This chapter describes the Digital Supply and Use Tables (SUTs) framework from the product perspective. Estimates for the production and final demand of Information and Communication Technology (ICT) goods and digital services illustrate how digitalisation is changing production processes and consumption habits. This chapter also defines two digital products that are of particular policy interest - cloud computing services and digital intermediation services - and discusses the challenges of compiling estimates for them.
OECD Handbook on Compiling Digital Supply and Use Tables
4. Digital products (the “what”)
Abstract
Introduction
The previous chapter (Chapter 3) discussed the transaction perspective (the “how”) within the Digital Supply and Use Table (Digital SUT) framework that was presented in Chapter 2. This chapter discusses the product perspective (the “what”). It proposes new totals and breakdowns to provide visibility of the impact that digital products have in the economy.
Except for two separately identified digital products, the Digital SUT aggregates all Information and Communication Technology (ICT) goods into one row and all digital services into another row. This provides straightforward indicators of the importance of digitalisation to production processes and growth in consumption of these products over time.
Two products, digital intermediation services (DIS) and cloud computing services (CCS) are separately identified, reflecting the fundamental role these products play in the digitalised production and value chains of both traditional industries and the new digital industries. The addition of these rows means that indicators associated with products in the conventional SUTs (such as intermediate and final consumption, import, exports etc.) can be generated for DIS and CCS specifically, as well as for the totals of ICT goods and of digital services.
This chapter examines each of these additional rows (ICT goods, digital services, DIS and CCS), providing more information on the definition of the specific products and aggregations. It also provides guidance on how countries can compile product-based estimates consistent with the Digital SUT framework.
ICT goods and digital services
Definition of ICT goods and digital services in the Digital SUTs
The product rows ICT goods and digital services are discussed together because the genesis of both rows is the same: the ICT product classification found in the Central Product Classification (CPC) Version 2.1, Part 5: Alternative structures (UNSD, 2015[25]). The goods and services included in this alternative ICT product classification (and thus in the rows of the Digital SUTs) consist of 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[25]). The classification includes both goods and services, but there is a split between ICT goods and digital services.
The row covering ICT goods within the Digital SUTs includes four types of ICT goods:
1. Computers and peripheral equipment;
2. Communication equipment;
3. Consumer electronic equipment; and
4. Miscellaneous ICT components and goods.
The row covering digital services includes the following broad categories:
1. Manufacturing services for ICT equipment;
2. Business and productivity software and licensing services;
3. Information technology consultancy and services;
4. Telecommunications services;
5. Leasing or rental services for ICT equipment; and
6. Other ICT services.
Detailed lists of the lower-level products included in these categories is provided in Annex 4.A.
These product rows correspond to the ICT sector in the International Standard Industrial Classification of All Economic Activities (ISIC) (UNSD, 2008[24]). They can be used as the basis for the digitally enabling industry outlined within the industry perspective of the Digital SUT framework (see Chapter 5).
The concept of a digital good or service is evolving. It should be considered independently of the nature of the transaction used to order or deliver it. Although many digital services must be delivered digitally, the fact that a service is delivered or ordered digitally does not automatically make that service a digital service. Some services such as publishing, gambling or education are increasingly being delivered digitally and the Digital SUT framework allows this change to be represented based on the nature of the transaction (see Chapter 3). However, rather than having a growing list of (potential) digital services, the Informal Advisory Group (IAG) on Measuring GDP in a Digitalised Economy took the decision to use the established classification within the CPC.
Why separately identifying ICT goods and digital services is important
Although the Digital SUT framework does not subscribe to the idea of a single definition of the digital economy and the Digital SUT framework allows for the digital economy to extend beyond the ICT sector, it is clear that many users and analysts consider ICT products as a fundamental component. Several definitions consider the digital economy to consist of “the effective use of information and communication technology (ICT)” (G20 DETF, 2016[67]) or “economic activities enabled by information and communication technologies (ICT)” (Pratt, 2016[68]). Even if not considered a proxy for the digital economy, it can be useful to separately identify these products for several reasons. By separating out ICT products, it is possible to track their growth and how they may be replacing the consumption of more traditional products in production as well as final consumption by households. Also, the link between ICTs and growth of value added and productivity has been established through a body of economic research (Spiezia, 2012[69]; Cette, Nevroux and Py, 2020[70]). Therefore, policy makers are keen to have more information on this trend.
Such insights and references have largely been made possible by the (relatively) neat and well-established delineation of ICT goods and digital services in economic statistics. In addition to there being a separate aggregate in the CPC, their use is consistent with the overall asset definition in the System of National Accounts (SNA).1 The SNA lists ICT Equipment and Computer Software as separate classes of fixed assets. Many countries are already publishing outputs related to this classification, so the inclusion of this aggregate within the Digital SUT framework appears quite achievable.
The established nature of the ICT product classification is reflected in the long time series that countries have available on investment in ICT Equipment as defined in the 2008 SNA. This definition is not as broad as the ICT goods and digital services aggregate included in the Digital SUTs, as it is limited to just equipment or, more specifically, “devices using electronic controls and also the electronic components forming part of these devices” (UNSD, Eurostat, IMF, OECD, World Bank, 2009[18]); but it shows that a lot of countries have detailed information available. The data shows an increase in investment in these goods over time (Figure 4.1), with most countries at least doubling the volume of investment in the past 15-20 years.
It is important to note that businesses are not just investing in ICT goods to leverage the digital transformation; they are also consuming more digital services in their everyday production processes. The multidimensional nature of the SUTs is useful to understand what products business are consuming. For example, the Digital SUTs are able to show the level of ICT goods and digital services being recorded as intermediate consumption by conventional industries in the production of non-digital goods and services.
Figure 4.1. Gross Fixed Capital Formation, ICT Equipment, selected OECD countries
Index (2000 = 100)
Source: (OECD, 2022[71]).
Measuring ICT goods and digital services
Since all of the products listed in the CPC ICT goods and digital services classification are already included in existing products, expenditure on these products is already captured in most existing economic business surveys that ask for information on products used in consumption. Normally, surveys asking for this type of information are undertaken on an annual basis. Therefore, much of the information required to populate the ICT goods and digital services rows is already being captured by national statistical offices and compilation of these rows becomes an exercise in separating out the ICT goods and digital services from other products.
In addition to the existing surveys, recent changes to business surveys initiated by the United Kingdom Office for National Statistics (ONS) in their Digital Economy Survey (See Chapter 3) has led to the inclusion of specific questions on the purchase of digitally delivered services by businesses (ONS, 2022[66]). Examples include:
What type of e-commerce purchases did your business make from suppliers located in the UK?
