Home | OECD iLibrary

Broadband infrastructure

Fixed and mobile broadband subscriptions continue to grow apace. The number of worldwide fixed broadband subscriptions increased by 72% in the last ten years, from 531.8 million in 2010 to 916.7 million in 2016. In OECD countries, fixed broadband subscriptions increased from 307.3 million in 2010 to 386.8 million in 2016, an increase of 26%. Mobile broadband growth by far outstripped fixed broadband with worldwide subscriptions increasing from 824.5 million in 2010 to 3 864 million in 2016. At the end of 2016, just over half the world’s population had a mobile broadband subscription. By way of contrast, the average for OECD countries was 99.3%. The pace of change can be rapid, however. Mobile broadband subscriptions in non-OECD countries registered a nine-fold increase over the last decade, with India adding almost 100 million broadband subscriptions in 2016 alone.

Machine-to-machine communication

The Internet of Things (IoT) refers to an ecosystem in which applications and services are driven by data collected from devices that act as sensors and interface with the physical world. This ecosystem could soon constitute a common part of the everyday lives of people in OECD countries and beyond. Important IoT application domains span almost all major economic sectors including: health, education, agriculture, transportation, manufacturing, electric grids and many more. Part of the underlying infrastructure of the IoT is machine-to-machine (M2M) communication. The Groupe Spéciale Mobile Association (GSMA) tracks the number of M2M subscriptions around the world. These data show the number of SIM cards embedded in machines, such as automobiles or sensors, which allow communication between such devices. Among G20 economies, the United States had the highest penetration (number of M2M SIM cards per inhabitant) in June 2017, followed by France and the United Kingdom. Between 2012 and Q2 2017, the number of subscriptions increased by 131% in OECD countries and 272% in the G20, although from a smaller base. The People’s Republic of China (hereafter “China”) had the largest share of worldwide M2M subscriptions (44%) at 228 million subscriptions in June 2017, representing three times the share of the United States.

Measuring the infrastructure for IoT using GSMA data on M2M

The GSMA’s definition of M2M is: “A unique SIM card registered on the mobile network at the end of the period, enabling mobile data transmission between two or more machines. It excludes computing devices in consumer electronics such as e-readers, smartphones, dongles and tablets”. The GSMA collects publicly available information about mobile operators that have commercially deployed M2M services. It then uses a data model based on a set of historic M2M connections reported at any point in time by mobile operators and regulators, along with market assumptions based on their large-scale survey of M2M operators and vendors. This pool of data is then reconciled by GSMA with their definition, normalised and analysed to identify specific M2M adoption profiles. These adoption profiles are then applied by the GSMA to all operators that have commercially launched M2M services, but do not publicly report M2M connections to produce national figures. For more information, see www.gsmaintelligence.com.

ICT technologies at the cutting edge

Technologies take time to develop and mature and may follow different development and adoption paths. Technologies that have several applications may at some point experience accelerated development – they may start to “burst”. Information and communication technologies (ICTs) are an example of bursting technologies. ICT products such as mobile phones and computers are renowned for their complexity and modularity, their rapid obsolescence, and their reliance on a wide array of continuously evolving technologies. A novel data-mining approach is used to monitor the extent to which different ICT fields emerge and develop, and to identify bursting technologies. Over 2012-15, five economies accounted for 69% to 98% of the top 20 bursting ICT technologies. Japan and Korea contributed to the development of all ICT fields whose development accelerated during this period, together accounting for 21% to about 70% of all patenting activities in these bursting ICT fields. The United States led the development of ICT technologies related to payment protocols (34%), transmission arrangement (28%) and digital video signal coding (28%). China was among the top five economies developing technologies in most bursting ICT fields, and was particularly active in light modulation and control inventions (28%). A few European economies, namely Sweden, Germany and France, also featured among the top five leaders of some bursting ICT fields.

