6. The diffusion of digital skill demands in labour markets
This chapter presents indicators of the pace by which the demand for digital skills has been permeating labour markets across ten countries: Belgium, Canada, France, Germany, Italy, the Netherlands, the United Kingdom, the United States, Singapore and Spain. The chapter focuses on the speed by which advanced data analysis skills, programming skills, automation and Internet of Things (IoT) related skills, as well as cybersecurity and business and sales digital skill demands have been diffusing across occupations over time. Results in this chapter show the significant diffusion of most digital skills across the demand of employers and provides information on areas where bottlenecks and labour market gaps are likely to emerge.
The use of digital technologies has increased exponentially in recent decades and their applications can now be found in a variety of different sectors of the labour market, going from mechanics and manufacturing to services and health care. The computing power of increasingly smaller digital devices such as mobile phones and tablets has allowed the development and implementation of a range of new digital services that are now channelled through mobile applications.
Digital services like Spotify or Netflix, just to mention some amongst the most famous contemporary digital platforms, have revolutionised the way in which music and entertainment are delivered to final customers. These digital services rely not only on the idea of bringing music and entertainment content online but, very importantly, on the specific technical development and deployment of digital technologies such as cloud computing, data analysis as well as machine learning. Netflix, for instance, uses adaptive bitrate streaming technology1 to adjust video and audio quality to match the broadband connection speed and network conditions of the viewers. These technical aspects of the service, in turn, are only possible thanks to the development of cloud storage that is used to save millions of terabytes of information online. Cloud computing itself, however, runs on a worldwide network of secure data-centres, which are regularly upgraded to the latest generation of fast and efficient computing hardware. Operating such complex architectures requires specific digital skills that go from database management skills to software development (usually also called DevOps) to the knowledge of data-oriented programming languages like Ruby, Python and Perl.
While the ones mentioned above are just examples of the widespread adoption of digital technologies in labour markets, it is clear that individuals and workers are nowadays using new digital technologies in an increasing number of jobs and even those who do not use them are seeing the nature of their jobs changing as tasks are increasingly automated by digital processes and by AI-powered technologies. Many of the technologies used today did not exist 10 to 15 years ago. Even from an anecdotal point of view, it is clear that the speed by which digital technologies have permeated labour markets and societies has been extraordinary.
This chapter aims at quantitatively assessing the pace by which the demand for key digital skills in different areas has been spreading across labour markets over time. It does so by leveraging the information contained in online job postings and building indicators that measure the extent by which the demand for digital technologies has been permeating the various different parts of the labour market over time in order to provide insights on current as well as potential future trends.
When using job postings to examine the diffusion of digital skills and technologies across labour markets, several previous studies have focused on counting the increase in the frequency with which the terms related to digital technologies have been mentioned across job postings.
Metrics based on the simple count of the frequency of digital skill mentions are, however, likely to miss whether such increase has been concentrated in a small number of sectors/occupations or if, instead, digital technologies and skill demands have actually spread across a wide variety of sectors and occupations, truly permeating labour markets. This latter question is arguably very important, as widespread diffusion of digital technologies across sectors and different job roles is what drive significant changes in the overall labour market that policy makers and firms need to adjust to.
In order to accurately capture the growth in the diffusion of skill demands in the digital economy across different sectors and occupations of the labour market, this chapter uses machine learning techniques applied to the analysis of online job postings to examine how much digital skills are interconnected with other skills across job vacancies and in employers’ recruitment requirements.2
More “interconnected” skills tend to co-occur in a wide variety of different work contexts and are being used and demanded widely in the labour market in different job roles. To give an example, the ability to operate MS Excel is becoming increasingly important “across” a variety of different jobs as this specific skill becomes more interconnected in a wide range of jobs and it is used in a variety of tasks. To put it differently, the frequency with which the basic knowledge of digital spreadsheets (for instance MS Excel) is required by employers has not only increased in its absolute frequency (i.e. the number of times it is mentioned) but it is also, and perhaps more importantly, required in a wider range of different jobs from the service sector (i.e. to handle requests from customers) to manufacturing (i.e. to analyse shipping of products) up to the health care sector (i.e. to keep track of patients’ records). This suggests that this particular skill demand is permeating the labour market and that its diffusion has increased over time.
