTs can also improve the quality of the core economic accounts, in particular when the accounts are affected by globalisation. An ESUT takes into account that larger enterprises may have different input structures due to their engagement in international trade as well as different economies of scale than smaller enterprises. The same holds for foreign-owned enterprises or enterprises with affiliates abroad, compared to purely domestically owned enterprises.

ESUTs also provide the basis for a better articulation of globalisation across various domains. Currently, national SUTs are the usual route to systematically and coherently combine national accounts, trade statistics and business surveys into a single integrated economic framework. ESUTs can provide additional information to better understand the various facets of globalisation, by combining information from SUTs with different data sources. Examples of such data sources are foreign affiliates statistics (FATS)/activities of multinational enterprises, commodity trade by enterprise characteristics and services trade by enterprise characteristics.

An ESUT and EIOT also provide valuable new policy-relevant insights on the different roles played by each enterprise category and how they interact with one another. For instance, information on foreign affiliates helps to quantify spillover effects from foreign direct investment and it allows extending analysis beyond value-added concepts such as the international profit (income) distribution of enterprises. Alternatively, the role of SMEs and indirect trade links through large enterprises can be studied.

Selected practical country examples

This section provides concrete country examples in compiling ESUTs and EIOTs. It highlights that there is no ideal enterprise grouping to be used to break down SUTs (Table 1.5). Rather, the choice is likely to depend on the country-specific policy needs, macroeconomic environment (e.g. trade openness) and data availability. The examples show that the methods are broadly replicable across countries.

Table 1.5. Several country experiences in selecting the relevant dimension to compile extended supply and use tables and extended input-output tables

	Criteria	Source of information	Dimension
Belgium	Economic relevance combined with data availability	Feasibility study	Ownership
China (People’s Republic of)	Economic relevance	Existing knowledge on enterprise behaviour	Ownership
Costa Rica	Economic relevance	Existing knowledge on enterprise behaviour	Ownership
Finland	Policy relevance	Demand from policy makers	Trading status and a combined enterprise size and group relation
Japan	Characteristics that present the highest level of heterogeneity	Statistical approach	Each industry is split by type of enterprise that is most relevant for this specific industry
Mexico	Economic relevance	Statistical approach	Size, ownership, exporter status
United States	Economic relevance	Academic literature	Size and ownership

Belgium

In Belgium, the Federal Planning Bureau (FPB) started to work on the construction of ESUTs in 2015, pursuing two major aims: 1) incorporating information on enterprise heterogeneity into the macroeconomically consistent framework of the SUT; and 2) producing policy-relevant analytical results.

The initial project focused on a disaggregation by exporter status for manufacturing industries in the 2010 Belgian SUT. It was considered as a test run and therefore limited in scope to keep the workload manageable. The exporter status criterion was chosen because of its relevance for a small and very open economy like Belgium and because it appeared relatively straightforward to implement.

In the wake of this successful initial project, the FPB received funding from Eurostat to pursue its work on ESUTs. This led to a follow-up project that began in 2019 and consisted of two parts: 1) a feasibility study of producing an ESUT for Belgium for each of the three main dimensions of enterprise heterogeneity, i.e. size, ownership and exporter status (Michel and Hambÿe, 2022[25]); and 2) an extended SUT for Belgium for 2015 for one of these heterogeneity dimensions. It was decided to construct an ESUT with a disaggregation by ownership given the strong and growing interest in this dimension. Analytical results based on the ownership-extended tables highlighted the importance of multinational groups in the Belgian economy (Hambÿe et al., 2022[26]).

A new Eurostat-funded project (Hambÿe, Michel and Trachez, 2023[29]) produced ownership-extended tables for 2019. It also investigated a disaggregation of industries by enterprise size class that takes into account whether enterprises belong to a domestic or a multinational group. It was found that, together, Belgian and foreign MNEs accounted for more than half of the total value of output in 2019. Their joint share in GDP stood at almost 45%, with 15% due to enterprises that belong to a Belgian MNE and almost 30% due to affiliates of foreign MNEs. The GDP share of domestic enterprises was 26%, while the non-disaggregated industries accounted for the remaining 30%. For employment, the shares of Belgian MNEs and foreign MNEs were lower, with, respectively, 11% and 18%, whereas the share of domestic enterprises was significantly higher (35%). Finally, Belgian MNEs, and in particular foreign ones, largely dominated Belgian exports of goods and services. Foreign MNEs accounted for 58% of total Belgian exports.

Michel and Hambÿe (2022[25]) combined the export-heterogeneous tables with employment. They found that 585 000 jobs, or 13% of economy-wide employment in Belgium, are sustained by manufacturing exports. This number would be overestimated by 4% if one were to use regular tables. Moreover, they identified who contributes to and who gains from exports for groups of enterprises rather than aggregated industries.

China (People’s Republic of)

There is a long tradition of incorporating enterprise heterogeneity into SUTs and IOTs in China (see, for example, Duan et al. (2012[30]); Hong, Wang and Zhu (2015[31]); and Yang, Wei and Zhu (2016[32]).

Yang et al. (2022[33]) compiled an EIOT where each industry is split into foreign-invested enterprises (FIEs) and domestically owned enterprises (DOEs). See Chapter 4 for a description of the methodology. FIEs and DOEs in China are different in many aspects, including production input, export pattern and impacts on the local economy. For example, FIEs are more export-oriented than DOEs. Besides, FIEs and DOEs play different roles in generating local value added, since a large part of value added from GVCs is generated by affiliates of multinational enterprises. FIEs and DOEs also exhibit different performance on technology dissemination and skill building.

The domestic value-added share in gross exports of DOEs was 82% in 2012, which is 17 percentage points higher than that of FIEs (Table 1.6). A possible reason is that DOEs have a higher share of non-processing exports than FIEs (88% and 33%, respectively). Not shown in the table: the share of FIEs’ contribution to GDP was 16%, and FIEs’ share in total value added in exports was 25%. FIEs’ dependence on foreign trade is up to 30%, 13 percentage points higher than for DOEs.

Table 1.6. Domestic and foreign content share of China’s exports in 2012

	Total exports (billion CNY)	DOEs’ share	FIEs’ share	Total exports	DOEs’ exports	FIEs’ exports
Exports	14 141	53%	47%
Domestic value added in exports	10 487	58%	42%	74%	82%	65%
Foreign value added in exports (vertical specialisation level)	3 654	36%	64%	26%	18%	35%

Note: CNY: Chinese yuan; DOE: domestically owned enterprise; FIE: foreign-invested enterprise.

