Policy Research Working Paper 10070 The Demand for Digital and Complementary Skills in Southeast Asia Wendy Cunningham Harry Moroz Noël Muller Aivin Solatorio Social Protection and Jobs Global Practice May 2022 Policy Research Working Paper 10070 Abstract As the economies of Southeast Asia continue adopting dig- evidence that highly digital occupations require not only ital technologies, policy makers increasingly ask how to digital skills, but also cognitive and socioemotional skills. prepare the workforce for emerging labor demands. How- Similarly, virtually all occupations, regardless of the digital ever, little is known about the skills that workers need to intensity of the job, require some basic or intermediate dig- adapt to these changes. Skills profiles in low- and middle-in- ital skills. Pairwise correlations and a factor analysis confirm come countries are typically derived from data collected in the complementarity between digital skills and different the United States, which is known to inaccurately reflect subsets of cognitive and socioemotional skills. The data also their occupational skills. This paper uses online job postings confirm that, even with the excitement about the digital data from Malaysia to identify the digital, cognitive, and revolution, the bulk of employment in Southeast Asia is in socioemotional skills required for digital and non-digital low- (around two-thirds) or medium-digital (around one- occupations. The skills profiles for each occupation are third) occupations. Only between 1 and 5 percent of jobs then merged with labor force survey data from Cambodia, are highly digital in the four countries studied. These find- Malaysia, Thailand, and Vietnam to sketch skills profiles of ings suggest that as education and training systems adapt the workforces in these countries. Using descriptive statis- to teach basic digital skills, they will need to continue to tics and linear probability model regressions, the paper finds foster cognitive and socioemotional skills. This paper is a product of the Social Protection and Jobs Global Practice. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http://www.worldbank.org/prwp. The authors may be contacted at wcunningham@worldbank.org. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team The Demand for Digital and Complementary Skills in Southeast Asia∗ Wendy Cunningham, Harry Moroz, Noël Muller, and Aivin Solatorio The World Bank JEL codes: J21, J23, J24, J63. Keywords: Digital skills, cognitive skills, socioemotional skills, employment, Cambodia, Malaysia, Thailand, Vietnam. ∗ We are grateful to Burning Glass Technologies for their assistance with data collection, to the Digital Development Partnership for financing this work, and to Victoria Levin, Andrew Mason, and Achim Schmillen for their feedback. 1. Introduction Digital technology has become common in the workplace. Nearly two-thirds of workers in the European Union reported using a computer or smartphone at work in 2015 (OECD 2019). Computer use at work is less widespread but still prevalent in low- and middle-income countries: 30 percent of respondents in 10 middle-income countries reported using computers at work in 2012-13, ranging from 34 percent in urban Vietnam to 55 percent in Yunnan, China (Saltiel 2020). Ride-hailing and digital payment apps have made smartphones an important tool in even the most elementary occupations, such as motorbike taxis and small shopkeepers, in developing countries around the world. The COVID-19 pandemic accelerated the adoption of digital technologies in workplaces. In response to the pandemic, firms in low- and middle-income countries, including in Southeast Asia, increased their use of and investments in digital platforms and solutions, such as e-commerce technology and online payment systems (Apedo-Amah et al. 2021; DeStefano and Timmis forthcoming, discussed in World Bank 2021). Firms looked for ways to facilitate remote work for workers who could do their jobs from home. Despite these trends, there is limited evidence on the skills needed to accompany the increasing digitalization at work, especially in low- and middle-income countries. While Southeast Asian countries are anticipating an Industrial Revolution 4.0 driven by the rise of digital technologies, there is little information about how these technologies will change the task content and skills requirements of jobs. Demand for digital skills is almost certain to rise. However, the types of digital skills required, and the extent to which occupations outside of the information and communications technology (ICT) sector will require digital skills, remains unclear. Furthermore, there is limited information on the extent to which other crucial skills, specifically cognitive skills (general knowledge and mental abilities) and socioemotional skills (behaviors and attitudes to manage emotions, relationships and personal goals), are needed to learn and manage digital technologies required in the workplace. Previous work using data from high-income countries provides evidence of the complementarity among digital, cognitive, and socioemotional skills. Research using the PIAAC surveys of adults in 30 countries (mostly high-income) shows that workers using more digital skills at work also have higher basic cognitive skills (e.g. numeracy) and perform more tasks that require cognitive and socioemotional skills, such as management and communication, accountancy and sales, and negotiation (OECD 2016, 2019). ICT occupations, such as software and application developers and database professionals, require more advanced digital skills (e.g. programming). However, the PIAAC data also show that workers in ICT occupations perform many tasks that require cognitive and 2 socioemotional skills (OECD 2019). Furthermore, workers in more digitally intensive sectors in those same countries earn higher wages when they have higher numeracy and self-organization skills (both cognitive skills) (Grundke et al. 2018). A similar result is found in the United States: using online job postings for professional occupations, Deming and Kahn (2018) find positive pairwise correlations of 0.5 between posted requirements for basic digital skills (computer literacy and office programs) and skills categories related to cognitive and socioemotional skills, as well as positive correlations of around 0.2 between advanced digital skills (programming and specialized software) and cognitive and socioemotional skills categories. 1 In this paper, we investigate the extent to which digital and non-digital (cognitive and socioemotional) skills are complements in digital and non-digital occupations and how these skills are distributed across the labor market in four Southeast Asian countries. Since detailed data on the skills requirements of occupations is not available for the study countries, we build a data set of skills requirements by occupation using data from more than half a million online job advertisements posted in Malaysia between 2016 and 2018. We classify the skills as digital, cognitive, socioemotional, and others. We use the resulting data set to create skills profiles at the occupation level, which we then match to employment data from nationally representative surveys in Cambodia, Malaysia, Thailand, and Vietnam. We define four levels of the digitalization of an occupation (very-low, low, medium, and high) based on the share of basic, intermediate, and advanced digital skills in each occupational profile. This allows us to answer two questions. First, we explore how digital, cognitive, and socioemotional skills correlate with each other within occupations, focusing on Malaysia where the primary source of data was collected. Second, we estimate the extent to which digital and nondigital skills are used in each country, given the occupational structure of the individual labor markets. In addition to providing the answers to these questions, the data and methodology add value in two ways. First, this is the first paper to use data from a Southeast Asian country to construct occupational skills profiles that include cognitive, socioemotional, and digital skills. Most studies use the occupational skills profiles reported in the U.S. Occupational Information Network (O*NET).2 However, O*NET tends to misrepresent the occupational skills in low- and middle-income countries due to differences in job tasks, skills, and production technologies across countries (Caunedo, Keller, and Shin 2021; Lewandowski et al. 2021; Lo Bello, Sanchez-Puerta, and Winkler 2019). The Malaysia 1 Deming and Kahn (2018) also find that people management (linked to supervision, leadership, non-project management, mentoring, staff), which can be classified as a socioemotional skill, has a much lower correlation with their measures of basic and advanced digital skills: 0.24 and -0.06, respectively (p. 351). 2 See for example Acemoglu and Autor (2011) for the U.S.; Goos, Manning, and Salomons (2009) for Western Europe; and Apella and Zunino (2018) for Latin American and Eastern and Central Europe countries. 3 data on occupational skills profiles is likely to proxy the other three Southeast Asian countries in the study more accurately than occupational skills profiles from the U.S. Second, the analysis identifies skills requirements from unstructured data – job postings – rather than from a predefined list of skills or tasks (as done in a firm survey, for example), allowing the data to “speak for itself” when creating skills categories. The analysis has three conclusions. First, virtually the entire employed population in the four-country sample are in low- and medium-digital occupations. Around two-thirds of the employed population is in low-digital occupations, which means that their tasks require limited use of even the most basic digital technologies. Less than 3 percent work in highly digital occupations, which require frequent use of digital technologies. Second, highly digital occupations require similar levels and kinds of cognitive and socioemotional skills as do low- and medium- digital occupations. Three exercises confirm this conclusion. Using data from Malaysia and regression analysis, we observe that low-, medium-, and highly digital occupations have different shares of digital skills but roughly the same shares of cognitive and socioemotional skills. The few exceptions are that medium- and high-digital occupations more frequently require thinking skills (one of the three cognitive skills subsets that we explore) and low-digital occupations more frequently require relationship skills (one of the three socioemotional skills subsets in our sample). Furthermore, pairwise correlations and an exploratory factor analysis confirm complementarities between levels of digital skills and subsets of cognitive and socioemotional skills. Similar analysis using occupational data from Cambodia, Thailand, and Vietnam show similar results. Third, all occupations are digital to some degree. In Malaysia, the share of skills that are digital in very low-, low-, and medium-digital occupations are 5, 9, and 15 percent. At least 1 percent of skills in every occupation in Malaysia is digital, ranging from 1 percent for cooks to 58 percent for software and applications developers and analysts. Results emerging from Cambodia, Thailand, and Vietnam occupational data show similar orders of magnitude. Overall, our findings show that the education and training sector will need to provide instruction that covers digital, cognitive, and socioemotional skills, regardless of the extent of technological change in the workplace. Since even low-digital occupations will need some level of digital skills, basic digital skills training is necessary for all. 4 2. Conceptual Framework and Definitions of Skills There are intuitive reasons why cognitive and socioemotional skills would be complements to digital skills. Cognitive skills linked to analytical and critical thinking help make sense of the information accessed through digital skills, say surfing the internet or analyzing data. Learning and problem solving can help to develop more digital skills and take advantage of complex computer processes. Socioemotional skills may help workers collaborate through digital technologies, use these technologies to produce creative outputs, and adapt as technologies evolve. 3 As such, we can think of cognitive and socioemotional skills as building blocks of digital skills, which themselves are divided in several levels of complexity. The levels of digital skills may also be pyramidal: tasks requiring more complex digital skills, for example, may not explicitly use less complex ones, but the underlying knowledge and skills encompassed in basic digital skills can be seen as pillars of more complex digital skills. Defining digital, cognitive, and socioemotional skills We define three main skills categories, each with three subsets of skills. The main categories are digital, cognitive, and socioemotional. Skills that do not fit these categories are defined as other technical and language skills. Digital skills are the skills that are needed to work with information and communications technology (ICT) software and devices. Digital skills allow people to access and use digital technologies. Digital skills are often mistaken for a narrow set of specialized skills used by ICT workers, such as programming or data science. However, digital skills encompass a wide range of complexity and are increasingly used across many kinds of tasks in a wide range of occupations (UNESCO 2017, IFC 2019, ITU 2020, DE4A 2021). Figure 1 presents the decision tree that we use to identify and classify digital skills. 3 Cognitive and socioemotional skills are themselves complementary to each other. In the U.S., job requirements for cognitive and social skills are positively correlated with pay and firm performance, and more so when both skills are required (Deming and Kahn 2018). Longitudinal data of adults, also for the U.S., show that wage returns to cognitive and socioemotional skills are increasing (Weinberger 2014, Deming 2017). Online job postings collected in Ukraine show that the shares of postings requiring cognitive and socioemotional skills within an occupation are very similar (Muller and Safir 2019). 5 Figure 1. Decision tree to identify digital skills and their levels Source: Own elaboration Following the framework in UNESCO (2017), we define three subsets of digital skills: basic, intermediate, and advanced (Table 1). At the basic level, workers use technology to perform basic tasks such as using Facebook to advertise products, sending and receiving emails, carrying out online transactions, using mobile apps for driving, and basic word processing. At the intermediate level, 6 workers use existing technologies for analysis, creation, management, and design, for example engineers using modeling software and architects using design software. The advanced level refers to specialized ICT skills such as software design, managing cybersecurity, and data science using big data analytics and artificial intelligence. Table 1. Definitions of digital skills levels Level Definition Examples of tasks 1. Functional use of digital devices Ability to access and use 2. Online communication via emails Basic digital technologies to 3. Using software for presentations, basic spreadsheet use perform basic tasks 4. Finding, managing and storing digital information and content (e.g. social media) Ability to use professional 1. Using professional software for analytics, accounting, project management software for analysis, Intermediate 2. Digital marketing, social media analytics creation, management, and 3. Web design, graphic design design 1. Computer programming 2. Cloud computing, network management Ability to perform 3. Artificial intelligence Advanced specialized ICT tasks 4. Data science, big data analytics 5. Cyber security 6. Web development, search engine optimization Source: adapted from UNESCO (2017) and IFC (2019). Cognitive and socioemotional skills are commonly identified as two fundamental broad skills categories that people use in their work and life. Cognitive skills are our mental abilities to think, learn, and solve problems (Almlund et al. 2011). They range from basic academic knowledge (literacy and numeracy) to more demanding tasks, such as critical thinking, problem-solving, and time management. Socioemotional skills are the attitudes and behaviors upon which we manage personal and social situations (Guerra, Modecki, and Cunningham 2014; Weissberg et al. 2015). Technical skills are commonly identified as a separate skillset, though they may also be thought of as a subset of cognitive and socioemotional skills that are the know-how to carry out one’s specific job, which includes digital skills. These three broad categories of skills are used on the job (World Bank 2018). We identify three subsets of cognitive skills and three subsets of socioemotional skills. Using lists of skills that employers commonly demand identified in previous work, we select three cognitive subsets: communication, organization, and thinking, which summarize 21 individual skills (Table 2) (Cunningham and Villaseñor 2016; Muller and Safir 2019). We also define three socioemotional 7 subsets: emotions, relationships, and personal growth, which summarize 25 socioemotional skills (Table 3). 