The following sections focus on the diffusion of 5 digital skills categories (and of the disaggregated skills therein):
The aim of the chapter is to assess the speed by which the demand for these skills has been diffusing within the labour markets covered in this report.
The remainder of this chapter presents results for Anglophone countries for which longer time series are available starting in 2012 and where the time dynamics are more visible. Results for EU countries, instead, can be calculated for a relatively shorter time span in between 2018 and 2021, a period where labour markets have also been heavily affected by the impact of the COVID-19 pandemic. Due to some of these limitations results, for the EU countries should be considered with some caution and are discussed in a separate section at the end of the chapter.
The text contained in online vacancies can be fed to NLP algorithms that transform the semantic information contained therein into mathematical vectors that can be understood and analysed by a machine (see Annex 6.A). Those mathematical vectors (which are meant to retain the meaning of the words they represent) occupy a specific place in a mathematical high-dimensional space, this latter commonly referred to as a “graph”.
It is then possible to assess when vectors in a graph are connected with each other (when keywords co-occur in a specific job vacancy) or disconnected (when they never co-occur in the same vacancy). In graph theory, the “eigenvector centrality” and the “local clustering coefficient” are two measures that are commonly used to assess the influence of a node in a network or, in other words, to measure the degree and quality of connections of a keyword with the rest of words in the text under exam.
Originally, these measures were developed by researchers in Google and used in the PageRank algorithm to quantify the importance of the connections among web pages based on the textual/semantic information contained in it. The same measures can, however, be used to capture the number of connections that a skill keyword has with other skills as well as the “quality” of those connections across job postings, where higher-quality connections are those with other skills that are also highly connected to the rest of the skills across job requirements. This paper uses the linear combination of the eigenvector centrality and the local clustering coefficient (see Annex 6.A) to measure how well-connected (i.e. pervasive) each skill is in the labour market in a specific point in time. This, in turn, allows to calculate the change in the extent of such connections over time (i.e. the “diffusion index”), and to unveil whether a certain skill has become more or less connected (i.e. more or less pervasive) during the period under exam. These measures are used to indicate whether skills are diffusing across jobs and at what speed.
Every day, humanity generates an incredible two and a half quintillion bytes of data (Marr, 2018[1]). Google alone processes more than 20 petabytes of data every day, which includes around 3.5 billion search queries. These already astonishing figures are poised to increase in the future as more and more digital devices connect to the internet and produce/collect data.
Advanced data analysis skills are, not surprisingly, at the core of the development and adoption of a variety of different digital technologies that leverage the use of the available digital data. The ability to understand data, which often comes in the form of unstructured text (i.e. web searches) or images (i.e. photos or videos uploaded to streaming platforms), is becoming a key skill in high demand in current (and, very likely, in future) labour markets.
In this chapter, five specific skills have been selected and grouped under the umbrella of ‘advanced data analysis’ skills3 and relate to a peculiar aspect of advanced data analysis:
Big data skills, for instance, refer to the analytical techniques used to investigate very large and diverse big data sets that include structured, semi-structured and unstructured data from different sources, and in different sizes from terabytes to zettabytes. Machine learning refers to the use and development of computer systems that are able to learn and adapt their analyses without following explicit instructions, by using algorithms and statistical models to analyse and draw inferences from patterns in data. Similarly, data science uses scientific methods, processes, algorithms and systems to extract knowledge and insights from noisy, structured and unstructured data. Data science is also at the core of the development of artificial intelligence, that is the ability of using new algorithmic techniques to analyse data and produce outputs that mimic human intelligence. Finally, data visualisation is an interdisciplinary field that deals with the graphic representation of data, essential to analyse massive amounts of information and make data-driven decisions based on them.
