Chapter 3. Regional policy is facing disruptive technologies
This chapter discusses some of the most important consequences of expected technological change for regional policy. First, it focuses on economic outcomes. In particular, the chapter discusses the consequences of automation on regional labour markets. Second, the chapter analyses the implications of selected technologies for regional policy. This part of the chapter will look beyond economic policies and will discuss the consequences of new technologies on many other important dimensions of regional policy. It looks at the effects of new technologies for public service delivery and discusses how to deal with the emergence of autonomous vehicles.
The statistical data for Israel are supplied by and under the responsibility of the relevant Israeli authorities. The use of such data by the OECD is without prejudice to the status of the Golan Heights, East Jerusalem and Israeli settlements in the West Bank under the terms of international law.
Regions face a continuous challenge to adapt to technological change. This challenge might become more difficult in the future, because the speed of technological change is accelerating. This chapter discusses some of the most important consequences of expected technological change for regional policy. First, it focuses on economic outcomes. In particular, the chapter discusses the consequences of automation on regional labour markets. Second, the chapter analyses the implications of selected technologies for regional policy. This part of the chapter will look beyond economic policies and will discuss the consequences of new technologies on many other important dimensions of regional policy. For example, it looks at the effects of new technologies for public service delivery and discusses how to deal with the emergence of autonomous vehicles.
None of the technological changes discussed in this chapter will have only positive or only negative implications. Unmanaged or poorly managed, they can create more harm than good, but with the right policy responses, they have the potential to improve economic outcomes and quality of life in all regions.
Job automation will have asymmetric impacts on regions
From an economic perspective, automation is likely to be the most important implication of technological progress in the coming years. If automation proceeds as many experts expect, it will have two opposing consequences. On the upside, automation offers a path to revive productivity growth that has been lacklustre in many OECD countries in recent years. Productivity growth raises aggregate incomes and reduces the prices of goods and services. In the long term, it is the only source for sustainable growth in living standards. On the downside, automation also creates the potential of large-scale job losses. This is especially a concern in the short and medium term, as the economy might shed jobs faster in rapidly automating sectors than it can create them in other sectors. This section provides an overview of the regional dimension of automation based on OECD (2018[28]).
The use of manufacturing robots is increasing rapidly
In 2009, the estimated world production of industrial robots was 60 000. In 2017, more than six times as many units (381 000) were sold (IFR, 2018[88]). Most of these are used in industries that manufacture mechanically complex goods, such as cars or electrical equipment. Countries leading in industrial production in these sectors are also the ones that invest the most heavily. Among the five countries with the largest investment, four are OECD countries; Germany, Japan, Korea and the United States. However, since 2015, investment has been the highest in the People’s Republic of China (hereafter “China”). Automation by robots affects sectors beyond manufacturing. Logistics and distribution centres are seeing a rapid increase in automation with, for example, automated carry robots moving goods between fixed workstations and whole automated assembly lines sorting and distributing goods for shipping. The market for these tools remains smaller than for industrial robots, but is expanding rapidly (IFR, 2018[89]).
Both supply and demand contribute to the rise of robots in manufacturing and services. Technologies are constantly improving and, crucially, become cheaper as the industry matures. Especially in fast-growing emerging economies, rising wages make it more attractive to substitute robots for human labour. This is the case for OECD countries where the use of robots is already very much prevalent, but also for countries, such as China, that are still catching up. In Korea, estimates suggest that for each 1 000 manufacturing workers 71 robots are in use; for Germany and Japan the estimates are 32 and 31 robots, respectively; while in China less than 10 robots are in use (IFR, 2018[88]). The absolute price of robots has decreased significantly (De Backer et al., 2018[90]) and is likely to decline even more in the future. As wages tend to rise, the relative decrease in the price of robots compared to wages is even higher. This relative decline means that even in low-wage sectors, such as logistics, automated solutions based on robots become cheaper than human labour.
Job automation will have asymmetric impacts on regions
Automation due to greater use of robots is bound to proceed. However, the most important factor determining the magnitude of job losses due to automation is arguably the evolution of artificial intelligence (AI). Two scenarios provide the plausible range of future developments in this respect. On the one hand, it is possible that the development of AI stalls at the current level (Marcus, 2018[60]). In this scenario, minor progress could be made due to increasing computing power as well as by optimising current applications of AI. However, no new technologies based on AI could be developed in the near future. Productivity growth at the regional and national level could still be achieved, but would rely on innovation related to other technologies and processes.
On the other hand, there is a possibility that the development of AI accelerates drastically. Such a scenario could occur, for example, if AI becomes able to develop new algorithms to solve fundamentally different problems than those for which it was created. In its most dramatic form, the process could culminate in AI being able to simulate all human cognitive processes, an event with unpredictable consequences for human civilisation. According to a survey of 550 researchers on artificial intelligence, a majority estimates that there is a 90% chance that this crucial point will be reached before the end of the current century (Müller and Bostrom, 2016[91]). In such a scenario, the substitutability of human and machine labour would increase significantly over the coming years and decades. AI could reach a point where it could replace workers in most jobs well before it can fully simulate human cognitive processes.
From today’s perspective, neither of the two scenarios can be ruled out in the medium term, but neither is there a consensus that one of them is particularly likely. In response to this uncertainty, this section discusses a middle-ground scenario of the evolution of AI. It is based on expert judgements on the most likely evolution of AI in Frey and Osborne (2013[52]) and presents regional estimates for the number of jobs at risk of automation from OECD (2018[28]).
Figure 3.1 shows the number of jobs that are at risk of automation at the national level if the development of artificial intelligence follows the expected path. Across the OECD, 14% of all jobs are estimated to consist of more than 70% of tasks that are likely to be automated, whereas another 32% of all jobs consist of 50-70% of tasks that are likely to be automated (Nedelkoska and Quintini, 2018[53]).
Previous waves of technological breakthroughs have shown that automation does not spread evenly across space. This is due to the fact that automatable tasks are more prevalent in certain occupations and sectors, and neither occupations nor sectors are evenly distributed within national borders. Thus, areas with a higher proportion of jobs relying on easily automatable routine tasks are likely to experience more disruption, whereas places where jobs involve more complex tasks are less at risk.
Notes: “High risk of automation” refers to the share of workers whose jobs contain at least 70% tasks that are likely to be automated. “Significant risk of change” reflects the share of workers whose jobs contain to 50-70% tasks that are likely to be automated.
Source: OECD (2018[28]), Job Creation and Local Economic Development 2018: Preparing for the Future of Work, https://dx.doi.org/10.1787/9789264305342-en based on Nedelkoska, L. and G. Quintini (2018[53]), “Automation, skills use and training”, https://doi.org/10.1787/2e2f4eea-en.
Regional labour shocks caused by automation will be concentrated in some regions (Figure 3.2). Regions that rely largely on basic manufacturing will be particularly affected. In contrast, many urban economies that have a high share of service sector jobs are less likely to be affected by automation. In addition, while the jobs lost may be concentrated in a few regions, new jobs may well emerge in entirely different regions. Inter-regional migration is one way in which these regional labour market imbalances can be resolved within national borders. However, there are several factors limiting the effectiveness of inter-regional migration as an adjustment mechanism. First, while mobility can be an important structural adjustment mechanism in the long term, it is rarely a short-term solution. People may find themselves out of a job and struggle to find a new one; but they also have family obligations, friends, financial responsibilities, etc. that are tied to where they currently live.
Second, geographical mobility is more restricted for low-skilled workers. This is due to the monetary and non-monetary fixed costs of moving that are proportionally higher relative to income gains from moving for workers with low incomes (Kennan and Walker, 2011[92]). The costs of moving are relatively similar for workers at all income levels. They include monetary costs, for example related to transporting furniture, as well as non-monetary costs, such as the effort required to find new friends. For high-income workers, these costs are often outweighed by the financial gains of finding a new job rather than staying unemployed. However, for low-income workers, the financial gains from moving are frequently not enough to make up for the costs. This is especially a problem in countries where house prices and rents are elevated in economically successful areas and much of the financial gains from higher wages would be absorbed by higher housing costs.
Thus, even under the most optimistic assumptions, it is unlikely that labour market mobility can make up for the uneven impact of automation across local labour markets. Thus, public policy needs to respond to shocks at the local and regional level with targeted measures that take the concrete local impact from automation into account.
How high is the risk of automation at the regional level?
There are large within-country differences in the number of jobs at risk of automation, but it is not straightforward to quantify them. This section presents estimates of the share of jobs at risk of automation at the regional level that are based on OECD (2018[28]). The methodology to produce subnational estimates of the share of jobs at risk of automation builds on several previous pieces of work. The general approach is based on Frey and Osborne (2013[52]), who estimate the risk of automation by occupation based on expert judgements on the expected future capabilities of artificial intelligence. Nedelkoska and Quintini (2018[53]) refine this approach by drawing on information from the OECD’s PIAAC survey on the tasks and required skills within occupations. To derive regional estimates, these numbers are disaggregated by OECD (2018[28]) using the data on the sectoral composition of regional economies to calculate the share of occupations by region. Interested readers are referred to these studies for further information on the underlying methodology.
Notes: High risk of automation refers to the share of workers whose jobs face a risk of automation of 70% or above. Data from Germany correspond to 2013. For Flanders (Belgium), sub-regions are considered (corresponding to NUTS2 level of the European Classification).
Source: OECD (2018[28]), Job Creation and Local Economic Development 2018: Preparing for the Future of Work, https://dx.doi.org/10.1787/9789264305342-en based on Nedelkoska, L. and G. Quintini (2018[53]), “Automation, skills use and training”, https://doi.org/10.1787/2e2f4eea-en.
Figure 3.2 shows the regional disparities in the share of jobs at risk of automation based on OECD (2018[28]). Several countries, including the Czech Republic, France, the Slovak Republic and Spain, display considerable differences in the share of jobs at high risk of automation. In Spain, the country with the largest regional disparity, the difference between the region with the most and least risky job profile is roughly 12 percentage points. In contrast, other countries such as Austria, Canada and Italy show much smaller disparities in the risk of automation. Furthermore, Figure 3.2 reveals an important pattern. Many of the regions with the lowest risk of automation are home to a large urban area. This pattern is due to the concentration of service sector jobs in the urban economy, which are generally less exposed to automation than other occupations.
