17 Author’s calculations based on data in Dingel, et al, 2020.
18 De Vries, et al, 2020.
19 Author’s calculations based on data in Dingel, et al, 2020.
20 Zhao, 2020.
21 Job-seekers searching for roles on the LinkedIn platform using built-in remote job filters, normalized against changes to all job searches.
22 The share of job postings, which use number of keywords (i.e. ‘remote work’, ‘work from home’, home office’) in 10 different languages, as well as built-in remote job filters.
23 LinkedIn analysed data from job search behaviour and job postings of full-time roles and its changes due to COVID-19 during the period of 11 February to 1 July. Analysts utilized the ‘remote work’ filter and a set of searchable key words such as ‘remote work’, ‘work from home’, ‘homeoffice’ in 10 different languages. The index is the start of the analysis period, 11 July. Results are normalized for platform growth as well as in the case of job searchers against the volume of job searches. The daily figures represent a seven-day smoothed proportion.
24 Kimbrough, 2020.
25 Mongey, et al, 2020.
26 World Bank, 2020.
27 Cook, et al, 2019.
28 ADP provides human capital management services to significant numbers of US companies. Its data can therefore act as a reliable proxy for changes to the American labour market.
29 Workers are considered to have dropped out of employment if they disappear from the ADP database. While some of those variations can reflect worker movements to companies which do not use ADP’s services, the scale of that effect is not typically as large; therefore, on the basis of past trends we can deduce that what we are reporting are reach changes to employment.
30 Data from FutureFit AI combines over 50 data sources on workforce demand and supply, translating a range of taxonomies of jobs and skills. Supply-side sources include over 350 million talent profiles listing 30,000 skills clusters, 80,000 job titles, hundreds of industries, thousands of learning opportunities and millions of companies worldwide. The data set used comes from worker profile information sourced from resumes and online professional profiles. It also includes key data points for the analysis—such as employers, start and end dates, job role, industries and employment sequence, among others.
31 This metric covers approximately 300,000 young professionals in the United States, defined here as those who have graduated with an upper secondary or tertiary (undergraduate) degree no earlier than 2008, and have held 15 or less positions and have not been in the labour market for longer than 20 years. These professionals have, on average, eight years of work experience after or during a student’s first degree. The average work experience tenure following graduation is 6.7 years. The overwhelming majority of this sample are in their first working decade.
32 Agopsowicz, 2019.
33 See, for example: Arntz, Melanie, Terry Gregory and Ulrich Zierahn, The risk of automation for jobs in OECD countries: a comparative analysis, OECD Social, Employment and Migration Working Papers No 189, OrganizationOrganization for Economic Cooperation and Development (OECD), 2016; McKinsey Global Institute, A Future That Works: Automation, Employment, and Productivity, McKinsey Global Institute (MGI), 2017; PwC, Will robots really steal our jobs? An international analysis of the potential long term impact of automation, 2018. For a range of relevant additional considerations, see: van der Zande, Jochem, et al., The Substitution of Labor: From technological feasibility to other factors influencing job automation, Innovative Internet: Report 5, Stockholm School of Economics Institute for Research, 2018.
34 Ding and Saenz Molina, 2020.
35 World Economic Forum, 2020a.
36 For more details on how the clusters are computed please refer to World Economic Forum, 2020a.
37 For an in-depth analysis of emerging jobs please see World Economic Forum, 2020a.
38 According to Coursera data from individuals completing reskilling and upskilling on its platform, working towards a new skill in Cloud Computing could take on average 106 full calendrical days; in Content, 24 days; in Data and AI professions, 60; in Engineering, 77 days; in Marketing, 39; People and Culture, 36; Sales. 37; and in Product Development professions, 44. We take the average month to have 21 working days.
39 Sweetland, 1996.
40 Hsieh, et al., 2019.
41 IMF, 2020.
42 Atlantic Council, 2020.
43 Gentilini, et al, 2020.
44 Economic Security Project, 2020.
45 OECD, 2020b.
46 Cahuc, et al, 2006, and Carroll, et al., 2016.
48 Skills Future Enhanced Training Support Package”, https://www.enterprisejobskills.sg/content/upgrade-skills/enhanced-training-support-for-SME.html.