Evidence presented in Figure 6.1 presents estimates of the speed by which, on average, advance data analysis skills have been permeating labour markets over time. Results indicate that the demand for advanced data analysis skills has been growing considerably faster than the average skills across the labour markets examined in between 2012 and 2021. In Canada, for instance, results show that the demand for advance data analysis skills has been diffusing across jobs published online twice as fast as the demand for the average skill in the same economy. Results indicate that advance data analysis skills are not only mentioned much more frequently in online job postings than a decade ago, but also, that they are mentioned in a far wider range of jobs and work contexts, signalling that their use has spread from narrow contexts (for instance in the IT sector) to a much more varied set of sectors and jobs, from health care to finance for instance.
Results in Figure 6.1 shows significant heterogeneity across countries in the speed by which advanced data analysis skills have permeated labour markets. In Singapore, for instance, the speed with which the demand for advanced data analysis skills has been diffusing across occupations is almost five times that of the average skill.4 Results are larger in the United Kingdom where the demand for advanced data analysis skills has spread 10 times faster than the demand for the average skill and in the United States, a country usually at the forefront of the technology adoption, where advanced data analysis skill demands have diffused more than 15 times faster than the average skill in its labour market between 2012 and the end of 2021. All in all, the analysis of job postings published online between 2012 and 2021 confirms that data analysis skills have become key in a variety of jobs and sectors. These trends are likely to continue in the future, fuelled by the increasing availability of datasets of different sorts, from text to images up to data recorded and collected by interconnected devices and appliances driven by the internet of things.
Note: Values on the horizontal axis represent the speed by which the examined skill has been diffusing in the overall labour market (captured by online job postings) relative to the average skill. A value of 5, for instance, means that the examined skill has been diffusing 5 times faster than the average. The speed of diffusion is calculated using the eigenvector centrality and local clustering coefficient measures as detailed in Annex 6.A. Skills are ordered by the increasing speed of diffusion. Advanced Data Analysis is the average of the other skills listed in the figure.
Source: OECD calculations based on Lightcast data on online job postings.
Among the skills sub-components belonging to the group of advanced data analysis skills, Figure 6.1 reveals that machine learning has been growing the fastest over the period 2012 to 2021 across the countries analysed with the exception of the United States (where the demand for data visualisation and data science skills have been spreading at an even faster pace). Machine learning applications, particularly suited to analyse big data sets, are hence expected to grow at a fast pace, triggering a more widespread use of AI in different areas from health care to services and entertainment (see Box 6.1). As data become integral part of business decisions, also the visualisation of results based on big data will become increasingly important, spurring the demand for professionals able to create intuitive representations of complex systems. This is also reflected in the results in Figure 6.1.
The seminal work by Alan Turing (Turing, 1950[2]) was the first academic contribution to pose the question “Can machines think?” and to suggest the idea that machines could perform cognitive task, replicating the behaviour of humans, in such an accurate way that machines would be eventually indistinguishable from their creators. Since then, a number of different definitions have been provided for the concept of “Artificial Intelligence” (AI henceforth).
The OECD AI’s Experts Group (AIGO) (OECD, 2019[3]) defines an AI system as “a machine-based system that can, for a given set of human-defined objectives, make predictions, recommendations or decisions influencing real or virtual environments. It uses machine and/or human-based inputs to perceive real and/or virtual environments; abstract such perceptions into models (in an automated manner e.g. with machine learning (ML) or manually); and use model inference to formulate options for information or action. AI systems are designed to operate with varying levels of autonomy”.
In recent years, a lively debate has emerged as to whether AI should be treated differently from “traditional” automation technologies that perform narrow routine tasks (Autor, Levy and Murnane, 2003[4]). (Muro, Whiton and Maxim, 2019[5]) argue for instance that AI, differently from more traditional automation technologies, is making significant progress in replicating a particular aspect of intelligence, namely “prediction”, this latter being central to decision making and an essential aspect of high-skilled jobs in the health care or business sector. Similarly, (Felten, Raj and Seamans, 2019[6]) and (Webb, 2019[7]), stressed AI’s ability to perform non-routine cognitive tasks through their ability to autonomously “acquire” and “apply” knowledge in problem solving contexts.
