Content uploaded by Mick Silver
Author content
All content in this area was uploaded by Mick Silver on Jul 04, 2018
Content may be subject to copyright.
EUROSTAT REVIEW
ON NATIONAL ACCOUNTS
AND MACROECONOMIC
INDICATORS
1/2018
EUROSTAT REVIEW
ON NATIONAL ACCOUNTS
AND MACROECONOMIC
INDICATORS 1/2018
Printed by Imprimerie Centrale in Luxembourg
Manuscript completed in June 2018
Neither the European Commission nor any person acting on behalf of the Commission is
responsible for the use that might be made of the following information.
Luxembourg: Publications Oce of the European Union, 2018
© European Union, 2018
Reuse is authorised provided the source is acknowledged.
The reuse policy of European Commission documents is regulated by Decision 2011/833/
EU (OJ L 330, 14.12.2011, p. 39).
For any use or reproduction of photos or other material that is not under the EU copyright,
permission must be sought directly from the copyright holders.
For more information, please consult: http://ec.europa.eu/eurostat/about/policies/
copyright
The information and views set out in this publication are those of the authors and do not
necessarily reect the ocial opinion of the European Union. Neither the European Union
institutions and bodies nor any person acting on their behalf may be held responsible for
the use which may be made of the information contained herein.
Theme: Economy and nance
Collection: Statistical books
Print ISSN 2443-7832 Cat.: KS-GP-18-001-EN-C
PDF ISSN 1977-978X Cat.: KS-GP-18-001-EN-N
Contents
EURONA — Eurostat Review on National Accounts and Macroeconomic Indicators 3
Contents
Editorial 5
R&D capitalisation: where did we go wrong?
Mark de Haan and Joseph Haynes 7
How to measure hedonic property price indexes better
Mick Silver 35
Mixed-form indices: a study of their properties
Bert M. Balk 67
Dynamic pricing as a challenge for consumer price statistics
Christian Blaudow and Florian Burg 79
In memoriam: Peter Hill (1929-2017) 95
Aims and scope
EURONA — Eurostat Review on National Accounts and Macroeconomic Indicators4
Aims and scope
EURONA is an open access, peer-reviewed, scholarly journal dedicated to National Accounts
and Macroeconomic Indicators. EURONA aims at providing a platform for researchers,
scholars, producers and users of macroeconomic statistics to exchange their research
findings, thereby facilitating and promoting the advancement of National Accounts and
Macroeconomic Indicators.
EURONA publishes empirical and theoretical articles within the scope of National Accounts
and Macroeconomic Indicators, as well as articles on important policy uses of these statistics.
They may relate to both users’ and producers’ interests, present subjects of general relevance
or investigate specific topics.
EURONA is non-partisan and applies the highest standards to its content, by emphasising
research integrity, high ethical standards, validity of the findings and cutting edge results.
EURONA gives room to all viewpoints.
The articles published in EURONA do not necessarily reflect the views or policies of the
European Commission.
Website: http://ec.europa.eu/eurostat/web/national-accounts/publications/eurona
Contact: ESTAT-EURONA@ec.europa.eu
Editors
Paul Konijn (1), Eurostat
Editorial board
Silke Stapel-Weber, Eurostat
Albert Braakmann, Statistisches Bundesamt
Gerard Eding, Centraal Bureau voor de Statistiek
Rosmundur Gudnason, Statistics Iceland
Robert Inklaar, University of Groningen, the Netherlands
Sanjiv Mahajan, Office for National Statistics
Gabriel Quiros, International Monetary Fund
Philippe Stauffer, Federal Statistical Office
Peter van de Ven, Organisation for Economic Co-operation and Development
(1) Paulus.Konijn@ec.europa.eu
Editorial
EURONA — Eurostat Review on National Accounts and Macroeconomic Indicators 5
Editorial
Three of the four articles in this issue of EURONA were presented at the meeting of the Ottawa
group on price indices in Germany, May 2017. The Ottawa group is one of the so-called ‘city
groups’ that work under the auspices of the United Nations Statistical Commission and has
been discussing and developing the theory and practice of price indices since 1994.
Mick Silver’s article discusses the measurement of (residential and commercial) property
price indices. The main difficulties in measuring property price indices are the infrequency
of transactions and the heterogeneity of properties. Hedonic regression methods are
recommended in such circumstances and the author discusses the best practices to apply
these methods.
Christian Blaudow and Florian Burg discuss the challenges posed to consumer price statistics
by the increase in dynamic pricing practices for consumer price indices. Dynamic pricing
refers to the frequent and automated adaptation of prices charged for products (mainly sold
online) in reaction to, for example, changes in demand or supply. The authors analyse data
obtained by web scraping, in other words, the automatic collection of price data on the
internet, to examine the extent of such practices.
One session of the 2017 Ottawa group meeting was dedicated to the memory of Peter von
der Lippe (1942-2016), who devoted his working life to economic statistics with a special
interest for price indices. He was especially passionate about the subject of chain indices, of
which he was one of the most vocal opponents. Bert Balk’s commemorative paper deals with
‘mixed-form’ indices, in other words, indices that behave as direct (fixed base) indices in the
short run (on a monthly basis) but as chain indices over the long run (on an annual basis).
However, this issue opens with an article from Mark de Haan and Joseph Haynes on an
important and topical national accounts issue: the capitalisation of research and development
expenditures, specifically in the context of globalisation. The main question is how to decide,
in a world of global value chains, at which location a specific intellectual property product (for
example, an R&D asset) is used in production. The authors describe and analyse two specific
cases of multinational enterprises that clearly demonstrate the challenges and risks for the
quality of economic statistics. To mitigate these risks, the authors conclude that enhanced
cooperation and data sharing between national statistical institutes is necessary.
Finally, the issue closes by paying tribute to T. Peter Hill (1929-2017), who made numerous
contributions to national accounts and price statistics during his career, most notably as the
driving force behind the SNA 1993. The obituary, written by his son Robert J. Hill, is reproduced
courtesy of the Review of Income and Wealth.
Paul Konijn
Editor of EURONA
Editorial
R&D capitalisation:
where did we go wrong?
MARK DE HAAN, JOSEPH HAYNES
Abstract: this paper is an attempt to contribute to the discussion of research and development
(R&D) capitalisation in the system of national accounts. The paper first spells out under which
conditions knowledge creation truly leads to fixed assets in the national accounts sense.