Identifying acceleration in technological development

Patents protect novel inventions and technologies, and patent data can help investigate a number of policy-relevant issues related to innovation and technological development. A new data mining approach called “DETECTS” (see Dernis et al., 2016), exploits information contained in patents to identify technologies whose development increases sharply (i.e. “bursts”), compared to previous levels and to the development of other technologies, and maps the time it takes for such dynamics to unfold. A technology field is said to burst or accelerate when a substantial increase in the number of patents filed in the field is observed. DETECTS monitors such acceleration in relative terms (i.e. compared to past development patterns in the field and relative to the pace of development in other fields). Monitoring fields in which accelerations occur is vital for policy making, as developments tend to persist in these areas over the short and medium term. Furthermore, information contained in patents about the technologies themselves and the geographical location of patent owners and inventors enables the identification of economies leading such technology developments, and can shed light on the generation of new fields arising from the cross-fertilisation of different technologies (e.g. ICT and environmental technologies).

A burst analysis focusing on ICT-related fields over the period 2000-14 reveals the sequence of technological developments occurring during these 15 years, the extent to which some ICT fields saw their development accelerated and the length of the period during which such bursts were sustained (the “duration of the burst”). At the start of the 2000s, activities burgeoned in the field of digital data processing, editing and optical recording, whereas the late 2000s saw accelerations in semi-conductor devices and wireless communications. Since 2012, inventions patented in the five top IP offices (IP5) and related to digital data transfer experienced a persistent acceleration of unprecedented intensity, reaching about 24 000 IP5 patent families in 2012-14 alone. During the last part of the period considered, open-ended bursts are underway in various domains linked to organic materials devices, image analysis, connection management and payment protocols. Compared to those observed at the beginning of the period, recent bursts seem to last longer and consist of a higher number of inventions.

How to read this figure

The size of the bubble indicates the intensity of the burst (i.e. the pace at which they accelerate), and the different shades indicate different technologies that start to burst at the same time. The X axis indicates the year in which technologies start to burst, and the Y axis displays the number of years after technologies stop bursting and continued their development at a very much slower pace. For example, acceleration in the development of patented technologies related to optical recording and reproduction (top-left) was first observed in 2001 (X axis), and lasted for four years (Y axis), until the end of 2004. Bubbles located along the diagonal line on the right-hand side of the figure represent open-ended bursting technologies (i.e. technologies still developing at an accelerated pace at the end of the sample period). Among ICT technologies that began to burst in 2012 are those related to digital data transfer, organic materials devices and image analysis. While developments in these fields were characterised by a varying number of patents – with digital data transfer accounting for the highest amount – inventive activities in all fields continued to occur at an accelerated pace up to the end of 2014.

Artificial intelligence

Artificial Intelligence (AI) is a term used to describe machines performing human-like cognitive functions (e.g. learning, understanding, reasoning or interacting). It has the potential to revolutionise production as well as contribute to tackling global challenges related to health, transport and the environment. The development of AI-related technologies, as measured by inventions patented in the five top IP offices (IP5), increased by 6% per year on average between 2010 and 2015, twice the average annual growth rate observed for patents in every domain. In 2015, 18 000 IP5 patent families related to AI were filed worldwide. Japan, Korea and the United States account for over 62% of AI-related patent applications during 2010-15, down from 70% in 2000-05. Over the same period, Korea, China and Chinese Taipei increased their number of AI patents compared to rates observed in 2000-05. EU 28 countries contributed to 12% of the total stock of IP5 AI-related inventions in 2010-15, down from 19% in the previous decade. AI technological breakthroughs such as “machine learning” coupled with emerging technologies such as big data and cloud computing are strengthening the potential impact of AI.

How to measure AI technologies?

Measuring the development of AI technologies is challenging as the boundaries between AI and other technologies blur and change over time. The indicators presented here make use of technology classes (i.e. the International Patent Classification, IPC, codes) listed in the patent documents to identify AI-related inventions. All inventions belonging to the “Human interface” and “Cognition and meaning understanding” categories listed in the 2017 OECD ICT taxonomy (see Inaba and Squicciarini, 2017) are here considered as being AI-related.