New examples, showing the ability of AI to perform such tasks, are being developed at a fast pace. GPT-3 is, for instance, one of the most sophisticated AI-powered Natural Language Processing (NLP) algorithm to this date. The current version of GPT-3 is able to answer complex medical questions and to correctly identify a disease from the simple description of its underlying symptoms, even suggesting the necessary treatment for the disease at hand. Notably, GPT-3 capabilities are transversal, ranging from its ability to write new software code to that of programming mobile applications or to autonomously produce poems and journal articles when prompted with a few lines of text.
It is clear that the increasing ability of machines to perform cognitive tasks as effectively as humans is poised to have an enormous impact on the way services are delivered, products are manufactured and innovation itself is created. In short, AI is expected to revolutionise the way people interact with machines for leisure and at work in ways that are, however, still difficult to predict.
The fast-paced adoption of AI-powered technologies will certainly have fundamental repercussions on the type of skills that individuals will need to master in at least two separate ways. On the one hand, individuals will need to develop adequate digital and cognitive skills to interact with AI. AI, in fact, does not operate in a vacuum and much of its potential is determined by how well humans are able to interact with it by supplying the correct inputs and understanding the outputs that are produced in return. Similarly, individuals will need to be educated on how to detect biases, fakes and mistakes that could result from the misuse of AI.
Among the digital skills that are on the rise in labour markets’ demands, programming skills occupy a key role as they are fundamental requirements for a variety of jobs that have been growing significantly in the last decade and that are expected to grow even further in the coming years (see Chapter 4). “Programming skills”, as defined in this chapter, encompass a variety of different aspects related to the creation, management and use of software code. Among the skills grouped under this label there are programming principles (i.e. the basic principles, concepts and methods, for how a computation or algorithm is expressed), software development principles (i.e. the set of recommendations that engineers should follow during programme implementation to write clear and maintainable code), software development methodologies (i.e. the processes used in software development that define the strategies and phases used to organise and write the software code, encompassing different approaches such as Agile, Waterfall or Lean), and scripting languages skills (i.e. the knowledge of specific computer languages that can be used to give instructions to other software, such as a web browser, server, or standalone applications as many of today’s most popular coding languages are scripting languages, such as JavaScript, PHP, Ruby, Python, and several others).
Results in Figure 6.2 indicate that the demand for programming skills has been diffusing at a very significant pace across all labour markets analysed. The speed of diffusion, however, varies across countries. The demand for programming skills has been diffusing in the United States and the United Kingdom at a particularly rapid pace (in between 6 and 9 times faster than the average skill) while the diffusion in other countries such as Canada and Singapore has been relatively slower, tough strong.
When looking at the skill sub-components in Figure 6.2, the demand for scripting languages skills, that is the ability to produce code in different programming languages such as JavaScript or Python, has been diffusing significantly faster than the average across all labour markets. In the United States, for instance, the demand for scripting languages skills has permeated the labour market at a pace that is up to 17 times faster than the one for the average skill. To put it in other words, results suggest that the firms and employers in the United States across a wide variety of sectors have been seeking workers able to operate Python, Ruby and many other scripting languages and that such increase in demand has spread across all sectors significantly faster than the average. Just taking the example of Python, this latter is nowadays used in a variety of different programming scenarios, from games to web applications which span across virtually all productive and service sectors.
Similar results, though to a lesser extent, can be found in the rest of the countries analysed, where scripting languages skills show up amongst the fastest diffusing programming skills. Interestingly, results also show that, while scripting languages skills have been diffusing very rapidly in labour markets, the demand for “simpler” programming principles (i.e. the basic principles of programming) have also diffused faster than the average skill, but at a much slower pace if compared with other programming skills. This seems to suggest that demand for more complex and technical programming skills (such as the ability to work with specific scripting languages) is increasing and diffusing at a faster pace than more basic programming skill demands and that, in the future higher-level digital skills are likely to be needed to fill forthcoming digital gaps.
Note: Values on the horizontal axis represent the speed by which the examined skill has been diffusing in the overall labour market (captured by online job postings) relative to the average skill. A value of 5, for instance, means that the examined skill has been diffusing 5 times faster than the average. The speed of diffusion is calculated using the eigenvector centrality and local clustering coefficient measures as detailed in Annex 6.A. Skills are ordered by the increasing speed of diffusion. Programming skills is the average of the other skills listed in the figure.