Source: Yang et al. (2022[33]).

Looking at the industry breakdown, manufacturing as a whole is characterised by a lower value-added ratio in gross exports both for DOEs and FIEs than the average for these types of enterprises (79% and 62%, respectively; with averages of 82% and 65%, respectively). DOEs export most in “wearing apparel” and “chemical products”. FIEs export most in “communication equipment and electronic equipment” followed by “electrical machinery”. As for the value-added ratio in gross exports, DOEs and FIEs differ greatly in “electrical machinery”, “communication equipment and electronic equipment”, and “measuring instruments and meters”.

Costa Rica

Costa Rica is a highly open economy. In recent years, exports and imports have each amounted to one third of GDP, while foreign share enterprises (FSEs) have accounted for approximately 64% of exports (Steller et al., 2021[34]). FSEs include multinational corporations that primarily target foreign markets and are connected to different stages within GVCs, resulting in dissimilar levels of interaction with domestic markets. Some of these enterprises have no connection to domestic markets while others have domestic control enterprises (DCEs) as their main suppliers. This dynamic causes heterogeneity within the economy in many areas, such as income payments to the rest of the world, production functions and foreign content ratios. It also creates a need for more granular data about income, employment, supply and demand relationships, and the linkages between export and domestic enterprises.

To meet the needs of policy makers and researchers, the Central Bank of Costa Rica has created an ESUT, an EIOT and institutional sector accounts that present data about FSEs and DCEs. These provide enhanced tools for economic analysis, research and projections.

In 2018, FSEs accounted for 26% of value added, 17% of employment and 66% of the country’s exports. They were primarily focused on activities oriented towards external markets, such as medical devices, foods, drinks, and professional and scientific services. Salaries in FSEs were 35% higher than those in the rest of the economy.

FSEs produced a diverse range of products, such as medical appliances, bananas, pineapples, food products, tires and plastic products. Exports of goods carried out by DCEs were more diverse, nonetheless. Almost half of the services exported by FSEs consisted of head offices and management services, followed by computer programming services and administrative and support services. Service exports from DCEs were more diverse than those from FSEs.

The imported component in manufacturing FSEs was double that of DCEs (33% and 16%, respectively) while in services it was quite similar (9% and 8%, respectively). At a more disaggregated level, the economic activities of manufacturing and services exhibited behaviours similar to those of the aggregates. Manufacturing FSEs had a slightly lower value-added/output ratio than the DCEs (37% versus 38%), but the ratio for FSEs in the case of services exceeded that of DCEs by 7 percentage points (62% versus 55%). Differences in these ratios were especially notable in the cases of medical devices and processed fruit and vegetables, where the ratios for DCEs were substantially higher than those of FSEs, and in administrative and support services, where the ratio for FSEs was much higher than that for DCEs.

Finland

Finland is a small open economy with significant exposure to foreign trade, and better knowledge of how trade affects the domestic economy is pivotal to improving policy making and securing economic stability and growth. ESUTs can be a powerful tool to design economic policy and can interest many stakeholders in government offices and research organisations.

The domestic need for better information on trade in value added was initially expressed by the Finnish Ministry for Foreign Affairs in 2017. The ministry stressed the need for information on where Finnish value added is exported, where it is consumed and what kind of imports are important for Finnish enterprises. Other stakeholders expressed specific needs for information as well. Policy makers already had several tools at hand, but the need for detailed and timely data has been ever growing (Box 1.2). Official statistics, and analytical frameworks that build on official statistics, are important to ensure reliable policy making.

In 2020, Statistics Finland and the OECD jointly expanded the scope of TiVA statistics to gain a more detailed and timelier picture of Finland´s role in global production. The project benefited from previous efforts carried out by the Nordic countries in co-operation with the OECD (Statistics Denmark and OECD, 2017[22]). It was financially supported by several domestic stakeholders, including the Finnish Ministry for Foreign Affairs, the Ministry of Economic Affairs and Employment of Finland, the Finnish Prime Minister´s Office, and the Confederation of Finnish Industry and Employers Foundation. Outputs included a report (OECD and Statistics Finland, 2020[35]) and regularly updated indicators (Statistics Finland, 2024[36]).

Box 1.2. Policy demand for extended supply and use tables in Finland

More granular information about globalisation, in the form of Trade in Value Added (TiVA) indicators by type of enterprise – such as the value added by SMEs due to exports from large enterprises – was highly sought after. Stakeholders from government offices, research organisations and academia showed interest in the data during the development phase and actively took part in the development of the project. The Trade in Value Added (TiVA) indicators are freely accessible on Statistics Finland’s website and a comprehensive list of users can, therefore, not be compiled. However, Statistics Finland has been contacted by various organisations since the indicators have been released, including several ministries, business associations and think tanks:

• The Ministry for Foreign Affairs was especially interested in more detailed trade information and information on trading partners. It funded two follow-up reports produced jointly by Statistics Finland and the OECD, which were released in May 2021.

• The Parliament Committee for the Future requested recommendations on how the new indicators can be used for policy making (e.g. trade policy) and was also interested in developing new sustainability indicators and indicators related to the European Green Deal. The Parliament Committee for the Future commissioned Statistics Finland to write a “handbook” on how to interpret and use granular TiVA indicators. The work included mapping of future development possibilities for the framework.

• Business Finland was especially interested in the role of multinational enterprises in domestic and global value chains. Stakeholder interest in the specific heterogeneity breakdown pushed Statistics Finland to accelerate the planned release for additional indicator packages.

• ETLA Economic Research (a private, non-profit economic research institute) was moving towards the use of TiVA in the coming years and indicated it will likely make use of Statistics Finland’s granular indicators when building and using its new analytical frameworks.

• The Finnish Prime Minister’s Office co-ordinates the government’s analysis, assessment and research activities to generate information that supports decision making and evidence-based policies, thus improving working practices. After the first national TiVA publication in November 2020, the Prime Minister’s Office launched funding for a research project entitled “Trends in International Trade and Economic Sensitivity” to support the government in policy making. The joint report by the OECD and Statistics Finland (OECD and Statistics Finland, 2020[35]) was mentioned as a possible base for the analysis.

Source: Statistics Finland.