4 Table 2. Subsets of cognitive skills Subset skill category Skills 1. Analytical Skills 2. Creative Problem Solving 3. Critical Thinking Thinking 4. Decision Making 5. Independent Thinking 6. Problem Solving 7. Communication Skills 8. Oral Communication 9. Presentation Skills Communication 10. Public Speaking 11. Verbal / Oral Communication 12. Written Communication 13. Goal Setting 14. Meeting Deadlines 15. Multi-Tasking 16. Organizational Skills Organization 17. Planning 18. Prioritizing Tasks 19. Scheduling 20. Strategic Planning 21. Time Management Table 3. Subsets of socioemotional skills Subset skill category Skills 1. Coping Strategy Emotions 2. Detail-Orientated 3. Due Diligence 4. Articulate 5. Building Effective Relationships 6. Conflict Management 7. Cultural Awareness 8. Listening 9. Mentoring 10. Negotiation Skills 11. People Management Relationships 12. Persuasion 13. Staff Management 14. Supervisory Skills 15. Teaching 16. Team Building 17. Team Management 18. Teamwork / Collaboration 19. Leadership 20. Thought Leadership 21. Creativity 22. Initiative Personal growth 23. Positive Disposition 24. Self-Motivation 25. Self-Starter 4 Although we base our classification on previous work (Cunningham and Villaseñor 2016; Muller and Safir 2019), we acknowledge that alternative categorizations are possible. 8 3. Data We build a data set of occupational skills profiles for four Southeast Asian countries that we call the Southeast Asia Digital [SEAD] data set. The SEAD is created from online job postings data collected in Malaysia that we link to employment data from household and labor force surveys in Malaysia, Cambodia, Thailand, and Vietnam. The online job postings data are used to identify the types of skills (digital, cognitive, socioemotional, and others) associated with each occupation in Malaysia. These occupational requirements are then extrapolated to the employed population of Malaysia and the three other countries selected because of similar interests in digital development, potential synergies from being in the same sub-region, and the availability of data. 5 The household and labor force surveys provide information on the distribution of employment by occupation in each country. 3.1. Deriving skills requirements from online job postings data The World Bank-Burning Glass Online Job Postings data in Malaysia We begin with a data set of more than half a million online job postings collected by the firm Burning Glass Technologies (Burning Glass hereafter) in Malaysia from May 2016 to December 2018. 6 In collaboration with the World Bank, Burning Glass collected more than 600,000 unique online job postings. 7 More than 95 percent of vacancies listed at least one skill requirement for a total of 8,221 unique skills. The raw data, in the form of job postings, include the job title, a free-text description field that lists required skills, and other job characteristics. Burning Glass parsed the text of each job vacancy, coded keywords and phrases as skills, and categorized the job titles into the 2013 Malaysia Standard Classification of Occupations (MASCO). 8 Burning Glass grouped individual skills into 633 clusters (groups of similar skills commonly learned together or substitutable), 9 which were then further grouped into 28 cluster families that are roughly aligned with sector-occupations (Burning Glass 2019). The World Bank-Burning Glass Malaysia Job Postings data is not representative of the distribution of jobs in Malaysia for two main reasons. First, the job postings collected by Burning Glass between 2016 and 2018 are a subset of all job vacancies during this period: they represent one-fifth of the number of 5 Cambodia and Vietnam are lower-middle income countries and Malaysia and Thailand are upper-middle income countries. 6 For recent studies using Burning Glass data for the U.S., see Deming and Kahn (2018), Acemoglu et al. (2020), and Deming and Noray (2020). For the U.S. and five other countries (Australia, Canada, New Zealand, Singapore, and the United Kingdom), see Cammeraat and Squicciarini (2021) and Squicciarini and Nachtigall (2021). 7 See appendix 6 of CSC (2019, pp. 126–134) for details on the online job postings data collected by Burning Glass in Malaysia. The data we use in this paper uses one additional month of vacancies (December 2018). The preliminary version of this data set was used to identify demand for occupations in Malaysia (CSC 2019). 8 Only 1.8 percent of postings could not be mapped to occupation codes and are thus excluded from the sample from which we draw skills requirements for occupations. 9 Similar skills commonly learned together include, for example, Microsoft Word and Excel in a course on Microsoft Office. 9 vacancies reported by a public online job board and in an employer survey in Malaysia (CSC 2019).10 Second, as in other countries, the job postings are biased towards higher-skilled occupations. Three- quarters of the job postings are for high-skilled occupations (managers, professionals, and technicians corresponding to ISCO-08 1-digit codes 1 to 3). In contrast, in the four countries we study between 7 percent (Cambodia) and 27 percent (Malaysia) of the employed population are in these occupations (Figure 2). In fact, 27 occupations in the labor force surveys of the four countries do not appear at all in the Burning Glass Malaysia job postings data. 11 Figure 2. Distribution of job postings and employment across occupations Share of job postings in the Burning Glass Malaysia data set and employment across 1-digit occupations in Cambodia, Malaysia, Thailand, and Vietnam Sources: World Bank-Burning Glass data set of online job postings in Malaysia (2016-18), Cambodia 2020 Socioeconomic Survey, and 2017 employment surveys of Malaysia, Thailand, and Vietnam. Notes: Occupations are classified with the ISCO-08 (ILO 2012), here at the 1-digit level. The category “skilled agriculture, forestry and fishery workers” includes subsistence and other low-skilled workers and can basically be grouped with the “elementary occupations” category. The discrepancies between the occupational distributions of job postings and current employment are expected for a few reasons. First, job postings represent employers’ search for new hires, which do not necessarily align with the level of current employment in an occupation since some occupations may 10 This is typical of online job postings data. For example, CareerBuilder.com, one of the largest job search sites in the United States, was found to capture 35 percent of all vacancies in January 2011 (Marinescu and Wolthoff 2020). 11 The discrepancy between the occupational distributions of job postings as compared to employment for the Burning Glass Malaysia data is similar to that found in six other countries (Cammeraat and Squicciarini 2021) and Ukraine (Muller and Safir 2019). However, this discrepancy is not always observed, as Marinescu and Wolthoff (2020) find similar distributions of employment in the U.S. and of job postings in of the main job website in the U.S. 10 have higher turnover (and thus vacancies) than others. Second, online job postings mostly target more skilled occupations, often published by formal firms looking for skilled workers who are comfortable using online sites for job search. Third, low-skilled work in South-East Asia frequently consists of self- employed jobs that would not appear on online job search websites (Cunningham 2018). The selectivity of vacancies is not an issue for the purpose of our study as long as there are a sufficient number of vacancies for a given occupation to construct a skills profile. Since we are not analyzing dynamics of vacancies but only using them to define skills requirements, the relative magnitude of the number of vacancies by occupation is irrelevant. Instead, we only need enough information to capture a representative picture of the skills profile for each occupation. 12 Building skills categories We categorize the 8,221 skills in the World Bank-Burning Glass Malaysia data set into our three main skills categories of interest (digital, cognitive, and socioemotional) and nine subsets of the skills categories. Skills that do not fit these three categories are defined as “other technical and language skills.” We categorize the skills manually by assigning the 633 skills clusters to each skills category. When all skills within a cluster belong to one category, that cluster is assigned to the respective category. If the individual skills within a skill cluster cannot be assigned to one category, we assign each skill within the cluster to the appropriate category. 13 Skill clusters that are left unclassified by Burning Glass, as is the case for about a third of skills (2,592), are assigned manually, facilitated by a maximum likelihood method to estimate most likely skills clusters and cluster family. 14 About a third of the total skills in the Burning Glass Malaysia data are digital, cognitive, or socioemotional according to our classification. The 20 most frequently cited skills are a mix of cognitive (6), socioemotional (4), and other (7) skills, as well as three basic digital skills (computer literacy, Microsoft Office, and Excel) (Figure 3). We identify 2,622 unique digital skills: 24 are basic digital skills, 794 are intermediate digital skills, and 1,804 are advanced digital skills. Among the nondigital skills, 24 are cognitive and 29 are socioemotional. The remaining two-thirds of skills are 12 We acknowledge that the skills listed in vacancies may diverge from those actually employed by workers in firms; indeed, enterprise surveys from the region show large gaps between employers’ notional skills demands and workers’ supply of skills – digital and otherwise – in local labor markets (Cirera et al. 2021). 13 To assist with the categorization of digital skills, we also use the O*NET-based methodology for calculating digital occupation scores in Muro et al. (2017). However, we only use this measure as an initial categorization, followed by a manual review the categorization of every skill. 14 To infer the cluster and cluster family for the 2,592 skills that were not assigned to a cluster by Burning Glass, we implemented a nearest neighbor search of similar skills with available information to the skills with missing information. For each skill with missing information, we assign the most common cluster and cluster family of the top 30 skills similar to it. Similarity is defined as the cosine similarity of the vectors representing the skills. The skills are assigned a 100-dimensional vector by training a Word2Vec model where we represent the job postings as “documents” and the skills as the “words” within the document. The Word2Vec model learns semantic relationships between skills as they co-occur across job posts and embed this information into their respective vectors. 11 other technical skills and languages. Cognitive and socioemotional skills have a higher frequency in the database than digital skills, and basic digital skills appear more frequently than advanced digital skills. Figure 4 shows the frequencies of the top 20 skills in each category. Figure 3. Top 20 skills in the Malaysia’s job postings data set Source: Authors’ calculations based on World Bank-Burning Glass online job postings data in Malaysia (2016- 18). 12 Figure 4. Top skills across skills categories in Malaysia’s job postings data Number of postings for each skill A. Advanced digital skills (top 20) B. Intermediate digital skills (top 20) C. Basic digital skills (all 24) D. Cognitive skills (all 24) E. Socioemotional skills (all 29) F. Other skills (top 20) Source: Authors’ calculations based on WB-Burning Glass online job postings data in Malaysia (2016-18). 13 Measuring skills requirements by occupation We calculate two measures of skills requirements. The first measure is the skills count, that is the average number of times a skill from a given skill category appears in job postings mapped to a given occupation. We use this measure to explore complementarities among the nine subsets of skills categories (advanced, intermediate, and basic digital skills; three subsets each of cognitive and socioemotional skills and their subsets) within occupations. Our second skills measure is the average relative frequency with which a skill category appears in an occupation. The measure is calculated by averaging for each occupation the number of times a skill category appears in the job posting mapped to that occupation divided by the total count of skills in the job postings for that particular occupation (that is, expressed as a share of total skills rather than the raw skill count). 3.2. Creating occupational skills profiles and digital occupation levels Using the online job postings data to create occupational skills profiles We take two additional steps to create our final data set. First, we map each job posting classified with the 2013 Malaysia Standard Classification of Occupations (MASCO) and its related skills profile to the 2008 International Standard Classification of Occupations (ISCO-08; see ILO 2012), the international occupational classification system. 15 This gives us occupational skill profiles for the occupational codes in the employment data from other countries. Second, we impute the occupational skills profiles of 27 occupations. Among the 127 3-digit ISCO-08 occupations, the online job postings data did not include job postings for 16 occupations — primarily agricultural workers and street vendors — and 11 did not have sufficient job vacancy postings to generate a skills profile for that occupation. 16 These 27 occupations roughly represent between 20 and 40 percent of employment in the four countries studied. 17 Omitting them would distort the distribution of digital skills in a country. Instead, we impute their skills profiles by identifying similar 15 Job vacancies are mapped to ISCO codes at the 4-digit level. 16 We defined a threshold of at least 45 job postings per occupation. 17 The imputed occupations with employment shares exceeding 1 percent in any of the four countries (and the ISCO code) in the sample are: street and market salesperson (521), market gardeners and crop growers (611), animal producers (612), fisheries workers/hunters/trappers (622), subsistence crop farmers (631), subsistence livestock farmers (632), subsistence fishers/hunters/trappers/gatherers (634), street vendors (excluding food) (952). The full list of imputed occupations, and the occupations used to impute their skills profiles, are provided in Annex 1. 14 occupations in the job postings data and use the skills profiles of those occupations as a proxy for the skills profile of missing occupations. 18 Description of the occupational skills profiles The average number of, and types of, skills listed in (or imputed from) the job vacancy data differs by occupation. Occupations require an average of 6 skills, ranging from 2.2 to 12 skills (Table 4). Heavy truck and bus drivers require the fewest skills (2.2) while software and applications developers and analysts are at the other end of the distribution requiring 12 defined skills. On average, occupations require one digital skill, less than one cognitive and socioemotional skill each, and more than three other skills, which include languages and other technical skills, highlighting the high degree of job- specific skills. Table 4. Count and share of requirements for skills categories and their subsets across the 127 occupations Variable Mean Std dev Minimum Maximum Average number of skills Total 5.6 1.8 2.2 12.0 (count) Digital 1.0 1.1 0.0 7.6 Advanced 0.3 0.8 0.0 6.4 Intermediate 0.3 0.4 0.0 2.5 Basic 0.4 0.2 0.0 1.9 Cognitive 0.7 0.3 0.2 1.3 Thinking 0.1 0.1 0.0 0.3 Communication 0.3 0.1 0.1 0.6 Organization 0.3 0.1 0.1 0.6 Socioemotional 0.5 0.2 0.1 1.2 Emotions 0.0 0.0 0.0 0.2 Relationships 0.3 0.2 0.1 1.0 Personal growth 0.1 0.1 0.0 0.6 Others 3.4 0.9 1.5 6.5 Average share among total Digital 12.7 10.6 1.0 58.1 skills Advanced 3.2 6.4 0.0 47.5 Intermediate 3.9 4.8 0.0 28.5 Basic 5.6 5.2 0.3 51.0 Cognitive 10.9 2.9 3.6 18.5 Thinking 1.2 0.7 0.1 3.9 Communication 5.3 1.9 1.0 10.3 Organization 4.2 1.5 1.2 9.2 Socioemotional 8.3 5.0 2.1 42.2 Emotions 0.5 0.5 0.0 4.3 Relationships 5.8 4.5 1.3 38.2 Personal growth 2.0 1.5 0.3 8.6 Others 68.2 10.6 28.6 84.3 18 We use two sources to define similar occupations. ILO (2012) lists related occupations in ISCO-08. The O*NET Career Changers Matrix lists occupations in the US that have similar skills and experience and that workers could transfer between with minimal additional preparation (O*NET 2021). We only use these two sources to identify potential similar occupations. Once identified, the imputed occupations take the average share of skills requirements for each skills category of their similar occupations in the Malaysia data (or the skill profile of the similar occupation when the similar occupation is only one). 15 Source: Authors’ calculations based on World Bank-Burning Glass online Malaysia job postings data (2016-18). The means and ranges for the main skills categories (e.g. digital or cognitive) are higher than the statistics for the subsets (e.g. advanced or basic) since an occupation may require skills from more than one subset within a main skills category. The average relative frequency of digital, cognitive, and socioemotional skills is in the range of 8-13 percent. The relative frequency of digital, cognitive, and socioemotional skills is similar (8-13 percent) while nearly three quarters of skills are classified other skills. Basic digital skills have a higher relative frequency than intermediate or advanced digital skills, though occupations requiring intermediate or advanced digital skills will implicitly require basic digital skills, as well. When ranking occupations by their digital skill level, the ordering is similar to a ranking of these occupations based on the O*NET digital skill level, but with significant variance. In other words, our measure of the demand for digital skills in occupations generally tracks, but differs from, the O*NET measures. Thus, our estimates of the digital intensity of occupations may return results that are closer to the Malaysian – and perhaps the Southeast Asian – reality than the O*Net would have allowed. 