This chapter defines advanced science, technology, engineering and mathematics (STEM) skills as disciplinary, interdisciplinary, transversal and professional competence applied to complex scientific and technological problems and responsible innovation, including ethical, societal and environmental implications. It explains why these skills matter, but also why reported shortages require careful interpretation. It examines the benefits and limitations of using STEM as a policy construct and introduces a framework for diagnosing and assessing five interconnected bottlenecks in advanced STEM skills: entry into higher education; persistence and completion in higher education; progression to advanced levels; career entry and effective skills use; and skills renewal across working life.
Policies for Developing Advanced STEM Skills in Higher Education
Abstract
Advanced STEM skills comprise four dimensions of competence: disciplinary, interdisciplinary, transversal and professional. These skills develop cumulatively across education and working life, and often their development is not linear as learners may move between education, research and employment, interrupt and resume learning, or move across fields and occupations.
The purpose of using advanced STEM skills as a concept is not to introduce a new statistical category or draw a fixed boundary around particular qualifications or occupations. It is to provide an analytical lens for examining how these skills are developed, deepened, connected, used and renewed; where barriers arise; and which policy interventions may address them effectively.
Shortages of advanced STEM skills are difficult to measure and interpret. Recruitment difficulties may reflect gaps, but also demand for experienced specialists, narrow recruitment criteria, working conditions or employers’ limited capacity to fully recognise and use available skills.
Five interconnected bottlenecks can constrain the development and effective use of advanced STEM skills. These concern entry into STEM higher education; persistence and completion; progression to advanced levels; career entry and effective skills use; and skills renewal across working life. The framework helps identify where a bottleneck becomes visible and examine the individual, institutional, organisational and system conditions that may produce it. Constraints can accumulate across education and working life or be transferred from one bottleneck to another.
Opportunities to develop, demonstrate, use and renew advanced STEM skills are unequally distributed. Gender, socio-economic background, geography, disability and migration background can interact and reinforce one another, shaping access to preparation, education, professional experience during study and access to suitable jobs. Addressing these barriers while maintaining high standards can broaden the talent base in STEM education, research and innovation.
Discovery often begins without a clear view of whether or how the resulting knowledge will find practical application. Curiosity-driven inquiry enables people to investigate the unknown and question existing explanations, producing knowledge whose importance may become clear only over time. Many advances that transform health, technology, the economy and daily life originate in attempts to understand phenomena rather than meet an immediate practical need.
Science and technology expand what societies are capable of doing, but they do not determine which possibilities should be pursued, how competing objectives should be reconciled or how benefits, risks and responsibilities should be distributed (Feynman, 1955[1]). These decisions require ethical and societal judgement as well as scientific and technical knowledge. They also require openness about uncertainty, critical scrutiny of assumptions and evidence, and a willingness to identify and correct error. These requirements become particularly important as discoveries give rise to technologies that move from research into complex economic and societal settings, where their consequences depend on how they are designed, interpreted, introduced and governed.
Consider a hospital introducing an artificial intelligence (AI) system to support clinical decisions. Computer scientists and statisticians may develop the model, but clinical specialists must determine whether its outputs are medically meaningful. Engineers and data experts must connect it with existing systems and assess whether the underlying information is reliable. Legal and ethical expertise is needed to address privacy, accountability and bias. Health professionals must understand when to use the system, when to question its outputs and how to explain decisions to patients. Psychologists can contribute knowledge of how people interpret and respond to automated advice. Designers, artists and specialists in visual communication can help make the human-machine interface comprehensible and examine how the presentation of information may influence behaviour. Managers must co-ordinate these contributions and determine how the technology fits into clinical practice.
None of these individuals need to possess every relevant form of expertise. General AI literacy may enable participants to understand that an AI system has limitations and should not be followed uncritically. Advanced STEM skills are required to determine whether the data and model are appropriate for the clinical purpose, validate performance across relevant populations, integrate the system with clinical and organisational processes, interpret uncertainty, redesign the system or its use where necessary, and assume professional responsibility for consequential decisions. The institutional and system conditions matter for the development of advanced STEM skills, and this is where policymaking can make a difference by investing in education and research, widening access to substantive learning and research opportunities, and sustaining the people, infrastructure and institutional capacity on which scientific and technological progress depends.
What are advanced STEM skills?
Advanced STEM skills concern the depth, combination and application of competence needed to address complex scientific and technological problems. They are not limited to specific fields of study, qualifications or occupations. Based on analysis of existing literature and evidence from practice, this report defines advanced STEM skills as a combination of four dimensions of competence:
Disciplinary competence provides depth in a scientific, technological, engineering or mathematical field, including command of its concepts, methods and standards of evidence.
Interdisciplinary competence enables people to integrate concepts, methods, evidence and perspectives from different fields around a shared problem.
Transversal competence enables people to communicate, collaborate, think critically, solve problems and continue.
Professional competence enables people to apply knowledge and methods responsibly in occupational, organisational and regulatory settings. It includes the capacity to contribute to innovation, entrepreneurship and knowledge transfer, and to work at the interfaces between science, policy, regulation and public administration, including by translating scientific and technical evidence for different purposes.
The relative importance of the four dimensions of competence varies across roles and situations. Education can develop all four dimensions through disciplinary and interdisciplinary study, authentic problems, research, practical experience and engagement with professional settings. Their further development and integration continue through research and work, as people apply their competence under specific organisational, institutional and regulatory conditions and assume greater autonomy and responsibility. Professional competence may, for example, involve developing and scaling an innovation, transferring knowledge between research and application, creating or managing an enterprise, advising policymakers, designing or implementing regulation, or applying scientific evidence in public administration.
Opportunities to undertake substantive work, exercise judgement and assume increasing responsibility enable people to develop more advanced levels of this competence. Its continued development therefore depends not only on the competence individuals possess, but also on whether organisations create roles and working conditions in which it can be used and extended. It requires people to recognise uncertainty and the limits of their competence, consider the ethical and societal consequences of their decisions, understand the institutional and regulatory contexts in which decisions are made, and remain accountable for how they interpret, communicate and use evidence.
Different levels of STEM skills can be distinguished according to the depth of knowledge involved, the complexity and familiarity of the context in which knowledge is applied, and the degree of integration, autonomy and judgement required (Figure 1.1.).
Advanced STEM skills differ from foundational and intermediate STEM skills in the depth, integration and judgement with which the competence dimensions are applied to a task. They enable people to work with greater disciplinary depth and to integrate knowledge across fields, address unfamiliar or complex problems, evaluate uncertainty and limitations, exercise autonomy and professional judgement, and develop, adapt or govern new knowledge and technologies. The staircase illustrates how these dimensions may be developed, combined and renewed across education, research and work.
Source: Goos, Carreira and Namukasa (2023[2]), “Mathematics and interdisciplinary STEM education: Recent developments and future directions”, https://doi.org/10.1007/s11858-023-01533-z Archer et al. (2015[3]), “‘Science capital’: A conceptual, methodological, and empirical argument for extending Bourdieusian notions of capital beyond the arts”, https://doi.org/10.1002/tea.21227; Global STEM Alliance (2025[4]), STEM Education Framework Research Base, https://www.nyas.org/wp-content/uploads/2025/05/STEM-Education-Framework-Research-Base.pdf; and OECD (2026[5]), PISA 2025 Assessment and Analytical Framework, https://doi.org/10.1787/86c36975-en.
Scientific literacy and AI literacy provide essential foundations for advanced STEM skills. They enable individuals and organisations to assess the basis, reliability and limitations of scientific claims, reported evidence, and importantly outputs generated by AI. Definitions for both literacies exist. Scientific literacy includes understanding how scientific knowledge and evidence are produced and evaluated (OECD, 2026[5]), while AI literacy includes understanding, using and critically evaluating AI systems and considering their ethical and societal implications (OECD/European Commission, 2026[6]; UNESCO, 2024[7]; Long and Magerko, 2020[8]). Basic scientific literacy is measured internationally among 15-year-old learners through the OECD Programme for International Student Assessment (PISA), while frameworks and assessment instruments for AI literacy remain under development, including for the planned PISA 2029 assessment of media and AI literacy.
The purpose of using advanced STEM skills as a concept is not to introduce a new statistical category or draw a fixed boundary around particular qualifications or occupations. It is to provide an analytical lens for examining how disciplinary, interdisciplinary, transversal and professional competence is developed, deepened, connected, used and renewed; where barriers arise; and which policy interventions may address them effectively.
Disciplinary depth and integration across fields
Many contemporary challenges in STEM fields and beyond have technical, environmental, social and organisational aspects that cannot be addressed adequately through a single disciplinary perspective. Addressing them effectively (and efficiently) requires people to extend their disciplinary competence into adjacent fields or combine it with knowledge and methods from other domains. Interdisciplinary learning commonly involves integration around substantive problems, inquiry, design and collaboration (Portillo-Blanco et al., 2024[9]; Rana, Aitken and Chimoriya, 2025[10]).
Interdisciplinary competence can be described as a set of skills that allow an individual to apply disciplinary knowledge while integrating concepts, methods, evidence and perspectives from different fields around a shared problem, research question or design challenge. It requires more than exposure to several disciplines. Essential is to understand how the contributing fields define problems, produce and evaluate evidence, and assess possible solutions, as well as how their contributions can be combined. Interdisciplinary competence therefore depends on sufficient disciplinary depth to judge the quality and limitations of each contribution (Ashby and Exter, 2018[11]; Portillo-Blanco et al., 2024[9]).
Mathematics can support integration of knowledge, methods and evidence from different fields through modelling, abstraction, quantitative reasoning, and the analysis of relationships, variability and uncertainty. However, it may be treated only as a service tool within work framed by another discipline. Effective integration requires mathematics to contribute to problem formulation, model construction, analysis and interpretation rather than only to calculation (Goos, Carreira and Namukasa, 2023[2]; Makonye and Moodley, 2023[12]). This explains the central role of mathematics in advanced STEM skills.
There has been quite some development in the introduction of interdisciplinary programmes in STEM higher education. Table 1.1 distinguishes disciplinary, multidisciplinary, interdisciplinary and transdisciplinary provision as analytical categories and not part of a hierarchy. Programme titles do not typically provide information on the disciplinary depth developed, how knowledge and methods are integrated or which educational and professional routes a qualification provides access to. It is therefore needed to make explicit what programmes contain.
Source: Based on Kelley and Knowles (2016[13]), “A conceptual framework for integrated STEM education”, https://doi.org/10.1186/s40594-016-0046-z; Tytler, Anderson and Williams (2023[14]), “Exploring a framework for integrated STEM: Challenges and benefits for promoting engagement in learning mathematics”, https://doi.org/10.1007/s11858-023-01519-x; and Goos, Carreira and Namukasa (2023[2]), “Mathematics and interdisciplinary STEM education: Recent developments and future directions”, https://doi.org/10.1007/s11858-023-01533-z.
Advanced STEM skills as a strategic policy priority
Advanced STEM skills create economic and societal value through their effective development, renewal and application. For this, people need to be able to enter and complete high-quality STEM education, progress to advanced levels, gain access to substantive professional experience and suitable employment, and renew their skills as knowledge, technology and professional practice change. Employers need to be able to interpret relevant competence, recruit people with appropriate foundations, provide effective onboarding and organise work so that advanced STEM skills inform professional practice and organisational decisions. There are several bottlenecks that effective policymaking needs to address.
Demand for advanced STEM skills is increasing across countries with implications for education and employment. The 2024 European Employment Services (EURES) report on labour shortages and surpluses, for example, identified widespread shortages in engineering, software development and other information and communication technology occupations, as well as in construction and healthcare. Employer surveys, vacancy data and sectoral assessments can help policymakers identify where pressures are concentrated. However, these sources do not always distinguish clearly among insufficient supply of specific skills, broader labour shortages, mismatches between employer requirements and candidate profiles, and recruitment difficulties associated with pay, working conditions or location (EURES, 2024[15]; European Commission, 2023[16]; OECD, 2022[17]).
In the United Kingdom, for example, the 2025 Quantum Skills Taskforce identified shortages in engineering and technical occupations needed to develop and diffuse quantum technologies. Many of these occupations are also in demand elsewhere in the economy, creating competition among quantum organisations and other sectors for people with relevant skills. The Taskforce therefore emphasised specialised education and training, stronger connections among education, research and industry, and the development of technical roles supporting the wider adoption of quantum technologies (UK Government, 2025[18]).
In Japan, the adoption of AI has increased demand for workers able to develop, implement and use AI systems. A 2025 Linux Foundation survey found a positive net hiring effect from the introduction of AI among the organisations surveyed but also reported that more than 70% of employers lacked sufficient employees with AI-related skills. Japan’s response places considerable emphasis on skills renewal among the existing workforce, including opportunities for workers to develop digital and AI-related competence during their careers. This reflects the importance of responding to technological change through continuing education and workplace learning as well as through initial education (The Linux Foundation, 2025[19]).
Korea has combined domestic skills development with measures intended to attract international researchers and postgraduate learners in strategic fields, including semiconductors, biotechnology, quantum technologies and AI. The government has expanded its long-running Brain Pool programme and introduced financial and immigration-related incentives intended to attract international researchers, specialists and postgraduate learners (Byeongseon, 2026[20]).
Australia faces reported shortages across a range of specialised engineering fields, including aeronautical, biomedical, electrical, environmental, geotechnical, marine, mining, petroleum and water engineering. Engineers Australia has called for a national engineering workforce strategy connecting government policy, higher education funding and changing industry needs (Engineers Australia, 2025[21]).
The OECD Research and Innovation Careers Observatory provides a complementary evidence base for examining these questions across countries. Its comparative indicators cover the size and composition of the research and innovation workforce, talent renewal, employment across different parts of the labour market, career outcomes, working conditions and international mobility. This evidence can help policymakers distinguish constraints affecting the development, circulation, employment and retention of research and innovation personnel rather than interpreting reported pressures solely as shortages of graduates (OECD, 2026[22]).
Europe’s Education Strategic Plan: Skills for Competitiveness and Innovation
For the European Union, advanced STEM skills have gained particular importance. They are considered crucial to Europe’s capacity to raise productivity, reduce innovation gaps and advance the green and digital transitions. Demographic change adds to this challenge by reducing the working-age population and increasing the importance of broadening participation, supporting skills renewal and making effective use of the competence already available.
The European Commission’s 2025 STEM Education Strategic Plan: Skills for Competitiveness and Innovation, adopted as part of the Union of Skills, reflects the growing strategic importance attached to STEM education in Europe (European Commission, 2025[23]). Its actions are organised around three objectives: establishing STEM as a strategic pillar of European education, skills and economic policy; strengthening and broadening the STEM talent base; and reducing barriers to participation, particularly for girls and women (Box 1.1).
The European Commission adopted the Communication “A STEM Education Strategic Plan: Skills for Competitiveness and Innovation” in March 2025 as part of its Union of Skills strategy. The Plan responds to concerns about Europe’s productivity and innovation performance, changing demand for skills in key technology sectors, a shrinking working-age population, declining performance in mathematics and science, persistent gender disparities and difficulties attracting and retaining STEM talent. It connects STEM education and training with competitiveness, the green and digital transitions, preparedness, technological leadership and strategic autonomy, and organises action around three objectives:
Lead: establish STEM as a strategic pillar of EU education, skills and economic policy. Proposed actions includes encouraging national and regional STEM strategies, strengthening intelligence on emerging skills needs and graduate outcomes, and creating a European STEM Executive Panel to advise on strategic issues and improve cross-portfolio policy co-ordination.
Level up: develop a stronger and more inclusive STEM talent base. The Plan seeks to improve the quality and relevance of STEM education, strengthen connections between education, research and employment, and support learning in fields important to Europe’s future competitiveness. Proposed actions include developing an EU STEM competence framework, supporting joint programmes and specialist training in strategic sectors, and exploring a European degree for engineers.
Lift barriers: widen participation and address obstacles affecting underrepresented groups.
The Plan proposes EU-level targets for 2030 to guide progress at EU level and to be supported by national or regional strategies.
Area | Participation target | Inclusion target |
|---|---|---|
STEM in initial medium-level vocational education and training | At least 45% of learners | At least 25% female participation |
STEM in tertiary education | At least 32% of students | At least 40% female participation |
ICT doctoral education | At least 5% of students | At least one-third female participation |
Source: European Commission (2025[23]), A STEM Education Strategic Plan: Skills for Competitiveness and Innovation, COM(2025) 89 final, European Union, Brussels, eur-lex.europa.eu/legal-content/EN/TXT/?uri=celex:52025DC0089.; European Commission (2025), Communication from the Commission to the European Parliament, the Council, the European Economic and Social Committee of the Regions: The Union of Skills, COM(2025) 90 final, European Union, Brussels, eur-lex.europa.eu/legal-content/EN/TXT/?uri=celex:52025DC0090.
The STEM Education Strategic Plan illustrates the value of treating advanced STEM skills as a strategic priority at European level. A common framework can raise visibility, establish shared objectives and support co-ordination across education, research, innovation and employment. Its aggregate targets can provide direction and a basis for monitoring progress across the European Union. Their interpretation and implementation nevertheless require more detailed diagnosis at national, regional and institutional levels. Similar aggregate outcomes may reflect different constraints across STEM fields, qualification levels, locations and population groups, while progress towards a participation target does not by itself show whether learners can persist, complete, progress to advanced levels or use and renew their advanced STEM skills. Member States therefore need to translate the Plan’s shared objectives into priorities and responses suited to their education systems, labour markets, institutional capacity and identified constraints.
Recent evidence from the United States demonstrates the policy importance of treating STEM talent development as an interconnected system rather than as separate interventions in education, research, employment and immigration matters. Zegart and Johnston (2026[24]) find that weaknesses in domestic preparation and progression increase reliance on international recruitment, while growing global competition and immigration constraints make attraction and retention less certain. International students earned 51% of US STEM master’s degrees and 39% of STEM research doctorates, yet the proportion remaining after graduation fell below stated intentions. These mutually reinforcing pressures show why policy needs to address the full set of conditions through which people develop, apply and continue to renew advanced STEM skills, rather than concentrating on enrolment or graduate supply in isolation. The report also illustrates the value of combining evidence across education, research, employment and migration, since no single indicator or dataset can establish whether a country is developing and using advanced STEM skills effectively. The five interconnected bottlenecks examined below in the chapter provide a framework for assessing these relationships across the development, renewal and use of advanced STEM skills.
Workforce shortages refer broadly to imbalances between the workers available and those sought for particular work. A labour shortage arises when too few people are available or willing to work under prevailing pay and working conditions, while a skills shortage arises when employers cannot obtain particular skills (Brunello and Wruuck, 2021[25]; Polanco-Jimenez and Witte, 2025[26]). Labour shortages can be assessed through indicators such as vacancy rates, recruitment times and wage trends.
Skills shortages are harder to assess because the skills available and required cannot usually be observed directly, qualifications and occupations provide only imperfect proxies, skill requirements change over time, recruitment difficulties may have causes other than insufficient skills, and employment indicators do not show whether advanced STEM skills are used effectively (OECD, 2022[17]).
Qualifications could be used as an indicator, but they are an imperfect proxy for skills for two reasons. On the one hand, credentials signal the skills developed through education, but they do not always accurately reflect individuals’ actual capabilities in practice. Weak alignment between qualifications and skills, due to differences in learning quality, specialisation, and practical experience, makes it difficult for employers to assess what workers can do and, in turn, to measure real skills shortages (Spence, 1973[27]; Arrow, 1973[28]). On the other hand, the demand for skills is constantly evolving. As technologies and work practices change, the competencies required by employers also shift, particularly in advanced STEM fields (Acemoglu and Restrepo, 2019[29]). This can create gaps between what is taught and what is needed in practice: graduates may hold relevant degrees but lack up‑to‑date or applied skills if education and training systems do not keep pace with technological change. Together, imperfect signalling and rapidly changing demand make it challenging to define, identify, and measure skills shortages with precision.
Advanced STEM skills are also used beyond occupations formally classified as STEM occupations, making their supply and use more difficult to assess. Examples illustrate this. The widespread adoption of digital technologies and AI has made skills in areas such as data analysis, modelling, systems integration and technology-informed decision making relevant across a growing range of occupations and sectors. Emerging fields such as quantum technologies provide another example. Their research, development and adoption involve scientists, engineers and technicians, but also managerial and other professional roles in sectors such as finance, healthcare and advanced manufacturing. Analysis confined to conventional STEM occupations may therefore overlook workers and skills (EPO/OECD, 2025[30]).
