Several countries added information from other domains to ESUTs and EIOTs. These additions vary from value added and income flows to employment, physical ESUTs and emissions. The chapter also provides suggestions about the integration of ESUTs and productivity by industry statistics, regional ESUTs and digital SUTs.
Handbook on Extended Supply and Use Tables and Extended Input‑Output Tables
Abstract
Extended supply and use tables’ (ESUTs) and extended input-output tables’ (EIOTs) contribution to economic analysis can be further extended by adding information from other domains. This chapter provides several examples from theory and practice, from different countries, providing an international perspective on (possible) implementation of these extensions. These topics can be seen as inspiration; the relevance of each topic for policy makers and other users will vary from country to country. The chapter starts with a link to income flows since value added in the domestic economy, in particular that generated by multinational enterprises (MNEs) or their affiliates, can flow abroad as property income. The chapter then considers ESUTs’ and EIOTs’ links to employment data, providing a means to illustrate how employment is related to the activities of different types of enterprises, including, in particular, their exports, which remain an area of high policy interest. This is followed by two examples from the environmental realm, namely, looking at ESUTs through prisms of material efficiency and emissions intensity. For these examples, experiences from practice are shared: why these topics were deemed relevant, which data and methods were used, and the results obtained. The chapter concludes with three examples that are either more theoretical (since they have not yet been carried out in practice) or strongly related to ESUTs. The first considers the integration of ESUTs with KLEMS-type productivity frameworks. The second looks at regional extended supply and use tables (ESUTs) (i.e. national ESUTs disaggregated into regional components). The third example considers digital SUTs, which, although technically not ESUTs, are closely related in many aspects.
Value added is often considered when thinking about economic impact. It is well-known how domestic production of a given industry, a type of enterprise or type of final expenditure translates into value added. However, part of that value added may flow abroad through increased foreign direct investment (FDI) and internationalisation of ownership. This can happen directly, for example because foreign-owned MNEs transfer profits to the parent enterprise abroad. It can also happen indirectly, for example because a small and medium-sized enterprise (SME) pays interest to a bank that uses this income to pay out dividends to foreign stockholders. As a consequence, national value added related to activities in the domestic economy is not the same as national income related to the same activities, just as gross domestic product (GDP) is not equal to gross national income (GNI).
For policy makers, it is relevant to know how much national income is related to activities of each industry, different types of enterprises and different types of final expenditure. For example, governments worldwide have limited resources available to stimulate the economy. They can aim to stimulate household consumption, government consumption, investments and exports. They might consider specific goods and services that are “greener” or related to specific industries. They could also attract foreign MNEs. In all cases, domestic value added in the form of wages and profits is generated, but how much of those profits remain in the country and how much leave the country in the form of income transfers? This is a difficult question; it is not surprising that there is a broad literature on fiscal multipliers and international linkages (see, for example, the literature review in Botev and Mourougane (2017[1])). Yet, an analysis using an EIOT and information about income flows would provide valuable new information for the policy debate.
This section will discuss data, methods and examples showing how to achieve this.
To set the scene, there are various types of primary income flows (codes in National Accounts in brackets):
Interest before adjusting for implicit financial services on loans and deposits. These implicit services were called financial intermediation services indirectly measured (FISIM) in the 2008 System of National Accounts (D.41a).
Dividends (D.421).
Withdrawals from the income of quasi-corporations (D.422).
Reinvested earnings on FDI (D.43).
Investment income attributable to insurance policy holders (D.441).
Investment income payable on pension entitlements (D.442).
Dividends attributable to collective investment fund shareholders (D.4431).
Retained earnings attributable to collective investment fund shareholders (D.4432).
Rent (D.45).
Adjustments for implicit financial services on loans and deposits (P.119C).
Remuneration of cross-border employees (D.1, possibly including both wages and salaries in cash or in kind (D.11) and employer’s actual and imputed social contributions (D.121 and D.122) (see Eurostat (2023[2])).
In practice, data on dividends, reinvested earnings on FDI and compensation of cross-border employees will be most prominent.
Public data sources
There are various public data sources for income flows. The most common is FDI statistics, such as FDI income payments about reinvested earnings by industry, as seen in OECD (2016[3]; 2016[4]; 2015[5]). The authors use the total income payments rather than only the distributed earnings, because it would better reflect how much of the domestic value added can be decided upon by foreign MNEs. Bohn, Brakman and Dietzenbacher (2021[6]) use the World Bank’s Bilateral Remittances Database for flows of labour income. They also use bilateral data on direct and portfolio investment positions from different sources to map capital income flows. These are provided by the OECD, the International Monetary Fund (IMF) (Coordinated Portfolio Investment Survey), and UN Trade and Development (UNCTAD). UNCTAD’s data concern the ultimate, not the direct, investors on a bilateral basis (Casella, 2019[7]). Meng et al. (2022[8]) use several data sources to estimate bilateral FDI flow and stock data by industry. These are the Global Trade Analysis Project bilateral multi-region multi-sector FDI stocks database; UNCTAD’s bilateral FDI stocks; the IMF Coordinated Direct Investment Survey with detailed data on FDI stocks; OECD FDI stocks by country by industry; and the Chinese Statistical Bulletin with the People’s Republic of China’s (hereafter “China”) outward FDI stocks by industry.
Meng et al. (2022[8]) estimate data for 35 economies (including “Rest of World”) and 20 industries. OECD (2016[3]) and Bohn, Brakman and Dietzenbacher (2021[6]) provide results for 27 and 42 countries, respectively. Data are, thus, widely available.
Duan et al. (2021[9]) consider China alone, although their methods might be applicable to other countries. They rely on foreign capital shares in stocks by industry and by type of enterprise (domestic enterprise or foreign invested enterprise) and use this to split capital income into national and foreign capital income. Their data sources are the China Industrial Economy Statistical Yearbook, the China Economic Census Yearbook and the China Statistical Yearbook. Duan et al. (2021[9]) note that the share of labour income remaining in China is expected to be very high since the United Nations Global Migration Database shows the share of foreigners in the total population of China is very small. This might be similar in other countries.
Confidential data sources
A national statistical office will have more detailed information available, such as data used for or produced from the Sector Accounts. These describe monetary flows between different sectors in the economy and abroad, providing a quantitative description of several coherent economic processes in an economy. These economic processes are described for groups of units with a more or less homogeneous economic behaviour: the sectors. The main sectors are non-financial corporations, financial corporations, government, households and non-profit institutions serving households (NPISH). The Sector Accounts show not only the economic relations of these sectors with each other but also the transactions of the domestic sectors with the “Rest of World”. The following processes are distinguished: production, income generation, income distribution, income redistribution, income expenditure, capital formation and financing.
The compiling process of the Sector Accounts uses a large number of sources from different areas such as production, final expenditure, income and finance. Examples include statistics on balance sheets and profit and loss accounts of non-financial corporations, reports on financial corporations from the national central bank, government administration of ministries and municipalities, budget surveys, and international trade and balance of payments statistics. These data sources could be useful for income analysis in an ESUT/EIOT setting as well. It is advisable to stay as close as possible to the Sector Accounts. Their underlying system is an integration framework where the different data sources are confronted and integrated, just like the National Accounts.
This section will now describe the data sources by sector in more detail.
For non-financial corporations (S11), the most relevant flows are interest and dividends paid to or received from abroad, retained earnings on FDI, and income received from foreign subsidiaries. These may be available through surveys of statistics of finances, (corporation) tax data, annual reports of housing corporations or in declarations of listed firms.
Financial corporations (S12) often account for the majority of property income flows in an economy and with “Rest of World”. Surveys, conducted either by the national statistical office or the national central bank, can provide information about interest and dividends. Sometimes information about ownership of debt securities or quoted equity is available. Several types of income can be easily linked to industries. For example, investment income attributable to insurance policy holders is related to insurance and pension funding (industry 65 in ISIC Rev. 4). Some of the interest is related to output of services for owner-occupied dwellings.
Flows between general government (S13) and its counterparts might be found in annual reports of the ministries. Similar information might be found in annual reports of municipalities. Ideally, such information is split by policy area (e.g. energy networks, usage of public pavement, maintenance of public transport), which allows for assigning to industries.
The property income transactions of households (S14) and NPISH (S15) are mostly with financial corporations. Other relatively large flows are dividends, withdrawals from the income of quasi-corporations abroad and retained earnings as collective investment fund shareholders. It is assumed that these are received by households in their capacity as investors/consumers.
Compensation of cross-border employees (non-residents) might be estimated using a linked employer-employee database, wage information and the population register.
The general guidelines, “use what you have and use common sense”, apply. Ideally, data are available. Otherwise, discussions with experts might provide information. When the data do not fit immediately, the largest discrepancies should be solved by hand and general methods (such as the GRAS method, an iterative scaling method described in Chapter 5) should be used to solve smaller discrepancies.
Methods
When considering FDI income flows in an input-output setting, it is advised to use an EIOT, since FDI income only concerns MNEs. However, there are other ways as well.
OECD (2016[3]; 2016[4]; 2015[5]) all combine trade in value added (TiVA) data with information at industry level about output and exports by domestically owned enterprises and foreign enterprises in the domestic economy. They also use FDI income data by industry. They first estimate the export intensity of foreign-owned enterprises by dividing their exports by turnover, at industry level. They then multiply this ratio with the FDI income in that industry. This yields an estimate of the FDI income related to the direct exports of foreign-owned enterprises in that industry. This is income that a country earns from its exports but leaves the country anyway. It can be compared to the direct domestic value added of the industry to put it into perspective.