Source: OECD calculations based on Lightcast data on online job postings.
Digital technologies, the development and implementation of automation and of the internet of things (IoT) are tightly linked. Chapter 2 in this report discusses briefly the key role that automation and IoT are having on labour markets. The adoption of automation technologies across a variety of different sectors, from manufacturing to services, is contributing to the significant shift in skill demands observed in labour markets. Many workers today perform different tasks than a decade ago as they rely on automated machines to carry out the most repetitive operations while humans focus on more cognitive tasks. These changes are happening at a fast pace and spreading across a wide range of sectors, requiring workers to adapt to new technologies and interact with them in areas of the labour market that were unthinkable just a decade ago. Hotel check-in automation, for instance, allow today’s tourists to access their rooms without having to interact with the hotel front desk, by simply showing a bar code. While this reduces the time spent by hotel staff to deal with receipts and documents’ controls, it allows them to work on resolving other issues and emergencies when they happen. For instance, keeping guests’ details in a digital form also means hotels can tailor their experience to them, including notes on possible allergies, preferences, or requests.
This type of automation is linked to the development of the so-called Internet of Things (IoT). IoT is best described as the technology underlying the connection of systems and devices (say the door of the hotel) to sensors, software, and other technologies that enable those devices to exchange data with other devices and systems over the Internet.
In the consumer market, IoT technology is synonymous with products for “smart homes” (i.e. interconnected lighting or heating systems) and “wearables” (i.e. smart watches). The IoT is, however, expanding rapidly to other areas such as “smart cities” (i.e. connecting cars to parking spaces, managing the efficient use of resources etc.). In the agriculture sector, John Deere (an American corporation that manufactures agricultural machinery, heavy equipment, forestry machinery), recently acquired Silicon Valley-based Blue River Technology to further the company’s goal of applying IoT to their activities. IoT, in that context, is used to monitor moisture levels, air and soil temperature and wind speed and convey the collected data to farmers. The company’s tractors and other types of equipment are outfitted with satellite-connected guidance and tracking systems that cull data allow for what’s called “precision farming,” which greatly increases the efficiency of fertilizers and pesticides.
Results in Figure 6.3 confirm the rapid diffusion of skill demands related to automation and IoT in the countries analysed. On average, automation and IoT skill demands diffused up to 6 times faster than other skill demands in labour markets, with a particularly fast pace in the United Kingdom and the United States. Despite their fast diffusion, when compared to other skill demands analysed above (advanced data analysis or programming skills), results seem to suggest that automation and IoT skill demands are still relatively concentrated in a narrower set of jobs. This is to say that, while automation has certainly spread across different sectors, the range of jobs that require a deep knowledge of automated systems or the implementation of new IoT systems seem to remain relatively smaller than in the case of other, more mainstream, digital-related skills.
Note: Values on the horizontal axis represent the speed by which the examined skill has been diffusing in the overall labour market (captured by online job postings) relative to the average skill. A value of 5, for instance, means that the examined skill has been diffusing 5 times faster than the average. The speed of diffusion is calculated using the eigenvector centrality and local clustering coefficient measures as detailed in Annex 6.A. Skills are ordered by the increasing speed of diffusion. Automation and IoT is the average of the other skills listed in the figure.
Source: OECD calculations based on Lightcast data on online job postings.
In a world that is reliant on interconnected devices and where large amounts of sensitive data are stored to improve decision making, cyber threats and data breaches pose significant risks for governments and businesses. In order to reduce vulnerability to cyberattacks, organisations are increasingly investing in cybersecurity and IT risk management. Reducing organisational level vulnerability to cyber threats requires investments in a cybersecurity ecosystem that takes a proactive and not only reactive approach to the management of cyber threats. This involves training cybersecurity professionals, as well as develop organisation-wide cybersecurity risk managerial decision-making, and a broad workforce that operates in a way that reduces exposure to cyber-attacks.