Purchases of goods.
Purchases of digitally delivered services.
Purchases of non-digitally delivered services.
What was your business's expenditure on digitally delivered services from suppliers located outside the UK?
Similar questions are asked in the Canadian Survey of Digital Technology and Internet Use, which asks firms for a breakdown of the type of goods or services ordered over the internet, specifically asking for a separation between those goods and services that were digitally delivered and “other services” (Statistics Canada, 2022[72]).
Neither the United Kingdom nor Canadian survey asks specifically for purchases of the products listed in the CPC classification, and digital service products as defined in the ICT CPC classification and services that can be “digitally delivered” are not exactly the same. However, many of the service products listed in the classification are only able to be delivered digitally (network management, website hosting and the entire telecommunications category). As such, the inclusion of questions such as these would assist countries in breaking up the existing product rows in order to create the aggregate rows of ICT goods and digital services.
The level of ICT goods and services provides a simple metric that can be reproduced across countries. A recent joint exercise between the OECD and the United States Bureau of Economic Analysis (BEA) showed that in the United States, certain industries have observed significant increases over the past 10-15 years in their consumption of ICT goods and digital services (see Chapter 6), regardless of the final product the industry was creating.
Cloud computing services
Definition of cloud computing services in Digital SUTs
The first step in appropriately measuring cloud computing services2 (CCS) is to clearly identify what it is. In this regard, the definition of CCS has continued to develop as new products associated with cloud technology became available, expanding the scope of cloud computing.
In 2014, the OECD provided a high-level definition of cloud computing as “a service model for computing services based on a set of computing resources that can be accessed in a flexible, elastic, on-demand way with low management effort” (OECD, 2014[73]). This definition is useful in setting out that cloud computing involves the provision of computing services using computing resources that can be accessed and scaled flexibly, in other words “on-demand”. However, this definition does not make clear the nature of the access, which occurs over networks (mostly the internet), or the types of resources accessed (such as networking, storage and computer processing power). The definition was expanded upon in a guidance note covering CCS prepared as part of the update of the 2008 SNA led by the Inter Secretariat Working Group on National Accounts (ISWGNA). This contained the following definition, which will be used for the Digital SUT framework:
“Cloud computing services consist of computing, data storage, software, and related IT services accessed remotely over a network, supplied on demand and with measured resource usage that allows charging on a pay-per-use basis” (ISWGNA, 2022[74]).
On-demand delivery is a defining aspect of cloud products. Access over a network alone is not sufficient for a service to be considered CCS. With on-demand delivery, a cloud service user can obtain the computing resources they require over the network and without the need for manual intervention on the part of the cloud services provider. Furthermore, the volume of computing power and data storage they have access to is practically unlimited because of the flexible and elastic nature of cloud services arising from pooling ICT resources across multiple users.
An additional interpretation consideration involves the exact meaning of “on-demand”. The 2008 SNA update guidance note suggests that “services delivered under contracts for a fixed period of access, such as a year, are ordered in advance, not supplied “on-demand.” Software subscriptions, for example, involve a license to access the software over a fixed time period” (ISWGNA, 2022[74]). However, although some CCS are billed based directly on usage,3 others, such as data storage, are usually billed based on the customer having access to a certain amount of resources over a given billing period (e.g. a month or a year). The customer is free to use as much or as little of the purchased capacity as they wish and, importantly, can usually add additional capacity rapidly if needed.
This type of pricing may not be precisely usage-based as specified in the definition. However, since the business is undertaking the same fundamental outsourcing of hardware and software, ideally the economic statistics should reflect this. Therefore, expenditure that reduces investment in favour of flexible, on-demand hardware and software accessed remotely is considered as CCS. With the evolution of both technology and pricing structures, compilers will need to make judgement calls on when a product does or does not meet the definition of CCS. As countries begin to undertake more measurement of CCS, that some of these challenges will continue to be discussed and reporting conventions are likely to emerge.
The definition outlined above excludes services related to co-location. This practice consists of placing firms’ hardware, such as their servers, in a common location in order to leverage some collective technological advantages such as increased network security, improved connectivity, or reliability of power supply without undertaking a full cloud migration. As pointed out in the SNA update guidance note, the servers (or other hardware) are still owned by the firms that use them, and therefore such co-location falls outside the definition of CCS; instead, it should be treated in the same way as conventional capital investment (ISWGNA, 2022[74]).
Why separately identifying cloud computing services is important
The use of CCS is growing across industries. Businesses can derive significant benefits from the flexibility that comes with “on-demand access, over the internet, to ICT resources, such as computing power, data storage capacity, operating system functionality, and software applications” (Baer, Lee and Tebrake, 2020[75]). Figure 4.2 shows that between 2014 and 2020, the use of cloud computing by businesses has grown in every OECD country for which data is available, doubling in more than half of the countries. It is important to be able to measure the changes taking place and analyse their impacts on economic growth and productivity.
Some of the growing preference for CCS can be explained by the cost saving that can be achieved by businesses, since cloud computing providers are able to leverage the gap between the low utilisation rate of on-premises servers and software on the one hand, and the high utilisation rate that cloud computing data centres can achieve by taking advantage of scale of economies and virtualisation on the other hand (Cisco, 2018[76]).
Figure 4.2. Proportion of businesses purchasing cloud computing services, OECD countries
% of businesses
Note: All business with 10 employees or more.
Source: (OECD, 2022[77]).
From a production function point of view, the expenditure on CCS is not just another input cost that needs to be measured, rather it is a fundamental shift from the traditional model of ICT provision, in which firms directly invest in ICT hardware, software, and complementary specialist labour, to a model where such services are provided by external companies.
The IMF (Baer, Lee and Tebrake, 2020[75]) identified a range of economic impacts that can be expected as an increasing share of firms take up CCS:
a softening in output and investment in IT equipment;
investment in IT equipment to become increasingly industrially and geographically concentrated;
increases in the cross-border flow of commercial services;
increases in the relative size of IT services industries;
increased concentration of IT related labour in those industries providing cloud services;
potential gains in productivity due to rationalisation associated with cloud computing (lower levels of unused capacity);
appearance of new products and processes made possible by cloud computing; and
change in the geographic concentration of imports of ICT equipment.
In practical terms, a move to CCS implies outsourcing some of the hardware and software on which a business (or other organisation) runs its applications and maintains its databases. For many businesses, expensive items such as servers and multi-year software licences can be replaced with smaller, regular payments for ICT services from external providers.4 These external providers are likely to be concentrated in the ICT services industry.