Bohn, Brakman and Dietzenbacher (2021[6]) link income in one country to consumption in others. They first construct a GDP-GNI matrix, showing how value added in country A ends up as income in country B. They do that by estimating how much value added stays in the own country, and using bilateral information about workers remittances (from the World Bank) and investment data to split the value added that leaves the country. In this way, production in country A leads to value added in country A, which leads to income in country B. After the construction of the GDP-GNI matrix, they combine it with calculations of value-added exports to estimate the income of each country that is embodied in foreign final use. In other words, consumption in country A is traced back to production (and value added) in country B, which leads to income in country C. This yields a country’s “GNI footprint”, showing the income that country C has due to consumption in country A.
Meng et al. (2022[8]) use the OECD Analytical Activity of Multinational Enterprises data set, a time series of multi-country IOTs where each industry is split into domestically owned enterprises and foreign-owned ones. They use it to estimate trade in value added by type of enterprise. Using different sources of FDI data, they estimate a very detailed matrix with bilateral FDI stocks and flows, by industry at country level. Given a host country and an industry, this shows the size of the investing country and its returns to capital, which allows linking production with value added and capital income, by host country, industry and investing country. Combining this information with the estimates of bilateral trade in value added, Meng et al. (2022[8]) arrive at estimates for “trade in factor income”. Among others, they obtain estimates for the US-owned factor-income induced by China’s final use.
Duan et al. (2021[9]) use tripartite IOTs for China. In these tables, each industry is split into three parts: 1) domestic enterprises meeting domestic demand; 2) enterprises in processing exports; and 3) enterprises involved in ordinary exports and/or foreign owned. In the third category, in each industry, capital income to output ratios and the share of “paid-in capital” (the asset value of enterprises) held by foreign shareholders are known. For RMB 1 of output in a given industry, they first derive the corresponding capital income. This is split into a foreign-owned and domestically owned part using the share of foreign shareholders in total paid-in capital. Together with an input-output analysis, this yields estimates of income related to exports that is flowing abroad/staying at home.
Lemmers (2023[10]) points out that besides income related to production, there is also income related to property, e.g. in financial services. Consequently, part of the outflow is financed by the inflow, and part of the outflow is financed with value added. Furthermore, value added due to production at a domestically owned enterprise may end up, via a bank and its profits, as dividends distributed to foreign and domestic entities. Figure 9.1 shows such income flows.
Source: Lemmers (2023[10]).
This approach first compiles an IOT where each industry is split by type of MNE. It then compiles tables of income flows between each type of MNE in each industry, several sectors and “Rest of World”. This flow of funds table is similar to an EIOT, as shown in Table 9.1.
The starting point is the integrated sector-to-sector property income matrix for 2015 from Dutch sector accounts. For both the use and supply of property income, there is limited information on counterparties on a detailed enterprise group to enterprise group basis. Therefore, in many instances, the production value (P.1) of enterprise groups was used to distribute property income flows to the counterpart industry and type of enterprise group. Since a large part of income flows occurs via the financial sector, pass through flows from Rest of World to the country and back to rest of the world that are not related to the domestic production process are problematic. These payments should, therefore, be removed from the data.
Examples of results
Table 9.2 provides an example of the results from OECD (2016[3]). It shows how much direct domestic value added in exports for six countries was expatriated via FDI payments. For example, Argentina had over USD 45 billion of direct domestic value added (DDVA) embodied in exports, i.e. by the enterprises that export directly. The total FDI income payments amounted to USD 11 billion. Since the average export intensity of enterprises was 0.18, 0.18 * 11 billion equals USD 2 billion of those FDI income payments that were related to the domestic value added of direct exporters. This means that 4% of domestic value added due to direct exports is expatriated. Note that this share is higher for countries where exports have a higher share in production and FDI income payments are relatively high compared to domestic value added (e.g. Belgium and China). For most countries (not shown in the table), the share of FDI income payments in DDVA in exports was below 10%. Exceptions were Estonia, Hungary, Ireland, Luxembourg and the Slovak Republic. The share was over 40% for Luxembourg due to its very open economy.
Lemmers (2023[10]) uses information for the Netherlands in 2015 and obtained three main messages. First, a substantial part of value added, 15%, flows abroad. Second, a substantial part first flows to a domestic entity and only then leaves the country. It is not only value added of foreign MNEs, but also of domestically owned enterprises. Third, there are substantial differences between industries and types of enterprises. It is necessary to accommodate this heterogeneity in the analysis.
As shown in Table 9.3, non-MNEs had a total value added of EUR 199 billion. Of this, EUR 120 billion was embodied in domestic final use and ultimately ended up in the Netherlands. EUR 3 billion was embodied in domestic final use and was transferred abroad via income transfers. EUR 74 billion was embodied in exports and ultimately ended up in the Netherlands. EUR 2 billion was embodied in exports and transferred abroad via income transfers. Note that there is substantial heterogeneity between the different types of enterprises. Of the value added of foreign-owned MNEs, about a third ultimately ends abroad. It is only about 2.5% for non-MNEs and 10% for domestically owned MNEs. This is not surprising as foreign MNEs will transfer profits abroad to the parent enterprise. Domestically owned enterprises do not have foreign parent enterprises. Dutch MNEs might pay dividends to foreign entities, but non-MNEs are less likely to do so. However, domestically owned enterprises, whether MNEs or not, will pay interest to banks that may pay dividends to foreign entities.
There is considerable interest in globalisation’s impact on jobs, especially in developed countries where labour costs can be high relative to developing economies. A key concern in this respect reflects labour-intensive activities in developed economies, where higher use of capital to compensate for relative price differences in labour costs can have only a marginal impact. The initial wave of global value chains (GVCs) saw many firms capitalise on labour savings through international outsourcing, resulting in declines in the labour-intensive, and typically lower skilled, parts of their workforces. More recent years have seen an acceleration of these trends to other, higher, parts of the value chain, often affecting higher skilled workers. Export-oriented enterprises typically hire more highly skilled workers, and their exports engage disproportionally higher skilled employees from upstream enterprises. Accessing foreign markets is not only costly due to existing export barriers, but enterprises are also subject to more competition and need to invest significant resources to stay ahead of the game, their workforce being a pivotal resource in this regard. Outsourcing enterprises have been able to improve their international competitiveness, increase exports and, consequently, generate jobs through specialisation in higher parts of the value chain. New enterprises have been created precisely because they were able to specialise in niche activities that international fragmentation has provided. Within this churning of jobs, there have been winners and losers. Even if the losers may only be temporary, as workers reskill to engage in new activities, this churning is an important element behind the broad backlash against globalisation. This is an important reason why better information on the effects of GVCs on employment is needed.
There are significant differences in how workers of different genders or different skills participate in the labour force. In addition to the segregation being seen as a problem in itself, it also affects, for example, how different types of employees take part in GVCs, how well different employees can take advantage of the benefits of globalisation and how economic shocks will affect employees.
Country example: Belgium
In a small open economy like Belgium, precise information on the contribution of exports to employment is of great value for policy makers. This contribution encompasses jobs in export-producing enterprises as well as jobs in domestic suppliers in the export value chain. Within this framework, a precise estimation of employment embodied in exports should account for heterogeneity between exporters and non-exporters, particularly in terms of employment intensity, productivity and input structures. Furthermore, this provides an estimate of the bias that would appear if one does not account for this heterogeneity.
Data and methods
Michel and Hambÿe (2022[11]) present calculations of export-related employment based on an export-extended IOT and employment data for Belgium in 2010.1 These data break down manufacturing industries into export-oriented enterprises and domestic market enterprises. Export-oriented enterprises are those that sell at least 25% of their output to foreign markets.
The disaggregation of industry-level manufacturing employment from the national accounts into export- and domestically oriented enterprises is based on enterprise-level data from the National Social Security Office. This administrative data source provides near-comprehensive information on the number of employees, which make up the bulk of manufacturing employment (more than 95%). When linked to the business register through a unique identifier for all manufacturing enterprises, it allows to determine the shares of export-oriented enterprises in total industry-level employment. Hence, for this disaggregation, it is implicitly assumed that the distribution of self-employed by exporting intensity is the same as for employees in each manufacturing industry. An additional disaggregation by skill category is obtained through enterprise-level data on educational attainment from the “social balance sheet”, which is a separate section of the annual accounts.
Results
According to the calculations with the exporter-extended IOT, export-related employment amounted to 1.32 million jobs, or 29.5% of total employment in Belgium. Manufacturing exports sustained 0.59 million jobs in Belgium, of which about half were directly in the exporting enterprises and half were indirectly in domestic supplier enterprises. With regular IOT and employment data, the number of jobs sustained by manufacturing exports would have been overestimated by 4%. This is essentially an overestimation of low-skilled jobs sustained by exports. The underlying reason for this overestimation is that the labour intensity, value ‑added-to-output ratio and import propensity are higher for domestically oriented manufacturing enterprises than for export-oriented manufacturing enterprises.
Digging deeper into the extended data, and presenting results by type of enterprise, reveals that exports of export-oriented manufacturers generate a substantial number of jobs in non-manufacturing enterprises. These are, in particular, the service providers (229 000 jobs). The reverse relationship does not hold, as exports of enterprises in service industries generate jobs almost exclusively in these industries and only 6 000 jobs at export-oriented manufacturers. This is illustrated in Table 9.4.
Overall, the results show that the combination of an EIOT and employment data allows us to confirm that within industries, the production process of export-oriented manufacturers is different from that of other enterprises and to improve the estimation of export-related employment.
