In a context where the alignment of training supply and demand has become crucial to meet the changing needs of today’s labour markets, few studies analyse whether training providers offer relevant course content and how training is delivered. This chapter aims to address the knowledge gap using a new cross-country dataset of education and training courses for adults in Australia, Germany, Singapore and the United States and applying a novel methodology to classify courses as green-related or AI‑related. The chapter also provides insights into the flexibility and accessibility of training supply in these four countries.
Training Supply for the Green and AI Transitions
Equipping Workers with the Right Skills
Abstract
Studies about training needs often focus on the demand side – the skills and knowledge that employers are seeking or struggling to find in the labour market. However, there is relatively little literature exploring the supply side of training needs – specifically, whether training providers offer relevant course content and how training is delivered. This chapter aims to address the knowledge gap regarding the nature of existing training provision. To do so, it uses a new cross-country dataset of education and training courses for adults in Australia, Germany, Singapore and the United States to characterise current training supply. Moreover, the study employs a novel methodology for classifying courses as related to the green or AI transitions based on text analysis of course titles and descriptions. The chapter begins by describing the course databases in each country and the methodology used to classify courses as green-related or AI-related. Subsequently, it provides insight into the current state of training provision across the four countries, including, in particular, insights into the flexibility and accessibility of training supply.
Data were gathered from catalogues of a mix of formal and non-formal education and training courses in Australia, Germany, Singapore and the United States. Catalogues provide prospective adult learners with information about available courses within each respective country. These countries were chosen due to a combination of data availability and language, with preference given towards training catalogues in English, to align with the English keyword lists used in the text analysis. The catalogues obtained from the countries span across the entire adult education system, with the range of courses compiled in these catalogues varying across countries, therefore providing insights into the supply of courses across different education sectors. For instance, in Australia, the catalogue encompasses all vocational education and training courses, whereas the catalogue in the United States includes general education courses spanning secondary to post-secondary levels. Despite differences in coverage, each catalogue offers an overview of education and training opportunities for adults. Table 2.1 below outlines the data fields extracted for each country based on availability. The remainder of this section describes in more detail the coverage of each country’s catalogue.
Note: (*) Denotes that there were too few observations to conduct reliable analysis using this data field.
Source: OECD compilation.
Australia
Data for Australia were provided by the National Centre for Vocational Education Research (NCVER). The NCVER collects data annually (and quarterly for government-funded training) about the provision of vocational education and training (VET) in the country (called the National VET Provider Collection). The NCVER also has a Total VET Students and Courses (TVA) database which gathers information on learner characteristics.1
The OECD obtained deidentified unit record data for the above two data sources in November 2023 (for the academic year 2022) and created a linked dataset using the subject ID number, a unique code used to identify a unit of competency. In the Australian VET sector, units of competency are nationally agreed statements of the skills and knowledge required for effective performance in a particular job or job function, developed through a process of national consultation with industry. Units of competency are endorsed components of training packages and can be combined into nationally recognised qualifications that comply with the Australian Qualifications Framework (AQF). Throughout this report, the term ‘courses’ will be used synonymously with the term ‘units of competency’.
Germany
Data for Germany has been sourced from DIPF, the Center for Educational Research and Educational Information, which hosts the IWWB database.2 This database consolidates vocational and educational training offers from across the country. IWWB collects its data from search portals that aggregate training offers from education providers across all German Bundesländer (states). The information is regularly and automatically updated. Currently, 79 education and training databases contribute data to IWWB, covering the large majority of VET courses in Germany, with 14 of these databases being privately operated. The IWWB database includes only courses open to the adult public and excludes courses leading to a bachelor’s or master’s degree, although some courses are offered by higher education institutions.
DIPF exported the IWWB data for the OECD in March 2024. Due to data confidentiality reasons, DIPF conducted the text matching process in-house, after receiving the OECD keyword lists described below. The OECD received a dataset which contained the courses identified to be green, AI or both green and AI, with their accompanying keyword matches.