Furthermore, employer-reported recruitment difficulties do not necessarily indicate a skills shortage. They may also reflect pay and working conditions, geographical constraints, narrow or unrealistic recruitment requirements, demand for experienced specialists, or limited employer investment in onboarding, training and continued development (Sostero and Fernández-Macías, 2021[31]). Recruitment difficulties therefore need to be interpreted alongside evidence on job requirements, candidate profiles, wages, location, working conditions and employer practices.
Employment in a STEM occupation or a role requiring a STEM qualification does not necessarily mean that advanced STEM skills are used effectively. Workers may undertake tasks offering limited complexity, autonomy, problem solving, professional judgement or opportunities for continued development. Conversely, graduates employed outside their field of study or outside an occupation formally classified as STEM may make extensive use of scientific, quantitative, computational or engineering competence.
In sum, distinguishing among these forms of underutilisation affects how apparent shortages are interpreted (Box 1.2). Evidence on tasks and actual skills use is therefore needed to determine whether advanced STEM skills are being applied effectively across occupations and sectors (Xu, 2012[32]; More and Rosenbloom, 2020[33]; Yin and Wang, 2026[34]).
Skills underutilisation refers broadly to a situation in which workers possess skills that are not fully required, applied or rewarded in their work. It can be examined across four related but distinct dimensions:
Qualification mismatch compares a worker’s educational attainment with the qualification level required for the job. Overeducation occurs when attainment exceeds the level normally required.
Field-of-study mismatch concerns whether the content of a worker’s education corresponds to the field in which they work. Employment outside the field of study does not necessarily mean that advanced STEM skills are unused, since scientific, quantitative, computational and other transferable competence may be applied elsewhere.
Skills mismatch compares a worker’s competence with the skills required by the job. Skills underutilisation occurs when existing competence is used insufficiently, for example because work provides limited task complexity, autonomy, problem-solving or professional judgement.
Employment underutilisation concerns the amount or conditions of employment and may include involuntary part-time work, insufficient hours, low earnings or limited career progression. It does not necessarily imply that the worker’s skills are poorly matched to the tasks performed.
These dimensions may overlap, but they identify different groups of workers and should not be used interchangeably. Qualification correspondence, field relevance and occupational classification provide indirect evidence, while information on tasks, autonomy and actual skills use gives a more direct indication of whether advanced STEM skills are being applied.
STEM is not a monolithic category
STEM as a policy concept gained influence partly in response to concerns about workforce needs, productivity, innovation and competitiveness (Box 1.3). In many countries, it provides a practical framework for bringing together science, technology, engineering and mathematics and for connecting education policy with research, employment and industrial development. It can raise the visibility of scientific and technical education, support co-ordination across policy areas and provide a basis for differentiated investment where programmes require costly infrastructure, specialist staff or additional teaching capacity.
However, the apparent coherence of STEM as a priority selection of disciplines can conceal substantial differences. Definitions may differ among education, employment and statistical systems. Education authorities may identify STEM through fields of study, while labour-market authorities use occupations, competencies or job tasks. For example, the EU STEM Education Strategic Plan (see Box 1.1) uses a programme- and field-based classification covering the ISCED-F (2013) fields of natural sciences, mathematics and statistics; information and communication technologies; and engineering, manufacturing and construction, as well as related interdisciplinary programmes. Importantly, common classifications of STEM do not include agriculture, forestry, fisheries and veterinary; and health and welfare – also fields which develop advanced STEM skills. Regardless of which fields are grouped under the acronym, they tend to have very different provision costs, capacity constraints, progression routes and patterns of demand.
Treating STEM as a strategic priority should not place it in opposition to the humanities, arts and social sciences as responsible innovation and sustainable growth requires knowledge and skills that draws on both STEM and non-STEM fields. Scientific and technological developments take shape within social, cultural, economic, legal and political contexts, and decisions about their purposes and use require knowledge and judgement from across these fields. Presenting investment in STEM as requiring reduced attention to other disciplines may narrow the knowledge available to define problems, assess consequences and guide the responsible application of science and technology.
The term science, technology, engineering and mathematics (STEM) can refer to programmes of study, qualifications, occupations, competencies or research activity. These approaches overlap, but they capture different populations and serve different analytical purposes. Countries differ in how they define and use STEM, reflecting differences in industry, education systems, labour markets, and regulations. As a result, there is no single global definition.
The boundaries of STEM have also broadened in some policy and educational contexts through the use of STEAM or STE(A)M, which makes explicit the contribution of the arts, humanities and social sciences to creativity, design, communication and the responsible application of technology.
STEM education programmes are usually identified through classifications of fields of study. OECD comparisons generally include natural sciences, mathematics and statistics; information and communication technologies; and engineering, manufacturing and construction. National classifications may differ, however, and the treatment of interdisciplinary programmes and fields related to agriculture, health, sustainability or the green transition is not uniform.
STEM qualifications certify completion of assessed programmes identified as STEM. They provide evidence of formal attainment, but qualifications with similar titles or at the same level may differ in content, specialisation, practical experience and assessment. They therefore do not provide a complete measure of the competence available or used.
STEM occupations are identified through occupational classifications, sometimes supplemented by information on the competencies or tasks associated with particular jobs. National and international classifications do not always draw the same boundaries. Moreover, advanced STEM skills are used in occupations not formally classified as STEM, while people within the same occupational category may undertake tasks requiring different forms and levels of expertise.
STEM research is generally classified through disciplinary fields, publications, patents or research activities. Emerging and interdisciplinary areas may not fit established categories, and classifications of research do not necessarily correspond to classifications of education or employment.
Despite the definitional variability, STEM remains a useful policy construct. It can focus attention on the capabilities required for scientific and technological development, help actors from education, research, employment and industry identify shared concerns, and support co-ordination across policy areas. However, its use should not imply that STEM is a single homogeneous category with fixed boundaries.
Meeting changing needs for advanced STEM skills requires education systems to draw on learners with a wide range of backgrounds who have aspirations, preparation and potential. This involves more than widening access and expanding participation. A broader talent base also expands the perspectives and questions represented in research and innovation. Large-scale empirical studies indicate that researchers’ characteristics are associated with the topics they study and that under-represented groups produce comparatively high rates of scientific novelty. However, their novel contributions may be less readily taken up, recognised and rewarded when they originate outside established scientific networks and perspectives (Hofstra et al., 2020[35]; Kozlowski et al., 2022[36]).
A group is under-represented when its participation falls below an appropriate benchmark. Which benchmark is appropriate depends on the policy question. Participation may be compared with the general population, learners at the preceding educational level, those who meet relevant entry requirements or those eligible for a specific opportunity. For example, the share of women entering doctoral education may be compared with their share among eligible master’s graduates in the same field. Both representation and transition rates should be considered, since similar participation shares may conceal differences in application, selection, continuation or progression.
Broadening the talent base must be accompanied by conditions that enable different perspectives to influence innovation and receive appropriate recognition. It also means ensuring that access to the educational opportunities needed to develop and demonstrate advanced STEM skills does not depend unnecessarily on learners’ starting points or circumstances. The OECD concept of inclusive excellence in science and the research workforce provides a useful basis for this approach. Inclusive excellence means maintaining high standards while embedding equity in the conditions through which people gain access to education, training, opportunities and recognition, so that differences unrelated to their ability to contribute do not determine their prospects of success (OECD, 2025[37]). Applied to advanced STEM education, this means preserving the depth and quality of educational pathways while identifying and addressing avoidable constraints on who can develop advanced STEM skills.
Broadening participation and perspectives in advanced STEM
Gender is the most visible and comparatively well-documented dimension of under-representation (Eurostat, 2026[38]; European Commission, 2025[39]). Figure 1.2 shows that across countries women are under-represented in core STEM. Women account for 31% of bachelor's enrolment on average, and no country reaches gender parity at any level of study. National differences are nevertheless substantial, ranging from more than 40% of bachelor’s enrolment in Sweden, New Zealand and Canada to around one-quarter in Hungary, Lithuania, Luxembourg, Belgium and Switzerland. Under-representation is particularly pronounced in short-cycle tertiary education, where women account for only 19% of enrolment on average.
Representation increases at higher levels of study, rising from 31% at bachelor's level to 39% at master’s level and 40% at doctoral level. However, this should not be interpreted as evidence of equitable progression. The figure compares enrolment at different educational levels in a single year rather than following individuals over time. It therefore cannot show whether women progress at the same rate as men or identify where differences emerge in application, admission, continuation or completion. Moreover, changes in the composition of enrolment across STEM fields may increase the female share at higher levels even where representation remains stable or declines within individual fields. Low representation of women in short-cycle tertiary STEM programmes may reflect differences in the disciplinary composition, occupational orientation and role of these programmes across national systems. However, the available data do not allow the causes of these patterns to be identified, and further analysis is required.
Women make up roughly one-third of STEM graduates at bachelor’s and master’s levels, but only around one-fifth of doctoral graduates in ICT and less than one-third in engineering. Women’s representation among doctoral graduates has declined since 2018 in several narrower STEM fields, including physics and mathematics, where they were already underrepresented. Women also hold less than one-third of senior academic positions and around one-quarter of institutional leadership roles (European Commission, 2025[39]). Participation at one level therefore does not ensure equitable progression to the next. OECD evidence further indicates that gender differences continue across research and innovation careers and may concern participation in different fields and sectors, international mobility, employment conditions and longer-term career outcomes. Examining representation at successive education levels is therefore insufficient: analysis also needs to follow who enters research and innovation work, the roles and conditions they experience, and whether they can progress and remain in these careers (OECD, 2026[22]).
Female students as a percentage of all students enrolled in core STEM, by level of education, 2024
Note: Values show female students as a percentage of all students enrolled in core STEM at the level indicated. Core STEM comprises natural sciences, mathematics and statistics (ISCED‑F 05), information and communication technologies (ISCED‑F 06) and engineering, manufacturing and construction (ISCED‑F 07). Countries are ranked in descending order of the female share of bachelor's enrolment. Data on short-cycle tertiary education are not available for all countries, and short-cycle provision is limited or absent in several systems.
Source: OECD (2026[40]), Number of enrolled students, graduates and new entrants by field of education, (database), OECD Data Explorer, https://data-explorer.oecd.org/s/5ky.
Examining equitable progression requires also information on who can access the learning, mentoring, research and professional opportunities that support educational and career transitions. The STEM Atlas shows how mapping initiatives can complement official statistics by making the distribution of such opportunities more visible. Developed by STEM Women and Global STEM Women, the approach has been used first in Spain, then also in, Brazil, Canada, Colombia Italy and Portugal, mapping more than 700 initiatives. While official data track enrolment, graduation and employment, they provide less information on outreach, mentoring, extra-curricular programmes, professional networks and similar activities. The Atlas collects information on the groups and educational or career stages reached, geographical coverage, activities and funding, and compares it with demographic and education data (Global STEM Women, 2026[41]; Díaz, 2026[42]). It can show where provision is concentrated and where particular places, groups or transitions receive limited support.
The wider issue is that access to mentoring, paid research experience, internships, professional networks, leadership and entrepreneurship opportunities is uneven. Mapping can make some of these differences visible where enrolment and graduation data cannot. However, the presence of an initiative does not establish its accessibility, quality or effects. An initiative may not reach learners facing the greatest constraints or provide substantive opportunities to develop advanced STEM skills.
Assessing whether policy responses reflect this range of connected constraints requires looking across the full education and professional route. The UNESCO STEM and Gender Advancement (SAGA) project illustrates how a systemic approach can be used to identify gaps in the conditions supporting participation and progression (UNESCO, 2018[43]). It was launched in 2015 and maps policies and indicators across seven objectives covering social norms and stereotypes, primary and secondary education, higher education, career progression, research content and practice, policymaking, and entrepreneurship and innovation. Across the pilot exercises, career progression was addressed by the largest number of identified policies and instruments, followed by participation and retention in higher education. These are areas in which governments, higher education institutions, research organisations and employers have relatively direct instruments, including recruitment and promotion procedures, scholarships, mentoring, parental-leave provisions and childcare support.
Social norms and stereotypes were widely identified in the pilot exercises as a major constraint but were addressed by relatively few instruments. This reflects the difficulty of assigning responsibility for influences that operate across families, schools, institutions, workplaces and wider society and whose effects may become visible only over time. Research content and practice and science- and technology-based entrepreneurship and innovation also received comparatively limited attention, particularly in countries where these had only recently emerged as policy priorities.
The pilot exercises show why broadening the talent base requires attention not only to the number of measures adopted, but also to their distribution across the stages and settings in which inequalities arise. A concentration of measures in one area does not establish that it is the most important source of inequality or that it is being addressed adequately, while limited policy activity may indicate that a constraint is difficult to assign to a single actor or address through a discrete instrument. Assessing the contribution of interventions such as SAGA, or UNESCO’s Campus Africa initiative, which brings together higher education institutions, governments and other partners to strengthen research and innovation, doctoral study and mobility, and higher technical education across Africa (UNESCO, 2026[44]), requires looking beyond the existence and scale of provision to examine which learners and institutions it reaches, the quality of their participation, and whether it produces sustained educational, research and employment outcomes
Unequal access and participation extend beyond gender
Although gender is the most visible and best-documented dimension of under-representation in STEM, similar questions arise for other groups whose opportunities to participate and progress may be shaped by their circumstances and the barriers they encounter.
Socio-economic background can affect whether learners encounter STEM early, gain the preparation and information needed to pursue it, and can afford the direct and indirect costs of advanced study (Betancur, Votruba-Drzal and Schunn, 2018[45]; Morgan et al., 2016[46]). Geography can limit access to relevant programmes, research infrastructure and employment opportunities. Learners with disabilities may encounter inaccessible teaching, assessment, laboratories or infrastructure (Borrego et al., 2025[47]). Learners with migration backgrounds may face barriers related to language, access to information and guidance, and the recognition of qualifications and previous learning (European Commission, 2024[48]).
The effects of these factors depend on how individual circumstances interact with educational, institutional and labour market conditions. Barriers can become more significant when advanced study requires relocation, extensive laboratory attendance or unpaid professional experience; when teaching and infrastructure are inaccessible; or when language requirements, unfamiliar qualifications and limited networks restrict access to educational and professional opportunities.
These circumstances can intersect and reinforce one another (Crenshaw, 1989[49]; Bauer et al., 2021[50]). Financial constraints may combine with disability-related costs, caring responsibilities or distance from suitable provision. Migration status may interact with qualification recognition, language barriers and limited access to networks. Policy should therefore examine how learners’ circumstances interact with educational, institutional and system conditions to shape access to preparation, substantive learning opportunities and progression.
Evidence is often thinnest where several dimensions intersect. Data availability and public discourse differ substantially across countries, and some countries do not collect data on ethnicity or migration background, limiting both national analysis and cross-country comparison (European Commission, 2025[51]). Small sample sizes can restrict reliable analysis, while data categories and legal frameworks differ. Administrative records may show participation without revealing the conditions that shaped it. Analysis should therefore be explicit about these limitations and, where comparable quantitative data are unavailable, use qualitative and initiative-level evidence cautiously to identify constraints that aggregate indicators may conceal (European Commission, 2025[51]; Bauer et al., 2021[50]).
In Europe, joined efforts across countries and stakeholders have resulted in a growing evidence and policy infrastructure. “She Figures”, the European Commission’s flagship statistical publication on gender equality in research and innovation, provides comparable, disaggregated indicators that follow women’s and men’s trajectories from doctoral education through research employment to senior academic, leadership and decision-making positions. It also covers working conditions and research and innovation outputs, enabling countries and institutions to identify disparities and monitor change over time. Institutional action is supported through so-called Gender Equality Plans. Under Horizon Europe, having such a plan is an eligibility condition for public bodies, higher education institutions and research organisations when applying for public funding (European Commission, 2025[52]; Karataş, Löther and Weber, 2025[53]).
No equivalent EU-wide statistical or funding frameworks exist for socio-economic background, disability, migration background or other relevant circumstances. Evidence and policy responses in these areas are more difficult to compare. Most countries address these groups primarily through general higher education access and student support measures.
These differences in opportunity and policy coverage cannot be located at a single stage. They may arise in preparation and entry into STEM education; access to substantive learning opportunities; persistence and completion; progression to advanced levels; transition into suitable employment; early-career development; or skills renewal. They can also accumulate across the transitions among these points. Understanding where the talent base is unnecessarily narrowed therefore requires a framework that distinguishes the principal locations and mechanisms of constraint while recognising their interdependence.
Advanced STEM skills are developed, used and renewed through routes connecting education and employment. These routes may begin with foundations acquired during school education and continue through vocational education and training, higher education, advanced study, employment and continuing education.
Figure 1.3 distinguishes the initial development of advanced STEM skills from their use and renewal across working life. Initial development typically includes entry into higher education, transition and progression between levels and fields, completion and career entry. Skills renewal may involve updating or deepening existing advanced STEM skills, moving into an adjacent field or preparing for substantial occupational change. It may include a return to education at undergraduate or postgraduate level, depending on the learner’s previous competence and intended destination. These routes are not necessarily linear. Learners may move between education sectors, change fields, interrupt and resume study, enter employment before completing a qualification, or return to education as the requirements of their work or career direction change. The same course may therefore contribute to initial skills development for a learner completing a first higher education qualification and to skills renewal for a learner returning with previous education or professional experience.
Constraints can become visible at different points along these routes and manifest in bottlenecks in entry into STEM higher education, persistence and completion, progression to advanced levels, career entry and effective skills use, and skills renewal across working life (Figure 1.3). Diagnosing these bottlenecks requires evidence from across education and employment. Education evidence can reveal patterns in preparation, entry, persistence, completion, progression and attainment, while employment evidence can show where individuals work and where demand may be changing. Evidence from employers and workers can provide information on recruitment, changing tasks, competence requirements, working conditions and actual skills use. No single perspective establishes whether the relevant advanced STEM skills are being developed and used in sufficient depth, in the relevant fields and locations, or for the purposes concerned.
Source: Authors based on literature review, and discussions in the Mutual Learning Group on Advanced STEM skills development in higher education.
Many policies examined in this report are relevant to STEM education more broadly as the development of advanced STEM skills is cumulative: access to strong foundations, high-quality teaching, progression, research and professional experience is necessary before advanced competence can be developed and demonstrated. The focus on advanced STEM skills guides policy analysis to assess whether a system allows learners and workers to develop, use and renew these skills.
Where the causes of a bottleneck may lie
The point at which a bottleneck becomes visible does not necessarily show what causes it. To analyse the latter, it is helpful to consider that constraints may arise in four analytically distinct but connected locations: through individuals’ access to educational and professional opportunities, within higher education institutions, within employers, and in wider system arrangements (Table 1.2).
Note: The four locations are analytical categories, not fixed levels of responsibility. System arrangements shape the responsibilities, resources and incentives of higher education institutions and employers. Institutional and employer practices, together with financial, geographical and personal circumstances, determine which opportunities individuals can access. A constraint may therefore become visible in one location while originating partly or wholly in another, and outcomes generally arise through interactions among them.
The role and capacity of higher education
Higher education has a distinctive role in the development of advanced STEM skills because it develops and assesses advanced disciplinary and methodological learning, connects education with research, provides access to specialist infrastructure and recognised qualifications, and supports progression to advanced study and skills renewal. Its contribution depends also on the wider system arrangements within which institutions operate.
The capacity of higher education institutions includes:
pedagogical capacity, including curriculum design, teaching, assessment, feedback, supervision and academic support
organisational capacity, including programme leadership, co-ordination, guidance, connected learner services, partnerships, recognition, and the collection and use of evidence for improvement
resource capacity, including qualified academic, technical and support staff, specialist infrastructure and research environments, staff time, data and analytical capacity, and sustainable funding
These forms of capacity are interdependent. The adequacy of institutional capacity depends not only on the number of available programme places, but also on student-to-staff ratios, class and laboratory group sizes, teaching intensity, supervisory workloads, staff expertise, and access to technical and learner-support staff. Additional places will not increase the supply of advanced STEM skills if staffing, infrastructure and organisation are insufficient to provide high-quality teaching, practical learning, supervision, feedback and support. These requirements may be particularly demanding in fields relying on specialist teaching, laboratories, research environments and disciplinary communities, and thus more challenging for smaller systems.
Specialist infrastructure requires continuing investment. Laboratories, equipment, computing and data environments, and physical facilities need to be operated, maintained, calibrated, repaired, replaced and updated as methods, technologies and safety requirements change. Practical teaching creates recurring costs for laboratory materials, software, data access, technical preparation, safety and waste management. Initial investment therefore creates sustainable capacity only when institutions can also finance operation, maintenance and renewal.
Staff time and institutional incentives are an important part of capacity. Curriculum renewal, pedagogical improvement, collaborative teaching, feedback, programme leadership and partnership-based provision require sustained work alongside research, supervision and other responsibilities. Where workload allocation and career evaluation give limited recognition to teaching, curriculum development and educational leadership, these functions may depend on temporary funding or a small number of committed staff. Institutions therefore need to allocate sufficient time and recognise substantive contributions to teaching and programme improvement in appraisal and career progression.