Country example: Finland
Finland is a small open economy with significant exposure to foreign trade. Better knowledge of how trade affects the domestic economy is pivotal to ensure good policy making. For example, the Ministry of Economic Affairs and Employment faces questions regarding employment and economic growth. It is, therefore, necessary to delve deeper into employment structures and GVC dependencies of enterprises and employees. For instance, does every type of gender or every type of skill benefit in the same way from domestic and foreign demand?
Data and methods
During a co-operative project in 2019-20, Statistics Finland and the OECD developed more information on employment and GVCs. First, they set up a system to compile regularly EIOTs. This facilitated calculating standard trade in value added indicators by type of enterprise. They incorporated employment statistics in the same system by creating a linked employer-employee database. The necessary data come from employee-level register data, containing employee characteristics such as age, gender and level of education. This enables detailed analyses, e.g. allowing for estimations of GVC dependencies of different groups of employees by enterprise type. Such employment indicators are included in Statistics Finland’s experimental trade in value added publication (Statistics Finland, 2024[12]).
Employees are divided into three groups based on their level of education: low skill, medium skill and high skill. The groupings are based on the International Standard Classification of Education maintained by the UNESCO Institute for Statistics. The education classifications are explained in detail in UIS (2012[13]). Employees with education levels in early childhood education (code 0), primary education (code 1), lower secondary education (code 2) or unknown level of education (code 9 or no code) are considered low-skilled employees. Employees with education levels in upper secondary education (code 3) or post-secondary non-tertiary education (code 4) are considered medium-skilled employees. Employees with education levels in short-cycle tertiary education (code 5), Bachelor’s or equivalent level (code 6), Master’s or equivalent level (code 7), or doctoral or equivalent level (code 8) are considered high-skilled employees.
Results
Finland has one of the smallest gender gaps in employment rates among OECD countries (OECD and Statistics Finland, 2020[14]), but sectoral and industrial segregation is a common topic in public discussions. The traditionally important role of manufacturing enterprises in GVC integration, simultaneously a reason for the discussion on gender segregation in the labour market, is reflected in the GVC dependency of employees. Over 70% of the employees dependent on exports are men (Figure 9.2). In total, more than 500 000 Finnish workers are either directly or indirectly dependent on foreign demand for their employment. The number remained fairly stable between 2013 and 2018, declining slightly at the beginning of the period but climbing back up over the last two years.
Source: Lindroos and Myllymäki (2021[15]).
The difference in participation rates between male and female workers in different sectors and industries is not only reflected in how many men and women participate in GVCs, but also in the way they participate (Figure 9.3). As women are disproportionately employed in non-exporting enterprises, their direct contribution to Finland´s exports is significantly lower than that of their male counterparts. However, non-exporting enterprises employ relatively more female workers. These enterprises provide upstream goods and services to exporting enterprises, which leads to jobs of female employees embodied in indirect exports.
Note: Private sector excluding agriculture (A), finance and insurance (K), real estate (L), education (P), health and social work (Q), and part of other service activities (S).
Source: Lindroos and Myllymäki (2021[15]).
Manufacturing remains the most important gateway to foreign markets, but service industries provide pivotal inputs through domestic value chains. This is also reflected in the export dependencies of employees by industry, as can be seen in Figure 9.4. But the Finnish economy has steadily transitioned towards a stronger dependency on service production over the last decade. Employment has decreased in several manufacturing industries while service industries are well represented among the industries showing strong employment growth (OECD and Statistics Finland, 2020[14]). The transition towards more service-based production indicates an upcoming change in GVC participation by gender. In 2018, women represented 57% of the workforce in service industries.
Source: Lindroos and Myllymäki, (2021[15]).
Besides the gender dimension of employment, the skill dimension is relevant as well. The Finnish labour force has undergone a steady transition towards more medium- and high-skilled employees. In 2018, about 35% of the Finnish labour force was high-skilled and 54% medium-skilled, having seen increases of about 3 and 5 percentage points since 2008, respectively. During the ten-year period, the number of employees classified as low-skilled declined by more than 100 000 in the private sector and the decline was spread broadly across industries.
The lean towards more educated workers is especially visible in Finnish two-way traders, but the workforce structure in exporters and importers is also significantly different from enterprises that do not participate in foreign markets. Two-way traders employed, at 49%, proportionally more high-skilled workers as a share of their workforce than other enterprises in 2018. The share of low-skilled employees in two-way traders was also noticeably small compared to other enterprises. The share of low-skilled employees is fairly even in other enterprises, but exporters and importers hire proportionally more high-skilled employees than non-traders.
Note: Private sector excluding agriculture (A), finance and insurance (K), real estate (L), education (P), health and social work (Q), and part of other service activities (S).
Source: Lindroos and Myllymäki, (2021[15]).
The transition towards more service-based production is also likely to affect the skills composition of exporting enterprises in the coming years. Knowledge-based activities such as information technology and information services, which accounted for over 15% of growth in value-added exports between 2013 and 2019 hired proportionally more highly educated employees than many traditional manufacturing industries. The contribution of service industries to exports is mostly indirect. But these industries continue to strengthen their role in Finnish direct exports as well.
International institutions such as the United Nations and the European Union see the transition to a more circular economy as a means to achieve a more sustainable economy (UNECE, 2021[16]; European Commission, 2021[17]). However, moving part of the production process abroad via international trade might lead to a less sustainable economy. It might lead to more pressure on natural resources and more pollution. It is also possible, however, that globalisation could lead to more efficient production processes, development and diffusion of new resource-friendly technologies (Fortanier and Maher, 2001[18]). The question is what net effect will these developments have on the economy? It is then necessary to know which materials are used by which industries and in what quantities.
Besides environmental policies, policies concerning secured supply of and dependency on materials are on many agendas as well. Materials such as lithium, cobalt, gallium and germanium are necessary for the energy and/or digital transition. Some countries might extract these materials themselves; many others are dependent on imports. The COVID-19 pandemic and growing geopolitical instabilities showed vulnerabilities and dependencies of countries when it concerned materials. The European Commission (2023[19]) developed policies to diminish the dependency on critical raw materials from certain countries. Again, it is necessary to know which materials are used by which industries and in what quantities.
Physical supply and use tables contain exactly that information. They capture extraction, imports, production, consumption, exports and other use at product level. They also contain flows of emissions and waste, while covering the entire economy. The System of Environmental Economic Accounting (United Nations, 2012[20]), describes how to compile such tables. Since it is firmly embedded in national accounts, it allows the development of indicators using both the physical and the monetary information such as resource efficiency and resource dependency. Among others, it enables effective monitoring of the circular economy and other resource policies at macro and meso level, which is necessary, since policies can be very product-specific. This leads to a better understanding of the location of inefficiencies/dependencies and how material use drives other environmental fields such as emissions.
It is meaningful to compile physical ESUTs as well, for several reasons. First, international trade plays an important role, and large enterprises, MNEs and traders are more involved in international trade than SMEs (small and medium-sized enterprises), non-MNEs and non-traders. Second, it is expected that different types of enterprises have different production processes. If certain types of enterprises are less resource-intensive, there might be best practices to be learnt for the other types of enterprises towards more dematerialisation. Third, it is relevant to know which type of enterprise is involved with each product since it may indicate vulnerabilities. For example, Germany and Poland nationalised natural gas firms tied to the Russian Federation (AP, 2022[21]).
Country experience: The Netherlands
Graveland (2016[22]) describes an approach to derive ESUTs in physical terms for the Netherlands for the reporting year 2012. He uses the existing material accounts for the Netherlands, a SUT in physical terms, as the starting point. Subsequently, he uses employment of domestically owned and foreign-owned enterprises to split material supply and use in an industry by ownership. While this approach is generally not recommended, as it relies on strong assumptions that will be explained in the methodological section, this example provides information about the magnitude of the involvement of foreign MNEs. Graveland estimated that about a third of material supply and use for Dutch production is by foreign-owned enterprises. This is concentrated in specific industries and corresponding products. The following sections describe his data, methods and results in more detail.
Data
The starting point of the analysis is the material flow monitor of the Netherlands in 2012. These are SUTs in physical (kilogramme) instead of monetary (EUR) terms. The tables contain about 300 products and 125 industries. The material flow monitor is based on existing national economic-environmental accounting statistics and is updated every two years.
The information about employment, production and foreign ownership was obtained in a similar way as described in the earlier chapters. The General Business Register provided the industry classification and the employment of each enterprise. Production was derived from the Structural Business Statistics (SBS) survey (for the enterprises in the survey), from estimates using this survey (for the enterprises in the SBS domain but outside the survey) and by combining employment with the ratio employment by production for enterprises outside the SBS. The delineation between domestically owned/foreign-owned enterprises was obtained from the Inward FATS (Foreign Affiliates Statistics). It compiles the UCI, the Ultimate Controlling Institutional Unit, which is the enterprise, up the chain of the Dutch enterprise’s control, which is not under the control of another enterprise, and the country of the UCI. If the country tagging is the Netherlands, the enterprise is domestically owned, if this country is not the Netherlands, the enterprise is foreign-owned. Sources for the country of the UCI are Statistics Netherlands’ surveys (such as the financial statistics of large non-financial enterprises or the Community Innovation Survey) and external sources (such as the Dunn & Bradstreet Database).
Methods
The previous chapters have shown how to compile ESUTs using a bottom-up approach (similar to the compilation of the regular SUTs) and a top-down approach (using information about production, value added, imports and exports at total level). The approach used in this study uses yet another method, namely splitting an industry into foreign-owned and domestically owned parts by employment or production. For example, if the foreign-domestic proportion in employment at metal manufacturing is 40/60, this method splits the material supply and use in metal manufacturing of each product by 40/60 as well. To arrive at the proportions, employment and production were aggregated by industry and ownership.