Singapore
Data for Singapore include adult training courses published by SkillsFuture Singapore on its public website.3 This national skills portal helps citizens understand their own current skills, identify their skills gaps toward their target careers, and receive training recommendations. Individual education institutions supply information about their courses to SkillsFuture Singapore for publication. All courses advertised on the public website are eligible for the SkillsFuture Credit, which entitles adults aged 25+ to a subsidy of SGD 500 towards upskilling with the possibility of further top-ups. In May 2024, a new SkillsFuture (Mid-Career) top up of SGD 4 000 will further support adults aged 40 and above, available to use for selected courses, towards upskilling and reskilling. SkillsFuture Singapore exported the data for the OECD in July 2024.
United States
Data for the United States is collected by Credential Engine,4 a non-profit organisation committed to improving transparency about credentials and training opportunities across the country. There are two ways that Credential Engine collects data: through partnerships with regional, state and federal governments that gather information about training opportunities in their jurisdiction and upload them to Credential Engine (to date, there are over 30 such partnerships), and by inviting education and training providers to upload information directly. To facilitate data collection and transparency, Credential Engine developed a common language for describing credentials – the Credential Transparency Description Language (CTDL). This is available on their website for anyone to use with databases, tools, systems, for the benefit of learners, businesses and researchers.
The landscape of credentials and training opportunities published by Credential Engine’s partners is vast and includes those offered by post-secondary education institutions, non-academic providers, massive open online course providers, secondary schools and other types of organisations. Training opportunities in the dataset fall into three categories: courses (35%), learning opportunities (65%), or learning programmes (<1%). A ‘course’ denotes a singular, organised sequence of educational activities designed to cultivate specific competencies. A ‘learning opportunity’ encompasses both structured and unstructured learning experiences, drawing from direct involvement, formal and informal study, observation, and participation in discourse and practical application. Meanwhile, a ‘learning program’ constitutes a collection of learning opportunities that culminate in a defined outcome, typically a credential such as a degree or certificate. To avoid overlap and to facilitate comparison with other countries in this study, only information about learning opportunities is used. The data for this project was exported by Credential Engine in November 2023.
This study analyses data on training courses and whether these courses deliver content to support upskilling and reskilling for the green transition and for the greater adoption of AI at work. Training courses are classified as green-related or AI-related whenever the course title or course description contains at least one keyword from a list of pre‑identified green or AI keywords. In cases where the same course was offered in multiple education institutions or multiple regions, this course appeared multiple times in the database. No attempts were made to eliminate such duplicates in order to ensure all offered courses were included comprehensively. Some minor text cleaning was performed on the course descriptions before conducting the text matching process, including the conversion of all text to lowercase and the removal of punctuation.
Identifying green-related training courses
To identify green-related training courses, a keyword list was developed building on the O*NET Green Topics list. The O*NET Green Topics list was developed by the Green Economy programme of O*NET – the Occupational Information Network, sponsored by the United States Department of Labor. In 2022, the 72 green topics identified in this list were linked to 31% of occupations (up from only 14% of occupations in 2009), and to 20% of educational programmes in a United States database (Lewis, Morris and Gregory, 2022[1]).
The O*NET Green Topics list was chosen as a starting point for this study because it is well-known and widely used in the literature. Furthermore, keywords contained in this list include general concepts and knowledge areas rather than specific skills or tasks. This makes it relevant for an analysis of training programmes, as course descriptions can refer to both broad knowledge areas as well as specific skills.
For the scope of this research however, a few modifications were made to the O*NET Green Topics list. Some keywords were added from the European Skills, Competences, Qualifications and Occupations (ESCO) classification of green skills, another internationally recognised classification. For instance, ‘sustainable energy’ is not present in the O*NET Green Topics list but is in the ESCO list and was thus added as it was deemed relevant. Instead, the word ‘energy’ on its own was removed as this was deemed too broad to reliably identify green-related courses. For similar reasons, ‘health and safety’ was removed. The complete list of green keywords used in this study, including those which were added or removed from the Green Topics list, is contained in Annex B. A total of 125 keywords are used in this study.