Institutions also need the data and analytical capacity to identify where learners encounter barriers, how programme experiences and outcomes differ, and whether changes improve learning and progression. This requires more than collecting participation, completion or satisfaction data. Programme leaders and institutional decision makers need evidence that can connect learner outcomes with curriculum, teaching, assessment, workload, support and access to practical and research opportunities. Data should support learners and programme improvement rather than operate primarily as a means of classifying individual risk.
Higher education capacity affects all five bottlenecks. Increasing entry without sufficient staff, infrastructure and learner support may restrict access to practical learning and weaken teaching and feedback. Expanding advanced study without adequate supervision, research infrastructure and funding may reduce quality or delay completion. Skills-renewal provision may remain inaccessible if programmes, recognition arrangements and learner services are designed mainly for recent school-leavers studying full-time. Research and professional experience may produce uneven benefits if access, learning objectives, responsibilities, supervision and quality assurance are unclear.
The importance of policy dialogue and how it was applied in preparing this report
Constraints differ across fields, qualification levels, population groups, institutions and locations, and the same observed outcome may arise from different mechanisms. These differences make policy dialogue important. Quantitative indicators can identify patterns, and research can provide evidence on the mechanisms that may produce them. Interpreting their policy relevance, however, also requires knowledge of how education systems, higher education institutions, employers and public authorities operate in different contexts. Policymakers, researchers and higher education practitioners bring complementary perspectives on where constraints arise, which actors can influence them, what implementation requires and why similar measures may produce different results. Dialogue across countries can help identify common bottlenecks while distinguishing their context-specific causes. It can also support comparison of policy priorities and areas for policy action without assuming that the same response should be transferred directly between systems.
This principle was applied in preparing this report. The report was developed through an iterative process that combined literature review, comparative data analysis, desk-based policy analysis and structured dialogue through the Mutual Learning Group on Advanced STEM Skills in Higher Education, organised by the OECD Secretariat (see Annex A for more information). The group brought together policymakers, researchers, higher education practitioners and other experts with experience from Australia, Austria, Belgium, Bulgaria, Canada, Croatia, Denmark, Estonia, Finland, Germany, Ireland, Japan, Korea, Poland, Portugal, Spain and Trinidad and Tobago, as well as participants from the European Commission’s Joint Research Centre, UNESCO and the European Organization for Nuclear Research (CERN).
The Mutual Learning Group provided a lens through which the literature, data and policy analysis could be interpreted for their policy relevance, implementation requirements and applicability across contexts. This helped bring the different sources of evidence together and organise the findings into a framework intended to inform policymakers.
Discussions in the group drew attention to institutional and system conditions that may be less visible in published research and evaluations, including fragmented responsibilities, limited staff time, unstable funding, insufficient infrastructure, difficulties in co-ordinating actors and the practical limits of transferring an approach between settings. The dialogue also helped test the relationships among bottlenecks, policy priorities and areas for policy action, including whether similar bottlenecks appeared across countries, which mechanisms produced them and how action at one point could create or relieve pressure elsewhere.
The resulting framework distinguishes five points at which constraints on the development and effective use of advanced STEM skills become visible.
Bottleneck 1: Entry into STEM programmes in higher education
Entry into STEM higher education varies substantially across countries and fields. Core STEM accounts for 28% of new tertiary education entrants on average across both EU and OECD Member States, but national shares range from 21% to 38%. Across countries, growth in entry has been limited overall. OECD and EU data indicate that the average number of new STEM entrants changed only modestly between 2015 and 2023, although trends differed substantially across countries and fields. Growth was stronger in information and communication technologies than in engineering or the natural sciences.
Figure 1.4 shows the share of new tertiary entrants in the three core STEM fields and in two additional science-intensive fields: agriculture, forestry, fisheries and veterinary; and health and welfare. Countries are ranked by the combined share of entrants across all five fields. National figures for new entrants into core STEM fields range from 38% in Germany, followed by Finland at 34%, Estonia at 33%, and Sweden and Luxembourg at 32% each, to 21% in Belgium and Poland and 22% in the Netherlands. This 17‑percentage-point range among systems facing broadly similar technological and demographic pressures directs attention towards conditions in particular national contexts rather than a uniform European pattern.
The combined share by which countries are ranked in Figure 1.4 should not be interpreted as a measure of core STEM entry. Finland records the highest combined share at 59%, but health and welfare alone accounts for 24% of its new entrants, the highest value in the comparison, while its core STEM share of 34% remains below Germany’s. Germany ranks sixth on the combined measure at 48%, while recording both the highest core STEM share, at 38%, and the lowest share entering health and welfare, at 8%. Belgium illustrates the same point more sharply: its combined share of 44% is close to the OECD average, but its core STEM share of 21% is the lowest in the comparison because health and welfare accounts for 21% of new entrants. Aggregating fields with different patterns of entry, provision costs and capacity constraints can therefore produce a measure that moves independently of core STEM entry.
New entrants to core STEM and science-intensive fields as a share of all new tertiary entrants, 2024
Note: Countries are ranked in descending order of the combined share of all five fields. Data refer to 2023 for Bulgaria, Czechia, Denmark, Finland, France, Greece, Hungary, Italy, Japan, Lithuania, Portugal, Sweden, the OECD average, and the European Union (25 countries). Data on information and communication technologies are not available for Japan, and its combined share is correspondingly understated. Core STEM fields comprise natural sciences, mathematics and statistics (ISCED‑F 05), information and communication technologies (ISCED‑F 06) and engineering, manufacturing and construction (ISCED‑F 07). Science-intensive fields comprise Agriculture, forestry, fisheries and veterinary (ISCED‑F 08) and health and welfare (ISCED‑F 09).
Source: OECD (2026[54]), Distribution of enrolled students, new entrants and graduates by field of education, (database), OECD Data Explorer, https://data-explorer.oecd.org/s/550.
Composition also varies within core STEM. Entry into natural sciences, mathematics and statistics ranges from 3% in Latvia to 11% in Luxembourg and 10% in France and Italy, while entry into information and communication technologies ranges from 3% in Italy to 13% in Luxembourg. Engineering, manufacturing and construction, the largest core STEM field in most systems, varies less proportionately but more in percentage-point terms, from 8% in Luxembourg and the United Kingdom to 22% in Germany. Information and communication technologies account for a comparatively small share of entry in most countries, with a median of 7%, compared with 13% for engineering. Italy, where core STEM entry of 30% exceeds the EU average but entry into information and communication technologies is the lowest in the comparison, shows that a relatively high aggregate share does not imply balanced entry across fields.
Entry data reflects both learner demand and institutional supply. Learner demand concerns whether prospective learners are interested in applying to STEM programmes and have the necessary preparation, while institutional supply concerns whether sufficient, accessible and high-quality programme places are available. These point to different sets of constraints and policy responses.
Earlier subject choices and unequal preparation can restrict academic eligibility
School education shapes learners’ interest in STEM and develops foundational competence. The structure of secondary education also affects access to higher education. Some systems require early specialisation, while others maintain a broader curriculum for longer. Early specialisation may limit later options, especially for learners who develop an interest in STEM later or have had limited access to relevant subjects (OECD, 2025[55]).
Advanced mathematics and science subjects frequently operate as formal prerequisites for entry into higher education STEM programmes. Learners and their families may not understand that discontinuing advanced mathematics or science subjects in secondary education can exclude entry to certain higher education fields. Access to these subjects, the quality of their teaching, the timing of specialisation and the flexibility of secondary-higher education pathways can therefore open or close later opportunities. The effects may be particularly strong in systems that use early tracking, have rigid subject combinations or provide limited opportunities to supplement previous learning. Some schools may also be unable to offer the relevant subjects because of teacher shortages, inadequate facilities or cohorts that are too small.
Evidence from England (United Kingdom) indicates that a reform encouraging greater participation in mathematics during the final two years of secondary education increased entry to and completion of STEM undergraduate degrees. However, gender and socio-economic differences in STEM participation remained unchanged. The findings suggest that widening access to prerequisite subjects can increase progression into STEM but may be insufficient on its own to overcome the wider barriers affecting participation among under-represented groups (Morando, 2024[56]).
Curriculum reforms can also have unintended effects. Following the reform of general upper-secondary education implemented in France from 2019, many students discontinued mathematics between the first and final years, with discontinuation more prevalent among female students. Mathematics provision was subsequently strengthened for students who did not take the mathematics specialisation. While these developments do not by themselves establish the reform’s causal effects on entry to higher education, they highlight the importance of assessing how subject choices shape preparation and eligibility for subsequent programmes across population groups, and the need to monitor participation before and after reform (Dauphin, 2024[57]; Ministry of National Education and Youth, 2023[58]). Chapter 3 discusses the methodological challenges involved in distinguishing the effects of an intervention from wider changes, examining differences in implementation and outcomes across groups, and using the resulting evidence to inform policymaking.
Formal prerequisites may be academically justified where later learning depends on cumulative disciplinary foundations. However, there is a risk that requirements unnecessarily exclude learners who develop an interest in STEM after making earlier subject choices, enter through vocational education and training, move between education systems or return as adults. Bridge provision, meaning additional learning designed to address identified gaps, can provide another route into STEM programmes without removing substantive academic requirements. It may include supplementary mathematics courses, foundation years or preparatory courses before entry (Bokhove et al., 2019[59]).
Teachers’ professional backgrounds, beliefs and instructional practices are also associated with students’ attainment, motivation and persistence in mathematics and science. These outcomes can, in turn, shape students’ plans to pursue STEM-related studies and careers (Ekmekci and Serrano, 2022[60]). In the natural sciences, practical activities can strengthen conceptual understanding and practical skills, but their effectiveness depends on how they are designed and taught (Oliveira and Bonito, 2023[61]).
Mathematics experiences shape whether learners consider STEM attainable and attractive
Learners’ interest in mathematics, confidence in their mathematical abilities and anxiety about the subject can shape their engagement and decisions about STEM education and careers. Mathematics self-efficacy refers to confidence in one’s capacity to perform particular mathematical tasks successfully. Mathematics anxiety refers to feelings of tension, apprehension or fear associated with engaging in or anticipating mathematical learning, assessment and problem-solving. Mathematics interest is understood as enjoyment of, curiosity about and positive engagement with the subject (Bokhove et al., 2019[59]).
Longitudinal evidence suggests that mathematics interest is a particularly consistent predictor of later STEM choices, while the relationships between self-efficacy, anxiety and STEM participation overlap with prior achievement and other dimensions of learners’ mathematical identities (Cribbs, Huang and Piatek‐Jimenez, 2021[62]; Ferdinand, Malanchini and Rimfeld, 2024[63]). This does not make self-efficacy or anxiety irrelevant as potential constraints. Confidence may strengthen expectations of success and willingness to undertake challenging mathematics, while anxiety may encourage avoidance, reduce perceived control and make mathematically intensive programmes appear more difficult or personally costly. Longitudinal studies find that mathematics anxiety and self-efficacy are associated with subsequent STEM choices, but the available evidence does not establish that either factor independently causes later choices. Their influence may operate through or alongside interest, attainment and the learning environments in which these develop (Cuder et al., 2024[64]; Ferdinand, Malanchini and Rimfeld, 2024[63]). Anxiety is associated with less positive attitudes, avoidance and, in some contexts, weaker performance. These factors can reinforce one another over time: anxiety may encourage avoidance and reduce opportunities to develop proficiency, while weaker achievement and unsuccessful experiences may contribute to greater anxiety (Lau, Ansari and Sokolowski, 2024[65]).
It is important to understand that interest, self-efficacy and anxiety are not fixed personal characteristics. They develop through learning experiences and may be influenced by curriculum, teaching practices, achievement, feedback and peer comparison, as well as socio-economic resources expectations about belonging, and stereotypes. Research on mathematics self-efficacy identifies reciprocal relationships with learning behaviour and achievement and highlights the role of environmental influences, including instructional practices and teacher feedback. Hence, learners with similar levels of performance may develop different levels of interest, confidence and anxiety and may view the same STEM route differently (Zakariya, 2022[66]; Street, Malmberg and Schukajlow, 2024[67]).
Aspirations lead to entry only when learners perceive that they can succeed and belong
Learners’ decisions about STEM depend also on the futures they hope to pursue and whether they see a place for themselves in this by pursuing a certain study programme or career. Learners are more likely to choose STEM when they can envision a desirable future in the field and believe they can succeed and belong within it. These judgements are shaped by prior achievement, subject access, social recognition, role models, family and peer expectations, institutional environments and the perceived social meaning of different STEM fields (Archer et al., 2012[68]; Archer et al., 2015[3]).
Aspirations and perceived fit influence one another over time. Aspirations refer to the education or career futures learners aim to pursue, while perceived fit refers to whether they see a STEM field as attainable, valuable and compatible with their interests, identity and sense of belonging. Previous academic and social experiences shape how learners understand their abilities, interests, values and identities in relation to STEM. These perceptions affect whether they pursue relevant subjects and apply to STEM programmes. Later experiences may strengthen an aspiration, redirect it towards another field or route, or weaken it. Furthermore, learners may enjoy doing science without seeing themselves, or being recognised by others, as someone who could become a scientist (Archer et al., 2010[69]; Carlone and Johnson, 2007[70]).
Unequal access to preparation, encouragement, information, networks and recognition matter for aspirations and perceived fit. Gender, race or ethnicity, socio-economic background, family, school experience and role models can influence whether advanced STEM skills are considered achievable. Interest may therefore be insufficient where dominant images of STEM remain narrow or exclusionary (Archer, DeWitt and Osborne, 2015[71]; Archer et al., 2016[72]). Also, learners may decide not to apply for a programme because they doubt their ability to meet its perceived academic demands. Studies found that in physics academically qualified learners may opt out because they see the field as exceptionally difficult, masculine or suited only to a narrow group (DeWitt, Archer and Moote, 2018[73]; Archer et al., 2016[74]).
Information and guidance can widen options learners consider, but cannot remove barriers
OECD analysis of the 2022 PISA finds that 39% of 15-year-olds across OECD countries were uncertain about their future occupation and that occupational expectations remained concentrated in a narrow range of traditional professions, with substantial gender differences in aspirations towards engineering and information technology. Although 76% had explored careers online, only 55% had spoken with a career adviser, 45% had visited a workplace or participated in job shadowing, and 35% had completed an internship. These findings suggest that many learners lack sustained guidance and direct exposure to workplaces and people in work that could help them understand the diversity of STEM fields, the educational choices they require and whether these routes are attainable (OECD, 2025[75]).
Information and guidance influence study choices by shaping awareness of available options, understanding of programme requirements, expectations about careers and confidence in navigating applications. Their contribution depends on whether they are timely, accessible, comprehensible and connected with opportunities that learners can realistically pursue (Rinn and Kettler, 2020[76]). Furthermore, information and guidance expand the perceived choice set rather than persuade a learner to select a particular programme (Lent, Brown and Hackett, 1994[77]). This is particularly relevant at initial entry because prospective learners often possess uneven knowledge of STEM fields. Medicine, engineering and computing may be familiar, while materials science, geoscience, data science, environmental engineering and biotechnology may be less known.
Effective guidance is not a one-off intervention as learners require information that helps them consider:
academic preparation: whether previous subjects, qualifications and skills correspond to programme requirements, and whether gaps can be addressed
interests: whether the content and activities of a particular field correspond to what they enjoy and want to explore
values: whether a field offers opportunities for social contribution, creativity, collaboration, financial security or environmental improvement
belonging: whether people with different identities and backgrounds participate and are recognised in the programme and related occupations
progression: which routes are available after completion and what additional conditions apply.
employment prospects: the range and quality of possible outcomes, without reducing the decision to salaries or aggregate demand
Effective guidance requires long-term efforts. However, access to information and guidance is unequal. School resources, career stereotypes, socio-economic background, family knowledge of higher education and peer choices influence what learners know and whether they can act on it. Learners may follow familiar routes or consider only programmes that seem suitable and attainable for “people like me”, even when they have similar interests and aspirations to others (DeWitt and Archer, 2015[78]; Tompsett and Knoester, 2023[79]; Andersen and Hjortskov, 2019[80]; Rubineau et al., 2023[81]; Archer et al., 2016[74]; Du and Wong, 2019[82]).
Role models can make STEM routes more visible, challenge stereotypes and provide practical information about study and careers. Their influence depends on whether learners see their success as relevant and attainable. However, role models alone are unlikely to make a difference, particularly when programmes remain unavailable, unaffordable or difficult to enter (Gladstone and Cimpian, 2021[83]; De Gioannis, Pasin and Squazzoni, 2023[84]).
Financial and practical costs can make a preferred STEM route unfeasible
STEM programmes are study-intensive and can therefore impose additional costs for learners. Family resources, borrowing capacity, access to paid employment and the timing of financial support affect whether a preferred choice is feasible in practice (Liu and Kringos, 2025[85]).
Financial and geographical constraints tend to reinforce one another. Specialised STEM programmes and research infrastructure may be concentrated in a limited number of locations, while laboratory work, projects and placements may require substantial in-person participation. Where long-distance commuting is impracticable and affordable student housing is limited, learners may be unable to access programmes for which they are otherwise qualified.
Financial support can enhance STEM participation in two ways. It may make STEM study more attractive by reducing its costs and financial risks or help learners who already prefer STEM to act on that preference, and it can reduce the need to work during study, delay entry, choose a less suitable institution or forgo higher education (Zhao, Perez-Felkner and Hu, 2024[86]; Romero, Cobian and Martín, 2025[87]).
There are different forms of financial support and they address different barriers (Box 1.4). Need-based grants can make study affordable and reduce reliance on paid work. Merit-based scholarships can lower costs and recognise achievement, but demanding renewal conditions may discourage entry or continuation. Tuition waivers do not cover living costs, transport, childcare, equipment or lost earnings. Maintenance support can give learners more time for classes, laboratory work, projects and placements. The 2026 Erasmus+ policy experimentation call illustrates a more integrated approach: it is testing scholarships for academically strong students, particularly those from disadvantaged backgrounds, that cover the full costs of participation in selected strategic fields and combine study support with mobility and traineeships, internships, mentorships or industry placements (European Commission, 2026[88]).
Financial support can facilitate entry, persistence and completion by reducing financial pressure and reliance on paid work. Policymakers should identify which barrier an instrument addresses, for which learners and at what stage.
Need-based grants provide non-repayable support based on household resources or assessed financial need. They can make study feasible, reduce reliance on paid employment and help learners manage financial shocks. This is particularly relevant in STEM programmes with substantial scheduled teaching, laboratory work, projects and placements that limit opportunities to work alongside study.
Merit-based scholarships allocate non-repayable support according to attainment, performance, demonstrated potential, or other indicators of achievement. They reduce the cost of study and recognise achievement. However, strict renewal conditions may make demanding STEM programmes appear financially risky and encourage learners to avoid difficult courses or leave after a temporary decline in performance.
Tuition and fee waivers remove all or part of tuition or programme-specific charges and may be universal or targeted. They directly reduce the listed price of study but do not cover accommodation, food, transport, equipment, childcare or forgone earnings.
Living costs and maintenance grants contribute to housing, food, transport, childcare, equipment and other living costs. They can reduce financial pressure and free time for study, laboratories, projects and placements. They are particularly important where specialised provision requires relocation and affordable housing is limited.
Institutional capacity can limit both the number and quality of available places
Constraints in the capacity of higher education institutions can be particularly important at advanced levels and in fields requiring intensive teaching, supervision and research infrastructure as they may limit admission directly or lead institutions to restrict recruitment where additional enrolment would exceed the capacity supported by available staff, infrastructure and funding or weaken the quality of provision (OECD, 2025[89]).
Expansion can produce unintended effects if enrolment grows faster than institutions’ staffing, infrastructure and learner-support capacity. Time-limited or targeted funding may create unsustainable programmes, relabel existing provision or redistribute rather than expand capacity, while geographic concentration and employer funding may restrict access or narrow curricula. Capacity should therefore be assessed by relating the number of available and filled places to the staffing, infrastructure, supervision and learner support required for a high-quality learning experience. Indicators such as student-to-staff ratios, laboratory and equipment availability, class and group sizes, supervisory workloads and access to learner support can help determine whether existing or planned enrolment is adequately supported. Expansion is appropriate where demand is sustained, existing capacity is insufficient and additional investment can support high-quality provision.
Capacity constraints may be geographically concentrated where specialist facilities and staff are available in only a small number of institutions. Regional strategies may build capacity around existing research and economic strengths, while concentration in a small number of institutions may create the critical mass required for expensive or highly specialised fields. The appropriate balance depends on the field, the geographical distribution of learners and employment, and the extent to which mobility or shared provision is feasible.