Figure 3.2 refers to the share of jobs at high risk of automation and all further estimates relate to this measure. However, the share of jobs at risk of automation, or at least at risk of significant change, is much higher. Instead of varying between 4.3% and 39.3% across OECD regions, it varies between 28.8% and 70.0% across OECD regions.
Some occupations have a particularly high risk of automation. Table 3.1 provides an overview of the occupations with the highest risk of automation and shows the total share of jobs in these occupations. Jobs in these occupations are more likely to be automated on a large scale than any other occupations. About 10% of workers across all regions are employed in the five “riskiest” occupations. Food preparation assistants; drivers and mobile plant operators; labourers in mining, construction, manufacturing and transport; machine operators; and refuse collectors face a particularly high risk of automation. As technology develops, their jobs are likely to be the first to suffer significant alterations. Targeted reskilling efforts should therefore be focused on these individuals in regions where a large share of workers are active in the occupations.
The evolution of regional employment and risk of automation over time
Regions can be classified into four categories depending on whether they gain or lose jobs and whether the gains or losses occur in sectors with high or low risk of automation. Table 3.2 shows a classification based on OECD (2018[28]) that divides regions according to whether or not they created jobs between 2011 and 2016 and according to whether job creation occurred predominantly in occupations with high or low risk of automation.
Regions that create jobs in occupations with a low risk of automation (Column A) improve their job situation in the short term and also reduce their long-term risk of unemployment from automation. In contrast, regions that create jobs in occupations at high risk of automation (Column B) improve their short-term job situation, but do so at the expense of moving towards a riskier job profile in the future. Regions that are losing jobs primarily in areas that are at high risk of automation (Column C) have the typical profile of regions in the process of undergoing a structural change caused by automation. While jobs are being lost to automation today, the risk of further job losses due to automation decreases. Lastly, regions that are losing jobs predominantly in occupations that are at low risk of automation (Column D) face the greatest challenge. They suffer current job losses combined with an increasing risk of further job losses in the future due to automation.
Several regions managed to transition towards low-risk jobs in the period 2011-16. Generally, a majority of regions in Europe have been creating new jobs in the aftermath of the financial crisis. Exceptions to this rule include some of the areas which were hit harder by the economic downturn: those in southern European countries along with Slovenia and parts of France. In addition, in most countries, more than half of the regions have been shifting towards employment that is at lower risk of automation (Table 3.2).
A few countries, such as the Czech Republic and Norway, managed to generate overall employment growth and shift towards less risky occupations in all regions. However, most countries experienced either a decline in employment or a move towards more risky jobs in some regions. Five countries had regions where both trends occurred in parallel, i.e. overall employment declined while the share of risky jobs increased.
Without action, automation is likely to result in stronger economic gains in more prosperous regions (and in particular for the highly educated) while creating losses for workers with low and intermediate levels of education in less prosperous regions (see Box 3.1 for a model-based simulation of results for EU regions). While the aggregate gains from automation are positive, the change in the way firms produce will hence have strong adverse consequences for inequality, both across and within regions. Labour demand will shift towards the high skilled, with associated increases in their wages relative to workers with low or intermediate skill levels.1
Building on an economic model of European TL2 regions developed by the Joint Research Centre of the European Commission (JRC), the JRC and the OECD have projected the impact of automation on the regional economy in the European Union.
The efficiency of capital is assumed to rise with technological progress in automation, i.e. the increase in the number of tasks that robots can perform and the quality with which they handle these tasks. This is modelled by an increase in the productivity of capital, i.e. as robots become “smarter”, the amount of output that can be produced for a given investment in capital will increase. This positive productivity shock will lead to an increase in output. The improvement in capital efficiency not only translates into direct output gains as the same amount of capital can now produce more output, but it also creates additional gains from increased capital investment. On average, households benefit from this development. Increased efficiency of the capital stock that workers use to produce output increases their productivity and therefore the wages for those in employment. Households also benefit from lower prices given that productive efficiency increases.
However, not all workers and regions will benefit the same. How much they benefit depends on the assumptions of the underlying model. For the simulation exercise, the working assumption is that technological change will be complementary to the skills that highly educated workers possess, while workers with low and intermediate levels of education are both (imperfect) substitutes for a combination of capital and high-skilled labour, i.e. workers with low or intermediate levels of education can be replaced by both more capital and more high-skilled workers. Consequently, automation creates stronger benefits in regions with a more educated workforce (typically more developed regions) and those with higher capital intensity, which tends to favour less developed regions (the capital share is 45% in less developed regions as opposed to 39% in more developed regions and 38% in transition regions). Overall, the positive effect of having a larger percentage of skilled workers dominates. The total labour income generated in more developed regions increases by about 0.12% for a 5% shock in capital productivity, while less developed and transition regions lose a labour income share by about 0.8% and 0.11%, respectively. Highly educated workers increase their labour share in all three types of regions, by 1% in less developed and transition regions and by more than double that percentage in more developed regions. Workers with low and intermediate levels of education lose income shares in all types of regions, but most in less developed regions where their income shares drop by 1.5% and more.
Simulating the impact of automation on European regions
The RHOMOLO model is a spatial general equilibrium model that is used for policy impact assessment and provides sector-, region- and time-specific simulations to support EU policy on investments as well as reforms covering a wide array of objectives (Lecca et al., 2018[93]). The standard model is combined with tailored estimates on the risk of automation for workers in different industries with different levels of education in European NUTS2 regions by combining results from OECD (2018[28]) with the European Labour Force Survey. The estimated risk of automation augments the elasticity of substitution between capital and labour. To estimate the impact of automation on the economic equilibrium, the simulation considers the deviation from the initial steady-state baseline driven by a gradual and permanent increase of capital productivity by 5%.
Source: Lecca, P. et al. (2018[93]), RHOMOLO V3: A Spatial Modelling Framework, https://ec.europa.eu/jrc/en/publication/rhomolo-v3-spatial-modelling-framework.
More developed regions will benefit more than less developed or transition regions (less developed regions defined by having per capita GDP below 75% of the EU average, and transition regions between 75% and 90%). More prosperous regions where a larger percentage of workers has high levels of education will see the strongest gains from automation as their economy is already well-prepared to reap the benefits through the combination of capital investment and increased use of workers with complementary high skills. In contrast, automation undermines the price competitiveness of less developed regions that often hinges on the low cost of labour. As capital becomes more efficient (and therefore relatively cheaper compared to workers’ wages), production is projected to “reshore” to locations with higher wages and skills. The type of jobs created will, of course, be very different than the ones that will be lost.
Factors that explain the risk of automation at the regional level
The uneven distribution of risks linked to automation raises the question of which kinds of regions will be most affected by it. Identifying the characteristics of these regions will help policy makers concerned with inclusive growth to target policy interventions to the most disadvantaged areas.
Highly automatable jobs are more likely to be concentrated in regions where productivity is low. At least partially, this is because regions with low productivity make less use of advanced machines. Since automation tends to increase labour productivity, regions with low levels of productivity also tend to have low levels of automation. This implies that these regions have more potential for further automation and hence a higher risk of future job losses.
Places with highly educated workforces are less affected by automation. With some exceptions, the risk of automation decreases as educational attainment required for the job increases. Thus, it is no surprise that regions that have a highly educated workforce have a low share of jobs at risk of automation. Figure 3.3 (left panel) shows the share of jobs at risk of automation for three types of regions: those with less than 25% of the workforce with a tertiary education, those with between 25% and 40% of the workforce with a tertiary education, and those with more than 40% of the workforce with a tertiary education. There is a negative relationship between the risk of automation and the share of workers with a tertiary education. Regions that have the highest share of jobs at risk of automation also have the lowest share of workers with a tertiary education. Reducing the risk of automation in those regions will therefore require efforts in training and education.
Note: Data reported in the education chart correspond to regions (TL2) in the Czech Republic, Denmark, Estonia, Germany, Greece, Ireland, Italy, Lithuania, Poland, the Slovak Republic, Slovenia, Spain and the United Kingdom.
Source: OECD (2018[28]), Job Creation and Local Economic Development 2018: Preparing for the Future of Work, https://dx.doi.org/10.1787/9789264305342-en based on Nedelkoska, L. and G. Quintini (2018[53]), “Automation, skills use and training”, https://doi.org/10.1787/2e2f4eea-en and national Labour Force Surveys.
Rural economies are especially at risk of automation. Figure 3.3 (right panel) shows that regions which have a low share of the population living in urban areas have a higher share of jobs at risk of automation. Rural economies have a lower share of service sector jobs, which are better protected from automation. Smaller towns and rural areas are also more likely to be highly reliant on a handful of employers or on a single industry. While this does not necessarily increase the risk of automation in and of itself, it makes it more difficult to absorb displaced workers if one of the employers automates on a large scale.
Automation creates a policy dilemma for economically struggling regions
Automation creates a dilemma for policy makers. On the one hand, economically struggling regions are often in particular need of productivity growth to restore their external competitiveness. In such cases, greater automation by firms is a key measure to increase the required productivity growth. On the other hand, automation threatens to raise unemployment in the short and medium term. This is a threat especially in those regions that are already facing high levels of unemployment combined with a large number of jobs at risk of automation. In those regions, any steps that increase unemployment have particularly severe social consequences and are politically challenging.
The challenge for low-productivity regions is especially daunting because they often have high levels of unemployment in parallel with low productivity levels. These regions have to provide jobs in the short term, but also have to encourage efforts to increase labour productivity to ensure high employment levels and prosperity in the long term.
This dilemma can only be solved by taking two considerations into account. First, policy makers need to embrace automation insofar as it is an important mechanism to increase labour productivity and thus, an important source of wage growth and long-term prosperity. Attempts to prevent automation today only lead to increased risks of job losses in the future. Such job losses could either occur if regional firms rapidly embrace new automation methods that have been pioneered elsewhere or they could occur if regional firms are pushed out of business by more productive competitors from other regions that have embraced automation.