During the project, Statistics Finland and the OECD developed a standardised process for disaggregating domestic SUTs using enterprise-level microdata, generating EIOTs, balancing the extended tables and calculating extended TiVA indicators. Enterprise grouping was chosen to respond to policy demand. Indicators were broken down by enterprise trading status (exporter, importer, two-way trader and non-trader), enterprise size and group relation (independent micro, dependent micro, independent small, dependent small, independent medium, dependent medium, large, and other), and ownership (domestic non-multinational, domestic multinational, foreign multinational and other). In addition to the standard TiVA indicators, the OECD and Statistics Finland also developed employment indicators. These are further broken down by gender and level of education, providing deeper insights into employment structures and GVC dependencies of enterprises and employees.

With the standardised process, Statistics Finland has set up regular publication of granular TiVA indicators at t+17 months after the end of the reference period. Statistics Finland’s current publication contains TiVA indicators for 2017-22 and employment indicators for 2017-21. The flexible and standardised production process allows for easy integration of new breakdowns when needed. The extensive register- and survey-based enterprise-level data of the national statistical office can be linked using unique business IDs, which in turn allows for the exploration of various enterprise heterogeneity breakdowns. This led to the publication of indicators by enterprise ownership (domestic non-multinational enterprise, domestic multinational and foreign multinational) in 2024, as there was an increasing need for information on the dynamics of enterprise ownership.

Domestic non-multinational enterprises are found to owe less of their value added due to foreign demand than multinationals do (Figure 1.1). For example, domestic non-multinational enterprises have 9% of their value added due to their own exports, whereas this is 45% for foreign multinationals. Differences in the share of value added due to indirect exports, producing goods and services in the supply chain of an exporter, are much smaller.

Figure 1.1. Finland value added due to direct and indirect exports, 2022

Share of total value added, by enterprise ownership

Source: Calculations based on Statistics Finland (2025[37]), Table 131a.

Japan

Hagino and Kim (2021[38]) document how they decided on the type of EIOT the most appropriate for Japan. They started by considering splitting industries into exporters and non-exporters, domestically owned and foreign-owned enterprises, and large and small enterprises as well as enterprises with and without foreign subsidiaries. Subsequently, they examined which type of heterogeneity is the most relevant for Japanese industries, heterogeneity being measured in terms of differences in the ratio of imported intermediate goods to total output, using enterprise-level microdata from the Basic Survey of Japanese Business Structure and Activities.

They found that overall, in Japan, the distinction between enterprises with and without foreign subsidiaries is relevant (Figure 1.2). In Japan, enterprises with foreign subsidiaries account for more than 95% of all exports and imports, which is considerably higher than in France and the United Kingdom. By contrast, foreign-owned enterprises do not play a pivotal role in international trade, therefore, this distinction may be less relevant in Japan than in other countries.

Figure 1.2. Share of enterprises with and without foreign subsidiaries in exports and imports, 2017

Source: For France and the United Kingdom: Trade by Enterprise Characteristics, (OECD, 2024[39]). For Japan: enterprise-level data from the (METI, 2023[40]).

The next step was to identify which breakdown was the most relevant for each industry by calculating the differences in the intermediate import ratio between different types of enterprises (Figure 1.3). In the metal and paper industries, the differences between enterprises with and without foreign subsidiaries are larger than those between exporters and non-exporters, as well as those between small and large enterprises. This reflects that metal and paper corporations, which need to import materials, have established subsidiaries to explore and mine raw materials or grow and harvest wood. In contrast, distinguishing between exporting and non-exporting enterprises is relevant for assembly industries such as the electronic and automobile industries, as widely discussed in the literature. For the chemicals industry, distinguishing between large and small enterprises is the most relevant.

Figure 1.3. Comparisons of intermediate import ratios of enterprises in Japan, 2015

Note: Large enterprises are those with a paid-in capital of at least 1 billion JPY.

Source: Hagino and Kim (2021[38]), based on enterprise-level data of the METI (2023[40]).

Following those results, each industry is split by type of enterprise that is most relevant for this specific industry (Table 1.7). Since differences in industries’ intermediate import ratios are mainly due to the import of goods that the industry produces itself, it is assumed that such differences stem from differences in within-industry imports. The differences in intermediate import ratios are reflected in the diagonal cells (the shaded parts in Table 1.7).

Table 1.7. Extended domestic input-output table and import table for Japan, 2015

	3		4		5		6		7-11
	Exporting	Non-exporting	Exporting	Non-exporting	With foreign subsidiaries	Without foreign subsidiaries	Large firms	Small firms	All
Extended domestic input-output table
Output, billion JPY
1	1 964	3 864	0	0	69	169	3	1	7
2	0	0	0	0	2	4	210	52	36
3	1 694	3 332	0	0	5	12	92	23	1
4	8	17	159	380	12	30	63	16	220
5	298	587	18	10	1 052	2 563	518	128	876
6	380	748	209	119	309	753	11 666	2 872	7 019
7	173	340	5	3	80	194	541	133	27 942
8	0	0	0	0	4	10	51	12	4 978
9	0	0	0	0	1	2	1	0	9 773
10	0	0	0	0	0	0	0	0	22 796
11	14	27	8	4	32	77	96	24	495
Total output	12 920	25 421	2 281	1 305	4 926	12 001	52 557	12 937	177 001
Extended import table
Import, billion JPY
1	559	1 101	18	11	36	88	163	40	56
2	4	7	1	0	14	34	8 129	2 001	1 289
3	1 065	884	5	3	1	2	87	21	10
4	4	9	175	131	12	28	43	11	74
5	12	24	1	1	390	372	33	8	62
6	81	159	84	48	29	71	4 582	1 213	819
7	2	4	1	0	9	22	45	11	4 429
8	0	0	0	0	0	0	0	0	1 535
9	0	0	0	0	1	2	0	0	5 145
10	0	0	0	0	0	0	0	0	2 060
11	2	4	20	11	2	5	5	1	163
Total import	1 730	2 191	305	205	494	624	13 087	3 307	15 644

Note: 1: Agriculture; 2: Mining; 3: Food; 4: Textiles; 5: Paper; 6: Chemicals; 7: Metal; 8: Machinery; 9: Electronics; 10: Transport equipment; 11: Other manufacturing. Imports by an industry from the same foreign industry.

Source: Calculations of Hagino and Kim (2021[38])) based on the Benchmark IOT (MIC, 2024[41])and enterprise-level data of the Basic Survey of Japanese Business Structure and Activities, Ministry of Economy, Trade and Industry (METI, 2023[40]).