Collaborative arrangements may allow institutions to share facilities, curricula, staff expertise and research environments. Deep, long-term and structured co-operation, such as the example of the European Universities alliances, can enable institutions to pool expertise, knowledge and resources and combine complementary strengths that may not be available within a single institution. Such arrangements can widen the educational and research opportunities available to learners, including in STEM and other strategic fields, but require aligned programme structures, compatible access and recognition arrangements, and sustainable organisational capacity across participating institutions (see example of European Universities alliances in Policy Priority 3.1).
Implications for diagnosing an entry bottleneck
Diagnosing an entry bottleneck requires evidence from learners, schools and higher education institutions, while responsibility for addressing identified constraints is distributed among schools, higher education institutions and public authorities, with employers contributing where constraints concern occupational information used in guidance and outreach.
Different aspects are relevant.
Diagnosis should follow learners through the stages leading to enrolment. Evidence on access to relevant subjects, attainment, eligibility, applications, offers, acceptance and enrolment can help identify where participation declines. Information from learners, schools and higher education institutions can then clarify whether the observed pattern reflects preparation, interest, confidence, programme awareness, perceived fit, application and admission arrangements, costs or institutional capacity. Analysis should examine the timing of constraints and whether they originate earlier than the point at which they become visible, and which groups are concerned. Curriculum and admission reforms should be monitored before and after implementation to assess their intended and unintended effects on subject participation, eligibility and entry.
Interest, confidence, anxiety, aspirations and perceived fit should not be treated as fixed characteristics of individual learners. They develop through prior learning, achievement, feedback, social recognition, stereotypes and educational environments. Learners may value a STEM field but doubt that they can meet its academic demands, belong within it or pursue the forms of work and social contribution they value. Where these factors constrain entry, responses may need to improve learning experiences, provide constructive feedback, widen representations of STEM study and work, and develop inclusive environments in which a wider range of learners can participate successfully.
Information and guidance should be assessed in relation to when it is provided, who can access it and whether learners can act on it. It should help learners understand programme content and requirements, possible routes into and beyond a programme, and the range and quality of associated employment opportunities. Information, guidance and role models can widen the options learners consider, but cannot compensate for routes that are inaccessible, unaffordable or unavailable.
Formal qualifications and prerequisites provide only partial evidence of preparation. Learners holding the same qualification may differ in the content and depth of their prior learning, the teaching they experienced and their opportunities to apply knowledge. Where preparation constrains entry, responses may need to widen access to relevant subjects and high-quality teaching, clarify how subject choices affect later eligibility and provide bridge provision and academic support for learners with identified gaps. Substantive prerequisites should be maintained where they are necessary for later learning, but learners need equitable opportunities to meet them.
Financial and practical barriers should be assessed beyond tuition costs. Diagnosis should consider participation and opportunity costs, including living costs, materials, transport, relocation, housing, childcare, lost earnings and the time demands associated with laboratory work, projects and placements. It should also examine whether geographical concentration and requirements for in-person participation restrict access. Financial support should address the specific barrier, learners and decision point concerned. Its coverage, amount, timing, accessibility and conditions should enable participation without making demanding STEM study appear financially risky.
The capacity of higher education institutions should be assessed by relating demand and the number of available and filled places to the staff, infrastructure, supervision, practical learning opportunities and learner support required for high-quality provision. Where capacity is insufficient, responses may include stable investment, shared provision or regional concentration, depending on the field, the critical mass required and the feasibility of learner mobility. Proposed expansion should be assessed for quality, sustainability, geographical accessibility and whether it creates additional capacity rather than relabelling or redistributing existing provision.
Bottleneck 2: Persistence and completion in STEM programmes
Persistence and completion are shaped by learners’ preparation and circumstances, and by the conditions they encounter within programmes, including curriculum sequencing, teaching, assessment, feedback, belonging, learner support and institutional capacity. A bottleneck in persistence and completion arises when learners who have the interest and potential to meet programme requirements cannot continue or complete because of avoidable constraints.
STEM programmes often combine cumulative learning, sequenced prerequisites and substantial project and laboratory work. Research by Seymour and Hunter (2019[90]) shows that losses during the early undergraduate years can affect also students who entered STEM with strong academic preparation. Their work challenges interpretations that attribute departure primarily to insufficient ability or commitment and directs attention towards the learning experiences students encounter. In particular, fixed beliefs about intelligence can support a selective view of teaching in which introductory courses are expected to identify the relatively few students presumed to have the capacity to succeed, rather than enable a wider range of learners to continue. Fixed beliefs can lead to resistance to reconsidering teaching and assessment. Diagnosis should therefore examine not only which learners leave or change route, but also how programmes define ability, organise early learning and assess whether teaching enables students to make progress.
It is important to distinguish constructive changes of route from displacement by avoidable barriers. Transfer may lead to a programme better aligned with a learner’s interests, preparation or objectives, while temporary interruption may allow a learner to manage their personal circumstances and later return. Measures of persistence and completion should therefore distinguish withdrawal from transfer, interruption, continued enrolment and completion in another programme or field.
Figure 1.5 shows the proportion of full-time new entrants to bachelor’s programmes who had completed a qualification three years after the theoretical end of their programme, in STEM and in health and welfare. Because the indicator covers full-time entrants only, it does not provide a complete picture in systems where part-time study is common. The figure presents an aggregate completion rate; complementary indicators are needed to distinguish continued enrolment, completion after changing field and departure from higher education among learners who had not completed by the observation point.
Completion rates of new entrants into bachelor's programmes by the end of the theoretical duration of their programme plus three years (2023)
Note: The students included in this survey are those who were new entrants to a tertiary level of education and who were full-time students at the time they entered the programme. The year of reference (2023) corresponds to a period three years after the theoretical end of the programmes these students entered, 2021 to one year after the theoretical end and 2020 to the theoretical end. The year of reference for Colombia is 2022.
Source: OECD (2025[91]), Education at a Glance 2025: OECD Indicators, Table B5.3, https://doi.org/10.1787/1c0d9c79-en.
Across OECD countries, 68% of full-time new entrants to STEM bachelor’s programmes had completed three years after the theoretical programme duration, compared with 63% across the EU25. Outcomes vary widely, from 84% in the Republic of Türkiye and 83% in Canada to 49% in Chile. Among the EU Member States shown, the range is narrower but remains substantial, from 71% in Poland and the Netherlands to 60% in Estonia, 59% in Sweden and 55% in Austria. Completion in STEM is lower than in health and welfare in almost every country shown. The average difference is 12 percentage points across OECD countries and 15 percentage points across the EU25, where 63% of STEM entrants completed compared with 78% in health and welfare. The gap reaches 22 percentage points in Sweden and Austria and 21 in Spain. It is small in the United Kingdom, absent in Iceland and reversed in New Zealand, where STEM completion is 3 percentage points higher than for health and welfare.
The differences between STEM and health and welfare may reflect variation in the characteristics and prior preparation of entrants, admission arrangements and the conditions learners encounter within programmes. Relevant programme conditions include curriculum structure and sequencing, prerequisite requirements, practical learning and institutional capacity.
The aggregate STEM rate also conceals differences among its constituent fields. Separate field-level data for the countries for which they are available indicate that completion in information and communication technologies is often below the national average across fields. Natural sciences, mathematics and statistics also record comparatively low completion, while outcomes in engineering, manufacturing and construction are more mixed. The difference between completion in any field and completion in the field of initial entry also indicates that some learners complete after changing field.
Explaining these patterns requires programme-level and learner-level evidence on where difficulties arise, which learners are affected and how they experience the conditions within their programmes. The following sections examine these conditions, beginning with programme culture and its influence on whether learners participate, seek help and recover from setbacks.
Programme cultures and collaborative learning influence participation, belonging and persistence
Sense of belonging can shape both whether learners choose a field and whether they persist once enrolled. It is a concept for whether a learner feels accepted, respected, supported and valued. It can influence willingness to participate, seek assistance and continue after setbacks (Strayhorn, 2018[92]). Learners are more likely to persist when they develop meaningful relationships with peers, educators and the professional communities associated with their field.
Belonging is shaped by everyday experiences, including classroom climate, interactions with peers and educators, teaching practices, opportunities to participate, and recognition of learners’ identities and contributions. It is an individual experience and an outcome of pedagogical and institutional arrangements (Duncan, Holt and Keenan, 2023[93]; Valdez and Kelp, 2023[94]). Highly competitive or exclusionary environments may discourage questions, collaborative study and contact with academic staff. Programme and institutional cultures can weaken belonging when they present difficulty and help-seeking as evidence of unsuitability, treat high attrition as proof of academic rigour or imply that successful learners should progress without assistance (Seymour and Hunter, 2019[90]).
Interventions to address sense of belonging should not seek to encourage learners to become more resilient or confident, as this places responsibility on individuals for conditions that programmes and institutions can change, they need to address structural barriers. Teaching and collaborative learning can strengthen belonging, engagement and academic participation by enabling learners to exchange reasoning, build relationships and make peer support and help-seeking more routine.
Collaborative learning may include problem-solving in pairs, laboratory teams, study groups, design projects, peer instruction and interdisciplinary assignments. Its contribution depends, however, on how activities are designed and facilitated and whether all learners participate meaningfully (Duncan, Holt and Keenan, 2023[93]; Tharayil et al., 2018[95]; Hodges, 2018[96]; Møgelvang and Nyléhn, 2022[97]). Some learners may dominate discussion, while others may be assigned less visible or peripheral tasks. Research on engineering design teams shows, for example, that gendered task allocation can place women disproportionately in non-technical roles and signal that their technical contributions are less expected or valued. Masculinised understandings of engineering expertise may also lead women to understate their skills and receive less access to central technical tasks (Schauer et al., 2025[98]).
Effective collaborative learning requires institutional capacity. Educators need time and expertise to design genuinely interdependent tasks, establish transparent expectations, distribute and monitor responsibilities, support productive interaction and provide constructive and actionable feedback. Assessment should make individual learning and contribution visible alongside the collective output. Programme design, facilitation and assessment should give every learner access to substantive disciplinary and technical reasoning and opportunities to undertake different forms of technical and leadership responsibility.
Gateway courses and assessment can identify learning needs or reinforce exclusion
Gateway courses develop and assess the disciplinary foundations required for later study. Failure in introductory courses can delay progression or lead learners to leave STEM. Where these courses are used to reduce cohort size rather than support learning, they can function as “weeding out” courses (Basu, Malik and Graunke, 2025[99]; Seymour and Hunter, 2019[90]). Competitive cultures and certain teaching and assessment practices can then contribute to departure, affecting also learners with strong academic records (Seymour and Hunter, 2019[90]).
Where substantial numbers of learners encounter the same difficulty, the explanation may not lie solely in their prior preparation. Programme analysis should consider whether prerequisites correspond to the knowledge required, whether core concepts are introduced and practised adequately, whether assessment reflects the stated learning outcomes, and whether curriculum sequencing allows competence to develop progressively. Recurring difficulties in gateway courses should therefore be examined as possible indicators of problems in prerequisites, teaching, assessment or programme design, with diagnostic assessment linked to timely learning support rather than additional selection.
However, when gateway courses are used for development, diagnostic assessment can help learners identify gaps, understand programme expectations and improve their study strategies. It can also show educators where difficulties occur and which learners are affected. Regular, lower-stakes assessment provides opportunities to address problems before a high-stakes examination puts continued participation at risk (Morris, Perry and Wardle, 2021[100])
Mentoring relationships are valuable, but quality matters
Mentoring and supportive relationships can help learners interpret academic expectations, respond to setbacks, identify research and professional opportunities and envisage themselves as future members of a field. Contact with peers, academic staff and professionals may provide technical guidance, encouragement and recognition, and access to opportunities beyond the formal programme (McKean et al., 2024[101]; Ravishankar et al., 2024[102]).
However, access to and use of these relationships are unevenly distributed among learners. Those who are already connected with academic and professional networks may find mentors and research opportunities informally, while others may have fewer opportunities or less time in which to establish such relationships (Reeves et al., 2022[103]; Zuckerman, Lo and Juavinett, 2024[104]; Torres-García, Zamudio-Orozco and Meráz García, 2026[105]). It requires institutional capacity to organise mentoring at sufficient scale, connect learners with suitable mentors, prepare and support mentors, monitor whether relationships are effective and provide alternative sources of support where necessary.
Financial and practical pressures can interrupt otherwise viable study
Financial and practical pressures can interrupt study even when learners are academically able and committed to continuing. As discussed under Bottleneck 1, costs for fees, accommodation and living, and forgone earnings can affect whether a STEM route remains financially feasible after enrolment. Financial and academic difficulties can reinforce one another. Financial pressure may reduce the time available for study, while failure in a prerequisite course can lengthen the programme and increase its cost. A temporary loss of income, housing or care arrangements can therefore lead to academic difficulty, and academic delay can create further financial pressure.
Paid employment does not necessarily impede persistence. Its effects depend on the number and predictability of working hours, the learner’s circumstances and whether the work is related to the programme. Intensive employment is generally associated with weaker academic outcomes and a greater risk of departure, while relevant work can provide income, professional contacts and opportunities to apply learning (Kocsis, Alter and Pusztai, 2022[106]; Choi, 2017[107]).
Part-time study, adaptable schedules, hybrid participation, temporary interruption and structured return can help learners continue during periods of employment, caring responsibilities, illness or other disruption. However, flexible routes may remain difficult to use if required courses are offered infrequently, practical sessions require full-time attendance or part-time learners have less access to projects, research and professional opportunities (El Galad, Betts and Campbell, 2024[108]).
An interruption does not necessarily indicate a loss of interest or commitment. Learners may stop because of financial difficulties, caring responsibilities or problems in particular courses and later seek to return. Institutions therefore need clear arrangements for interruption, re-entry and guidance rather than treating all interruptions as academic disengagement (West, Chen and Boika, 2024[109]). Making these arrangements usable requires reliable records of completed learning and learner status, clear responsibility for maintaining proportionate contact during the interruption, and timely academic and practical guidance before return. This can be difficult where learners do not enrol for one or more academic years, particularly if information and responsibilities are fragmented across programme, administrative and support functions.
Institutional capacity determines whether student support is timely, connected and sustainable
Persistence and completion also depend on whether institutions have the capacity to help learners with different preparation and circumstances meet programme standards.
A programme may be difficult to navigate as a whole. Programme leadership is needed to connect courses, co-ordinate teaching and assessment, align support with curriculum requirements and identify recurring barriers to progression (Roehrig et al., 2021[110]; Frost and Ackrill, 2025[111]). Staff acting as programme leaders need sufficient time, authority and evidence to improve programme coherence (Aalde and Staal Jenset, 2023[112]).
Student support includes academic, personal and psychosocial support. The aim is to help learners understand expectations, address difficulties and respond to setbacks, particularly at predictable transition points. Academic support includes tutoring, learning guidance and assistance with course requirements. Personal support helps learners address practical circumstances affecting participation, including financial, administrative, accessibility, health and caring-related needs. Psychosocial support includes counselling and other specialist assistance with well-being, adjustment, relationships and belonging.
Student support needs to be accessible, connected with programme requirements and co-ordinated across academic and specialist services. Fragmented services may require learners to navigate several offices when they have limited time or capacity to do so. Academic programmes and specialist services therefore need clear connections, responsibilities and referral arrangements, while preserving confidentiality and professional boundaries. Staff providing student support need sufficient time and recognition for advising and mentoring (Hofstra et al., 2022[113]).
Early-warning systems can reveal patterns in performance, participation and credit accumulation, but they are useful only when institutions can understand the causes and provide timely support (Akçapınar, Altun and Aşkar, 2019[114]); (Rets, Herodotou and Gillespie, 2023[115]). These systems should be used carefully. Learner data do not explain why difficulties arise, and risk classifications may reproduce inequalities or label learners. When many learners struggle in the same course or transition, the problem may lie in teaching, assessment, curriculum sequencing or programme organisation. Learning analytics also raise concerns about privacy, transparency, consent, bias and institutional responsibility (Cerratto Pargman and McGrath, 2021[116]; Tzimas and Demetriadis, 2021[117]).
Evidence should therefore be examined at both learner and programme levels. It can identify learners who need immediate support and reveal aspects of programme design that require improvement. Responsibility for acting on this evidence is distributed across institutional and disciplinary structures. The Student Assessment of their Learning Gains, developed by Seymour et al. (2000[118]), was designed for this purpose. It focuses on students’ reported gains in understanding, skills, attitudes and the integration of knowledge, rather than general satisfaction with the educator or course.
Implications for diagnosing a persistence and completion bottleneck
Diagnosing a bottleneck in persistence and completion needs to distinguish withdrawal, transfer, temporary interruption, continued enrolment, delayed completion and completion in another programme or field. Transfer or interruption may represent an appropriate change of route rather than a failure to persist. Completion indicators should also be disaggregated by field, programme and population group and interpreted in relation to the observation period and the learners covered. Indicators limited to full-time entrants, for example, do not provide a complete picture where part-time study is common.
Diagnosis requires evidence from learners, programme teams, higher education institutions and public authorities. Learners can provide information on the circumstances and experiences underlying transfer, interruption, delay or withdrawal; programme teams and institutions can identify recurring difficulties in courses, transitions and support arrangements; and public authorities can enable analysis across institutions, education sectors and periods of study. Responsibility for addressing identified constraints lies primarily with higher education institutions and programme leaders, supported by public authorities where constraints concern funding, regulation, system structures or the secure linkage of data.
Diagnosis should identify when and where difficulties arise, whether they are concentrated in particular courses or transition points, and how outcomes differ across programmes and learner groups. Prior preparation needs to be considered alongside programme culture, teaching, assessment, feedback, curriculum sequencing, belonging, collaborative learning, mentoring, financial and practical pressures, flexible study arrangements, access to academic and psychosocial support, and institutional capacity. Differences in completion between fields do not by themselves show whether the causes lie in entrants’ preparation, admission arrangements or the conditions encountered within programmes.
Following learners across institutions, education sectors and periods of study is necessary but can be difficult. An institution may record a learner as having withdrawn when the learner has transferred, resumed study elsewhere or subsequently completed another programme. A unique student identifier can support secure linkage of enrolment, credit, transfer and completion records and help distinguish permanent departure from movement, interruption followed by return and completion elsewhere. Its usefulness depends on consistent reporting, data quality, interoperability and clear rules governing access, linkage, privacy and permissible uses. An identifier links records but does not explain why a learner changed route and should therefore be combined with programme-level evidence and information from learners.
Programme culture, belonging and collaborative learning should be examined together as dimensions of programme quality. Evidence should show whether learners feel accepted, supported and recognised; can participate substantively, seek help and receive constructive feedback; and have a reasonable opportunity to recover from setbacks without difficulty being treated as evidence of unsuitability. Analysis should examine classroom climate, interactions with peers and educators, the organisation and facilitation of collaborative work, and whether expectations, responsibilities and assessment criteria are transparent. In collaborative activities, it should determine whether all learners participate in substantive disciplinary and technical reasoning or whether some dominate while others are confined to peripheral or less visible tasks. It should also assess whether particular groups receive less recognition, have weaker access to informal advice or are given fewer opportunities to undertake technical and leadership responsibilities.
Recurring difficulty in a gateway course may reflect gaps in prior preparation, but it may also indicate misaligned prerequisites, inadequate opportunities to practise, ineffective teaching, unsuitable assessment or problems in curriculum sequencing. Diagnosis should examine whether gateway courses develop and assess the foundations required for later study or function mainly to reduce cohort size. Diagnostic and lower-stakes assessment can help identify difficulties early, but only if it is connected with timely support and a reasonable opportunity to meet transparent academic standards.
Financial and practical pressures should be examined throughout study rather than only at entry. Diagnosis should consider direct costs, living and housing costs, caring responsibilities, health, employment and the time demands of laboratory work, fieldwork, projects and placements. It should also examine how academic and financial difficulties reinforce one another, since failure in a prerequisite course can extend the duration and cost of study, while financial pressure can reduce the time available for learning. Paid employment should not automatically be treated as a barrier: its effects depend on the number and predictability of working hours, the learner’s circumstances and whether the work contributes to professional learning.
Flexible study, adaptable schedules, temporary interruption and structured return should be assessed for whether they enable learners to continue or resume study while preserving coherent course sequences, substantive practical and collaborative learning, academic guidance and access to research and professional opportunities. Diagnosis should determine whether apparently flexible routes are usable in practice, including whether required courses are offered sufficiently often and whether practical activities are accessible to part-time learners. Institutions also need reliable records of completed learning and learner status, clear responsibility for proportionate contact during interruption where possible, timely guidance before return and arrangements that allow learners to resume without unnecessary loss of relevant previous learning.