Second, policy makers have to help their local workforce and businesses to deal with the potential downsides of automation. They should consider both worker skills and firm upgrading. Training and reskilling programmes can target people in jobs at high risk of automation. Engaging employers in skills development is important in identifying the set of skills required for the local labour market. Policies that facilitate the transition to new economic activities with higher value added, particularly in regions relying on high automation risk sectors, are also essential. In particular, small and medium-sized enterprises (SMEs) can benefit from training programmes targeted at the management level that inform about the possibilities of digital technologies and provide advice on how to transition to a greater use of digital technology.
Digital innovation in cities
The emergence of artificial intelligence in particular and increasing digitalisation more generally have important consequences for urban and rural areas that go beyond labour market effects. At city level – which is the focus of this section – digitalisation has been at the centre of attention in discussions about “smart cities”. While the definition of smart cities has evolved over time, a smart city can be characterised as using digital innovation as a tool to help local governments boost economic growth, foster well-being and facilitate civic engagement.
Much of the literature on smart cities has focused on the opportunities that digital innovation can create for cities. Local governments are, for instance, at various stages of adopting smart metres to better manage water and energy consumption, or smart sensors to improve traffic flow. Yet digital innovation can also present challenges and risks for cities that need to be better understood and addressed. New technologies disrupt the existing regulatory and policy environment: for example, Uber and Airbnb have transformed the traditional landscape for mobility and short-term rental housing in cities. New technologies can pose challenges to legal and regulatory frameworks, consumer protection, taxation, labour contracts, or fair competition.
Digital technologies, if implemented, would therefore profoundly reshape urban development and management. To analyse their possible implications for cities, this section is organised in two parts. The first sub-section identifies the main opportunities (i.e. objectives) of key smart technologies in cities. The second identifies the main risks, challenges and trade-offs of smart urban technologies.
Smart city objectives: The opportunities of digital innovation in cities
Among the numerous emerging new technologies, several are predicted to have particularly strong implications for urban development and management. These include additive manufacturing (3D printing), AI, big data analytics, Blockchain, civic technology, the Internet of Things (IoT) and unmanned aerial vehicles (drones) (see Table 1.1). In the intermediate future, autonomous vehicles (AV) are also primed to have a strong impact on cities, as will be discussed in the subsequent section.
From a public policy perspective, digital technologies can enable municipal administrations to be more efficient, more responsive and more sustainable. In terms of efficiency, digital technologies can enable public sector interventions to have a larger impact by using fewer resources, including through greater integration of public services. For example, big data availability on transport flows, energy, water and waste systems allows unprecedented depth of analysis and facilitates targeted real-time interventions for a better management of urban systems. The electricity grid is a good example of an increasingly integrated system through information and communication technology (ICT) and real-time data. A key aspect of such “smart grids” is demand- and supply-side management, enabled by smart metres that contribute to energy savings.
Likewise, IoT technologies can support the efficiency of public service delivery. It enables street objects (street lamps, parking metres) to communicate, which allows a continuous monitoring of their performance and scheduling maintenance only when it is needed – or predict when there is danger of a breakdown. McKinsey & Company estimates that the application of this technology could reduce maintenance costs by up to 25%. And, by 2025, the IoT could have a total economic impact of USD 3.9 trillion to USD 11 trillion per year (Manyika and Chui, 2015[98]).
As to responsiveness and transparency, digital technologies can improve cities’ communication with citizens through virtual platforms. The expansion of digital government (e-government) services and civic technology enables a broader range of the population to access public information and services, take better and informed decisions, and express their opinions through online platforms, online petitions or online voting. Civic technology, therefore, could allow greater participatory and democratic engagement around urban issues. In addition, governments increasingly use crowdsourced data to gain real-time detailed information on public service delivery and infrastructure needs, and facilitate appropriate real-time responses. For instance, citizens can report and inform city employees about the location of potholes, broken traffic lights, stray garbage or any other urban challenge they face on a daily basis through smartphone applications. Governments could also better identify individuals in disadvantaged conditions and determine target groups for policy instruments through the completion of online surveys, primary data collections and IoT technologies. For example, wearable devices, telemedicine or e-health could send early warnings of citizens’ health conditions, which would improve the responsiveness of the healthcare system and reduce medical expenses by avoiding emergency care and unplanned hospitalisation. Crowdsourced data, moreover, could assist disaster management in cities.
Digital technologies can also bring opportunities for sustainability and resilience in cities. Unmanned aerial vehicles, for instance, could allow geospatial surveying, and more accurate and cost-efficient air and water pollution monitoring, where information can be shared with citizens in real time. Similarly, early warning systems for floods and other types of natural disasters could improve preparedness and immediate responses. Smart metres and dynamic pricing on electricity have the potential to drastically change the energy consumption patterns of firms and households. They can provide incentives to adapt energy consumption to energy demand. Thereby, they facilitate the use of renewable energy, which tends to have greater supply fluctuations than traditional sources of energy Moreover, electrically powered cars, bicycles and scooters could considerably reduce air and noise pollution. Such a shift towards electric transport modes should be incentivised by a favourable policy environment (tax breaks and exemptions, waivers on road tolls, or subsidy programmes), improvements in the scale and power of the charging infrastructure, as well as uniform standards for charging stations and plugs for all vehicle manufacturers.
Risks, challenges and trade-offs of digital innovation in cities
There are important risks associated with citizen privacy. In an era of open data, big data analytics and the Internet of Things, personal information could be shared with undesirable parties or for unwanted purposes. Such privacy concerns are particularly relevant for health and medical data. In addition, there are risks that open data and big data analytics, which enable information to be tailored to specific groups according to their personal characteristics, could be manipulated by third parties (see (Glancy, 2012[99]), (Helbing, 2015[100]), (European Research Cluster on the Internet of Things, 2015[101]), (Piniewski, Codagnone and Osimo, 2011[102])). Hence, from a public policy perspective, crucial challenges need to be addressed as to the type of data cities should collect and publish as well as for how long it will be stored. In this respect, political considerations, regulatory frameworks, interests and values will be useful to influence, guide and implement citizen privacy-related policies. The OECD has published specific privacy guidelines to advise/inform policy making (OECD, 2013[103]). Meanwhile, the University of Rotterdam (Netherlands) has developed a decision model to help urban policy makers to determine whether and how a data set should be published for reuse (Gemeente Rotterdam,(n.d.)[104]).
In addition to citizen privacy, security breaches that put data and safety at risk are also a challenge and should be a priority for public policy makers. In fact, as digitalisation is increasingly mainstreamed into urban infrastructure, services and activities, public administrations and the private sector are at higher risk of cyberattacks. For example, the 2018 cyberattack on several critical systems in Atlanta in the United States affected the police department, the judicial system, water management and other citizen services. Similarly, in 2017 a number of European hospitals, telecoms and railways were damaged by a co-ordinated cyberattack (Greenemeier, 2018[105]).
As a response, the current regulatory frameworks must be adapted. Along these lines, the European Union proposed a new data protection reform through the 2016 General Data Protection Regulation (European Parliament, 2016[106]) that strengthens privacy and improves the control of citizens over personal data. This will: 1) oblige privacy policies to be written in a clear and straightforward language; 2) require users to give an affirmative consent before their data can be used by businesses; 3) increase transparency regarding data transfers and the purpose of business data collection; 4) give stronger rights to users to access copies of their data held by businesses; move their data to other platforms; have their data deleted (right to erasure); sue companies who process, collect or own private data; and 5) strengthen the enforcement of data privacy laws through higher fines and greater co-operation between data protection authorities.
Current regulatory frameworks must also be adapted to new ways of doing business. In particular, technology companies often control a very large share of their markets, which raises the question to what degree they are monopolies with the potential to harm consumers. Furthermore, regulation is often uneven in areas where digital business models compete with traditional business models. On the one side, newcomers complain that rules and regulations designed for traditional market practices are being applied to newly evolved business models in inappropriate ways. On the other side, there is a gap of rules and regulations for new business models for traditional market players, giving them an unfair advantage.
Not all cities have the human, technological and governance capacity (within local governments) to adapt to new business models in technologically driven environments. In many cases, municipal governments lack the necessary human and infrastructure capacity to develop and adopt comprehensive smart city initiatives, in particular when attempting to incorporate integrated, systems-approaches to urban services within municipal administrations that are often strongly divided by policy area (Kleinman, 2016[95]). For instance, many local governments lack the requisite capacity and skills for collecting, storing and analysing data given the depth and scale required, nor the infrastructure and computing power needed to store and process the data. Building in-house capacity with data scientists is not easy for many cities, given that similar skills are of great value in the private sector as well. Regarding infrastructure and computing power, many cities will not have the financial means or know-how to build and maintain local servers.
Lastly, although smart cities would increasingly rely on data for policy design and implementation, more data do not necessarily translate into better policy making. As Kleinmann (2016[95]) points out, “data are not information”, and reliance on big data may still only provide a piece of the bigger puzzle. Examples of data-driven policy inefficiencies are smartphone applications inviting citizens to report problems on city streets: one study found that the map of potholes reported by citizens systemically corresponded to areas with younger, wealthy residents who owned smartphones rather than an accurate portrayal of the broader street network’s problems. Another study found that social media alerts generated in the aftermath of Hurricane Sandy overrepresented the challenges experienced in Manhattan (given the high density of smartphone users who reported storm-related problems), compared to the challenges in coastal communities that were in reality harder hit (Kleinman, 2016[95]).
Further research is needed to understand the trade-offs between competing policy objectives of efficiency, transparency and environmental sustainability. Big data and smart city technologies depend on sizeable infrastructures (servers, data centres, cabling and power supplies) that consume significant amounts of energy and leave a sizable carbon footprint. Thus, further research should measure the environmental, social and economic impacts of digital innovation in cities. By extension, there is concern that digital innovation might exacerbate existing disparities among social groups that use digital technologies regularly and those that do not.
Autonomous vehicles will transform mobility in cities and beyond
Autonomous vehicles will become available in the intermediate future. Such self-driving cars will drastically increase comfort and safety of road travel. They will also improve the mobility of people who cannot drive today, for example because of disabilities or age. While self-driving cars have the potential to improve many aspects of daily life, they could also create a series of undesired consequences if they are inadequately regulated.