The extension of IOTs incorporating differences in intermediate import ratios leads to very different estimates of vertical specialisation than estimates obtained with a non-extended SUT (Table 1.8). Vertical specialisation is defined as the amount of foreign value added embodied in exports and measured as the ratio of imported intermediate goods embodied in exports, following (Hummels, Ishii and Yi, 2001[5]). The vertical specialisation indicator, defined as the amount of VS divided by exports, based on the EIOT (34.3%) is 70% greater than that based on the non‑extended IOT (20.5%).

Table 1.8. Vertical specialisation indicators calculated from the extended and non-extended input-output tables for Japan, 2015

		Calculation based on extended IOT			Calculation based on non-extended IOT
Industry	Extension elements	Industry total of VS coefficient	Exports	Amount of VS	Industry total of VS coefficient	Exports	Amount of VS	Domestic value added included in imported intermediates	Amount of VS after deducting domestic value added
			billion JPY	billion JPY		billion JPY	billion JPY	%	billion JPY
Agriculture		0.17	0	0	0.12	0	0	0.6	0
Mining		0.10	33	3	0.07	33	2	0.5	2
Food	Exporting	0.43	155	66	0.18	155	29	1.2	28
	Non-exporting	0.38	0	0
Textile	Exporting	0.96	131	126	0.24	131	32	1.6	31
	Non-exporting	0.41	0	0
Paper	With foreign subsidiaries	0.47	289	135	0.16	375	60	0.9	59
	Without foreign subsidiaries	0.42	86	36
Chemical	Large	0.54	6 811	3 673	0.36	7 521	2 679	1.8	2 630
	Small	0.55	711	388
Metal	With foreign subsidiaries	0.35	2 866	1 014	0.16	4 535	739	1.9	725
	Without foreign subsidiaries	0.31	1 669	521
Machinery	Exporting	0.34	8 685	2 943	0.16	8 685	1 348	7.1	1 253
	Non-exporting	0.36	0	0
Electronics	Exporting	0.43	14 294	6 200	0.25	14 294	3 536	8.2	3 245
	Non-exporting	0.52	0	0
Transport equipment	Exporting	0.42	22 919	9 644	0.21	22 919	4 756	4.3	4 552
	Non-exporting	0.48	0	0
Other manufacturing	Exporting	0.52	526	275	0.16	526	86	7.9	79
	Non-exporting	0.51	0	0
Electric, gas and water		0.31	23	7	0.30	23	7	0.8	7
Construction		0.08	23	2	0.10	23	2	1.5	2
Wholesale and retail		0.09	18 051	1 602	0.15	18 051	2 650	1.2	2 618
Transportation and warehouse		0.12	158	20	0.08	158	13	2.7	13
Finance and insurance		0.08	0	0	0.06	0	0	1.1	0
Real estate and leasing		0.05	16	1	0.04	16	1	1.0	1
Community, society and individual services		0.09	278	25	0.04	278	21	2.4	20
Total		9.5	77 725	26 680	2.87	77 725	15 961	2.1	15 632

Note: VS: vertical specialisation.

Source: Calculations of Hagino and Kim (2021[38]) based on the Benchmark IOT (MIC, 2024[41]) enterprise-level data of the Basic Survey of Japanese Business Structure and Activities (METI, 2023[40]) and OECD TiVA indicators (OECD, 2024[42]).

Mexico

In Mexico, the National Institute of Statistics and Geography (INEGI) produced its first ESUT in 2018 (reporting year 2013) and a new ESUT in 2023 (reporting year 2018). The 2023 update was in line with the new base year of National Accounts in Mexico, namely 2018. The reason for compiling an ESUT was to obtain more detailed information about the types of enterprises that are economically relevant.

The main sources of information for the Mexican ESUT are the Economic Census (which takes place every five years), the Foreign Trade Database and the regular SUTs. The advantage of using the Census is that it provides a large amount of information at the establishment level. This makes it feasible to characterise them according to the criteria considered appropriate for the extensions to be developed. Thanks to the Census, INEGI obtained maximum detail in its tables. There are different ESUTs, namely with the following focuses:

• Export status, where establishments are classified into exporter, formal non-exporters and informal non-exporters.

• Ownership, where exporting establishments are classified as domestic owner affiliates, foreign owner affiliates, domestic owner and foreign owner.

• Size of the economic unit, grouping the economic units into small, medium and large. This extension covers the total economy, not only the exporting establishments.

• The integrated focus that in a single tabulation distinguishes between export status and ownership and size at the same time. For example, it contains information about the formal non-exporters that are foreign-owned and large establishments.

INEGI’s internal ESUTs are very detailed but contain confidential information. Therefore, the published data only contain 20 industries. These industries are according to the North American Industry Classification System (NAICS). The process of compiling the ESUTs involves taking all information from the regular SUT as fixed and a subsequent aggregation towards the specific extension. For this reason, re-aggregating each of the extensions leads to the regular SUT again.

The economic unit that the Mexican ESUTs measure is the establishment, not the company. Chapter 2 emphasises that the compiler decides whether all the establishments that form part of a legal unit (or company) should be classified into the same category. In the case of Mexico, for practical reasons, it was decided that the ideal measurement to extend the SUT is at the establishment level since this is also the way the regular SUT are measured. See Teran-Vargas (2024[43])) for more details about the methodology.

To give an example of the results about the new base year 2018: in manufacturing, 70% of the production is produced by exporting establishments. Of this production by exporters, 44% is at domestically controlled establishments and of this production, 78% is by large establishments (which employ more than 250 persons). The remaining 56% of production in manufacturing by exporters is by foreign-owned establishments and of this production, 88% is by large establishments.

United States

Efforts in the United States to develop ESUTs have focused primarily on adding information on multinational enterprise and ownership status. SUTs have been extended to show breakdowns within each industry for domestically owned multinational enterprises, foreign-owned multinational enterprises and non-multinationals.

Among others, academic literature was a reason for doing so. Melitz (2003[44]) motivated the development of models of international trade that allow for different levels of enterprise‐level productivity. Helpman, Melitz and Yeaple (2004[45]) found that only the most productive enterprises engage in foreign activities and the most productive of these enterprises engage in foreign direct investment. Consistent with Melitz’s model, there is also evidence that the bulk of trade in both goods (Bernard, Jensen and Schott, 2009[46]) and services (Barefoot and Koncz-Bruner, 2012[47]) has involved MNEs. Therefore, extending SUTs by MNE categorisation is an important way to account for enterprise heterogeneity.