Institutional capacity should be assessed in relation to the scale, timeliness and coherence of teaching and learner support. Effective teaching, practice, formative assessment, feedback, supervised laboratory and project work, academic guidance and psychosocial support depend on sufficient academic and technical staffing, infrastructure and organisational capacity. Diagnosis should examine whether support is accessible and connected with mainstream study or fragmented across services that learners must navigate independently. It should also assess whether academic programmes and specialist services have clear responsibilities and referral arrangements for responding to academic difficulties, financial and personal pressures, discrimination and exclusion, while preserving confidentiality and professional boundaries.
Bottleneck 3: Progression to higher levels of STEM studies
Progression to higher levels in STEM can take several forms and does not necessarily follow a linear sequence. Learners may progress from a short-cycle programme to a bachelor’s programme, from bachelor’s to master’s study, or from master’s to doctoral education. They may also enter employment after an earlier qualification and develop advanced STEM skills through professional practice, work-based learning or a later return to education. A progression bottleneck arises when learners with the interest and potential to benefit from advanced study cannot progress because of avoidable academic, financial, informational, geographical or institutional barriers.
Figure 1.6 provides context for examining progression. Panel A presents core STEM graduates as a share of all graduates at each level, while Panel B presents internationally mobile students as a share of core STEM enrolment. The panels have different denominators, do not follow the same learners over time and are not independent, since the graduates in Panel A include internationally mobile graduates. Considering both indicators together helps assess the relative prominence of core STEM at advanced levels and the extent to which advanced provision draws on domestic and internationally mobile learners.
Core STEM graduates as a percentage of all graduates, and internationally mobile students as a percentage of core STEM enrolment, by level of education, 2024
Note: Panel A shows graduates in core STEM as a percentage of all graduates at the level indicated. Panel B shows internationally mobile students as a percentage of all students enrolled in core STEM at the level indicated. The two panels have different denominators and are not independent, since graduate counts in Panel A include internationally mobile graduates. Core STEM comprises natural sciences, mathematics and statistics (ISCED‑F 05), information and communication technologies (ISCED‑F 06) and engineering, manufacturing and construction (ISCED‑F 07). Mobile students are identified on the basis of country of prior education and include nationals of the reporting country who completed their prior education abroad. Countries are ranked in descending order of the core STEM share of doctoral graduates, and the same order is used in both panels. France is not shown, as it does not report the number of international students. Doctoral data on international students are not available for the Netherlands. Values for Graduates in Sweden and Portugal refer to 2023.
Source: OECD (2026[54]), Distribution of enrolled students, new entrants and graduates by field of education, (database), OECD Data Explorer, https://data-explorer.oecd.org/s/5kw; OECD (2026[40]), Number of enrolled students, graduates and new entrants by field of education, (database), OECD Data Explorer, https://data-explorer.oecd.org/s/5ky).
Core STEM accounts for 24% of bachelor’s graduates and 43% of doctoral graduates on average across the countries shown. Its share is higher among doctoral than bachelor’s graduates in almost every country, although the difference ranges from 45 percentage points in Luxembourg and 35 percentage points in Italy to three percentage points or less in Croatia and Bulgaria.
The greater representation of core STEM emerges primarily at doctoral level rather than through a steady increase across all three levels. Core STEM accounts for 25% of master’s graduates on average, almost the same as its share among bachelor’s graduates, and its share is lower among master’s than bachelor’s graduates in 17 of the 32 countries shown. The difference is particularly pronounced in Korea, where core STEM accounts for 32% of bachelor’s graduates but 24% of master’s graduates, and is also visible in Croatia, Hungary, Romania and Bulgaria. These cross-sectional differences show how the composition of graduates varies by level, but do not indicate whether individual learners progress from one level to the next.
Internationally mobile students represent larger shares of core STEM enrolment at more advanced levels. Across EU Member States, they account for 8% of core STEM bachelor’s enrolment, 21% of master’s enrolment and 30% of doctoral enrolment. The largest increase is therefore between bachelor’s and master’s levels. Australia, the United Kingdom and Canada differ from the general pattern because the international share is highest at master’s level, reaching 82%, 76% and 62%, respectively.
Reading the panels together illustrates that countries with similar core STEM shares among doctoral graduates can differ substantially in the international composition of core STEM doctoral enrolment. Italy and Switzerland both record high core STEM shares among doctoral graduates, at 55% and 46% respectively, but internationally mobile students account for 11% of core STEM doctoral enrolment in Italy and 77% in Switzerland. Poland and Greece combine core STEM shares among doctoral graduates of 44% and 40% with international shares of around 1%, while Denmark, Finland, Norway and Belgium record broadly comparable graduate shares alongside international shares above 50%. These comparisons do not establish a relationship between the two indicators, but they show that a similar representation of core STEM among doctoral graduates can be associated with very different patterns of international participation.
International mobility is an important part of the development and circulation of advanced STEM skills. As global competition for learners, researchers and experienced professionals intensifies, countries and higher education institutions may seek to attract international talent, particularly in specialised and emerging fields. The patterns in Figure 1.6 raise different questions for diagnosis. Where internationally mobile students account for a small share of advanced core STEM enrolment, analysis may focus on domestic participation and progression, the availability and attractiveness of advanced study, and institutional capacity. Where the international share is high, analysis should also examine access and support for internationally mobile students, recruitment conditions and the exposure of advanced provision to changes in international mobility. In both cases, diagnosing a progression bottleneck requires learner-level evidence on movement between levels, fields and institutions and on the conditions shaping decisions to continue, change route, enter employment or return to education. Diagnosis should examine attraction, recognition, integration, retention, return and international collaboration as connected parts of mobility, while also considering whether dependence on international recruitment substitutes for investment in domestic talent development or transfers shortages between countries.
For short-cycle programmes to support progression, occupational and educational routes need to be connected
Short-cycle qualifications in post-secondary education generally develop occupation- and industry-specific skills. Their value depends not only on immediate employment outcomes, but also on whether graduates can adapt as technologies, occupations and skills requirements change. Focused preparation may support rapid entry into work but lose value over time if it provides limited foundations for adaptation and further learning. The initial employment advantage associated with vocational education may therefore decline over working life (Hanushek et al., 2016[119]).
Evidence on short-cycle post-secondary education needs to be interpreted in context. Much of the empirical literature concerns community colleges and other public two-year institutions in the United States (Soliz, 2023[120]). Programmes and institutions differ across countries in their duration, admission arrangements, curricular orientation, occupational standing and progression rights (Cavallaro, 2023[121]). Community colleges, higher technical institutions, universities of applied sciences and dual vocational arrangements should therefore not be treated as interchangeable. Evidence from one model needs to be interpreted in relation to the governance, qualification system, labour market, employer involvement and transfer arrangements of the relevant system (Li and Pilz, 2021[122]).
Many countries across the OECD area have introduced short-cycle in STEM. Ideally, these programmes should provide credible routes to further study. Not all learners will wish or need to progress immediately, but they should be able to return later to deepen their advanced STEM skills or respond to changing occupational requirements. However, formal progression rights do not always make this possible in practice. Credits may count only as electives rather than towards the sequenced requirements of a bachelor’s programme, requiring learners to repeat content, complete additional prerequisites or extend their studies. This creates a bottleneck in progression.
Progression from bachelor’s to master’s study depends on the value, visibility and accessibility of further specialisation
Bachelor’s programmes fulfil a double role as they prepare graduates for employment and provide foundations for master’s and doctoral study. The balance between these two functions varies across STEM fields and occupations. A bachelor’s degree is sufficient for many occupations, while research-intensive and specialised roles may require or favour an advanced qualification. This varies by field, programme and graduate characteristics (Okahana and Hao, 2019[123]; Skjelbred and Nesje, 2022[124]).
Learners need to assess whether a master’s qualification is necessary for their intended career, whether a suitable specialisation is available and whether its expected value justifies the additional time, cost and forgone earnings. Some may enter employment, combine work and study or return after gaining professional experience. A bottleneck arises when learners who would benefit from master’s study cannot progress because of avoidable barriers rather than an informed decision to follow another route.
Grades influence admission, financial support and access to research opportunities, but they need to be interpreted in context. Applicants may have studied in programmes with different curricula, assessment practices, institutional resources and opportunities for practical or advanced work. Evidence from biomedical sciences indicates that grades and standardised test scores have limited or variable relationships with later doctoral outcomes (Walters et al., 2022[125]). Holistic admissions approaches seek to consider grades alongside research experience, relevant competence, previous opportunities and programme fit (Young et al., 2022[126]; Posselt et al., 2023[127]).
When a master’s qualification is a formal or practical requirement for doctoral study in a field, barriers to master’s entry and completion can also restrict access to doctoral education. The availability, cost and geographical distribution of suitable master’s programmes, together with admission requirements and access to research experience, can therefore shape who is able to pursue a research route.
Undergraduate research can help learners understand how scientific knowledge is produced, develop research skills and decide whether advanced study suits their interests. Experience in laboratory work, course-based research, independent projects or research placements is associated with stronger scientific understanding, confidence and preparation for postgraduate study. It is also associated with persistence and completion, although its causal effects are difficult to establish because participation is not randomly assigned (Seymour et al., 2004[128]; Lopatto, 2007[129]; Haeger, Bueno and Sedlacek, 2024[130]).
Access to research experience and specialised programmes is uneven (Pierszalowski, Bouwma-Gearhart and Marlow, 2021[131]; Young et al., 2022[126]; Crumley-Effinger, 2023[132]; Topliceanu, Blanchard and Collier, 2026[133]). Learners may lack information or contacts, face informal selection processes or be unable to accept unpaid or poorly paid positions. Opportunities may be concentrated in research-intensive institutions or a limited number of locations. Relocation or international mobility can widen learners’ options but introduce costs and barriers related to accommodation, language, immigration requirements and qualification recognition.
Progression across disciplines requires transparent routes
Many scientific and technological developments depend on integrating knowledge and methods from several disciplines. Enabling learners to progress across disciplines can widen access to emerging fields, connect established areas of expertise and support the development of new combinations of advanced STEM skills. Such progression does not replace disciplinary depth: learners need sufficient foundations to understand, evaluate and integrate the contributions of the fields involved.
Progression across disciplines can support the development of advanced STEM skills, especially in emerging fields that combine several disciplines. Quantum technologies draw on physics, mathematics, computing and engineering; data science combines mathematics, statistics, computing and domain knowledge; and robotics connects engineering with control, sensing and AI. Learners may enter these fields by building on one discipline or through an interdisciplinary programme.
A key challenge is to recognise relevant previous learning while ensuring that learners have the foundations required by the new field. Programme titles do not always show the depth developed in each discipline, how the disciplines are integrated or which further study and employment routes the programme supports. Learners and receiving institutions therefore need clear information about assumed prior knowledge, areas of disciplinary depth, the purpose of integration, additional preparation, assessment and progression routes (Mazzeo Ortolani, 2025[134]). Finding a balance between requirements and bridging is important. Rigid admission requirements may exclude applicants with suitable competence and potential, while admission without adequate foundations or bridge provision may lead to difficulties later in the programme.
Interest and success in doctoral education depends on viable routes and sufficient capacity
A bottleneck arises when qualified and interested candidates cannot progress because funded positions, suitable supervision or research infrastructure are unavailable, or because study and career conditions make participation impractical.
Applicants to doctoral education programmes have different levels of research preparation. Admissions criteria such as grades, publications, references and institutional affiliation may reflect applicants’ previous opportunities as well as their competence. Research-oriented master’s programmes may include extensive methods training, thesis work and laboratory experience, while professionally oriented programmes may focus on specialised employment. Limited previous research experience should therefore not automatically be interpreted as limited research potential (Young et al., 2022[126]). More holistic assessment criteria considers academic achievement alongside relevant competence, research experience, prior opportunities and programme fit but requires clear and consistent criteria (Posselt et al., 2023[127]; Orellana et al., 2025[135]).
In many STEM fields doctoral studies involve substantial forgone earnings while offering uncertain financial and career returns. However, doctoral education is essential to the renewal of research, innovation and higher education capacity (Ganapati and Ritchie, 2021[136]). This suggests that doctoral routes should accommodate movement between employment and research. Professional experience can develop advanced STEM skills and generate research questions, but returning candidates may face limited recognition, unsuitable funding or difficulty combining study with employment and caring responsibilities. Part-time and other flexible routes can widen access, provided they maintain the depth of research training and offer appropriate supervision and support (Gardner and Gopaul, 2012[137]).
Progression in doctoral studies also depends on institutional capacity. Limited funding, supervision or research infrastructure may restrict admission even when qualified candidates are available. Capacity includes manageable workloads, prepared supervisors, consistent guidance, access to research communities and infrastructure, professional development, and clear procedures for addressing supervisory difficulties (Sherman et al., 2021[138]; Bahnson and Abane, 2026[139]; Bach, Armstrong and Jones, 2025[140]).
Implications for diagnosing a progression bottleneck
Diagnosing a progression bottleneck is challenging. In a nutshell it means that learners who are interested in and could benefit from higher-level study face avoidable barriers to an appropriate route. Available indicators need to be interpreted carefully. Graduate shares at different qualification levels do not track individual movement between levels, while graduate shares and the international composition of enrolment concern different populations.
Diagnosis requires evidence from multiple sources, including learners, programme teams, higher education institutions, employers and public authorities. Learners can provide information on their objectives, decisions and the academic, financial and practical barriers they encounter. Programme teams and higher education institutions can identify constraints related to preparation, admissions, recognition, provision, funding, supervision and research opportunities. Employers can provide evidence on occupational destinations, the value attached to advanced qualifications and opportunities to move between employment and education. Public authorities can enable analysis across institutions, qualification levels and education sectors and identify constraints arising from funding, regulation, recognition systems, geography and international mobility. Responsibility for addressing identified constraints is distributed among these actors according to the route and condition concerned.
Progression should be examined as a connected route rather than as a series of isolated admission decisions. The following are key aspects.
For progression from short-cycle to bachelor’s study, diagnosis should examine curriculum alignment, guidance, admission requirements, credit recognition and access to additional preparation. Formal progression rights are insufficient if credits do not meet the sequenced requirements of the receiving programme, learners must unnecessarily repeat previous learning or required courses are inaccessible. Analysis should also reflect differences among national qualification systems, institutional models, labour markets and forms of employer involvement rather than assuming that short-cycle programmes operate similarly across countries.
For progression from bachelor’s to master’s study, diagnosis should consider the purpose and expected value of further specialisation, the availability of suitable programmes, admission requirements, direct costs, forgone earnings and opportunities to combine study with employment or return later. Entry into employment should not be treated as evidence of constrained progression when a bachelor’s qualification provides appropriate preparation for the learner’s intended occupation. The relevant question is whether learners who would benefit from master’s study can identify, access and finance an appropriate route.
Access to research experience requires particular attention because it can shape learners’ understanding of research, development of research competence and preparation for postgraduate study. Laboratory work, course-based research, independent projects and research placements can help learners assess whether postgraduate study suits their interests. Access may nevertheless depend on information, academic contacts, informal selection, institutional location and the ability to accept unpaid or poorly paid opportunities. For doctoral progression, diagnosis should include among others sustainable funding, supervision capacity, and research infrastructure. Learners also need preparation for careers within and beyond academia (OECD, 2023[141]).
Progression across fields requires flexible entry combined with adequate foundations. Rigid requirements may exclude applicants with relevant competence and potential, while admission without the required foundations or suitable bridge provision may create difficulties later. Programme titles alone may not indicate disciplinary depth, how knowledge and methods are integrated or which further educational and professional routes are available. Recognition arrangements should reduce unnecessary repetition while preserving the competence required at the receiving level.
Financial, geographical and administrative barriers can affect progression across all routes. Diagnosis should consider tuition and living costs, forgone earnings, relocation and accommodation, the availability of suitable programmes, language and immigration requirements, qualification recognition and the feasibility of combining study with employment or caring responsibilities. International mobility can widen access to specialised provision but can also introduce additional costs and administrative barriers. Where advanced provision depends substantially on internationally mobile learners, analysis should consider their access and participation alongside domestic progression and the exposure of provision to changes in international mobility.
Bottleneck 4: Career entry and effective skills use
A career entry and effective skills use bottleneck arises when people with relevant advanced STEM skills cannot obtain suitable work or use and develop those skills in employment. The availability of suitable work depends partly on the structure and technological intensity of economic and research activity. Countries and regions with fewer research-intensive, capital-intensive or technologically complex organisations may generate less demand for some advanced STEM profiles, while demand may be concentrated in particular sectors, organisations or locations (OECD, 2025[37]). This does not necessarily indicate a failure of education or recruitment. Diagnosis therefore needs to consider the industrial, research and innovation context alongside the circumstances of individuals and the practices of employers.
Policymakers need to consider two connected groups of constraints affecting career entry and the effective use of advanced STEM skills (Table 1.3). Constraints affecting individuals concern unequal access to substantive professional experience, occupational information, professional networks and suitable employment. Constraints arising on the employer side concern the interpretation and validation of advanced STEM skills, recruitment criteria, onboarding, work organisation, continued development, working conditions and advancement. These constraints can reinforce one another: for example, unequal access to professional experience can limit the evidence candidates can present, while narrow experience requirements can further restrict entry to suitable work.
The policy concern is not only whether graduates find employment, but whether they obtain suitable and sustainable work, apply their advanced STEM skills and gain access to feedback, learning and greater responsibility. Increasing graduate numbers may have limited effects if employers cannot recognise relevant competence, use narrow recruitment criteria or organise work in ways that restrict skills use and development.
Unequal access to substantive professional experience
Professional experience during study is an important bridge between advanced STEM education and suitable employment. It can help learners understand possible roles, make informed employment choices and demonstrate to employers that they can apply disciplinary knowledge in professional settings (Luk and Chan, 2021[142]).
Access to professional experience during study uneven. Unpaid or poorly paid placements may be impracticable for learners who need paid employment or must meet additional transport, accommodation or dependent-care costs. Existing work and family responsibilities may create further barriers. Access can also depend on institutional and professional networks. Established relationships between selective universities and prestigious employers may give some learners easier access to internships and subsequent employment, while excluding others with comparable potential (Moss-Pech, 2021[143]). Participation alone does not establish that an experience is accessible or educationally valuable. Outcomes depend on the relevance and quality of the work, the responsibilities assigned, the supervision and feedback provided, and opportunities to connect workplace experience with academic learning. Placements may provide substantive technical and professional work, but they may also consist mainly of routine, administrative or poorly defined tasks (Luk and Chan, 2021[142]; Vlachopoulos and Pachni Tsitiridou, 2026[144]; Schneider et al., 2024[145]). Higher education institutions and employers share responsibility for quality.
Unequal access can have lasting effects because employers may later treat professional experience as evidence of employability (Luk and Chan, 2021[142]). Learners who obtain substantive placements can strengthen their position in recruitment, while others may appear less prepared even when they have comparable disciplinary foundations and potential. Inequality in access to professional experience during study can therefore reinforce employer-side difficulties in interpreting and validating advanced STEM skills.
Financial, geographical and personal constraints can limit access to suitable employment
Employment outside a person’s field of study or below their qualification level may reflect limited access to suitable work rather than a lack of relevant skills or interest. Access depends not only on an individual’s qualifications and advanced STEM skills, but also on whether relevant work is available and realistically accessible.
Housing, relocation and transport costs limit where people can work. Low-paid or temporary roles and long recruitment processes may be impractical for those who need a secure income. Caring responsibilities, disability-related needs and immigration conditions may further restrict their options. These constraints are especially important when suitable work is concentrated in particular regions, sectors or organisations. Such concentration may reflect the location of research infrastructure, capital-intensive production, technologically advanced firms, public research organisations or related economic clusters. In regions where these activities are limited, fewer roles may provide opportunities to use and develop particular advanced STEM skills.
Working conditions also shape whether suitable roles are accessible. Temporary contracts, mobility requirements, uncertain progression and long or unpredictable hours may exclude people unable to accept the associated risks. People may choose a more accessible and secure role over one that better matches their qualification but requires relocation, offers insecure income or conflicts with other responsibilities.
Constraints may continue after employment begins. Relocation, irregular hours, extended travel or lower short-term earnings may limit access to specialised experience, research collaboration, substantial technical assignments and greater responsibility. The relevant question is therefore whether people have a viable opportunity to obtain and progress in work that uses and develops their advanced STEM skills.
Unequal access to occupational information and professional networks
Occupational information and professional networks can affect whether individuals identify suitable employment, understand recruitment requirements and connect their advanced STEM skills with different roles and sectors. Information about tasks, working conditions, career prospects and competence requirements may come from programme staff, employers, former placement providers, professional associations, alumni and peers, as well as public vacancy information.