Almost all major car manufacturers as well as several tech companies are working on the development of autonomous vehicles. Advanced prototypes exist that can autonomously drive in most traffic situations. As of mid-2018, Waymo, a leading developer of self-driving cars, announced that its test-fleet drives more than 40 000 vehicle kilometres per day autonomously on public roads (Lebeau, 2018[107]).
In a few cities, pilot programmes allow private customers to use self-driving cars already today. For example, selected customers in Phoenix (United States) can hail rides in self-driving cars through a conventional ride-hailing app (Randall and Bergen, 2018[108]). However, as of the time of writing, the technology is not ready for a universal adoption. This was shown by an accident in which a self-driving car driven in autonomous mode struck and killed a pedestrian in early 2018 during a test drive.
Today’s prototypes of self-driving cars still struggle to handle complex traffic conditions and adverse weather conditions that blind their sensors. Predictions on when the first fully autonomous vehicles go on sale to the general public vary. On the one hand, the most optimistic experts and car manufacturers predict that fully autonomous vehicles will be on sale by 2021 (Walker, 2018[109]). On the other hand, a few experts argue that fully autonomous vehicles will not be available in any foreseeable future (Wolmar and Sutherland, 2017[110]). However, such extreme predictions seem to be outliers. Most experts expect autonomous vehicles to become available at some point during the next decade.
As of late 2018, cars are on the market that can drive autonomously in some circumstances. On the most common definition of vehicle automation, these cars have reached Level 3 (see Table 3.4). In comparison, the cars used in the large field test in Phoenix mentioned above are probably at or close to Level 4. However, the most transformative effects of self-driving cars are likely to emerge once cars reach Level 5, i.e. when cars will be able to drive in all traffic conditions autonomously. Only once this technology is available will cars be able to drive while the occupant is asleep or even drive without any occupant. Such cars will not require a steering wheel and other control instruments of today’s cars. The subsequent section will discuss the implications of such fully autonomous vehicles that are able to drive in all traffic conditions on public roads without a driver.
Fully automated cars will probably not become available in all places at the same time. Most likely, cities or countries with advantageous climate conditions (e.g. no snow and little rain), orderly traffic and a favourable regulatory environment will see an earlier introduction than other places. Such a staggered introduction offers policy makers two advantages. First, the timeline for the introduction of self-driving cars becomes more predictable once a large-scale rollout begins in some places. Second, policy makers in most places will be able to learn from the experience of the early adopters and can adjust their policies accordingly to deal with any undesired consequences.
It is uncertain how quickly self-driving cars will become widespread once they are available. First, it is not yet foreseeable for how long manually operated cars will be sold once self-driving cars become available. Second, the lifespan of today’s cars is well over a decade. Depending on whether these cars are driven until the end of their economic lifespan, the widespread adoption of self-driving cars may take more or less time. In any case, even if the transition to self-driving cars happens quickly, it will take several years from the time the first self-driving vehicles become available to their widespread use. Given these constraints, it appears unlikely that self-driving cars will make up a majority of all cars before the early 2030s.
Autonomous vehicles will change where people live and how they use cars
Once cars can operate without driver, vehicle occupants will be able to use their time during a trip for various activities. For example, occupants might work, sleep or read during a commute to and from work. Thus, commuting by car could become much less of a hassle than it is today. This effect might be amplified by a fundamentally revised design of cars. Once the need for a traditional car cockpit disappears, cars could be redesigned to resemble small living rooms, offices or other spaces designed to maximise work productivity or comfort (Litman, 2018[112]).
If commuting becomes more pleasant, people may decide to live much further away from cities than they do nowadays in order to live in larger homes or be surrounded by more green space (Metz, 2018[113]). A commute of 90 minutes might not be so daunting if the time can be used to answer emails in the morning and to read or watch TV in the evening. As a consequence, self-driving cars are likely to lead to a renewed suburbanisation process in the absence of policy interventions to prevent this. Most likely, this suburbanisation process would be felt most strongly in ex-urban areas around cities that are beyond today’s commuter belt of cities. These areas would experience price increases and could see significant new construction. In contrast, house prices in suburbs close to city centres might decline in relative terms because the advantage of living close to a city centre becomes smaller relative to more remote locations. Even the higher costs of longer commutes would not necessarily provide a deterrent from living further away from places of work in city centres. At least in cities with expensive housing prices, the savings from lower housing costs further away from cities could compensate for the higher costs of longer commutes.
Better planning at the metropolitan level is necessary to prevent uncontrolled suburbanisation. Where metropolitan planning or other co-ordination mechanisms at the metropolitan or regional level exist, their geographic boundaries might have to be adjusted to account for the fact that people will commute longer distances into core cities. Where planning is not co-ordinated within metropolitan areas, the need to do so will increase because of self-driving cars.
Autonomous vehicles are likely to have strong effects on employment. In the United States alone, there are an estimated 3.8 million professional drivers who transport goods or people (Beede, Powers and Ingram, 2017[114]). As soon as self-driving cars become widespread, many of these jobs could become redundant. At the same time, new jobs are likely to emerge related to the servicing of autonomous vehicles, but also in professions that do not yet exist.
Given that the employment effects of autonomous vehicles are likely to be similar to those of other technologies that lead to automation, they are not discussed in this section. Interested readers are referred to Section 0 of this chapter, which discusses the labour market effects of automation more broadly.
Car sharing will become widespread
Self-driving cars are likely to lead to a widespread adoption of car-sharing services (ITF, 2015[115]). Shared self-driving cars would operate similarly to current ride-hailing services. Instead of picking up a car at a designated location, a shared self-driving car could drive to the user on demand. Once it is possible to order a car on demand at any time, the need to own a private car will decline drastically. For many people, using shared cars would also offer a financial advantage, as it would be much cheaper to rent a shared car occasionally than to own a private car. Car sharing will be especially attractive for people who use cars only infrequently or drive short distances. For those who drive long distances regularly, savings will be smaller because the costs of owning and sharing a car converge, the more a car is used. Furthermore, cost savings of shared cars would probably be lower for commuters who have to rent a shared car during rush hour. It is likely that car-sharing companies will charge higher prices during these times to account for increased demand.
Car-sharing services have the potential to drastically reduce the number of cars that are needed to transport a given number of people. By increasing the time that a car is in use, the same number of cars can serve a much larger number of trips than individually owned cars. However, at least initially, the widespread introduction of car-sharing services could lead to some frictions. For example, car-sharing companies need to ensure that there is a near-constant availability of cars to be attractive to customers. If many car-sharing companies enter the market at the same time, this could lead to an oversupply of cars on the road.
Self-driving cars will threaten established public transport models
Ride-hailing schemes based on shared self-driving cars could pose a significant threat to existing public transport systems. Currently, it is prohibitively expensive to use taxis and other ride-hailing schemes on a daily basis for most of the population in OECD countries. However, a large share of the costs of a taxi is due to labour costs of the driver. In comparison, the costs of operating a private vehicle per kilometre tend to be much lower than the costs of a taxi. For example, the German Automotive Club estimates the total operating costs for a typical middle class car to be approximately EUR 0.45 per kilometre (ADAC, 2018[116]). This figure includes capital costs, depreciation, fuel, maintenance, taxes and other costs.
Once self-driving cars are widespread, it is unlikely that the kilometre costs of a self-driving taxi that is ordered on demand will be much higher than the costs of operating a private vehicle today. Competition between operators will push prices towards their marginal costs and operating costs of autonomous cars will not be much different than operating costs of cars today. Bösch et al. (2018[117]) estimate that the costs of a vehicle kilometre in an individual autonomous taxi would be CHF 0.41 in Switzerland (approximately USD 0.42). Other estimates fluctuate between USD 0.2 and USD 0.6 per kilometre (Litman, 2018[112]). If autonomous taxis operate in a car-pooling mode with several passengers on board, the kilometre costs per passenger would be even lower. As a consequence, the price for a trip of a few kilometres in a self-driving taxi would be approximately as high as the price of a single public transport ticket in many cities in OECD countries.
Given the low costs, it is likely that many public transport users would switch to self-driving taxis that are ordered on demand. Such a shift would threaten the financial viability of many public transport systems and could lead to increased road traffic and congestion. To prevent a major shift away from public transport, two sets of policy measures should be introduced.
First, self-driving cars should be integrated with public transport systems. In areas that are well-served by public transport, self-driving cars are direct competitors for the public transport system. However, in less densely populated areas that have an insufficient public transport coverage, self-driving cars could complement public transport by helping to cover the last “mile” that is not covered by public transport. To make this attractive for users, the interchange between self-driving car and public transport should be made as seamless as possible. At a suburban train station, this can, for example, mean providing infrastructure for a large number of self-driving cars that will wait for passengers arriving on a train during evening rush hour.
Second, congestion pricing and other tax and regulatory instruments should be used to limit congestion and prevent large shifts away from public transport to individual transport (see the following section). Taxes on trips in self-driving cars may be adjusted to encourage multi-modal trips. For example, taxes per kilometre on trips in which self-driving cars are used to cover the last mile to and from a train station could be lower than taxes for trips that take place entirely in self-driving cars.
Self-driving cars need to be regulated and taxed to avoid congestion
Given the convenience of self-driving cars that are ordered on demand and their low cost, it appears likely that the number of trips would increase considerably. To counteract the expected increase in demand for trips, congestion pricing and other taxes on the externalities of self-driving cars should be introduced. They could take into account the time and route of a trip in the calculation of the tax. Such pricing schemes should encourage the use of carpooling and the above-mentioned integration of self-driving cars with public transport services.
In densely populated city centres, strongly increasing demand for trips in self-driving cars can only be accommodated without increasing congestion if carpooling becomes widespread. In such a scenario, passengers with similar itineraries share a vehicle and are picked up and dropped off on the way. With a sufficiently large number of users, passengers can be selected automatically by an algorithm so that their itineraries largely overlap. This works better the greater the number of people using a carpooling service, because it becomes easier to find users with matching itineraries for a vehicle. In an ideal scenario, carpooling services would take only slightly more time than using a single-occupancy vehicle. In order to incentivise carpooling, public authorities should tax it at lower rates than single-occupancy use of cars.