As a next step, R&D capitalisation is examined in the context of globalisation. One of the serious
problems that multinational enterprise (MNE) groups present for macro-economic measurement
is the issue of assigning economic ownership of R&D, and intellectual property (IP) more generally,
to the various fractions of a global value chain and therefore to domestic economies. This is an
issue for which international guidance is currently incomplete and still under research by national
accountants. In this paper the discussion of IP focuses largely on R&D.
By analysing real world companies and their production processes this paper aims to
highlight some of the issues with the current recording treatment around IP. This translation
of information on the MNE group’s business structure to the national accounts framework will
give an indication of real world distortions that national accountants may encounter when
measuring the activities of MNE groups on a domestic economy basis.
All the information contained within this paper relating to these MNE groups is taken from
previously published publically available sources. There may be deficiencies in the way
the characteristics of these MNE group structures are being revealed by these sources. We
nevertheless take these available sources as the starting point of this paper with the main purpose
of highlighting the complexities of recording these structures in the national accounts.
This paper offers a number of proposals for improvements though definite solutions to the
issues are not possible in one paper alone. Perhaps the greatest contribution of this paper is
in highlighting clearly the need for openness and data sharing between national statistics
institutes (NSIs). Accurate recording of the activities of MNE groups requires cooperation and
data sharing at a far greater level than NSIs have previously been willing to do.
JEL codes: E01, F62
Keywords: national accounts, intellectual property, globalisation
(1) National accounts department, Statistics Netherlands.
R&D capitalisation: where did we go wrong?
EURONA — Eurostat Review on National Accounts and Macroeconomic Indicators8
1
1. Introduction
(2) The misplace d conceptual argument in which pu blic R&D is compared with public
infrastructure is discussed later on in this paper.
A significant innovation in the latest System of National Accounts (SNA) update (2008 SNA)
was the capitalisation of expenditure on research and development (R&D). In the process of
the SNA update, Statistics Netherlands produced several papers on this issue (de Haan and
van Rooijen-Horsten (2004) and van Rooijen-Horsten et al. (2007)). These papers highlighted
several data issues such as: the translation of Frascati Manual (OECD (2015)) based R&D
statistics to national accounts data; assessing service lives of R&D assets; and dealing with
possible overlaps between R&D and computer software. This kind of guidance was later
formalised in the OECD’s Handbook on Deriving Capital Measures of Intellectual Property
Products (2009). While the 1993 SNA implementation included the introduction of computer
software capitalisation for which the first country results showed a disparity of applied
methods and results, the introduction of R&D capitalisation was ‘managed’ in a more careful
way. Unfortunately, we cannot conclude that R&D capitalisation in the national accounts has
been totally successful.
In the papers produced by Statistics Netherlands, two conceptual concerns were brought to
attention:
• R&D in the public domain does not necessarily comply with the general definition of an
asset in the SNA sense. Economic ownership of public knowledge cannot be claimed by
one particular economic agent;
• Guidance on how to account for R&D flows and stocks inside multinational enterprises
(MNEs) is totally lacking.
Supporters of the first proposition (for example, representatives from Statistics Denmark,
Statistics Netherlands and the United Kingdom’s Office for National Statistics) ’lost the battle’.
Ultimately, it was decided that R&D expenditure, both public and private, should be treated
equally as fixed assets in the 2008 SNA. The arguments supporting this choice were pragmatic
rather than conceptual. Our impression is still that publicly available knowledge contrasts with
the general SNA definition of an economic asset (2). This broad demarcation of R&D assets
is also ambiguous and creates implausible outcomes. Therefore, we revisit this issue in the
subsequent section of this paper before moving on to the issue of globalisation.
In recent years, a second issue on R&D within MNE groups and globalisation has received
increasing attention. For national accountants, one of the key challenges of economic
globalisation is explaining how capital services of intellectual property (IP) enter globally
organised production chains. Several developments are complicating this globalisation
puzzle. Firstly, the international fragmentation of production chains, inside or outside MNE
structures, may imply that business functions such as R&D and software development (in
other words, product development and design, development of software inputs) are being
separated and (spatially) disconnected from the process of physical transformation (the actual
manufacturing of the good embedding the IP). Secondly, production chain fragmentation
may also enter the stages of physical transformation. Examples of highly fractured and
specialised manufacturing webs are those found in the automobile or aircraft industries.
R&D capitalisation: where did we go wrong?
EURONA — Eurostat Review on National Accounts and Macroeconomic Indicators 9
1
Nowadays, some manufacturers entirely offshore the physical transformation stages of
production; such ‘production arrangers’ are also called factoryless goods producers (FGPs).
The issue of FGPs was intensively discussed in the UNECE task force on global production
(2015). Questions about their economic classification and the kinds of transaction these
companies are generally engaged in were, unfortunately, not brought to a final conclusion.
Both issues are closely linked to recording R&D or, more generally, IP flows and stocks.
R&D capitalisation suggests that IP products can be accounted for like any other fixed asset in
the national accounts. Our view on globalisation is that this is not the case. This point is picked
up in Section 3 of this paper.
An additional complicating factor is that IP, or intangible assets more broadly, may become
a vehicle for tax planning. MNE groups may locate their IP and report related IP revenues (in
other words, royalties) in low tax jurisdictions and subsequently charge affiliated companies,
which report substantive shares of the group’s turnover, for the use of the IP. Such tax
planning arrangements may involve a range of special purpose entities (SPEs) located in a
variety of countries. A national accountant is usually able to observe only fragments of the
tax planning arrangement and is easily misled by the information being obtained at the level
of individual SPEs, or other entities in a tax planning arrangement. Judgements on substance
or divergences in legal vis-à-vis economic ownership are extremely difficult. This is the main
issue covered in Section 4.
Section 5 winds up with (tentative) conclusions and suggestions for future work.
2. The wheel of knowledge and IP
creation
Knowledge cannot be valued in money terms. Any attempt to do so is doomed to fail as
the importance of knowledge to society cannot be comprehensively evaluated in terms
of all ‘capital services’ obtained by society from our common knowledge base. One crucial
characteristic of knowledge is its use for purely scientific reasons, in other words, building up
new knowledge. Knowledge creation inherently depends on existing knowledge. We call this
the ‘wheel of knowledge’ (which also happens to be a videogame).