Identifying AI-related training courses
To identify AI-related training courses in this report, lists of AI keywords used in two previous OECD working papers examining the demand for AI skills in jobs (Squicciarini and Nachtigall, 2021[2]; Borgonovi et al., 2023[3]) were combined. Each of these lists has been validated by experts. While the AI keywords list used in Squicciarini and Nachtigall (2021[2]) builds on previous OECD work assessing AI-related developments in science and technology, the AI keywords list used in Borgonovi et al. (2023[3]) exploits the AI skills classification developed by Lightcast. The complete list of AI keywords used in this study can be found in Annex B. There are 107 AI keywords in the first list and 218 AI keywords in the second list. There are 285 AI keywords in the combined list which is used in this study. Given the recent rise of AI chatbots, the terms ‘bard’, ‘bert’ and ‘ChatGPT’ were also added for completeness.
Validating keyword lists
After the initial text matching, random quality checks were conducted to validate the two keyword lists and identify potential false positive matches. It was found that the green keyword ‘wind’ often referred to ‘window’ and therefore a modification was made to require a space after the word ‘wind’ during the text search. Additional false matches, such as the AI keyword ‘bee colony’ (which is a heuristic algorithm for AI but it is also referenced in many agricultural related course descriptions), were also removed from the final data. Overall, this validation process did not significantly change the share of courses that were found to be green- or AI-related as modifications required were minor.
The descriptive analysis in this section draws on data for 22 393 VET courses in Australia for the 2022 academic year, 3 488 651 vocational and further educational courses in Germany as of March 2024, 47 560 adult learning courses in Singapore as of July 2024, and 32 307 training opportunities in the United States as of November 2023.5 As noted earlier in the Chapter, data for each country represents a different sector of education and training; as a result, cross-country comparisons should be performed and interpreted with caution.
A varied landscape of training providers ensures choice
Courses across the countries in our dataset are offered by a range of providers. This reflects the overall breadth of the adult learning landscape, with courses provided by both the public and private sectors, and across vocational, higher and non-formal education and training. The range of choice allows learners to compare and contrast courses and make training decisions based on their learning needs and desires.
Provider type
The course catalogues for Australia and the United States provide insights into the type of providers offering adult training. A range of providers deliver adult learning, including both public and private institutions. In Australia, the majority (64%) of courses are delivered by privately-operated registered training providers.6 Public institutions like technical and further education (TAFE) institutes – Australia’s largest, government-run vocational education and training provider – deliver 18% of courses in the catalogue. Schools and universities deliver 5% of courses, while another 5% are offered by enterprises. A minority of courses (8%) are delivered by other organisations including community-based adult education providers, industry associations and enterprises.
In the course catalogue for the United States, over half (57%) of courses are delivered by public institutions. Private not-for-profit organisations deliver around 39% of courses and private for-profit organisations deliver 5% of courses. The public institutions delivering training courses in the United States are mostly post-secondary schools (63%), with 32% offered by general education and training providers, and 2% offered by vendors and quality assurance organisations.
Flexible training options are essential to boost participation
A flexible adult learning system allows learners to better balance training alongside work or other commitments (OECD, 2023[4]). Indeed, a lack of time due to work or family responsibilities are the primary reasons adults report for not participating in adult learning (OECD, 2019[5]). The provision of more flexible learning options would also encourage greater participation in training by a wider group of adults, thus improving the inclusiveness of the system.
Factors that affect the flexibility of training include alternative study load options (such as the choice to study part-time) and the mode of delivery (for example, the possibility to take online courses). Other important aspects of flexibility – for example, whether courses are available on weekends or whether content was delivered in a modular format – are not available in the current dataset and therefore remain unstudied in this report.
Study load
The provision of part-time study options is one way to address barriers related to time and financial cost, which many adults experience, as reported in the OECD Survey of Adult Skills. Low-skilled adults in particular are more likely than medium- or high-skilled individuals to report a shortage of time due to family responsibilities as a barrier to training (OECD, 2023[4]). The option to study part-time can allow these individuals to maintain other commitments while training. Furthermore, the ability to continue working full-time while studying part-time can help to engage the 16% of adults who report a lack of financial resources as a barrier to their participation in training (OECD, 2019[5]).