As inventions protected by patents can be assigned to a number of technology classes at the same time, it is possible to investigate the extent to which AI is combined with other technologies by examining the “co-occurrence” of IPC codes in patent families (i.e. the listing of several IPC codes in the same patent document). The figures presented here show technologies that are more often combined with AI, and are displayed in accordance with the WIPO IPC-Technology concordance (2013) and the ICT taxonomy.

An examination of all technology fields in which AI-related patents are filed shows that AI technologies are frequently associated with a variety of digital technologies used for big data analytics. These include digital data processing and transfer as well as applications used for transport and health. For example, a closer look at medical technologies reveals that up to 30% of inventions used for medical diagnosis (e.g. eye testing or general medical examinations) incorporate embedded AI-related components.

Science landscape

The world’s top R&D performer is the United States, which surpassed USD 500 billion of domestic R&D expenditure in 2015. The second biggest performer of R&D is China (USD 409 billion PPP), which overtook the combined EU28 area (USD 386 billion PPP) in 2015. Israel and Korea have the highest ratio of R&D expenditures to GDP owing to rapid increases in recent years. OECD partner economies account for a growing share of the world’s R&D, measured in terms of total researchers and R&D expenditures. In most economies personnel costs, including researchers, account for the bulk of R&D expenditures. This explains the close relationship between R&D as a percentage of GDP and the number of researchers as a percentage of total employment. Variations can be related to differences in the relative prices of different R&D inputs (including researcher remuneration), the degree of R&D specialisation in each economy, and R&D capital expenditures relating to research infrastructures being developed for the future.

Top science

The global volume of scientific production, as indexed in the private bibliometric database Scopus, grew significantly over the 2005-16 period. Indicators of “scientific excellence” focus on the changing contributions of countries to the top cited publications. China increased its production of highly-cited scientific output and so its share in the world’s top 10% most-cited publications from less than 4% in 2005 to 14% in 2016, making it the second largest country behind the United States. The combined EU area maintained its global share of high quality scientific production, surpassing the United States as a scientific powerhouse. However, as the second figure shows, the average “excellence” of EU research is still lagging at about 12%, lower than both the United States and the United Kingdom, which maintain their status as countries with high shares of high-quality scientific research (14%). Starting from a low base, the Russian Federation also saw its average performance increase to over 4% over the period.

How to read these figures

Figure 11 depicts the area or country share of the world’s top 10% most-cited documents within their class (articles, reviews and conference proceedings) and publication year. For example, more than 30% of top-cited documents are produced by EU-based authors. Figure 12 illustrates the percentage of documents produced within each country that attain a top 10% cited status. For the EU area, this is close to 12%. A citation-based measure of journal influence, the Scimago Journal Rank, has been used to rank documents with identical numbers of citations. Because more recent documents attract fewer citations, values for recent years will be more influenced by this adjustment. The same applies to fields where citations take longer to occur.

R&D trends

Gross domestic expenditure on R&D (GERD) in the OECD area grew 2.3% in real terms from 2014-15 to reach USD 1.14 trillion. This increase furthered the recovery of R&D expenditure in the aftermath of the 2008-09 global and financial crisis. Since 2013, OECD GERD has remained stable as a percentage of GDP at 2.4%. Recent growth has been driven primarily by businesses, which account for around 70% of all R&D. Private non-profit institutions’ R&D (which includes most charities) also grew strongly over 2013-15, although this represents only a small share of total R&D (2.4%). Government-performed R&D rebounded slightly, while the pace of growth of R&D undertaken by higher education (the second biggest R&D performing sector) slowed. Among countries covered in the OECD Main Science and Technology Indicators (http://oe.cd/msti), R&D intensity was highest in Israel and Korea, the latter of which has experienced fast growth since 2002 – driven primarily by increasing business R&D. This is also the case in China where GERD as a share of GDP surpassed the EU28 share in 2012 and continued to grow towards the OECD level (2.4%), reaching 2.07% in 2015. The higher education sector is a significant contributor to R&D performance in most countries, particularly with respect to fundamental basic research. However, in China, higher education institutions’ R&D accounts for only 7% of GERD, markedly below the OECD and EU28 levels (18% and 23%, respectively).