A further cause of congestion could be self-driving cars that circulate empty in wait for passengers. While this helps to assure prompt availability of a car in an on-demand service, it also adds to congestion and pollution. In cities where costs of parking are high, it could even be cheaper to have a car circulate permanently on the road instead of parking it for a fee. To prevent excessive circulation of empty vehicles, it should be monitored and, if necessary, limited. Authorities could impose special charges on cars that circulate empty or prohibit extensive empty circulation entirely.
The increase in car usage due to self-driving cars and the available road capacity will vary strongly from city to city. Taxes, fees and other regulatory instruments need to be adapted to the local and regional circumstances. Thus, legal frameworks at the national level should allow for regionally and locally variated taxation and regulation.
Eventually, self-driving cars could increase road capacity because shorter reaction times of the computers guiding the vehicles could mean that a much lower safety distance between vehicles is needed. Using so-called platooning, a group of vehicles could circulate in a close formation similar to wagons of a train. This requires autonomous vehicles to communicate with each other. To facilitate the development of this technology, policy makers should implement suitable regulations and standards (Fagnant and Kockelman, 2015[118]).
However, platooning works best on roads where only autonomous vehicles operate. It is unclear if platooning will be a viable option to increase road capacity during a transition period in which self-driving and manually driven cars operate. In cities, the technique is unlikely to be widely applicable anytime soon because roads are shared with pedestrians, cyclists and other users. Thus, unless proven otherwise, policy makers should assume that gains in road capacity from automated vehicles will not be sufficient to compensate for an increased use of cars and prepare measures to limit congestion.
Self-driving cars could free up much of the existing parking space in cities
Once self-driving cars constitute a majority of cars, the need for parking space would decline substantially. Partly, fewer parking spaces would be needed because of the above-mentioned shift towards car-sharing models. Widespread car sharing implies that fewer cars are needed to serve a population than today. Estimates show that the number of cars could decrease by up to 80% (ITF, 2015[115]). Furthermore, those cars would be in use for a much higher percentage of their time than today’s cars. Thus, they will spend less time being parked.
A second factor reducing the need for parking within cities is the fact that self-driving cars can drive autonomously to a parking space. Instead of having to park close to the destination of the vehicle occupant, cars could park where they create the fewest nuisances. For example, cars could park in empty parking lots around stadiums, on brownfield sites outside of cities or in purpose-built parking garages at strategic locations.
The combination of those two factors could make parking space within cities almost obsolete. While there would still be space required to drop-off people, no space for parking would be needed along roads. If self-driving cars do not lead to a widespread adoption of car sharing, parking space in cities could still be reduced, but to a lesser degree. The limiting factor for removing inner-city parking would be the time it takes for an autonomous vehicle to reach its parking destination and the negative effects that the additional traffic creates. Without a shift away from private car ownership, eliminating on-street parking in large cities would probably require the construction of large-scale parking garages at strategic locations.
As Shoup (2017[119]) documents in much detail, parking creates significant monetary and non-monetary costs for society. Thus, a reduction in parking space in valuable locations could provide social benefits, provided the freed up space is put to good use. This would primarily benefit dense city centres, where space is most valuable. Freed-up on-street parking space could be used for other purposes. It would be possible to enlarge sidewalks, build cycle lanes, create green space, or install outdoor seating for cafés and restaurants. Likely, there are other, yet undiscovered, uses for the space, too.
In addition to on-street parking, large parking lots within the urban fabric are common in some OECD countries, such as the United States (OECD, 2018[120]). Often, this parking space is owned by businesses and provided to customers or employees. In some instances, its provision is required by land-use regulations. Most of the parking space would become obsolete under a scenario of widespread adoption of self-driving cars. If it becomes available for development, this space would have significant economic value. Furthermore, developing the space would bring important co-benefits, for example related to the environmentally beneficial impacts of densification.
To use this potential, parking space within cities should be progressively reduced as self-driving cars become available. As a first step, minimum parking requirements for new developments should be abolished. Their downsides outweigh their benefits even without self-driving cars. Once self-driving cars become widespread, they will have even less justification. Thus, these requirements should be reduced or abolished entirely as soon as possible (OECD, 2017[121]). In parallel, the public provision of parking space should gradually be reduced. Given that parking space in many cities already exceeds the socially optimal amount, this process should be started today (Shoup, 2017[119]).
In parallel, alternative uses for the freed-up space should be explored. City-wide plans should be developed that show how the space is allocated across neighbourhoods and across competing uses. Since it is impossible to predict at this point how much and how quickly parking space can be freed up, the plans should include prioritisations and stages that allow a step-by-step reduction in on-street parking. Furthermore, new planning policies should be developed to allow the conversion of inner-city parking lots. Often, these parking lots are in prime locations and offer great economic potential for development as well as the possibility to increase the attractiveness of the urban fabric.
In parallel, planning should begin for how to provide parking for self-driving cars. Given that one shared self-driving car can replace several privately owned cars, the number of new parking spaces needed is likely not as high as the number of parking spaces that will be abolished in city centres. Nevertheless, large cities will eventually have to provide space to park hundreds of thousands of cars outside of rush hour and especially at night. Currently underused spaces such as the parking lots around sport stadiums outside of cities could be suitable sites for such parking. Given that the shift towards autonomous vehicles is likely to coincide with a shift towards electric vehicles, such parking spaces would need to be equipped with large-scale charging infrastructure.
Self-driving cars will increase the attractiveness of rural areas
While the use of cars has declined in urban areas, it remains high in rural areas, where few alternative solutions exist (van Dender and Clever, 2013[122])]; (ITF, 2017[123]). Public transport tends to be costly and inefficient in low-density areas, with long waiting times for passengers and an underutilisation of the system. Under such conditions, self-driving cars will be an important complement to public transport in rural areas. A fleet of self-driving vehicles operating in ride-sharing mode could eliminate the need for traditional public transport in small towns and villages (ITF, 2015[115]).
Self-driving cars will increase the attractiveness of rural areas. As mentioned above, rural areas close to cities will become more attractive as places to live for workers who commute into cities. Greater ease of travelling and improved mobility will also improve quality of life in places further away from cities. Nevertheless, the technology has potential downsides also in rural areas. Important amenities for local communities, such as small shops and restaurants, will face increased competition from competitors in larger cities that will become more accessible. This could contribute to a further decline of activity in the centres of small towns and villages.
Autonomous vehicles will also decrease the costs of freight transport and reduce delivery times. The most immediate savings will come from saving the labour costs of drivers. Autonomous freight trucks will also be able to run around the clock, without the need to respect rest times. This will reduce costs because it allows better capital utilisation. Shorter delivery times will also be a major benefit to producers of perishable goods as well as to companies that rely on just-in-time delivery. For example, producers of quickly perishable food that are located in remote areas will be able to access much larger markets than today or to forego costly air transport.
Good regulation is essential for the successful introduction of self-driving cars
Well-regulated self-driving cars offer the potential of making cities more liveable, greener and more inclusive. They will also improve mobility in rural areas. However, these advantages will not materialise automatically. It will require adequate regulation to prevent unintended consequences. In this respect, self-driving cars are similar to traditional cars that also provided huge benefits, but had severe undesired effects. Regulations need to minimise these side effects, while still encouraging the adaption of the technology.
Once the technology for fully self-driving cars becomes available, there is the possibility that urban mobility will change drastically within a few years. As of today, it is still unclear how and when the transformation towards self-driving cars will occur. As a consequence, it is yet too early to determine the exact policy framework for self-driving cars at the local level. However, it is almost guaranteed that certain steps need to be taken to prevent undesired effects and use opportunities. Policy makers need to be prepared for this change and need to have the right tools to steer this transformation. This includes, in particular, congestion charges and other measures to limit an increase in car use. Since many of these measures will eventually have to be introduced at the local level, national governments should already provide the legal basis to allow local governments to introduce them when necessary.
Last, but not least, it should be mentioned that the successful introduction of autonomous vehicles requires further measures, such as harmonised regulatory frameworks at national and international level. They need to cover aspects such as safety standards and liability regulations. Since these frameworks do not have an explicit regional dimension, they go beyond the scope of this report and are not further discussed.
New technologies for regional development in rural regions
A number of technological changes are likely to shape how rural areas can succeed in a more complex, dynamic and challenging environment. Digitalisation will open up wider possibilities to engage in regional, national and international markets. Along with the spread of high-speed broadband, innovation can create new educational opportunities in rural areas (e.g. e-education), increase social connections, boost productivity (e.g. 3D printing), and change the ways in which land is managed and services delivered (e.g. automated farms or e-health). Other technologies that are likely to have large effects on rural areas are self-driving cars (as discussed above) and better communication techniques.
Better communications techniques
Technologies to work remotely have facilitated the delocalisation of jobs and created organisational changes in many industries (Moriset, 2011[124]). Teleworking, co-working spaces, virtual teams, freelancing and online talent platforms are all on the rise. For instance, in the United States, the share of workers who primarily work from home has more than tripled over the past 30 years, representing currently 2.4% of the workforce (Bloom et al., 2015[125]). Home-based workers have a wide spectrum of jobs ranging from academics and software engineers to managers and sales assistants.
Some OECD countries have considered teleworking as a policy strategy to revitalise rural areas. Gers, for instance, is an organisation that promotes teleworking in 47 towns in the south-west of France. The organisation maintains a network of entrepreneur teleworkers. Together with participating local governments, it provides support for their installation and helps them to integrate into communities (Moriset, 2011[124]).
In the future, the emergence of augmented reality (AR) and virtual reality (VR) technologies could further increase the possibilities for working remotely. AR projects virtual elements into actual surroundings, whereas VR projects an entirely virtual reality. Both technologies have been advancing and are getting closer to the point where virtual imagery becomes indistinguishable from real surroundings. Furthermore, improvements are being made concerning the inclusion of auditory and tactile elements. The technologies could, for example, serve as a much closer substitute for face-to-face meetings than current teleconferencing technologies. Businesses in rural communities, therefore, could particularly benefit from AR and VR as virtual face-to-face meetings could be used to improve connections with customers and suppliers.