In Australia, the overwhelming majority (94%) of learners enrolled in a VET course in 2022 completed their studies part-time. Almost three‑quarters of learners were actively working while undertaking training – with 42% employed full-time, 25% employed part-time, and 5% self-employed or working in a family business. Of the learners who study full-time, an equal share work part-time and full-time (about one‑quarter each). Of the learners who study part-time, 44% work full time and 25% worked part-time. This suggests that many learners are balancing both work and study, in some cases both full-time. Furthermore, unemployment or a break from work may be one factor related to the decision to study full time, with trainees enrolled full-time more likely than part-time learners to report that they were unemployed or out of the labour force. This may also reflect the choice to not work while studying full-time.
In Singapore, more than half (64%) of all courses are offered on a part-time basis, 18% are offered on a full-time basis, and 18% of courses allow the learner to decide between part-time and full-time. The latter option provides learners with the most flexibility in terms of study load, allowing them to study at a pace that suits their schedule and needs. These findings are consistent with the nature of Singapore’s continuing education and training system, which primarily caters to adult learners, who have a high preference for flexible learning schedules.
Delivery mode
Offering a variety of delivery modes can accommodate for a range of learning style preferences, and can reduce the time and cost of training by for instance reducing or eliminating commute times to and from classrooms (OECD, 2023[4]). Moreover, amongst adults who want to participate in adult learning, around 12% report an inconvenient time or place as a barrier to doing so (OECD, 2019[5]). Offering courses online, in hybrid, blended or self-paced formats, and allowing learners to study asynchronously, are all options to facilitate greater participation. Figure 2.1 highlights the variability of online learning provision across the countries studied.
In Australia, 12% of courses are taught entirely online (either self-paced or in real-time), while 42% are entirely classroom-based. Despite this relatively low rate of courses taught entirely online, about one in five (22%) courses are delivered using a combination of modes. Moreover, 9% of courses are offered as entirely work-based learning (i.e. conducted in the workplace but delivered either by the employer or the training provider). Consistent with this finding, about 10% of VET courses are undertaken by learners as part of an apprenticeship or traineeship. This shows that some learning is conducted while learners simultaneously work and study, allowing learners to gain useful and practical experience whilst potentially earning an income, rendering these training options highly accessible to lower income adults. Apprenticeships may also partly explain the high share of students both studying and working full-time in Australia.
In the United States, most courses are offered either completely online or in-person (41% each), with the remaining 18% adopting a blended delivery format. Blended (or hybrid) learning has become increasingly popular over recent years, allowing learners to benefit from face‑to-face interaction with their teacher and other learners, while also enjoying the flexibility of online learning (OECD, 2023[4]). Moreover, the high share of training offered online supports the overall trend post-COVID towards online learning (OECD, 2020[6]).
Training in the United States is also highly flexible regarding where and when learners learn, with almost half (49%) of courses following a self-paced format. Self-paced courses tend to be offered online, consistent with the high share of online‑only courses in the dataset. Learners undertaking self-paced courses have the benefit of studying asynchronously (where the instructor and learner(s) are not interacting with each other in real-time), allowing them to study at a time or place that suits them (OECD, 2023[4]). Finally, about one‑third (31%) of courses follow a more traditional, lecture‑based format, while 18% incorporate a combination of learning methods.
Learning is also delivered in a variety of formats in Singapore. About 31% of all courses are delivered online, with 12% offered asynchronously, and 14% offered synchronously (learning happens at the same time as teaching). Over half (52%) of courses are still delivered as traditional classroom facilitated training, while 13% are delivered on-the‑job or as practical training.
In Germany, the vast majority of courses (81%) are offered in presence (for instance in the form of seminars). This is followed by courses that are web- or computer-based (9%) or offered as distance learning courses – i.e. all types of learning at a distance and can take place both traditionally and digitally – (2%).
Note: Given underlying differences in data, cross-country comparisons are not advised.
Source: Australia’s NCVER vocational training database; Germany: IWWB course catalogue; Singapore: SkillsFuture Singapore course database; United States: Credential Engine database.
Expanding access promotes a more inclusive adult learning landscape
Making adult learning accessible for low-skilled and low-income adults is imperative to supporting a more inclusive future of work. While the changing nature of work affects everyone, adults with lower skills are likely to be most affected. This dataset allows us to measure a number of factors that affect the accessibility of training: access to funding, course duration, and whether a course has pre‑requisites. Moreover, learner-level data from Australia allows some reflection on the diversity of learner profiles.