This method is not recommended, as it uses two very strong assumptions about input and output for the production process. First, it assumes that the inputs for production of the two types of enterprises are the same (with a possible exception for value added) in each industry. They use the same products in the same proportions. Second, they are assumed to produce the same products in the same proportions as well. If one were to compile an extended physical IOT from these ESUTs, without any extra information, even the origin and destination (domestic/foreign) of input and output of the two types of enterprises would be the same in each industry. Therefore, results only show the possibilities of physical ESUTs and a rough order of magnitude.
Results
This section contains two examples of the results. See Graveland (2016[22]) for more examples.
Figure 9.6 shows that material use is concentrated in a few industries. These are construction; manufacture of coke and petroleum; manufacture of food, beverages and tobacco products; chemistry; and pharmaceuticals. The service industry uses a large quantity of materials as well, but this is not surprising given that the industry encompasses 80% of the Dutch economy.
Ownership of the materials varies significantly among industries. There are hardly any foreign-owned construction enterprises in the Netherlands, hence the foreign share of material use in this industry is low. However, in manufacture of coke and petroleum, and of chemistry and pharmaceuticals, foreign-owned enterprises have a majority. This holds for several other industries as well, e.g. the manufacture of basic metals and the supply of energy and water.
Source: adapted from Graveland (2016[22]).
The total share of foreign-owned enterprises in Dutch material use was 36%. This is much greater than their shares in domestic production and employment. The reason is that they have substantial shares in material-intensive industries, whereas they are relatively small in material non-intensive industries such as services. Services also include government services, education and healthcare services, all with a low (or zero) share of foreign ownership and a low material intensity.
The total share of foreign-owned enterprises in Dutch material supply was 39%. As explained previously, foreign-owned enterprises have large shares in several material-intensive industries. They use a sizeable share of total materials in the Dutch economy, and therefore produce (supply) a large share as well. Differences in the share in supply and use are because the ESUT was not balanced. Graveland (2016[22]) provides detailed results for material supply.
The type of products used by domestically owned/foreign-owned enterprises is strongly related to the industries where they are active. Figure 9.7 shows that products with the most weight in the Dutch economy are fossil energy carriers, other mining products (such as sand and ores), coke and refined petroleum products, and chemical and pharmaceutical products. With the exception of other mining products, foreign-owned enterprises use at least half of the quantities of these four products.
Source: adapted from Graveland (2016[22]).
The supply of Dutch-produced materials is similar, since this is related to the industries as well. For example, a large part of the manufacture of coke and petroleum is foreign-owned. In the Netherlands, this industry transforms fossil energy carriers to coke and refined petroleum products. Hence, supply of coke and refined petroleum products will be foreign-owned for a large part as well.
Other possible indicators
Following Delahaye et al. (2015[23]); Hanemaaijer et al. (2021[24]); and Delahaye, Tunn and Tukker (2022[25]), this section points out other indicators related to resource efficiency, resource dependency, material substitution, recycling and environmental impact, respectively.
Using natural resources efficiently is necessary since they are limited, and extraction might cause harm. For enterprises, efficiency is necessary since it reduces costs and dependencies. It is considered best to consider resource efficiency at industry level. Material intensity is defined as the ratio of material use (in kilogrammes [kg]) to value added. Material productivity is defined as the ratio of the quantity that is produced with 1 kg of materials. Differences between types of enterprises can be due to, among others, different production processes or use of different materials. Both types of differences can lead to sharing best practices.
Countries will usually not produce every single product; they may lack ores or fossil fuels and therefore be resource-dependent. Sometimes substitution is possible, but often it is not. For a given product, a country is independent from others when import plus extraction minus export is covered completely by domestic material extraction and/or production. It is relevant to distinguish between imports of raw materials, intermediate products and final goods, since this indicates which stage of the production process takes place in the own country. A material footprint of imports is calculated as the amounts of raw materials necessary to produce imported goods. A physical SUT will show which industry is dependent on which critical raw material to produce. An ESUT will show whether there are differences in this respect between different types of enterprises. A time series will show whether part of the production process is moved abroad or not.
Measuring substitution of non-organic materials (such as fossil fuels) by biomass to produce the same product monitors the transition to a circular economy. For example, it could concern producing plastic from plant-based oil instead of from fossil oil. Another way to monitor the transition is by defining biotic materials as those that come from renewable biological resources and abiotic materials as “the rest”. Combining this with bio-based products allows to estimate the size of the bio-based economy and monitor it over time. An ESUT will indicate whether there are differences between types of enterprises and whether there is potential for one of them to improve.
Recycling mitigates the depletion of natural resource reserves and reduces dependencies on countries which supply materials. It is not meaningful to compare industries since they have very different material use/supply and production processes. For example, electricity suppliers will use recycled biomass whereas paper producers will use recycled paper. Comparing the use of recycled materials by different types of enterprises in the same industry may lead to sharing best practices.
One can estimate which material flows to and via a country have the most environmental impact abroad. Since different types of enterprises in the same industry will differ in their use of domestic/foreign materials, their environmental impact abroad will differ as well. This is why Guilhoto and Howell (2024[26]) recommend using an extended IOT or an extended multi-country IOT when estimating the carbon dioxide (CO2) footprint of FDI. For the same reason, it is recommended to use an extended IOT or multi-country IOT when estimating the material footprint of MNEs, SMEs, etc.
Input-output models have been widely used in studying trade-related economic and environmental issues. However, the traditional input-output model fails to capture enterprise heterogeneity. As mentioned in Chapter 2, production structure, technology and trade (Bernard et al., 2012[27]; Chong et al., 2019[28]) vary substantially among different types of enterprises. This is also the case for energy consumption and carbon emissions. Only by taking this heterogeneity into account can proper estimates of the economic gains and environmental costs related to exports of different types of enterprises be obtained.
Country example: China
This section is based on Zhang and Yang (2022[29]). One typical feature of China’s economy is the enormous number of SMEs, accounting for 99% of its total number of enterprises in 2018. SMEs contributed to more than 60% of GDP, 50% of taxation, 70% of technological innovation and 80% of urban employment, demonstrating that they are a dominant force in the Chinese economy. At the same time, SMEs also have a considerable environmental impact. Recent evidence by Meng et al. (2018[30]) shows that more than half of China’s total CO2 emissions were attributed to SMEs in 2010.
Policies are aimed at enterprises, but a one-size-fits-all policy is bound to fail. Zhang and Yang’s work provides scientific insights for policy makers to choose effective industrial structure adjustment strategies. Different policies for different types of enterprises can lead to China’s trade progress while mitigating trade-related emissions.
Data
The data sets used to compile the EIOT can be divided into three categories: 1) a regular non-competitive IOT; 2) enterprise-level data; and 3) other official statistics.
China’s regular IOT from the National Bureau of Statistics (NBS) is the basis for the compilation of the EIOT.
Enterprise-level data include the Chinese Annual Survey of Industrial Firms (CASIF) and Chinese Customs Trade Statistics. Specifically, the CASIF data obtained from the NBS covers more than 300 000 enterprises each year, accounting for approximately 95% of total Chinese industrial output. At enterprise level, the CASIF provides information of industrial output; value added; export value; and other basic characteristics such as enterprise code, industry code, etc. The data from the Chinese Customs Trade Statistics are compiled by the General Administration of Customs of China, which records all transactions (imports and exports) passing through Chinese customs. It contains enterprise identification (name, address, ownership), product code (eight-digit Harmonized Commodity Description and Coding System [HS] code), value of imports and exports, etc.
China’s National Economic Census and other official statistics from NBS are also used for the total constraints.
Methods
Zhang and Yang (2022[29]) constructed an extended non-competitive input-output model for China that distinguishes between large, medium and small-sized enterprises. The classification of an enterprise into the different enterprise size categories is based on the NBS’ “Division Standard of Large/Medium/Small Sized Enterprises”, which varies across sectors. Table 9.5 shows the criteria for the industrial sector. These are related to the number of employees and an enterprise’s revenue. To be classified as large or medium-sized, enterprises should meet both the number of employees and revenue criteria. Otherwise, the enterprise will be classified into the next lower category of enterprise size.
Emissions by size class are estimated using their dependence on the intermediate inputs from energy sectors. The study distinguishes five energy sectors: 1) coal mining and processing; 2) crude petroleum and natural gas extraction; 3) petroleum processing and coking; 4) electricity and steam production and supply; and 5) gas production and supply. Note that other emissions, for example at the enterprise itself, are not taken into account. It is recommended to include such information if it is available.
Results
Table 9.6 shows the share of each size class in output, value added, export and CO2 emissions in the industrial sector for 2012. Overall, the total output and value added of SMEs accounted for more than 60% in the industrial sector, demonstrating that SMEs play a crucial role in China’s economic development. In terms of exports, large enterprises accounted for half of China’s industrial sector’s total exports, indicating that large enterprises have certain advantages over SMEs in international trade.
Next, the value added and CO2 emissions generated per unit export of different size classes are calculated. This enables comparing economic gains and environmental costs related to exports of different size classes in China. The results imply that considering the heterogeneity of enterprise size matters when measuring the exports-related value added and CO2 emissions in China. Table 9.7 shows that the value added per RMB 1 000 of exports of large enterprises is the lowest (RMB 627), while that of small enterprises is the highest (RMB 791), and the ratio of medium-sized enterprises lies in between. The pattern of embodied CO2 emissions per RMB 1 000 of exports by size class is similar to that of value added. Large enterprises exported cleaner than SMEs. Possible explanations may be the difference in export product structure and the import intensity of intermediate inputs across different sized enterprises.
