Chapter 3. Perspectives on innovation policies in the digital age
Most innovations today are new products and processes made possible by digital technologies or embodied in data and software. This transformation took place first in digital sectors (e.g. software) but has now spread to all sectors, including services (e.g. retail and education) and manufacturing (e.g. automotive). It results in new dynamics, with data as core inputs to research and innovation, more service innovation, the blurring of boundaries between services and manufacturing (servitisation), and greater speed and collaboration in innovation. Innovation policies need to adapt, so as to address data access issues, to become more agile, to promote open science, data sharing and co-operation among innovators, and to review competition and intellectual property policy frameworks. This chapter first assesses the economic mechanisms that characterise digitalisation and reviews the impacts of the digital transformation on innovation in the digital age. It then discusses how these changes affect business dynamics. Based on these insights, it draws lessons for innovation policies and concludes by providing perspectives on the future.
Most innovations today are new products and processes, enabled by digital technologies or embodied in data and software. These digital innovations are an outcome and a component of digital technologies, which allow collecting, processing, manipulating, storing and diffusing data automatically, using machines. Progress in electronics (Moore's law) and data science have introduced a new way of using technologies. Advances in artificial intelligence (AI), a set of technologies that emulate certain aspects of human intelligence, promise further progress in the manipulation of digitalised information and knowledge.
These changes are driven by advances in science and innovation, and are themselves drivers of science and innovation. Many dimensions of the digital world differ from the physical, tangible world, and innovation processes and outcomes are being transformed as a consequence. Although this transformation first occurred in the digital sectors, it is now widespread and involves many tangible sectors, such as the agro-food and automotive industries. For example, the Internet of Things (IoT) connects the physical and digital worlds, allowing every object and location in the physical world to become part of the digital world.
With those sweeping transformations under way, it is pertinent to evaluate whether – and in what directions – policy support for innovation should adapt. This chapter assesses the economic mechanisms that characterise digitalisation and reviews the impacts of the digital transformation on innovation in the digital age. The chapter discusses how these changes affect business dynamics. Based on these insights, it draws lessons for innovation policies and provides perspectives on the future.1
Digital technologies have lowered information-related production costs and changed the characteristics of innovation (Figure 3.1).
The processes and products embodying or implementing digital technologies are characterised by their “fluidity”. Fluidity means that data can circulate, and be reproduced, shared or manipulated instantaneously, at any scale and at no cost. Once available, digitised knowledge (i.e. knowledge that takes the form of data) can be shared instantaneously between any number of actors, notwithstanding geographic distance and other (natural or institutional) barriers, with each actor having full access to it (OECD, forthcoming). This characteristic affects all economic processes, like the commercialisation of new products and the diffusion of knowledge. Fluidity allows scaling up to serve entire markets much more rapidly, i.e. achieving “scale without mass”, facilitating both competition by new entrants and “winner-takes-all (or most)” market dynamics. This ease of scaling digital goods contrasts with tangible goods, which are subject to physical production and distribution constraints (e.g. manufacturing and transportation costs).
Digital technologies have drastically reduced several types of costs, notably: 1) the marginal costs of producing intangible-intensive goods and services; 2) the costs of searching, verifying, manipulating and communicating information and knowledge; and 3) the costs of launching new goods and services – specifically those with high information and knowledge content – on the market (Haskel and Westlake, 2017). The costs of verifying the reputation and trustworthiness of potential partners through digital technology such as blockchain are lower. This increases the chances a successful search will result in actual matches (between supply and demand of labour, inputs, products, etc.) on the market, thereby reducing production costs and improving product quality (Goldfarb and Tucker, 2017).
Digital technologies are also increasingly embedded in many tangible products. They transform them into smart, connected products (e.g. connected cars and agricultural machinery equipped with sensors) that are able to produce and exchange data about their own status and performance, or the environmental conditions around them (the Internet of Things, IoT). Based on the data they generate, these products are key enablers of a wide range of services and process innovations. For instance, IoT applications can be used to track in real time the trajectory and storage conditions of food throughout the supply chain.
New possibilities for handling data have made them core inputs of innovation in all sectors of the economy (OECD, 2015a). Data feed into innovations in multiple ways; for example, data on consumer behaviour can be used to customise services or to develop entirely new services (such as on-demand mobility services like Uber, which rely on instantaneous information about demand and supply to organise transportation). Data generated in production processes (e.g. managerial and technical data), public-sector data (e.g. transportation and patient files) and research data (e.g. experimental data) are less visible, but equally important. All these types of data are relevant – albeit to different degrees – to innovation.
In this context, the deployment of AI and machine learning further increases the expected value of data. Machine learning requires large numbers of observations before the software is able to perform the expected task, although much research is currently taking place in AI to reduce the amount of data needed to train a program. The development of the IoT also means that data generation is increasing steadily, as more devices and activities are connected.
Because of data’s growing importance, many businesses make large investments to access data, whether by setting up data-gathering systems, acquiring data-rich companies (Microsoft notably acquired LinkedIn to take control of its data) or contracting with partners. At the same time, many businesses still need to develop best applications and data analytics infrastructures to bring value from data analytics to their business.
The digital transformation also creates opportunities for innovation in services as digital technologies reduce costs, while allowing greater fluidity in reaching and interacting with consumers, and tracking their behaviour. In particular, innovation opportunities arise for: 1) new services, such as predictive-maintenance services using IoT data, on-demand transportation services and web-based business services; 2) renting as a service or sharing instead of selling equipment; and 3) customising products (i.e. adapting products to each customer's specific needs, thanks to software and data capabilities).