Other factors that affect the accessibility of training, such as whether adults have access to personalised advice and guidance, whether recognition of prior learning is possible, if modular learning opportunities are available, or if adults can have paid time off from work to study, are not captured in the data.
Funding source
As noted above, 16% of adults report cost pressures as a reason for not participating in adult learning. Additionally, data from the OECD Priorities of Adult Learning (PAL) dashboard indicate that around one‑fifth of individuals contribute – either fully or partially – to the cost of their own training. Public funding of training can therefore alleviate some of the financial burden on individuals and widen access to training. Public support is also important to reduce the private cost pressures of training in countries where financial incentives for firms to offer training to their employees are limited (OECD, 2019[5]).
In Australia, 74% of courses are fee‑for-service programmes, meaning that learners or their employers pay for training. While this is a high share of courses which require private funding, some of these learners may be eligible for government subsidies or loans whereby costs are partially covered or deferred. About one‑quarter (26%) of courses attract government funding, usually those offered by TAFE institutes (i.e. the public sector).
Whilst data on funding is not available for other countries in our study, cost data is available for Singapore and the United States. In Singapore, the mean total cost of a course per learner is SGD 3 454 and the median is SGD 1 152. All courses in the SkillsFuture registry are eligible for the SkillsFuture Credit, a scheme which provides all Singaporeans aged 25 years and above with an opening credit of SGD 500.7 This means that for the median priced course, residents can receive a discount of almost half of the cost of learning. Adult learning subsidy schemes such as the SkillsFuture Credit increases the accessibility of adult learning, providing many learners with lower cost training options. In the United States, courses have a mean total cost of USD 9 840 and a median total cost of USD 5 800. This relatively high total cost can be explained by the larger share of higher education degree programmes captured in the course registry.
Duration
Lack of time remains the biggest factor limiting participation in adult learning across the OECD. Therefore, shorter courses may enable access to a wider range of learners. Figure 2.2 depicts the share of courses which last 50 hours or less across the three countries.8
In Australia, courses on average last 38 hours, while the median duration is 40 hours.9 Moreover, the vast majority of courses are short – with 96% of all courses lasting 100 hours or less, and 75% of all courses lasting 50 hours or less. This relatively short workload makes training an attractive and feasible option for many learners.
Similar to Australia, most courses in the Singaporean course catalogue are short, with a significant majority (69%) lasting only 50 hours or less. The average duration of courses is 124 hours, or about three weeks at a full-time study load, while the median is much shorter at 24 hours. This likely reflects the fact that many courses in the Singaporean course catalogue are shorter, adult learning courses.
By contrast, the United States course catalogue has a high share of full-length, higher education degree programmes which renders course duration quite high. On average, courses take 53 weeks to complete (the median is 39 weeks). Moreover, only less than 1% of all courses have an average duration of 50 hours or less, rendering courses in the United States much lengthier than other countries in this report.
Note: Given underlying differences in data, cross-country comparisons are not advised.
Source: Australia’s NCVER vocational training database; Singapore: SkillsFuture Singapore course database; United States: Credential Engine database.
Pre‑requisites
Whilst only 3% of adults report not having the required pre‑requisites as a barrier to training, this barrier disproportionately affects adults with low formal qualifications (OECD, 2019[5]; OECD, 2023[4]). Offering courses across a range of pre‑requisite levels would widen access to training. In Singapore (the only country for which we have data on entry requirements), 17% of courses have either no course pre‑requisite or only require learners to have a primary-level education, suggesting very low-skilled learners can reasonably access training. Another 11% require only a secondary school diploma to enter. The remaining courses require either a diploma (8%), a degree (8%), a certificate (13%), or a license (0.4%).10
While data on pre‑requisites is not available for Australia, the dataset has information on the previous educational attainment of learners, which can be an indicator for the level of difficulty or accessibility of courses. This data show that vocational education in Australia appears to be highly accessible for learners that come from lower skilled backgrounds, with only 20% of learners reporting that they have obtained a bachelor’s degree or higher level degree before enrolling into a VET course. Instead, most learners have either a diploma or certificate – one that can be typically awarded through the VET system – or have not achieved any further education beyond high school.