As with other types of investment, expenditures on R&D and innovation are pro-cyclical (positively related to economic performance). Business-financed R&D is particularly affected by varying finance availability and aggregate demand. The major drop in GDP and business R&D in 2008-09 was partly balanced by growing government-funded R&D. Since 2010, business-funded R&D has recovered, while direct government funding of R&D has declined – mainly due to budget consolidation policies. Since 1985, the three types of R&D have evolved differently: applied research and experimental development, which account for most of R&D expenditure (21% and 62% of GERD, respectively, in 2015; reaching a combined 95% in China) have more than doubled in real terms since 1985. Basic research (17%) has nearly quadrupled over the same period, driven by sustained growth in R&D within higher education. Considerable differences across sectors and countries underlie the general trends presented. For example, relative increases in business-performed basic research are also a factor in some countries including the United States, which has seen this rise from 3% to 5% of GERD between 2005 and 2015.

Measuring R&D and its components

R&D activity is measured by summing all relevant expenditures incurred in performing R&D as defined in the Frascati Manual (OECD, 2015a). R&D comprises basic research (creating new knowledge with no specific application in view), applied research (creating new knowledge with a specific practical aim), and experimental development (of new products or processes). Separating these components is challenging in some countries and sectors, leading to coverage gaps. Financial incentives, especially government funding decisions and priorities, may also affect survey respondents’ reporting of R&D projects as basic or applied research, impacting measures of sector and/or industry specialisation in different types of R&D.

Concentration of business R&D

R&D is a highly concentrated activity: within countries a small number of firms are responsible for a large proportion of total business R&D (BERD). This is corroborated by a new analysis of R&D performance across a number of OECD countries at the enterprise level. The 50 largest domestic R&D performers account for 40% of BERD in Canada and the United States, 55% in Germany and Japan, and 70% in Denmark and New Zealand. Broadening the analysis to the top 100 R&D performers leads to a relatively moderate increase in the cumulative share of BERD accounted for by large R&D performers. These figures should be considered in relation to the size of the country and the total number of business R&D performers. In New Zealand, for example, the top 50 performers represent 4% of all R&D performing enterprises, whereas in France or Germany they represent a much smaller fraction. Understanding the concentration of business R&D has immediate implications for the allocation and potential targeting of public support for business R&D, which is prone to be skewed towards large R&D performers.

microBeRD: the OECD microdata project on impact and incidence of public support for business R&D

The OECD has launched the microBeRD project to analyse the extent and impact of public support for business R&D at the micro level. microBeRD seeks to facilitate policy learning by exploring the wide heterogeneity in companies’ eligibility and use of government support – both within and across countries. The project adopts a coordinated, distributed approach to the analysis of microdata across different jurisdictions, undertaken in collaboration with national experts with access to R&D and public support microdata. The majority of these experts already collaborate with the OECD on longstanding activities to set measurement standards for R&D and develop internationally comparable aggregate R&D indicators.

The use of a common, adaptable code facilitates consistent, multi-country analysis of heterogeneity in the uptake and impact of public support for business R&D across firms. This approach also preserves data confidentiality (only aggregated, non-disclosive data are shared with OECD), while addressing questions that cannot be explored through analysis within a single country or with publicly available data sources alone.

A series of indicators derived from R&D microdata can inform the policy analysis of markets and policy drivers of R&D performance and their impacts. Indicators on the concentration of R&D performance within OECD countries can help understand the role of competition, for example, through comparison with other measures of economic concentration at industry or country level. Furthermore, a comparison of the actual concentration of R&D performance with microdata-based measures of the concentration of public support for R&D can help identify the existence of potential biases and consistency with the stated rationales for allocating support. While it is broadly acknowledged that R&D is a highly concentrated activity, there is only limited internationally comparable evidence available on the degree of R&D concentration within OECD countries. microBeRD seeks to help close this evidence gap.

For more information on the microBeRD project, see http://oe.cd/microberd.