Servitisation is disruptive to business practices, as it removes the boundaries between manufacturing and services, and requires entirely new business models. Many manufacturing firms’ strategy and innovation activity now follows the “3 S” model: sensors, software and service. For instance, Bosch has installed software-monitored sensors on many of its car parts, which allow the company to offer its customers better maintenance services. Conversely, service firms like Amazon and Google are also entering the manufacturing industry, producing home appliances, mobile phones, computer chips, etc.
The lower cost of launching new products and processes using the Internet and online platforms facilitates versioning and experimenting products for differentiated customers. Lower costs can also produce more frequent innovation: software can be updated daily (or even more frequently), with new versions downloaded from the Internet. The changes are widespread, extending far beyond the purely digital sectors. In the automotive industry, although the hardware (the car itself) might last for years with little change, the software is frequently updated. In the music industry, the reduced cost of disseminating music through the Internet has generated an increase in creation, to satisfy consumers’ highly differentiated and fast-changing tastes.
In addition to the reduced costs of launching and diffusing products, another driver of the digital transformation is the cumulative nature of upgrades, reducing product cannibalisation (i.e. the creative destruction of its own product by a company): when a firm issues an innovation, it may simply add to products that are already on the market and it can be downloaded as an “add-on”. Contrary to a new car model, for example, the new digital product will not replace the firm’s existing products; rather, it will enhance them.
The acceleration in versioning and innovation is not synonymous with more rapid technological progress and productivity. Many of these improvements are small. Technical change may have become more staged and continuous, but is not necessarily more rapid. Nonetheless, access to these incremental innovations benefits end consumers as they have access to advanced versions. If consumer feedback on versions is integrated effectively in innovation processes, then versioning may also boost innovation.
Where “superstar” effects are in place, a small advantage over competitors might allow a firm to seize all of the market – hence increasing the expected reward in case of a successful (even minor) advance. This also increases the risk for firms, as a setback or a lag – however small – could mean losing all of the market. This creates competitive pressure and, consequently, firms have an interest in updating and launching new versions to gain or maintain lead positions, even at the margin.
Thanks to the reduced costs of (and greater need for) collaboration, innovation has become more collaborative. The reduced costs come from the growing role of data in collaboration, whereas the greater need comes from the evolution of demand (e.g. addressing grand challenges, or designing mobile phones to integrate knowledge from various fields). This enhanced collaboration can take different forms and follow different paths: data sharing, open innovation, innovation ecosystems, platforms (hubs), mergers and acquisitions (often driven by the need to combine various types of competences), and global value chains (which integrate technology in successive stages, along an ordered line).
Successfully harnessing the potential of digital technologies requires combining different technologies used for varying purposes into coherent systems. Actors may also engage in collaborative innovation processes to hedge against the risks from disruptive innovations by competitors; these risks will be higher in the context of general-purpose technologies (GPTs).
New forms of open innovation allow collaborating much more actively than previously with large communities of experts and consumers. External sourcing practices (procurement) involving tournaments, collaborations, open calls and crowdsourcing are new ways for firms to address innovation challenges; some of these practices could become permanent, while others could be one-off only. Examples of corporate initiatives include the BMW Customer Innovation Lab, IBM InnovationJam, Dell IdeaStorm, Procter & Gamble’s Connect+Develop and GE Fuse (Board of Innovation, n.d.[5]). These practices are also conducted through intermediary online platforms, such as Innocentive, IdeaConnection, Innoget, Hypios and NineSigma.
Finally, digital technologies can be characterized as relatively young, far-ranging and fast-evolving GPTs, affecting all sectors of the economy. Hence, their current and future development generates much uncertainty. This is particularly true of AI, a set of technologies that can emulate functions normally accomplished by human intelligence, based on pattern recognition and prediction. Not only is AI expected to transform economic activity, it also raises complex societal and ethical issues. However, this transformation could take some time, as the number of possible applications is far greater than the number of current applications (Brynjolfsson, Rock and Syverson, 2017). Although recent research points to decreasing productivity of innovative activities over the past few decades (Bloom et al., 2017), some scholars expect AI to reverse this trend (Cockburn, Henderson and Stern, 2018).
The transformations in innovation processes and outcomes affect business dynamics and market structure, with consequences on the distribution of performance and rewards among businesses, individuals and regions.
On the one hand, as large volumes of data are fluid and potentially available to everyone at a low marginal cost (notwithstanding obstacles to data access, which can be substantial, but are due to market actors, not to physical costs), the costs of market entry and expansion for new firms requiring such data are lowered. Hence, digitalisation potentially creates a more level playing field in terms of access to data inputs (providing that no regulatory or strategic barriers are in place). Increasingly digitalised information and knowledge become accessible to all, creating more equality of opportunities. This applies not only to many scientific or public-sector databases, but also to certain valuable private-sector data (e.g. scientific publications subject to copyright). For example, the US National Institutes of Health database2 allows researchers to access information on privately and publicly funded clinical studies from around the world, including study protocols, purposes and results. The database of Genotypes and Phenotypes also provides access to data and results from studies that have investigated the interaction of genotype and phenotype in humans (Sheehan, 2018). Such potentially widespread and free access contrasts with physical goods that do not allow for such widespread access and use.
This increased access to data has spurred dynamic entrepreneurial activity based on digital innovation in several markets. These include the transportation sector (with the emergence of platform-based car-sharing and ride-hailing applications) and retail (with the emergence of start-ups specialised in data analytics, to optimise inventories and personalise sales). Many highly successful start-ups have been created by students using digital technologies and data to illustrate these new dynamics of the intangible economy. Famous examples include Facebook (Mark Zuckerberg), Snapchat (Evan Spiegel), Dropbox (Arash Ferdowski and Drew Houston) and Invite Media (Nat Turner).