The study found apparent disparities in the export structure among different size classes. Specifically, large enterprises mainly export high-tech or sophisticated products, such as “electronic and telecommunication equipment” and “electric equipment and machinery”. These amounted to 45% and 10%, respectively, of total industrial exports for large enterprises in 2012. These sectors tend to use relatively more imported intermediates and therefore usually have a lower value added per unit of exports ratio. Meanwhile, they also have a very low CO2 emissions intensity compared with resource-intensive and labour-intensive sectors.
Furthermore, in 2012, 22% of the total intermediate input of large enterprises consisted of imported products. For medium-sized and small enterprises this was only 16% and 7%, respectively. This indicates that larger enterprises use more imported intermediate products, which leads to a higher intensity of foreign raw materials and intermediate input products. Thus, these enterprises would create lower domestic value added and lower embodied domestic CO2 emissions in exports.
This section, based on Samuels (2021[31]), motivates integrated ESUTs and KLEMS (capital, labour, energy, materials and services; see, for example, O’Mahony and Timmer (2009[32])) productivity statistics. Merging the two approaches with TiVA analysis yields insights into the economic forces driving the origins and evolution of GVCs; the sources of cross-country economic competitiveness; and improves the analytical usefulness of the ESUTs, KLEMS and TiVA statistics. This section first highlights the benefits of integrating KLEMS and TiVA statistics (without ESUTs). Subsequently, it briefly discusses the data needs for and the advantages of integrating KLEMS with ESUTs.
There has been substantial progress in the international statistical community in the production of broadly comparable (across countries) KLEMS statistics and IOTs for TiVA decompositions. Merging the two parses TiVA estimates from total trade in value added to its origins across factors of production. This is useful for analysing the origins of value chains across countries, instead of taking them as given. It provides estimates of underlying factor flows that are not evident in standard TiVA estimates without the integrated KLEMS detail.
When a country engages in GVCs, its trade in value added reflects and is driven by its production structure relative to others in the GVC. When a country exports value added, this value added embeds contributions from different types of labour and capital services. Integrating KLEMS with TiVA allows identifying these different sources of value-added trade. For example, a more advanced country that may be intensive in educated labour and research and development (R&D) likely has a different trade structure from a country intensive in production that does not require highly educated labour and R&D.
Taking a concrete example, consider the identification of the role of R&D in US value-added exports. Using standard TiVA estimates, R&D has a limited apparent role in US trade even though the United States is relatively R&D-intensive. On the other hand, US KLEMS statistics include information on the role of R&D services in US production (BEA, 2023[33]). R&D embedded in US value-added exports can be measured by identifying the share of R&D directly embedded in each value-added export and combining this with the R&D embedded in the contributions of US upstream industries that participate in the domestic supply chain. Table 9.8 shows a preliminary estimate of the factor content of US value-added exports in 2019 and demonstrates the usefulness of integrating KLEMS and TiVA statistics. The results show that embedded R&D is a significant contributor to US exports that would not be evident in traditional trade or TiVA trade measures. With existing data, producing KLEMS and TiVA-integrated statistics is relatively straightforward.
A promising future direction is to integrate KLEMS and ESUTs. The premise of the ESUTs is that further decomposing industries in the supply-use tables (e.g. into exporters and non-exporters) helps understand the origins of TiVA. It refines the TiVA estimates as well by isolating the input structure of globally engaged producers from other producers.
Integrating KLEMS with ESUTs has the same basic motivation as integrating KLEMS and the standard TiVA estimates discussed above. However, it is even more powerful, because the exporters and non-exporters are separated, thus providing refined TiVA estimates, and more detailed origins and sources of value added across the global value chain.
The data necessary for such an account that includes KLEMS-type productivity estimates would be demanding. Investigative work would have to be done to evaluate the prospects for estimating the necessary components. The basic components would be output prices that reflect differences in prices charged by globally engaged versus other producers. On the input side, comparable information would need to be assembled: the number and compensation rates for workers classified by demographic group (e.g. workers with a college and non-college degree and other dimensions) along with the capital stock, asset composition and rates of return in globally engaged versus purely domestically engaged producers. While the data challenges are significant, the benefits of such an account are important in understanding the sources of economic competitiveness as the global value chain and its evolution dominates much of economic development throughout the world.
One of the primary factors behind the development of TiVA-type indicators was to better understand the impact of globalisation, in particular in relation to the fragmentation of GVCs. An important focus of that work centres on “who benefits” and, in particular, the evidence that trade is not a zero-sum game. However, while there is significant evidence that trade has been, and remains, an important driver of growth across the globe, within countries the evidence, albeit largely anecdotal, suggests that those gains are not always spread equally, potentially exacerbating what have become known as geographies of discontent (OECD, 2023[34]). Regional ESUTs provide a powerful tool that can shed insights on how regions within countries have benefited, or otherwise, from globalisation, and to inform policies, especially emerging new industrial policies, many of which have a strong place-based focus. The Inflation Reduction Act and the CHIPS and Science Act in the United States, as well as the European Chips Act and Korea’s K-Chips Act, are recent examples of efforts to foster supplier diversification and partial onshoring of production processes through a mix of industrial and innovation policies.
This section is based on Horvát and Amann (2025[35]). It provides more information on various types of regional SUTs along with challenges and possible solutions pertaining to their construction. The fact that regional SUTs are rare, and regional ESUTs even rarer, illustrates the nature of the challenges involved, particularly the considerable challenge of measuring transactions concerning institutional units across regional borders. As such, the focus below is mostly on regional SUTs, with a more theoretical discussion on the possibility of extending them to regional ESUTs.
The spatial scale matters for measuring interdependencies between markets. Larger regions typically possess more of the necessary resources, infrastructure and capacities to meet many of their internal needs effectively and rely less on external resources. Conversely, smaller regions are more likely to concentrate on specific activities or tasks where they have a comparative advantage, making them more reliant on trade and co-operation. To illustrate, the latest OECD TiVA estimates show that 17% of European Union domestic value added is embodied in its gross exports (outside the European Union), which is, on average, half the size of domestic value added embodied in gross exports of individual member states. There are sizeable differences between larger economies such as France and smaller ones such as Ireland.
Another motivation for considering regions is the large variety in socio-economic and environmental performance between regions. This is because, for example, their demographic make-up, their industry agglomeration and specialisation, as well as their links within and across regions (Monsalve et al., 2020[36]). Furthermore, they can determine a region’s position in GVCs (Meng et al., 2016[37]) and, ultimately, how regions are affected by global economic developments.
Typologies for regional supply and use tables
Typologies underpinning the construction of regional SUTs can broadly be categorised in three ways:
Administrative typologies (ARSUTs) are generally based on historical, economic, statistical and political boundaries representing some form of subnational government, such as a US state, German Länder, Spanish Autonomous Community or any other form of subnational government. Administrative typologies enable a more accurate and detailed depiction of regional characteristics and link the potential analysis with local stakeholders. Examples of such typologies are the OECD territorial grids, the territorial levels classification (OECD, 2023[38]) or the Nomenclature of territorial units for statistics (NUTS) (Eurostat, 2022[39]) classification for EU member countries. ARSUTs improve the understanding of regional impacts and dependencies by increasing the spatial granularity and accuracy of the relationships between and within regions.
Analytical typologies are based on descriptive categories that follow more functional forms of areas, e.g. metropolitan and non-metropolitan within countries (UNSD, 2020[40]; Dijkstra, Poelman and Veneri, 2019[41]), which may result in spatial aggregations of regions that cross regional administrative boundaries. These can help better understand specific spatial factors that drive economic performance, such as agglomeration effects. Examples of functional forms include labour market areas (OECD, 2020[42]) and functional urban areas (OECD, 2018[43]). Functional urban areas, sometimes called “metropolitan areas”, encompass cities along with their adjacent commuter regions.
Combined typologies blend administrative and characteristic typologies. They combine administrative boundaries with descriptive categories with spatial dimensions (Monsalve et al., 2020[36]; Sato and Narita, 2023[44]; Zheng et al., 2019[45]).
Compiling regional supply and use tables
Methodological approaches
Subnational statistics can either be derived top-down from national aggregates using economic indicators or econometric and mathematical models (European Commission: Joint Research Centre, Rueda-Cantuche, J., López-Alvarez, J. and Galiano Bastarrica, 2025[46]), or obtained bottom-up from microdata sources. Hybrid approaches, which blend top-down tools with available microdata, are most frequently used in practice.2 Similarly, compiling regional ESUTs can be done using bottom-up tools or can be derived from national tables (using top-down methodologies) or a combination of the two.3 Top-down methodologies can produce distorted or inaccurate regional projections resulting from modelling constraints. In turn, bottom-up methods are complex and more data-intensive (Davidson et al., 2022[47]).
Ideally, regional ESUTs would be constructed bottom-up. However, no country has complete regional SNA accounts (European Commission et al., 2009[48]), and the availability of economic measures usually decreases with increasing regional granularity. A particular challenge relates to the unit for analysis that may be used for surveys. In many countries, this is the level of the enterprise and not the establishment, which means that underlying data for the production and consumption of a multi-establishment enterprise may not always be available. While National Statistical Offices try to apportion activity across regions in constructing regional national accounts data, this phenomenon can give rise to headquarters biases in regional statistics, especially with respect to activities with a high dependence on knowledge-based capital.
Whether using a bottom-up or top-down approach, breakdowns of regions that are equivalent to single or collections of administrative units are simpler to compile than those based on functional areas, given the data availability.
Challenges compiling regional supply and use tables bottom-up
The central challenges when compiling regional SUTs bottom-up relate to subnational production accounts, interregional trade, central government activity and sampling.