19 Defining occupation digital levels We use cluster analysis to assign each occupation to a digital occupation level. The cluster analysis allows the data to endogenously sort the occupations into groups with similar digital skills profiles. Since cluster analysis is an unsupervised machine learning technique, informed selection of the variables must be done to increase the likelihood of the clustering model to generate meaningful groups. We use the 13 variables as input to the model. Five “raw” variables are included. The values for the “share of advanced, intermediate, and basic digital skills” (1) are the fundamental characteristics that encode an occupation’s digital level. The sum of these three, i.e., the “total share of digital skills" (2), provides a summary of how much digital skills is required by an occupation across levels of digital skills complexity. We also include three variables that represent the share of combinations of digital skills levels: advanced and intermediate digital skills share (3), advanced and basic digital skills share (4), and intermediate and basic digital skills share (5). Notably, all 127 occupations have at least 1 percent of their skills that are digital skills. Several derived variables are included to increase the complexity of the model. 20 In particular, the ratio of a level of the digital skill share (i.e., the ratio of advanced digital skills to the total share of digital skills) intuitively provides a quantitative measure of the intensity of the digital skill (e.g., advanced) 19 It was not possible to directly compare the digital scores and occupational ranking by digital scores using the Malaysia data to that used by O*Net due to differences in methodology for generating digital scores between the two data sets. 20 The complexity of the model, in this context, pertains to the additional information that is used to fit the model. For example, a regression model with 100 variables is more complex than another regression model that uses only 10 variables. 16 compared with the other digital skills (e.g., intermediate and basic). These variables will inherently discriminate between occupations having the same total share of digital skills, but their respective digital skill components are different. 21 We include six variables, each a ratio of advanced, intermediate, basic or a combination thereof. 22 The final two variables are rank variables. We expect that adding the variables corresponding to the “rank of the occupations with respect to the share of advanced digital skills” further reinforces the signal that we want the model to learn. Using the rank variables discourages the model from grouping together occupations that belong in opposing regimes of the share of advanced digital skills. This rationale holds true for the second variable, namely the “rank of the occupations with respect to the total share of digital skills.” We apply K-means clustering using the above 13 variables. The resulting five clusters of occupations can be ranked in their degree of digitalization by averaging the values of the “share of total digital skills” and “share of advanced digital skills” (Figure 5.A). We then analyze the resulting clusters to assess the quality of the groupings. Our analysis suggests that two of the five clusters generally share similar characteristics. To simplify the presentation and given the similarity of their digital skills, we combine the second and third clusters (numbers 2 and 3 in Figure 5.A) into a single cluster, resulting in four digital skills levels (Figure 5.B). 23 The full list of occupations assigned to each of the four levels, and the share of skills in each occupation that are advanced, intermediate, and basic digital skills, is provided in Annex 2. 21 For example, two occupations both have 50% total share of digital skills but the first has (40% Advanced, 5% Intermediate, 5% Basic) digital skills while the other has (5% Advanced, 5% Intermediate, 40% Basic) digital skills. 22 The variables are: (1) ratio of Advanced digital skills to the total share of digital skills, (2) ratio of Intermediate digital skills to the total share of digital skills, (3) ratio of Basic digital skills to the total share of digital skills, (4) ratio of the sum Advanced and Intermediate digital skills share to the total digital skills share, (5) ratio of the sum Advanced and Basic digital skills share to the total digital skills share, and (6) ratio of the sum Intermediate and Basic digital skills share to the total digital skills share. 23 The empirical results when using the four-cluster classification are similar to those when using the five-cluster classifications (Annex 7). 17 Figure 5. Occupations’ clusters and digital levels according to their total and high digital skills A. Share of total and advanced digital skills of occupations by the original five clusters B. Share of total digital and high digital skills of occupations by their digital level (final four clusters) Source: Authors’ calculations based on World Bank-Burning Glass online Malaysia job postings data (2016-18). 18 The “high digital” occupational cluster includes 13 occupations (Table 5). All require all levels of digital skills. All also have a high share of digital skills among total skills, ranging from 28 to 58 percent. The high-digital cluster includes, for example, software developers, a high-skilled occupation, and keyboard operators, a medium-skilled occupation. Both have a similar share of their skills that is digital (56-58 percent) but software developers mostly require advanced digital skills while keyboard operators mostly require basic digital skills. The bulk of these 13 occupations are consistent with the ISCO taxonomy of information technology occupations and most of these 13 occupations also have high scores on an index of digitalization created for occupations in the United States based on O*NET data (Muro et al. 2017). 24,25 Table 5. Digital skill and number of occupations of digital levels based on clusters Average share of digital skills (%) Cluster Digital Number of Total Advanced Intermediate Basic number level occupations 0 Very Low 4.8 0.8 1.0 2.9 38 1 8.1 2.1 1.7 4.2 20 Low 2 11.5 1.1 3.5 7.0 21 3 Medium 15.0 3.0 5.0 7.1 35 4 High 38.3 15.8 13.3 9.3 13 Source: Authors’ calculations based on World Bank-Burning Glass online Malaysia job postings data (2016-18). Notes: Darker red colors show higher percentage. Thirty-five occupations map to the “medium digital” occupation cluster. These occupations have low requirements for advanced digital skills, but a moderate demand for intermediate- and basic-digital skills. This category includes machine operators, technical education teachers, and office clerks, who regularly use specialized software to carry out their jobs. Most occupations can be classified as low digital. To provide some nuance to this group, we maintain the two clusters that emerged from the exercise and name them “low digital” and “very low digital” occupations. The exercise maps 41 occupations to the low-digital occupation category. These occupations require a moderate use of basic-digital skills. A broad range of occupations are in this category, including transport specialists (like Grab drivers who use digital phone apps to identify their next transport gig job), chief executives who may use MS Office for daily work, or salespeople who operate cash registers and inventory control software. The very low digital jobs include 38 24 Eight of the thirteen occupations are information technology occupations as defined by the ISCO occupation classification (ILO 2012, pp. 25-26). The two occupations we define as high digital that are not ICT occupations are mathematicians, actuaries, and statisticians (ISCO code 212) and engineering professionals (excluding electrotechnology) (ISCO code 214). 25 The occupation digital score is based on occupations’ three digitalization aspects as measured by O*Net for occupations in the US —computer & electronics, programming, and interacting with computer— following the methodology by Muro et al. (2017). 19 occupations. These occupations require few digital skills, and the digital skills that they do require are mostly basic digital skills. Cooks, who might use computers or smartphone for purchasing or inventory management, and Early Child Development teachers, who may use digital technologies to access or provide lessons, are examples of occupations in the very low digital occupations category. 26 3.3. Matching the occupational skills profiles with country employment data: The Southeast Asia Digital (SEAD) data set To analyze the skills profile of the employed population in the four countries we study, we create the Southeast Asia Digital (SEAD) data set that matches the occupation skills profiles to country employment data. We use the occupation variable to do the match with data from four Southeast Asian countries: the 2020 Cambodia Socioeconomic Survey and the 2017 labor force surveys in Malaysia, Thailand, and Vietnam. Each country data set is representative of the population at the national level, allowing us to quantify the number of people working in an occupation. As such, we have a data set of occupation skills profiles for 127 occupations and weights representing the occupations’ employment share in the four countries. 4. Methodology To explore the complementarity of digital and other skills and the levels of occupation digitalization, we use descriptive statistics, pairwise correlations, a factor analysis, and linear probability model (LPM) regressions. Similarity of skills required in very low-, low-, medium-, and highly digital occupations After an inspection of summary statistics, we apply an LPM to estimate the probability that a skill i in occupation o is found in a particular digital occupation group g. We use an LPM to estimate marginal effects of binary outcomes following Friedman (2012) and Bellemare (2015, 2018). The LPM is as follows: = + ∑ + (1) where yog is a binary variable representing occupation o dependent on the (digital occupation level) groups g of interest (e.g. medium- vs. low-digital occupations, high- vs. low-digital occupations, etc.). Doi is a dummy variable for skill i of occupation o taking the value of one if the skills count of a skills category of an occupation is above the mean. βoig is the coefficient for the groupings g quantifying the 26 An alternative methodology that uses a smaller set of variables to assign occupations to high, medium, and low digital categories resulted in a similar mapping. Results are available from the authors. 20 effect of the dummy variable for skill i in occupation o. α is a constant, and ε is the error term. The dummy variable for a given skill i of occupation o is derived as follows: � 1, ≥ = � (2) � 0, < � is the mean of xoi over all where xoi represents the average count of skill i in occupation o and occupations, where i is a vector of six skills subsets. Three of the skills subsets are cognitive skills: thinking, communication, and organization. The other three are socioemotional skills: emotions, relationships, and personal growth. The dummy variable is used for easier interpretation of the LPM’s coefficients than the raw skills count variables would allow. This transformation, in the context of LPM, allows us to interpret the coefficients as percentage point increase (or decrease) towards the probability of being a given occupation digital level if the raw skills count for the variable is above (or below) the mean of the skills count over all occupations. 27 We run the above analysis to be representative of the employed population of Malaysia (around 14 million of workers). Since we are interested in understanding the skills required by employers in a country-specific job market, we use the Malaysia subsample of the SEAD, with each occupation being weighted by the percentage of the employed population in it. 28 We carry out two additional analyses using different subsamples. First, we run the LPM with an unweighted sample to understand the correlation between skills within occupation, irrespective of the occupational composition of the workforce. Second, we carry out a similar analysis using the Cambodia, Thailand, and Vietnam subsamples. The results are similar to those from the Malaysia subsample so, in the interest of parsimoniousness, mostly present the results from the Malaysia subsample. 29 Results for the occupation skills profile (unweighted by the distribution of occupations in a particular country) and replications for other countries are presented in Annex 3. 27 Since the outcome is binary and we are using a probability linear model, we can interpret the coefficient of a dummy variable to be a percentage point contribution towards the probability of the outcome. For illustration, assume a PLM model with a constant value of zero. If we analyze one non-zero dummy variable and let all the other variables take a value of zero, then the coefficient of the non-zero dummy variable will equate to the probability of the outcome. Adding another non-zero dummy variable will linearly increase or decrease the probability according to the value of the coefficient. A regression using the raw skills count is available from the authors. 28 Studies on job skills demand and technological similarly combine data on job tasks at the occupational level from the O*NET database and individual-level employment survey that weight each occupation by the percentage of the employed population in each occupation (Autor, Levy, and Murnane 2003; Hardy, Keister, and Lewandowski 2018; Caunedo, Keller, and Shin 2021). 29 By construction, the variation between the results with occupational skills profile and those for the employed populations are only due to differences in the distribution of the employed population across occupations. 21 Complementarities between digital and non-digital (cognitive, and socioemotional) skills Next, we investigate complementarities between digital, cognitive, and socioemotional skills in Malaysia’s jobs. 30 We first examine pairwise correlations. The pairwise correlations show the magnitude of relationships between two subsets of skills. We use Pearson correlations and test the significance of the correlation through t-test. We then carry out a factor analysis using the skills count of the nine digital, cognitive, and socioemotional skills subsets (three each) detailed in section 2. The exploratory factor analysis allows us to model the relationships among skills subsets by reducing them to fewer variables thus highlighting stronger links between several skills subsets. We use a standard exploratory factor analysis that finds a number q of factors that linearly reconstruct our nine original variables according to the following relationship: = 1 + 2 + ⋯ + + (3) where is the value of the ith observation on the jth variable, is the ith observation on the kth common factor, is the set of linear coefficients called the factor loadings, and is similar to a residual but is known as the jth variable’s unique factor (StataCorps 2019). We use varimax rotation to produce orthogonal factors, 31 to get maximized factor loadings. Again, we use the SEAD data for Malaysia, weighted by the distribution of employment across occupations, to study those relationships within the existing job structure. Occupation-level results (unweighted by the distribution of occupations in a particular country) and replications for other countries are also estimated and presented in Annex 4. Occupation digitalization and skills requirements in Cambodia, Malaysia, Thailand, and Vietnam Finally, we use descriptive statistics to calculate the levels of occupation digitalization and skills requirements across the four countries we study. We use the SEAD data set to understand the distribution of skills across the four countries of interest. We also generate occupation-weighted data sets over several years for Malaysia, Thailand, and Vietnam to understand how the distribution of skills across each country evolves over a 4-6 year period. 32 30 We repeat this analysis for unweighted occupations, and weighting for the Cambodian, Thai, and Vietnamese occupational distributions. Again, results are similar to those when using the Malaysia-weighted data. 31 By definition, orthogonal factors are not correlated with each other. 32 We were unable to access multiple years of the Cambodian SES. 22 5. The Skills Profile of Very Low-, Low-, Medium-, and High Digital Occupations: The Case of Malaysia The level of the digital occupation is positively correlated with the share of digital skills required by employers in Malaysia. 