Entrepreneurial activity linked to disruptive business models has also helped improve consumer welfare in ways that are not always easy to assess. For example, digital maps, encyclopaedias and social media have massively improved consumer welfare. However, the disruptive business models behind those services mean that routinely used metrics – such as gross domestic product (GDP) – are no longer adequate to capture the improvements, requiring novel approaches to track them (Box 3.1). Work conducted in the context of the OECD-wide Going Digital project documents often unmeasured contributions of the digital economy to well-being.
Erik Brynjolfsson (Massachusetts Institute of Technology [MIT] and National Bureau of Economic Research) and Avinash Collis (MIT)
One of the fundamental objectives in economics is to assess people’s well-being. Economists, policy makers and journalists routinely use changes in GDP and metrics derived from it – such as productivity – as proxies for changes in well-being. However, GDP was never meant to be a measure of welfare. It is a measure of production. In some cases, GDP and welfare are correlated, but in many other situations, this is not the case. In fact, in some cases, the change in GDP can even have the opposite sign from the change in welfare.
Treating GDP as a proxy for welfare is particularly problematic for digital goods, such as online encyclopaedias, search engines, social media and digital maps. Most of these are available at zero price to consumers and are therefore largely excluded from GDP. As the production and consumption of such goods grows, GDP does not change, but welfare does increase. A growing number of goods are transitioning from traditional physical goods to free digital goods. While these types of goods were counted in GDP measures, they are excluded from GDP once they transition to free digital goods. The encyclopaedia industry offers an excellent illustration of such a transition. Previously, people bought and paid for physical copies of encyclopaedias, such as Encyclopaedia Britannica, and these transactions contributed to GDP. Over the past 15 years, however, Wikipedia has replaced Encyclopaedia Britannica as the premier reference source. Because it has zero price, Wikipedia is excluded from GDP measures. As a result, the contribution of encyclopaedias to GDP has decreased, because people have shifted from paying for Encyclopaedia Britannica to consuming Wikipedia for free. Nonetheless, consumers are clearly better off.
In theory, consumer surplus is a better measure of consumer welfare than GDP. In practice, it is challenging to measure consumer surplus in a scalable manner, since this requires estimating demand curves. In Brynjolfsson, Eggers and Gannamaneni (2016), we propose a new way of directly measuring consumer welfare, using massive online choice experiments while staying within the neoclassical framework. Our approach takes advantage of the fact that in recent years, it has become much easier to collect data online on a large scale. These advances have been essential to creating alternative measures of the economy, including ours. Our approach can be scaled easily to hundreds of thousands of goods, by running several thousand choice experiments every day. This approach can be implemented more frequently than the consumer price index and can be used to track changes in well-being over time. Moreover, goods – including non-market goods – can be easily added or removed from the basket.
The system of national accounts centred on GDP was one of the greatest inventions of the 20th century. In the 21st century, the proliferation of digital data, combined with an infrastructure that allows surveying millions of people easily, cheaply and quickly, provides an opportunity to develop new measures of welfare. These can be used to supplement and extend existing national accounts.
Digital platforms also enable entrepreneurship by lowering set-up costs for newcomers. For example, e-commerce platforms (e.g. Alibaba, Amazon and eBay) allow new ventures to offer products to the market without paying extra for marketing. Such platforms also gather very accurate information on the activities of the companies that use them (e.g. who their customers are, how their sales are evolving, and what they spend on marketing); this puts them in a favourable position to provide funding to these companies, as the information asymmetry (a usual barrier to funding small and medium-sized enterprises [SMEs]) is minimal. For example, Amazon proposes a range of financial products to businesses trading on its platform (OECD, forthcoming).
But the fluidity of data may contribute to industry concentration thanks to three factors. One factor is the natural advantage of platforms (defined as Internet-based structures that organise the interaction between different actors) in increasing market efficiencies. Important efficiency gains can be derived from combining data to exploit optimally the information and knowledge they contain, providing natural advantages to large aggregators of data. Similarly, providing combined services on a single platform, and bringing together a larger group of users, offers considerable consumer benefits. In other words, several small platforms that provide fewer services, have fewer users each and build on less data would be much less efficient than a single, large and more diversified platform. Such economies of scale are characteristic of a natural monopoly.
The second factor promoting concentration is “scale without mass”, a consequence of the increasingly intangible composition of products. The larger the intangible component, the easier it is to expand production to the entire market, at little or no supplementary cost. In the case of software, the cost of producing an additional unit is close to zero, as no further set-up costs are involved. The much smaller number of employees relative to the sales of certain digital companies compared to companies operating in traditional industries illustrates this dynamic. At the same time “scale without mass” allows successful competitors to grow quickly, as fewer overhead costs are incurred even as production is expanded to the full market.
A third factor is the scarcity of certain factors – notably skills – that are complementary to data and are required to exploit data efficiently (OECD (2017a) and Nedelkoska and Quintini (2018)). Such scarcity also tends to favour concentration: up to a certain group size, skilled workers are more efficient when employed jointly (in certain firms or regions) thanks to intra-team knowledge exchanges.