Any economic changes should be identifiable in economic statistics. However, the IMF points out that the shift towards CCS creates challenges in a statistical system currently adapted to measure the traditional model of ICT provision. This is another reason why it is important to separately identify and publish data on cloud computing. Without an understanding of this shifting dynamic in ICT investment, estimates of capital stock by industry, for example, may be distorted. The 2008 SNA update guidance note on cloud computing illustrates this using data for the United States (Figure 4.3).
Figure 4.3. Real capital stock of ICT hardware
Index (2000=100)
Note: Calculations of the index for all industries excluding data processing (etc.) use capital compensation as weights.
Source: (ISWGNA, 2022[74]) based on BEA-BLS industry-level production account, expanded capital detail (2021 release).
The suppliers of cloud services may be located outside the user’s country. If the level of ICT investment continues to be used as an indicator of digital intensity, some firms, industries and countries may appear to be de-digitalising (or at least digitalising more slowly) because they have outsourced their computing requirements. Therefore, there is a need to measure expenditures on CCS in order to capture these services as an input to production, including those that are imported, and to observe the extent to which they are substituting traditional ICT expenditures.
As shown by the increase in cloud use in every OECD country (Figure 4.2), this trend appears widespread. Therefore, it is vital that statistical offices can provide accurate and internationally comparable estimates from the beginning of this change rather than once it is fully established. Ultimately, the production of more granular statistics on the consumption of CCS will permit better understanding of the (change in) use of ICT by businesses.
Estimating cloud computing services: challenges
Conceptually, the production and consumption of cloud services are more straightforward to estimate than other areas of digitalisation. The services are usually produced by the businesses that sell them, and they are sold directly to the consumer at a market price.5 Most businesses and households that consume cloud services would be recording them in the same way as other inputs, making them relatively easy to measure.
The biggest practical challenge for measuring cloud services is the classification issue: deciding which product classification to use. This is a non-trivial issue because the characteristics of cloud computing – accessed over a network, provided on demand, and charged based on usage – can be applied to many different services. Therefore, there are likely to be elements of cloud computing in a range of current product categories.
An OECD Working Paper published in 2021 included an examination of the statistical product classes likely to contain cloud services products (Ker, 2021[78]). Table 4.1 shows the CPC 2.1 sub-classes that the paper identified as containing services associated with CCS.
Table 4.1. CPC sub-classes containing services associated with cloud computing services
Source: (Ker, 2021[78]).
The paper points out two important considerations. Firstly, while all the services products identified in Table 4.1 are likely to be delivered online, they are not necessarily delivered using the cloud infrastructure giving rise to the key features of CCS (rapid elasticity, measured service, pooled resources, etc.). For example, while sub-classes 84392 and 84391 mention that payment “may be by subscription or pay-per-play”, it does not need to be. A consumer may purchase a game or piece of software online as a once-only transaction. They then own this game or software meaning that the transaction is accessed via a network, but it fails several other cloud computing features such as measured service and on demand. The concern is not that the product does not contain services related to cloud services, but rather that they can be delivered in a way that is not consistent with cloud services. The classes can therefore, at best, be described as “cloud-containing product classes” (rather than as cloud services products).
Secondly, some of the sub-classes listed such as “other hosting and IT infrastructure provisioning services” or “online games”, which are present in both the CPC and Classification of Products by Activity (CPA),6 contain types of services that meet the definition of CCS (e.g. data storage) as well as types of services that do not meet the definition regardless of how the service is delivered. For example, data mining and data modelling are fundamentally different digital services to those included within the CCS definition (they are outside the scope of CCS).
Such issues are likely to exist whenever pre-existing classifications are used. A final report by the Eurostat Task Force on “Price and volume measures for services activities”, finalised in June 2018, included the broad recommendation that “The supply of SaaS should be classified with other software: CPA 58.2 (Software publishing services). PaaS is most likely CPA 62.01 (Computer programming services) while IaaS is CPA 63.11.1 (Data processing, hosting, application services and other IT infrastructure provisioning services)” (European Commision, 2018[79]).7 This means that these CPA categories include cloud services that will need to be separated out in compiling Digital SUTs to present estimates of CCS separately.
The 2008 SNA update guidance note on cloud computing suggests that “management and support services for CCS are an important part of the cloud computing industry” and that “hosting of servers and software [to be accessed through networks] is a related activity that may be useful to aggregate with cloud computing” (ISWGNA, 2022[74]). Such a broad interpretation would avoid the need to break down certain sub-classes that contain both cloud and non-cloud output; but it implies a reduced focus on CCS specifically.
Box 4.1 provides more detail on the challenges of separating cloud services from the existing CPC categories. Overall, when attempting to estimate the value of CCS in their economy, compilers are strongly encouraged to focus on production that as strictly as possible meets the definition of CCS in this handbook – those that involve computing, data storage, software and IT services – rather than including all services that have some characteristics of cloud technology.
Box 4.1. Challenges of using product classifications to estimate CCS
When considering which CPC sub-classes should be included in an estimate of CCS for the Digital SUTs, a balance must be found between including more classes so that all types of cloud services can be included while limiting as much as possible the inclusion of non-cloud products.
More and more web-hosting contracts, which are part of CPC sub-class 83151: “Website hosting services” (see Table 4.1), are underpinned by cloud infrastructure; but not all of them are. Any output that is produced using a single server model should not be regarded as a “cloud hosting service” and therefore should not be included in an estimate of CCS.
CPC 2.1 sub-class 83159: “Other hosting and IT infrastructure provisioning services” includes some activities such as data storage services that are cloud services, as well as activities that are simply customers of cloud services. For example, video and audio streaming services are part of this sub-class. These companies may rely on cloud infrastructure to store and serve content, but they tend to use separate cloud service providers to power their content delivery networks. For example, Netflix uses AWS and Microsoft Azure to power its video streaming products. Subscribers buy passive access to a library of content where they can rapidly access content on demand; but they do not have access to computing/data storage/software resources directly. It is the streaming service company, rather than the end user, that is the customer of the cloud service.
The inclusion of audio and video streaming services may imply a precedent for treating products that use cloud services as a key input as equivalent to cloud services themselves. Many other companies and products might be equally reliant on CCS, such as online ride sharing and delivery platforms or social networks.