33 Using simple summary statistics, we find that one-third of the skills required in Malaysia’s highly digital occupations are digital skills, while 16 percent of the skills required in Malaysia’s medium-digital occupations can be classified as digital skills (Figure 6, panel A). An even lower share of digital skills is required in low- and very-low-digital occupations (9 and 5 percent, respectively) in Malaysia. These trends are observed in the other three sample countries, with slight differences in point estimates (Annex Table A.3.1). First, Thailand’s highly digital occupations require a higher share of digital skills (37.1) than Malaysian highly digital occupations (33.3 percent). Second, Malaysia’s low digital occupations require the fewest digital skills among the four countries (9.4 percent) compared to 9.7 percent (Vietnam), 9.8 percent (Cambodia), and 10.1 percent (Malaysia). Highly digital occupations require the most digital skills across the three skill levels. The relative importance of advanced digital skills in Malaysia is much higher for highly digital occupations: 13 percent of their required skills are advanced digital as compared to 1-3 percent for very-low, low, and medium digital occupations (Figure 6, panel A). A similar trend and point estimates are observed in the other countries, suggesting robust proportionality in advanced digital skills across digital occupation categories. However, highly digital occupations also require a larger share of intermediate digital skills than medium digital occupations require (14 vs. 3.8 percent). 34 That might be because medium-digital occupations, while requiring some digital skills, mostly require skills in a job specialty (e.g. finance or marketing). This trend emerges in the other countries (Annex Table A.3.1). Similarly, the share of basic digital skills is less demanded for very-low and low-digital occupations than medium digital occupations in Malaysia (Figure 6, panel A). 35 This reflects that while basic digital skills are a required in those occupations, they mostly require other skills. 33 These results deviate from those in Table 5, which presents the average share of digital skills – total and by level – for each of the four digital occupation levels, due to the weighting of occupations by country. Table 5 assigns an equal weight to each occupation with the country-specific estimates use country weights. Thus, occupations with certain skills mixes, even within digital occupation level, may play a more prominent role in the country averages than in the unweighted average. 34 Cambodia is the only country where intermediate digital skills play a larger role than advanced digital skills in highly digital jobs. Annex Table A.3.1 shows that 17.5 percent of skills in Cambodia’s highly digital occupations are classified as intermediate digital skills, as compared to 5.2 percent in medium occupations. At the same time, only 8 percent of advanced digital skills are required in Cambodian highly digital occupations. Thus, Cambodia’s 13 highly digital occupations require more intermediate, rather than advanced, digital skills. 35 Basic digital skills are also more demanded in highly digital occupations in Cambodia, Vietnam, and especially Thailand (8.8 percent). 23 Figure 6. Relative importance of skills categories across digital occupation levels in Malaysia A. Composition of skills categories across occupation digital levels B. Composition of skills categories and subsets across occupation digital levels Source: Authors’ calculations based on SEAD data. Notes: The share of skills splits between digital, cognitive, and socioemotional, and other skills. Digital skills levels and cognitive and socioemotional skills subsets are subcategories counted in the total. All four digital occupation levels in Malaysia require cognitive and socioemotional skills that are relatively similar in share and type. About 8-12 percent of skills required in all four digital occupation levels are classified as cognitive skills, though Malaysia’s highly digital occupations are those that rely less on cognitive skills as compared to other occupations in the category. There is slightly more variance in the share of skills defined as socioemotional: while 6 percent of skills defined in highly digital occupations are socioemotional, 12 percent of skills required by very-low-digital occupations are socioemotional (Figure 6, panel A and B), as compared to 5-8 percent of skills for the other digital occupation levels. Breaking down by skills subsets suggests some small differences within categories. For example, relationship skills (a socioemotional skill) are a few percentage points more expected in 24 very-low-digital occupations than in highly digital occupations (Figure 6, panel B). These trends also emerge for the other three sample countries with a few exceptions (Annex Table A.3.1). 36 The similar requirement for cognitive and socioemotional skills across digital occupation levels in Malaysia is confirmed when controlling for potential correlations between variables. The regression analysis finds that the required subsets of cognitive and socioemotional skills are quite similar across very-low, low-, medium-, and highly digital occupations. 37 The most notable exception is that, on average, occupations that require more thinking skills (a subset of cognitive skills) are 209 percentage points more likely to be a highly digital occupation relative to a very low-digital occupation (Table 6). This correlation was strongly identified for the other three countries, as well. A weak positive correlation is also found between thinking skills and low (versus very low) and high (versus low) digital occupations in all countries except Thailand (Annex Table A.3.2). While more emotional skills (a subset of socioemotional skills) are expected from medium-digital occupations compared to low digital occupations in Malaysia, the size and imprecision of the estimates for the other occupation types do not reveal more general patterns. However, this relationship emerged with more precise estimates for the other three countries. Relationship skills are weakly negatively correlated with low (versus very low) and high (versus low) digital occupations, though correlations are stronger for Cambodia. A few other weak correlations emerge, but most subsets are not statistically different across occupation levels at the 5 percent significance level or below, meaning that there is no discernable difference in demand for those skills across digital occupation levels. 38 36 Namely, relationship skills are less important for very low digital occupations in the other three countries, as compared to Malaysia, while personal growth is more prevalent among highly digital occupations. 37 These associations should not be interpreted as causal but rather as conditional correlations for at least two reasons. First, there are factors beyond cognitive and socioemotional skills that could influence the digital occupation level. Second, subsets of cognitive and socioemotional skills may be correlated with each other to some degrees, especially those in the cognitive subsets, and thus may bias the conditional correlations with digital occupation levels (see Table 7). 38 Similar results emerge from LPM estimates using the unweighted occupational skills profile sample and the country- specific weighted occupations for Vietnam and Thailand. The communications, relationships, and personal growth variables play a greater explanatory role in the Cambodia -and, to some extent, the Vietnam - estimates, with the coefficient signs and magnitudes were similar to the Malaysia-weighted sample (Annex Table A.3.2). 25 Table 6. Conditional correlations between digital occupation level and cognitive and socioemotional skills requirements in Malaysia LPM regressions of digital occupation levels on dummies taking the value of one if the skills count of a skills category of an occupation is above the mean Occupation digital level Low vs. High vs. Medium High vs. High vs. Very Very vs. Low Medium Low low low (1) (2) (3) (4) (5) Cognitive Thinking 0.36* 0.22 0.01 0.53* 2.09*** (0.17) (0.13) (0.15) (0.23) (0.40) Communication 0.47* -0.19 0.01 -0.17 -0.09 (0.21) (0.14) (0.33) (0.24) (0.16) Organization 0.00 -0.16 0.12 -0.13 -0.60 (0.16) (0.15) (0.10) (0.14) (0.41) Socioemotional Emotions -0.00 0.48** -0.30 0.12 0.81 (0.15) (0.15) (0.21) (0.19) (0.41) Relationships -0.30* -0.35 0.08 -0.58* -0.67 (0.14) (0.18) (0.38) (0.24) (0.43) Personal growth -0.16 -0.03 0.36* 0.34 -0.55 (0.16) (0.17) (0.18) (0.20) (0.41) Constant 0.44** 1.47*** 2.22*** 1.39*** 0.74 (0.14) (0.14) (0.22) (0.33) (0.54) N 79 75 47 54 51 R-sq 0.36 0.36 0.13 0.15 0.44 Source: Authors’ calculations based SEAD data. Note: The regressors for the outcome are dummy variables taking the value of one if the skills count of a skills category of an occupation is above the mean. The 2017 Malaysian labor force survey does not report any workers in one of the 127 occupations from our occupational skills profile (paramedical practitioners, ISCO code 224). *p < 0.1 **p < 0.05 ***p < 0.01 6. Complementarities among Digital, Cognitive, and Socioemotional Skills: The Case of Malaysia Cognitive and socioemotional skills correlate more strongly with basic digital skills than with intermediate and advanced digital skills in Malaysia. Basic digital skills correlate highly with all three subsets of cognitive skills and with one socioemotional skill in Malaysia. The correlation coefficient between basic digital skills and thinking, communication, organization (cognitive), and emotions (socioemotional) is around 0.5 (Table 7). Basic digital skills are not correlated with relationships or personal growth in Malaysia. In contrast, intermediate digital skills correlate with thinking and 26 personal growth, around 0.4, and to a lesser extent with organization and emotions, around 0.3. 39 By contrast, advanced digital skills only meaningfully correlate with thinking skills (0.5) in Malaysia. In other words, thinking skills strongly correlate with all levels of digital skills while complementarities with other socioemotional and cognitive skills are particular to the level of digital skills. The correlation between advanced digital skills and other cognitive skills is positive, but weak, while the correlation is close to zero for most socioemotional skills in Malaysia. The findings are robust across the other country samples and the unweighted occupational sample, though the correlation point estimates are higher in these other country samples than in Malaysia (Annex Table A.4.1). Table 7. Correlations between skills in Malaysia Pairwise correlations based on the number of skills counts of skills categories Source: Authors’ calculations based SEAD data. Notes: Darker red and blue colors show higher and lower magnitudes, respectively. *p < 0.1 **p < 0.05 ***p < 0.01 The Malaysia findings are consistent with evidence from online job postings for professional occupations in the U.S. Deming and Kahn (2018) find that basic digital skills correlate more highly with cognitive and socioemotional skills than do advanced digital skills. Basic digital skills (computer literacy and office programs) have a correlation with skills categories related to cognitive and socioemotional skills of 0.5 while advanced digital skills (programming and specialized software) have also positive but lower correlations around 0.2. While cognitive skills subsets highly correlate among themselves in Malaysia, this is less the case for digital skills subsets, and almost nonexistent among socioemotional skills subsets. The three cognitive subsets have high correlation coefficients of around 0.7 for Malaysia, which is consistent with theories and empirical work showing that different specific cognitive skills tend to be highly correlated 39 In Cambodia, organization and emotions are more strongly correlated with intermediate digital skills than thinking and personal growth (Annex Table A.4.4). 27 (Almlund et al. 2011) (Table 7). The three socioemotional skills, by contrast, correlate little between each other in Malaysia, except for personal growth and relationship skills that correlate around 0.3, reflecting that they mostly capture different facets of socioemotional skills. The correlations between the digital skills subsets are modest for Malaysia but significant between basic and intermediate (0.3), much higher between intermediate and advanced (0.5). The correlation is close to zero between basic and advanced digital skills in Malaysia. 40 This may be due to employers who wish to hire workers with advanced digital skills assume that the potential hires already have basic digital skills, so they do not mention the basic skills in the job postings. The last correlation is lower than estimated for professional occupations in the U.S. that have a correlation of 0.2 between their basic and advanced digital skills (Deming and Kahn 2018). An alternative methodology to consider not only pairwise relationships between skills subsets but possible systematic association between more subsets shows stronger complementarities. 41 Exploratory factor analysis finds that three groups of skills break out across the 9 digital, cognitive, and socioemotional subsets for Malaysia. The first factor is a mix of the three cognitive skills subsets, relationship skills (socioemotional) and basic digital skills (Table 8). It appears to be a factor of less digital skills for a broad range of occupations in Malaysia. The occupations with more of this skills mix are low- and medium digital occupations, such as government professionals and plant and machine operators (Annex Table A.5.1). The second factor is based on intermediate and advanced digital skills and to a lesser extent thinking skills (cognitive) and personal growth (socioemotional). 42 The typical occupations in Malaysia requiring a high weight on these kinds of skills are highly digital occupations, such as software developers and network professionals. The third factor heavily loads on the socioemotional skills subsets of emotions and personal growth. 43 Typical Malaysian occupations using this skill set include occupations from all three levels of occupation digitalization, such as architects and primary school teachers. Similar factors emerge for the other three countries. 44 40 The other three countries and the unweighted occupation profile data find similar results (Annex Table A.4.1). 41 Very similar results are found when not rotating factors and using oblique rotations (i.e. allowing for correlations across factors) (results available upon request). 42 The socioemotional skills that weighed most heavily in this factor in Vietnam was organization and in Thailand was relationship skills (Annex Table A.4.2). 43 The third factor is the least stable across the alternative subsamples. For example, in Cambodia, it only weighs heavily on advanced digital skills. In Thailand, the highest factor loadings are on personal growth and a lower loading on intermediate digital skills, though very low basic skills, as in the Malaysia-weighted sample. The Vietnam-weighted sample also weighs very heavily on personal growth and very low on all other skills (Annex Table A.4.2). 44 A notable exception is that an “advanced digital” factor emerges for Cambodia. No other skills weight positively heavily in this factor while socioemotional skills are particularly negative (Annex Table A.4.2). 28 Table 8. Exploratory factor analysis of skills in Malaysia Exploratory factor analysis based on skills counts of skills categories B. Factors with heavy loadings on A. Factor loadings and uniqueness specific skills Sources: Authors’ calculations based on SEAD data. Note: The factor analysis is based on varimax rotation for orthogonal factors. Darker red and blue colors show higher and lower magnitudes, respectively. The factor analysis thus confirms that in Malaysia, cognitive and socioemotional skills tend to associate more with basic digital skills than with intermediate and advanced digital skills, with some differences in the pairwise correlations. Both approaches confirm that basic digital skills are strongly associated with cognitive and socioemotional skills subsets. However, the two methods find different results in the kinds of socioemotional skills that associate with basic digital skills. For example, basic digital skills load highly in a factor with high relationship skills factor loadings, while their pairwise correlation is modest. Similarly, in Malaysia, basic digital skills have a low loading in factors where emotional skills load high, while both strongly correlate. 45 Some disconnects are observed in cognitive skills, as well. For example, thinking skills (cognitive) load lightly in the factor of intermediate and advanced digital skills but they are strongly correlated and highly significant Pearson correlations. The lower association between cognitive and socioemotional skills, on one hand, and intermediate and advanced digital skills, on the other hand, may be due to the lower prevalence of the latter in Malaysia (and the other sample countries). Indeed, since basic digital skills are widespread, they associate with cognitive and socioemotional skills, which are also widespread. This is illustrated by the fact that, when only considering the occupation skills profile, i.e. not projected to Malaysia’s employed population, advanced digital skills load highly with thinking skills and intermediate digital skills with personal growth (Annex Table A.4.2). 45 Emotions and basic digital skills load highly in the same factors in the other three countries (Annex Table A.4.2). 