Fetzer et al. (2023[23]) compiled such data and demonstrate how enterprises in different categories play different roles in the production of US exports. They showed that about half the value of US exports in 2005 and 2012 was created by non-multinationals, about 40% by domestically owned and foreign-owned multinationals combined, and the remaining 10% coming from foreign production via imported intermediate inputs. They also underlined the important role that foreign-owned multinationals play more generally in domestic supply chains. Of the value added by foreign-owned multinationals, 30% was embedded as an input into production by other types of enterprises.

Additionally, the imported content of exports as a share of exports varied notably by enterprise type within most industries. The imported content of exports was concentrated in a few industries, the largest being petroleum manufacturing. Most domestic content of exports by enterprises in goods-producing industries was from US MNEs and most domestic content of exports in services industries was from non-MNEs.

References

[48] 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/chapters/c14136.

[47] Barefoot, K. and J. Koncz-Bruner (2012), “A profile of US exporters and importers of services: Evidence from new linked data on international trade in services and operations of multinational companies”, Survey of Current Business, Vol. 92/6, pp. 66-87, https://apps.bea.gov/scb/pdf/2012/06%20June/0612_MNC.pdf.

[46] Bernard, A., J. Jensen and P. Schott (2009), “Importers, exporters and multinationals: A portrait of firms in the US that trade goods”, in Producer Dynamics: New Evidence from Micro Data, pp. 513-552, University of Chicago Press.

[12] Chen, Q. et al. (2019), “Distinguishing China’s processing trade in the world input-output table and quantifying its effects”, Economic Systems Research, Vol. 31/3, pp. 361-381, https://doi.org/10.1080/09535314.2018.1534225.

[24] 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.

[14] 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.

[13] Duan, Y. et al. (2023), “Processing trade in Chinese interregional input-output tables: Construction and application”, Economic Systems Research, Vol. 35/4, pp. 566-585, https://doi.org/10.1080/09535314.2021.2012430.

[30] Duan, Y. et al. (2012), “Does domestic value added induced by China’s exports really belong to China?”, China & World Economy, Vol. 20/5, pp. 83-102, https://doi.org/10.1111/j.1749-124X.2012.01303.x.

[4] Feenstra, R., G. Grossman and D. Irwin (eds.) (1996), Foreign Investment, Outsourcing and Relative Wages, The Political Economy of Trade Policy: Papers in Honor of Jagdish Bhagwati, pp. 89-127, The MIT Press, Cambridge.

[23] 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. (ed.), Challenges of Globalization in the Measurement of National Accounts, University of Chicago Press.

[21] Fortanier, F. et al. (2020), “Accounting for firm heterogeneity in global value chains”, Journal of International Business Studies, Vol. 51, pp. 432-453, https://doi.org/10.1057/s41267-019-00282-0.

[53] Greene, W. (2019), Econometric Analysis. Eighth edition, Pearson Education Ltd. Essex, United Kingdom.

[38] Hagino, S. and J. Kim (2021), “The usefulness of 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, Federal Planning Bureau Belgium Report 12932.

[26] Hambÿe, C. et al. (2022), “Les groupes multinationaux en Belgique: Structure et activité économique”, Working Paper 5-22, Federal Planning Bureau, Brussels, https://www.plan.be/sites/default/files/documents/WP_202205_FR.pdf.

[45] Helpman, E., M. Melitz and S. Yeaple (2004), “Export versus FDI with heterogeneous firms”, American Economic Review, Vol. 94/1, pp. 300-316, https://doi.org/10.1257/000282804322970814.

[31] Hong, M., Z. Wang and K. Zhu (2015), “Domestic value-added in China’s exports and its distribution by firm ownership”, Journal of Comparative Economics, Vol. 43, pp. 3-18.

[5] Hummels, D., J. Ishii and K. Yi (2001), “The nature and growth of vertical specialization in world trade”, Journal of International Economics, Vol. 54/1, pp. 75-96, https://doi.org/10.1016/S0022-1996(00)00093-3.

[28] INEGI (2023), Cuadros de oferta y utilización extendidos 2018: Fuentes y metodologías, National Institute of Statistics and Geography, Mexico.

[15] INEGI (2018), Cuadros de oferta y utilización extendidos 2013: Fuentes y metodologías, National Institute of Statistics and Geography, Mexico.

[11] Jiang, X. et al. (2016), “Revisiting the global net carbon dioxide emission transfers by international trade: The impact of trade heterogeneity of China”, Journal of Industrial Ecology, Vol. 20/3, pp. 506-514, https://doi.org/10.1111/jiec.12404.

[6] Johnson, R. and G. Noguera (2012), “Accounting for intermediates: Production sharing and trade in value added”, Journal of International Economics, Vol. 86/2, pp. 224-236, https://doi.org/10.1016/j.jinteco.2011.10.003.

[9] 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.

[44] 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.

[40] METI (2023), Basic Survey of the Japanese Business Structure and Activities, Japanese Ministry of Economy, Trade and Industry, https://www.meti.go.jp/english/statistics/tyo/kikatu/index.html.

[41] MIC (2024), Input-Output Tables for Japan, Japanese Ministry of Internal Affairs and Communication, https://www.soumu.go.jp/english/dgpp_ss/data/io/index.htm.

[25] Michel, B. and C. Hambÿe (2022), “Export-sustained employment: Accounting for exporter-heterogeneity in input-output tables”, Economic Systems Research, Vol. 34/2, pp. 215-233, https://doi.org/10.1080/09535314.2020.1869701.

[2] Michel, B. et al. (2019), “Extended supply and use tables for Belgium: Where do we stand”, Eurostat Review on National Accounts and Macroeconomic Indicators (Eurona), Vol. 2/2019, pp. 51-71.

[52] Miller, R. and P. Blair (2022), Input-Output Analysis: Foundations and Extensions. Third edition., Cambridge University Press, Cambridge, United Kingdom.

[18] Mitoma, H. (2023), “Carbon footprint analysis considering production activities of informal sector: The case of manufacturing industries of India”, Energy Economics, Vol. 125, https://doi.org/10.1016/j.eneco.2023.106809.

[19] Mitoma, H. and N. Yamano (2024), Estimating an Extended Informal Sector Input-Output Table for India with an Employment Account Based on Workforce Characteristics, presented at the 30th International Input-Output Association Conference, Santiago, Chile, https://www.iioa.org/conferences/30th/papers/files/4994.pdf.

[39] OECD (2024), Trade by Enterprise Characteristics (TEC), OECD publishing, https://www.oecd.org/en/data/datasets/trade-by-enterprise-characteristics-tec.html.