The concern is that as more and more service provision, especially ICT services, becomes not only enabled by cloud technology, but is provided and consumed in a manner that shares similar characteristics to CCS (on demand through a network, flexible, charged on a pay-per-use basis), this would result in all ICT services gradually being thought of as CCS. This would reduce the analytical usefulness of the CCS estimates. Therefore, it is the view of the IAG on Measuring GDP in a Digitalised Economy that countries should try to limit CCS estimates to activities involving computing, data storage, software and IT services. This will enable users to obtain the best information possible. While such a recommendation may require additional modelling, this may be better than the alternative of broad assumptions regarding certain product sub-classes.
Estimating cloud computing services: examples
Statistics Netherlands estimates the production of CCS for their Digital SUTs using the CPA categories recommended by Eurostat's Task Force on “Price and volume measures for services activities”. Depending on the product, they allocate to CCS all production (e.g. for data processing, hosting, application services) or a portion of production, which varies based on the industry producing it (e.g. for online software and for computer programming, consultancy and related services). In this way, Statistics Netherlands has identified specific combinations of certain products, produced by specific industries as representative of cloud services. The benefits of this approach are that estimates can be produced for imports and exports of priced CCS and it is also possible to show a breakdown of high-level CPA product categories within the CCS estimate. Figure 4.4 shows that the largest product category is information services.
Figure 4.4. Composition of priced cloud computing services, Netherlands, 2018
% of total
Source: (Statistics Netherlands, 2021[43]).
In its November 2022 Digital Economy Satellite Account publication (Bureau of Economic Analysis, 2022[47]), the BEA updated its estimate of cloud services for the United States. Previously, the BEA used product data from the Economic Census which were “grown using private industry reports and public financial information from companies engaged in cloud computing” (Bureau of Economic Analysis, 2022[80]). However, further investigation found that this estimate contained unrelated internet and data products, which may have resulted in an overestimate of cloud services. The revised method uses data purchased from the International Data Corporation for current years with years prior to 2013 back-cast using publicly available Securities and Exchange Commission filings for major cloud producers. The results suggest that components of CCS within the United States have grown rapidly, increasing 232% from 2015 to 2020 (Figure 4.5). This estimate includes co-location services, data management services, video and audio streaming services, and information and document transformation services so the BEA may be including some elements with cloud characteristics that go beyond the definition of CCS. This highlights some of the measurement difficulties faced by compilers.
Figure 4.5. Gross output of cloud services products, United States, 2005-2021
Billion US dollars
Note: Includes the following products: co-location services; data management services; video and audio streaming services; and information and document transformation services. Estimated by extrapolating product data form the 2012 US Economic Census using private industry reports and public financial information from companies engaged in cloud computing. May include some similar products not meeting the definition of cloud services.
Japan’s publication on “Service areas of product classification 2019” (Japanese Ministry of Internal Affairs and Communications, 2019[81]) also attempts to estimate the level of production of CCS, developing specific product categories for IaaS and PaaS. These include sub-product 37002406 “ICT infrastructure shared services” which is defined as “services that provide fundamental functions for systems/applications to be built upon, including services provided through networks whereby users share equipment or facilities (e.g. servers or storages in data centres)” and 40102100 “ICT applications shared services”, described as “services as using networks, servers in data centres, equipment, and storage facilities shared with other users to provide application services”. Surveys using these classifications were used in the 2021 economic census, with the intention of data being used in the compilation of the next benchmark SUT in 2025. The initial estimate of CCS was similar to that reported elsewhere.
Separating the underlying survey data, which does not specifically distinguish services meeting the definition of CCS from other products recorded in the same class, is still a work in progress for most countries. As such, early estimates may include some sales of non-cloud service products. However, despite their slightly different definitions and methodology, the results from countries that have already produced estimates of CCS for inclusion in their Digital SUT outputs are broadly similar as a proportion of total output (Table 4.2).
Table 4.2. Production of cloud computing services
Note: Canada‘s estimates are for output of North American Industry Classification System (NAICS) class 51821; data processing, hosting and related services. Japan’s estimate is a proportion of total income from sales.
Final considerations for cloud computing services
This section has demonstrated that there is a clear demand for data that shows the production and consumption of CCS in a manner consistent with other products in the national accounts. As more countries produce estimates consistent with or similar to cloud services as defined in the Digital SUT framework, it will be vital to understand country practices for deriving CCS estimates from the product categories in the established international product classifications or their manner of estimating consumption and production of CCS independently.
As noted in the SNA guidance note, there will be a need for “additional collaboration with classification experts on refinement of classifications for the cloud computing and related industry and products” (ISWGNA, 2022[74]). Updates of the CPC and related product classifications should provide opportunities to improve estimation of CCS. They should also aim to avoid unintended consequences for CCS estimation, such as the expansion of categories containing products associated with CCS (currently used by some countries to estimate CCS) to include additional products that are not predominantly CCS products.
Digital intermediation services
Definition of digital intermediation services in Digital SUTs
Digital intermediation services (DIS) are produced by Digital Intermediation Platforms (DIPs) when they match a buyer/consumer of a good or service with a seller/producer. While DIS is not currently listed in international product classifications, it is not new. Existing activities such as “sales on a commission basis” and “reservations services” usually involve businesses facilitating transactions between two independent parties, and this may be done digitally. However, with digital intermediation now occurring in more and more industries, the CPC and other product classifications are being revised to better record such activities.
Since DIS is seen as the product produced by DIPs, there is a complimentary definition in Chapter 5, where DIPs are defined as:
Businesses that operate an online interface that facilitate, for a fee, the direct interaction between multiple buyers and multiple sellers, without the platform taking economic ownership of the goods or rendering the services that are being sold (intermediated).
The product of these units, the DIS, will be defined as:
Online intermediation services that facilitate transactions between multiple buyers and multiple sellers in exchange for a fee, without the online intermediation unit taking economic ownership of the goods or rendering the services that are being sold (intermediated).
The definition goes into detail regarding ownership of the good or services being sold in order to make clear that the DIS is a product in itself and should not be considered a margin added on top of the value of the underlying product. Margins are usually represented as an additional retail or distributional cost added on top of the producer prices to arrive at the price paid by the consumer. Rather, as discussed below, the DIS is a product usually consumed by the producer and considered another cost of production.8 This difference is fundamental in separating out the DIS from retail and wholesale trade.
The definition used for the Digital SUT framework specifically mentions online facilitation of transactions. Intermediation services can be provided on a non-digital basis, but the Digital SUT framework specifically focuses on digital output. This may result in a difference between the definition used here and that used in updates of the product classifications, as product classifications have to be broad enough to cover digital and non-digital intermediation services.