The balance between the factors favouring and hampering concentration varies over time and sectors, and is influenced by policies. Polarised market structures, characterised simultaneously by the dynamics of concentration and massive new entry, are also possible. Such market structures have a few giants, with a long tail of smaller and fast-changing niche producers, and a shrinking space for medium-sized businesses. Using data from a retailer with both online and offline channels, Brynjolfsson, Hu and Smith (2010) show that the variety of products available and purchased online is higher than for those offline, reflecting more opportunities for niche products in the online economy.
Similar distribution and dynamics apply to other economic variables, i.e. the incomes of individuals (with diverse skills, positions and employers) and the wealth of places (with large cities increasing their advantage over rural regions). Skewness is reinforced by the fact that markets are now globally integrated; in the past, national borders shielded places, people and firms from foreign competition, limiting global concentration.
Creating value out of data requires complementary assets – namely, individual skills, collective and organisational competencies (i.e. the right institutional setting to exploit information), and data-assessment tools. In the digital age, data are the main input to many production processes; these data are fluid, contrary to the physical inputs that prevailed previously and limited mobility. The best performers can access and use many of the data available (whereas they could hardly access and use all of the physical resources available), leveraging their advantage more than in the past, where the lowest performers could still secure easier access to certain resources. Any entrepreneur can potentially access a wide range of data and leverage their efficiency advantage, however small (as the whole market becomes integrated). This is true at the individual level, allowing top entrepreneurs to command larger production teams and take decisions with key data (Garicano and Rossi-Hansberg, 2006); at the organisational level, allowing firms with the strongest capacities to leverage data better; and at the geographical level, as the top cities or regions worldwide can access and exploit a wide range of available data to build their prosperity (Kerr and Kominers, 2015). The growing prosperity of cities also reflects the complementarity of non-codified social knowledge with codified, digital knowledge. Gaspar and Glaeser (1998) suggest that the reduced communications costs may most benefit those that already communicate much, meaning that falling costs would benefit cities most, further driving concentration.
The new context and features of innovation require changes to the targets, mechanisms, instruments of innovation policies and to the policy mix of innovation. This is because, as discussed in the previous sections, digitalisation is affecting essentially all mechanisms that drive innovation, exactly those mechanisms that innovation policies are targeting. Therefore, all innovation policy instruments are affected (Figure 3.2). Some instruments will adapt their target or content to digital innovation while essentially preserving their processes; that includes for instance policies supporting entrepreneurship, SMEs or generic technologies. Other domains will go through in-depth transformations, including sometimes of their rationale: this includes science policy, with its move towards open science and or policies supporting university-industry linkages, with a move towards co-creation.
Source: Guellec and Paunov (2018).
This section discusses eight principles for the design of innovation policies in the digital age (Figure 3.3).
First, as data now constitute new input to innovation, access to data – and to the tools that gather and help interpret them – will influence who participates in digital innovation, and in what ways. Innovation policy must therefore address data access. The goal should be to ensure the broadest access to those data and knowledge that facilitate competition (e.g. through alternative uses of the same data), re-use (i.e. producing a gain in efficiency) and transparency (e.g. creating the ability to check the validity of results obtained on a given dataset). However, data-access policy has to take into account the diversity of data, as access issues differ across data categories, as well as economic and non-economic constraints. This includes incentives to produce the data in the first place, competition, intellectual property, privacy and ethics.
Certain data (e.g. customer data or product-design data) are trade secrets. They cannot be shared without endangering the firm’s competitive position, or even its very existence. Opening access to such data might allow firms with the most effective data-processing capabilities to take control of the relevant markets, turning established companies into suppliers and possibly reducing competition, as data-based markets are more prone to “winner-takes-all (or most)” dynamics than other markets.
Government should also create the appropriate conditions to promote the emergence of data markets. The development of knowledge markets, which previously focused on intellectual property (IP) rights and now encompass all data, has been viewed positively by economists (Yanagisawa and Guellec, 2009). Not only does trading data facilitate the exchange of data for innovation purposes, but it also allows putting a price tag on data generation and curation for future use, facilitating the generation of more data.
Second, accelerated innovation cycles owing to digital innovation should be matched by adequate policy experiments to support innovation, which means rethinking the types of instruments used and their implementation.
Approaches to ensure rapid and agile policy responsiveness include policy experiments that operate in “start-up mode” where experiments can be deployed, evaluated and modified, and then scaled up or down, or abandoned quickly. Using digital tools to design innovation policy and monitor policy targets is another option to spur faster and more effective decision-making. For instance, some governments use “agent-based modelling” (a form of AI) to anticipate the impact of policy variants on different types of businesses. Another approach is to shift emphasis from instruments that target specific groups of recipients or technologies to ones that are more flexible. Such instruments include tax reliefs, certain regulations, and intellectual property (IP) rights, as well as simplified innovation support schemes (e.g. ‘sector-agnostic’ and “single-window” grant application processes). Mission-oriented programmes that set a goal, but do not impose the means to reach it, could help. Of course, the specific drawbacks of such instruments (e.g. the lack of selectivity, resulting in a deadweight loss) compared to targeted instruments should be considered and weighted against the advantage of greater flexibility. Another option is to provide the necessary autonomy and agility to choose the proper technological avenues to achieve a stated policy objective. In the United States, the Defense Advanced Research Projects Agency has successfully boosted fundamental defence research, thanks to its organisational flexibility at the administrative level and the significant authority granted to programme directors (Azoulay et al., 2018). Similar programmes have been adopted by other countries, including Canada, to spur game-changing technological breakthroughs.