[42] OECD (2024), Trade in Value Added (TiVA) indicators 2023 edition, OECD, Paris, https://data-explorer.oecd.org/?pg=0&bp=true&snb=14&tm=TIVA.

[49] OECD (2023), OECD Handbook on Compiling Digital Supply and Use Tables, OECD Publishing, Paris, https://doi.org/10.1787/11a0db02-en.

[20] OECD (2015), Firm Heterogeneity and Trade in Value Added, OECD, Paris, https://one.oecd.org/document/STD/CSSP/WPTGS(2015)23/en/pdf.

[7] OECD (2013), Interconnected Economies: Benefiting from Global Value Chains, OECD Publishing, Paris, https://doi.org/10.1787/9789264189560-en.

[35] OECD and Statistics Finland (2020), Globalisation in Finland: Granular Insights into the Impact on Businesses and Employment, OECD and Statistics Finland, https://doi.org/10.13140/RG.2.2.22456.16640.

[56] Rodrigues, J. and J. Rueda-Cantuche (2013), “A two-stage econometric method for the estimation of carbon multipliers with rectangular supply and use tables”, Ecological Economics, Vol. 95, pp. 206-212, https://doi.org/10.1016/j.ecolecon.2013.09.001.

[51] Rueda-Cantuche, J. (2011), “Econometric analysis of European carbon dioxide emissions based on rectangular supply-use tables”, Economic Systems Research, Vol. 23/3, pp. 261-280, https://doi.org/10.1080/09535314.2011.604768.

[54] Rueda-Cantuche, J. and A. Amores (2010), “Consistent and unbiased carbon dioxide emission multipliers: Performance of Danish emission reductions via external trade”, Ecological Economics, Vol. 69/5, pp. 988-998, https://doi.org/10.1016/j.ecolecon.2010.01.003.

[17] Saborío, G. (2015), Costa Rica: An Extended Supply-Use Table, Paper presented at the 2015 IIOA conference in Mexico City.

[27] Sallusti, F. and S. Cuicchio (2023), Towards the Compilation of ESUT for Italian Economy, Paper presented at the 2023 IIOA conference in Alghero.

[22] 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.

[37] Statistics Finland (2025), Value added trade principal indicators by enterprise ownership, 2017-2022.

[36] Statistics Finland (2024), “Trade in value added”, Experimental Statistics, https://pxdata.stat.fi/PxWeb/pxweb/en/Kokeelliset_tilastot/Kokeelliset_tilastot__tiva (accessed on 28 August 2024).

[55] Stehrer, R. et al. (2024), “Wrapping input-output multipliers in confidence intervals”, Journal of Economic Structures, Vol. 13/17, https://doi.org/10.1186/s40008-024-00331-4.

[34] Steller, A. et al. (2021), Costa Rica: Extended Supply and Use Tables and Non-financial Private Sector Accounts to Address Globalization, Central Bank of Costa Rica, https://www.bccr.fi.cr/indicadores-economicos/DocCuentasNacionales2017/ESUT_COSTA_RICA.pdf.

[50] Ten Raa, T. and J. Rueda-Cantuche (2007), “Stochastic analysis of input-output multipliers on the basis of use and make tables”, Review of Income and Wealth, Vol. 53/2, pp. 318-334, https://doi.org/10.1111/j.1475-4991.2007.00227.x.

[43] Teran-Vargas, J. (2024), Extended Supply and Use Tables of Mexico. Base Year 2018.

[1] United Nations (2018), Handbook on Supply, Use and Input-Output Tables with Extensions and Applications, United Nations, New York, NY, https://doi.org/10.18356/9789213582794.

[8] United Nations (forthcoming), Handbook on Integrating Business and Trade Statistics, United Nations, New York, NY, forthcoming.

[16] Yamano, N. et al. (2022), “Development of the OECD Inter-Country Input-Output Database: Sources and methods”.

[3] Yamano, N. and C. Webb (2022), Accounting for Within-industry Firm Heterogeneity in an Inter-country Input-output System, Paper presented at the 2022 IIOA conference in Langkawi, https://www.iioa.org/conferences/28th/papers/files/4466.pdf.

[10] Yang, C. et al. (2015), “Processing trade biases the measurement of vertical specialization in China”, Economic Systems Research, Vol. 27/1, pp. 60-76, https://doi.org/10.1080/09535314.2014.955463.

[33] Yang, C. et al. (2022), Extended Supply-Use Tables by Firm Heterogeneity for China, Paper presented at the 2023 IIOA conference in Alghero, https://iioa.org/conferences/29th/papers/files/4561.pdf.

[32] Yang, C., R. Wei and K. Zhu (2016), Compiling SUTs with Firm Heterogeneity: Methods and Some Empirical Results for the Case of China, Paper presented at the 2016 IIOA conference in Seoul.

Annex 1.A. Extensions of supply and use tables related to globalisation that are not extended supply and use tables

Ahmad (2023[48]) puts forward four extensions of regular supply and use tables (SUTs), relevant for measuring globalisation, which are not extended supply and use tables (ESUTs):

1. Estimates of goods for processing, merchanting or re-exports. Processing trade involves manufacturing services on physical inputs owned by others and/or the value of the goods imported, and the customs value of the processed goods exported. Merchanting is the purchase of goods by a resident (of the compiling economy) from a non-resident combined with the subsequent resale of the same goods to another non-resident without the goods being physically moved in and out of the compiling economy. Re-exports are goods produced in other economies, and previously imported, that are exported with no substantial transformation from the state in which they were previously imported.

2. Residents’ expenditure abroad and non-residents’ expenditure in the domestic territories.

3. Conversion of imports of goods from cost, insurance and freight price to free on board (FOB) prices. Multi-country SUTs typically require imports to be valued at FOB prices, just as exports already are in a regular SUT. Furthermore, information on tariffs/duties paid by product is also desirable to help build import matrices (using the proportionality assumption) in basic prices and analysing the impact of tariffs on global value chains.

4. Complementary information on import flow matrices broken down by partner. The geographical breakdown of the import matrix by country of origin within the SUT framework is an essential step to produce global IOTs and may sometimes be useful for the compilation of national SUTs, too.

SUT-based thematic accounts

SUTs provide a powerful foundation for the creation of thematic accounts. These can be focused on a wide range of issues from travel and tourism to arts and cultural production to education, health, and outdoor recreation. Construction of these accounts typically entails identifying specific industries and products within the SUT framework that are wholly or partially in scope for the topic of interest. SUT data are then reconfigured to highlight the portions of those industries and products relevant for the chosen topic.