An update of CPC Version 2.1 is expected to be presented to the United Nations Statistical Commission in 2024. Early discussion on the update suggests that intermediation products will be shown together at division (2-digit) level with the lower-level groups, classes and sub-classes reflecting the nature of the goods and services facilitated by the intermediation. For example, the intermediation service provided by a DIP when it facilitates a transaction between a consumer of accommodation services with a person looking to rent out their home would be in a different sub-class to the intermediation service provided by a DIP that facilitates a transaction between a driver with a consumer looking for a ride; but both would belong to in the same division.
Whatever the outcome of the CPC update discussion, in the Digital SUT framework the service of facilitating a transaction between a consumer and a producer in exchange for a fee is considered the same product regardless of the underlying good or service being intermediated. In the Digital SUTs, all output of intermediation products should be recorded in a single product row: the DIS row.
Why separately identifying digital intermediation services is important
Thanks to the internet and other digital tools, there is now continual communication between sellers and DIPs, with much increased availability of products, regardless of the geographical location of the producer or the consumer. The ability of buyers to compare similar products, at the touch of a button and for no explicit cost, has made intermediation platforms desirable for many consumers. Overall, as will be elaborated on further in Chapter 5, the growing use of DIPs is of considerable interest to users.
Bringing all instances of DIS into a single product category provides a clear interpretable number which can reflect the impact of DIPs on the economy. Furthermore, the aggregation of all DIS into a single row is consistent with the treatment of DIPs in the Digital SUTs. From the industry perspective (as outlined in Chapter 5) all DIPs are aggregated into a single “digital industry” to provide a clearer picture of their contribution to value added of the economy. Therefore, within the Digital SUT framework, all DIPs are contained within a single industry column and the product they produce, the DIS, is contained within a single product row.
In some situations, a component of the fee associated with DIS is separately invoiced to the consumer, on these occasions this portion is considered as household consumption. However, on almost all occasions, the DIS product is consumed as intermediate consumption as the producer is usually responsible for all or part of the fee. As such, a split based on those industries that are consuming the DIS product and shown as intermediate consumption in the use table will provide additional analytical detail on industries that are using intermediation platforms to reach final consumers.
A final reason why it is important to separately identify DIS is that including a third party that provides value added (DIS) to the production chain, but does not take ownership of the product, challenges the traditional producer/seller and consumer/buyer paradigm that features in almost every transaction recorded in the accounts. This leads to two measurement issues. The first is that the additional player creating output can distort the picture of where price changes are coming from. It is important to attempt to separate changes in the price of DIS from changes in the underlying product9. The second is that the flow between the additional player and the producer and/or consumer must be appropriately recorded. This is discussed further in the section on consumer and producer approaches to measuring DIS.
Estimating digital intermediation services: examples
The recording of DIS goes hand in hand with the recording of output from DIPs. Conceptually, this is quite straightforward. The output is (usually) produced by formal units, which charge a market price in direct exchange for the provision of a service (the intermediation service). Therefore, output can be calculated in a conventional way, by identifying and surveying the units that produce the service.
However, there are practical concerns in the measurement of DIS that may not exist for other products. So far, in work undertaken to try and estimate DIS there appear to be three approaches. These are:
Creating estimates of DIS on the assumption that units identified as DIPs are solely responsible for the production of DIS, so that output of DIPs and DIS are the same.
Modelling estimates of DIS from known business information and surveys.
Separately identifying production of DIS within existing product estimates.
Of these three options, there are examples of countries undertaking the first two approaches. These are discussed below. The last option is also discussed even though the IAG is not aware of any countries that have used this approach to estimate DIS.
Examples based on output of DIPs
For this approach, units considered to be producing DIS are surveyed, with their output calculated in the standard way. The most important assumption for this approach is that the units considered part of the DIP industry (see Chapter 5) only produce DIS. While this seems to be a reasonable assumption, it is also possible for other units beyond DIPs to produce DIS. For example, a unit may, in addition to reselling goods they own or selling goods that they produced themselves, also facilitate the purchase of other goods and services. Unless the unit predominately produces DIS, they will be classified in an industry other than DIPs. Therefore, this approach is likely to produce an underestimate of DIS. However, this underestimate is likely to be small as it appears that output of DIS from the large DIPs accounts for most of the DIS produced.
As discussed in Chapter 5, arguably the biggest challenge with this approach is the practical identification of DIPs in the business register. However, once identified, these units can be surveyed, and estimates can be compiled in a manner similar to estimates of other products. Such an approach was undertaken by Statistics Netherlands following the successful identification of DIPs using a systematic machine learning approach (see Chapter 5). Their compilation of DIS estimates is presented in Box 4.2.
Box 4.2. Producing DIS estimates in the Netherlands
Using conventional business surveys to compile estimates of production of DIS based on DIP output
Following the identification of potential DIPs, including those that charge an explicit fee, Statistics Netherlands obtains information from their annual Structural Business Survey. This is used to split the platforms between those where more than 50% of employees work on intermediation and those that fall below this threshold. The former are considered predominately DIPs and are reallocated to the DIP industry. Statistics Netherlands notes that of those units identified as DIPs, “the largest businesses are fully specialised digital intermediary platforms”. (Statistics Netherlands, 2021[43])Calculations are undertaken to separate the output from these DIPs that is considered DIS from the production of more traditional products. Those platforms where fewer than 50% of employees work on intermediation are left in their “conventional” industry. A fraction of their output, estimated according to the products that different industries are producing, is considered DIS and is reallocated to the DIS product.
Statistics Netherlands is also able to apply data from their international trade in services survey to estimate what percentage of DIS is being exported. However, because DIS estimates depend on production data from resident businesses, there is no information on the amount of DIS imported.
Source: (Statistics Netherlands, 2021[43]).
Using the data collection and methodology outlined in Box 4.2, Statistics Netherlands has calculated that in 2018, DIS made up around 12.2% of all digital products produced within the Netherlands, with a value of 15.5 billion euros (see Table 4.3). Most of total DIS output in the Netherlands was exported (see Table 4.4), reflecting the fact that there are a large number of multinational platforms based in the Netherlands for fiscal reasons.
Table 4.3. Supply of ICT goods and digital services, Netherlands, 2018
Billion euros
Note: * Other supply e.g. import tariffs, other use e.g. accumulation of stock.
Source: (Statistics Netherlands, 2021[43]).
Table 4.4. Use of ICT goods and digital services, Netherlands, 2018
Billion euros
Note: * Other supply e.g. import tariffs, other use e.g. accumulation of stock.