Third, traditional support tools for research and innovation should be revisited to ensure their effectiveness. Service innovation, which receives little support from traditional instruments, is progressing and sector boundaries are increasingly blurred; technological change can take unexpected directions, owing to the novel application of digitalisation to traditional technological fields, which can generate surprising and sudden changes in the technological trajectory. To provide an example of services tools, the Netherlands has implemented an experimental scheme called ‘service design vouchers for manufacturing SMEs’ to support manufacturing SMEs in developing services that are related to their products.
The functioning of the intellectual property (IP) system is also changing and requires policy attention. To take but one example, AI can create patentable inventions. This raises the question of who should own them: the original AI programmer, the user of the AI software that generated the invention, or the owners of the data to which AI is applied? In addition, patent grants require that the invention be “non-obvious to a person skilled in the art”. If an AI system is considered to be such a “person”, this might put the bar much higher for patentability in certain domains where AI is now a major research tool (e.g. pharmacy or combinatorial chemistry). However, trademarks may gain new importance as anchors for online search (Bechtold and Tucker, 2014).
Fourth, policy should support the development of core generic (or multi-purpose) digital technologies to facilitate downstream innovation and address societal challenges. Businesses are currently investing heavily in these technologies. Initial technological developments were primarily sponsored by governments. This is true not only of the Internet, but also of AI – which was developed almost exclusively through academic research for more than five decades, before businesses got involved in the late 2000s. Hence, governments need to keep investing in core technologies to prepare future waves of innovations. They also need to ensure these multi-purpose digital technologies are developed to serve not only commercial purposes, but also social and environment purposes. Public research is often best placed to do just that. Such investments benefit from collaboration in technology development and around AI’s economic, ethical, policy and legal implications. Institutions such as the Digital Catapult Centre in the UK were created to promote the early adoption of advanced digital technologies by innovative firms, for instance, by facilitating access to advanced technology testbeds to experiment and prototype new IoT products and services; and providing the computational power and expertise needed to develop AI solutions.
Aside from development, technology diffusion and adoption also deserve specific policy attention, with differences across firms and sectors requiring the application of suitable diffusion support services. An example of a policy initiative is Germany’s SME 4.0 Competence Centres that support SMEs to be aware of, test and adopt new digital technology solutions for their business, each centre focusing on specific technologies or application areas. Another example is the CAP’TRONIC programme in France which aims to help SMEs enhance their competitiveness by integrating digital solutions and embedded software in their products. SMEs participating in the programme can access technical seminars, trainings and workshops, counselling services and expert support to develop their digital innovation projects.
Along the same lines, governments should apply digital technologies to their own activities, including public research (e.g. data gathering, analysis, sharing, simulation etc.). This includes the following:
Increasing data access: data are a core driver of open science, which is widely seen as a way of increasing the quality and reducing the cost of research. Open access allows reusing data, reproducing results, testing a diversity of hypotheses on the same empirical basis, facilitating cross-disciplinary collaboration, etc. (OECD, 2015a; and Dai, Shin and Smith, 2018).
Offering specific training and capacity-building activities: scientists need to master digital tools (e.g. data curation, simulation and deep learning), so that they can either implement them or collaborate with team members who are using them. For example, enhancing researchers’ digital skills is one of the key objectives of Norway’s digitalisation strategy for the higher education sector, 2017-20 (Government of Norway, 2018).
Developing research tools and infrastructures: new instruments (e.g. data-sharing platforms and super-computing facilities for AI) may be critical to research and require new investments. Japan’s High Performance Computing Infrastructure programme, for example, requires an annual investment of more than USD 120 million (US dollars) to build a high-performance computing infrastructure that universities and public research centres can use to conduct R&D in various fields.
Engaging in partnerships: research organisations should partner with industry to leverage industry progress in advanced digital technologies, with a view to applying it to public research.
Fifth, growing interactivity and collaboration in innovation justify policies supporting co-operation and open innovation between industry and academia, but also among businesses. The reduced cost of collaboration stemming from digitalisation has not reduced the barriers to collaboration (such as differing regulatory regimes and diverging incentives), but it has made the social cost of not collaborating higher, as more opportunities are lost. Such policies need to consider new forms of collaboration towards innovation. Online platforms, in particular, support small-scale entrepreneurship, by offering opportunities to identify adequate niche markets. Many governments have created platforms where public research and universities can advertise their inventions, knowledge and capacities, and businesses can post their own needs. The two sides can then interact and agree on deals. Other ways to support collaboration include new types of cluster policies, such as Canada’s Innovation Superclusters Initiative.
Sixth, support for competition and entrepreneurship is needed to find the right balance in the digital age between static efficiency – where scale benefits are important – and dynamic efficiency – which drives innovation. This is a complex area, where the fundamentals of competition policy are called into question by digital innovation in the presence of network effects, standards, etc. (OECD, 2016, 2017b, and 2018a). For instance, it is difficult to determine exactly what constitutes a “dominant position”, as market positions are permanently threatened by new entrants. Arguably, digital innovation requires firms to be large, in order to achieve economies of scale; hence, weakening dominant firms (e.g. through aggressive anti-trust action) could weaken innovation. Data concentration may also shape competition dynamics (OECD (2016). On the other hand, several small firms and regulators have complained that large companies engage in certain behaviours (e.g. product tie-ins or preventive takeovers) that may hamper competition and innovation, as they prevent small players from accessing the market. Policies that recognise economies of scale, while ensuring equal access to markets and resources, would help support the long tail of firms (particularly SMEs) and regions (including rural areas with limited innovation capacities) (see the report by Planes-Satorra and Paunov (2017) on inclusive innovation policies).