Another important problem to confront is that knowledge itself does not depreciate. Codified
knowledge may get lost in the course of catastrophic losses (for example, a fire in a library or
a computer crash), which is according to the SNA not the same as depreciation. Crucial too
in the process of knowledge creation is that the complementary tacit knowledge, or human
capital, is being maintained, or even expanded, by our educational systems.
In the process of developing an electric automobile for the 21st century one cannot say that
the required knowledge obtained in ancient times, say the invention of a wheel millennia ago,
is less significant to the car than more recent inventions, for example, the development of
powerful batteries. As such, we cannot argue that the invention of a wheel is at this point of
time (partly or fully) depreciated. We are still enjoying, as ever, the fine properties of a wheel.
R&D capitalisation: where did we go wrong?
EURONA — Eurostat Review on National Accounts and Macroeconomic Indicators10
1
Equally, we cannot say that contributions from ancient philosophers like Pythagoras or
Socrates to contemporary thinking have become less relevant and should therefore be
depreciated. But, if knowledge does not depreciate then the ‘wheel of knowledge’ becomes
larger and larger, year after year.
How does this thinking contribute to national accounting? The last two versions (1993, 2008)
of the SNA underscored rightfully the increasing significance of knowledge as a production
factor. Business value and profits increasingly rely on tacit knowledge (human capital) and
codified knowledge (IP products). This is why computer software, artistic originals, mineral
exploration and R&D were included in the SNA list of fixed assets (not human capital which is
another story).
This issue of whether IP products have equal properties as other (tangible) fixed assets is
picked up in the subsequent sections of this paper. The minimum requirement is that IP
products should comply with the general definition of an asset: they are subject to economic
ownership and provide future benefits to their owner. In addition, a fixed asset must be the
outcome of production.
With respect to intangible assets these conditions should be given careful consideration. In
relation to R&D performed by businesses we can safely assume that companies are able to
claim the benefits from the R&D they fund or carry out themselves. As high-tech companies
may spend up to 10 % of their turnover on R&D, it is quite likely that these companies will be
receiving a reasonable return to R&D capital and are capable of claiming R&D ownership by
patenting or other ways of limiting access.
In the context of globalisation, this paper explains that at the level of a multinational
enterprise the concepts of ownership and obtaining related benefits are conceptually sound
and applicable. When stepping down to the level of affiliated companies, or when assessing
ownership and R&D returns at the level of the country where these affiliates are resident, both
concepts become fuzzy and less easily applicable.
We think this is a serious issue. If national accountants are not able to explain how R&D is
linked to production and output, they are not capable of accounting properly for R&D flows
and stocks. These concerns are picked up in the subsequent sections of this paper.
De Haan et al. (2004) raised the question of what are the conditions under which R&D
complies with the general SNA asset definition (at least at the level of a multinational
enterprise). They concluded that due to the exclusive access to knowledge acquired from R&D,
the owner may exert a certain level of market power which has a clear and distinct market
value. This knowledge may be translated into products with, in the eyes of the consumer,
unique and much appreciated properties, not found in the products offered by rival
suppliers. The service obtained from knowledge assets will deteriorate in line with the loss
in monopolistic power that the owner will inevitably experience over time. Competitors will
eventually be able to copy the invention or may develop variants themselves, by way of new
R&D projects, with product properties which outperform previous product innovations.
This loss in market power causes the knowledge asset to depreciate over time. This
depreciation is by definition the outcome of obsolescence as R&D or IP generally will not be
subject to wear and tear. The knowledge itself will not disappear, it may generate a positive
R&D capitalisation: where did we go wrong?
EURONA — Eurostat Review on National Accounts and Macroeconomic Indicators 11
1
contribution to society for many years, yet its commercial value will inevitably decline. This
distinction between knowledge and its possible commercial value is of crucial importance.
The knowledge as obtained from R&D will not depreciate. However, access exclusiveness and
its potential commercial value will depreciate. Depreciation refers to the fact that a patent
(or exclusive user rights more generally) is time limited and the progression of technology
inevitably implies advancing obsolescence.
As a thought experiment it may be worth considering the (part fictional) story of the
discovery of penicillin by Alexander Fleming and his refusal to take out a patent, believing
that the discovery was too important to limit its use. As national accountants, the question
we should be asking is whether the discovery of penicillin therefore led to a fixed asset? If
neither Fleming nor anyone else could claim economic ownership and accrue future benefits
due to the knowledge being freely available and usable then there is no fixed asset. Instead,
there is only knowledge. However, had Fleming opted to obtain a patent then there would
have been an economic owner and a fixed asset. This example shows that it is the patent, or
more generally obtaining exclusive ownership, that gives rise to the fixed asset and not the
knowledge or discovery itself. Where knowledge is not protected by any means, a patent or
secrecy, a fixed asset cannot be recognised.
Sharing profitable knowledge incurs a cost as it may delimit the monopolistic power of the
initial owner. One should be aware that commercial success is often the combination of
codified knowledge (the R&D asset) and tacit knowledge (the complementary human capital
required to translate knowledge into successful product blueprints). Copying tacit knowledge
may be harder than copying R&D assets. This means that exclusive ownership of scientific
knowledge is not necessarily safeguarded by patenting but can equally be obtained by way
of secrecy or by exclusive access to the complementary tacit knowledge.
The service lives of patents in the various scientific areas (for example, pharmaceuticals,
electronic appliances, information technology (IT)) may be a reasonable proxy for assessing
service lives of patented and non-patented R&D projects. This is how many national statistical
institutes (NSIs) go about assessing service lives of R&D assets. As unsuccessful projects are
unavoidable in the process of seeking commercial success, capitalising expenditure on both
successful and unsuccessful projects is defendable in the attempt to approximate the overall
market value of business R&D capital.
We have seen that the 2008 SNA recommends all R&D to be capitalised, including business
research and non-commercial research (for example, university research). The argument used
in the 2008 SNA for also capitalising the latter type of research is that university R&D is a public
good which is beneficial to society for a longer time period, similar to public roads or bridges.
The arguments below speak against this analogy. The 2008 SNA (paragraph 10.98) explains
that ’the knowledge remains an asset as long as its use can create some form of monopoly
profit for its owners. When it is no longer protected […] it ceases to be an asset’. Yet, this
wording could be read such that the 2008 SNA itself already rejects the idea of publically
shared knowledge as an asset in the SNA sense.