Subnational production accounts
As noted above, identifying and measuring economic activity (and the location where they originate) may be complicated whenever the enterprise has multiple establishments located in different regions, which can be further complicated if the establishments produce significant secondary products or services that differ from the enterprise’s primary activity. If production accounts at the establishment level are not available or exploited, apportionment methods are typically used to assign measures of the economic activity of enterprises to establishments using secondary information such as the number of employees or any other available measures.4 Table 9.9 describes preferred apportionment indicators for production and generation of income account measures, by decreasing level of precision and anticipated complexity of implementation.
Interregional trade
Interregional transfers of goods and services are a central feature of any interregional framework at the subnational level: goods and services can cross regional borders at multiple stages of the production process, yet such regional transfers are generally not captured, at least readily, in the national statistical system. Consistent trade flows between regions can, however, be obtained or estimated by populating trade flow matrices using initial flow information sourced from either one or a combination of different sources (Table 9.10). These commodity flows are then balanced using information on regional production and use of products and services together with final consumption. Note that Table 9.10 only contains the possible sources for flows that can identify the type of enterprise. All sources in the table can, therefore, be related to the enterprise level.
Notes:
1: Headquarter bias refers to the potential allocation of the disproportionate activity to headquarters as they are considered as owning knowledge-based capital of the enterprise.
2: B2G corresponds to business to government, B2B business to business and B2C business to consumer transactions.
3: The list of country examples is non-comprehensive and includes current as well as past projects and initiatives.
Source: Horvát and Amann (2025[35]).
Central government activity
While the location of production of central government can be readily determined (as well as its consumption) through administrative data, this is less trivial with respect to the location of consumption. In this sense, distribution on a per-capita basis is imperfect, as it assumes homogeneous regional demand for central government activity. Another approach would be to allocate it based on the location of production, as in the case of Canada (Davidson et al., 2022[53]), recognising that this would have little impact on input-output coefficients. However, regardless of how consumption is allocated, identifying from which regions central government purchased its intermediate consumption would still be necessary. Advancements in electronic government procurement systems are, however, helping to make these data more readily available (Cocciolo, Samaddar and Fazekas, 2023[54]).
Sampling
Available microdata are typically sampled to guarantee representative results at the national level. Therefore, the data may be ill-suited to produce representative regional SUTs, especially for regions with specific industry profiles. Furthermore, the availability of regional surveys and administrative or commercial data sets to remedy these challenges will vary on a case-by-case basis. In some cases, the lack of data may render a bottom-up construction of regional SUTs infeasible.
Digital SUTs are technically not ESUTs, since they do not split industries by type of enterprise. However, they do split SUTs by type of industry and by type of product, and several of the methods and data sources may be of use to compilers of ESUTs. This section is based on Mitchel (2022[55]) and Sakuramoto (2022[56]) and presents digital SUTs. First, it explains the general background and history of digital SUTs. Subsequently, it introduces the general framework, addressing the where (within or outside the production boundary), how (digitally ordered/delivered or not), what (digital/non-digital products) and who (digital/non-digital industries). It then presents the most sought-after indicators, concluding with three country examples.
Why do we need the digital supply and use tables
Digitalisation has fundamentally altered the production and consumption of goods and services worldwide. Enterprises have been able to leverage digitalisation to disrupt established markets while also improving the efficiency of their production processes. At the same time, the digital transformation has permitted consumers to access a wider variety of goods and services while exercising greater control over the characteristics of the transaction processes.5 Despite the rise of digitalisation, and the inclusion of digital transactions within the SNA production boundary, many of the digital trends occurring within the economy are not explicitly observable in the national accounts.
This absence of these key trends associated with digitalisation within the national accounts has, at times, caused confusion about what is (and is not) included in the production boundary of the national accounts6 and who is (or is not) benefiting from the digitalisation taking place. This need for more information has been recognised by other international groupings. For example, the G20 Digital Economy Task Force (DETF) requested a greater focus on measuring digitalisation and its impact on the economy. This includes a direct call to develop “satellite national accounts” focusing on the digital economy (G20 DETF, 2018[57]), creating the “G20 Roadmap toward a Common Framework for Measuring the Digital Economy” that explicitly included the digital SUTs as a method to improve the visibility of digitalisation in the national accounts (G20 DETF, 2020[58]; OECD, 2020[59]). The G20 DETF (2021[60]) reaffirmed and emphasised the need for co‑operation and sharing of best practices among national statistical offices.
There is currently no single, generally accepted definition of what the digital economy entails. This absence of agreement could be attributed to its multi-dimensional nature, since digitalisation has affected the production, ordering, delivery and consumption of almost all goods and services. From a measurement point of view, one can derive a picture of the digital economy by aggregating certain products or industries, e.g. the ICT products in the Central Product Classification (UNSD, 2015[61]) or the ICT sector in the International Standard Industrial Classification of All Economic Activities, Revision 4 (UNSD, 2008[62]). From a policy point of view, however, these definitions are often considered to be too narrow, insomuch that they miss digitalisation’s impact on the production of traditional goods and services. Therefore, it is likely that the output of these “narrow” interpretations of the digital economy understates the overall impact of digitalisation on the economy.
The digital SUT framework
To improve the visibility of increasing digitalisation in macroeconomic statistics, a digital SUT framework was developed (Figure 9.8). The framework’s fundamental point of delineation is the nature of transaction (“how”). However, to provide outputs that highlight the points of policy interest, the framework also includes additional variables. These include the good or service being ordered and delivered (“what”), the actor/sector involved in the transaction (“who”). The Figure clarifies which interactions/transactions are within the SNA production boundary (shown by a dotted line separating digital production from non-monetary digital flows).
Note: 1. DIPs = Digital Intermediation Platforms.
1. There are currently seven new digital industries; the last column in Figure 9.8 shows examples. The full list is provided later in the chapter.
Source: OECD, (2023[63]).
The multi-dimensional nature of the digital economy requires a framework that can produce outputs reflecting both the production and consumption of digital products as well as the production and consumption of non-digital products obtained through digital means, whether they are digitally ordered, digitally delivered or both. The SUTs are uniquely positioned to do this as the regular SUTs record not only what was produced and consumed but also who produced and consumed it. The SUTs are also flexible enough that additional products and industries can be added to provide more detail on specific topics without disrupting the balance already occurring within the SUTs.
With this in mind, digital SUTs contain the following additions from regular SUTs:
Five additional rows under each product (and aggregate), separating transactions by whether they are digitally or non-digitally ordered, with digitally ordered transactions further broken down into ordered directly from the counterparty, ordered via a resident digital intermediation platform and ordered via a non-resident platform.
Two additional columns delineating the nature of the delivery of the service as either digitally delivered or not digitally delivered.
Seven additional industry columns, aggregating enterprises based on characteristics related to the transaction nature or how they are leveraging the digitalisation.
Four additional rows representing the explicitly identified digital products of cloud computing services and digital intermediation services as well as an aggregation of all ICT goods and digital services that fall within the SNA production boundary.
Three additional rows representing data and digital service products that are currently outside the SNA production boundary.
More detailed definitions within the framework
Nature of the transaction (How)
In the digital SUTs, transactions in goods and services are broken down into five types, as shown in Table 9.11, which presents an example for the product row representing accommodation services. Theoretically, such a breakdown is conceivable for each product in the SUTs, but it is unlikely that such a breakdown will be compiled at the detailed level for all products.
The product (“What”)
While all goods and services produced in the economy are theoretically included in the digital SUTs, the framework focuses on ICT goods and digital services and non-digital goods and services that are most likely to be digitally ordered and/or digitally delivered. Examples of this include travel services, transport, accommodation and food services. Therefore, while non-digital products that are rarely if ever transacted digitally (i.e. trade in primary commodities, wholesale business services, etc.) are within scope of the digital SUTs, splitting the nature of transaction for these products is a lower priority.
The alternative product perspective within the framework relates to aggregates of ICT goods and digital services. Simply put, this includes all products that “must primarily be intended to fulfil or enable the function of information processing and communication by electronic means, including transmission and display” (UNSD, 2015[61]). As such, it coincides with the alternative classification of ICT products, as included in the Central Product Classification 2.1 (UNSD, 2015[61]). Like units that will populate the new digital industries, the production of these products is likely recorded in a large number of product rows. Ideally, portions of these product rows will be aggregated together to form two high-level rows: ICT goods and digital services.7 The exceptions to this aggregation are the two digital products explicitly identified in the digital SUTs, i.e. digital intermediation services (DIS) and cloud computing services (CCS). These two products, both of which are not explicitly identified in existing product classifications, are of particular interest to users, as they both represent the production and consumption of a digital service that has fundamentally altered the way businesses operate.
A final inclusion from the product perspective within the digital SUTs framework concerns three products that are outside the current SNA production and asset boundary. These are data, zero-priced digital services provided by enterprises and zero-priced digital services provided by the community. The measurement of these products is addressed as part of the overall revision of the SNA.
Digital industries (Who)
Like its absence in regular SUTs, there is no sector perspective set within the digital SUT framework. Rather, the additional (“who”) perspective provided in the digital SUTs relates to the creation of seven new digital industries:
1. digitally enabling industries
2. digital intermediation platforms charging a fee
3. data- and advertising-driven digital platforms
4. producers dependent on intermediation platforms
5. e-tailers
6. financial service providers predominantly operating digitally
7. other producers only operating digitally.