Seventh, preparing individuals for the digital transformation is essential to increase the pool of skilled workers and empower their participation. It is important that innovation authorities collaborate with those in charge of education and labour market policies to ensure the right skills needed for digital innovation are being developed. Innovation authorities have an important role to play in informing other government authorities of new skills demands as businesses engage in digital innovation and that arise with rapid and broad technological change. There are often new mixes of skills for innovation, e.g. innovation in the automotive industry increasingly requires strong capabilities in software engineering and AI, in addition to traditional core competences in mechanical and electronic engineering. Fostering interdisciplinarity (particularly of computer sciences with specific traditional disciplines) is increasingly important, requiring interdisciplinary degrees with an important digital component (see, for example, MIT undergraduate degrees on computer science and biology, and on computer science, economics and data science) (MIT, 2018).
Eighth, data fluidity creates the need to set national policies targeting global markets. Digitalisation facilitates the circulation of knowledge, including across national borders, reducing governments’ ability to restrict the benefits of policies to their own country. While data sharing clearly generates benefits at a global level, data distribution across countries is not equal. Governments must facilitate data access across borders, while ensuring that ethical and economic standards are respected.
Responding to the new imperatives of the digital transformation, several STI strategies place objectives related to digital transformation at the core of their strategic orientations, often in active consultation with the public (Box 3.2). Developing these strategies also requires engaging with the public to establish a social licence, by demonstrating the beneficial aspects of these technologies and addressing public concerns through better information and appropriate action (e.g. protecting privacy and developing certain applications for the public good). A lack of engagement with society creates the risk of a significant future backlash, with negative impacts on the development and deployment of these technologies.
Germany’s New High-Tech Strategy sets priorities for research and innovation, listing the “digital economy and society” as its first priority. The High-Tech Strategy supports science and industry’s implementation of Industry 4.0. It considers the successful development and integration of digital technologies within industrial application sectors as key to the country’s future competitiveness. It also supports smart services, big-data applications (particularly focusing on SMEs), cloud computing, digital networks, digital science, digital education and digital-life environments.
The Estonian Research and Development and Innovation Strategy 2014-20, “Knowledge-based Estonia”, aims to increase the economy’s knowledge intensity and competitiveness. It identifies information and communication technologies (ICT) (e.g. their use in industry, cybersecurity and software development) as one of three key priority areas for investment in research, development and innovation. The other two priority areas are resource efficiency, and health technologies and services.
France Europe 2020: A Strategic Agenda for Research, Technology Transfer and Innovation places research at the centre of France’s policy priorities. It views research (including basic research) as key to addressing the main emerging scientific, technological, economic and social challenges, and promoting competitiveness. France Europe 2020’s priorities include strengthening research in breakthrough technologies, and investing in digital training and infrastructures.
Slovenia’s Smart Specialisation Strategy includes Industry 4.0 as one of the three key priority areas for action. It highlights the need to optimise and digitalise production processes, and to apply a range of enabling technologies (e.g. robotics, nanotechnologies and modern production technologies for materials) to specific priority areas (e.g. smart buildings, the circular economy and mobility).
Austria’s Open Innovation Strategy is the country’s response to the challenges of digital transformation and globalisation. Its main objective is to open up, expand and further develop the innovation system in order to boost its efficiency and output orientation, and improve innovation actors’ digital literacy. The Open Innovation Strategy formulates 14 measures around 3 action areas: 1) developing a culture of open innovation and teaching open-innovation skills among all age groups; 2) creating heterogeneous open-innovation networks and partnerships across disciplines, industry branches and organisations; and 3) mobilising resources and creating adequate framework conditions for open innovation.
Japan’s Fifth Science and Technology Basic Plan emphasises the importance of achieving “Society 5.0”, also defined as a “super-smart society”. To that end, it sets the development of cutting-edge ICTs and the IoT as top science and technology policy priorities. The Basic Plan also encourages further developing AI, while minimising risks and limiting automated decision-making.
With the rise of AI, digital innovation will continue to expand and even accelerate, involving all fields of technology and all types of innovation. AI allows using large quantities of data more effectively and applying digitalisation to new areas, like driving vehicles. The coupling of AI and the IoT will create virtual twins of most real-world processes, creating a basis for more innovation. Consequently, the trends identified in this chapter will continue: the role of data in the innovation process will continue to grow, innovation will accelerate, and knowledge will become more fluid. Changing innovation policies to reflect these trends will become even more important.
More than a piecemeal approach, a broader strategy is needed, which factors in the profound changes to innovation caused by digitalisation. This strategy would reshape innovation policies and link them more closely with other policy areas. What was formerly a “plus” is now a “must”: with digitalisation, most products are new products, and innovation becomes ubiquitous. Consequently, innovation is directly affected by all policy domains, and what happens in other domains could affect innovation. Better linking policy areas to sustain innovation is a major challenge for governments in the digital era. Several governments have become aware of this issue (Box 3.2), but they are still at an early stage in conceptualising and developing integrated responses, and much learning will be needed in the future.
Governments will also benefit from digitalisation, using digital technologies to adapt their policies and improve policy design, implementation, monitoring and evaluation (Chapter 12 on digital science and innovation policy). The availability of larger amounts of data, and the ability to analyse them more rapidly, will strengthen policy processes. Data are available on all aspects of the innovation process, i.e. technologies, firms, innovation projects, innovation funding, business creation, and, crucially, government policies and programmes themselves. By implementing the appropriate analytical tools, governments will be able to improve their diagnosis (e.g. of technological trends and obstacles to innovation across corporate categories), in order to adopt and evaluate the corresponding policies. They could do this rapidly, facilitating policy experimentation. The way forward also requires developing strategies to leverage and interpret different data sources, to achieve informed decision-taking in a fast-changing environment (Brynjolfsson and Mitchell, 2017).