An SUT that has been reorganised to support a thematic account may include industry decompositions and may be superficially similar to an ESUT in other ways as well. However, the primary goal of these accounts is to highlight activity in specific areas of the economy. Any split of activities by type of enterprise, which distinguishes an ESUT, is only incidental to that purpose.

Thematic accounts involve instead the rearrangement of existing information from National Accounts to enable an area of economic and/or social importance to be analysed in much greater detail, with additional dimensions (United Nations, 2018[1]). The 2008 System of National Accounts (SNA) distinguishes two types of thematic accounts. The first may be seen as an extension of the core National Accounts without changing the underlying concepts of the SNA in any fundamental way. These mostly cover accounts specific to certain fields, such as tourism, education and environmental protection expenditures. The second type is mainly based on concepts that are alternatives to those of the SNA, also possibly with changes in classifications, too. These include, for example, a different production boundary, an enlarged concept of consumption or capital formation, an extension of the scope of assets, etc. (United Nations, 2018[1]). However, none of them can be considered to be an ESUT.

Digital SUTs

The SUT framework has also been discussed extensively as a mechanism for organising information necessary to construct a digital economy thematic account. To this end, the OECD’s framework for Digital SUTs (OECD, 2023[49]) outlines how SUTs may be transformed into a digital SUT by adding and modifying various rows and columns. As with other SUT-based thematic accounts, a digital SUT has many superficial similarities to an ESUT. However, any split of industries by enterprise type is again only incidental to the primary purpose of highlighting digital activity in the economy. Therefore, a digital SUT is not an ESUT. See Chapter 9 for a more detailed discussion of digital SUTs.

Annex 1.B. Testing enterprise heterogeneity for extended supply and use tables

This annex outlines, for the first time in literature, an econometric test to help measure the degree of heterogeneity in an enterprise’s surveyed data used for the compilation of extended supply and use tables (ESUTs). The test helps ESUT compilers determine for every product whether a type of enterprise (e.g. multinationals) uses a statistically different technological production structure1 compared to the average of the industry. This will help countries to identify those industries that would need an additional breakdown by size, ownership, etc. and avoid breaking down all industries and/or products to construct ESUTs.

The first section provides the methodological background and context for an econometric test of enterprises’ data heterogeneity. The second section describes the econometric tests applied to supply and use tables (SUTs) in the literature, while the third section develops a new heterogeneity test for ESUTs. It concludes with a few remarks and recommendations for ESUT compilers.

Background and context

This section describes the methodological background and the context of a new econometric test of enterprises’ data heterogeneity. Enterprises’ data can be theoretically linked directly to input-output multipliers without the need to compile input-output tables (IOTs), as shown by ten Raa and Rueda-Cantuche (2007[50]). Later, Rueda-Cantuche2 (2011[51]) named this method the Supply and Use Based (SUBE) econometric approach but using SUTs instead of IOTs. This framework is appropriate for rectangular SUT-type of enterprises’ data with more enterprises/industries (broken down by size, ownership, etc.) than products, as may be the case for ESUTs.

By using enterprises’ data and appropriate regressions, input-output multipliers can be derived econometrically, thus, testing whether those multipliers are significantly different across enterprises in terms of size, ownership, exporter status, etc. In other words, this tests the product technology assumption that all enterprises produce one product using the same technology for each product.

Econometric input-output multipliers using enterprises’ data

This section discusses the theoretical way in which input-output multipliers (or the elements of the Leontief inverse matrix) are directly linked to enterprises’ data on their supply and use of products. For each surveyed enterprise, the data tell how much they produce of each product and how much they use of each product as intermediate input.

Annex Figure 1.B.1. From supply and use tables to input-output multipliers

Note: SUT: supply and use table.

As shown in Annex Figure 1.B.1, this approach proposed by ten Raa and Rueda-Cantuche (2007[50]) connected two different bodies of literature. The first body of literature discusses the methods for constructing IOTs from SUTs; the second discusses stochastic input-output analysis, the measurement of uncertainty in the technical coefficients and in the consequent Leontief inverse matrix. The integration permits input-output analysts to estimate directly economic impacts from the enterprises’ data3 underlying the SUTs, without the need to estimate IOTs. We now explain why.

Consider the Leontief quantity model (Miller and Blair, 2022[52]) for estimating demand-driven employment impacts, defining employment multipliers as the number of jobs generated per unit (million EUR, USD, etc.) increase of the final use of products. Mathematically,

$\mathrm{ }\mathrm{ }\mathrm{ }\mathrm{ }\mathrm{ }\mathrm{ }\mathbf{ }\lambda =l{(\mathbf{I}-\mathbf{A})}^{-1}$ (1)

where l and λ are the row vectors of labour coefficients4 and the employment multipliers, respectively. Bearing in mind that the product technology assumption implies that $\mathbf{A}={\mathbf{U}\mathbf{V}}^{-\mathrm{T}}$ and $\mathbf{l}={\mathbf{L}\mathbf{V}}^{-\mathrm{T}}$, replacement of A and l in (1) yields:

$\lambda ={\mathbf{L}\mathbf{V}}^{-\mathrm{T}}{(\mathbf{I}-{\mathbf{U}\mathbf{V}}^{-\mathrm{T}})}^{-1}=\mathbf{L}{\left(\left(\mathbf{I}-{\mathbf{U}\mathbf{V}}^{-\mathrm{T}}\right){\mathbf{V}}^{\mathrm{T}}\right)}^{-1}=\mathbf{L}{\left({\mathbf{V}}^{\mathrm{T}}-{\mathbf{U}\mathbf{V}}^{-\mathrm{T}}{\mathbf{V}}^{\mathrm{T}}\right)}^{-1}=\mathbf{L}{\left({\mathbf{V}}^{\mathrm{T}}-\mathbf{U}\right)}^{-1}$

And therefore,

$L=\lambda \left({\mathbf{V}}^{\mathrm{T}}-\mathbf{U}\right)$ (2)

By using as independent variables the rows of the so-called net output matrix (V^T- U), for each establishment, enterprise or unit, the following multiple linear econometric model5 is formulated:

$L_i={\lambda }_1{x}_{1i}+{\lambda }_2{x}_{2i}+{\lambda }_3{x}_{3i}+\cdots +{\lambda }_k{x}_{ki}+{\epsilon }_i$ $\mathrm{ }\mathrm{ }\mathrm{ }\mathrm{ }\mathrm{ }\mathrm{ }\mathrm{ }\mathrm{f}\mathrm{o}\mathrm{r}\mathrm{ }i=1,\mathrm{ }2\dots m$ (3)

where the respective regression coefficients (${\lambda }_i)\mathrm{ }$are input-output multipliers; k is the number of products; m is the number of surveyed establishments, enterprises or units; and ${\epsilon }_i\mathrm{ }$is an independent random disturbance error, assumed to be normally distributed with zero mean and constant variance. The theoretical model contains no constant term.