The demand for workers with cybersecurity skills has been increasing significantly in recent years (see Chapter 4) and it is expected to grow even faster in the near future. Several commentators expect that this trend will continue in the future and create bottlenecks and shortages in labour markets, and is already doing so in several countries as geopolitical tensions increase as well as cyber-attacks.
The analysis of job postings shows that the demand for cybersecurity skills has diffused very rapidly within labour markets and across occupations. The knowledge of cybersecurity is required in different roles spanning from Information Security Analysts (in charge of planning implementing, upgrading, or monitoring security measures for the protection of computer networks and information) to Security Management Specialist and Network and Computer Systems Administrators (see Table 6.1).
Across countries, results in Table 6.2 show that the demand for Cybersecurity skills has been diffusing at a fast pace in particular in the United Kingdom and in the United States (in between 6 to 10 times faster than the average skill in those countries respectively). The diffusion, despite being faster than the average, has been relatively slower, instead, in Singapore and Canada where cybersecurity skill demands have been increasing but remained requested with high relevance in a narrower set of jobs.
Digital technologies are nowadays used by businesses in virtually all productive sectors. Increasingly in the last decade, digital tools have been used to spur businesses’ productivity by, for instance, by streamlining accounting operations and supporting sales through business intelligence or through the creation of social media business accounts reach out new and old customers.
In a recent Forbes’ article (Higgings, 2021[8]), for instance, highlights how technologies such as cloud-based data management, process automation and advanced analytics are actually poised to elevate business accountants in new and empowering ways as new digital technologies will support rather than replace workers in those business positions. In the accounting work context, “centralizing data management, particularly through the use of cloud technology, reduces waste and lowers costs considerably by improving communication and collaboration. Standardization and a cohesive datasphere make it easier to capture, access, share and analyze data. Transparency improves as data silos are dismantled, and data quality rises, rather than falls, with data quantity”. In a highly digitalised business, accountants can then put their uniquely human skills to transforming the insights extracted from high-quality data into more effective financial planning and reporting.
Results in Figure 6.4 show that the demand for business accounting digital tools has spread over time significantly faster than other skills. The demand for accounting and finance software related skills has diffused in between 4 to 6 times faster than the average skill in the United Kingdom and the United States. Similar pattern is observed for Canada and Singapore, though the diffusion of the demand for those skills has been relatively slower.
Along with accounting and finance digital tools, businesses have increasingly started using new sales strategies to reach their customers, many of which heavily rely on digital technologies. Among those, a key role is played by businesses’ social media which, channelled through digital platforms and technologies, can significantly spur productivity and sales. (Pourkhani et al., 2019[9]) highlight how “today social media platforms such as Twitter and Facebook enable the creation of virtual customer environments (VCEs) where online communities of interest form around specific firms, brands, or products”. Most large brands today (but also a quickly increasing share of small, medium as well as micro enterprises) have a Twitter, Facebook or Instagram account that they use to interact with customers, manage orders, shipping and requests as well as complaints. Social media are also used to gather information from customers. On Instagram, for instance, businesses can collect feedback from followers by posing questions or polls while other digital tools can also be used to monitor the sentiment of customers by looking for instance at the use of specific keywords (or hashtags) in social media posts.
Results in Figure 6.4 that the uptake of social media has been very fast in the period between 2012 and 2021. The demands for social media skills has diffused up to 3 times faster than the average skill in Canada and Singapore, while the diffusion has been stronger (up to 14 times faster) in the United Kingdom and in the United States, suggesting that a rapidly increasing number of businesses are now hiring (or searching for) workers with skills in the area of social media management.
Note: Values on the horizontal axis represent the speed by which the examined skill has been diffusing in the overall labour market (captured by online job postings) relative to the average skill. A value of 5, for instance, means that the examined skill has been diffusing 5 times faster than the average. The speed of diffusion is calculated using the eigenvector centrality and local clustering coefficient measures as detailed in Annex 6.A. Skills are ordered by the increasing speed of diffusion. Business and Sales Digital skills is the average of the other skills listed in the figure.