This chapter covered five dimensions of change for innovation in the context of the digital transformation and identified policy implications. Based on this assessment, the chapter outlined several key implications for innovation policy. The “In my view” box, by Luc Soete (Box 3.3), describes other dimensions of change in the digital age and outlines the new challenges posed by digitalisation. It also discusses challenges for innovation policy as it aims to become more agile and contribute to achieving wider societal goals.
Luc Soete, University of Maastricht
How has current digitalisation affected innovation processes and outcomes? Let me just add to the five dimensions listed in this chapter a few of what I would call “low hanging digital-fruit opportunities”. First, the ubiquitous use of data as core input presents opportunities going beyond pure consumer behaviour and is now willingly encroaching on other aspects of human behaviour, i.e. social interactions, attention-seeking, interactive entertainment, health diagnostics, political choices and many more. Second, advertising, has a new and central role, now fully transformed from a supply-based attention-seeking activity, to an information-service and participatory activity. Third, opportunities exist to exploit hidden and underutilised sources of capital – both physical capital (trading, sharing and renting out flats; driving services; second-hand goods; and equipment of all sorts) and human capital (activating underutilised talents and skills). We observe this almost daily in our perception of the current digital age.
The coming digital age, however, raises many more new challenges. These can be best described in terms of the further diffusion and development of some key GPT features of digitalisation, such as AI, robotics and machine learning. These GPTs are likely to lead to further optimisation in production, distribution and service provision, and to increased predictability, also allowing full autonomy. The extensive use of – and access to – data as the core, essential input is likely to cover all sectors – not just personal data on social media, customers and transaction data, and patient data, but also data on education and learning, on delivered public-administration services (such as taxes and social security), and all sorts of behavioural data. The sky is the limit. The likely impact will now go way beyond consumers and economics, influencing citizens in everything they do, including their employment, possible deskilling or reskilling and job security.
The chapter considers eight principles of innovation policy that are crucial to the coming digital age. While this list may seem complete, it resembles a mixture of well-known policy challenges, mostly unrelated to “digitalisation” and specific new digital issues – such as the first issue, access to data.
Moreover, some of the objectives may be hard to reach, e.g. there are limits to the “agility” of innovation policy when it comes to developing speedy and agile policies. How experimental can regulatory policy be? The country examples provided when discussing outcome-focused and anticipatory regulation are interesting, but can they be generalised? The “innovation principle” proposed by the European Commission also comes to mind, but it is quite difficult to implement in reality.
In terms of policies, I would propose an alternative approach, focusing more explicitly on the possible conflicts or trade-offs between, on the one hand, the current policy challenges discussed in the chapter (i.e. privacy/protection; public data sharing versus private data ownership; regulatory boundaries when data go beyond consumer or customer data, such as in the case of patient data) and on the other hand, the future digital challenges and opportunities (i.e. how to enable production and distribution optimisation across the board; how to develop machine learning and predictability, including autonomy, and within which sets of rules and responsibilities; what kind of public-private interactions using individual data; and how to address future employment concerns).
When confronted with such intertemporal potential conflicts or trade-offs in policy making, it would be best to focus on the future digital opportunities in achieving societal goals, such as the “grand challenges” or the Sustainable Development Goals, as guiding principles for the future digital “direction” the coming digital age should take. In a certain sense, the rate and speed of the digital transformation is “out of control”. There is very little that governments or policy makers can do, apart from facilitating its further diffusion through increased training and education in relevant areas of data analysis, AI, robotics and machine learning. However, many citizens across the OECD member countries and beyond are asking more fundamental questions, such as the purpose of these new technologies. What is AI good for? What problems will machine learning solve? In my view, the emergence of the new digital age presents policy makers with a unique opportunity to focus innovation policy on the “direction” of technical change. That direction is ultimately a public responsibility, which governments should pursue readily and actively.
References
Azoulay, P. et al. (2018), “Funding Breakthrough Research: Promises and Challenges of the “ARPA Model”, NBER Working Paper, No. 24674, National Bureau of Economic Research, Cambridge, MA, https://doi.org/10.3386/w24674.
Bechtold, S. and C. Tucker (2014), “Trademarks, Triggers, and Online Search”, Journal of Empirical Legal Studies, Vol. Forthcoming, https://doi.org/10.2139/ssrn.2266945.
Bloom, N. et al. (2017), “Are Ideas Getting Harder to Find?”, NBER Working Paper, No. 23782, National Bureau of Economic Research, Cambridge, MA, https://doi.org/10.3386/w23782.
Board of Innovation (n.d.), Open Innovation & Crowdsourcing Examples, https://www.boardofinnovation.com/list-open-innovation-crowdsourcing-examples/ (accessed on 12 April 2018).
Brynjolfsson, E., F. Eggers and A. Gannamaneni (2016), “Using Massive Online Choice Experiments to Measure Changes in Well-being”, NBER Working Paper, No. 24514, National Bureau of Economic Research, Cambridge, MA, https://doi.org/10.3386/w24514.
Brynjolfsson, E., Y. Hu and M. Smith (2010), “Research Commentary: Long Tails vs. Superstars: The Effect of Information Technology on Product Variety and Sales Concentration Patterns”, Information Systems Research, Vol. 21/4, pp. 736-747, https://doi.org/10.1287/isre.1100.0325.
Brynjolfsson, E. and T. Mitchell (2017), “What can machine learning do? Workforce implications”, Science, Vol. 358/6370, pp. 1530-1534, https://doi.org/10.1126/science.aap8062.