While large firms account for the bulk of business R&D in most of the countries considered, small and medium-sized firms still account for a significant share of BERD, ranging from 21% in Belgium to 56% in Norway. Within each size category, most R&D is performed by firms established five or more years ago. With the exception of the Czech Republic and Italy, most of the R&D performed by younger firms (established less than five years ago) is attributable to small companies with 10-49 employees, vis-à-vis medium-sized (50-249 employees) and large (250 and more employees) enterprises. The countries with the largest share of R&D performed by younger firms are Israel (9.3%), Norway (8.6%) and the Czech Republic (7.6%).

Across countries, there are significant differences in the extent to which firms of different size and age rely on external sources of R&D funding. In Belgium and Norway, external sources of funding account on average for at least 15% of R&D expenditure in every size and age category, while in the Czech Republic and Israel, external sources make up less than 7%. Overall, small R&D performers tend to rely more heavily on external R&D funding. Government funding is particularly important for small R&D performers, while its relative importance for young versus old small companies varies across countries. Funds from abroad play a more important role for medium-sized and large R&D performers.

Top corporate R&D players

Top corporate R&D investors are companies at the technology frontier that account for a substantial amount of innovation-related investment and output. Their headquarters are concentrated in a few economies, in particular the United States, Japan and China. On average, each top 2 000 R&D investor has affiliates in 21 economies and is active in 9 different industries. R&D expenditure as well as innovative output in the form of patents and trademarks also appears to be highly concentrated. In 2014, the top 10% of these corporate R&D investors (i.e. the top 200 companies with their affiliates) accounted for about 70% of R&D expenditure, 60% of IP5 patent families (inventions patented in the five top IP offices), 53% of designs and 38% of trademarks. Industry-specific dynamics, product complexity and market differentiation strategies, among others, help to explain differences among companies in the use of intellectual property types. Top R&D investors play a leading role in the development of digital technologies. They account for the ownership of about 75% and 55% of global ICT-related patents and designs, respectively, while about 21% of their affiliates operate in ICT industries, on average. Patents protecting ICT-related developments represent 44% of the total patent portfolio of top R&D investors in the ICT sector. However, the share of ICT-related patents owned by non-ICT corporations varies substantially, reaching 70% or more in the case of companies operating in the “Finance and insurance” and the “Administrative and support services” industries.

Top corporate R&D investors in the “Computers and electronics” industry are, by far, the most reliant on intellectual property (IP) rights and account for about one-third of total patent filings by top R&D investors. “Transport equipment”, “Machinery and Chemicals” are also emerging as patent-intensive industries. Companies differ in the extent to which they rely on various IP assets. Among ICT corporations, top R&D investors such as Samsung or Sony rely on patents and designs to almost the same extent, while others such as Fujitsu and Toshiba rely more on technological developments than design, and yet others, e.g. Microsoft and Apple place a greater emphasis on design than patents.

How to read the word clouds

The word clouds are assembled using information about the distribution of patent and design portfolios of top corporate R&D investors. The font size of the company names reflects the relative size of the patent or design portfolios of the company vis-à-vis those of other companies in the sample. The names of top corporate R&D investors active in the ICT sector appear in dark blue bold characters, whereas those from other sectors are shown in light blue. The position and orientation (i.e. vertical vs horizontal) of names in the word clouds has no meaning, aside from ensuring that the names are clearly visible.

Who are the world’s top corporate R&D investors?

Top R&D investors worldwide are companies that are either parents of (a number of) subsidiaries or independent entities. In the former case, the R&D spending figure used for the ranking is that which appears in consolidated accounts and includes spending made by subsidiaries. Among top R&D investors in 2014, 82% of the companies also appear in the 2012 list (see Dernis et al., 2015). Notable differences between the lists include a smaller number of ‘Computer and electronics’ companies and a higher number of ‘Pharmaceuticals’ corporations in 2014, as compared to 2012. Asia-based companies emerge as the biggest patent assignees among the sample. Out of the top 50 IP5 assignees, 30 are headquartered in Asia of which 19 are located in Japan and 6 in Korea. Top R&D investors headquartered in the European Union, the United States and Japan specialise in a relatively broad number of technologies. EU and US companies often focus on technologies that play a fundamental role in addressing key societal challenges, such as health or the environment. Companies headquartered in China and Korea specialise almost exclusively in ICT-related technologies. More than half of top R&D investors employ the full IP bundle (patents, trademarks and designs). However, IP strategies vary depending on the target market and the industry in which the companies operate. More information about these companies and their patenting, design and trademarking activities can be found in Daiko et al. (2017).