Source: OECD calculations based on Lightcast data on online job postings.
Relative to the Anglophone countries analysed above, data on online job postings for EU countries are available for a shorter period of time in between 2018 and 2021. Such a shorter time series makes the detection of long-term trends in skill diffusion comparatively more difficult for EU countries. In addition, most of the time series for EU countries coincide with the period marked by the unprecedented COVID-19 crisis which, as pointed out in Chapter 4, has had important repercussions on labour markets across all countries. Despite these methodological challenges, the analysis of EU data still brings interesting elements to the understanding of the diffusion of digital skill demands across countries.
Table 6.3 shows the list of top 5 digital skills for each EU economy analysed in this report over the period 2018-21, ranked by the speed of diffusion of their demand across online job postings within each country.
The knowledge of how to operate Ubuntu, an open source Linux-based operating system designed for computers, smartphones, and network servers shows among the digital competencies that have diffused faster across EU countries, in particular in Belgium, France and the Netherlands in between 2018 and the end of 2021. In this context it is interesting to notice that Linux has gradually improved its market share in the last few years and that notable growth in its adoption (perhaps triggered by its open-source nature compared to alternatives with significant licencing costs) has been the strongest in 2020 which is also reflected in the speed by which Ubuntu has been diffusing in job requirements across labour markets.
The demand for Javascript skills (a programming language that is one of the core technologies of the World Wide Web) has also been diffusing particularly rapidly across EU countries. As highlighted in Chapter 5 of this report, Java is particularly relevant for software developers and programmers but those skills have been diffused rapidly also in other fast growing digital roles such as computer systems engineers and web developers and UI / UX designers / developers. This is reflected in the results in Table 6.3 where Javascript skills are amongst the digital skills that have diffused faster in Germany, the Netherlands and France. Javascript skills are also related to other technical skills that have been diffusing very fast across the demands of employers in online job postings such as the knowledge of Ajax (a set of web development techniques that uses various web technologies to create web applications), Cascading Style Sheets (CSS, a cornerstone technology of the World Wide Web, alongside HTML and JavaScript) or AngularJs (AngularJS is a toolset for building the framework most suited to application development) which have diffused at a rapid pace in all EU countries analysed.
Results in Table 6.3 also show that several business-related digital skills have been on the rise in EU countries’ demands from employers. The demand for SAP (Systems, Applications and Products in Data Processing) related skills has been on the rise, particularly in Belgium, Germany and Italy, but generally digital competences related to digital marketing as well as Google Analytics and Sales Force have been permeating labour markets much at a fast pace, signalling the adoption by firms of digital tools used to manage their customer and enterprise relationship management.
References
[4] Autor, D., F. Levy and R. Murnane (2003), “The Skill Content of Recent Technological Change: An Empirical Exploration”, The Quarterly Journal of Economics, Vol. 118/4, pp. 1279-1333, https://doi.org/10.1162/003355303322552801.
[6] Felten, E., M. Raj and R. Seamans (2019), “The Variable Impact of Artificial Intelligence on Labor: The Role of Complementary Skills and Technologies”, SSRN Electronic Journal, https://doi.org/10.2139/ssrn.3368605.
[8] Higgings, M. (2021), The Future of accounting: how will digital transformation impact accountants, https://www.forbes.com/sites/forbestechcouncil/2021/05/19/the-future-of-accounting-how-will-digital-transformation-impact-accountants/.
[1] Marr, B. (2018), “How Much Data Do We Create Every Day? The Mind-Blowing Stats Everyone Should Read”, Forbes, https://www.forbes.com/sites/bernardmarr/2018/05/21/how-much-data-do-we-create-every-day-the-mind-blowing-stats-everyone-should-read/.
[5] Muro, M., J. Whiton and R. Maxim (2019), Brookings metropolitan policy program.
[3] OECD (2019), “Scoping the OECD AI principles: Deliberations of the Expert Group on Artificial Intelligence at the OECD (AIGO)”, OECD Digital Economy Papers, No. 291, OECD Publishing, Paris, https://doi.org/10.1787/d62f618a-en.