Vacancy descriptions may not explain what the work involves, which requirements are essential, how academic or professional experience will be interpreted, or whether a role offers opportunities to use and develop advanced STEM skills. Programme and occupational labels can be particularly unclear for roles that combine several fields or apply advanced STEM skills outside conventional STEM sectors.
Access to this information and to professional networks is uneven (Moss-Pech, 2021[143]). Established relationships between selective institutions and employers may give some learners earlier information about opportunities, greater familiarity with recruitment practices and easier access to professional references. Candidates outside these networks may have less contextual support when interpreting requirements and presenting their experience These differences become especially important where recruitment relies on informal recommendations, knowledge of particular sectors or the ability to present qualifications and experience in forms employers readily recognise.
Furthermore, substantial experience requirements can create a circular barrier: candidates need experience to obtain their first role but cannot gain recognised experience without access to substantive work. Unequal access to placements, professional networks and information can reinforce this problem. Lengthy recruitment, repeated internships, insecure initial employment and mobility requirements may further exclude candidates who cannot absorb the financial or personal costs.
Employers may have difficulty interpreting and validating advanced STEM skills
Employers may find it difficult to judge what applicants’ qualifications, skills and experience show about their ability to perform a role. Because employers cannot observe performance before appointment, they often use qualifications, previous experience and other evidence as signals of potential (Spence, 1973[27]).
The main issue is whether employers can interpret and validate relevant advanced STEM skills without relying unnecessarily on narrow proxies, that is, familiar institutions, employers or directly comparable experience (Stans et al., 2025[146]). Particularly difficult to interpret can be interdisciplinary programmes, international qualifications, and movement between fields (Tomlinson and Anderson, 2020[147]). Work samples, structured technical tasks and competence-based assessment can provide more direct evidence but may increase recruitment costs or disadvantage candidates with fewer opportunities to produce recognised evidence.
Doctoral and interdisciplinary graduates, people moving between fields, international graduates and candidates from less familiar institutions may be especially affected because employers can find their competence harder to interpret (Ganapati and Ritchie, 2021[136]; Sinche et al., 2017[148]; Skakni et al., 2025[149]).
Employers differ in their willingness and capacity to support learning after recruitment
Starting a professional role involves further learning, particularly when advanced STEM skills are concerned. Graduates may have strong disciplinary, analytical and research foundations but still need to learn the organisation’s processes, technologies, expectations, regulations and working practices through onboarding. These are not necessarily gaps in higher education because much of this knowledge is specific to the workplace (Mitschelen and Kauffeld, 2025[150]).
Onboarding typically combines formal training, supervised work, feedback, mentoring and informal learning. Effective onboarding should clarify the role, connect previous learning with workplace tasks, develop organisation- and sector-specific knowledge, explain relevant standards and responsibilities, provide feedback and support, and enable progression towards greater autonomy and responsibility. Weak onboarding can leave employees uncertain about their responsibilities, how to apply their skills and where to seek support. This can delay learning, weaken performance and job satisfaction, and increase intentions to leave (Cingillioglu and Schoettner, 2025[151]; Bauer et al., 2024[152]).
Claims that graduates are not sufficiently “work ready” may indicate gaps in education but may also reflect an expectation that they already possess organisation-specific know-how, which requires effective onboarding (OECD, 2021[153]; OECD, 2026[154]). Organisations may employ people with advanced STEM skills but lack the expertise, management practices or structures needed to use them effectively. This reflects an organisation’s ability to recognise, understand and apply relevant knowledge, often described as its absorptive capacity (Cohen and Levinthal, 1990[155]). Organisations with low absorptive capacity of advanced STEM skills may struggle to assess employees’ skills, design suitable work or support their continued development.
The scope for public policy intervention is limited because employers are primarily responsible for work organisation. Public authorities and intermediary organisations may support management development, knowledge exchange and collaboration, but employers determine whether advanced STEM skills are used through substantive tasks, professional judgement, participation in decisions and progression towards greater responsibility.
Working conditions and advancement structures affect retention
Suitable employment at entry does not ensure that early-career workers can remain, develop and progress. Employment security, pay, working hours, organisational culture and access to significant assignments affect whether they gain experience and move towards greater autonomy and responsibility. This is an area for public policy intervention.
Temporary contracts, uncertain career prospects and mobility requirements can make STEM careers difficult to sustain. These conditions may also limit access to the assignments, relationships and experience through which advanced STEM skills become recognised. In academia, insecure employment is associated with stress, exhaustion and uncertainty (Solomon and Du Plessis, 2023[156]).
Career paths may diverge after parenthood. Evidence from the United States shows that both mothers and fathers are more likely to leave full-time STEM employment after becoming parents, but departures are more common among mothers. Women are also more likely to move to part-time work or leave employment altogether (Cech and Blair-Loy, 2014[157]). Similar differences are found in progression to permanent academic positions (Lassen and Ivandić, 2024[158]).
Implications for diagnosing a career entry and effective skills use bottleneck
Diagnosing a career entry and effective skills use bottleneck requires more than employment rates, qualification matching and occupational classifications. Relevant opportunities may be concentrated in particular sectors, organisations and regions and depend on the economic, research and innovation structure of a country or region.
Multiple sources of evidence are needed for diagnosis. Individuals can describe their access to opportunities, recruitment experiences, work and development; employers can provide evidence on competence requirements and employment practices; and institutions and public authorities can identify patterns across programmes, sectors, regions and population groups. Responsibility depends on the constraint concerned, although employers are primarily responsible for recruitment and for the conditions under which advanced STEM skills are used and developed at work.
The following are key aspects of diagnosing a bottleneck in career entry and effective use of advanced STEM skills.
Financial, geographical and personal circumstances may restrict the placements and roles that individuals can realistically pursue. Relevant constraints may include unpaid or poorly paid opportunities, transport and relocation costs, limited availability of suitable housing, dependent care, disability-related requirements, immigration conditions, insecure employment and lengthy recruitment processes.
Where professional experience is organised as part of a study programme, it should be assessed for its substance and quality, not simply by whether learners participated. Diagnosis should examine the relevance and complexity of the work, the technical and professional responsibilities assigned, the supervision and feedback provided, and the connection with academic learning. It should determine whether placements enable learners to develop and demonstrate disciplinary, interdisciplinary, transversal and professional competence. Evidence should also show who gains access to meaningful projects, references and work samples that employers subsequently recognise, since unequal access to these opportunities can produce unequal credibility during recruitment.
Employer recruitment practices should be examined alongside candidate supply. Recruitment difficulties may indicate that relevant competence is unavailable, but they may also reflect difficulty interpreting evidence of advanced STEM skills, narrow selection criteria or requirements for directly comparable experience. Policymakers have an interest in distinguishing among these causes because treating all recruitment difficulties as shortages may lead to expanding education provision when the constraint lies in how employers identify, assess or recruit candidates. Better diagnosis can help direct policy towards the relevant constraint and connect investment in advanced STEM skills more effectively with their use in the economy.
Diagnosis should examine whether employers can interpret qualifications, grades, projects, work samples and previous employment in relation to the role concerned. Programme titles alone may not indicate disciplinary depth, how knowledge and methods are integrated or how learning has been applied. Access to recognised projects, professional experience and references is also unequally distributed. Improving how this evidence is interpreted can help employers widen the pool of suitable candidates, reduce reliance on familiar but imperfect proxies and improve the match between candidates’ advanced STEM skills and the work they will perform.
Diagnosis should then examine the work employees actually perform rather than relying on occupational titles or qualification requirements. Relevant evidence includes task complexity and variety, access to suitable technologies and infrastructure, autonomy, professional judgement, collaboration across fields, participation in decisions and opportunities for technical expertise to influence organisational practice. Limited use of advanced STEM skills may reflect organisational or managerial constraints rather than insufficient employee competence. Examining skills use can show employers whether recruitment investments translate into performance and organisational learning and help policymakers distinguish insufficient skills supply from limited opportunities to apply available skills effectively within organisations.
Working conditions and advancement structures should be considered integral to skills use, development and retention. Diagnosis should examine pay, employment security, working hours, mobility expectations, organisational culture, professional networks and transparent advancement criteria. It should assess whether parental leave and flexible working can be used without reducing access to promotion or substantive work, and whether discrimination, harassment or exclusion affect development and retention. Organisations may recruit workers with relevant advanced STEM skills but fail to retain them if working conditions, advancement structures or access to suitable work remain inadequate. Diagnosing these conditions can reveal avoidable losses of workers and accumulated experience that might otherwise be interpreted as insufficient external supply.
Bottleneck 5: Skills renewal and return to higher education
Technological change has been altering tasks within occupations and the competence required to perform them (Frey and Osborne, 2017[159]). For example, AI, robotics and automation have already replaced some tasks and created others making skills renewal an essential part of the development of advanced STEM skills (Gmyrek et al., 2025[160]).
In this report, skills renewal covers updating or deepening advanced STEM skills in an existing field or role, moving into an adjacent field, such as from engineering to robotics or from mathematics to data science, and preparing for a more substantial occupational change. These forms may overlap, but they require learning of different depth and duration. Updating may involve a short, specialised course, while movement into another field or occupation may require additional foundations and a longer, coherent sequence of learning rather than disconnected courses (Li, 2022[161]). Skills renewal can take place through continuing education, workplace learning, professional training and other structured learning. Higher education has a distinctive role when learners need advanced disciplinary knowledge, research-informed teaching, specialist infrastructure, credible assessment or supervised work on complex problems. Access may nevertheless remain difficult when programmes and support services are designed mainly for recent school-leavers studying full time.
A skills-renewal bottleneck arises when people cannot access or complete the learning needed to update, deepen or extend their skills, or when they cannot apply the renewed skills at work. Responsibilities are shared but differentiated. Public authorities can support accessible provision, finance, information, quality assurance and recognition, while higher education institutions and other providers are responsible for coherent learning and credible assessment. Employers influence whether workers have the time, financial support and flexibility to participate, whether renewed competence is recognised, and whether work provides opportunities to apply and continue developing it. Addressing a skills-renewal bottleneck may consequently require connected action across education and employment rather than provision of additional learning opportunities alone.
Participation in adult learning differs substantially across countries (Figure 1.7). Participation in non-formal, job-related learning ranges from less than 10% in Korea to more than 50% in Norway, compared with an OECD average of approximately 37%. Formal learning generally involves fewer than 15% of adults. Most adult learning therefore takes place through non-formal, job-related provision rather than through a return to formal education. Expanding higher education provision alone is unlikely to remove a skills-renewal bottleneck. Diagnosis should also consider employer-provided learning, professional and specialist provision, and whether adults have the time, resources and support needed to participate.
Share of adults who participated in formal or non-formal job-related learning in the 12 months prior to the survey, %
Note: Adults aged 25-65. OECD is an unweighted average of all participating member countries. See source's Box 1.1 for a precise definition of job-related adult learning.
Source: OECD (2025[162]) Trends in Adult Learning: New Data from the 2023 Survey of Adult Skills, https://doi.org/10.1787/ec0624a6-en.
Skills renewal depends on provision that meets different learning needs
Adults seeking to develop advanced STEM skills differ in their prior education, professional experience, existing competence and goals.
Four broad groups of learners in skills renewal can be distinguished:
recent STEM graduates seeking to update or deepen specialist competence
STEM professionals moving into an adjacent field
graduates and professionals seeking to move into STEM from another field
professionals without previous higher education seeking to develop advanced STEM skills
These groups need different routes for skills renewal. A course for experienced STEM professionals may be unsuitable for someone entering from another field, while requiring experienced professionals to repeat extensive foundational learning may add unnecessary time and cost. Adults without previous higher education may need recognition of vocational and workplace learning, support to develop missing foundations and a clear route towards a higher education qualification.
The appropriate provision depends on the purpose of skills renewal. Focused updating may require a short, specialised course. Deeper specialisation, movement into an adjacent field or substantial occupational change may require additional foundations and a longer, coherent sequence of learning. Higher education has a distinctive role where learners need advanced disciplinary knowledge, research-informed content, specialist infrastructure, formal assessment or a recognised qualification. Employers, professional bodies and specialist providers may be better placed to address changes in particular tools, standards and workplace practices.
Cost, time, caring responsibilities, location, scheduling and limited employer support may prevent participation. Learning may also be constrained by missing foundations, weak sequencing, limited access to specialist educators or infrastructure, insufficient feedback and few opportunities to apply learning to complex problems (OECD, 2025[163]). These constraints may reinforce one another. Adults who need additional preparation or support may also find it difficult to reduce their working hours or attend courses during the working day (Radovan, 2024[164]; Park and Su, 2026[165]). Flexible schedules and evening, hybrid, online, modular or part-time provision can help address these constraints by making learning more compatible with employment, caring responsibilities and geographical constraints. However, flexibility alone does not ensure effective learning.
Short learning supports skills renewal only when it is coherent, recognised and sustainable
Micro-credentials have huge potential in skills renewal (OECD, 2023[166]). They certify specific competence developed through short, targeted learning and are increasingly used to respond to technological and labour-market change and are generally intended to complement rather than replace larger qualifications (OECD, 2023[166]). Their main advantage is responsiveness. Micro-credentials can help learners acquire current and specialised knowledge in areas such as digital technologies, engineering and green innovation.
However, for advanced STEM skills development short learning provides only a partial response. A short course may update or extend existing advanced STEM skills but cannot compensate for missing foundations. A stand-alone micro-credential may be appropriate for focused updating, while connected components may support deeper specialisation or movement into an adjacent field. More substantial occupational change will generally require additional foundations and a longer, coherent sequence of learning. Where micro-credentials are intended to accumulate towards a larger qualification, their value depends on how they are connected with wider education and training systems (Toledo-Navarrete et al., 2026[167]).
Public policy can establish common information requirements, quality-assurance arrangements and frameworks for recognition and accumulation. It can also support co-ordination among providers, qualification authorities and employers. Education providers remain responsible for coherent learning and credible assessment, while receiving institutions and employers determine whether credentials are recognised for further study or employment. Without coherent recognition arrangements, learners may complete several useful courses without developing, or being able to demonstrate, the combination of competence required for advanced study or a new professional role (Varadarajan, Koh and Daniel, 2023[168]; Tamoliune et al., 2023[169]; OECD, 2026[170]).
Recognition is still uneven across providers, institutions and labour markets (Gauthier, 2020[171]; Kässi and Lehdonvirta, 2022[172]).OECD reviews (e.g (OECD, 2026[170]) of micro-credential systems suggest that an accumulation route towards a larger qualification should specify:
the previous learning required for entry
the level and volume of each component
the learning outcomes and assessment
how the components relate to one another
which combinations can count towards a larger qualification
who is responsible for recognition
whether recognition is automatic or individually assessed
which further educational or occupational opportunities follow
Implications for diagnosing a skills-renewal bottleneck
Diagnosing a skills-renewal bottleneck requires evidence on whether adults can access suitable learning and develop, demonstrate, apply and gain recognition for the intended advanced STEM skills. Low participation may reflect limited interest among potential learners, but may also arise from unsuitable provision, cost, time, caring responsibilities, location, scheduling or limited employer support. Participation and completion rates therefore need to be interpreted alongside evidence on learning and its subsequent recognition and application. Indicators should distinguish different forms and purposes of learning and, where possible, examine participation and outcomes across fields, learner groups and employment situations.
The following are key aspects in diagnosing a skills-renewal bottleneck.
Diagnosis should first identify the purpose for which skills renewal is needed, the intended learners, their starting points and the educational or occupational outcomes the provision is expected to support. Updating or deepening advanced STEM skills in an existing field, moving into an adjacent field and preparing for substantial occupational change require learning of different depth, duration and coherence. Differences in previous education, professional experience and existing competence also affect the provision required. A short, specialised course may meet a focused updating need, while movement into another field or occupation may require additional foundations and a longer, coherently sequenced route.
Diagnosis should also consider which provider or combination of providers is appropriate for the learning need. Higher education has a distinctive role where learners require advanced disciplinary knowledge, research-informed teaching, specialist infrastructure, credible assessment, supervised work on complex problems or a recognised qualification. Employers, professional bodies and specialist providers may be better placed to address changes in particular tools, standards and workplace practices. Expanding higher education provision alone is therefore unlikely to address all skills-renewal constraints.
Information should be assessed for whether it enables learners seeking skills renewal to understand the prior knowledge required, the purpose and level of the learning, the competence developed and assessed, and the routes available after completion. Where learners combine components from several providers, analysis should examine whether information explains how those components relate, whether they can contribute to a larger qualification, who is responsible for recognition and how education institutions and employers are likely to interpret them. It should also determine whether learners must construct their own routes from fragmented information and uncertain recognition arrangements.
Short courses and micro-credentials may meet focused updating needs, but they should not be assumed to provide the foundations or coherence required for deeper specialisation, movement into an adjacent field or substantial occupational change. Diagnosis should examine whether their scale, content and assessment correspond to the competence required for the intended destination. Where components are intended to accumulate towards a larger qualification, analysis should also consider their entry requirements, level and volume, relationships with one another, recognised combinations and subsequent opportunities. Responsibilities and procedures for recognition should be clear, and accumulation should represent a coherent progression in learning rather than simply the completion of several separate components. Diagnosis should establish whether the provision addresses an identified skills-renewal need, is recognised and sustainable, and supports meaningful educational or occupational progression.
Diagnostic assessment and recognition arrangements should be examined for whether they identify relevant competence developed through education, vocational learning and professional experience, the essential gaps that remain and what can be developed during the programme. Analysis should assess whether recognition reduces unnecessary repetition without removing the disciplinary, interdisciplinary, transversal or professional competence required for successful learning and subsequent practice. For adults without previous higher education, diagnosis should examine whether substantial professional learning is recognised and whether they can access the foundations required for an advanced STEM qualification.
Financial and practical barriers should be examined in combination. Diagnosis should consider tuition and participation costs, forgone earnings, the feasibility of reducing working hours, caring responsibilities, location, scheduling and access to specialist facilities. It should also examine whether employer support is available across occupations, organisations and forms of employment. Adults who need additional preparation or support may face longer and more costly routes while also having the least flexibility to reduce employment or attend provision during the working day.
Flexible schedules and evening, hybrid, online, modular or part-time provision can widen participation, but flexibility does not itself establish accessibility or quality. Diagnosis should assess whether these formats preserve coherent sequencing, access to specialist educators, timely feedback, credible assessment and learner support. Where advanced STEM learning requires laboratories, software, equipment, data environments, collaborative problem-solving or sustained participation in a disciplinary community, analysis should determine whether learners can access these opportunities in practice.
The opportunity to apply learning is part of skills renewal. Diagnosis should follow learners beyond participation and completion to examine whether they develop demonstrable advanced STEM skills, gain recognition for them and apply them in further education or work. Relevant evidence includes changes in tasks and responsibilities, access to suitable technologies and infrastructure, recognition by managers and colleagues, progression to more complex work and movement towards the intended educational or occupational destination. Analysis should assess employer support not only in terms of time and financial assistance for participation, but also in terms of opportunities to apply and further develop the competence acquired. Monitoring should also examine the level and coherence of the learning, the credibility of assessment and whether learners receive the teaching, support, infrastructure and practical experience needed to achieve the intended outcomes.
Advanced STEM skills combine disciplinary, interdisciplinary, transversal and professional competence and are developed, demonstrated, used and renewed across education, research and employment. The five bottlenecks examined in this chapter identify connected points at which their development and effective use may be constrained: entry into STEM higher education; persistence and completion; progression to advanced levels; career entry and effective skills use; and skills renewal across working life. These bottlenecks are neither independent nor strictly sequential. A constraint may originate before the point at which it becomes visible, affect several transitions or be transferred to a later point. Increasing entry without sufficient teaching and learner support, for example, may weaken persistence and completion, while increasing graduate supply without suitable employment, effective onboarding and substantive work may lead to the underuse of advanced STEM skills.
Broadening the talent base is part of this challenge. It requires systemic barriers in education and work to be addressed so that access to the conditions through which advanced STEM skills are developed and demonstrated does not depend unnecessarily on learners’ circumstances. A broader talent base can increase the competence on which employers, research organisations and wider society can draw and bring a wider range of perspectives to STEM education, research, innovation and professional practice.
References
[112] Aalde, O. and I. Staal Jenset (2023), “Study Program Leaders’ Perceptions of Coherence and Strategies for Creating Coherent Teacher Education Programs”, Journal of Teacher Education, Vol. 75/3, pp. 334-346, https://doi.org/10.1177/00224871231208683.
