OCDE - Making comparable TVET data work for evidence-informed policymaking de 2026

 1

MAKING COMPARABLE TVET DATA WORK FOR EVIDENCE-INFORMED POLICYMAKING © OECD 2026

21 September 2026 Key messages • Policymakers often lack comparable evidence on who participates, what learners acquire and how programmes perform, limiting their ability to target reform and investment – yet, well-performing technical and vocational education and training (TVET) can help c ountries respond to changing skill needs by preparing young people and adults for occupations central to the green and digital transitions. • The global TVET evidence base remains fragmented. The Global TVET Data (GTD) mapping of 31 international datasets finds strong coverage of access and participatio n but weaker coverage of work-based learning, labour-market outcomes, quality, financing, governance and the teaching and training workforce. Data are also often difficult to disaggregate by sex, age, region and programme type and have different study design and different operationalisation mainly due to other priority population groups. • Availability does not guarantee use. Policymakers and other stakeholders must navigate different definitions, reference populations and reporting periods, while data are often not sufficiently timely, dispersed across institutions and often lack access ible metadata or the identifiers needed to link participation, completion and employment outcomes. • Closing the gap requires co-ordinated action. Agreeing a minimum set of policy-relevant indicators, strengthening national statistical and analytical capacity, and investing in transparent platforms and knowledge-brokering functions that translate comparable evidence into decisions. Together, these efforts can support evidence-informed policymaking and better-pe rforming TVET systems.

What’s the issue? Technical and vocational education and training (TVET) is central to prepar ing young people and adults for rapidly changing labour markets . It supplies many of technical and middle-level skills and competences in construction, manufacturing, health, care, transport and other occupations. Its importance Making comparable TVET data work for evidence-informed policymaking2 

adults for rapidly changing labour markets . It supplies many of technical and middle-level skills and competences in construction, manufacturing, health, care, transport and other occupations. Its importance Making comparable TVET data work for evidence-informed policymaking2 

MAKING COMPARABLE TVET DATA WORK FOR EVIDENCE-INFORMED POLICYMAKING © OECD 2026

extends beyond the green and digital transitions: many economies already face pers istent labour shortages and regional mismatches between supply of and demand for workers and s kills, while raising income and structural transformation are increasing demand for skilled workers in many economies (Fiest, 2024[1]) (Arias Ortiz et al., 2020 [2]). Better information on the distribution of skills and workers, supported by more harmonised indicators, can help policymakers identify shortages and mis matches, anticipate changing skills needs and learn from how other systems respond. In 2023, 75% of employers across OECD reported difficulties filling roles, with skilled trades among the profiles most difficult to find (United Nations, Dept of Economic and Social Affairs, 2019 [3]; ManpowerGroup, 2024 [4]). These shortages do not make TVET the only policy response, but they increase the need for systems that can identify changing occupational requirements and adjust programmes accordingly. Sound, evidence-informed governance is therefore essential for TVET systems to fulfil this role . Because TVET spans schools, workplaces, and adult learning, steering it r equires evidence from both education and labour-market systems. Reliable national data already provides key information for policy decisions; relevant, high-quality and timely data can support skills-needs analysis, agenda setting, funding decisions, monitoring, and evaluation (OECD/European Commission, 2025 [5]). Internationally comparable data adds a further dimension by helping policymakers and other stakeholders identify areas requiring action, benchmark performance, learn from peers and interpret national outcomes in a wider context . Yet the international TVET evidence base remains limited and is not consistently used in policymaking. The challenge has two dimensions. First, the availability of international harmonised TVET

action, benchmark performance, learn from peers and interpret national outcomes in a wider context . Yet the international TVET evidence base remains limited and is not consistently used in policymaking. The challenge has two dimensions. First, the availability of international harmonised TVET indicators is constrained by uneven thematic and country coverage, system complex ity, and methodological inconsistencies – for example in defining TVET and reference populations (see Box 1) (OECD, forthcoming[6]). Second, where data are available, accessibility barriers, institutional fragmentation and limited analytical capacity can prevent their effective use (Steiner-Khams i et al., 2024 [7]; Selwaness et al., 2022[8]). Why does it matter? When comparable and timely TVET data are unavailable or insufficiently us ed, policymakers face difficulties in identifying priority areas for action, comparing performance, asses sing whether investment produces results and learning from peers. While these challenges affect all systems because of the overall scarcity of comparable TVET data, their impact is particularly pronounced in countries with limited national statistical and analytical capacity, where opportunities to complement internat ional comparisons with robust national analysis are often more restricted. Without comparable TVET data, benchmarking and international policy learning are severely constrained. In general education, internationally harmonised data has proven a powerful policy lever. For example, the OECD Programme for International Student Assessment (PISA) provides a common reference point for identifying performance gaps and informing reform. TVET has no equivalent instrument with comparable global reach. This makes it harder to determine whether weak outcomes r eflect limited access, programme quality, occupational mismatch or wider labourmarket conditions. The OECD’s PISA-VET initiative is being developed to address part of this by assessing professional and employability skills in selected occupations, but it is still at an early stage (OECD, 2024 [9]). Limited comparability also weakens accountability for TVET investment and outcomes. Without consistent measures of completion, employment and earnings, governments can struggle to compare programmes, assess returns and justify funding decisions. In Scotland (United Ki ngdom), for example,