The industries have been separated out from their conventional industry columns to quantify specific aspects of digital activity currently unidentifiable within the existing industry allocation of SUTs. Importantly, the new industries identified include enterprises that are classified based on how they use digital technologies within their business models or to interact with consumers, rather than the fundamental type of economic activity undertaken,8 which is the basis for classification in the regular SUTs.9 For example, a retailer becomes an e-tailer if it receives most of its orders digitally. In practice, this means that two economic entities that are currently classified in two separate ISIC industries, due to their fundamental activity, may be placed in the same digital industry within the digital SUTs if they are leveraging digitalisation in the same manner. For example, a bookmaker (gambling services) and a tertiary education provider (education services) would be classified in separate columns in the regular SUTs but would be placed together in “other producers operating digitally” in the digital SUTs if they are both only delivering their services digitally.
High-priority indicators of the digital SUT
The digital SUT framework is highly ambitious, and while the additional rows and columns mentioned previously in this chapter are added to all products for consistency, it is not expected that any country will be populating all rows and columns associated with every product.10 Because of this, the advisory group proposed that a set of high-priority indicators should be considered as a first set of targets for countries compiling digital SUTs, focusing on the most important outputs of the digital SUTs from a user perspective:
Expenditures split by nature of the transaction. The following indicators are proposed to monitor their developments:
total household final consumption expenditure digitally ordered
total imports digitally ordered
total exports digitally ordered.
Output and/or intermediate consumption of DIS, CCS and total ICT goods and digital services. The indicators show the evolution of the digital transformation across industries. An increasing percentage of intermediate consumption of ICT goods and digital services relative to other products provides a good indicator for increasing digitalisation. Intermediate consumption of CCS and DIS is of relevance to better understand which industries are being most disrupted by the use of intermediation platforms or require the use of more flexible data storage to undertake their business.
Output, gross value added and its components, of digital industries, preferably in basic prices.
Current examples by countries
Although ambitious, countries are still making significant strides in the development and publication of outputs consistent with the digital SUTs. A clear example is the initial set of digital SUTs published by Statistics Canada (2021[64]). The experimental estimates (Table 9.12) revealed insights into digital activity such as estimates of total value of production that was digitally ordered and digitally delivered, as well as the gross value added of several “digital industries”.
Statistics Netherlands (2021[65]) has also produced an initial publication that includes outputs related to the high-priority indicators. The proportion of gross value added contributing to the economy by the digital industries (8%) is much higher than the equivalent figure in Canada (5.5%). Table 9.13 shows that this is reflected in the higher share from both digital intermediation platforms (10% compared to 2.7% in Canada) and e-tailers (23% in the Netherlands compared to 4.4% in Canada).
Japan (ESRI, 2021[66]) published a preliminary digital SUT which classified the entire Japanese economy into digital/non-digital industries and digital/non-digital products. It capitalised on the existing benchmark SUT and the Economic Census for Business Activity, which directly covered e-commerce, as well as the Survey of the Household Economy, the E-commerce Market Survey and the Basic Survey on the Information and Communications Industry. ESRI (2021[66]) points out the lack of information sources about smaller transactions in digital services. These are needed to capture, for example, intermediaries in Ireland, Airbnb intermediate services, YouTube rewards and advertising, etc.
The results for Japan are more comparable to those of the Netherlands. The size of the total digital industry is similar (also 8% of gross value added), as is the size of the main contributor: digitally enabling industries.
The comparison that is available between these three publications is evidence of the benefit of compiling estimates in a manner consistent with an agreed-upon framework. Users will know that they are comparing similar concepts and be able to quantify clear digital trends that are impacting the economy.
Overall, it is useful to look upon the compilation of digital SUTs as a continually evolving process in which countries can complete elements of the tables as source data become available. Countries are, therefore, encouraged to complete what they can, as soon as they can, with the idea of continually sharing emerging practices. In this way, digital SUTs can act as a roadmap, providing clear targets for countries to aim for when dealing with the challenges of making digital transformation more visible in economic statistics. These targets also provide a shared frame of reference when developing data sources that can be comparable across countries.
References
[21] AP (2022), “Germany, Poland take over natural gas firms tied to Russia”, Associated Press, https://apnews.com/article/europe-business-germany-economy energy-industry-9b46b2333c10a074bddf652cc1a7a38a.
[49] Avonds, L., B. Hertveldt and B. Van den Gruyce (2021), “Élaboration du tableau entrées-sorties interrégional pour l’année 2015: Sources de données et méthodologie”, Working Paper 7-21, Bureau fédéral du Plan, Brussels, https://www.plan.be/fr/publications/elaboration-du-tableau-entrees-sorties.
[33] BEA (2023), Integrated Industry-Level Production Account (KLEMS), Bureau of Economic Analysis, Suitland, MD, https://www.bea.gov/data/special-topics/integrated-industry-level-production-account-klems.
[27] 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.
[6] Bohn, T., S. Brakman and E. Dietzenbacher (2021), “From exports to value added to income: Accounting for bilateral income transfers”, Journal of International Economics, Vol. 131, https://doi.org/10.1016/j.jinteco.2021.103496.
[1] Botev, J. and A. Mourougane (2017), “Fiscal consolidation: What are the breakeven fiscal multipliers?”, CESifo Economic Studies., Vol. 63/3, pp. 295-316.
[7] Casella, B. (2019), “Looking through conduit FDI in search of ultimate investors: A probabilistic approach”, Transnational Corporations Investment and Development, Vol. 26/1, pp. 109-146.
[28] 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. 8/8, pp. 1-24, https://doi.org/10.1186/s40008-019-0139-1.
[54] Cocciolo, S., S. Samaddar and M. Fazekas (2023), “Government analytics using procurement data”, in The Government Analytics Handbook: Leveraging Data to Strengthen Public Administration, World Bank Group, Washington, DC, https://openknowledge.worldbank.org/server/api/core/bitstreams/bd31ce7f-96f9-4e98-8cdd-a87488625e8e/content.
[53] Davidson, S. et al. (2022), Improving the Quality of Regional Economic Indicators in the UK: A Framework for the Production of Supply and Use and Input Output Tables for the Four Nations, Economic Statistics Centre of Excellence, London, https://www.escoe.ac.uk/publications/a-framework-for-the-production-of-supply-and-use-and-input-output-tables-for-the-four-nations.
[47] Davidson, S. et al. (2022), Building a Suite of Subnational Socioeconomic Indicators for the United Kingdom: Opportunities, Challenges and Recommendations, Economic Statistics Centre of Excellence, London, https://www.escoe.ac.uk/publications/building-a-suite-of-subnational-socioeconomic-indicators-for-the-united-kingdom-opportunities-challenges-and-recommendations.
[23] Delahaye, R. et al. (2015), Material Flow Monitor: A Time Series, Statistics Netherlands, The Hague/Heerlen, https://www.cbs.nl/en-gb/publication/2016/05/material-flow-monitor-a-time-series.
[25] Delahaye, R., V. Tunn and A. Tukker (2022), “Developing a material flow monitor for the Netherlands from national statistical data”, Journal of Industrial Ecology, Vol. 27/2, pp. 408-422, https://doi.org/10.1111/jiec.13365.
[41] Dijkstra, L., H. Poelman and P. Veneri (2019), “The EU-OECD definition of a functional urban area”, https://doi.org/10.1787/d58cb34d-en.
[9] Duan, Y. et al. (2021), “How much did China’s emergence as “the world’s factory” contribute to its national income?”, China Economic Review, Vol. 69, https://doi.org/10.1016/j.chieco.2021.101658.
[66] ESRI (2021), The OECD Digital SUTs for 2015 and 2018 in Japan, Paper presented at the fifth meeting of the Informal Advisory Group on Measuring GDP in the Digital Economy, Economic and Social Research Institute, Cabinet Office, Government of Japan.
[19] European Commission (2023), Critical Raw Materials Act, https://single-market-economy.ec.europa.eu/sectors/raw-materials/areas-specific-interest/critical-raw-materials/critical-raw-materials-act_en.
[17] European Commission (2021), “Circular economy action plan”, web page, https://ec.europa.eu/environment/strategy/circular-economy-action-plan_en.
[50] European Commission (2013), “eInvoicing in Italy”, https://ec.europa.eu/digital-building-blocks/sites/display/DIGITAL/eInvoicing+in+Italy (accessed on 6 November 2024).
[48] European Commission et al. (2009), System of national accounts 2008.
[46] European Commission: Joint Research Centre, Rueda-Cantuche, J., López-Alvarez, J. and Galiano Bastarrica (2025), L.A., FIGARO-REG multi-regional input-output (MRIO) tables, European Commission, Seville.
[2] Eurostat (2023), “Glossary: Compensation of employees”, https://ec.europa.eu/eurostat/statistics-explained/index.php?title=Glossary:Compensation_of_employees.
[39] Eurostat (2022), Statistical Regions in the European Union and Partner Countries: NUTS and Statistical Regions 2021: 2022 Edition, Publications Office of the European Union, Luxembourg, https://ec.europa.eu/eurostat/documents/3859598/15193590/KS-GQ-22-010-EN-N.pdf/82e738dc-fe63-6594-8b2c-1b131ab3f877?t=166.
[69] Eurostat (2013), Manual on regional accounts methods, https://doi.org/10.2785/33649.
[18] Fortanier, F. and M. Maher (2001), “FDI and sustainable development”, Financial Market Trends, Vol. 79, pp. 107-130.
[60] G20 DEFT (2021), Declaration of G20 Digital Ministers, G20 Digital Economy Task Force, http://www.g20.utoronto.ca/2021/210805-digital.html.