First, looking at the resemblance of public research and public bridges or roads there
is generally no confusion about economic ownership of the latter (we leave aside the
complexity of public-private operations which is not relevant to this discussion). The
R&D capitalisation: where did we go wrong?
EURONA — Eurostat Review on National Accounts and Macroeconomic Indicators12
1
government is responsible for maintaining the road and may even be liable for damages
to users caused by deficiencies. The government has decision-making power: it may, for
example, decide to sell the road to a private operator or put the underlying land to another
(public) use. In this sense public infrastructure meets the definition of a fixed asset. This may
not always be the case for R&D in the public domain. Once in the public domain the R&D
asset has become a pure public good. To consider this more fully we first break down, non-
exhaustively, the kinds of research projects that are carried out in the public domain.
Government bodies may conduct scientific research for various reasons. Some of this
research may be linked to commercial purposes and may even be patented (for example,
supporting agriculture or enhancing the circular economy or, more generally, improving
the environmental performance of businesses). This type of research is quite comparable to
business R&D. When businesses are able to claim the (commercial) revenues of this public
research, one may argue that this R&D has been transferred to them. This exclusivity gives rise
to economic ownership and therefore is an indicator that such public R&D should be recorded
as a fixed asset. Given its purpose this dedicated R&D is likely to be subject to obsolescence
as newer techniques may replace older ones. So, this R&D depreciates in an economically
meaningful way. Crucial in this context is whether or not the government unconditionally
grants all parties access to this knowledge. If so, the knowledge is in fact a public good and
cannot be an economic asset in the SNA sense.
Another example is defence-related research. This research may be performed either by
commercial or government bodies. One may expect that this research is conducted under
strict secrecy since its key purpose is obtaining a military advantage over (potential) enemies.
In relation to dedicated military research there will generally be no misunderstanding
about ownership and the beneficiaries of this research. By not publicising such research the
government maintains a quasi-monopoly position and is the economic owner of a fixed asset.
In the arms race, equal steps taken by potential enemies will inevitably lead to diminishing
the defensive advantages of research projects over time, again implying this research can be
depreciated in a meaningful way, even though the purpose of this R&D may be (partly) non-
commercial.
Another part of R&D performed in the public domain is purely non-commercial, scientific,
university-based research. Obviously the origin of scientific research is being claimed by
their authors in scientific journals. This is not the same as claiming economic ownership.
The main purpose of this research is extending science which requires, among other things,
allowing full access to scientific results, for verification purposes or to allow other scholars
to build on published findings. The main purpose of university research is feeding scientific
debate. In the strict context of university research, notions such as economic ownership and
economic revenue become meaningless. Scientific results are shared and applied by others
for the sake of conducting new research. Once academic research has been published the
revealed knowledge immediately becomes not only a pure public but also a free good (3). A
pure public good cannot be a fixed asset as no single owner exists who can claim economic
ownership and earn any future benefits. Therefore, this element of public R&D does not meet
the definition of a fixed asset as it is not subject to economic ownership.
(3) A public goo d is one where individuals cannot b e effectively excluded from its use, while
its use by one individual does not reduce av ailability to others. Publi c R&D is also a free
good as its us e is principally unlimited and not su bject to depreciation.
R&D capitalisation: where did we go wrong?
EURONA — Eurostat Review on National Accounts and Macroeconomic Indicators 13
1
This paper has already argued that the depreciation of business R&D is the outcome of two
factors. First, competitors in the market may catch up (dispersion or sharing of knowledge).
Second, new research and innovations may outperform previous innovations which will
inevitably lead to its obsolescence. Following this line of thinking one may argue that the R&D
assets as owned by companies will eventually be transformed into R&D in the public domain.
At that moment the R&D ceases to be an asset in the SNA sense as it has become public
knowledge.
This leads to the following conclusions. The main purpose of most academic research is
generating public knowledge over which ownership cannot be claimed by one economic
agent, not even a government. The outcome (we hesitate to call this revenue) of research
is commonly shared by academia. Therefore, academic research, once published, does not
meet the definition of an asset. Furthermore, academic research and knowledge in general is
not subject to economic depreciation as service lives are, in principle, indefinite. Depreciation
functions applied to academic research lack any conceptual underpinning.
The intrinsic inconsistency of such calculations can be underscored by the following
representation of a production function of academic research (in ISIC Rev.4 Division 85). In
case of public education and research, the SNA convention is to value output (X) as the sum
of costs. Let us assume a purely scientific research institute (perhaps allied to a university).
Its main current costs are the salaries of researchers (L). According to the 2008 SNA the
output of this research institute is R&D which is recorded as gross fixed capital formation. Its
depreciation feeds back into the production account of the research institute. We assume that
the salaries and labour input are constant over time. We also assume geometric depreciation
(d). The production function is represented by equation (1). The capital accumulation function
is represented by equation (2).
(1) Xt = L + d × R&Dt
(2) R&Dt = (1−d) × R&Dt−1 + Xt−1
(3) Xt − Xt−1 = d × L
So the remarkable outcome of the SNA convention is that while labour input (L) remains
constant over time, each year the R&D output of this research institute will increase linearly by
d × L while the R&D capital stock will expand on an annual basis by L.
What is modelled by equations (1) and (2) is the ‘expanding wheel of knowledge’ which has
nothing to do with economic accounting. According to equations (1) and (2), government
consumption would increase annually by d × L according to the SNA convention of non-
market output valued at sum of costs and ignoring labour productivity changes, while
intuitively one would agree that given constant labour input the research institute would
generate constant output.
In other words, the R&D output of this research institute should be recorded directly as
government consumption and not as gross fixed capital formation. It should be emphasised
that either the consumption or investment option will have a similar impact on GDP. Though
the investment option leads to the undesirable disturbance of recursive GDP additions as the
R&D capitalisation: where did we go wrong?
EURONA — Eurostat Review on National Accounts and Macroeconomic Indicators14
1
consumption of fixed capital will additionally add to the output of the government sector,
measured as the sum of costs.
(4) http://www.oecd.org/sti/ind/global-value-chains.htm.
(5) https://www.digitaltrends.com/mobile/IPhone- cost-what-apple-is-paying/.
(6) http://www.aeronewst v.com/en/industry/commercial-aviation/3707-boeing-787-
dreamliner-structure-part s-from-around-the-globe.html.