Beyond the possibility to improve teleworking experiences, AR and VR will have other economic implications. Estimates suggest that the market for AR and VR has grown fourfold between 2015 and 2018 (Hall and Ryo, 2017[126]). Distinct enterprise applications are now emerging across a variety of tasks. For example, within professional education, the technology is used to simulate workplace environments in various professions, such as quality control, healthcare and driving. In other professions, such as surgery and mechanical maintenance, AR is already used to provide guidance to workers in the workplace by superimposing directions into their field of vision. In the future, AR and VR could also enter mass markets, for example to improve online shopping experiences by providing more realistic impressions of goods (Glazer et al., 2017[127]).
Using AR and VR in rural areas requires access to fast broadband at affordable prices. Thus, policy frameworks should reflect the need for a wider diffusion of digital networks. Ensuring competition in broadband provision, promoting private investments and setting minimum broadband speed are strategies that have been tremendously effective in extending broadband coverage in OECD countries (OECD, 2018[128]). Implementing universal service frameworks to provide telecommunication access when the costs exceed commercial returns has also been widely used to provide services for low-density areas.
Smart service provision in rural areas: Future of health and education
Better ICT and digital connectivity also facilitates a wider provision of public services in rural areas by allowing remote delivery of services. Education is more costly in rural areas and in many cases of lower quality than in urban areas. For example, students in small towns score 31 points lower on average in science than their peers in large cities in the OECD’s Programme for International Student Assessment (PISA) tests (OECD, 2017[129]). While many factors influence learning outcomes of students, long travel distance to schools, limited curriculum options, as well as difficulties in attracting and retaining teachers make education provision more difficult in rural areas.
Policies to integrate digital technologies in schools are common in most OECD countries, but they can be further expanded. Beyond helping students learn crucial ICT skills, they can also improve and expand the educational profile of schools. For example, schools in rural Finland provide elective classes through teleconferencing technologies. Lessons in one school are streamed to classrooms in other schools, where students interact remotely with the teacher. This allows the schools to offer elective classes, such as foreign language classes, which do not have a sufficiently large number of students in a single school (OECD, 2017[70]).
Importantly, many of the challenges related to the introduction of new technologies into online education and other online public services are not related to the technologies themselves, but to organisational processes. For example, integrating video conferencing equipment into classrooms is technologically simple. However, in order to use it to provide lessons remotely in multiple schools, it is necessary to synchronise each school’s timetable. Likewise, teachers need to know how to use the equipment and have to adjust their teaching styles and methods to the new teaching environment. This requires the provision of adequate teacher training (OECD, 2017[129]).
Beyond public service delivery, new technologies also allow new private providers of education to service rural areas. Long-distance education (or online courses), e-learning, podcasting, interactive television teaching tablets, modular coursework and self-directed learning can enrich curriculum opportunities in remote areas (OECD, 2017[129]). For example, online courses can be effective in terms of peer-to-peer interactions and free up teachers’ time. Massive open online courses (MOOCs) are nowadays more common and have created large online communities. The online platform Coursera, for example, has more than 22 million course enrolments across 190 countries.
The second area of service delivery that can be particularly improved in rural areas through new technologies is healthcare. So-called e-health (i.e. the use of information and communications technologies for healthcare provision) is about improving the flow of information through digital means. As of 2016, 43 out of 73 analysed countries had developed a national e-health strategy (World Health Organization, 2016[130]). E-health services can be especially beneficial in remote areas where doctors and other health service providers can be difficult to access. For example, remote consultations of specialists are becoming more and more common across the OECD.
Beyond specialist care, new technologies can also improve day-to-day healthcare provision in areas that are underserved by healthcare providers. Smartphone-based health apps, for instance, is one of the ways in which this trend has progressed the most. Between 2013 and 2015, mobile health apps doubled, reaching 165 000 available apps in 2015 (OECD, 2017[131]). They can perform various healthcare functions, such as the continuous monitoring of patients, interactions between patients and health professionals beyond traditional settings, and communications with systems that can provide real-time feedback from prevention to diagnosis, treatment and monitoring (OECD, 2017[131]).
New health technologies not only create new treatment options, they also modify the procedures for healthcare delivery (OECD, 2017[131]). As in the case of education, using them is not only a question of mastering the technology; it is equally important to adjust organisational structures to integrate them into existing processes. For example, medical specialists need to be available to interact remotely with patients when needed and solutions have to be found to provide urgent intensive care to emergency patients. Furthermore, new technologies are also changing how individuals and communities engage with healthcare providers. Thus, it is important to involve the general public in the development of these technologies and their implementation into existing healthcare systems. Otherwise there is a risk that they will not be accepted by the population.
3D printing: Decreasing reliance on supply chains
Additive manufacturing (often called 3D printing) is a process of making three-dimensional solid objects by adding layers of material on top of each other. It has the potential to transform traditional manufacturing processes based on large centralised factories into a decentralised manufacturing process that integrates large parts of the value-added chain. Decentralised manufacturing technologies have the potential to make small-volume production much cheaper relative to mass production. This could allow some goods to be produced in small volumes directly in regions, rather than be shipped from large factories to rural areas. Eventually, it could allow small businesses and even consumers to design and assemble final products.
The technology is already available today and the 3D printing market is growing rapidly. 3D printers are already capable of producing products from a variety of materials and 3D-printed goods are sold in various sectors including aerospace, jewellery and medical devices (Beyer, 2014[132]). While mass production using 3D printing is still less common, the technology is already significantly altering the market for some machined plastic and metal parts. For instance, Boeing has already replaced machining with 3D printing for over 20 000 pieces (OECD, 2017[133]).
Mainstreaming 3D printing will largely depend on the evolution of the cost of switching from mass-manufacturing methods to 3D printing. The small size of current printers and quality requirements of input materials (plastics, resin, ceramic and metals) still pose barriers to the widespread production of some goods. However, the technology is maturing quickly and is likely to become more common for the production of various goods at competitive prices (OECD, 2017[134]).
3D printing creates new opportunities for economic development in peripheral regions
From a regional development policy perspective, 3D printing offers two major opportunities. First, it reduces the dependence of businesses on established supply chains. Such supply chains are often clustered geographically and businesses located far away from them face more challenging logistics. By integrating the production process, 3D printing reduces the logistical complexity of production, which could potentially benefit businesses in remote regions the most. 3D printing can also bridge supply gaps for the delivery of time-critical parts in the production process. Thus, companies could become less reliant on just-in-time delivery. Likewise, urgent medical goods can be produced and delivered much faster through 3D printing than through centralised production processes. For example, hospitals in rural areas can use 3D printing to prepare tailor-made casts or implants without the need to send specifications to specialised centres and wait for the final prosthesis to be delivered. All these benefits are most valuable in remote regions, where distances are greater.
Second, 3D printing reduces the costs of prototyping and small-scale production (Conner et al., 2014[135]). This is a particular advantage for start-ups and SMEs, which could increase innovation and firm creation. For example, in Colombia, 3D printing has been applied in innovative ways in fashion design (Ishengoma and Mtaho, 2014[136]). By reducing the advantages from economies of scale, this effect also encourages production tailored to local needs. As a consequence, production could become more decentralised. Smaller firms that are located within regions would often be best placed to produce for local markets because they know the demand best.
Although 3D printing is a technology that provides opportunities especially for rural areas, some challenges to use these opportunities remain. There is a lack of professionals qualified to operate and maintain 3D printers. Since these professionals are in high demand across most national economies, rural areas struggle to attract and retain experienced workers (OECD, 2018[137]). Unless this skills gap is overcome, there is a risk that 3D printing will benefit primarily regions with highly educated workforces. Enhancing the knowledge about the technology’s possibilities is critical to allow rural business to prepare and plan production processes. In some cases, government agencies and research institutes provide 3D printing services directly to businesses. For example, regional public agencies such as the Institute for Entrepreneurial Competitiveness in Hidalgo (Mexico) are already offering 3D-printed prototypes to local entrepreneurs, mainly from the textile industry (OECD, forthcoming[138]).
Unmanned aerial vehicles: Improving productivity
Drones or unmanned aerial vehicles are unmanned aircrafts that are remote-controlled or operate autonomously. Drones are already undertaking complex, time-consuming or dangerous tasks in industries such as agriculture, construction, retail, insurance and entertainment. For example, drones are used to check remote infrastructure, such as oil pipelines, count wildlife and monitor forest fires (Rao, Gopi and Maione, 2016[139]). In farming, drones are used to monitor livestock and crops, allowing farmers to survey large areas more quickly. This process can be further automated through intelligent systems that do not require farmers to monitor video feeds but rather flag anomalies that need to be investigated. Beyond monitoring, drones are also used to intervene directly in the agricultural production process, for example by spraying pesticides (OECD, 2018[137]).
Automated drone-based deliveries are still in their infancy, but could transform postal services, especially in sparsely populated areas. Amazon Prime Air, DHL and Google have already conducted tests of deliveries with drones, and in 2015, the US Federal Aviation Administration approved the first commercial drone delivery (Xu, 2017[140]). Amazon has projected that once the service is fully deployed it will be able to deliver more than 80% of its goods through air (Rao, Gopi and Maione, 2016[139]). Drone-based deliveries may become commercially available within five to ten years. They are likely to be firstly deployed in rural areas since it is far more difficult for drones to navigate buildings and infrastructure in more densely populated cities (Xu, 2017[140]). Once widespread, drone-based deliveries would not only improve the supply of goods in rural areas, they could also make it easier for rural producers to reach new markets with their products.
Rural areas can also further benefit from economic activities around the development and testing of drones, which cannot be done in urban areas. As most of the regulatory framework in OECD countries prevents the use of drones in dense urban settings (OECD, 2018[137]), the technology can only be tested in sparsely populated areas. This could make rural areas attractive for technology and R&D companies, which, if well managed, could generate knowledge spillovers in local communities.
However, regulation around the use of drones is often not adapted to conditions in rural areas. At the moment, there are mostly national guidelines on the use of drones, which do not always take regional conditions into account. Sometimes, guidelines are also imprecise about the areas where certain kinds of drone use are permitted. As a consequence, drone users and developers operate in legal grey zones (Levush, 2016[141]). Providing more clarity about the areas and conditions in which certain types of uses are permitted could help the development of an industry built around drones.