Limited comparability also weakens accountability for TVET investment and outcomes. Without consistent measures of completion, employment and earnings, governments can struggle to compare programmes, assess returns and justify funding decisions. In Scotland (United Ki ngdom), for example, apprenticeship monitoring has relied heavily on participation targets and shortterm self-reported outcomes, prompting the OECD to recommend linked administrative data and longer-term monitori ng (OECD, 2022 [10]). In Benin, limited programme-level labour-market data prevent robust estimates of the returns to technical secondary education. Strengthening national data systems while improving alignment 3

MAKING COMPARABLE TVET DATA WORK FOR EVIDENCE-INFORMED POLICYMAKING © OECD 2026

with international reporting frameworks is therefore an important policy lever. Wi thout comparable evidence, the basis for performance management, investment decisions and cross-country learning remains fragmented (World Bank, UNESCO and ILO, 2023 [11]). All countries bear the cost of fragmented evidence, but lower-capacity systems face a greater risk of exclusion from policy dialogue and are less able to benchmark their performance. Ministries may lack the technical infrastructure, skilled personnel and data management systems ne eded to translate complex international datasets into actionable policy recommendations (UNESCOIESALC, 2025 [12]). Strong national data systems are the foundation for evidence-informed TVET policy. Com parable international data can add further value by enabling benchmarking, peer learning and cr oss-country analysis, but only when countries have the capacity to produce, interpret and use the underlying evidence. The limited availability of TVET-specific comparable data leaves a structural gap in the global education evidence landscape. Unlike general education, where decades of international co-ordination have produced robust monitoring systems, TVET has not benefited from such investments in com parable data infrastructure (World Bank, UNESCO and ILO, 2023 [11]). As a result, policy decisions with significant consequences for learners, labour markets and economies are made with less evidence tha n those consequences warrant.

Understanding the gaps: Why comparable TVET data falls short of policy needs

data infrastructure (World Bank, UNESCO and ILO, 2023 [11]). As a result, policy decisions with significant consequences for learners, labour markets and economies are made with less evidence tha n those consequences warrant.

Understanding the gaps: Why comparable TVET data falls short of policy needs The consequences outlined above stem from a dual challenge constraining the glo bal TVET data landscape for policymaking. On one hand, the availability of internationally comparable TVET data remains limited, shaped by persistent gaps in coverage and by technical constraints roo ted in the complexity of TVET systems (see Box 1). On the other, where comparable data exist, practical and institutional barriers prevent usage in policymaking. This section sets out the practical causes of these challenges.

What the Global TVET data mapping reveals so far Research conducted under the forthcoming Global TVET Data Report maps 31 international datasets with stakeholder input from a technical workshop and semi-structured interviews (OECD, forthcomi ng[6]). The mapping documents geographic and thematic coverage, time coverage, data sources, methodo logical standards and the scope for disaggregation. It confirms that policymakers do not draw on a single harmonised source: they combine datasets that differ in coverage, comparability, granularity and timeliness, often supplementing them with national statistics, qualitative evidence and expert judgement. Geographic and thematic coverage remain uneven. Ten of the 31 datasets have global coverage, while nine focus exclusively on Europe. European countries therefore benefit from a denser ecosys tem of established regional sources, including several datasets maintained by Eurostat and C edefop, whereas equivalent multi-country sources are less extensive in Asia-Pacific, the Middl e East and North Africa, and Sub-Saharan Africa . Access and participation (e.g. entry into participation in TVET) appear in 84% of datasets and relevance and outcomes (e.g. alignment with labour-market needs and learner outcomes) in 58%, while the teaching and training workforce (e.g. teachers, trainers and their characteristics), financing (e.g. resources and expenditure), quality (e.g. standards, processes, learning conditi ons and

datasets and relevance and outcomes (e.g. alignment with labour-market needs and learner outcomes) in 58%, while the teaching and training workforce (e.g. teachers, trainers and their characteristics), financing (e.g. resources and expenditure), quality (e.g. standards, processes, learning conditi ons and assessments) and governance (e.g. institutional arrangements and system management) are covered in only 42% to 48%. Opportunities to disaggregate indicators by sex, age, region, p rogramme type or other priority population groups are also inconsistent. These gaps closely match th e information needs raised by stakeholders, particularly on work-based learning, labour-market outcomes and the quality of provision. A complementary indicator-level analysis shows that similarly labelled ind icators are not necessarily comparable, underscoring the need for greater alignment across international definition s and classifications. For example, indicators measuring enrolment or participation in vocational education differ4 