3. Corporate R&D property and global
R&D networks
A. Introduction
At least two complicating factors limit our understanding of how the services of R&D capital
enter the global production chain. The first one is the global fragmentation of production
and, within the so-called global value chain, the disconnected supply of physical and
intangible inputs. The second is that R&D creation itself can be subject to interlinked global
research networks. Both issues are considered in this section.
B. Globally fragmented value chains
Global production contrasts with the idea of ‘national’ accounting and this is why so much
effort has recently been put into developing guidance supplementing the 2008 SNA (UNECE
(2011) and UNECE (2015), Eurostat (2014)). As explained by the OECD, international production,
trade and investments are increasingly organised within global value chains, where the
different stages of the entire production process, from product design all the way to product
distribution and after sales services, are located across different countries (4).
IP and IT play a fundamental, enabling, role in the global value chain. For example,
communication networks enable product development and design to be geographically
disconnected from goods fabrication.
The well-known value added breakdown of an iPhone indicates that the physical parts and
assembling costs represent roughly half the iPhone’s retail price (5). All of the remaining value
added generated by the iPhone’s production is connected to intangible inputs such as R&D,
design, marketing and presumably activities such as supply-chain management. The income
is generated in different regions of the world.
Graphic presentations of global supply chains show well the geographic distribution and
clustering of manufactured parts and assembling making up the iPhone, a motor car or an
airplane (6). How R&D feeds in to these global value chains is harder to explain. This issue is
often ignored as analysis of global production networks often limit themselves to the physical
transformation segments of global production.
However, if according to the 2008 SNA R&D is a fixed asset, like any other (tangible) fixed asset,
the national accounts should be able to explain which entities inside the MNE structure are
actually investing in R&D and consuming the concomitant R&D services. In other words, we
R&D capitalisation: where did we go wrong?
EURONA — Eurostat Review on National Accounts and Macroeconomic Indicators 15
1
should be able to explain which (affiliated) entity (in which country) owns the R&D asset and
is accountable for its depreciation or more generally the costs of using the R&D asset. Similarly,
the accounts should be able to explain how R&D and IP contribute to output and multifactor
productivity on a country-by-country basis.
There are several reasons why these questions are difficult to answer:
1. Basic and applied research provide capacity-enhancing technologies which facilitate
product innovation but will not directly result in blueprints of new products (7). In other
words, in contrast to product development, basic research misses a direct link to the
goods and services outputs. This being the case, the head office of an MNE seems to be
the most obvious candidate for economic owner of this truly corporate R&D property. It is
quite likely that head offices take the (funding) decisions on basic research investments in
line with the overall corporate innovation strategy. The latest Frascati Manual (OECD (2015),
par. 3.11) confirms this view: ’In large and complex organisations, decisions concerning
the strategic direction and financing of R&D activities units tend to occur at a higher
organisational level than does the day-to-day management of R&D operations. (...) These
decisions can cut across national borders, thus raising a challenge for the statistical
authorities and agencies, whose responsibility is often limited to gathering information
from resident units’. In other words, allocation of basic and applied research or allocating
its capital services, to the goods manufacturers inside the MNE is inherently without
economic meaning.
2. R&D is different from most activities performed by a corporation in the process of its
operation. Research is typically not performed with the expectation of immediate profit.
Instead, it is focused on the long-term profitability of a company. As such, the way in
which R&D feeds into the production function is unlike other fixed asset categories. Even
for computer software, its presence in a local computer or in the cloud is needed in the
course of the transformation process in order to deliver its capital services. Obviously, a
similar presence is also required for tangible capital items. In contrast, once a potentially
successful recipe for a new medical drug, or the technical design of a new motor car, has
been developed, the production process will be set up according to this new blueprint,
after which the R&D capital has delivered its contribution to output. This does not imply
there is no return to R&D capital involved in the course of producing the medical drug or
motor car. However, this different mechanism by which R&D contributes to output implies
that the R&D asset is not necessarily found in the balance sheet of the entity engaged
in the transformation, in other words, the actual fabrication of the drug or motor car.
Instead, the R&D asset may be on the balance sheet of an affiliated company (in a low tax
jurisdiction) or may not feature on a balance sheet at all, as corporate accounting rules are
generally quite restrictive in capitalising R&D.
3. Inside or outside the MNE group’s scope, a production network is not just the sum of its
component parts. Product development and design are activities typically carried out
by the arrangers or principal entities inside global production networks. So these entities
are often the main R&D investors inside the global value chain. This is also according to
the explanation of factoryless goods producers (FGPs) in the Guide to Measuring Global
Production (UNECE (2015)). In this regard FGPs and head offices of MNE groups carry out
(7) Basic and applie d research represents 20 % of total busine ss R&D in the United States:
https://www.nsf.gov/statistics/2017/nsf17320/.
R&D capitalisation: where did we go wrong?
EURONA — Eurostat Review on National Accounts and Macroeconomic Indicators16
1
similar tasks: they both manage global supply chains with the aim of optimising network
synergy. They are both expected to bring together the intangible and physical stages
of global production. The main difference is that FGPs have outsourced the physical
transformation activities while inside the MNE these activities are (partly) carried out by
affiliated companies. Also different from an FGP, a head office will not necessarily report
turnover from sales of goods. Alternatively, this turnover is expected to be reported by
one or several of the MNE group’s affiliated goods producers. As product and process
innovations obtained from R&D may affect several stages in the production network, from
a holistic point of view it seems defendable that the FGP or head office is the typical stage
where R&D enters the global production chain. It does not seem feasible to assign R&D
inputs to the separate transformation stages in the production chain. One R&D asset, or
one piece of knowledge, may lead to multiple product innovations and the enhancing of
profits of several business units inside a single MNE group.
4. In the context of an FGP arrangement, R&D may lead to innovations of products
assembled and supplied by non-affiliated contract producers in various parts of the
world. The value added and profits generated by these contract producers will typically
omit the return to R&D assets as their production costs, and thus their output prices,
will not include R&D costs. The R&D returns are directly captured by the principal of the
global production arrangement. Discussions in the global production taskforce (UNECE
(2015)) showed that, in the case of an FGP, national accountants have great difficulties
in explaining the nature of the transaction between the contract manufacturer and
the principal: the purchase of a good or the purchase of a (manufacturing) service. Our
conclusion is that in economic terms the good purchased from the contractor differs
fundamentally from the good sold to consumers, even though in physical terms no
distinction can be made. This may have implications for the commodity classification in
the national accounts and the balance of payments. In the classifications of goods not
only are the physical characteristics of the product relevant, but also the conditions under
which the product is transferred from one economic owner to another.