29 7. Levels of Occupation Digitalization and Skills Requirements in Cambodia, Malaysia, Thailand, and Vietnam In the four countries studied, between 40 and 66 percent of the employed population are in very low digital occupations, 20-31 percent in low digital occupation, 13-24 percent in medium-digital occupations, and 1-5 percent in highly digital occupations (Figure 7). 46 Malaysia, the country with the highest GDP per capita of the four countries, also has the highest share of employment in medium- and highly digital occupations (24 and 5 percent, respectively) (Figures 7 and 8). Thailand and Vietnam are relatively similar in their distribution of employment across digital occupation groups. Cambodia, the country with lowest GDP per capita of the four countries, has only 0.8 percent of its employed population in highly digital occupations. While methodologies for classifying the digitization of occupations differ across studies, Muro et al. (2017) also find a low proportion of employed people in highly digital occupations in the US in 2002: 56 percent in low-digital occupations, 40 percent in medium-digital, and 5 percent in high digital. By 2016, the distribution in the US became 30 – 48 – 23 percent, in low-, medium-, and highly digital occupations, respectively. Figure 7. Employment across digital occupation level, by country Source: Authors’ calculations based SEAD data. 46 Annex Table A.6.1. shows the employment share of the occupation in each country and its digital level. 30 Figure 8. Share of intermediate and highly digital occupations by country’s GDP per capita A. Share of high digital occupations and GDP per capita B. Share of medium digital occupations and GDP per capita Source: Authors’ calculations based SEAD data and World Bank’s World Development Indicators. Note: the size of the bubbles represents the employed population in millions: 9.3 for Cambodia, 14.0 for Malaysia, 36.6 for Thailand, and 52.4 for Vietnam. Digital, cognitive, and socioemotional skills, as aggregate categories 47 have roughly the same relative importance, around 8-10 percent of skills (Figure 9), however the average share of digital skills differs slightly across countries. Malaysia has a higher average share of digital skills; ten percent of Malaysian workers’ skills are digital while eight percent are digital in the other three countries. This suggests that 47 Namely, all skills types and not disaggregating by digital occupation level to allow for cross-country comparison. 31 a higher share of Malaysia’s workforce is engaged in occupations that require at least some digital skills. In all four countries, the share of jobs requiring basic digital skills is greater than the share requiring intermediate digital skills, which, in turn, is greater than the share requiring advanced digital skills. Given that digital skills are likely cumulative, the share of jobs requiring basic digital skills is likely better represented by the “total digital skills” category than the “basic digital skills” category. The demand for cognitive skills is very similar across countries. The bulk of jobs require the subsets communication (oral and verbal communication) and organization (planning, multitasking, time management, etc.) (4-5 percent). The third cognitive skill — thinking skills (problem solving, critical thinking, decision making, etc.) — is required by less than 1 percent of jobs in each country. Socioemotional skills, though a little less demanded, still show a strong presence. The most requested subset of socioemotional skills is relationship skills (teamwork, management, leadership, etc.) around 5-7 percent, twice the showing of personal growth skills (creativity, initiative, self-motivation, etc.), and well above emotional skills (coping strategy, detail-orientation, diligence, etc.). The average share of socioemotional skills and their subsets also shows minor differences across countries. The fact that the shares of cognitive and socioemotional skills look so similar across countries despite differences in their employment distribution across occupations may reflect that both cognitive and socioemotional skills are required in most types of jobs. Figure 9. Relative importance of skills categories across countries Average share of a skill category among total skills, by country Source: Authors’ calculations based SEAD data. Note: Since occupational skills profiles are derived from the same source (World Bank-Burning Glass data set of job postings), cross-country differences are due to the distribution of employment across occupations of each country. The distribution can be seen at 1-digit level in figure 2 and at 3-digit level in Annex Table A.6.1. 32 A short time series finds a slow transition to digital occupations in Southeast Asia, with only Vietnam showing a trend toward greater digitalization. When expanding the sample to annual labor force surveys in Malaysia (2011-17), Thailand (2014-20), and Vietnam (2014-15, 2017-18), we find that any changes in the occupational structure of labor markets in Malaysia and Thailand did not affect the distribution of the workforce across digital skills levels (Figure 10). In contrast, the data from Vietnam finds a decline in very low digital occupations (from 70 percent to 64 percent over the five-year period) and a 3-percentage point increase in low- and medium digital occupations. The micro-data explain these changes, where the share of workers in occupations related to subsistence agriculture — very low digital occupations — declined significantly over the period with slow increases in craft occupations (assemblers, construction, machine operators) and services (shop salespeople) which are classified as low or medium digital occupations. 33 Figure 10. Employment across digital occupation level, by year and country, in Malaysia, Thailand, and Vietnam A. Malaysia (2011-17) B. Thailand (2014-20) C. Vietnam (2014-2018) Source: Authors’ calculations based SEAD data matched with and employment surveys of Malaysia, Thailand, and Vietnam for corresponding years. 34 8. Conclusion Our study confirms that digital skills are common in Southeast Asia. The job postings data from Malaysia in 2016-18 reveals that all occupations require at least some digital skills. Highly digital occupations, however, still represent a very small share of employment in the four countries studied: 5 percent in Malaysia, around 2 percent in Thailand and Vietnam, and 1 percent in Cambodia. Instead, the bulk of employment is in very-low-digital occupations (from 40 percent in Malaysia to 66 percent in Cambodia) and low-digital occupations (from 16 percent in Vietnam to 31 percent in Malaysia). These shares are changing very slowly over time, with the exception of Vietnam’s somewhat rapid shedding of very low digital occupations in favor of low and medium digital occupations We find that occupations from all four digital occupations levels need cognitive and socioemotional skills. While the share of intermediate and advanced digital skills differs across digital occupation levels in Malaysia, all four levels of occupations require roughly the same share of basic digital skills (5-9 percent), cognitive skills (8-12 percent), and socioemotional skills (6-12 percent). Cognitive and socioemotional skills are even more important than digital skills for very-low, low-, and medium- digital occupations (nearly the entire employed population). The share of cognitive and socioemotional skills combined is greater than digital skills: 23 versus 5 percent for very-low digital occupations, respectively, 20 versus 9 percent for low-digital occupations, respectively, and 16 versus 15 percent for medium-digital occupations, respectively. Conditional correlations estimated through regressions show that medium- and highly digital occupations require more thinking skills (cognitive) than very low digital occupations but require similar levels of the other cognitive and socioemotional subsets as do occupations that are less digitally intensive. Similar patterns emerge for Vietnam, Cambodia, Thailand. When disaggregating cognitive and socioemotional skills, subsets of these skills categories are found to associate more closely with certain levels of digital skills. Using an exploratory factor analysis for Malaysia, we find three clusters of skills from nine digital, cognitive, and socioemotional subsets of skills. Two of these factors are a mix of digital skills and subsets of cognitive and socioemotional skills: 1) basic digital skills with the three subsets of cognitive skills (thinking, communication, and organization), and socioemotional skills related to relationships; and 2) intermediate and advanced digital skills with (to a lesser extent) thinking skills and personal growth. The third cluster does not include digital skills, instead being limited to socioemotional skills, namely with emotions and personal growth. This shows that basic digital skills can be considered a generic skill as much as 35 cognitive and socioemotional and that digital skills are associated to some degree with cognitive and socioemotional skills. Pairwise correlations among those subsets confirm those patterns. Similar patterns emerge for Vietnam, Cambodia, and Thailand. Our estimations have three main limitations. First, we proxy a skills profile for 20 percent of occupations that are not sufficiently listed in the job postings data set used to create a skills profile. These occupations represent a sizeable share of employment in the four studied countries: from 14 percent in Malaysia to 43 percent in Cambodia. We address this shortcoming by imputing skills profiles for these occupations based on similar occupations. Second, we derive skills use from requirements in job postings, which may omit skills that employers implicitly require rather than explicitly request. Employers do not specify the level of skills proficiency or use, so we equally weight all skills listed in the job posting. As a result, our skills profiles may not be complete, particularly omitting lower-level skills or under- (or over-) weighting the importance of certain skills. Third, we may be over-estimating the degree of digital-ness of the 127 occupations. Since we draw our skills data from on-line jobs vacancy postings, the jobs within each of the 127 occupations that are listed on the internet – and used to construct a skills profile that is a proxy for all jobs within an occupation – may be those that are more inclined to use digital technologies. Overall, our findings show that all occupations require digital, cognitive, and socioemotional skills, suggesting that policy makers may need to reform the “basic package” provided by education systems in response to digital technologies in the workplace. Education system still need to teach technical and language skills, since those are in very high demand by employers. However, they may also want to rethink the “basic package” of education, namely, to continue teaching cognitive skills, increasing emphasis on socioemotional skills, and adding basic digital skills to the general curriculum. Learning to learn will become increasingly important as new technologies enter the workplace and workers need to adapt to them. Finally, while advanced digital skills are increasingly emphasized in policy circles, they are only needed for a very small subset of jobs. The growth in demand for these jobs is growing very slowly in Southeast Asia, as comparison to the US. Even as economies grow into the next level of development, advanced skills will still be a minority of the skillset employers value. So, while advanced digital training may be a goal, the reality of the evolving landscape of jobs and technologies is more modest with a continued demand for the traditional cognitive, the increasingly explicit socioemotional, and the newest basic and intermediate digital skills. 36 9. References Acemoglu, D., and D. H. Autor. 2011. “Skills, Tasks and Technologies: Implications for Employment and Earnings.” In O. Ashenfelter and D. E. Card (Eds.) Handbook of Labor Economics, Volume 4: 1043–1171. Amsterdam: Elsevier. Aedo, C., J. Hentschel, J. Luque, and M. Moreno. 2013. “From Occupations to Embedded Skills: A Cross-Country Comparison.” World Bank Policy Research Working Paper 6560. Washington, DC: World Bank. Almlund, M., A. L. Duckworth, J. J. Heckman, and T. Kautz. 2011. “Personality Psychology and Economics.” In E. A. Hanushek (Ed.). Handbook of the Economics of Education, Vol. 4, 1–181. Amsterdam: North Holland. Apedo-Amah, M. C., B. Avdiu, X. Cirera, M. Cruz, E. Davies, A. Grover, L. Iacovone, U. Kilinc, D. Medvedev, F. O. Maduko, S. Poupakis, J. Torres, and T. T. Tran. 2021. “Unmasking the Impact of COVID-19 on Businesses: Firm Level Evidence from Across the World.” World Bank Policy Research Working Paper 9434. Washington, DC: World Bank. Apella, I., and G. Zunino. 2018. “Nonstandard Forms of Employment in Developing Countries: A Study for a Set of Selected Countries in Latin America and the Caribbean and Europe and Central Asia.” World Bank Policy Research Working Paper 8581. Washington, DC: World Bank. Autor, D. H., F. Levy, and R. Murnane. 2003. “The Skill Content of Recent Technological Change: An Empirical Exploration.” Quarterly Journal of Economics 118(4): 1279–1333. Bellemare, M. F. 2015. “A Rant on Estimation with Binary Dependent Variables.” Personal blog. ______. 2018. “Causal Inference with Observational Data: Nonstandard Dependent Variables.” Lecture. University of Copenhagen, Denmark. Burning Glass Technologies. 2019. “Mapping the Genome of Jobs: The Burning Glass Skills Taxonomy.” White Paper. Boston, MA: Burning Glass Technologies. Cammeraat, E., and M. Squicciarini. 2021. “Burning Glass Technologies’ data use in policy-relevant analysis: An occupation-level assessment.” OECD Science, Technology and Industry Working Papers 2021/05. Paris: OECD. Caunedo, J., E. Keller, and Y. Shin. 2021. “Technology and the Task Content of Jobs across the Development Spectrum.” NBER Working Paper 28681. Cambridge, MA: National Bureau of Economic Research. CSC (Malaysia’s Critical Skills Monitoring Committee). 2019. Critical Occupations List 2018/2019: Technical Report. Kuala Lumpur: TalentCorp. Available at: https://bit.ly/3uZ6G4E Cirera, X., A. D. Mason, F. de Nicola, S. Kuriakose, D. S. Mare, and T. T. Tran. 2021. The Innovation Imperative for Developing East Asia. World Bank East Asia and Pacific Regional Report. Washington, DC: World Bank. Cunningham, W. 2018. “Meeting the Jobs Challenge in Southeast Asia: Lessons from Cambodia, Myanmar, and Vietnam” in World Bank (eds). Navigating Uncertainty: East Asia and Pacific Economic Update. World Bank East Asia and Pacific Regional Report. Washington, DC: World Bank. 37 Cunningham, W., and P. Villaseñor. 2016. “Employer Voices, Employer Demands, and Implications for Public Skills Development Policy Connecting the Labor and Education Sectors.” World Bank Research Observer 31 (1): 102–34. DE4A (Digital Economy for Africa). 2021. Digital Skills: The Why, the What, and the How - Methodological Guidebook for Preparing Digital Skills Country Action Plans for Higher Education and TVET (V2.0). Washington, DC: World Bank. Deming, D. J. 2017. “The Growing Importance of Social Skills in the Labor Market.” Quarterly Journal of Economics 132(4): 1593–1640. Deming, D. J., and L. B. Kahn. 2018. “Skill Requirements across Firms and Labor Markets: Evidence from Job Postings for Professionals.” Journal of Labor Economics 36(1): 337–69. Deming, D. J., and K. Noray. 2020. “Earnings Dynamics, Changing Job Skills, and STEM Careers.” Quarterly Journal of Economics 135(4): 1965–2005. DeStefano and Timmis forthcoming. “Firm Digital Adoption during COVID-19.” Manuscript. Discussed in World Bank (2021, pp. 28–29). Friedman, J. 2012. “Whether to probit or to probe it: in defense of the Linear Probability Model.” Development Impact Blog. Washington, DC: World Bank. Goos, M., Manning, A., Salomons, A. 2014. “Explaining Job Polarization: Routine-Biased Technological Change and Offshoring.” American Economic Review: Papers & Proceedings 99(2): 58–63. Grundke, R., L. Marcolin, T. L. B. Nguyen, and M. Squicciarini. 2018. “Which skills for the digital era?: Returns to skills analysis.” OECD Science, Technology and Industry Working Papers 2018/09. Paris: OECD. Guerra, N., K. Modecki, and W. Cunningham. 2014. “Social-Emotional Skills Development across the Life Span: PRACTICE.” World Bank Policy Research Working Paper 7123. Washington, DC: World Bank. Hardy W., R. Keister R., and P. Lewandowski. 2018. “Educational upgrading, structural change and the task content of jobs in Europe.” Economics of Transition 26(2): 201–231. IFC (International Finance Corporation). 2019. Digital skills in sub-Saharan Africa: Spotlight on Ghana. Washington, DC: IFC. ILO (International Labour Organization). 2012. International Standard Classification of Occupations 2008. Volume 1: Structure, group definitions and correspondence tables. Geneva: ILO. ITU (International Telecommunication Union). 2020. Digital Skills Assessment Guidebook. Geneva: ITU. Lewandowski, P., A. Park, W. Hardy, and Y. Du. 2021. “Technology, Skills, and Globalization: Explaining International Differences in Routine and Nonroutine Work Using Survey Data. World Bank Economic Review (forthcoming). Lo Bello, S., M. L. Sánchez-Puerta, and H. Winkler. 2019. “From Ghana to America: The Skill Content of Jobs and Economic Development.” World Bank Policy Research Working Paper 8758, Washington, DC: World Bank. 38 Marinescu, I. E., and R. Wolthoff. 