Source: (Statistics Netherlands, 2021[43]).
Examples of estimating DIS from known business information and surveys
Another alternative involves attempting to identify the output of goods and services ordered via a DIP and then using this amount to estimate the DIS being produced.
Since every transaction undertaken via a DIP involves the production of DIS, the level of DIS produced is related to the transactions taking place. The pricing structure associated with DIS is often quite complicated. Some DIPs charge a set fee regardless of the value being transacted, while others charge a percentage of the value over a minimum fee. The difference in price may be due to a variety of factors, including competition and quality of the service provided. Therefore, the value as well as the quantity of the transactions are important; with both of these pieces of information, an “effective DIS rate” can be calculated and used to estimate the total DIS produced.
An example of this work is provided in Chapter 5, where data that is publicly available from Airbnb is used to estimate both the value added of the owners of the accommodation being rented out and the output of the intermediation service provided by the DIP. Research showed that the platform fee charged to the guests is 6% to 12% of the rent, the exact rate depending on the amount of the rent (Tobiassen, 2021[84]). Since it is separately invoiced to the consumer, it is considered final household consumption of DIS. Depending on the location of the platform, this may also show up in imports and export of these services.
Another example is the approach being developed by the United States BEA for the ride share business (see Box 4.3). They apply indicators to the conventional product rows in the existing SUTs to calculate the value of the output associated with platforms. Then they apply publicly available information from platforms and companies’ annual financial reports to work out how much of the final output remained with the platform, representing the value of the DIS.
Box 4.3. DIS associated with rideshare in the United States
Currently, the BEA’s Digital Economy Satellite Account (DESA) does not include estimates for DIS resulting in an incomplete picture of the digital economy. BEA has proposed the first phase in developing estimates of DIS that focuses initially on two service areas of immediate interest to their users, person-to-person (P2P) accommodation and P2P ride shares.
The main data sources proposed for the estimates are the economic census and company annual financial reports. Estimates are derived by working out a percentage for “electronic revenue” from the economic census1 and applying this to output from existing SUTs (Figure 4.6). A proportion of the electronic revenue (or customer’s expenditure on transport via a DIP) is considered to belong to the supplier of the underlying service, e.g. the taxi company. In these instances, no DIS is produced as no transaction is facilitated between producer and consumer.
Finally, a proportion of the customer’s expenditure on transport facilitated via a DIP is considered to represent the DIS provided by the DIP. This proportion is based on information taken from the financial reports. While only focusing on two products for the time being, this model provides a basis on how estimates of DIS may be compiled.
Figure 4.6. Example of modelling digital intermediation services
Note: GO = gross output; Estimates are provided for illustration purposes only.
1. Electronic revenue is defined as sales of goods and services where the buyer places an order, or the price and terms of the sale are negotiated, over Internet, a mobile device (M-commerce), extranet, Electronic Data Interchange (EDI) network, electronic mail, or other comparable online systems (United States Census Bureau, 2022[45]).
Examples based on separating DIS from existing product estimates
The output of DIS is already included in estimates of output of existing units within the national accounts, so estimates of DIS are included in those of other products. This means that one way to estimate DIS would be to separate out the relevant values from the existing product classes. No countries are currently doing this at present, but it is an option.
There are several classes in the CPC where DIS may be included. While often the CPC categories do not explicitly mention the term intermediation, the explanatory notes may provide clues that the product is similar to intermediation services. For example:10
CPC class 8551: Reservation services for transportation
CPC class 8552: Reservation services for accommodation, cruises and package tours
CPC class 8553: Other reservation services
These three classes incorporate services that include “obtaining a client's requirements, advising on alternatives, assisting in the client's choice and may include issuance of tickets on behalf of the service provider. They are often provided in person, by telephone or over the Internet. Respective reselling services are also included” (UNSD, 2015[25]).
While the explanatory notes mention acting “on behalf of the service provider” which is similar to the proposed definition of intermediation services, the notes also incorporate “reselling services”, which is fundamentally different to intermediation services. However, since services by definition are not able to be held as inventory, it is likely that the CPC interpretation of “reselling” is closer to intermediation as defined in the Digital SUTs than to the traditional retail business model.
There are also several products that explicitly include intermediation in the explanatory notes in a way that is consistent with the proposed definition. These include CPC class 8511: Personnel search and referral services and CPC class 7223: Land sales on a fee or contract basis. Both of these describe intermediation services between a buyer and seller of specific goods or services.
While not often published at such a low level of breakdown, many NSOs may already have this information internally. If this is the case, it should be possible to separately identify those services produced digitally, in exchange for a fee, without taking any ownership of the underlying product. In some cases, it may be decided that the entire class meets these three DIS criteria.
It should be noted that the classes currently identified in the CPC in relation to intermediation services mainly reflect the products that were purchased via intermediation when CPC Version 2.1 was published in 2015, that is, travel and tourism services. Intermediation services are now used more widely; but using the CPC product classes as a starting point for estimation may provide a way to produce initial estimates.
Recording DIS in the accounts - the consumer and producer approaches
Transactions involving the production and consumption of DIS, can be recorded from either the “consumer” or “producer” perspective (Box 4.4). The difference between the two is whether all transactions between the parties involved are recorded as they occur in the real world or if certain flows are re-routed between parties in order to better reflect the actual flow of services occurring. The difference may be important, especially if the DIP involved is a non-resident, as the choice may have a significant impact on trade statistics.
There is general agreement in the national accounts community that recording from the producer perspective is the appropriate treatment. This includes endorsement at the 12th meeting of the ISWGNA Advisory Expert Group on National Accounts (ISWGNA, 2018[85]) and the global consultation on DIPs for the update of the 2008 SNA (ISWGNA, 2022[86]). The global consultation provided overwhelming support for the producer approach to measuring DIPs and DIS in the accounts (ISWGNA, 2022[87]). It has also been included as the preferred approach in both the Handbook on Measuring Digital Trade (IMF, OECD, UNCTAD, WTO, 2023[13]) and the OECD/G20 Roadmap toward a Common Framework for Measuring the Digital Economy (OECD, 2020[6]).
The balance of payments community also takes this approach in its guidance note for the update of the 7th Balance of Payment Manual (BPM7) on “Merchanting and Factoryless Producers; Clarifying Negative Exports in Merchanting; and Merchanting of Services” (IMF: BOPCOM, 2022[88]). As well as discussing similar arguments, it includes the point that by definition services “are not separate entities over which ownership rights can be established [and as] they cannot be traded separately from their production” (see SNA §6.17 (UNSD, Eurostat, IMF, OECD, World Bank, 2009[18]). In theory, the same holds for intermediation of services, that is, it is impossible for the DIP to purchase and then resell a service. Thus, the producer approach is the only viable option for recording DIS.