5. In the context of an MNE, the output price of the affiliated contract producer may
indeed include the return to R&D capital as its output may be directly distributed to
end consumers. However, the required R&D assets may, or may not, be found on the
balance sheet of the affiliated manufacturer. It is still possible that headquarters, in their
role as global production arrangers, provide the R&D inputs, possibly without any intra-
company flows of R&D services being observed. In such a situation the R&D profits will be
repatriated to the headquarters via property income (dividends or retained earnings).
6. The latter point shows that corporate funding of R&D is not necessarily linked to how and
where the R&D is translated into commercial success. Ignoring tax planning for a moment,
from the MNE group’s perspective a spatial allocation of generated R&D income is
irrelevant as this income will eventually reach the MNE’s shareholders wherever generated.
Discussions with a number of R&D managers of Dutch multinational companies led to
the conclusion that cost redistribution is not common practice (de Haan & van Rooijen-
Horsten (2004)).
R&D capitalisation: where did we go wrong?
EURONA — Eurostat Review on National Accounts and Macroeconomic Indicators 17
1
7. Ironically, R&D cost accounting (IP-related royalty payments) within the MNE is particularly
observed in the context of tax planning arrangements. Fair competition authorities,
tax authorities and statisticians alike have to evaluate to what extent IP cost accounting
arrangements have economic substance. Looking at recent events one must conclude
that tax planning arrangements of MNE groups may place national accountants in a very
difficult position. This issue is further discussed in Section 4 of this paper.
To conclude, (national) IP economic ownership in the context of global production is still not
a well understood concept. The arguments above indicate that IP economic ownership seems
to usually coincide with the decision-making entities in the global value chain. These are the
entities that are expected to manage overall the intangible and tangible inputs of production.
However, such a view has several implications that require further examination:
• Assigning economic R&D ownership to headquarters on behalf of the MNE requires,
amongst other things, a careful examination of cross-border R&D flows as they are reported
in international trade in services statistics. R&D conducted by foreign affiliates may, or may
not, be (partly) funded by headquarters (or by sister companies) or may even have been
purchased. This means that the practicalities of such an approach need to be carefully
thought through. Some guidance is already provided by the Frascati Manual in showing a
data collection scheme for R&D expenditure at the MNE level (Figure 11.2 in OECD (2015)).
• The central product classification (CPC) should be further examined to address the
economic characteristics and output of contract producers in FGP type arrangements. For
example, the CPC should underscore that the iPhone delivered by a contract producer is a
totally different product from the iPhone purchased by a consumer.
C. Global R&D networks
R&D statistics based on the Frascati Manual (OECD (2015)) provide information on R&D
expenditure. This is without any doubt crucial information for the purpose of measuring R&D
investment. The assumption that R&D expenditure is overall a reasonable approximation of its
commercial benefits is not likely to be replaced by an alternative measurement method. The
costs of carrying out R&D and maintaining global R&D networks can be statistically observed
in a meaningful way on a country-by-country basis. The allocation of (economic ownership
of) investments of R&D networks on a country-by-country basis is a less clear concept. Of
course we can assume that the allocation of costs is representative for the allocation of
investments but this seems to be a rather shaky assumption.
R&D capitalisation: where did we go wrong?
EURONA — Eurostat Review on National Accounts and Macroeconomic Indicators18
1
Global R&D networks within MNE groups are best illustrated with the help of a few real life
examples. The technology firm Samsung has over 50 000 employees working in collaboration
on R&D spread across multiple R&D centres in South Korea as well as others in Russia, India,
China, Israel, Japan, Poland, the United States and the United Kingdom (8). Table 1.1 details
some of the R&D activities undertaken by Samsung outside of South Korea.
Another example is Philips, a leading technology company operating in the healthcare and
consumer electronics sector and one of the largest Dutch MNE groups, with its headquarters
located in the Netherlands. However, Philips also conducts R&D activities across the world as
shown in Table 1.2 (9).
(8) http://www.samsung.com/semiconductor/about-us/research-development/.
(9) https://www.philips.com/a-w/research/locations.html.
Table 1.1: The Samsung R&D network
Research institute Country Type of R&D activities
1Beijing Samsung Telecommunication China Mobile telecommunications standardisation and
commercialisation for China
2Samsung Semiconductor China R&D China Semiconductor packages and solutions
3Samsung R&D Institute India India System software for digital products, protocols for
wired/wireless net works, application and graphic
design
4Samsung Telecom Research Israel Israel Hebrew software for mobile phones
5Samsung R&D Institute Japan-Yokohama Japan Core next-generation parts and components, digital
technologies
6Samsung R&D Institute Poland Poland STB software platform development, EU STB/DTV
commercialisation
7Moscow Samsung Research Centre Russia Optics, sof tware algorithms and other new
technologies
8Samsung R&D Institute United Kingdom United Kingdom Mobile phones and digital TV sof tware
9Dallas Telecom Laboratory United States Technologies and products for next-generation
telecommunication systems
10 Samsung Information Systems America United States Strategic parts and components, core technologies
Table 1.2: The Philips R&D network
Research institute Country Type of R&D activities
1Philips Research Shanghai China Imaging systems
2Philips Research Suresnes France Healthcare
3 Philips Research Aachen Germany Healthcare
4Philips Research Hamburg Germany Imaging systems, biological modelling, computer
assisted detection
5Philips Research Asia India Healthcare
6Philips Research Africa Kenya Healthcare, design, user interface
7Philips Research Eindhoven Netherlands Healthcare and global headquar ters for all R&D
8Philips Research Cambridge United Kingdom Healthcare
9Philips Research North America United States Healthcare, artificial intelligence
R&D capitalisation: where did we go wrong?