Testing heterogeneity in input-output multipliers with enterprises’ data

This section builds upon the econometric model presented in the previous section and elaborates on a new test to measure the statistical significance of enterprise heterogeneity in survey data for the compilation of ESUTs.

The econometric model presented in (3) in the previous section can be used for two different purposes:

• The estimation of consistent and unbiased (backward) input-output multipliers (e.g. employment, output, value added and/or emissions), and their confidence intervals. It would also allow testing the statistical significance of each individual input-output multiplier by product.

• Testing econometrically if the estimated input-output multipliers differ significantly between enterprises6 versus the product technology assumption (by which all enterprises/industries produce the same product with the same technology for each product).

It is precisely the second purpose that can be easily linked to ESUTs and that can be used to test the heterogeneity of the enterprises’ data underlying the compilation of ESUTs. The main idea is as follows: from (3), add one dummy variable ($z_{ki})$ multiplicatively for each product k. This yields the following econometric model:

$L_i={\lambda }_1{x}_{1i}+{\lambda }_2{x}_{2i}+{\lambda }_3{x}_{3i}+\cdots +{\lambda }_k{x}_{ki}+$

$+{\beta }_1{x}_{1i}\mathrm{ }{z}_{1i}{+{\beta }_2{x}_{2i}\mathrm{ }{z}_{2i}+{\beta }_3{x}_{3i}\mathrm{ }{z}_{3i}\cdots {+\beta }_k{x}_{ki}\mathrm{ }{z}_{ki}+\mathrm{ }\epsilon }_i$ $\mathrm{ }\mathrm{ }\mathrm{ }\mathrm{ }\mathrm{ }\mathrm{ }\mathrm{ }\mathrm{f}\mathrm{o}\mathrm{r}\mathrm{ }i=1,\mathrm{ }2\dots m$

Assume now that we are interested in evaluating whether, for a particular product (say product 1), establishments/enterprises/units belonging to multinational enterprises (MNEs) do have a statistically different impact in terms of employment with respect to other non-MNE units producing product 1. In doing so, the interpretation of the regression coefficients is:

• ${\lambda }_1{x}_{1i}=E\left(L_i{z}_{1i}=0\right)$; expected employment impact (number of jobs) due to increases in the final use of product 1, being produced by non-MNE enterprises/units/establishments.

• $\left({\lambda }_1+{\mathrm{ }\beta }_1\right)x_{1i}=E\left(L_i{z}_{1i}=1\right)$; expected employment impact (number of jobs) due to increases in the final use of product 1, being produced by enterprises/units/establishments belonging to MNEs.

• ${\mathrm{ }\beta }_1{x}_{1i}=\mathrm{ }E\left(L_i{z}_{1i}=1\right)-E\left(L_i{z}_{1i}=0\right)$; expected difference in the employment impact (number of jobs) due to increases in the final use of product 1, between those produced by enterprises/units/establishments belonging to MNEs and those produced by non-MNE units.

By testing the null hypothesis7 ${\mathrm{ }{H}_0:\beta }_1=0$, we will test whether MNEs and non-MNEs have the same employment impact per currency unit (e.g. million USD) in producing one additional unit of product 1, with a pre-fixed confidence level, e.g. 95%. The same applies to all other products, one for each regression coefficient $\beta .$ Moreover, one can test whether MNEs and non-MNEs can have the same employment impacts for all products by testing the following null hypothesis8 ${\mathrm{ }{H}_0:\beta }_1=\cdots ={\beta }_k=0$.

Concluding remarks

Compilers of ESUTs need to choose the dimension (size, ownership, exporter status, etc.) to break down industries. Enterprises’ surveyed data and econometric analysis based on an SUT framework can help in selecting the industries requiring a break down. This choice is related to their statistically significant differences in product technologies and resulting different economic impacts driven by increases in the final use of products. This annex aims to inspire ESUT compilers to test enterprise heterogeneity with surveyed data to help with the decision about which industries/products to break down. Belgium, Mexico and the Netherlands have shown interest in following this approach but have not been able to work on it yet.

Notes

← 1.  A different technological production structure leads to a different matrix of technical coefficients as well as to a different Leontief inverse matrix and, therefore, to different input-output multipliers and different economic impacts.

← 2.  Rueda-Cantuche (2011[51]) extended this approach to SUTs, which were considered aggregated versions of the underlying enterprises’ data when the enterprises’ data were not available. Other relevant literature using SUTs are Rueda-Cantuche and Amores (2010[54]), Rodrigues and Rueda-Cantuche (2013[56]) and more recently, Stehrer et al. (2024[55]) with panel data econometrics. Overall, the use of enterprises’ data is more appropriate than the use of SUTs to avoid possible aggregation bias.

← 3.  The main caveat of this approach is that only statistical offices typically have access to the necessary enterprises’ data.

← 4.  As for technical coefficients, labour coefficients result from dividing employment (L) over the total output by industry. A is the technical coefficients matrix and I the identity matrix. U stands for the intermediate use matrix (product by industry) and V stands for the transposed of the supply table (product by industry) – also denoted in the literature as the make table. ^T denotes transposition and ^-1 the inverse of a matrix, being ^–T the inverse of a transposed matrix or vice versa.

← 5.  Further details of this econometric approach with enterprises’ data can be found in ten Raa and Rueda-Cantuche (2007[50]). In particular, it is important to note that the data entering such an econometric model should be transformed into basic prices, as described in the appendix of the same article (pp. 332-334). For further details of SUBE, see Rueda‑Cantuche (2011[51]).

← 6.  It can also be industries, following the SUBE approach developed by Rueda-Cantuche (2011[51]). However, aggregation may be an important caveat.

← 7.  This individual test is well known in the econometric literature and uses a Student’s t-distribution (Greene, 2019[53]).

← 8.  This test is well-known in the econometric literature and uses a F-Fisher distribution (Greene, 2019).