[29] Acemoglu, D. and P. Restrepo (2019), “Automation and New Tasks: How Technology Displaces and Reinstates Labor”, Journal of Economic Perspectives, Vol. 33/2, pp. 3-30, https://doi.org/10.1257/jep.33.2.3.
[114] Akçapınar, G., A. Altun and P. Aşkar (2019), “Using learning analytics to develop early-warning system for at-risk students”, International Journal of Educational Technology in Higher Education, Vol. 16/1, https://doi.org/10.1186/s41239-019-0172-z.
[79] Allen, W. (ed.) (2023), “Family socioeconomic status and college attendance: A consideration of individual-level and school-level pathways”, PLOS ONE, Vol. 18/4, p. e0284188, https://doi.org/10.1371/journal.pone.0284188.
[104] Alvares, S. (ed.) (2024), “Mentorship for Transfer Student Success in STEM Research: Mentor Approaches and Reflections”, CBE—Life Sciences Education, Vol. 23/2, https://doi.org/10.1187/cbe.23-08-0156.
[99] Amaral, L. (ed.) (2025), “A 24-year longitudinal study on a STEM gateway general chemistry course and the reduction of achievement disparities”, PLOS ONE, Vol. 20/2, p. e0318882, https://doi.org/10.1371/journal.pone.0318882.
[80] Andersen, S. and M. Hjortskov (2019), “The unnoticed influence of peers on educational preferences”, Behavioural Public Policy, Vol. 6/4, pp. 530-553, https://doi.org/10.1017/bpp.2019.14.
[135] Anders, J. (ed.) (2025), “Integrative review in PhD admissions: A case study of efficiently minimizing bias while maximizing the student narrative”, PLOS One, Vol. 20/6, p. e0323426, https://doi.org/10.1371/journal.pone.0323426.
[71] Archer, DeWitt and Osborne (2015), “Is Science for Us? Black Students’ and Parents’ Views of Science and Science Careers”, Science Education, Vol. 99/2, pp. 199-237, https://doi.org/10.1002/sce.21146.
[3] Archer, L. et al. (2015), ““Science capital”: A conceptual, methodological, and empirical argument for extending bourdieusian notions of capital beyond the arts”, Journal of Research in Science Teaching, Vol. 52/7, pp. 922-948, https://doi.org/10.1002/tea.21227.
[68] Archer, L. et al. (2012), “Science Aspirations, Capital, and Family Habitus”, American Educational Research Journal, Vol. 49/5, pp. 881-908, https://doi.org/10.3102/0002831211433290.
[69] Archer, L. et al. (2010), ““Doing” science versus “being” a scientist: Examining 10/11‐year‐old schoolchildren’s constructions of science through the lens of identity”, Science Education, Vol. 94/4, pp. 617-639, https://doi.org/10.1002/sce.20399.
[72] Archer, L. et al. (2016), “Stratifying science: a Bourdieusian analysis of student views and experiences of school selective practices in relation to ‘Triple Science’ at KS4 in England”, Research Papers in Education, Vol. 32/3, pp. 296-315, https://doi.org/10.1080/02671522.2016.1219382.
[74] Archer, L. et al. (2016), “The “Exceptional” Physics Girl”, American Educational Research Journal, Vol. 54/1, pp. 88-126, https://doi.org/10.3102/0002831216678379.
[28] Arrow, K. (1973), “Higher education as a filter”, Journal of Public Economics, Vol. 2/3, pp. 193-216, https://doi.org/10.1016/0047-2727(73)90013-3.
[11] Ashby, I. and M. Exter (2018), “Designing for Interdisciplinarity in Higher Education: Considerations for Instructional Designers”, TechTrends, Vol. 63/2, pp. 202-208, https://doi.org/10.1007/s11528-018-0352-z.
[140] Bach, L., E. Armstrong and M. Jones (2025), “Fostering inclusive and effective doctoral supervisory practice: the Let’s Talk About card game”, Journal of Learning Development in Higher Education 36, https://doi.org/10.47408/jldhe.vi36.1362.
[139] Bahnson, M. and T. Abane (2026), “Beyond attrition: advisor relationships, progress clarity, and laboratory switching in engineering doctoral programs”, Frontiers in Education, Vol. 11, https://doi.org/10.3389/feduc.2026.1770054.
[125] Bati, A. (ed.) (2022), “The impact of holistic review on correlations between doctoral student outcomes, and GPA and GRE scores in the biomedical sciences”, PLOS ONE, Vol. 17/12, p. e0279258, https://doi.org/10.1371/journal.pone.0279258.
[50] Bauer, G. et al. (2021), “Intersectionality in quantitative research: A systematic review of its emergence and applications of theory and methods”, SSM - Population Health, Vol. 14, p. 100798, https://doi.org/10.1016/j.ssmph.2021.100798.
[152] Bauer, T. et al. (2024), “New Horizons for Newcomer Organizational Socialization: A Review, Meta-Analysis, and Future Research Directions”, Journal of Management, Vol. 51/1, pp. 344-382, https://doi.org/10.1177/01492063241277168.
[45] Betancur, L., E. Votruba-Drzal and C. Schunn (2018), “Socioeconomic gaps in science achievement”, International Journal of STEM education, Vol. 5, https://doi.org/10.1186/s40594-018-0132-5.
[59] Bokhove, C. et al. (2019), “The Role of “Opportunity to Learn” in the Geometry Curriculum: A Multilevel Comparison of Six Countries”, Frontiers in Education, Vol. 4, https://doi.org/10.3389/feduc.2019.00063.
[47] Borrego, M. et al. (2025), “A scoping review on U.S. undergraduate students with disabilities in STEM courses and STEM majors”, International Journal of STEM Education, Vol. 12/1, https://doi.org/10.1186/s40594-024-00522-2.
[96] Brickman, P. (ed.) (2018), “Contemporary Issues in Group Learning in Undergraduate Science Classrooms: A Perspective from Student Engagement”, CBE—Life Sciences Education, Vol. 17/2, p. es3, https://doi.org/10.1187/cbe.17-11-0239.
[25] Brunello, G. and P. Wruuck (2021), “Skill shortages and skill mismatch: A review of the literature”, Journal of Economic Surveys, Vol. 35/4, pp. 1145-1167, https://doi.org/10.1111/joes.12424.
[20] Byeongseon, K. (2026), Ministry of Science and ICT Promotes “Brain to Korea” to Attract Outstanding Overseas Science and Technology Talent, https://view.asiae.co.kr/en/article/2026021211565015632 (accessed on 20 February 2026).
[70] Carlone, H. and A. Johnson (2007), “Understanding the science experiences of successful women of color: Science identity as an analytic lens”, Journal of Research in Science Teaching, Vol. 44/8, pp. 1187-1218, https://doi.org/10.1002/tea.20237.
[121] Cavallaro, M. (2023), “Universities of applied sciences’ EU research project participation through the lens of differentiation”, Science and Public Policy, Vol. 51/1, pp. 1-14, https://doi.org/10.1093/scipol/scad048.
[157] Cech, E. and M. Blair-Loy (2014), “Consequences of Flexibility Stigma Among Academic Scientists and Engineers”, Work and Occupations, Vol. 41/1, pp. 86-110, https://doi.org/10.1177/0730888413515497.
[116] Cerratto Pargman, T. and C. McGrath (2021), “Mapping the Ethics of Learning Analytics in Higher Education: A Systematic Literature Review of Empirical Research”, Journal of Learning Analytics, Vol. 8/2, pp. 123-139, https://doi.org/10.18608/jla.2021.1.
[107] Choi, Y. (2017), “Student Employment and Persistence: Evidence of Effect Heterogeneity of Student Employment on College Dropout”, Research in Higher Education, Vol. 59/1, pp. 88-107, https://doi.org/10.1007/s11162-017-9458-y.
[151] Cingillioglu, I. and M. Schoettner (2025), “Adapting to AI-mediated workplaces: how STEM trainees navigate new challenges and opportunities”, Information Technology & People, Vol. 38/8, pp. 251-275, https://doi.org/10.1108/itp-02-2025-0163.
[155] Cohen, W. and D. Levinthal (1990), “Absorptive Capacity: A New Perspective on Learning and Innovation”, Administrative Science Quarterly, Vol. 35/1, https://doi.org/10.2307/2393553.
[49] Crenshaw, K. (1989), “Demarginalizing the intersection of race and sex: A Black feminist critique of antidiscrimination doctrine, feminist theory and antiracist politics”, University of Chicago Legal Forum, Vol. Vol. 1989/1/Article 8, pp. 139-167, http://chicagounbound.uchicago.edu/uclf/vol1989/iss1/8 (accessed on 3 June 2026).
[62] Cribbs, J., X. Huang and K. Piatek‐Jimenez (2021), “Relations of mathematics mindset, mathematics anxiety, mathematics identity, and mathematics self‐efficacy to STEM career choice: A structural equation modeling approach”, School Science and Mathematics, Vol. 121/5, pp. 275-287, https://doi.org/10.1111/ssm.12470.
[132] Crumley-Effinger, M. (2023), “ISM Policy Pervasion: Visas, study permits, and the international student experience”, Journal of International Students, Vol. 14/1, https://doi.org/10.32674/jis.v14i1.5347.
[64] Cuder, A. et al. (2024), “The impact of math anxiety and self‐efficacy in middle school STEM choices: A 3‐year longitudinal study”, British Journal of Educational Psychology, Vol. 94/4, pp. 1091-1108, https://doi.org/10.1111/bjep.12707.
[57] Dauphin, L. (2024), “Les choix d’enseignements de spécialité et d’enseignements optionnels à la rentrée 2023” [Choices of specialisation and optional subjects at the start of the 2023 school year], Direction de l’évaluation, de la prospective et de la performance, Ministry of National Education and Youth, Paris, https://www.education.gouv.fr/depp/les-choix-d-enseignements-de-specialite-et-d-enseignements-optionnels-la-rentree-2023-452430 (accessed on 5 August 2026).
[84] De Gioannis, E., G. Pasin and F. Squazzoni (2023), “Empowering women in STEM: a scoping review of interventions with role models”, International Journal of Science Education, Part B, Vol. 13/3, pp. 261-275, https://doi.org/10.1080/21548455.2022.2162832.
[78] DeWitt, J. and L. Archer (2015), “Who Aspires to a Science Career? A comparison of survey responses from primary and secondary school students”, International Journal of Science Education, Vol. 37/13, pp. 2170-2192, https://doi.org/10.1080/09500693.2015.1071899.
[73] DeWitt, J., L. Archer and J. Moote (2018), “15/16-Year-Old Students’ Reasons for Choosing and Not Choosing Physics at a Level”, International Journal of Science and Mathematics Education, Vol. 17/6, pp. 1071-1087, https://doi.org/10.1007/s10763-018-9900-4.
[42] Díaz, E. (2026), STEM Atlas: Measuring, Monitoring and Strengthening the STEM Talent Pipeline, presentation to the fourth meeting of the OECD Mutual Learning Group on Advanced STEM Skills in Higher Education.
[93] Duncan, V., E. Holt and S. Keenan (2023), “Creating an equitable and inclusive STEM classroom: a qualitative meta-synthesis of approaches and practices in higher education”, Frontiers in Education, Vol. 8, https://doi.org/10.3389/feduc.2023.1154652.
[82] Du, X. and B. Wong (2019), “Science career aspiration and science capital in China and UK: a comparative study using PISA data”, International Journal of Science Education, Vol. 41/15, pp. 2136-2155, https://doi.org/10.1080/09500693.2019.1662135.
[60] Ekmekci, A. and D. Serrano (2022), “The Impact of Teacher Quality on Student Motivation, Achievement, and Persistence in Science and Mathematics”, Education Sciences, Vol. 12/10, p. 649, https://doi.org/10.3390/educsci12100649.
[108] El Galad, A., D. Betts and N. Campbell (2024), “Flexible learning dimensions in higher education: aligning students’ and educators’ perspectives for more inclusive practices”, Frontiers in Education, Vol. 9, https://doi.org/10.3389/feduc.2024.1347432.
[21] Engineers Australia (2025), Nation’s future workforce under strain as key engineering shortages persist, https://www.engineersaustralia.org.au/news-and-media/2025/11/nations-future-workforce-under-strain-key-engineering-shortages-persist (accessed on 20 February 2026).
[30] EPO/OECD (2025), Mapping the global quantum ecosystem, EPO / OECD, https://doi.org/10.65216/20251217-0001.
[15] EURES (2024), Report on labour shortages and surpluses 2024, Publications Office of the European Union, Luxembourg, https://www.ela.europa.eu/sites/default/files/2025-06/EURES_Report_on_labour_shortages_and_surpluses_2024.pdf.
[88] European Commission (2026), Policy Experimentation: Erasmus+ Scholarships, Call for Proposals ERASMUS-EDU-2026-POL-EXP-SSEF, European Education and Culture Executive Agency, Brussels, https://ec.europa.eu/info/funding-tenders/opportunities/docs/2021-2027/erasmus/wp-call/2026/call-fiche_erasmus-edu-2026-pol-exp-ssef_en.pdf (accessed on 14 September 2026).
[39] European Commission (2025), She Figures 2024 – Gender in research and innovation – Statistics and indicators, https://op.europa.eu/en/publication-detail/-/publication/7646222f-e82b-11ef-b5e9-01aa75ed71a1/language-en (accessed on 1 August 2026).
[52] European Commission (2025), She Figures 2024: Gender in Research and Innovation, Statistics and Indicators, Directorate-General for Research and Innovation, Publications Office of the European Union, https://doi.org/10.2777/592260 (accessed on 3 June 2026).
[51] European Commission (2025), “She Figures 2024: Policy Report”, https://doi.org/10.2777/934401.
[23] European Commission (2025), STEM Education Strategic Plan: Skills for Competitiveness and Innovation, Communication from the Commission to the European Parliament, the Council, the European Economic and Social Committee and the Committee of the Regions, COM(2025) 89 final, European Commission, Brussels, https://education.ec.europa.eu/sites/default/files/2025-03/STEM_Education_Strategic_Plan_COM_2025_89_1_EN_0.pdf (accessed on 3 June 2026).
[48] European Commission (2024), Refugee and migrant integration into education and training, https://education.ec.europa.eu/focus-topics/improving-quality/inclusive-education/migrants-and-refugees (accessed on 12 August 2026).
[16] European Commission (2023), Employment and social developments in Europe 2023, Publications Office of the European Union, Luxembourg, https://doi.org/10.2767/089698 (accessed on 17 November 2025).
[38] Eurostat (2026), Eurostat, https://ec.europa.eu/eurostat/databrowser/view/educ_uoe_grad02/default/table (accessed on 2026 August 01).
[94] Fankhauser, S. (ed.) (2023), “Student perceptions of inclusive pedagogy in undergraduate STEM classrooms”, Journal of Microbiology & Biology Education, Vol. 24/3, https://doi.org/10.1128/jmbe.00097-23.
[63] Ferdinand, R., M. Malanchini and K. Rimfeld (2024), “Mathematics interest, self-efficacy, and anxiety predict STEM career choice in emerging adulthood”, npj Science of Learning, Vol. 9/1, https://doi.org/10.1038/s41539-024-00275-1.
[1] Feynman, R. (1955), “The value of science”, Engineering and Science, Vol. 19/3, pp. 13-15, https://calteches.library.caltech.edu/1575/1/Science.pdf.
[159] Frey, C. and M. Osborne (2017), “The future of employment: How susceptible are jobs to computerisation?”, Technological Forecasting and Social Change, Vol. 114, pp. 254-280, https://doi.org/10.1016/j.techfore.2016.08.019.
[111] Frost, D. and R. Ackrill (2025), “The multiple dimensions of curriculum mapping: designing a comprehensive outcomes-based framework”, London Review of Education, Vol. 23/1, https://doi.org/10.14324/lre.23.1.17.
[137] Gardner, S. and B. Gopaul (2012), “The Part-Time Doctoral Student Experience”, International Journal of Doctoral Studies, Vol. 7, pp. 063-078, https://doi.org/10.28945/1561.
[171] Gauthier, T. (2020), “The value of microcredentials: The employer’s perspective”, The Journal of Competency-Based Education, Vol. 5/2, https://doi.org/10.1002/cbe2.1209.
[83] Gladstone, J. and A. Cimpian (2021), “Which role models are effective for which students? A systematic review and four recommendations for maximizing the effectiveness of role models in STEM”, International Journal of STEM Education, Vol. 8/1, https://doi.org/10.1186/s40594-021-00315-x.
[4] Global STEM Alliance (2025), STEM Education Framework Research Foundations, The New York Academy of Sciences, https://www.nyas.org/wp-content/uploads/2025/05/STEM-Education-Framework-Research-Base.pdf (accessed on 20 February 2026).
[41] Global STEM Women (2026), SWAR: STEM Women Annual Report, https://www.globalstemwomen.org/swar (accessed on 3 June 2026).
[160] Gmyrek et al. (2025), Generative AI and jobs, ILO, Geneva, https://doi.org/10.54394/hetp0387.
[2] Goos, M., S. Carreira and I. Namukasa (2023), “Mathematics and interdisciplinary STEM education: recent developments and future directions”, ZDM – Mathematics Education, Vol. 55/7, pp. 1199-1217, https://doi.org/10.1007/s11858-023-01533-z.
[119] Hanushek, E. et al. (2016), “General Education, Vocational Education, and Labor-Market Outcomes over the Lifecycle”, Journal of Human Resources, Vol. 52/1, pp. 48-87, https://doi.org/10.3368/jhr.52.1.0415-7074r.
[35] Hofstra, B. et al. (2020), “The Diversity–Innovation Paradox in Science”, Proceedings of the National Academy of Sciences, Vol. 117/17, pp. 9284-9291, https://doi.org/10.1073/pnas.1915378117.
[113] Hofstra, B. et al. (2022), “Diversifying the Professoriate”, Socius: Sociological Research for a Dynamic World, Vol. 8, https://doi.org/10.1177/23780231221085118.
[53] Karataş, M., A. Löther and L. Weber (2025), “Gender Equality Plans and Inclusiveness in the European Research Area”, Social Inclusion, Vol. 13, https://doi.org/10.17645/si.10092.
[172] Kässi, O. and V. Lehdonvirta (2022), “Do Microcredentials Help New Workers Enter the Market?”, Journal of Human Resources, Vol. 59/4, pp. 1284-1318, https://doi.org/10.3368/jhr.0519-10226r3.
[13] Kelley, T. and J. Knowles (2016), “A conceptual framework for integrated STEM education”, International Journal of STEM Education, Vol. 3/1, https://doi.org/10.1186/s40594-016-0046-z.
[106] Kocsis, Z., E. Alter and G. Pusztai (2022), “The Role of Student Employment in Persistence and Efficiency in STEM Higher Education”, International Journal of Education in Mathematics, Science and Technology, Vol. 10/4, pp. 831-848, https://doi.org/10.46328/ijemst.2177.
[36] Kozlowski, D. et al. (2022), “Intersectional inequalities in science”, Proceedings of the National Academy of Sciences, Vol. 119/2, https://doi.org/10.1073/pnas.2113067119.
[158] Lassen, A. and R. Ivandić (2024), “Parenthood and Academic Career Trajectories”, AEA Papers and Proceedings, Vol. 114, pp. 238-242, https://doi.org/10.1257/pandp.20241118.
[65] Lau, N., D. Ansari and H. Sokolowski (2024), “Unraveling the interplay between math anxiety and math achievement”, Trends in Cognitive Sciences, Vol. 28/10, pp. 937-947, https://doi.org/10.1016/j.tics.2024.07.006.
[77] Lent, R., S. Brown and G. Hackett (1994), “Toward a Unifying Social Cognitive Theory of Career and Academic Interest, Choice, and Performance”, Journal of Vocational Behavior, Vol. 45/1, pp. 79-122, https://doi.org/10.1006/jvbe.1994.1027.
[122] Li, J. and M. Pilz (2021), “International transfer of vocational education and training: a literature review”, Journal of Vocational Education & Training, Vol. 75/2, pp. 185-218, https://doi.org/10.1080/13636820.2020.1847566.
[161] Li, L. (2022), “Reskilling and Upskilling the Future-ready Workforce for Industry 4.0 and Beyond”, Information Systems Frontiers, Vol. 26/5, pp. 1697-1712, https://doi.org/10.1007/s10796-022-10308-y.
[85] Liu, Z. and N. Kringos (2025), “What prevents STEM universities from widening participation? A systematic literature review on global experiences”, International Journal of STEM Education, Vol. 12/1, https://doi.org/10.1186/s40594-025-00580-0.
[8] Long, D. and B. Magerko (2020), “What is AI Literacy? Competencies and Design Considerations”, Proceedings of the 2020 CHI Conference on Human Factors in Computing Systems, pp. 1-16, https://doi.org/10.1145/3313831.3376727.