EURONA — Eurostat Review on National Accounts and Macroeconomic Indicators 19
1
Although we did not undertake a full investigation, the literature on R&D management seems
to confirm that regional R&D facilities may support local product development as well as the
overall MNE’s longer-term research strategy. For example, Papanastassiou and Pearce (2005)
find that local R&D laboratories in the United Kingdom are mostly funded by the parent
company of the MNE group. This is considered as being powerfully indicative of the manner in
which such decentralised operations are now integral to the ways in which these companies
seek to apply existing core technologies and to regenerate and broaden the scope of these
crucial knowledge competences. It depicts a process of refocusing decentralised R&D away
from the short-term objective of assisting particular subsidiaries to apply existing technologies
to their specific competitive situation, towards positions integral to the more sustained
technological and competitive development of the MNE group. In contrast to independently
operating R&D facilities, close cooperation between the regional R&D units within an MNE is
expected to provide substantial externalities, in the form of systematic group-level spillover
benefits. Central financial participation in the funding of laboratories can be seen as crucial
in developing the necessary interdependencies between decentralised R&D units, and in
securing the cohesive growth of intra-group knowledge flows.
Some MNE groups like Apple follow quite aggressive strategies in obtaining the knowledge
required for strengthening global competitiveness. Recently Apple opened R&D units in
Berlin, the French Alps and New Zealand, all in the close neighbourhood of companies with
a strong record in certain scientific areas (for example, mapping or augmented reality). In
several cases these companies lost employees to Apple soon after Apple opened its new R&D
unit (10). This shows that the choice of location of newly-established R&D units is on occasion
solely driven by knowledge acquisition, the availability of human capital/tacit knowledge and
not by locating the R&D unit close to those MNE affiliates that are supposed to transform the
R&D into a product innovation, output and commercial success.
The existence of R&D networks within the MNE structure appears to have similar implications
for the national accounts as the existence of fragmented production chains. While the
geographical distribution of R&D costs within the MNE structure as reflected by Frascati
Manual (OECD (2015)) based statistics is likely to be reasonably well measured, the distribution
of (the economic ownership of) the created R&D assets inside the MNE is not well understood.
For smaller national firms, there will likely be a strong geographical correlation between R&D
activities and the obtained commercial gains. In those cases it is reasonable to assume that
the location of R&D activity coincides with R&D asset ownership. However, within the MNE
framework this assumption cannot generally be made on solid grounds. As R&D strategies
and R&D funding are expected to result from the overall corporate strategy, the choice of
considering R&D as genuine corporate property appears attractive. However, as mentioned
the practicalities of such a choice should be carefully considered.
When assigning R&D ownership to the head offices one should ensure that the production
accounts for each of the MNE group’s entities meaningfully represent the various fragments
of production encountered inside the MNE group. For example, each of the accounts should
sufficiently support productivity measurement (Schreyer (2018)). This implies that together
with R&D ownership, the R&D revenues need to be recorded in the accounts of the head
office. Equally, the R&D costs need to be assigned to the MNE groups’ affiliates. This is not a
(10) https://www.bloomberg.com/news/articles/2017-09-21/apple-s-global-web-of-r-d-labs-
doubles-as-poaching-operation.
R&D capitalisation: where did we go wrong?
EURONA — Eurostat Review on National Accounts and Macroeconomic Indicators20
1
new phenomenon as head offices will more broadly provide all sorts of intra-group services
to its affiliates, for example, supply-chain management services, financial services and
marketing activities.
One way to allocate all of these costs is by using allocation mechanisms such as the formulary
apportionment techniques used by Guvenen et al. (2017). The main goal of Guvenen et al.
is to allocate the generated income over those entities in the MNE which are carrying out
the actual production activities. This is an attempt to overcome the disturbances caused
by tax planning arrangements. In this paper we suggest allocating the sum of ‘overhead
costs’, or in other words all intra-group services provided by head offices, to those affiliated
companies which carry out part of the genuine economic activities. Obviously such allocation
requires a concerted action from all the NSIs involved. The outcome of this exercise should
be an economically sound allocation of the MNE group’s value added and gross operating
surplus leading to meaningful productivity statistics at the level of individual enterprises
or establishments inside the MNE group. This goal corresponds closely to formulary
apportionment allocation of profits as carried out by Guvenen et al. Please be aware that the
proposed exercise may also help to overcome some of the substantive bilateral asymmetries
witnessed for trade in services statistics today. Perhaps a concerted cost allocation of head
offices could also overcome some of the disturbances of transfer pricing.
The example presented in the annex to this paper is quite simple as all R&D costs are assigned
to one single affiliated company. But in essence it illustrates the cost reallocation proposed in
this paper.
4. Intellectual property and tax planning
One may argue that R&D capitalisation in the 2008 SNA revealed (but did not necessarily
cause!) the national accounts’ vulnerability to problems arising from globalisation, as MNE
groups may use IP assets as vehicles for tax planning. The goal of such tax planning is to
shift revenue to units within the MNE structure that are tax resident in low tax jurisdictions, a
consequence of which is that MNE groups can minimise their global tax liability. This is often
achieved through the use of royalty and licence agreements linked to IP assets. Units of an
MNE will typically be required to pay a royalty charge to another unit within the MNE for the
right to use assets intrinsic to the production process. In doing so, profit from sales in higher
tax jurisdictions can be transferred to units in lower tax jurisdictions, minimising the global tax
liability for an MNE. Such constructions are often used by MNE groups in high technology-
based industries where R&D and other forms of IP play a crucial role. The lack of a physical
presence of IP assets lends themselves to such constructions as they can be easily located
and relocated around the world at little cost. Under such conditions, the observable global
value chain of MNE groups reflects an artificial, tax-driven, reality rather than what could
be considered the true production process reflecting economic substance. We should also
note that movable tangible assets such as transportation equipment may also be subject to
tax planning arrangements as their (legal) ownership can be assigned to a leasing company
resident in a low tax jurisdiction.
R&D capitalisation: where did we go wrong?
EURONA — Eurostat Review on National Accounts and Macroeconomic Indicators 21
1
The two real life examples of Google and Nike explored in this section highlight the expected
consequences of following, as a national accountant, the legal reality as revealed in source
statistics, rather than looking around the legal reality and depicting the MNE group’s real
economic substance, which can only be seen once the entire ‘elephant’ has been observed.
It should be emphasised that all information on both cases has been obtained from public
sources that have previously been published such as news articles and business reports and
does not use information obtained for the purpose of official statistics.