2020. “Opening the Black Box of the Matching Function: the Power of Words.” Journal of Labor Economics 38(2): 535–68. Muller, N., and A. Safir. 2019. “What Employers Actually Want: Skills in demand in online job vacancies in Ukraine.” World Bank Social Protection & Jobs Discussion Paper 1932. Washington, DC: World Bank. Muro, M., S. Liu, J. Whiton, and S. Kulkarni. 2017. Digitalization and the American Workforce. Washington, DC: Brookings. OECD (Organisation for Economic Co-operation and Development). 2016. “Skills for a Digital World: 2016 Ministerial Meeting on the Digital Economy Background Report.” OECD Digital Economy Papers 250. Paris: OECD. ______. 2019. OECD Skills Outlook 2019: Thriving in a Digital World. Paris: OECD. O*NET (Occupational Information Network). 2021. Career Changers Matrix version 20.3. Excel file available at: https://bit.ly/3bTfjXh Pasquier-Doumer, L., X Oudin, and V. Nguyen Thang. 2017. The Importance of Household Businesses and the Informal Sector for Inclusive Growth in Vietnam. Vietnam Academy of Social Sciences and the French National Research Institute for Sustainable Development. Saltiel, F. 2020. “Who Can Work from Home in Developing Countries?” Covid Economics 6. London: Centre for Economic Policy Research (CEPR). Squicciarini, M., and H. Nachtigall, 2021. “Demand for AI skills in jobs: Evidence from online job postings.” OECD Science, Technology and Industry Working Papers 2021/03. Paris: OECD. StataCorp. 2019. Stata 16 Base Reference Manual. College Station, TX: Stata Press. UNESCO (United Nations Educational, Scientific and Cultural Organization). 2017. Digital Skills for Life and Work. Report by the Broadband Commission for Sustainable Development. Paris: UNESCO. World Bank. 2018. World Development Report 2018: Learning to Realize Education’s Promise. Washington, DC: World Bank. ______. 2021. East Asia and Pacific Economic Update, April 2021: Uneven Recovery. Washington, DC: World Bank. Weinberger, C. J. 2014. “The Increasing Complementarity Between Cognitive and Social Skills.” Review of Economics and Statistics 96: 849–861. 39 Annex 1. Imputations of skills requirements for occupations unrepresented by job postings data Table A.1.1. Employment share of the 27 imputed occupations Employment share (%) Occupation # Occupation name Cambodia Malaysia Thailand Vietnam code 1 Legislators and senior officials 111 0.5 0.1 0.7 0.4 2 Traditional and complementary medicine professionals 223 0.0 0.0 0.0 0.0 3 Traditional and complementary medicine associate professionals 323 0.0 0.1 0.0 0.0 4 Veterinary technicians and assistants 324 0.0 0.0 0.0 0.0 5 Secretaries (general) 412 0.0 0.1 0.0 0.0 6 Street and market salespersons 521 4.8 4.2 3.5 7.1 7 Market gardeners and crop growers 611 9.6 5.1 21.7 4.8 8 Animal producers 612 6.8 0.2 2.5 1.8 9 Mixed crop and animal producers 613 0.0 0.0 0.2 0.0 10 Forestry and related workers 621 0.4 0.0 0.2 0.0 11 Fishery workers, hunters and trappers 622 1.3 0.7 0.8 1.8 12 Subsistence crop farmers 631 12.2 0.0 3.1 1.1 13 Subsistence livestock farmers 632 3.5 0.0 0.0 0.1 14 Subsistence mixed crop and livestock farmers 633 0.0 0.0 0.0 0.1 15 Subsistence fishers, hunters, trappers and gatherers 634 1.1 0.2 0.2 0.0 16 Wood treaters, cabinet-makers and related trades workers 752 0.9 0.5 0.3 0.8 17 Other craft and related workers 754 0.0 0.1 0.5 0.5 18 Metal processing and finishing plant operators 812 0.0 0.2 0.3 0.2 19 Rubber, plastic and paper products machine operators 814 0.1 0.6 0.6 0.3 20 Food and related products machine operators 816 0.3 0.3 0.5 0.2 21 Wood processing and papermaking plant operators 817 0.0 0.5 0.1 0.2 22 Locomotive engine drivers and related workers 831 0.0 0.0 0.0 0.0 23 Ships' deck crews and related workers 835 0.0 0.1 0.1 0.2 24 Vehicle, window, laundry and other hand cleaning workers 912 0.7 0.2 0.4 0.1 25 Street and related service workers 951 0.0 0.0 0.0 0.1 26 Street vendors (excluding food) 952 0.4 0.3 1.1 0.2 27 Refuse workers 961 0.4 0.5 0.3 0.3 Total 43.0 14.0 37.1 20.3 Sources: Employment shares are from Cambodia 2020 Socioeconomic Survey, and 2017 employment surveys of Malaysia, Thailand, and Vietnam. 40 Table A.1.2. Similar occupations used for imputations Code of occupations used for # Occupation name Code imputations 112, 121, 122, 134, 141, 142, 143, 232, 1 Legislators and senior officials 111 242, 243, 261, 331, 332, 333, 335, 422, 522, 541 2 Traditional, and, complementary, medicine, professionals 223 222, 224, 225, 226 3 Traditional, and, complementary, medicine, associate, professionals 323 321, 322, 325 4 Veterinary, technicians, and, assistants 324 321, 324, 325, 516, 532 5 Secretaries, (general) 412 334, 421, 422, 441 6 Street, and, market, salespersons 521 7 Street, and, related, service, workers 951 512, 522, 524, 911, 941, 962 8 Street, vendors, (excluding, food) 952 9 Market, gardeners, and, crop, growers 611 10 Animal, producers 612 11 Mixed, crop, and, animal, producers 613 12 Forestry, and, related, workers 621 13 Fishery, workers,, hunters, and, trappers 622 921 14 Subsistence, crop, farmers 631 15 Subsistence, livestock, farmers 632 16 Subsistence, mixed, crop, and, livestock, farmers 633 17 Subsistence, fishers,, hunters,, trappers, and, gatherers 634 313, 711, 722, 723, 731, 732, 751, 752, 18 Wood, treaters,, cabinet-makers, and, related, trades, workers 752 811, 815, 818, 821, 921, 932 311, 312, 313, 315, 325, 421, 422, 441, 512, 513, 522, 524, 621, 711, 721, 722, 19 Other, craft, and, related, workers 754 723, 732, 741, 742, 753, 813, 818, 821, 833, 932 313, 512, 711, 722, 723, 732, 751, 811, 20 Metal, processing, and, finishing, plant, operators 812 813, 815, 818, 932 265, 313, 512, 711, 712, 721, 722, 723, 21 Rubber,, plastic, and, paper, products, machine, operators 814 731, 732, 751, 753, 811, 813, 815, 818, 821, 932 313, 722, 732, 751, 811, 813, 818, 821, 22 Food, and, related, products, machine, operators 816 932 23 Wood, processing, and, papermaking, plant, operators 817 512, 722, 732, 815, 818, 821, 932 24 Locomotive, engine, drivers, and, related, workers 831 511, 711, 811, 832, 833, 931 313, 711, 742, 751, 811, 813, 833, 834, 25 Ships', deck, crews, and, related, workers 835 931 26 Vehicle,, window,, laundry, and, other, hand, cleaning, workers 912 815 27 Refuse, workers 961 441, 711, 712, 832, 834, 921 Sources: Own elaboration based on ILO (2012) and O*NET (2021). Notes: ILO (2012) lists some related occupations to a given one. O*NET (2021) lists occupations in the United States that have similar skills and experience and which workers could transfer between with minimal additional preparation. 41 Table A.1.3. Name of occupations used for imputations Code Occupation name 112 Managing directors and chief executives 121 Business services and administration managers 122 Sales, marketing and development managers 134 Professional services managers 141 Hotel and restaurant managers 142 Retail and wholesale trade managers 143 Other services managers 222 Nursing and midwifery professionals 224 Paramedical practitioners 225 Veterinarians 226 Other health professionals 232 Vocational education teachers 242 Administration professionals 243 Sales, marketing and public relations professionals 261 Legal professionals 265 Creative and performing artists 311 Physical and engineering science technicians 312 Mining, manufacturing and construction supervisors 313 Process control technicians 315 Ship and aircraft controllers and technicians 321 Medical and pharmaceutical technicians 322 Nursing and midwifery associate professionals 324 Veterinary technicians and assistants 325 Other health associate professionals 331 Financial and mathematical associate professionals 332 Sales and purchasing agents and brokers 333 Business services agents 334 Administrative and specialised secretaries 335 Regulatory government associate professionals 421 Tellers, money collectors and related clerks 422 Client information workers 441 Other clerical support workers 511 Travel attendants, conductors and guides 512 Cooks 513 Waiters and bartenders 516 Other personal services workers 522 Shop salespersons 524 Other sales workers 532 Personal care workers in health services 541 Protective services workers 621 Forestry and related workers 711 Building frame and related trades workers 712 Building finishers and related trades workers 721 Sheet and structural metal workers, moulders and welders, and related workers 722 Blacksmiths, toolmakers and related trades workers 723 Machinery mechanics and repairers 731 Handicraft workers 732 Printing trades workers 741 Electrical equipment installers and repairers 742 Electronics and telecommunications installers and repairers 751 Food processing and related trades workers 752 Wood treaters, cabinet-makers and related trades workers 753 Garment and related trades workers 811 Mining and mineral processing plant operators 813 Chemical and photographic products plant and machine operators 815 Textile, fur and leather products machine operators 818 Other stationary plant and machine operators 821 Assemblers 832 Car, van and motorcycle drivers 833 Heavy truck and bus drivers 834 Mobile plant operators 911 Domestic, hotel and office cleaners and helpers 921 Agricultural, forestry and fishery labourers 931 Mining and construction labourers 932 Manufacturing labourers 941 Food preparation assistants 962 Other elementary workers Sources: ILO (2012). 42 Annex 2. Occupations classified by digital levels, and share of levels of digital skills in each Table A.2.1 Highly digital occupations Share of digital skills O*NET # Occupation name Code Total Advanced Intermediate Basic digital score 1 Software and applications developers and analysts 251 58.1 47.5 8.0 2.6 78 2 Database and network professionals 252 54.7 42.6 8.3 3.9 76 3 ICT service managers 133 39.7 23.9 11.7 4.1 59 4 ICT operations and user support technicians 351 46.8 23.3 18.6 4.8 68 5 Electrotechnology engineers 215 33.1 21.4 8.7 3.0 68 6 Electronics and telecommunications installers and repairers 742 27.7 12.3 10.3 5.1 55 7 Mathematicians, actuaries and statisticians 212 22.7 8.0 6.6 8.2 61 8 Architects, planners, surveyors and designers 216 44.2 7.9 28.5 7.9 54 9 Telecommunications and broadcasting technicians 352 26.9 6.1 12.8 8.0 52 10 Physical and engineering science technicians 311 24.0 4.3 14.0 5.6 47 11 Printing trades workers 732 30.4 3.3 17.5 9.6 48 12 Creative and performing artists 265 34.4 2.3 25.1 7.0 45 13 Keyboard operators 413 55.6 2.0 2.6 51.0 46 Table A.2.2. Medium digital occupations Share of digital skills O*NET # Occupation name Code Total Advanced Intermediate Basic digital score 1 Blacksmiths, toolmakers and related trades workers 722 26.9 1.6 23.2 2.0 48 2 Other clerical support workers 441 26.7 2.7 2.4 21.7 45 3 General office clerks 411 21.6 2.2 2.4 17.1 47 4 Numerical clerks 431 20.4 1.8 3.7 14.9 45 5 Chemical and photographic products plant and machine op. 813 18.7 3.4 4.2 11.1 47 6 Engineering professionals (excluding electrotechnology) 214 18.3 6.0 7.4 4.9 57 7 Financial and mathematical associate professionals 331 17.8 1.8 4.5 11.5 42 8 Administration professionals 242 17.4 4.3 4.4 8.7 45 9 Vocational education teachers 232 16.9 3.9 8.0 5.1 45 10 Sales, marketing and public relations professionals 243 15.8 2.0 6.3 7.5 49 11 Building frame and related trades workers 711 15.5 1.8 7.5 6.2 32 12 Authors, journalists and linguists 264 15.3 2.0 5.8 7.5 46 13 Secretaries (general) 412 15.3 2.1 2.1 11.0 46 14 Food and related products machine operators 816 15.2 2.7 7.0 5.5 40 15 Librarians, archivists and curators 262 14.3 3.5 4.6 6.2 62 16 Wood processing and papermaking plant operators 817 14.1 2.4 7.4 4.3 39 17 Regulatory government associate professionals 335 13.9 2.4 2.5 9.0 47 18 Protective services workers 541 13.8 5.1 1.3 7.4 36 19 Administrative and specialised secretaries 334 13.6 2.5 2.7 8.4 47 20 Paramedical practitioners 224 13.6 3.5 3.6 6.4 40 21 Rubber, plastic and paper products machine operators 814 13.5 2.3 6.7 4.5 37 22 Other craft and related workers 754 13.4 2.7 5.1 5.7 40 23 Ship and aircraft controllers and technicians 315 13.3 1.9 2.9 8.6 41 24 Metal processing and finishing plant operators 812 12.8 2.0 6.1 4.8 38 25 Manufacturing labourers 932 12.5 2.1 4.9 5.5 33 26 Wood treaters, cabinet-makers and related trades workers 752 12.4 2.3 6.0 4.1 36 27 Assemblers 821 12.3 5.8 2.7 3.8 37 28 Other health associate professionals 325 12.2 2.5 2.2 7.4 38 29 Ships' deck crews and related workers 835 12.1 2.9 4.3 4.9 37 30 Other stationary plant and machine operators 818 11.7 2.5 2.8 6.4 43 31 Process control technicians 313 11.2 3.2 4.7 3.3 42 32 Physical and earth science professionals 211 11.2 3.4 4.0 3.8 54 Sheet and structural metal workers, moulders and welders 33 721 11.0 5.4 3.4 2.3 37 (…) 34 Painters, building structure cleaners and related trades w(…) 713 10.7 3.6 2.9 4.2 35 35 Handicraft workers 731 10.4 3.9 4.9 1.6 35 43 Table A.2.3. Low digital occupations Share of digital skills O*NET # Occupation name Code Total Advanced Intermediate Basic digital score 1 Artistic, cultural and culinary associate professionals 343 17.3 0.6 11.3 5.5 40 2 Material-recording and transport clerks 432 16.1 1.3 3.3 11.5 43 3 Legal, social and religious associate professionals 341 14.1 0.6 0.6 12.8 47 4 Mining, manufacturing and construction supervisors 312 13.0 1.0 5.4 6.5 44 5 Transport and storage labourers 933 12.7 0.5 3.0 9.2 25 6 Manufacturing, mining, construction, and 132 12.6 1.5 4.7 6.4 50 distribution managers 7 Building finishers and related trades workers 712 12.6 0.8 6.3 5.5 36 8 Mining and construction labourers 931 12.1 0.7 6.7 4.7 42 9 Business services agents 333 11.9 0.6 2.7 8.5 45 10 Client information workers 422 11.9 1.4 2.0 8.5 47 11 Refuse workers 961 11.8 1.4 3.3 7.1 33 12 Professional services managers 134 11.5 1.6 2.6 7.3 43 13 Legislators and senior officials 111 11.1 1.9 2.9 6.4 44 14 Finance professionals 241 10.8 1.6 3.6 5.6 47 15 Business services and administration managers 121 10.8 1.4 3.0 6.3 45 16 Personal care workers in health services 532 10.4 2.0 0.8 7.6 36 17 Life science technicians and related associate 314 9.9 1.0 2.4 6.5 41 professionals 18 Mining and mineral processing plant operators 811 9.7 1.7 2.4 5.7 25 19 Life science professionals 213 9.7 2.4 3.1 4.2 49 20 Managing directors and chief executives 112 9.5 2.7 2.9 3.9 45 21 Sales, marketing and development managers 122 9.5 1.2 3.2 5.0 45 22 University and higher education teachers 231 9.3 4.0 2.4 2.9 49 23 Traditional and complementary medicine associate 323 9.2 2.0 1.6 5.6 42 professionals 24 Travel attendants, conductors and guides 511 9.1 0.2 0.6 8.3 33 25 Locomotive engine drivers and related workers 831 9.1 1.3 3.1 4.6 32 26 Tellers, money collectors and related clerks 421 8.8 2.0 1.3 5.5 46 27 Shop salespersons 522 8.7 1.0 2.0 5.7 43 28 Electrical equipment installers and repairers 741 8.7 3.0 2.7 2.9 37 29 Production managers in agriculture, forestry and 131 8.7 1.1 2.3 5.2 33 fisheries 30 Social and religious professionals 263 8.6 1.2 1.5 5.8 43 31 Machinery mechanics and repairers 723 8.6 1.9 3.8 2.9 39 32 Veterinary technicians and assistants 324 8.5 1.7 1.3 5.5 39 33 Medical and pharmaceutical technicians 321 8.2 1.7 1.6 4.8 46 34 Sports and fitness workers 342 8.0 3.4 0.9 3.7 41 35 Nursing and midwifery associate professionals 322 7.2 1.7 0.9 4.6 43 36 Cashiers and ticket clerks 523 7.1 1.8 0.8 4.6 35 37 Other teaching professionals 235 7.0 1.4 1.2 4.4 45 38 Traditional and complementary medicine 223 6.5 1.5 1.5 3.4 44 professionals 39 Other services managers 143 6.0 1.5 1.5 3.0 46 40 Medical doctors 221 4.4 2.5 0.3 1.6 47 41 Heavy truck and bus drivers 833 3.6 2.2 0.4 1.0 29 44 Table A.2.4. Very low digital occupations Share of digital skills O*NET # Occupation name Code Total Advanced Intermediate Basic digital score 1 Sales and purchasing agents and brokers 332 8.5 0.7 1.7 6.2 42 2 Garment and related trades workers 753 8.0 0.0 4.5 3.5 23 3 Legal professionals 261 7.4 1.2 1.4 4.8 43 4 Mobile plant operators 834 7.1 0.3 1.5 5.3 29 5 Other health professionals 226 6.9 0.5 2.2 4.2 45 6 Child care workers and teachers' aides 531 6.7 1.0 1.5 4.2 29 7 Hairdressers, beauticians and related workers 514 5.9 0.9 0.1 4.8 33 8 Other elementary workers 962 5.4 0.3 0.9 4.3 36 9 Other sales workers 524 5.4 0.9 1.3 3.2 41 10 Food preparation assistants 941 5.1 1.0 0.7 3.4 32 11 Street vendors (excluding food) 952 4.9 0.7 1.0 3.2 35 12 Street and market salespersons 521 4.9 0.7 1.0 3.2 35 13 Street and related service workers 951 4.9 0.7 1.0 3.2 35 14 Veterinarians 225 4.8 0.5 1.3 3.0 49 15 Hotel and restaurant managers 141 4.7 0.5 1.0 3.2 37 16 Retail and wholesale trade managers 142 4.6 0.4 0.8 3.4 42 17 Waiters and bartenders 513 4.5 1.2 0.4 2.9 35 18 Mixed crop and animal producers 613 4.5 1.4 1.0 2.2 27 19 Subsistence mixed crop and livestock farmers 633 4.5 1.4 1.0 2.2 27 20 Animal producers 612 4.5 1.4 1.0 2.2 27 21 Forestry and related workers 621 4.5 1.4 1.0 2.2 27 22 Fishery workers, hunters and trappers 622 4.5 1.4 1.0 2.2 27 23 Agricultural, forestry and fishery labourers 921 4.5 1.4 1.0 2.2 27 24 Subsistence crop farmers 631 4.5 1.4 1.0 2.2 27 25 Market gardeners and crop growers 611 4.5 1.4 1.0 2.2 27 26 Subsistence livestock farmers 632 4.5 1.4 1.0 2.2 27 27 Subsistence fishers, hunters, trappers and gatherers 634 4.5 1.4 1.0 2.2 27 28 Car, van and motorcycle drivers 832 4.5 1.3 1.1 2.1 30 29 Building and housekeeping supervisors 515 4.4 0.2 0.4 3.7 43 30 Textile, fur and leather products machine operators 815 4.0 0.7 1.0 2.3 29 31 Vehicle, window, laundry and other hand cleaning workers 912 4.0 0.7 1.0 2.3 29 32 Domestic, hotel and office cleaners and helpers 911 3.6 0.1 1.4 2.2 23 33 Food processing and related trades workers 751 3.6 0.6 1.1 1.9 33 34 Other personal services workers 516 3.2 0.4 0.6 2.2 35 35 Secondary education teachers 233 2.9 0.1 0.5 2.4 45 36 Nursing and midwifery professionals 222 2.5 0.6 0.2 1.7 41 37 Primary school and early childhood teachers 234 2.1 0.3 0.2 1.6 41 38 Cooks 512 1.0 0.7 0.0 0.3 33 Source: Authors’ calculations based on World Bank-Burning Glass Malaysia online job postings data (2016-18). Notes: Darker red colors show higher magnitudes within the “share of digital skills” or “O*Net digital score” column while darker blue colors show lower magnitudes within each set of columns. O*NET digital score, used as a benchmark here, is based on occupations’ three digitalization aspects in measured by O*NET for occupations in the U.S. —computer & electronics, programming, and interacting with computer— following the methodology by Muro et al. (2017). 