MAKING COMPARABLE TVET DATA WORK FOR EVIDENCE-INFORMED POLICYMAKING © OECD 2026

in the age groups and education levels covered, whether they include formal, non-formal or wor k-based learning, and whether results are reported as percentages or raw headcounts. Methodological differences therefore constrain comparison even where indicators appear to measure the s ame concept. UNESCO Institute for Statistics, ILOSTAT and UNESCO-OECD-Eurostat frameworks are the standards most frequently referenced, yet seven datasets rely on national definitions alon e (OECD, forthcoming [6]). National definitions preserve important context, but without transparent mappings to common classifications they limit benchmarking and longitudinal comparison. Box 1. Why TVET is difficult to measure consistently A common starting point: What counts as TVET? UNESCO defines TVET broadly as education, training and skills development rel ating to occupational fields, production, services and livelihoods, delivered as part of lifelong lea rning at secondary, postsecondary and tertiary levels and including work-based and continuing learni ng (UNESCO, 2016 [13]). National systems operationalise this broad concept differently. TVET may be school-based, workplacefields, production, services and livelihoods, delivered as part of lifelong lea rning at secondary, postsecondary and tertiary levels and including work-based and continuing learni ng (UNESCO, 2016 [13]). National systems operationalise this broad concept differently. TVET may be school-based, workplacebased or mixed; offered before entry to employment or to workers already employed; and delivered by public, private, employer or community providers (Kis, 2020 [14]; UNESCO Institute for Statistics, 2025[15]). These differences make it difficult to define a single statistical populat ion that captures all forms of TVET while still distinguishing them from general and pre-vocational education. Expanding on current conceptualisations and co-producing measurable and inclusive definiti ons offers a promising foundation for more comprehensive and comparable TVET data in the future. Differences in definitions and methodologies: How is TVET classified? Data reported under the label “TVET” may refer to different programme types and learner groups. Some datasets classify programmes through the International Standard Classification of Educati on (ISCED), while others use national definitions or occupational classifications. ISCED 2011 distinguishes between academic and professional education at levels 6-8 but it does not provide fully operational definitions , leading to inconsistent application and measurement of higher TVET (UNESCO Institute for Statistics, 2018[16]). Comparisons also change depending on whether the reference population is all young people, all learners, adults, apprentices or participants in particular programme levels. More consistent metadata documentation, combined with sustained international collaboration, can gradually strengthen the comparability and reliability of TVET statistics across countries. The main Global TVET Data Report will examine how more practical guidance, methodological explanations and country examples could help map national programmes, fields of study, occupations and sectors to international classifi cations such as ISCED, ISCO and NACE . For example, the informal expert team collaborating on reviewing the ISCED classifications, aims to clarify the definitions of higher profess ional education and workbased learning.

Non-formal and informal TVET: What is still excluded from the data?

such as ISCED, ISCO and NACE . For example, the informal expert team collaborating on reviewing the ISCED classifications, aims to clarify the definitions of higher profess ional education and workbased learning.

Non-formal and informal TVET: What is still excluded from the data?

In many countries, technical and vocational skills are acquired through traditiona l apprenticeships, on-the-job training and other non-formal or informal pathways. When these pathways do not lead to an officially recognised qualification, they generally fall outside ISCED-bas ed collections. The UNESCO-OECD-Eurostat international data collection also focuses on formal education, while l abourforce surveys do not yet consistently identify the type, duration or quality of nonformal training. As a result, a substantial part of TVET provision cannot be included reliably in international benchmarking or policy learning (UNESCO Institute for Statistics, 2025 [15]; UNESCO Institute for Statistics, 2018[16]; Kis, 2020[14]). However, experience from the measurement of lifelong and informal learning suggests that careful conceptualisation can progressively bring these pathways into the statistic al picture (OECD, 2026[17]; OECD, 2026[18]). 5