Box 4.4. Recording flows of DIS – consumer and producer perspectives
The consumer perspective focuses primarily on the flow of money from the consumer. In many cases, for the payment, the consumer interacts only with the DIP. The interaction with the producer is often only around the provision of the service: there is no monetary transaction. Additionally, since the consumer does not pay the producer directly, the producer must seek payment from the platform, which holds the payment in trust until the service is provided. A literal recording of this situation would likely consist of a flow from the consumer to the platform with a follow-up payment from the platform to the producer. Such a treatment would reflect the DIP as “buying” the product from the producer to resell to the final buyer. However, this does not reflect the actual role of the DIPs, and the significant difference that exists between DIPs and retail traders, i.e. the DIPs do not take ownership of the goods or services in question and have a reduced level of financial risk.
The producer perspective focuses primarily on who the producer is providing the underlying services to (the person making the payment) and who the producer pays in order to facilitate the transaction, allowing for the underlying service to be provided. As such, the DIS product is treated as intermediate consumption paid by the producer as a cost of producing the final product provided to the consumer. This is outlined in Figure 4.7.
Figure 4.7. Recording flows of DIS – consumer and producer perspectives
In this approach, the intermediation service product produced by the platform is consumed by the producer as intermediate consumption; and possibly in part by the consumer as final consumption, depending on the criteria of the platform. A transaction is recorded between the consumer and the platform only if the platform charges a separately invoiced fee for the intermediation service direct to the consumer. Whether the fee is paid by the consumer, the producer or both makes no difference to the ultimate level of value added being created by either the producer or the DIP. A numerical example is shown in Annex 4.B. Most importantly, the payment for the underlying good or service, which includes any amount paid by the producer to the platform, is reflected as going from the consumer to the producer, despite it most likely being paid via the platform and held in trust by the DIP.
Focusing on the producer has two clear advantages:
If the DIP is a non-resident, this treatment removes the possibility of a distortion in the level of imports and exports from the country in which the DIP resides, as the underlying product remains a flow between the resident producer and resident consumer.
It provides a better reflection of the products (and the industries producing them) being consumed via transactions facilitated by DIPs. Even if the DIP is a resident entity, a distortion might exist if the value of final consumption of all products purchased via a DIP is represented as being produced by the DIP industry regardless of the underlying product.
Practical challenges when measuring DIS
Despite the general agreement on how the DIS should be recorded, the results of the global consultation on DIPs (ISWGNA, 2022[87]) also emphasised that there are a number of practical issues that NSOs will face when attempting to implement the recommendations. There are three broad challenges:
1. The differences in business models of DIPs create difficulties in appropriately recording the flows of DIS.
2. The producer approach to measuring DIS still poses significant measurement challenges and data requirements.
3. If the DIP producing the DIS is a non-resident, compilers may need to rely on modelling or information sharing between NSOs at the firm level.
The different business models include the different ways that DIPs collect revenue. Some DIPs charge a percentage of the value of the product, whereas others charge a flat rate. Some only charge the producer, while others charge both the producer and the consumer. Table 4.5 outlines three different scenarios that may be encountered by compilers. If the fees are separately itemised on the invoice and attributable to the seller/producer and/or the consumer/buyer, they are referred to as “explicit”. If this is not the case, the fees are considered as “implicit” and compilers will need to make assumptions about both the value of the fee and who pays for it.
Table 4.5. Explicit and implicit fees paid to DIPs
Note: Explicit and implicit fees in this table can be understood as meaning what is known to the compiler.
Source: (IMF, OECD, UNCTAD, WTO, 2023[13]).
These differences can make modelling the amounts difficult. An assumption based on the business model of one platform may over-estimate or under-estimate the value of the product produced by another platform. Responding to such challenges will require greater understanding of DIPs and sharing of knowledge across NSOs.
The second set of challenges relates to the producer approach to measuring DIS. Due to the redirection of certain transactions as well as the possibility of imperfect information, applying the producer approach in the national accounts may require imputations to estimate the production of DIS and the output of DIPs. A similar situation applies to retail margins as, in the conventional SUTs, the margin has to be separated from the underlying value of the good. This calculation is usually undertaken by NSOs by taking the cost of a good away from the final charge to the consumer. It would depend on being able to obtain these two values from the DIPs. Such information should be available using traditional survey methodologies, as long as the DIP is a resident business and included on the business register.
The third set of challenges relates to non-resident DIPs. If the DIPs are non-residents, the DIS being consumed is imported and the exact values allowing for a residual calculation of the DIS will probably be unavailable.11 In these instances, calculation of DIS will have to be modelled using available household and business statistics.
Conclusion
In many of the initial attempts to measure the digital economy, NSOs created estimates based on selecting products considered to be digital (Barefoot et al., 2018[8]; Statistics Canada, 2019[9]; Australian Bureau of Statistics, 2019[10]). While these estimates did not provide the full picture of the digital economy desired by users, this does not mean that separating out specific products is not worthwhile. Recording the changing nature of output shows not only the evolving nature of the products required for use in production, but also where the production is coming from.
Since the compilation of the conventional SUTs involves the generation of estimates related to the production of various goods and services, separating out ICT goods and digital services within the Digital SUT framework is feasible. Doing so provides an indication of the level of digitalisation occurring in the economy, including in parts of the economy that are not predominantly digital.
Finally, by creating estimates of output related to specific aspects of digitalisation and key components of the digital economy, compilers can provide relevant insights into important digitalisation trends, such as the take up of CCS and production of DIS. The role of intermediation in production chains and the change from investing in physical goods and hardware to purchasing intangible and flexible ICT services are two key areas of focus for users.
The ideal product breakdowns outlined in the framework will probably require the development of new data sources and methods. As more and more countries produce estimates of these products, best practices will be shared that will assist other countries in their compilation efforts.
Annex 4.A. Information and Communication Technologies (ICT) as defined in the CPC
Annex Table 4.A.1. Information and Communication Technologies (ICT) as defined in Central Product Classification (CPC), Version 2.1, Part 5: Alterative Structures
Source: (UNSD, 2015[25])
Annex 4.B. The producer versus the consumer perspective
When a transaction that involves the production and consumption of DIS occurs across borders, the choice of recording the DIS transaction from either the consumer or producer perspective does not make a difference to the overall GVA level for either country. However, there is potentially a significant difference in the imports and exports estimates. Annex Table 4.B.1 shows an example.