45 Annex 3. Replications of Malaysia-focused analyses about the similarity of skills required in very-low, low, medium, and highly digital occupations for the occupational skills profile, Cambodia, Thailand, and Vietnam Table A.3.1. Share of skills categories across digital occupation levels in the occupational skills profile, Cambodia, Malaysia, Thailand, and Vietnam (replication of data behind figure 6) Occupational skills Cambodia Malaysia Thailand Vietnam profile Digital occupation Digital occupation Digital occupation Digital occupation Digital occupation levels levels levels levels levels Skills VL L M H VL L M H VL L M H VL L M H VL L M H Digital 4.8 9.9 15.0 38.3 5.0 9.8 15.5 32.7 4.8 9.4 15.9 33.3 4.5 10.1 15.3 37.1 4.6 9.7 14.8 34.4 Advanced digital 0.8 1.6 3.0 15.8 1.0 1.2 2.7 8.7 0.8 1.6 3.1 13.2 1.1 1.3 3.1 12.6 1.1 1.3 3.0 14.5 Intermediate digital 1.0 2.6 5.0 13.3 1.5 3.2 5.2 17.5 1.1 2.9 3.8 14.3 1.0 3.1 4.8 15.8 1.1 3.0 5.1 13.6 Basic digital 2.9 5.6 7.1 9.3 2.5 5.5 7.6 6.5 2.9 5.0 9.0 5.9 2.4 5.7 7.4 8.8 2.4 5.4 6.7 6.2 Cognitive 10.9 11.9 10.3 9.4 9.9 10.7 10.1 8.6 10.5 11.6 10.1 8.3 10.5 11.0 9.5 8.8 10.3 11.3 10.2 8.9 Thinking 0.6 1.3 1.3 2.0 0.4 1.0 1.0 1.6 0.6 1.2 1.2 1.8 0.4 1.1 1.2 1.8 0.4 1.1 1.2 2.0 Communication 5.6 6.0 4.7 3.9 5.2 5.6 4.8 3.5 5.5 5.7 4.6 3.4 5.5 5.4 4.4 3.6 5.3 5.6 4.8 3.7 Organization 4.4 4.5 4.0 3.4 4.2 3.9 4.2 3.4 4.3 4.5 4.1 3.0 4.4 4.2 3.7 3.2 4.4 4.3 4.1 3.1 Socioemotional 11.0 7.9 6.4 6.7 10.0 7.1 5.8 7.6 12.2 7.8 5.9 5.5 9.9 7.1 5.8 6.8 9.9 7.2 6.2 6.1 Emotions 0.5 0.5 0.6 0.7 0.7 0.3 0.8 0.5 0.4 0.4 0.6 0.4 0.3 0.4 0.6 0.6 0.3 0.4 0.7 0.4 Relationships 7.9 5.7 4.3 3.6 5.6 5.4 3.8 3.4 8.9 5.7 4.1 3.3 6.3 5.3 4.0 3.4 6.2 5.4 4.2 3.5 Personal growth 2.6 1.7 1.5 2.4 3.7 1.4 1.2 3.7 2.8 1.6 1.2 1.8 3.3 1.4 1.2 2.8 3.4 1.4 1.2 2.2 Other 73.4 70.4 68.3 45.6 75.0 72.4 68.6 51.2 72.5 71.2 68.1 52.8 75.1 71.8 69.4 47.3 75.2 71.8 68.8 50.6 Total 100 100 100 100 100 100 100 100 100 100 100 100 100 100 100 100 100 100 100 100 Source: Authors’ calculations based on SEAD data. Notes: “VL”, “L”, “M”, and “H” respectively stands for “Very low”, “Low”, “Medium”, and “High”. The share of skills splits between digital, cognitive, and socioemotional, and other skills. Digital skills levels and cognitive and socioemotional skills subsets are subcategories counted in the total. 46 Table A.3.2. Conditional correlations between digital occupation level and cognitive and socioemotional skills requirements (replication of Table 6) Ordinary Least Square (OLS) regressions of digital occupation levels on dummies taking the value of one if the skills count of a skills category of an occupation is above the mean A. Occupational skills profile Occupation digital level Low vs. Medium High vs. High vs. High vs. Very low vs. Low Medium Low Very low (1) (2) (3) (4) (5) Cognitive Thinking 0.24 0.19 0.23 0.65* 2.26*** (0.15) (0.13) (0.13) (0.26) (0.38) Communication 0.28* -0.19 0.15 -0.13 -0.33 (0.13) (0.14) (0.24) (0.24) (0.25) Organization 0.24 -0.05 0.05 0.06 -0.45 (0.13) (0.13) (0.17) (0.27) (0.61) Socioemotional Emotions -0.01 0.28* 0.06 0.40 0.91 (0.13) (0.13) (0.18) (0.27) (0.54) Relationships -0.28** -0.26 -0.25 -0.58* -0.45 (0.10) (0.18) (0.28) (0.28) (0.27) Personal growth -0.20* 0.02 0.18 0.13 -0.21 (0.09) (0.17) (0.18) (0.28) (0.26) Constant 0.45*** 1.46*** 2.05*** 1.22*** 0.32 (0.09) (0.09) (0.09) (0.17) (0.29) N 79 76 48 54 51 R-sq 0.30 0.15 0.13 0.24 0.61 47 B. Cambodia and Malaysia Cambodia Malaysia Occupation digital level Occupation digital level High High vs. Low vs. Mediu High vs. Low vs. Medium High vs. High vs. High vs. vs. Very Very m vs. Very Very low vs. Low Medium Medium Low Low low low Low low (6) (7) (8) (9) (10) (11) (12) (13) (14) (15) Cognitive Thinking 0.25 -0.04 0.20 0.49* 2.28*** 0.36* 0.22 0.01 0.53* 2.09*** (0.16) (0.13) (0.11) (0.21) (0.46) (0.17) (0.13) (0.15) (0.23) (0.40) Communication 0.79*** -0.24* 0.60** 0.04 -0.28 0.47* -0.19 0.01 -0.17 -0.09 (0.15) (0.09) (0.21) (0.16) (0.24) (0.21) (0.14) (0.33) (0.24) (0.16) Organization 0.07 -0.24* 0.01 -0.22 -0.06 0.00 -0.16 0.12 -0.13 -0.60 (0.12) (0.10) (0.05) (0.15) (0.10) (0.16) (0.15) (0.10) (0.14) (0.41) Socioemotional Emotions -0.06 0.71*** -0.03 0.25 0.10 -0.00 0.48** -0.30 0.12 0.81 (0.10) (0.10) (0.08) (0.21) (0.09) (0.15) (0.15) (0.21) (0.19) (0.41) Relationships -0.05 -0.16 -0.86*** -0.75** -0.09 -0.30* -0.35 0.08 -0.58* -0.67 (0.10) (0.15) (0.21) (0.24) (0.08) (0.14) (0.18) (0.38) (0.24) (0.43) Personal growth -0.14 -0.08 0.46** 0.60** -0.07 -0.16 -0.03 0.36* 0.34 -0.55 (0.12) (0.18) (0.15) (0.19) (0.08) (0.16) (0.17) (0.18) (0.20) (0.41) Constant 0.20 1.46*** 2.03*** 1.12*** 0.08 0.44** 1.47*** 2.22*** 1.39*** 0.74 (0.12) (0.14) (0.07) (0.10) (0.08) (0.14) (0.14) (0.22) (0.33) (0.54) N 79 75 45 52 49 79 75 47 54 51 R-sq 0.65 0.66 0.49 0.39 0.59 0.36 0.36 0.13 0.15 0.44 48 C. Thailand and Vietnam Thailand Vietnam Occupation digital level Occupation digital level Low vs. High vs. Low vs. High vs. Medium High vs. High vs. Medium High vs. High vs. Very Very Very Very vs. Low Medium Low vs. Low Medium Low low low low low (16) (17) (18) (19) (20) (21) (22) (23) (24) (25) Cognitive Thinking 0.26 0.21 0.08 0.46 2.11*** 0.24 0.25* 0.08 0.65* 2.59*** (0.18) (0.13) (0.09) (0.25) (0.40) (0.14) (0.12) (0.08) (0.25) (0.28) Communication 0.60* -0.26* 0.35 -0.03 -0.15 0.77*** -0.25 0.38 0.24 -0.19 (0.23) (0.12) (0.29) (0.20) (0.14) (0.17) (0.16) (0.29) (0.25) (0.15) Organization 0.10 -0.22 0.06 -0.19 -0.45 0.11 -0.28* -0.03 -0.42 -0.01 (0.12) (0.12) (0.09) (0.20) (0.37) (0.11) (0.13) (0.08) (0.21) (0.12) Socioemotional Emotions -0.08 0.51*** -0.04 0.29 0.45 -0.14 0.59*** 0.00 0.29 -0.01 (0.11) (0.12) (0.10) (0.21) (0.38) (0.08) (0.13) (0.09) (0.22) (0.09) Relationships -0.14 -0.29* -0.41 -0.53* -0.16 -0.09 -0.43 -0.49 -0.84** -0.09 (0.11) (0.14) (0.33) (0.24) (0.12) (0.07) (0.23) (0.31) (0.27) (0.07) Personal growth -0.13 0.03 0.27 0.37 -0.14 -0.16 0.14 0.20 0.43* -0.10 (0.14) (0.16) (0.17) (0.21) (0.14) (0.08) (0.21) (0.14) (0.19) (0.09) Constant 0.22 1.48*** 2.03*** 1.17*** 0.15 0.20* 1.57*** 2.03*** 1.17*** 0.12 (0.13) (0.11) (0.09) (0.14) (0.14) (0.08) (0.11) (0.06) (0.16) (0.10) N 79 76 48 54 51 79 76 48 54 51 R-sq 0.45 0.47 0.18 0.23 0.52 0.66 0.48 0.19 0.28 0.65 Source: Authors’ calculations based SEAD data. Note: Independent variables are dummies taking the value of one if the skills count of a skills category of an occupation is above the meian and thus can be interpreted as percentage point increase (or decrease) towards the probability of being a given occupation digital level. Malaysia does not have one of the 127 occupations of the occupational skills profile (paramedical practitioners, ISCO code 224) according its 2017 labor force survey, so its total number of occupations is 126. *p < 0.1 **p < 0.05 ***p < 0.01 49 Annex 4. Replications of Malaysia-focused analyses about the complementarities between digital, cognitive, and socioemotional skills for the occupational skills profile, Cambodia, Thailand, and Vietnam Table A.4.1. Correlations between skills (replications of Table 7) Pairwise correlations based on the number of skills counts of skills categories A. Occupational skills profile Digital Cognitive Socioemotional Personal Advanced Intermediate Basic Thinking Communication Organization Emotions Relationships growth Advanced 1.00 Digital Intermediate 0.45*** 1.00 Basic 0.05 0.31*** 1.00 Cognitive Thinking 0.55*** 0.40*** 0.39*** 1.00 Communication 0.22** 0.11 0.37*** 0.65*** 1.00 Organization 0.14 0.30*** 0.44*** 0.59*** 0.60*** 1.00 Socioemotional Emotions 0.15 0.28*** 0.64*** 0.46*** 0.40*** 0.44*** 1.00 Relationships 0.06 -0.04 0.08 0.37*** 0.59*** 0.49*** 0.17* 1.00 Personal growth 0.07 0.64*** 0.18** 0.07 0.13 0.35*** 0.29*** 0.16* 1.00 Others 0.02 0.15* 0.33*** 0.60*** 0.66*** 0.56*** 0.23*** 0.33*** 0.09 B. Cambodia Digital Cognitive Socioemotional Personal Advanced Intermediate Basic Thinking Communication Organization Emotions Relationships growth Advanced 1.00 Digital Intermediate 0.31*** 1.00 Basic 0.10 0.35*** 1.00 Cognitive Thinking 0.34*** 0.27*** 0.64*** 1.00 Communication 0.14 0.07 0.62*** 0.68*** 1.00 Organization 0.09 0.35*** 0.54*** 0.53*** 0.43*** 1.00 Socioemotional Emotions 0.01 0.46*** 0.48*** 0.33*** 0.08 0.31*** 1.00 Relationships -0.09 0.02 0.33*** 0.39*** 0.49*** 0.16* 0.45*** 1.00 Personal growth -0.07 0.15* -0.30*** -0.27*** -0.25*** 0.06 -0.06 -0.04 1.00 Others 0.08 0.12 0.58*** 0.62*** 0.82*** 0.43*** 0.05 0.29*** -0.13 50 C. Thailand Digital Cognitive Socioemotional Personal Advanced Intermediate Basic Thinking Communication Organization Emotions Relationships growth Advanced 1.00 Digital Intermediate 0.42*** 1.00 Basic 0.12 0.35*** 1.00 Cognitive Thinking 0.44*** 0.39*** 0.60*** 1.00 Communication 0.15* 0.15* 0.60*** 0.64*** 1.00 Organization 0.10 0.27*** 0.50*** 0.61*** 0.51*** 1.00 Socioemotional Emotions 0.15* 0.32*** 0.63*** 0.59*** 0.46*** 0.41*** 1.00 Relationships 0.01 -0.02 0.25*** 0.36*** 0.60*** 0.28*** 0.33*** 1.00 Personal growth 0.03 0.31*** -0.22** -0.18** -0.10 0.11 -0.14 -0.02 1.00 Others 0.07 0.21** 0.56*** 0.57*** 0.80*** 0.49*** 0.30*** 0.36*** -0.09 D. Vietnam Digital Cognitive Socioemotional Personal Advanced Intermediate Basic Thinking Communication Organization Emotions Relationships growth Advanced 1.00 Digital Intermediate 0.46*** 1.00 Basic 0.16* 0.46*** 1.00 Cognitive Thinking 0.44*** 0.44*** 0.67*** 1.00 Communication 0.20** 0.25*** 0.70*** 0.72*** 1.00 Organization 0.15* 0.39*** 0.55*** 0.60*** 0.60*** 1.00 Socioemotional Emotions 0.14 0.40*** 0.62*** 0.57*** 0.47*** 0.45*** 1.00 Relationships -0.00 -0.02 0.31*** 0.34*** 0.52*** 0.20** 0.33*** 1.00 Personal growth -0.03 0.07 -0.32*** -0.23*** -0.20** 0.05 -0.27*** -0.02 1.00 Others 0.12 0.28*** 0.60*** 0.69*** 0.77*** 0.57*** 0.32*** 0.23** -0.10 Source: Authors’ calculations based SEAD data. *p < 0.1 **p < 0.05 ***p < 0.01 51 Table A.4.2. Exploratory factor analysis of skills for the occupational skills profile, Cambodia, Thailand, and Vietnam (replications of Table 8) Exploratory factor analysis based on skills counts of skills categories A. Occupational skills profile Factors Skills category Skill subset 1 2 3 4 Uniqueness 1 Digital Advanced 0.03 -0.03 0.94 0.10 0.11 2 Digital Intermediate -0.08 0.20 0.48 0.78 0.11 3 Digital Basic 0.09 0.90 0.04 0.11 0.16 4 Cognitive Thinking 0.50 0.39 0.68 0.02 0.13 5 Cognitive Communication 0.79 0.33 0.24 -0.03 0.21 6 Cognitive Organization 0.68 0.41 0.12 0.29 0.27 7 Socioemotional Relationships 0.18 0.83 0.10 0.17 0.24 8 Socioemotional Emotions 0.91 -0.07 -0.02 0.05 0.17 9 Socioemotional Personal growth 0.14 0.09 -0.07 0.95 0.07 B. Cambodia Factors Skills category Skill subset 1 2 3 Uniqueness 1 Digital Advanced 0.24 0.18 0.79 0.29 2 Digital Intermediate 0.09 0.83 0.28 0.23 3 Digital Basic 0.78 0.35 -0.06 0.27 4 Cognitive Thinking 0.85 0.24 0.14 0.20 5 Cognitive Communication 0.87 -0.03 -0.05 0.25 6 Cognitive Organization 0.49 0.52 0.04 0.48 7 Socioemotional Relationships 0.28 0.68 -0.36 0.32 8 Socioemotional Emotions 0.51 0.16 -0.62 0.33 9 Socioemotional Personal growth -0.54 0.45 -0.12 0.49 C. Malaysia Factors Skills category Skill subset 1 2 3 Uniqueness 1 Digital Advanced 0.15 0.74 -0.07 0.43 2 Digital Intermediate 0.20 0.87 0.07 0.20 3 Digital Basic 0.76 0.09 -0.26 0.34 4 Cognitive Thinking 0.80 0.36 0.07 0.22 5 Cognitive Communication 0.82 -0.02 0.32 0.23 6 Cognitive Organization 0.80 0.14 0.23 0.29 7 Socioemotional Relationships 0.74 0.16 -0.06 0.42 8 Socioemotional Emotions 0.34 -0.25 0.78 0.22 9 Socioemotional Personal growth -0.12 0.43 0.75 0.24 52 D. Thailand Factors Skills category Skill subset 1 2 3 Uniqueness 1 Digital Advanced 0.03 0.78 -0.04 0.39 2 Digital Intermediate 0.19 0.78 0.35 0.24 3 Digital Basic 0.71 0.32 -0.26 0.32 4 Cognitive Thinking 0.73 0.49 -0.18 0.19 5 Cognitive Communication 0.87 0.02 -0.05 0.25 6 Cognitive Organization 0.71 0.20 0.22 0.41 7 Socioemotional Relationships 0.67 0.31 -0.20 0.41 8 Socioemotional Emotions 0.73 -0.31 0.12 0.36 9 Socioemotional Personal growth -0.07 0.09 0.95 0.08 E. Vietnam Factors Skills category Skill subset 1 2 3 Uniqueness 1 Digital Advanced 0.06 0.78 -0.03 0.38 2 Digital Intermediate 0.28 0.81 0.09 0.26 3 Digital Basic 0.76 0.28 -0.32 0.24 4 Cognitive Thinking 0.75 0.43 -0.18 0.22 5 Cognitive Communication 0.88 0.08 -0.10 0.21 6 Cognitive Organization 0.72 0.30 0.20 0.36 7 Socioemotional Relationships 0.66 0.24 -0.31 0.41 8 Socioemotional Emotions 0.70 -0.36 0.10 0.37 9 Socioemotional Personal growth -0.10 0.04 0.96 0.07 Sources: Authors’ calculations based on SEAD data. Note: The factor analysis is based on varimax rotation for orthogonal factors. 53 Annex 5. Occupations’ classification in exploratory factor analysis of skills categories Table A.5.1. Occupations’ scores of factors from exploratory factor analysis of digital, cognitive, and socioemotional skills subsets in Malaysia Occupation Employment Digitalization Factor Factor Factor share in Name Code level 1 2 3 Malaysia Architects, planners, surveyors and designers 216 High -0.5 7.8 3.9 0.4 Software and applications developers and analysts 251 High 0.5 7.2 -0.2 0.3 Database and network professionals 252 High 1.3 6.5 -0.6 0.3 Creative and performing artists 265 High -0.3 6.1 3.5 0.1 ICT operations and user support technicians 351 High 1.1 4.9 -0.4 0.3 Telecommunications and broadcasting technicians 352 High 0.3 4.5 2.3 0.1 Information and communications technology service managers 133 High 1.4 4.0 0.4 0.1 Electrotechnology engineers 215 High 0.4 3.3 -0.4 0.5 Printing trades workers 732 High -0.5 2.3 0.5 0.1 Electronics and telecommunications installers and repairers 742 High 0.1 2.0 -0.3 0.1 Mathematicians, actuaries and statisticians 212 High 2.3 1.7 -0.3 0.0 Physical and engineering science technicians 311 High -0.5 1.1 -0.9 2.3 Keyboard operators 413 High 2.3 -0.3 -3.3 0.1 Regulatory government associate professionals 335 Medium 2.9 -0.1 -0.5 0.9 Chemical and photographic products plant and machine operators 813 Medium 2.4 0.4 0.3 0.2 Administration professionals 242 Medium 2.3 0.6 0.4 0.3 Administrative and specialised secretaries 334 Medium 2.1 -0.2 0.1 0.9 Librarians, archivists and curators 262 Medium 1.6 -0.1 1.6 0.0 Numerical clerks 431 Medium 1.3 -0.2 -1.3 1.2 Authors, journalists and linguists 264 Medium 1.3 2.2 2.6 0.1 Sales, marketing and public relations professionals 243 Medium 1.1 1.4 1.3 0.2 Secretaries (general) 412 Medium 1.1 -0.5 -0.2 0.1 Ship and aircraft controllers and technicians 315 Medium 1.1 -0.2 0.0 0.1 General office clerks 411 Medium 1.1 -0.3 -1.5 4.8 Financial and mathematical associate professionals 331 Medium 0.9 0.1 -1.2 0.3 Engineering professionals (excluding electrotechnology) 214 Medium 0.7 1.2 -0.3 1.0 Protective services workers 541 Medium 0.6 -0.2 -0.5 2.8 Process control technicians 313 Medium 0.6 0.8 -0.4 0.1 Other clerical support workers 441 Medium 0.3 -0.4 -1.6 0.5 Ships' deck crews and related workers 835 Medium 0.1 0.4 -0.3 0.1 Other health associate professionals 325 Medium 0.1 0.0 -0.9 0.4 Physical and earth science professionals 211 Medium 0.0 0.6 -0.1 0.0 Building frame and related trades workers 711 Medium 0.0 0.9 -0.8 1.8 Other craft and related workers 754 Medium -0.1 0.2 -0.3 0.1 Food and related products machine operators 816 Medium -0.1 0.7 -0.3 0.3 Assemblers 821 Medium -0.2 0.1 -0.8 1.1 Vocational education teachers 232 Medium -0.2 0.4 -0.5 0.1 Metal processing and finishing plant operators 812 Medium -0.3 0.5 -0.3 0.2 Rubber, plastic and paper products machine operators 814 Medium -0.3 0.8 -0.1 0.6 Manufacturing labourers 932 Medium -0.4 0.2 -0.7 0.5 Paramedical practitioners 224 Medium -0.4 -0.1 -0.6 Wood treaters, cabinet-makers and related trades workers 752 Medium -0.5 0.5 -0.4 0.5 Sheet and structural metal workers, moulders and welders, (…) 721 Medium -0.6 0.6 -0.9 0.9 Painters, building structure cleaners and related trades workers 713 Medium -0.6 -0.1 -0.9 0.4 Wood processing and papermaking plant operators 817 Medium -0.7 0.5 -0.5 0.5 Handicraft workers 731 Medium -0.7 0.5 -0.3 0.1 Other stationary plant and machine operators 818 Medium -0.7 -0.2 -0.9 3.2 Blacksmiths, toolmakers and related trades workers 722 Medium -1.3 1.9 -1.4 0.1 Finance professionals 241 Low 2.4 0.2 0.0 1.5 Managing directors and chief executives 112 Low 2.1 0.0 1.7 1.1 Business services and administration managers 121 Low 2.1 -0.2 0.5 1.5 Other services managers 143 Low 2.0 -1.0 3.2 0.2 