[142] Luk, L. and C. Chan (2021), “Students’ learning outcomes from engineering internship: a provisional framework”, Studies in Continuing Education, Vol. 44/3, pp. 526-545, https://doi.org/10.1080/0158037x.2021.1917536.
[12] Makonye, J. and N. Moodley (2023), “Connecting mathematics to STEM education: interdisciplinary teaching and learning facilitation”, ZDM – Mathematics Education, Vol. 55/7, pp. 1365-1373, https://doi.org/10.1007/s11858-023-01522-2.
[134] Mazzeo Ortolani, G. (2025), STEM and STEAM education, and disciplinary integration: A guide to informed policy action, JRC Research Reports, No. JRC141438, European Commission, Brussels, https://publications.jrc.ec.europa.eu/repository/handle/JRC141438 (accessed on 5 August 2026).
[101] McKean, T. et al. (2024), “A multi-tiered mentoring community approach to research experiences for local students from disadvantaged and underrepresented minority backgrounds”, Frontiers in Education, Vol. 9, https://doi.org/10.3389/feduc.2024.1366063.
[136] Menezes, R. (ed.) (2021), “Professional development and career-preparedness experiences of STEM Ph.D. students: Gaps and avenues for improvement”, PLOS ONE, Vol. 16/12, p. e0260328, https://doi.org/10.1371/journal.pone.0260328.
[58] Ministry of National Education and Youth (2023), “Place des mathématiques de l’école au lycée: Une nouvelle dynamique pour les mathématiques” [The place of mathematics from primary school to upper-secondary education: A new dynamic for mathematics], Bulletin officiel de l’Éducation nationale, de la Jeunesse et des Sports, No. 2, 12 January 2023, https://www.education.gouv.fr/bo/23/Hebdo2/MENE2300946N.htm (accessed on 1 August 2026).
[150] Mitschelen, A. and S. Kauffeld (2025), “Workplace learning during organizational onboarding: integrating formal, informal, and self-regulated workplace learning”, Frontiers in Organizational Psychology, Vol. 3, https://doi.org/10.3389/forgp.2025.1569098.
[97] Møgelvang, A. and J. Nyléhn (2022), “Co-operative Learning in Undergraduate Mathematics and Science Education: A Scoping Review”, International Journal of Science and Mathematics Education, Vol. 21/6, pp. 1935-1959, https://doi.org/10.1007/s10763-022-10331-0.
[56] Morando, G. (2024), “*Mathematics Specialization at High School and Undergraduate Degree Choice: Evidence From England”, Educational Evaluation and Policy Analysis, Vol. 47/4, pp. 1010-1044, https://doi.org/10.3102/01623737241255348.
[33] More, S. and T. Rosenbloom (2020), “Job-Field Underemployment and Career Satisfaction: A Relationship of Cause and Effect”, International Journal of Business and Management, Vol. 15/10, p. 82, https://doi.org/10.5539/ijbm.v15n10p82.
[46] Morgan, P. et al. (2016), “Science Achievement Gaps Begin Very Early, Persist, and Are Largely Explained by Modifiable Factors”, https://doi.org/10.3102/0013189X1663318.
[100] Morris, R., T. Perry and L. Wardle (2021), “Formative assessment and feedback for learning in higher education: A systematic review”, Review of Education, Vol. 9/3, https://doi.org/10.1002/rev3.3292.
[143] Moss-Pech, C. (2021), “The Career Conveyor Belt: How Internships Lead to Unequal Labor Market Outcomes among College Graduates”, Qualitative Sociology, Vol. 44/1, pp. 77-102, https://doi.org/10.1007/s11133-020-09471-y.
[54] OECD (2026), Distribution of enrolled students, new entrants and graduates by field of education, (database), OECD Data Explorer, https://data-explorer.oecd.org/s/5kw (accessed on 7 August 2026).
[40] OECD (2026), Number of enrolled students, graduates and new entrants by field of education, (database), OECD Data Explorer, https://data-explorer.oecd.org/s/5ky (accessed on 11 August 2026).
[154] OECD (2026), OECD Employment Outlook 2026: Geographic Disparities in Jobs and Incomes, OECD Publishing, Paris, https://doi.org/10.1787/7e710f54-en.
[5] OECD (2026), PISA 2025 Assessment and Analytical Framework, PISA, OECD Publishing, Paris, https://doi.org/10.1787/86c36975-en.
[22] OECD (2026), Research and Innovation Careers Observatory 2026: Key Indicators on Talent, Mobility and Working Conditions, OECD Publishing, Paris, https://doi.org/10.1787/138ed294-en.
[170] OECD (2026), Towards a Harmonised Micro-Credential Ecosystem in Czechia, Higher Education, OECD Publishing, Paris, https://doi.org/10.1787/0be96df9-en.
[91] OECD (2025), Education at a Glance 2025: OECD Indicators, OECD Publishing, Paris, https://doi.org/10.1787/1c0d9c79-en.
[55] OECD (2025), Education GPS: Diagrams of Education Systems, Organisation for Economic Co-operation and Development, https://gpseducation.oecd.org/.
[37] OECD (2025), OECD Science, Technology and Innovation Outlook 2025: Driving Change in a Shifting Landscape, OECD Publishing, Paris, https://doi.org/10.1787/5fe57b90-en.
[89] OECD (2025), The Financial Sustainability of Higher Education: Insights from Policy in OECD Countries, Higher Education, OECD Publishing, Paris, https://doi.org/10.1787/f544ccfe-en.
[75] OECD (2025), The State of Global Teenage Career Preparation, OECD Publishing, Paris, https://doi.org/10.1787/d5f8e3f2-en.
[162] OECD (2025), Trends in Adult Learning: New Data from the 2023 Survey of Adult Skills, Getting Skills Right, OECD Publishing, Paris, https://doi.org/10.1787/ec0624a6-en.
[163] OECD (2025), What’s missing in adult learning - and how do we fix it?, OECD Publishing, Paris, https://doi.org/10.1787/58b9acfd-en.
[141] OECD (2023), “Promoting diverse career pathways for doctoral and postdoctoral researchers”, OECD Science, Technology and Industry Policy Papers, No. 158, OECD Publishing, Paris, https://doi.org/10.1787/dc21227a-en.
[166] OECD (2023), “Public policies for effective micro-credential learning”, OECD Education Policy Perspectives, No. 85, OECD Publishing, Paris, https://doi.org/10.1787/a41f148b-en.
[17] OECD (2022), Skills for Jobs Database 2022: Key Insights, https://www.oecdskillsforjobsdatabase.org/data/S4J2022_results.pdf.
[153] OECD (2021), Training in Enterprises: New Evidence from 100 Case Studies, Getting Skills Right, OECD Publishing, Paris, https://doi.org/10.1787/7d63d210-en.
[6] OECD/European Commission (2026), Empowering Learners for the Age of AI: An AI Literacy Framework for Primary and Secondary Education, OECD Publishing, Paris, https://doi.org/10.1787/65cd27d4-en.
[123] Okahana, H. and Y. Hao (2019), “Are They Worth it?: Master’s Degrees and Labor Market Outcomes in the STEM Workforce”, Innovative Higher Education, Vol. 44/3, pp. 165-185, https://doi.org/10.1007/s10755-019-9455-5.
[61] Oliveira, H. and J. Bonito (2023), “Practical work in science education: a systematic literature review”, Frontiers in Education, Vol. 8, https://doi.org/10.3389/feduc.2023.1151641.
[165] Park, J. and M. Su (2026), “Adult Learners’ Participation in Higher Education: A Systematic Review of Determinants and Barriers (2000–2025)”, Trends in Higher Education, Vol. 5/1, p. 19, https://doi.org/10.3390/higheredu5010019.
[131] Pierszalowski, S., J. Bouwma-Gearhart and L. Marlow (2021), “A Systematic Review of Barriers to Accessing Undergraduate Research for STEM Students: Problematizing Under-Researched Factors for Students of Color”, Social Sciences, Vol. 10/9, p. 328, https://doi.org/10.3390/socsci10090328.
[26] Polanco-Jimenez, J. and K. Witte (2025), ENESET - Factors influencing STEM participation and effective intervention strategies, https://doi.org/10.2766/6852280.
[9] Portillo-Blanco, A. et al. (2024), “A Systematic Literature Review of Integrated STEM Education: Uncovering Consensus and Diversity in Principles and Characteristics”, Education Sciences, Vol. 14/9, p. 1028, https://doi.org/10.3390/educsci14091028.
[127] Posselt, J. et al. (2023), “Redefining Merit Through New Routines: Holistic Admissions Policy Implementation in Graduate Education”, Educational Evaluation and Policy Analysis, Vol. 47/1, pp. 159-184, https://doi.org/10.3102/01623737231201612.
[164] Radovan, M. (2024), “Workplace Flexibility and Participation in Adult Learning”, Sustainability, Vol. 16/14, p. 5950, https://doi.org/10.3390/su16145950.
[10] Rana, K., S. Aitken and R. Chimoriya (2025), “Interdisciplinary Approaches in Doctoral and Higher Research Education: An Integrative Scoping Review”, Education Sciences, Vol. 15/1, p. 72, https://doi.org/10.3390/educsci15010072.
[102] Ravishankar, S. et al. (2024), “Empowering STEM students: A university‐wide mentorship program fostering retention and belonging”, Journal of Cellular Physiology, Vol. 239/7, https://doi.org/10.1002/jcp.31348.
[103] Reeves, A. et al. (2022), “A Pilot Graduate Student-Led Near-Peer Mentorship Program for Transfer Students Provides a Supportive Network at an R1 Institution”, Journal of Chemical Education, Vol. 100/1, pp. 134-142, https://doi.org/10.1021/acs.jchemed.2c00427.
[115] Rets, I., C. Herodotou and A. Gillespie (2023), “Six Practical Recommendations Enabling Ethical Use of Predictive Learning Analytics in Distance Education”, Journal of Learning Analytics, Vol. 10/1, pp. 149-167, https://doi.org/10.18608/jla.2023.7743.
[76] Rinn, A. and T. Kettler (2020), “Assessing and Developing Diverse STEM Talent: Introduction to the Special Issue”, Journal of Advanced Academics, Vol. 31/3, pp. 159-160, https://doi.org/10.1177/1932202x20928747.
[110] Roehrig, G. et al. (2021), “Understanding coherence and integration in integrated STEM curriculum”, International Journal of STEM Education, Vol. 8/1, https://doi.org/10.1186/s40594-020-00259-8.
[87] Romero, A., K. Cobian and P. Martín (2025), “Understanding the complexities of financial support for students in grant-funded STEMM training programs”, Frontiers in Education, Vol. 10, https://doi.org/10.3389/feduc.2025.1445151.
[81] Rubineau, B. et al. (2023), “Pathways of Peer Influence on Major Choice”, Social Forces, Vol. 102/3, pp. 1089-1110, https://doi.org/10.1093/sf/soad129.
[98] Schauer, A. et al. (2025), “Internal and external influences on role stereotype adherence and gender dynamics on engineering design teams”, International Journal of STEM Education, Vol. 12/1, https://doi.org/10.1186/s40594-025-00528-4.
[145] Schneider, J. et al. (2024), “Creating better internships by understanding mentor challenges: findings from a series of focus groups”, International Journal of STEM Education, Vol. 11/1, https://doi.org/10.1186/s40594-024-00518-y.
[118] Seymour, E. et al. (2000), “Creating a Better Mousetrap: On-line Student Assessment of their Learning Gains”, paper presented at the National Meeting of the American Chemical Society, San Francisco, https://www.colorado.edu/eer/sites/default/files/attached-files/seymoursalgbettermousetrap2000.pdf (accessed on 18 September 2026).
[90] Seymour, E. and A. Hunter (eds.) (2019), Talking about Leaving Revisited, Springer International Publishing, Cham, https://doi.org/10.1007/978-3-030-25304-2.
[128] Seymour, E. et al. (2004), “Establishing the benefits of research experiences for undergraduates in the sciences: First findings from a three‐year study”, Science Education, Vol. 88/4, pp. 493-534, https://doi.org/10.1002/sce.10131.
[138] Sherman, D. et al. (2021), “The Changing Landscape of Doctoral Education in Science, Technology, Engineering, and Mathematics: PhD Students, Faculty Advisors, and Preferences for Varied Career Options”, Frontiers in Psychology, Vol. 12, https://doi.org/10.3389/fpsyg.2021.711615.
[130] Shortlidge, E. (ed.) (2024), “Participation in Undergraduate Research Reduces Equity Gaps in STEM Graduation Rates”, CBE—Life Sciences Education, Vol. 23/1, https://doi.org/10.1187/cbe.22-03-0061.
[149] Skakni, I. et al. (2025), “PhD graduates pursuing careers beyond academia: a scoping review”, Higher Education Research & Development, Vol. 45/1, pp. 235-254, https://doi.org/10.1080/07294360.2025.2515211.
[124] Skjelbred, S. and K. Nesje (2022), “Differences in transition to the labour market across STEM graduates”, Research in Comparative and International Education, Vol. 18/1, pp. 79-103, https://doi.org/10.1177/17454999221134399.
[120] Soliz, A. (2023), “Career and Technical Education at Community Colleges: A Review of the Literature”, AERA Open, Vol. 9, https://doi.org/10.1177/23328584231186618.
[156] Solomon, S. and M. Du Plessis (2023), “Experiences of precarious work within higher education institutions: a qualitative evidence synthesis”, Frontiers in Education, Vol. 8, https://doi.org/10.3389/feduc.2023.960649.
[31] Sostero, M. and E. Fernández-Macías (2021), he Professional Lens: What Online Job Advertisements Can Say About Occupational Task Profiles, European Commission, http://hdl.handle.net/10419/245944.
[27] Spence, M. (1973), “Job Market Signaling”, The Quarterly Journal of Economics, Vol. 87/3, p. 355, https://doi.org/10.2307/1882010.
[146] Stans, R. et al. (2025), “The Impact of Higher Education on Employer Perceptions”, The Economic Journal, Vol. 136/674, pp. 602-625, https://doi.org/10.1093/ej/ueaf061.
[92] Strayhorn, T. (2018), College Students’ Sense of Belonging, Routledge, https://doi.org/10.4324/9781315297293.
[67] Street, K., L. Malmberg and S. Schukajlow (2024), “Students’ mathematics self-efficacy: a scoping review”, ZDM – Mathematics Education, Vol. 56/2, pp. 265-280, https://doi.org/10.1007/s11858-024-01548-0.
[169] Tamoliune, G. et al. (2023), “Exploring the potential of micro-credentials: A systematic literature review”, Frontiers in Education, Vol. 7, https://doi.org/10.3389/feduc.2022.1006811.
[95] Tharayil, S. et al. (2018), “Strategies to mitigate student resistance to active learning”, International Journal of STEM Education, Vol. 5/1, https://doi.org/10.1186/s40594-018-0102-y.
[19] The Linux Foundation (2025), State of Tech Talent Japan Report: Trends in Technical Hiring, AI Disruption, and the Skills Gap, https://www.linuxfoundation.org/hubfs/Research%20Reports/lfr_techtalent_jp2025_052325.pdf?hsLang=en (accessed on 19 February 2026).
[167] Toledo-Navarrete, A. et al. (2026), “A scoping review of national stackable micro-credentials ecosystems in higher education with implications for Ecuador”, Frontiers in Education, Vol. 11, https://doi.org/10.3389/feduc.2026.1793357.
[147] Tomlinson, M. and V. Anderson (2020), “Employers and graduates: the mediating role of signals and capitals”, Journal of Higher Education Policy and Management, Vol. 43/4, pp. 384-399, https://doi.org/10.1080/1360080x.2020.1833126.
[133] Topliceanu, A., M. Blanchard and K. Collier (2026), “Characteristics of International Graduate STEM Students in the United States and the Supports and Barriers They Experience: A Systematic Literature Review”, Trends in Higher Education, Vol. 5/2, p. 42, https://doi.org/10.3390/higheredu5020042.
[105] Torres-García, C., L. Zamudio-Orozco and M. Meráz García (2026), “A Qualitative Examination of Mentoring Practices for First-Generation, Low-Income STEM Students Through a Critical Intersectionality Lens”, Sage Open, Vol. 16/2, https://doi.org/10.1177/21582440261431861.
[14] Tytler, R., J. Anderson and G. Williams (2023), “Exploring a framework for integrated STEM: challenges and benefits for promoting engagement in learning mathematics”, ZDM – Mathematics Education, Vol. 55/7, pp. 1299-1313, https://doi.org/10.1007/s11858-023-01519-x.
[117] Tzimas, D. and S. Demetriadis (2021), “Ethical issues in learning analytics: a review of the field”, Educational Technology Research and Development, Vol. 69/2, pp. 1101-1133, https://doi.org/10.1007/s11423-021-09977-4.
[18] UK Government (2025), UK Quantum Skills Taskforce report, https://www.gov.uk/government/publications/uk-quantum-skills-taskforce-report/uk-quantum-skills-taskforce-report#skills-needs-of-the-sector (accessed on 20 February 2026).
[44] UNESCO (2026), UNESCO - Campus Africa, https://www.unesco.org/en/higher-education/campus-africa (accessed on 1 August 2026).
[7] UNESCO (2024), AI competency framework for students, UNESCO, https://doi.org/10.54675/jkjb9835.
[43] UNESCO (2018), Telling SAGA: Improving Measurement and Policies for Gender Equality in Science, Technology and Innovation, SAGA Working Paper 5, United Nations Educational, Scientific and Cultural Organization (UNESCO), https://doi.org/10.51584/IJRIAS.2026.11030005.
[148] van Rijnsoever, F. (ed.) (2017), “An evidence-based evaluation of transferrable skills and job satisfaction for science PhDs”, PLOS ONE, Vol. 12/9, p. e0185023, https://doi.org/10.1371/journal.pone.0185023.
[168] Varadarajan, S., J. Koh and B. Daniel (2023), “A systematic review of the opportunities and challenges of micro-credentials for multiple stakeholders: learners, employers, higher education institutions and government”, International Journal of Educational Technology in Higher Education, Vol. 20/1, https://doi.org/10.1186/s41239-023-00381-x.
[144] Vlachopoulos, D. and O. Pachni Tsitiridou (2026), “Preparing University Graduates for the Labour Market Through Employability Skills Development and University–Industry Collaboration: A Systematic Review”, Education Sciences, Vol. 16/3, p. 426, https://doi.org/10.3390/educsci16030426.
[109] West, C., Q. Chen and N. Boika (2024), “Bringing them back: using latent class analysis to re-engage college stop-outs”, Frontiers in Psychology, Vol. 15, https://doi.org/10.3389/fpsyg.2024.1297464.
[129] Williams, P. (ed.) (2007), “Undergraduate Research Experiences Support Science Career Decisions and Active Learning”, CBE—Life Sciences Education, Vol. 6/4, pp. 297-306, https://doi.org/10.1187/cbe.07-06-0039.
[32] Xu, Y. (2012), “Career Outcomes of STEM and Non-STEM College Graduates: Persistence in Majored-Field and Influential Factors in Career Choices”, Research in Higher Education, Vol. 54/3, pp. 349-382, https://doi.org/10.1007/s11162-012-9275-2.
[34] Yin, Y. and C. Wang (2026), “Overeducation, vertical skill mismatch, and employment outcomes among Chinese university graduates: the mediating role of skill utilization”, Frontiers in Psychology, Vol. 17, https://doi.org/10.3389/fpsyg.2026.1859874.
[126] Young, N. et al. (2022), “Rubric-based holistic review: A promising route to equitable graduate admissions in physics”, Physical Review Physics Education Research, Vol. 18/2, https://doi.org/10.1103/physrevphyseducres.18.020140.
[66] Zakariya, Y. (2022), “Improving students’ mathematics self-efficacy: A systematic review of intervention studies”, Frontiers in Psychology, Vol. 13, https://doi.org/10.3389/fpsyg.2022.986622.
[24] Zegart, A. and E. Johnston (2026), The 2026 US Talent Report: Benchmarking America’s STEM Pipeline Amid the Great Talent Competition, Hoover Institution and Stanford Institute for Human-Centered Artificial Intelligence, https://hoover-s3-website.s3.us-west-2.amazonaws.com/s3fs-public/research/docs/Hoover_HAI_TalentReport_Final.pdf (accessed on 29 July 2026).
[86] Zhao, T., L. Perez-Felkner and S. Hu (2024), “The Impact of Merit Aid on STEM Major Choices: A Propensity Score Approach”, Educational Evaluation and Policy Analysis, Vol. 47/3, pp. 939-959, https://doi.org/10.3102/01623737241254842.