[9] Pourkhani, A. et al. (2019), “The impact of social media in business growth and performance: A scientometrics analysis”, International Journal of Data and Network Science, pp. 223-244, https://doi.org/10.5267/j.ijdns.2019.2.003.
[2] Turing, A. (1950), “I.—Computing Machinery and Intelligence”, Mind, Vol. LIX/236, pp. 433-460, https://doi.org/10.1093/mind/lix.236.433.
[7] Webb, M. (2019), “The Impact of Artificial Intelligence on the Labor Market”, SSRN Electronic Journal, https://doi.org/10.2139/ssrn.3482150.
The vector representation of skill keywords in a n-dimensional space is also functional to assess the connections across skills and, as such, the degree by which skills are pervasive in the observed labour market published online. The connections between a group of keywords can be represented by a so-called skill graph. In such graph, the keywords extracted from online vacancies represent the vertices (also called nodes) which can be either connected when both vertices co-occur in a specific job vacancy, or disconnected when both vertices never co-occur in the same vacancy.
A so-called adjacency matrix can be built to represent these skill co-occurrences.5 Whenever a skill co-occurs with another skill in a certain job vacancy, the row corresponding to the skill “A”, and the column corresponding to the skill “B” will get the value 1. Note that the adjacency matrix is symmetric, meaning that the co-occurrence between skills is undirected and therefore commutative.
One can hence use this adjacency matrix to calculate the eigenvector centrality (EVC) and the local clustering coefficient (LCC) for each skill. The power iteration algorithm is used to derive the relativity score for each vertex v in the network. Given a graph G, and adjacency matrix A, the relative centrality score of a certain skill can be defined as:
Since this is an undirected graph, the local clustering coefficient can also be defined as:
Both measures serve as an important indicator for contextual diversity and the importance of certain skills as compared to other skills in the network. In graph theory, the “eigenvector centrality” and the “local clustering coefficient” are two measures that are commonly used to assess the influence of a node in a network or, in other words, to measure the degree and quality of connections of a keyword with the rest of words in the text under exam. Originally, these measures were developed by researchers in Google and used in the PageRank algorithm to quantify the importance of the connections among web pages based on the textual information contained in it. The same measures can, however, be used to capture the number of connections that a skill keyword has with other skills as well as the “quality” of those connections, where higher quality connections are those with other skills that are also highly connected to the rest of the skills in the vector space.
One can finally create a unidimensional measure of skill diffusion by normalizing and rescaling the eigenvector centrality and the local clustering coefficient into a single measure using the following:
The change over time of the Diffusion index is used in the analysis above to measure the degree by which skills have become pervasive in the labour market. The Diffusion index is computed for each skill keyword analysed in the database of online job postings and compared to the average diffusion across all skills in each economy where faster diffusion of a skill means an increase (above average) of the connections of that particular skill with other skill demands across job postings, hence an increase in how much that skill is permeating the labour market in a variety of different work contexts and job roles.
Notes
← 1. Adaptive bitrate streaming technology a method of video streaming over HTTP where the source content is encoded at multiple bit rates.
← 2. In particular, the analysis in this chapter applies the concept of eigenvector centrality and local clustering coefficients to the analysis of the information in online job postings. Both eigenvector centrality and local clustering coefficients have been used by companies like Google to assess connections between webpages and to identify networks and relationships between them.
← 3. The selection of skill keywords falling in each broader skill category has been done in consultation with experts from Randstad Research Italy, whose support is greatly appreciated. Indeed, more skill keywords could have added to the selected list, which is not meant to be exhaustive but representative of different aspects related to advanced data analysis.
← 4. As labour markets evolve rapidly, so do also skill demands. Each skill analysed in the database of online job postings may increase its diffusion (if it is mentioned in a wider variety of jobs) or decrease it (if it is mentioned in a narrower set of jobs). Notice that the measure of speed of diffusion for digital skills is computed relative to the average diffusion of skills across the labour market.
← 5. The extracted skill graph forms an undirected acyclic graph, meaning that skills do not co-occur with themselves. As a result the diagonal of the adjacency matrix is 0.