Professional services managers 134 Low 1.8 -0.3 0.3 0.2 Manufacturing, mining, construction, and distribution managers 132 Low 1.6 0.0 0.1 0.9 Sales, marketing and development managers 122 Low 1.5 0.4 2.1 0.4 54 Occupation Employment Digitalization Factor Factor Factor share in Name Code level 1 2 3 Malaysia Material-recording and transport clerks 432 Low 1.4 -0.2 -0.7 0.9 Legal, social and religious associate professionals 341 Low 1.4 -0.6 -1.2 0.2 Legislators and senior officials 111 Low 1.4 -0.1 0.6 0.1 Client information workers 422 Low 1.3 -0.4 -0.3 0.5 Business services agents 333 Low 1.2 -0.2 0.7 0.4 Social and religious professionals 263 Low 1.1 -0.6 0.6 0.2 Life science professionals 213 Low 1.0 0.0 0.4 0.1 Tellers, money collectors and related clerks 421 Low 0.8 -1.1 0.9 0.5 Travel attendants, conductors and guides 511 Low 0.6 0.3 1.4 0.1 Other teaching professionals 235 Low 0.6 -0.7 2.0 0.7 Mining, manufacturing and construction supervisors 312 Low 0.5 -0.2 -0.1 1.3 Transport and storage labourers 933 Low 0.4 -0.2 -1.0 0.2 Shop salespersons 522 Low 0.4 -0.3 0.4 5.7 Production managers in agriculture, forestry and fisheries 131 Low 0.4 -0.3 0.2 0.1 Mining and construction labourers 931 Low 0.4 0.1 -0.6 2.2 Life science technicians and related associate professionals 314 Low 0.4 -0.1 -0.7 0.1 Sports and fitness workers 342 Low 0.3 -0.2 1.4 0.1 Personal care workers in health services 532 Low 0.0 -0.7 -0.3 0.2 University and higher education teachers 231 Low -0.2 -0.1 0.5 0.6 Traditional and complementary medicine associate professionals 323 Low -0.2 -0.3 -1.1 0.1 Veterinary technicians and assistants 324 Low -0.2 -0.5 -0.5 0.0 Locomotive engine drivers and related workers 831 Low -0.2 0.1 -0.3 0.0 Traditional and complementary medicine professionals 223 Low -0.3 -0.5 -0.4 0.0 Medical and pharmaceutical technicians 321 Low -0.3 -0.5 -0.9 0.1 Nursing and midwifery associate professionals 322 Low -0.3 -0.5 -1.4 0.7 Mining and mineral processing plant operators 811 Low -0.3 0.2 -0.5 0.2 Medical doctors 221 Low -0.4 -0.7 -0.1 0.4 Artistic, cultural and culinary associate professionals 343 Low -0.4 2.5 2.3 0.3 Refuse workers 961 Low -0.6 0.1 -0.9 0.5 Cashiers and ticket clerks 523 Low -0.6 -0.3 -0.8 0.8 Machinery mechanics and repairers 723 Low -0.8 0.0 -0.8 2.2 Building finishers and related trades workers 712 Low -0.9 0.2 -0.9 0.9 Electrical equipment installers and repairers 741 Low -1.0 0.0 -0.8 1.2 Heavy truck and bus drivers 833 Low -1.1 -0.4 -0.7 3.2 Primary school and early childhood teachers 234 Very low -1.1 -0.9 2.8 1.9 Secondary education teachers 233 Very low -0.8 -1.1 2.7 1.4 Retail and wholesale trade managers 142 Very low 1.1 -0.8 1.7 0.5 Child care workers and teachers' aides 531 Very low 0.5 -0.4 1.6 1.6 Food processing and related trades workers 751 Very low -0.4 0.2 1.3 1.2 Hotel and restaurant managers 141 Very low 1.0 -0.9 1.2 0.5 Other sales workers 524 Very low 0.1 -0.7 0.9 1.6 Food preparation assistants 941 Very low 0.6 -0.7 0.8 0.5 Sales and purchasing agents and brokers 332 Very low 0.5 -0.6 0.6 2.3 Building and housekeeping supervisors 515 Very low 0.3 -1.0 0.5 0.2 Hairdressers, beauticians and related workers 514 Very low -0.7 -0.7 0.4 0.8 Market gardeners and crop growers 611 Very low -1.3 0.3 0.3 5.1 Animal producers 612 Very low -1.3 0.3 0.3 0.2 Mixed crop and animal producers 613 Very low -1.3 0.3 0.3 0.0 Forestry and related workers 621 Very low -1.3 0.3 0.3 0.0 Fishery workers, hunters and trappers 622 Very low -1.3 0.3 0.3 0.7 Subsistence crop farmers 631 Very low -1.3 0.3 0.3 0.0 Subsistence livestock farmers 632 Very low -1.3 0.3 0.3 0.0 Subsistence mixed crop and livestock farmers 633 Very low -1.3 0.3 0.3 0.0 Subsistence fishers, hunters, trappers and gatherers 634 Very low -1.3 0.3 0.3 0.2 Agricultural, forestry and fishery labourers 921 Very low -1.3 0.3 0.3 4.8 Waiters and bartenders 513 Very low -0.4 -0.6 0.3 2.6 Garment and related trades workers 753 Very low -0.2 0.3 0.3 1.1 Cooks 512 Very low -1.2 -0.3 0.2 1.5 Street and market salespersons 521 Very low -0.2 -0.5 0.2 4.2 Street and related service workers 951 Very low -0.2 -0.5 0.2 0.0 Street vendors (excluding food) 952 Very low -0.2 -0.5 0.2 0.3 55 Occupation Employment Digitalization Factor Factor Factor share in Name Code level 1 2 3 Malaysia Domestic, hotel and office cleaners and helpers 911 Very low 0.0 -0.8 0.0 2.1 Veterinarians 225 Very low -0.1 -0.6 0.0 0.0 Other personal services workers 516 Very low -0.6 -0.6 -0.1 0.1 Textile, fur and leather products machine operators 815 Very low -0.4 -0.2 -0.2 0.1 Vehicle, window, laundry and other hand cleaning workers 912 Very low -0.4 -0.2 -0.2 0.2 Legal professionals 261 Very low 1.8 -0.4 -0.3 0.2 Other health professionals 226 Very low 0.1 -0.5 -0.4 0.3 Car, van and motorcycle drivers 832 Very low -0.8 -0.2 -0.7 1.5 Nursing and midwifery professionals 222 Very low -0.6 -0.8 -0.7 0.2 Other elementary workers 962 Very low -1.0 -0.3 -1.1 1.5 Mobile plant operators 834 Very low -0.8 -0.3 -1.4 0.8 Source: Authors’ calculations based SEAD data. Notes: The factor analysis is based on varimax rotation for orthogonal factors of nine digital, cognitive and socioemotional subsets of skills (three each). Clusters are described in section 6. Darker red and blue colors show higher and lower magnitudes, respectively. 56 Annex 6. Distribution of occupations across job postings and employment in the four countries Table A.6.1. Share of job vacancies and employment by digital occupation level of the 127 occupations Share in Share in employment Digital ISCO Occupation name job occupation code Cambodia Malaysia Thailand Vietnam vacancies level Software and applications developers and analysts 251 6.0 0.1 0.3 0.1 0.1 High Database and network professionals 252 1.8 0.0 0.3 0.1 0.1 High Information and communications technology service managers 133 1.0 0.0 0.1 0.0 0.0 High ICT operations and user support technicians 351 1.7 0.0 0.3 0.2 0.1 High Electrotechnology engineers 215 1.2 0.0 0.5 0.1 0.2 High Electronics and telecommunications installers and repairers 742 0.1 0.1 0.1 0.3 0.3 High Mathematicians, actuaries and statisticians 212 0.1 0.0 0.0 0.0 High Architects, planners, surveyors and designers 216 2.3 0.0 0.4 0.3 0.1 High Telecommunications and broadcasting technicians 352 0.1 0.1 0.1 0.0 0.0 High Physical and engineering science technicians 311 1.7 0.2 2.3 0.4 0.3 High Printing trades workers 732 0.0 0.1 0.1 0.1 0.1 High Creative and performing artists 265 0.2 0.3 0.1 0.2 0.1 High Keyboard operators 413 0.1 0.1 0.1 0.0 High Engineering professionals (excluding electrotechnology) 214 5.9 0.0 1.0 0.3 0.5 Medium Assemblers 821 0.1 0.1 1.1 1.6 0.8 Medium Sheet and structural metal workers, moulders and welders (…) 721 0.1 0.9 0.9 0.6 1.0 Medium Protective services workers 541 0.3 1.2 2.8 1.4 0.9 Medium Administration professionals 242 2.7 0.0 0.3 0.3 0.8 Medium Vocational education teachers 232 0.0 0.0 0.1 0.1 0.1 Medium Handicraft workers 731 0.0 1.0 0.1 1.3 0.9 Medium Painters, building structure cleaners and related trades workers 713 0.2 0.1 0.4 0.4 0.3 Medium Librarians, archivists and curators 262 0.0 0.0 0.0 0.0 0.1 Medium Paramedical practitioners 224 0.1 0.0 0.0 0.0 Medium Chemical and photographic products plant and machine operators 813 0.0 0.1 0.2 0.2 0.0 Medium Physical and earth science professionals 211 0.2 0.0 0.1 0.0 Medium Process control technicians 313 0.0 0.0 0.1 0.1 0.1 Medium Ships' deck crews and related workers 835 0.0 0.1 0.1 0.2 Medium Other craft and related workers 754 0.0 0.1 0.5 0.5 Medium Other clerical support workers 441 0.1 0.1 0.5 0.4 0.9 Medium Food and related products machine operators 816 0.3 0.3 0.5 0.2 Medium Other health associate professionals 325 0.4 0.1 0.4 0.3 0.1 Medium Administrative and specialised secretaries 334 5.2 0.2 0.9 0.2 0.2 Medium Other stationary plant and machine operators 818 0.1 0.1 3.2 0.5 0.7 Medium Regulatory government associate professionals 335 0.1 0.0 0.9 0.2 0.5 Medium Wood processing and papermaking plant operators 817 0.0 0.5 0.1 0.2 Medium Wood treaters, cabinet-makers and related trades workers 752 0.9 0.5 0.3 0.8 Medium Rubber, plastic and paper products machine operators 814 0.1 0.6 0.6 0.3 Medium General office clerks 411 6.0 2.1 4.8 1.5 0.1 Medium Secretaries (general) 412 0.0 0.1 0.0 0.0 Medium Manufacturing labourers 932 1.0 0.4 0.5 1.3 1.3 Medium Sales, marketing and public relations professionals 243 4.9 0.2 0.2 0.4 0.5 Medium Authors, journalists and linguists 264 0.5 0.2 0.1 0.1 0.1 Medium Metal processing and finishing plant operators 812 0.0 0.2 0.3 0.2 Medium Ship and aircraft controllers and technicians 315 0.1 0.0 0.1 0.0 0.0 Medium Building frame and related trades workers 711 0.0 4.5 1.8 1.2 4.2 Medium Numerical clerks 431 0.8 0.1 1.2 0.5 0.1 Medium Financial and mathematical associate professionals 331 5.4 0.6 0.3 1.5 0.4 Medium Blacksmiths, toolmakers and related trades workers 722 0.3 0.2 0.1 0.5 0.1 Medium University and higher education teachers 231 0.3 0.0 0.6 0.2 0.2 Low Sports and fitness workers 342 0.1 0.0 0.1 0.1 0.0 Low Electrical equipment installers and repairers 741 0.1 0.4 1.2 0.9 0.5 Low Managing directors and chief executives 112 1.9 0.0 1.1 0.1 0.0 Low Medical doctors 221 0.1 0.1 0.4 0.1 0.2 Low 57 Share in Share in employment Digital ISCO Occupation name job occupation code Cambodia Malaysia Thailand Vietnam vacancies level Life science professionals 213 0.2 0.0 0.1 0.1 0.1 Low Heavy truck and bus drivers 833 0.1 0.4 3.2 0.7 0.9 Low Traditional and complementary medicine associate professionals 323 0.0 0.1 0.0 0.0 Low Personal care workers in health services 532 0.1 0.0 0.2 0.3 0.1 Low Tellers, money collectors and related clerks 421 0.2 0.5 0.5 0.3 0.2 Low Machinery mechanics and repairers 723 0.2 1.3 2.2 1.6 0.9 Low Legislators and senior officials 111 0.5 0.1 0.7 0.4 Low Cashiers and ticket clerks 523 0.1 0.1 0.8 0.4 0.1 Low Nursing and midwifery associate professionals 322 0.0 0.0 0.7 0.2 0.3 Low Medical and pharmaceutical technicians 321 0.2 0.0 0.1 0.1 0.1 Low Mining and mineral processing plant operators 811 0.0 0.0 0.2 0.1 0.2 Low Veterinary technicians and assistants 324 0.0 0.0 0.0 0.0 Low Professional services managers 134 1.0 0.1 0.2 0.3 0.0 Low Finance professionals 241 3.1 0.4 1.5 0.2 1.3 Low Traditional and complementary medicine professionals 223 0.0 0.0 0.0 0.0 Low Other services managers 143 0.5 0.1 0.2 0.1 0.3 Low Manufacturing, mining, construction, and distribution managers 132 4.6 0.0 0.9 1.0 0.1 Low Business services and administration managers 121 6.7 0.2 1.5 0.6 0.1 Low Client information workers 422 3.1 0.2 0.5 0.5 0.2 Low Other teaching professionals 235 0.1 0.3 0.7 0.2 0.2 Low Refuse workers 961 0.4 0.5 0.3 0.3 Low Locomotive engine drivers and related workers 831 0.0 0.0 0.0 0.0 Low Material-recording and transport clerks 432 1.4 0.1 0.9 0.8 0.3 Low Social and religious professionals 263 0.1 0.1 0.2 0.1 0.1 Low Sales, marketing and development managers 122 5.2 0.1 0.4 0.2 0.0 Low Production managers in agriculture, forestry and fisheries 131 0.0 0.0 0.1 0.0 0.0 Low Mining, manufacturing and construction supervisors 312 0.4 0.4 1.3 0.5 0.1 Low Shop salespersons 522 4.7 9.1 5.7 7.0 5.1 Low Life science technicians and related associate professionals 314 0.1 0.0 0.1 0.0 0.0 Low Building finishers and related trades workers 712 0.0 1.2 0.9 1.4 0.2 Low Mining and construction labourers 931 0.1 2.5 2.2 1.7 2.3 Low Legal, social and religious associate professionals 341 0.2 0.0 0.2 0.1 0.1 Low Business services agents 333 0.8 0.2 0.4 0.3 0.2 Low Artistic, cultural and culinary associate professionals 343 1.4 0.1 0.3 0.2 0.1 Low Transport and storage labourers 933 0.6 1.0 0.2 1.1 1.0 Low Travel attendants, conductors and guides 511 0.0 0.0 0.1 0.1 0.0 Low Market gardeners and crop growers 611 9.6 5.1 21.7 4.8 Very low Animal producers 612 6.8 0.2 2.5 1.8 Very low Mixed crop and animal producers 613 0.0 0.0 0.2 0.0 Very low Forestry and related workers 621 0.4 0.0 0.2 0.0 Very low Fishery workers, hunters and trappers 622 1.3 0.7 0.8 1.8 Very low Subsistence crop farmers 631 12.2 0.0 3.1 1.1 Very low Subsistence livestock farmers 632 3.5 0.0 0.0 0.1 Very low Subsistence mixed crop and livestock farmers 633 0.0 0.0 0.0 0.1 Very low Subsistence fishers, hunters, trappers and gatherers 634 1.1 0.2 0.2 0.0 Very low Agricultural, forestry and fishery labourers 921 0.0 4.8 4.8 2.8 30.0 Very low Car, van and motorcycle drivers 832 0.2 2.6 1.5 3.3 1.9 Very low Legal professionals 261 0.2 0.0 0.2 0.2 0.1 Very low Waiters and bartenders 513 0.2 0.7 2.6 1.0 0.5 Very low Child care workers and teachers' aides 531 0.1 0.1 1.6 0.2 0.2 Very low Food preparation assistants 941 0.3 0.2 0.5 0.3 0.2 Very low Hairdressers, beauticians and related workers 514 0.2 0.6 0.8 0.8 0.7 Very low Other sales workers 524 0.8 0.8 1.6 3.0 0.5 Very low Textile, fur and leather products machine operators 815 0.0 0.1 0.1 1.3 3.7 Very low Vehicle, window, laundry and other hand cleaning workers 912 0.7 0.2 0.4 0.1 Very low Cooks 512 0.1 0.6 1.5 1.9 0.7 Very low Sales and purchasing agents and brokers 332 6.1 0.1 2.3 0.5 0.1 Very low Street and market salespersons 521 4.8 4.2 3.5 7.1 Very low Street and related service workers 951 0.0 0.0 0.0 0.1 Very low Street vendors (excluding food) 952 0.4 0.3 1.1 0.2 Very low 58 Share in Share in employment Digital ISCO Occupation name job occupation code Cambodia Malaysia Thailand Vietnam vacancies level Food processing and related trades workers 751 0.1 1.2 1.2 1.7 1.3 Very low Nursing and midwifery professionals 222 0.4 0.2 0.2 0.4 0.1 Very low Veterinarians 225 0.0 0.0 0.0 0.0 0.0 Very low Other health professionals 226 0.7 0.1 0.3 0.3 0.1 Very low Hotel and restaurant managers 141 0.3 0.2 0.5 0.3 0.0 Very low Other personal services workers 516 0.1 0.1 0.1 0.1 0.4 Very low Retail and wholesale trade managers 142 0.4 0.0 0.5 0.5 0.1 Very low Mobile plant operators 834 0.0 0.5 0.8 0.5 0.3 Very low Other elementary workers 962 0.2 0.7 1.5 0.9 0.9 Very low Primary school and early childhood teachers 234 0.3 0.8 1.9 1.4 1.5 Very low Building and housekeeping supervisors 515 0.1 0.0 0.2 0.7 0.5 Very low Domestic, hotel and office cleaners and helpers 911 0.1 0.6 2.1 1.4 0.7 Very low Secondary education teachers 233 0.1 0.5 1.4 0.6 1.0 Very low Garment and related trades workers 753 0.0 9.3 1.1 0.6 2.1 Very low Source: Authors’ calculations based on SEAD data. Notes: The 27 occupations in bold are the ones whose skills profile is imputed from one or several similar occupations (see section 3). 59 Annex 7. Robustness Checks: Results using the 5 clusters based on ascending digital occupation levels without merging clusters 2 and 3 Table A.7.1 and figures A.7.1. and A.7.2. show that the results presented in Tables 6 and Figures 6 and 7 are robust to merging the clusters based on ascending digital occupation levels. Table A.7.1. Conditional correlations between clusters based on ascending digital occupation levels and cognitive and socioemotional skills requirements (replication of Table 6) Clusters based on ascending digital occupation levels 2 vs. 1 3 vs. 1 4 vs. 1 5 vs. 1 3 vs. 2 4 vs. 2 5 vs. 2 4 vs. 3 5 vs. 3 5 vs. 4 (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) Cognitive Thinking 0.51 0.82* 1.35*** 2.78*** 0.06 0.78** 1.43** 0.14 0.38 0.01 (0.26) (0.40) (0.27) (0.53) (0.31) (0.24) (0.47) (0.11) (0.39) (0.15) Communication 0.21 0.91* 0.18 -0.12 0.33 -0.19 -0.43 -0.25 -0.52 0.01 (0.16) (0.42) (0.27) (0.21) (0.18) (0.46) (0.71) (0.15) (0.35) (0.33) Organization -0.11 0.30 -0.27 -0.81 0.26 0.16 1.25 -0.35** -0.43 0.12 (0.19) (0.37) (0.27) (0.55) (0.22) (0.35) (0.76) (0.12) (0.34) (0.10) Socioemotional Emotions 0.30 -0.35 1.30*** 1.08 -0.33 0.78 0.04 0.49*** 0.47 -0.30 (0.16) (0.33) (0.25) (0.54) (0.19) (0.44) (0.66) (0.11) (0.33) (0.21) Relationships -0.43** -0.12 -1.44*** -0.89 0.26 -0.60 -1.00 -0.39 -0.80* 0.08 (0.14) (0.25) (0.30) (0.57) (0.26) (0.54) (0.53) (0.19) (0.30) (0.38) Personal growth -0.32* 0.09 -1.09*** -0.73 -0.05 -0.76 -0.92 -0.13 0.22 0.36* (0.13) (0.30) (0.29) (0.54) (0.31) (0.48) (0.68) (0.17) (0.31) (0.18) Constant 1.48*** 1.17*** 2.49*** 1.98** 2.22*** 2.90*** 2.69*** 3.79*** 4.11*** 4.22*** (0.15) (0.26) (0.33) (0.72) (0.13) (0.36) (0.65) (0.11) (0.61) (0.22) N 58 59 72 51 41 54 33 55 34 47 R-sq 0.34 0.50 0.67 0.44 0.36 0.34 0.26 0.55 0.36 0.13 Sources: Authors’ calculations based on SEAD data for Malaysia. Note: The regressors for the outcome are dummy variables taking the value of one if the skills count of a skills category of an occupation is above the mean. The 2017 Malaysian labor force survey does not report any workers in one of the 127 occupations from our occupational skills profile (paramedical practitioners, ISCO code 224). See subsection 3.2. for details on the clustering. *p < 0.1 **p < 0.05 ***p < 0.01 60 Figure A.7.1. Composition of skills categories and subsets across clusters based on ascending digital occupation levels (replication of Figure 6.B) Source: Authors’ calculations based on SEAD data. Notes: The share of skills splits between digital, cognitive, and socioemotional, and other skills. Digital skills levels and cognitive and socioemotional skills subsets are subcategories counted in the total. See subsection 3.2. for details on the clustering. Figure A.7.2. Employment across clusters based on ascending digital occupation levels across countries (replication of Figure 7) Source: Authors’ calculations based SEAD data. Notes: See subsection 3.2. for details on the clustering. 61