Drones also have several potential downsides that need to be addressed. Beyond obvious safety and privacy concerns requiring appropriate regulations, drone deliveries create competition for the local retail sector. This is a concern primarily in rural areas, where the local retail infrastructure is sparse and even the closure of a single shop can be a serious loss for a community.
New approaches to agricultural practices
In the agricultural sector, there is a wide range of innovations with the potential of substantially changing the way food, fibre and biofuel is grown and distributed. All of these developments hold the promise of achieving more resilient, productive, and sustainable agriculture and food systems and enabling comprehensive farm-to-fork traceability. At the core of such innovation lies the increasing capacity to capture, analyse and exchange agricultural data. So far, four key trends have been identified as part of the digitalisation and automation of farms: 1) the ratification of production processes; 2) data-driven decision making; 3) innovation from traditional suppliers as well as new actors entering the market; and 4) the reduction of information asymmetries between different actors (OECD, 2018[142]).
The data-driven technologies that are enabling the surge of “smart farming” or “e-farming” leverage ICT, sensors, the Internet of Things (IoT), robots, drones, big data, cloud computing, artificial intelligence and blockchain (OECD, 2018[142]). The integrated use of these technologies supports farming innovations such as satellite data to monitor crop growth and water resources, automated agricultural production, and ICTs to connect farmers in new ways (OECD, 2018[142]).
For example, precision farming is a pioneer technique that provides farmers with near real-time analysis of key data about their fields, which is paving the way for full automation of farms (OECD, 2017[143]). This technique uses big data analytics to provide productivity gains through an optimised use of agriculture-related resources including savings on seeds, fertiliser, irrigation and even a farmer’s time. The development began with yield mapping and later developed into technology that provides precision-guidance throughout the entire agricultural production cycle (OECD, 2017[143]).
These early products have since been enhanced by using a combination of sensors and GPS on tractors that not only drive themselves, but also use analytic systems that permit the vehicles to plant, water, harvest and communicate among themselves. It is estimated that autonomous tractors can plant or harvest 200-250 hectares per day (in comparison to the 40-60 acres that a single farmer can manage without automated technology) (OECD, 2017[143]). In 2017, the project “Hands Free Hectare” led by Harper Adams University in Shropshire (England) and the firm Precision Decisions, resulted in the first farm in the world to successfully plant, tend and harvest a crop in a completely automated way (Feingold, 2017[144]).
Automation of farms and the large-scale use of ICT systems require fast and reliable mobile Internet connections throughout rural areas. In order to seize the benefits of the deployment of data-collection technologies, policy making should address persisting issues regarding connectivity, particularly in remote regions (OECD, 2018[142]).
Agricultural data governance and regulation will be central to ensure that rural communities benefit from the automation of agriculture. The control of agricultural data by major agriculture technology providers has led to controversial discussions on the potential harm to farmers. The benefits of data-intensive equipment for farmers can be uncertain unless ownership of data is well-defined (OECD, 2017[143]). Local and regional governments should push for data governance regulation that empowers and involves local communities in the automation process, and takes into account local specificities. Some countries have already made step forwards on this. In the United States, the American Farm Bureau Federation met with major providers of precision farming technologies to produce the Privacy and Security Principles for Farm Data in 2016.
Long-term perspectives on the future of food
Several global trends are influencing food security and the overall sustainability of food and agricultural systems. In addition to a higher food demand from a larger world population, income growth in low- and middle-income countries adds pressure on the dietary transition towards higher consumption of meat, fruits and vegetables, relative to that of cereals, which would require shifts in output and pressure on natural resources (FAO, 2017[145]).
Innovative systems that protect and enhance the natural base while increasing productivity are hence needed. Synthetic meat production, insects or land-based fish farming are some examples of how innovative technologies can participate in the future of food. Synthetic meat is a niche technology that can attain the dual goal of coping with an increasing demand for food and protein while reducing the environmental impact of regular livestock (i.e. less land and water consumption). In the immediate future, progress will come from better technologies to use plant protein as a meat substitute. In the long term, meat grown from animal cells in vitro could replace meat from livestock (Alexander et al., 2017[146]). However, it is unclear when the technology will be ready for large-scale adoption: although it is possible to grow in vitro meat in small quantities, several important technological obstacles remain for commercialisation (Hocquette, 2016[147]).
Insects are another alternative source of protein that can be produced with lower levels of greenhouse gas emissions and water consumption. They are high in fat, protein and micronutrients. They have the advantage of having high production efficiency due to their rapid growth rates and maturity. Moreover, 100% of the production is edible – opposed to 40% for cattle (Alexander et al., 2017[146]). Products made from cricket flour are already on the market and several companies are actively researching the potential for the use of insects in protein production. However, while the technological challenges to the use of insects are manageable, cultural challenges concerning the acceptance of this food source are likely to be more severe.
Further technological developments in the field of aquaculture (or aquafarming), more specifically in land-based fish farming, are already changing aquaculture practices. Aquaculture is mostly responsible for the rapid growth of fish for human consumption. In 1974, it provided only for 7% of fish for human consumption and by 2004 it had increased to 34% (China represents more than 60% of global aquaculture production) (FAO, 2016[148]). However, current aquaculture practices are often inefficient, volatile, susceptible to disease and damaging to the environment (Hodgkins, 2017[149]). For instance, conventional aquaculture systems depend on flow-through of clean water from freshwater sources or coastal currents, thus depending on an ample supply of high-quality water. On the other hand, in recycling aquaculture systems, effluent water leaving the tanks is treated and refreshed before being returned, thus reducing water consumption (Kvernevik, 2017[150]). Other benefits include more flexibility for choosing location and species for farming as well as a high yield potential. While research is still ongoing, especially to implement land-based fish farming at an industrial scale, some firms such as Niri in Norway and Marvesta in the United States have begun using it for commercial purposes.
If any of the above-mentioned technologies become widespread, it would have large implications for agriculture. It would create an opportunity for rural regions to diversify their production of food and unlock new business opportunities that are more sustainable, but it would also threaten established modes of production and could harm established producers unless they adapt their business models.
References
[33] ADAC (2018), Autoberater, https://www.adac.de/infotestrat/autodatenbank/default.aspx (accessed on 10 August 2018).
[64] Alexander, P. et al. (2017), “Could consumption of insects, cultured meat or imitation meat reduce global agricultural land use?”, Global Food Security, Vol. 15, pp. 22-32, https://doi.org/10.1016/j.gfs.2017.04.001.
[31] Beede, D., R. Powers and C. Ingram (2017), “The Employment Impact of Autonomous Vehicles”, SSRN Electronic Journal , https://doi.org/10.2139/ssrn.3022818.
[50] Beyer, C. (2014), “Strategic Implications of Current Trends in Additive Manufacturing”, Journal of Manufacturing Science and Engineering, Vol. 136/6, p. 064701, https://doi.org/10.1115/1.4028599.
[25] Bloomberg (2018), Waymo's Self-Driving Cars Are Near: Meet the Teen Who Rides One Every Day - Bloomberg, https://www.bloomberg.com/news/features/2018-07-31/inside-the-life-of-waymo-s-driverless-test-family (accessed on 10 August 2018).
[42] Bloom, N. et al. (2015), “Does Working From Home Work? Evidence From A Chinese Experiment”, Oxford Journal, https://doi.org/\10.1093/qje/qju032.
[34] Bösch, P. et al. (2018), “Cost-based analysis of autonomous mobility services”, Transport Policy, Vol. 64, pp. 76-91, https://doi.org/10.1016/J.TRANPOL.2017.09.005.
[24] CNBC (2018), Waymo's autonomous cars log 1 million miles in a month, https://www.cnbc.com/2018/07/20/waymos-autonomous-cars-log-1-million-miles-in-a-month.html (accessed on 09 August 2018).
[53] Conner, B. et al. (2014), “Making sense of 3-D printing: Creating a map of additive manufacturing products and services”, Additive Manufacturing, Vol. 1-4, pp. 64-76, https://doi.org/10.1016/j.addma.2014.08.005.
[4] De Backer, K. et al. (2018), “Industrial robotics and the global organisation of production”, OECD Science, Technology and Industry Working Papers, No. 2018/03, OECD Publishing, Paris, https://dx.doi.org/10.1787/dd98ff58-en.
[39] Dender, K. and M. Clever (2013), Recent Trends in Car Usage in Advanced Economies-Slower Growth Ahead? Recent Trends in Car Usage in Advanced Economies-Slower Growth Ahead?, International Transport Forum, http://www.internationaltransportforum.org/jtrc/DiscussionPapers/jtrcpapers.html (accessed on 01 October 2018).
[23] European Parliament, C. (2016), Regulation (EU) 2016/679 of the European Parliament and of the Council of 27 April 2016 on the protection of natural persons with regard to the processing of personal data and on the free movement of such data, and repealing Directive 95/46/EC (General Data Protection Regulation), http://data.europa.eu/eli/reg/2016/679/oj.
[18] European Research Cluster on the Internet of Things (IERC) (2015), “Internet of Things: IoT, Privacy and Security Issues”, European Commission: Information, Society and Media,, http://www.internet-of-things-research.eu/pdf/IERC_Position_Paper_IoT_Governance_Privacy_Security_Final.pdf (accessed on 09 October 2018).
[35] Fagnant, D. and K. Kockelman (2015), “Preparing a nation for autonomous vehicles: opportunities, barriers and policy recommendations”, Transportation Research Part A: Policy and Practice, Vol. 77, pp. 167-181, https://doi.org/10.1016/j.tra.2015.04.003.
[63] FAO (2017), The future of food and agriculture: Trends and challenges, http://www.fao.org/3/a-i6583e.pdf.
[66] FAO (2016), The State of World Fisheries and Aquaculture 2016, http://www.fao.org/3/a-i5555e.pdf.
[62] Feingold, S. (2017), “Field of machines: Researchers grow crop using only automation”, CNN, https://edition.cnn.com/2017/10/07/world/automated-farm-harvest-england/index.html (accessed on 1 October 2018).
[7] Frey, C. and M. Osborne (2013), “The future of employment: How susceptible are jobs to computerisation?”, Technological Forecasting and Social Change, Vol. 114/C, pp. 254-280, https://econpapers.repec.org/article/eeetefoso/v_3a114_3ay_3a2017_3ai_3ac_3ap_3a254-280.htm (accessed on 28 February 2018).