Annex Table 4.B.1. Reconciliation of transactions involving DIPs, gross and net approach
Country A pays $100 for a service, purchased using a DIP. The DIP is a resident in Country B and charges an intermediation service fee of $20. However, this is not separately invoiced to the consumer. The producer of the services is also in Country A and after factoring in all input costs, including the $20 intermediation fee, charges $100 for their service.
From the producer perspective, the output of the producer ($100) is equal to the household consumption ($100), with the intermediation service fee ($20) being recorded as an import and then intermediate consumption for the producer. This results in Gross Value Added (GVA) of $80 and $20 for Country A and Country B respectively.
From the consumer perspective, the GVA estimate for each country remains the same as in the net approach. However, the import and export estimates are both significantly higher as the full output of the producer ($80) would have been considered an export from Country A and an import by Country B, with a subsequent export of $100 being recorded, representing the full value of the service being recorded as household consumption and an import into Country A after the DIP has applied their value added ($20). Arguably, the approach focusing on the consumer transaction distorts the true level of output coming from the digital intermediary service provider.
This treatment results in output from a service producer in Country A being recorded as output and an export from a different country in which the producer does not reside. While there is no distortion to the GDP estimate as this is offset by the service import also being recorded, it still appears counterintuitive, as the value of the underlying product would be “traded” across borders twice, despite the product having never left the country of the buyer and producer.
DIP fees paid by producers versus consumers
Specific flows vary greatly between DIPs. While the DIS benefits both the producer and the consumer, some DIPs charge a clearly invoiced fee to the consumer and the producer while others charge only the producer who, in effect, subsidises the consumer.12 A further complication is whether the fee is explicit and therefore easily obtainable by statistical compilers, or implicit and needs to be imputed (see Annex Table 4.B.2).
Regardless of whether the fee is explicit or implicit, the main transaction (for the intermediated product) between the consumer and the producer should reflect the full value that the buyer pays less the fee paid by the buyer to the DIP (if there is one). Annex Table 4.B.2 shows an example. The intermediation fee of $12 is split evenly between the producer (who pays $6, recorded as intermediate consumption) and the consumer (who pays $6, recorded as Household final consumption). The amount recorded as paid by the consumer to the producer reflects the full value that the buyer pays ($100) less the fee paid by the buyer to the DIP ($6).
Annex Table 4.B.2. DIS paid by both producer and consumer
Alternatively, if no fee is charged to the consumer or no information on the payment of fees is known, the payment between consumer and producer should reflect the value of the good or service being intermediated plus the intermediation fee ultimately paid by the producer as intermediate consumption. The example in Annex Table 4.B.3 shows the full $100 being paid to the producer as they are responsible for paying the full $12 intermediation fee.
Annex Table 4.B.3. DIS paid solely by producer
The total amount of Household Final Consumption and Gross Value Added is the same, regardless of if the fee is paid by both the producer and the consumer, as in the first example, or solely by the producer, as in the second example.
Notes
← 1. This overriding definition of an asset is if the “goods and services are used in production for more than one year” (SNA §10.33) (UNSD, Eurostat, IMF, OECD, World Bank, 2009[18]).
← 2. In this handbook, “cloud computing”, “cloud services”, and “cloud computing services” are treated as having the same meaning. In some conceptualisations, “cloud computing” is framed as the subset of cloud services in which cloud computers are used to perform computational tasks. In this way, software-as-a-service applications are a form of cloud computing. In that view, services such as cloud storage are outside the scope of cloud computing as they are mainly “passive” rather than focused on “active” computation. Nevertheless, cloud storage involves the provision of network connectivity, data storage and indeed some processing power so are considered the same product in the Digital SUT.
← 3. For example, cloud computing capacity may be billed for each second during which a user is running their code (although often a minimum charge of 60 seconds or 1 hour may apply).
← 4. Furthermore, the ICT products that firms continue to buy are likely to be getting cheaper in real and quality-adjusted terms.
← 5. There are examples where the cloud service is provided for free as a mechanism to assist in generating data that is then monetised. However, on most occasions these examples are more consistent with a freemium pricing model than with true free digital services, as discussed in the 2008 SNA update guidance note (https://unstats.un.org/unsd/nationalaccount/RAdocs/DZ3_GN_Free_Digital_Products_Core.pdf).
← 6. The CPA is the European Union’s official classification of products by activity. Within the CPA the sub class is 63.11.1 “Data processing, hosting, application services and other IT infrastructure provisioning services”.
← 7. SaaS refers to Software as a Service, PaaS refers to Platform as a Services, and IaaS refers to Infrastructure as a Service. The final report from the Eurostat Task Force provides more details on each of these definitions.
← 8. While the explicit fee charged by the DIP can be to either the producer or consumer, or to both, in almost all cases a charge is made to the producer. The consumer is charged in some but not all cases.
← 9. Work in this area was presented to the Voorburg Group on Service Statistics, see (Hernandez Santacoloma, 2022[118]; Boey, 2022[117]).
← 10. While CPC is explicitly mentioned, the product classes mentioned in the following paragraphs are also represented in the equivalent regional classifications such as CPA, NAPCS. For example, 79.11.1: Travel agency services for transport reservations, 79.11.2: Travel agency services for reservation of accommodation, cruises and package tours, and 79.90.3: Other reservation services n.e.c, in the CPA. The likely equivalent in the NAPCS would be 3150101: Reservation service for passenger transportation, 31502: Lodging reservation service, and 31803: Other leisure and travel services, n.e.c..
← 11. This challenge could be addressed via sharing of information between countries on MNEs’ revenue, including importantly, the source of the revenues. Such a solution has been discussed in regard to the challenges faced by globalisation, with initial programs beginning to be developed such as Eurostat’s Early Warning System, https://ec.europa.eu/eurostat/web/economic-globalisation/early-warning-system#:~:text=The%20EWS%20is%20a%20structured,work%20on%20concrete%20restructuring%20cases.
← 12. On top of the variance in business models, there can be occasions when separately charged fees are reduced or waived. Often this is promotional and should be viewed as a pricing decision rather akin to “buy-one-get-one-free”. On these occasions, the consumption is not treated as a transfer even though conceptually it could be considered as one. Rather the situation is incorporated as a price change. Some DIP business models include the payment of a rebate to consumers who purchase through the platform. Again, this should be considered as an adjustment to the market price paid (SNA §3.121) rather than an additional flow between DIP and consumers.