[58] G20 DETF (2020), G20 Digital Economy Ministers Meeting: Ministerial Declaration, G20 Digital Economy Task Force, https://g7g20-documents.org/fileadmin/G7G20_documents/2020/G20/Saudi%20Arabia/Sherpa-Track/Digital%20Economy%20Ministers/1%20Ministers%27%20Language/G20%20Digital%20Economy%20Ministers%20Meeting_22072020.pdf.
[57] G20 DETF (2018), “Toolkit for measuring the digital economy”, web page.
[52] Généreux, P. and B. Langen (2002), The Derivation of Provincial (Inter-regional) Trade Flows: The Canadian Experience, Paper presented at the 14th International Input-Output Techniques Conference, 10-15 October 2002, Montreal, https://www.statcan.gc.ca/en/statistical-programs/document/1401_D4_T9_V1-eng.pdf.
[22] Graveland, C. (2016), Foreign Ownership of Firms in Relation to their Material Use (in Dutch), Statistics Netherlands.
[26] Guilhoto, J. and K. Howell (eds.) (2024), Carbon Footprint of Foreign Direct Investment (FDI), International Monetary Fund, Washington, DC.
[24] Hanemaaijer, A. et al. (2021), Integral Circular Economy Report 2021, Environmental Assessment Agency, The Hague.
[35] Horvát, P. and J. Amann (2025), “Regional ESUT”, OECD, Paris.
[10] Lemmers, O. (2023), Measuring the Economic Contribution of Firms and Activities in Terms of National Income, Presented at the 2023 IIOA conference in Alghero, https://iioa.org/conferences/29th/papers/files/4603.pdf.
[15] Lindroos, P. and M. Myllymäki (2021), Beyond value added - employment (gender and skills), Statistics Finland.
[8] Meng, B. et al. (2022), “Trade in factor income and the US-China trade balance”, China Economic Review, Vol. 73, https://doi.org/10.1016/j.chieco.2022.101792.
[30] Meng, B. et al. (2018), “More than half of China’s CO2 emissions are from micro, small and medium-sized enterprises”, Applied Energy, Vol. 230, pp. 712-725, https://doi.org/10.1016/j.apenergy.2018.08.107.
[37] Meng, J. et al. (2016), “The impact of domestic and foreign trade on energy-related PM emissions in Beijing”, Applied Energy, Vol. 184, pp. 853-862, https://doi.org/10.1016/j.apenergy.2015.09.082.
[11] 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.
[55] Mitchell, J. (2022), “An extended summary of digital SUTs”, OECD, Paris.
[36] Monsalve, F. et al. (2020), “Nesting a city input-output table in a multiregional framework: A case example with the city of Bogota”, Journal of Economic Structures, Vol. 9/1, pp. 1-24, https://doi.org/10.1186/s40008-020-0185-8.
[32] O’Mahony, M. and M. Timmer (2009), “Output, input and productivity measures at the industry level: The EU KLEMS Database”, The Economic Journal, Vol. 119/538, pp. F374-F403, https://doi.org/10.1111/j.1468-0297.2009.02280.x.
[67] OECD (2024), Global Value Chain Participation in UK Regions: Technical Background Note, OECD Publishing, Paris.
[63] OECD (2023), OECD Handbook on Compiling Digital Supply and Use Tables, OECD Publishing, Paris, https://doi.org/10.1787/11a0db02-en.
[34] OECD (2023), OECD Regional Outlook 2023, OECD Publishing, Paris, https://doi.org/10.1787/92cd40a0-en.
[38] OECD (2023), OECD Territorials grid, OECD Publishing, Paris, https://www.oecd.org/content/dam/oecd/en/data/datasets/oecd-geographical-definitions/territorial-grid.pdf.
[59] OECD (2020), A Roadmap Toward a Common Framework for Measuring the Digital Economy; Report for the G20 Digital Economy Task Force, OECD Publishing, Paris, https://web-archive.oecd.org/2020-07-23/559604-roadmap-toward-a-common-framework-for-measuring-the-digital-economy.pdf.
[42] OECD (2020), Delineating Functional Areas in All Territories, OECD Territorial Reviews, OECD Publishing, Paris, https://doi.org/10.1787/07970966-en.
[68] OECD (2019), Guidelines for Supply-Use Tables for the Digital Economy, OECD Publishing, Paris.
[43] OECD (2018), OECD Regions and Cities at a Glance 2018, OECD Publishing, Paris, https://doi.org/10.1787/reg_cit_glance-2018-en.
[4] OECD (2016), Data Needs for Incorporating FDI Income Statistics into the Trade in Value Added Framework, OECD Publishing, Paris, DAF/INV/STAT(2016)14.
[3] OECD (2016), Integration of FDI Statistics in TiVA: Results and Data Challenges, OECD Publishing, Paris, COM/STD/WPTGS/DAF/WGIIS(2016)1.
[5] OECD (2015), Linking TiVA and FDI Statistics: First Results and Future Plans, OECD Publishing, Paris, DAF/INV/STAT(2015)5.
[14] 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.
[51] ONS (2023), “Industry-to-industry payment flows, UK: 2016 to 2023, experimental data and insights [data set]”, Office for National Statistics, https://www.ons.gov.uk/economy/economicoutputandproductivity/output/articles/industrytoindustrypaymentflowsuk/2016to2023experimentaldataandinsights (accessed on 28 February 2024).
[56] Sakuramoto, T. (2022), A Digital Economy Satellite for the Japanese Economy.
[31] Samuels, J. (2021), Integrating Extended Supply-use Tables and KLEMS Productivity statistics.
[44] Sato, I. and D. Narita (2023), “Use of nested multi-regional input-output analysis for the evaluation of subnational emission reduction policies: A case of Japanese prefectures”, Sustainable Production and Consumption, Vol. 37, pp. 294-305, https://doi.org/10.1016/j.spc.2023.03.003.
[64] Statistics Canada (2021), “Digital supply and use tables, 2017 to 2019”, https://www150.statcan.gc.ca/n1/daily-quotidien/210420/dq210420a-eng.htm.
[12] Statistics Finland (2024), “Trade in value added [data set]”, Experimental Statistics, https://pxdata.stat.fi/PxWeb/pxweb/en/Kokeelliset_tilastot/Kokeelliset_tilastot__tiva.
[65] Statistics Netherlands (2021), Research on Supply-Use tables for the Digital Economy in the Netherlands, Statistics Netherlands, https://www.cbs.nl/en-gb/background/2021/49/research-on-supply-use-tables-for-the-digital-economy-in-the-netherlands.
[13] UIS (2012), International Standard Classification of Education: ISCED 2011, UNESCO Institute for Statistics, Montreal, https://uis.unesco.org/sites/default/files/documents/international-standard-classification-of-education-isced-2011-en.pdf.
[16] UNECE (2021), “Circular economy”, web page, https://unece.org/trade/CircularEconomy.
[20] United Nations (2012), System of Environmental-Economic Accounting: Central Framework, United Nations, New York, NY.
[40] UNSD (2020), “A recommendation on the method to delineate cities, urban and rural areas for international statistical comparisons”.
[61] UNSD (2015), “CPC 2.1”, Central Product Classification, Version 2.1, Statistical Papers, Series M No. 77, Ver. 2.1, United Nations, New York, NY, https://unstats.un.org/unsd/classifications/unsdclassifications/cpcv21.pdf.
[62] UNSD (2008), “ISIC Rev. 4”, International Standard Industrial Classification of All Economic Activities (ISIC), Rev. 4, Statistical Papers, Series M., No. 4, Rev. 4, United Nations, New York, NY, https://unstats.un.org/unsd/publication/seriesM/seriesm_4rev4e.pdf.
[29] Zhang, J. and C. Yang (2022), “China’s LMS Table capturing firm size heterogeneity and its application”.
[45] Zheng, H. et al. (2019), “Linking city-level input-output table to urban energy footprint: Construction framework and application”, Journal of Industrial Ecology, Vol. 23/4, pp. 781-795, https://doi.org/10.1111/jiec.12835.
← 1. The project of compiling the corresponding EIOT started in 2016, based on the official Belgian SUT and IOT for 2010, which was then the most recent input-output reference year. This compilation used the full set of individual enterprise‑level data sources underlying the construction of Belgium’s official input-output reference year SUT and IOT.
← 2. OECD (2024[67]) provides a discussion on how these challenges can be addressed in the context of the United Kingdom.
← 3. For example, using regional indicators per multi-regional unit for regionalisation; see Eurostat (2013[69]).
← 4. Some statistical institutions collect data on establishment-level for wages, production or capital; each can be better suited for apportion of different measures.
← 5. The terms “digitisation”, “digitalisation” and “digital transformation” may sometimes appear to be used interchangeably; however, they each represent something slightly different to each other. “Digitisation” is the conversion of analogue data and processes into a machine-readable format. “Digitalisation” is the use of digital technologies and data as well as interconnection that results in new or changes to existing activities. “Digital transformation” refers to the economic and societal effects of digitisation and digitalisation (OECD, 2019[68]).
← 6. The production boundary of the national accounts includes “all production actually destined for the market, whether for sale or barter. It also includes all goods or services provided free to individual households or collectively to the community by government units or NPISHs” (2008 SNA §1.40). What is not included are goods or services provided for free by private enterprises.
← 7. Goods are still considered to be delivered on a non-digital basis only.
← 8. The exception is the digitally enabling industries, where units are classified based on the products they are producing.
← 9. This basis for classification in the SUTs follows similar practices in the classification of groups, sections and divisions in the international industry classification, whereby the “actual production process and technology used become less important as a criterion for grouping activities” (UNSD, 2008[62]).
← 10. For example, the OECD SUT database presents countries SUTs with over 90 products. But many of the rows represent goods that by assumption are unable to be digitally delivered.