[16] Glancy, D. (2012), “Privacy in Autonomous Vehicles”, Santa Clara Law Review, Vol. 52/4, https://digitalcommons.law.scu.edu/cgi/viewcontent.cgi?referer=&httpsredir=1&article=2728&context=lawreview (accessed on 09 October 2018).
[44] Glazer, E. et al. (2017), Virtual reality shopping experience., U.S. Patent and Trademark Office.
[22] Greenemeier, L. (2018), Urban Bungle: Atlanta Cyber Attack Puts Other Cities on Notice - Scientific American, Scientific American, https://www.scientificamerican.com/article/urban-bungle-atlanta-cyber-attack-puts-other-cities-on-notice/ (accessed on 09 October 2018).
[43] Hall, S. and T. Ryo (2017), Augmented and virtual reality: the promise and peril of immersive technologies | World Economic Forum, World Economic Forum, https://www.weforum.org/agenda/2017/09/augmented-and-virtual-reality-will-change-how-we-create-and-consume-and-bring-new-risks/ (accessed on 19 October 2018).
[17] Helbing, D. (2015), “Societal, Economic, Ethical and Legal Challenges of the Digital Revolution: From Big Data to Deep Learning, Artificial Intelligence, and Manipulative Technologies”, SSRN Electronic Journal, https://doi.org/10.2139/ssrn.2594352.
[65] Hocquette, J. (2016), “Is in vitro meat the solution for the future?”, Meat Science, Vol. 120, pp. 167-176, https://doi.org/10.1016/j.meatsci.2016.04.036.
[67] Hodgkins, K. (2017), "Shrimp from the Sahara sounds crazy, but it may be the future of aquaculture", Digital Trends, https://www.digitaltrends.com/cool-tech/future-of-food-biofloc-shrimp-farming/.
[88] IFR (2018), World Robotics: Industrial Robots 2018, International Federation of Robotics.
[89] IFR (2018), World Robotics: Service Robots 2018, International Federation of Robotics.
[54] Ishengoma, F. and A. Mtaho (2014), 3D Printing: Developing Countries Perspectives, https://arxiv.org/ftp/arxiv/papers/1410/1410.5349.pdf (accessed on 02 October 2018).
[40] ITF (2017), ITF Transport Outlook 2017, OECD Publishing, Paris, https://doi.org/10.1787/9789282108000-en.
[32] ITF (2015), “Urban Mobility System Upgrade: How shared self-driving cars could change city traffic”, International Transport Forum Policy Papers, No. 6, OECD Publishing, Paris, https://doi.org/10.1787/5jlwvzdk29g5-en.
[9] Kennan, J. and J. Walker (2011), “The Effect of Expected Income on Individual Migration Decisions”, Econometrica, Vol. 79/1, pp. 211-251, https://doi.org/10.3982/ecta4657.
[12] Kleinman, M. (2016), “Cities, Data, and Digital Innovation”, IMFG Papers on Municipal Financing and Governance, Vol. 24, https://munkschool.utoronto.ca/imfg/uploads/338/imfgpaper_no24_citiesdatadigitalinnovation_markkleinman_feb_10_2016.pdf (accessed on 09 October 2018).
[68] Kvernevik, T. (2017), Industrializing land-based fish farming for a protein-hungry future: An interdisciplinary approachto environmental and economic succes, https://www.oecd.org/tad/crp/2.%20Trond-Inge_KVERNEVIK-Fish-farming.pdf.
[93] Lecca, P. et al. (2018), RHOMOLO V3: A Spatial Modelling Framework, https://ec.europa.eu/jrc/en/publication/rhomolo-v3-spatial-modelling-framework.
[59] Levush, R. (2016), Regulation of Drones.
[29] Litman, T. (2018), Autonomous Vehicle Implementation Predictions Implications for Transport Planning, http://[email protected] (accessed on 10 September 2018).
[15] Manyika, J. and M. Chui (2015), By 2025, Internet of things applications could have $11 trillion impact, McKinsey & Company, https://www.mckinsey.com/mgi/overview/in-the-news/by-2025-internet-of-things-applications-could-have-11-trillion-impact (accessed on 30 November 2018).
[5] Marcus, G. (2018), “Deep Learning: A Critical Appraisal”, http://arxiv.org/abs/1801.00631.
[30] Metz, D. (2018), “Developing Policy for Urban Autonomous Vehicles: Impact on Congestion”, Urban Science, Vol. 2/33, https://doi.org/10.3390/urbansci2020033.
[14] Mohammed, F. et al. (2014), UAVs for smart cities: Opportunities and challenges, IEEE, https://doi.org/10.1109/ICUAS.2014.6842265.
[41] Moriset, B. and Bruno (2011), “Developing the digital economy in French rural regions?”, Netcom 25-3/4, pp. 249-272, https://doi.org/10.4000/netcom.175.
[6] Müller, V. and N. Bostrom (2016), “Future Progress in Artificial Intelligence: A Survey of Expert Opinion”, in Fundamental Issues of Artificial Intelligence, Springer International Publishing, Cham, https://doi.org/10.1007/978-3-319-26485-1_33.
[8] Nedelkoska, L. and G. Quintini (2018), “Automation, skills use and training”, OECD Social, Employment and Migration Working Papers, No. 202, OECD Publishing, Paris, https://dx.doi.org/10.1787/2e2f4eea-en.
[45] OECD (2018), BRIDGING THE RURAL DIGITAL DIVIDE OECD DIGITAL ECONOMY PAPERS, OECD Publishing, Paris, https://doi.org/10.1787/852bd3b9-en.
[60] OECD (2018), How digital technologies are impacting the way we grow and distribute food.
[1] OECD (2018), Job Creation and Local Economic Development 2018: Preparing for the Future of Work, OECD Publishing, Paris, https://dx.doi.org/10.1787/9789264305342-en.
[13] OECD (2018), OECD Blockchain Primer, OECD Publishing, https://www.oecd.org/finance/OECD-Blockchain-Primer.pdf (accessed on 05 October 2018).
[55] OECD (2018), Proceedings 11th OECD Rural Development Conference: Enhancing Rural Innovation, http://www.oecd.org/rural/rural-development-conference/outcomes/Proceedings.pdf.
[37] OECD (2018), The Environmental And Welfare Implications Of Parking Policies, OECD Publishing, Paris.
[49] OECD (2017), New Health Technologies: Managing Access, Value and Sustainability, OECD Publishing, Paris, https://doi.org/10.1787/9789264266438-en.
[61] OECD (2017), OECD Digital Economy Outlook 2017, OECD Publishing, Paris, https://doi.org/10.1787/9789264276284-en.
[47] OECD (2017), OECD Territorial Reviews: Northern Sparsely Populated Areas, OECD Territorial Reviews, OECD Publishing, Paris, https://doi.org/10.1787/9789264268234-en.
[38] OECD (2017), The governance of land use in OECD countries : policy analysis and recommendations., OECD Publishing, Paris, https://doi.org/10.1787/9789264268609-en.
[51] OECD (2017), The next production revolution: A report for the G20, https://www.oecd.org/g20/summits/hamburg/the-next-production-revolution-G20-report.pdf (accessed on 02 October 2018).
[52] OECD (2017), The Next Production Revolution: Implications for Governments and Business, OECD Publishing, Paris, https://doi.org/10.1787/9789264271036-en.
[46] OECD (2017), “Country roads: Education and rural life”, Trends Shaping Education Spotlight 9, http://www.oecd.org/education/ceri/spotlight9-CountryRoads.pdf..
[10] OECD (2016), OECD Science, Technology and Innovation Outlook 2016, OECD Publishing, Paris, https://doi.org/10.1787/sti_in_outlook-2016-en.
[11] OECD (2015), Drying Wells, Rising Stakes: Towards Sustainable Agricultural Groundwater Use, https://www.oecd-ilibrary.org/docserver/9789264238701-en.pdf?expires=1539090314&id=id&accname=ocid84004878&checksum=9593E8436B55909B6E882FB204176C9D (accessed on 09 October 2018).
[20] OECD (2013), The OECD Privacy Framework, http://www.oecd.org/sti/ieconomy/oecd_privacy_framework.pdf (accessed on 09 October 2018).
[56] OECD (forthcoming), Hidalgo Territorial Review, OECD.
[19] Piniewski, B., C. Codagnone and D. Osimo (2011), “Nudging lifestyles for better health outcomes: crowdsourced data and persuasive technologies for behavioural change”, JRC European Commission: Institute for Prospective Studies, http://ftp.jrc.es/EURdoc/JRC64206.pdf (accessed on 09 October 2018).
[57] Rao, B., A. Gopi and R. Maione (2016), “The societal impact of commercial drones”, Technology in Society, Vol. 45, pp. 83-90, https://doi.org/10.1016/J.TECHSOC.2016.02.009.
[21] Rotterdam, G.((n.d.)), Open data beslismodel, http://beslisboom.rotterdamopendata.nl/.
[36] Shoup, D. (2017), The High Cost of Free Parking, Routledge, https://doi.org/10.4324/9781351179782.
[28] Society of Automotive Engineers (2014), Taxonomy and Definitions for Terms Related to On-Road Motor Vehicle Automated Driving Systems.
[26] TO (2018), The Self-Driving Car Timeline - Predictions from the Top 11 Global Automakers, https://www.techemergence.com/self-driving-car-timeline-themselves-top-11-automakers/ (accessed on 10 August 2018).
[27] Wolmar, C. and R. Sutherland (2017), Driverless cars: fantasy or reality? | The Spectator, The Spectator, https://www.spectator.co.uk/2017/12/are-driverless-cars-really-the-future/ (accessed on 10 September 2018).
[48] World Health Organization (2016), Global Observatory for eHealth Global diffusion of eHealth: Making universal health coverage achievable, http://apps.who.int/bookorders. (accessed on 04 October 2018).
[58] Xu, J. (2017), Design Perspectives on Delivery Drones, http://www.rand.org/giving/contribute (accessed on 02 October 2018).
Note
← 1. Education in the context of the underlying economic model is taken as a proxy for the more abstract notion of “skills”.