A. The double Irish with a Dutch sandwich(11)
EXPLAINING THE CASE
The ‘double Irish with a Dutch sandwich’ is a name given to a legal business arrangement
which is designed to minimise the MNE’s global tax liability. This technique has most
prominently been used by technology companies, because these firms can easily shift large
portions of profits to other countries by assigning IP rights to subsidiaries abroad. From
2015 onwards, Irish tax legislation no longer allows companies to use the double Irish Dutch
sandwich for new tax plans; existing plans can be continued until 2020. The latter may have
severe repercussions for national statistics as in response MNE groups may restructure their
business and set-up alternative tax planning schemes. Business restructurings may also be a
response to recent United States tax reforms.
One of the MNE groups using the double Irish Dutch sandwich construction is Google (12). The
main ingredients, which are typical for the double Irish Dutch sandwich recipe, are as follows.
The parent company at the top of the corporate hierarchy is Alphabet Inc. This company is
based in Mountain View, California (United States). Although most of the ultimate parents of
MNE groups using the double Irish Dutch sandwich structure are resident in the United States,
this is not necessarily the case. Google Inc. sits below Alphabet Inc. in the hierarchy and is the
top of the structure for what can best be described as the everyday Google internet functions
such as its search engine, maps, e-mail. A large number of companies operating across the
world sit below Google Inc. in the hierarchy.
One of these is Google Ireland Holdings Unlimited, which is an Irish incorporated entity
managed and controlled from Bermuda — a common choice. This is a special purpose
entity (SPE) registered in Ireland but not liable for tax in Ireland. Rather, it is liable for tax in
Bermuda from where it is officially managed and controlled (13). This type of holding company
with only holding activities has no physical presence and zero employees, or only sufficient
employment to fulfil a strict legal requirement, in other words, the only employees are
directors or shareholders who are normally non-Irish residents.
(11) A detailed legal explanation of the double Irish with a D utch sandwich is given in Brothers,
J (2014), ‘From the Doub le Irish to the Bermuda Triangle’, Tax Analysis.
(12) https://fd.nl/ondernemen/1180304/google-sluisde-vorig-jaar-15-mrd-royalties-door-
nederland.
(13) Idem, see footnote (12).
R&D capitalisation: where did we go wrong?
EURONA — Eurostat Review on National Accounts and Macroeconomic Indicators22
1
Google Netherlands Holding B.V. is a Dutch resident company. It is an SPE type unit with no
employees and no activities other than financing and participating in affiliated companies (14).
This Dutch SPE receives royalty payments from Google units in Ireland and Singapore which
are directly transferred to Google Ireland Holdings Unlimited, minus a small amount for
administrative costs.
Google Ireland Limited is an Irish registered company that undertakes real economic activities
in Ireland. It also has a wider role outside Ireland of being the company that closes all deals for
Google AdWords across Europe. AdWords represents a large portion of Google’s revenue. It
has been estimated that as much as 88 % of Google’s non-U.S. revenue is recorded by Google
Ireland Limited (15). Together these Google affiliates, representing the double Irish Dutch
sandwich, operate as follows.
Google Ireland Holdings Unlimited owns various IP rights which it licences to Google
Netherlands Holding B.V. who in turn then sublicenses these rights to Google Ireland Limited.
Google Ireland Limited uses the sublicenses in its production process and generates revenue.
In doing so it is liable to pay royalty fees to Google Netherlands Holding B.V. as a result of
using the IP.
Google Netherlands Holdings B.V. is also liable to pay royalty fees to Google Ireland Holdings
Unlimited on account of the licencing agreement between the two. As such, the royalty
payments make their way from Ireland via the Netherlands back to an Irish registered
company which is however controlled, managed and liable to pay corporation tax in
Bermuda. Google Netherlands Holdings B.V. acts only to channel financial flows between
units. In comparison with the value of the royalty flows, little profit remains in the Netherlands.
The Dutch SPE is not an essential hub in the tax planning arrangement. Rather, it is an
additional insurance layer against potential withholding tax liabilities arising on direct royalty
payments. The zero rate of withholding taxes on incoming and outgoing royalty payments
between Ireland and the Netherlands allows this royalty flow to be seen as being taxed
already (though at a zero rate) meaning the potential tax liability is therefore removed.
Typically, the Dutch SPE will pay virtually identical royalty payments to the Irish holding unit as
it receives. In 2015, over 99.9 % of the royalties received by Google Netherlands Holdings B.V.
were repaid to Google Ireland Holdings (16). An overview of the Google structure is presented
in Figure 1.1.
(14) Google Ne therlands Holdings B.V., Annual repor t 2016.
(15) van Gees t, van Kleer and Smits (2015), pp. 64.
(16) As calculated based on data from Goo gle Netherlands Holding B.V., Annual repor t 2015,
publically av ailable at www.kvk.nl. Royalties received EUR 14 963 billion, roya lties repaid
EUR 14 951 billion.
R&D capitalisation: where did we go wrong?
EURONA — Eurostat Review on National Accounts and Macroeconomic Indicators 23
1
Figure 1.1: A double Irish Dutch sandwich: the Google case
NATIONAL ACCOUNTS IMPLICATIONS
There are several concerns when translating the information obtained from each of these
entities to national accounts statistics.
The arrangement requires that IP ownership is transferred from the ultimate parent (in the
United States) to the royalty and licence company in a low tax jurisdiction (Bermuda); in the
Google case this is Google Ireland Holdings. This apparent IP transfer raises several questions:
for example, would this be an IP purchase/sale, and if so, what would be a representative market
value of such an intra-MNE group transaction? But perhaps an even more fundamental issue
is whether or not this transaction has economic substance at all. Is Google Ireland Holdings,
besides the legal owner, also the economic owner of this IP? One may expect that, despite
this arrangement, strategic decisions about IP creation and allocation continue to be made
in the United States, even in cases where part of its IP ownership is transferred to an affiliated
company abroad. A practical question is whether such international intra-group IP transactions
will be recorded in all the countries involved in a symmetrical way. In other words, will the value
representing the export of the IP from the United States equal the import value as reported in
Bermuda/Ireland?
Another question is the country of residence of Google Ireland Holdings Unlimited, as this
company is registered in Ireland but managed and controlled in Bermuda and is also liable
Google Ireland Holdings Unlimited
Company
• Owns the right of IP use outside
the United States
• Sub-licenses IP rights to Google Netherlands
Google Ireland Limited
• R