Technology at the global frontier

Top corporate R&D investors worldwide lead the development of many emerging technologies. This is evident from an examination of the technology fields in which these companies intensified their inventive activities during 2012-14 and the contribution of top R&D investors to the overall development of these fields. Top corporate R&D investors accelerated their inventive activities in areas such as engines, automated driving systems, and information and communication technologies (ICT) related to connectivity, transmission and digital data transfer. In many of these fields, top R&D corporate investors own 80% or more of the worldwide portfolio of patents related to these technologies.

In which technologies are top R&D companies leading?

R&D activities undertaken by the world’s top corporate R&D investors result in the development of new technologies. The DETECTS methodology (see Dernis et al., 2016, for details) was applied to the portfolio of top 2000 R&D players to highlight technology fields experiencing an accelerated (“bursting”) development, compared to other technologies. Patent bursts are sudden and persistent increases in the number of patents in a given field, as compared to those observed in other fields, and are characterised here at the level of International Patent Classification (IPC) groups. The top emerging technologies are defined according to the IPC codes that follow open-ended bursting behaviour, i.e. a rapid acceleration in patenting, from the early 2010s onwards. Artificial intelligence refers to the “Human interface” and “Cognition and meaning understanding” categories in the ICT patent taxonomy as described in Inaba and Squicciarini (2017).

Top players in artificial intelligence

Top 2000 corporate R&D investors own 75% of the IP5 patent families related to artificial intelligence (AI). These investors are not necessarily global companies in the ICT sector; firms in each and every industry contribute to advancing AI, albeit to different extents. In addition to “Computers and electronics”, which accounts for 64% of the AI portfolio of top R&D players, corporations operating in “Transport equipment” and “Machinery” are responsible for high levels of inventive activities in the AI domain over the period considered (almost 1000 patents a year, on average). The development of AI technologies is fairly concentrated. R&D corporations based in Japan, Korea, Chinese Taipei and China account for about 70% of all AI-related inventions belonging to the world’s 2000 top corporate R&D investors and their affiliates, and US-based companies for 18%.

Research in machine learning

Research in the field of artificial intelligence (AI) has aimed for decades to allow machines to perform human-like cognitive functions. Breakthroughs in computational power and systems design have raised the profile of AI, with its outputs increasingly resembling those of humans. Such advances enabled IBM’s Deep Blue computer to beat world chess champion Garry Kasparov in 1997 and allowed computers to distinguish between objects and text in images and videos. A key driver has been the development of machine learning (ML) techniques. ML deals with the development of computer algorithms that learn autonomously based on available data and information. Drawing on the power of “big data” sources, algorithms can deal with more complex problems that were assailable only to human beings. Experimental bibliometric analysis shows remarkable growth in scientific publications related to ML, especially during 2014-15. The United States leads in this area of research both in terms of total publications and highly cited ones. Also worthy of note is the fast growth experienced by India, now the third largest producer of scientific documents on ML after China and fourth behind the United Kingdom on a quality-adjusted basis.

Which scientific documents have been identified as related to machine-learning?

These experimental estimates are based on a search for the text item “*machine learn*” in the abstracts, titles and keywords of documents published up to 2016 and indexed in the Scopus database. The accuracy of this approach depends on the comprehensiveness of abstract indexing, which implies a bias towards English-speaking journals. In a survey by Kalantari et al. (2017) of 142 big data experts asking for relevant keywords on big data research, “machine learning” was identified as relevant in 29% of cases. From the identified list of keywords, machine learning was retrieved in 40% of the 11 000 documents identified in the Web of Science database, covering the period 1980–2015, making this the most frequent category above “Data centre”, “Big data” and “Data warehouse”.