Brynjolfsson, E., D. Rock and C. Syverson (2017), “Artificial Intelligence and the Modern Productivity Paradox: A Clash of Expectations and Statistics”, NBER Working Paper, No. 24001, National Bureau of Economic Research, Cambridge, MA, https://doi.org/10.3386/w24001.
Cockburn, I., R. Henderson and S. Stern (2018), “The Impact of Artificial Intelligence on Innovation”, NBER Working Paper, No. 24449, National Bureau of Economic Research, Cambridge, MA, https://doi.org/10.3386/w24449.
Dai, Q., E. Shin and C. Smith (2018), “Open and inclusive collaboration in science: A framework”, OECD Science, Technology and Industry Working Papers, No. 2018/07, OECD Publishing, Paris, https://doi.org/10.1787/2dbff737-en.
Garicano, L. and E. Rossi-Hansberg (2006), “Organization and Inequality in a Knowledge Economy”, The Quarterly Journal of Economics, Vol. 121/4, pp. 1383-1435, https://doi.org/10.1093/qje/121.4.1383.
Gaspar, J. and E. Glaeser (1998), “Information Technology and the Future of Cities”, Journal of Urban Economics, Vol. 43/1, pp. 136-156, https://doi.org/10.1006/JUEC.1996.2031.
Goldfarb, A. and C. Tucker (2017), “Digital Economics”, NBER Working Paper, No. 23684, National Bureau of Economic Research, Cambridge, MA, https://doi.org/10.3386/w23684.
Government of Norway (2018), Digitalisation strategy for the higher education sector 2017-2021, Ministry of Education and Research, Oslo, https://www.regjeringen.no/en/dokumenter/digitalisation-strategy-for-the-higher-education-sector-2017-2021/id2571085/sec5 (accessed on 29 May 2018).
Guellec, D. and C. Paunov (2018), “Innovation policies in the digital age”, OECD Science, Technology and Industry Policy Papers, No. (forthcoming).
Haskel, J. and S. Westlake (2017), Capitalism without Capital : The Rise of the Intangible Economy, Princeton University Press, Princeton.
Kerr, W. and S. Kominers (2015), “Agglomerative forces and cluster shapes”, The Review of Economics and Statistics, Vol. 97/4, pp. 877-899, https://doi.org/10.1162/REST_a_00471.
MIT (2018), Interdisciplinary Undergraduate Degrees < MIT, MIT Course Catalog - Bulletin 2017-18, http://catalog.mit.edu/interdisciplinary/undergraduate-programs/degrees/ (accessed on 29 May 2018).
Nedelkoska, L. and G. Quintini (2018), “Automation, skills use and training”, OECD Social, Employment and Migration Working Papers, No. 202, OECD Publishing, Paris, https://doi.org/10.1787/2e2f4eea-en.
OECD (forthcoming), Vectors of the Digital Transformation, OECD Publishing, Paris.
OECD (2018a), Rethinking Antitrust Tools for Multi-Sided Platforms, http://www.oecd.org/competition/rethinking-antitrust-tools-for-multi-sided-platforms.htm (accessed on 24 September 2018).
OECD (2018b), Going Digital in a Multilateral World - Document for the meeting of the Council at the Ministerial Level, 30-31 May 2018.
OECD (2017a), Future of work and skills - Paper presented at the 2nd Meeting of the G20 Employment Working Group, http://www.oecd.org/els/emp/wcms_556984.pdf (accessed on 24 May 2018).
OECD (2017b), Algorithms and Collusion: Competition Policy in the Digital Age, http://www.oecd.org/daf/competition/Algorithms-and-colllusion-competition-policy-in-the-digital-age.pdf (accessed on 24 September 2018).
OECD (2016), Big Data: Bringing Competition Policy to the Digital Era, http://www.oecd.org/daf/competition/big-data-bringing- (accessed on 24 September 2018).
OECD (2015a), “Making Open Science a Reality”, OECD Science, Technology and Industry Policy Papers, No. 25, OECD Publishing, Paris, https://doi.org/10.1787/5jrs2f963zs1-en.
OECD (2015b), Data-Driven Innovation: Big Data for Growth and Well-Being, OECD Publishing, Paris, https://doi.org/10.1787/9789264229358-en.
Planes-Satorra, S. and C. Paunov (2017), “Inclusive innovation policies: Lessons from international case studies”, OECD Science, Technology and Industry Working Papers, No. 2017/2, OECD Publishing, Paris, https://doi.org/10.1787/a09a3a5d-en.
Planes-Satorra, S. and C. Paunov (forthcoming), “The digital innovation policy landscape in 2019 (working title)”, OECD Science, Technology and Industry Policy Papers, No. (forthcoming), OECD Publishing, Paris.
Sheehan, J. (2018), Digital Health Innovation:Policy and Standards, Presentation to the OECD Digital Health Innovations workshop, 12 April 2018, the Hague, https://www.innovationpolicyplatform.org/system/files/Panel%201.5.%20Sheehan.pdf (accessed on 01 June 2018).
Yanagisawa, T. and D. Guellec (2009), “The Emerging Patent Marketplace”, OECD Science, Technology and Industry Working Papers, No. 2009/9, OECD Publishing, Paris, https://doi.org/10.1787/218413152254.
Notes
← 1. The chapter builds on the digital and open innovation project of the OECD. Guellec and Paunov (2018) provides a more detailed discussion on the implications of the digital transformation on innovation policy. This work builds on and contributes to the OECD-wide Going Digital project.