Diversity of learner profiles
Finally, data on learner enrolments in Australia sheds light on the diversity of learner profiles and therefore how accessible vocational education is for a range of learners. In 2022, of the over 3 million learners enrolled into at least one vocational education course,11 32% were not born in Australia, 4% were international learners, and 16% spoke another language besides English at home, indicating some racial, ethnic and cultural diversity in the VET system. Moreover, men and women enrol into vocational educational at roughly similar rates suggesting a good gender balance. The learner pool is also geographically diverse with 30% of learners residing in regional or remote Australia – that is, outside a major city – and about 5% of learners self-report as having a disability. This diversity of learner profiles suggests that accessibility of vocational education in Australia is relatively high.
Finally, the average age of learners is relatively evenly distributed, suggesting a range of learner age profiles, with the mean learner age estimated at 35 years old. The peak average age of learners is around 17 years old, consistent with the above finding that many learners entering the VET system have only a high school level education. This suggests that the VET system is particularly accessible for younger learners of lower educational backgrounds.
References
[3] Borgonovi, F. et al. (2023), “Emerging trends in AI skill demand across 14 OECD countries”, OECD Artificial Intelligence Papers, No. 2, OECD Publishing, Paris, https://doi.org/10.1787/7c691b9a-en.
[1] Lewis, P., J. Morris and C. Gregory (2022), Green Topics: Identifying Linkages to Occupations and Education Programs Using a Linguistic Approach, The National Center for O*NET Development, Raleigh.
[4] OECD (2023), Flexible adult learning provision, OECD Publishing, https://www.oecd.org/els/emp/skills-and-work/adult-learning/booklet-flexibility-2023.pdf.
[6] OECD (2020), “The potential of online learning for adults: Early lessons from the COVID-19 crisis”, OECD Policy Responses to Coronavirus (COVID-19), OECD Publishing, Paris, https://doi.org/10.1787/ee040002-en.
[5] OECD (2019), Getting Skills Right: Future-Ready Adult Learning Systems, Getting Skills Right, OECD Publishing, Paris, https://doi.org/10.1787/9789264311756-en.
[2] Squicciarini, M. and H. Nachtigall (2021), “Demand for AI skills in jobs: Evidence from online job postings”, OECD Science, Technology and Industry Working Papers, No. 2021/03, OECD Publishing, Paris, https://doi.org/10.1787/3ed32d94-en.
← 1. This data is available to registered users on the VOCSTATS tool at www.ncver.edu.au/research-and-statistics/vocstats.
← 2. Available at www.iwwb.de/kurssuche/.
← 3. Available at www.myskillsfuture.gov.sg.
← 4. Available at https://credentialengine.org/.
← 5. Note, descriptive analysis in this section is conducted only on non-missing data.
← 6. In Australia, only registered training organisations can deliver nationally recognised training. Training which is not nationally recognised can be delivered by any training provider, and primarily includes locally developed programmes or skill sets – created to meet an identified training need – and non-accredited modules. This data therefore indicates that most training in the Australian vocational sector is delivered by registered training organisations.
← 7. More information on the SkillsFuture Credit scheme can be found here: www.myskillsfuture.gov.sg/content/portal/en/career-resources/career-resources/education-career-personal-development/SkillsFuture_Credit.html (accessed 14 March 2024).
← 8. The definition of a short course is unclear. The average 15‑week long university course in the United States requires a minimum of 3 contact hours per week, or 42 hours per semester (www.aic.edu/academics/credit-hours-calculator/). In the European Union, under the European Credit Transfer and Accumulation System, the average 3‑credit university course requires about 75 learning hours per semester, with 25 contact hours and about 50 hours of individual study time (https://data.europa.eu/doi/10.2766/87192). Therefore, contact hours for the average university course can vary across countries. Therefore, for simplicity, we deem a course to be short if it lasts less than 50 hours.
← 9. Duration refers to hours of supervised training deemed necessary to conduct the required training and assessment activities.
← 10. Due to data difficulties, we were unable to classify the entry requirements of a high share of courses (35%) in the Singaporean registry.
← 11. This includes learners who may have eventually dropped out of their programmes or discontinued their studies.