MAKING COMPARABLE TVET DATA WORK FOR EVIDENCE-INFORMED POLICYMAKING © OECD 2026

Barriers to using comparable TVET data for policymaking Stakeholder consultations and broader evidence on education policymaking identify five recurring barriers between the publication of comparable data and their use in decisions (Suazo -Galdames, Saracostti and Chaple-Gil, 2025 [19]). Figure 1 summarises these barriers; the examples below show how they arise specifically in TVET. Figure 1. Five barriers between comparable TVET and policy use

• Relevance. Comparable sources are strongest on access and participation, but less often c over the quality of work-based learning, informal apprenticeships (i.e. skills traini ng based on informal agreements between apprentices and experienced craftspeople) (Hofmann et al., 2022[20]),

• Relevance. Comparable sources are strongest on access and participation, but less often c over the quality of work-based learning, informal apprenticeships (i.e. skills traini ng based on informal agreements between apprentices and experienced craftspeople) (Hofmann et al., 2022[20]), completion, employment outcomes, financing, governance and the teaching and training workforce. This limits, for example, policymakers' ability to assess whether programm es lead to positive learner and labour-market outcomes. Stakeholders in OECD and non-O ECD countries identified these areas as priorities, indicating a mismatch between what is easiest to collect and what is most useful for policy (Steiner-Khamsi et al., 2024 [7]). • Accessibility. A published “TVET participation rate” may use all 15-24-year-olds, enrolled learners or participants in selected programme levels as its reference population. Reference pe riods may also differ, including a specific survey period or the current academic year. Unless such meta data are clearly presented alongside the data, users may compare rates constructed from different populations or timeframes (Jadeau and Fogarassy, 2025 [21]; OECD, 2020 [22]). Until metadata is systematically reported, including definitions, reference populations, programme coverage, reporting periods, calculation methods and known data-quality limitations, this will continue to be a barrier. Missing observations, differences in data-collection methods and other na tional methodological variations should also be clearly documented, as they can af fect the validity and interpretation of cross-country comparisons.

Available indicators do not always address priority policy questions or reflect the specific features of regional and national TVET systems.Relevance Data, definitions and metadata are dispersed across platforms and are not always easy for users to locate, retrieve or combine.Accessibility Responsibilities for TVET data are distributed across institutions that may use different classifications, identifiers and reporting practices, limiting interoperability and coherence. Institutional fragmentation Using comparable TVET data requires the skills, infrastructure and institutional resources needed to integrate sources, assess data quality and translate findings into policy advice.Analytical capacity

Responsibilities for TVET data are distributed across institutions that may use different classifications, identifiers and reporting practices, limiting interoperability and coherence. Institutional fragmentation Using comparable TVET data requires the skills, infrastructure and institutional resources needed to integrate sources, assess data quality and translate findings into policy advice.Analytical capacity Indicators with similar labels may refer to different populations, programme types, reference periods or methodologies, making interpretation and comparison difficult.Understandability6 

MAKING COMPARABLE TVET DATA WORK FOR EVIDENCE-INFORMED POLICYMAKING © OECD 2026

• Institutional fragmentation. TVET data are often held by, among others, education and labour ministries, training providers, qualifications authorities and social pa rtners. If these actors use different programme codes or learner identifiers, enrolment records cannot be linked reliably to completion, employer or employment data. The consequence is not merely duplicated effort: policymakers cannot reconstruct learner pathways across schools, TVET programmes, furth er education and employment, n or can they produce coherent national inputs for international reporting (Pham, 2026[23]; ETF (European Training Foundation), 2013[24]; Selwaness et al., 2022[8]).

As a result, barriers to progression and lifelong learning often remain difficult to identify. • Analytical capacity. TVET analysis often requires linking education and labour-market data across different classifications and time periods. Estimating completion and post-programme employment, for example, may require matching provider records to social-security or labour-force data while distinguishing school-based, apprenticeship and adult programmes. Many systems lack the staff, secure infrastructure or data-sharing arrangements needed for this work (UNESCOIESALC, 2025[12]; ILO; African Development Bank, 2023[25]; Selwaness et al., 2022[8]). • Understandability. Identical labels do not always describe identical measures. “TVET enrolment” can differ by age, ISCED level, formal status and inclusion of work-based comp onents;

• Understandability. Identical labels do not always describe identical measures. “TVET enrolment” can differ by age, ISCED level, formal status and inclusion of work-based comp onents; “Employment after TVET” can vary by follow-up period and definition of employment. Without clear and accessible metadata, users cannot assess whether a comparison is valid (Kis, 2020 [14]).

Together, these barriers increase the time and expertise required to turn a pu blished indicator into a defensible policy conclusion. Improving data supply without addressing these barriers will therefore have limited impact on policy use. Efforts underway Current initiatives can be grouped along the three pathways shown in Figure 2: expand ing comparable data, strengthening national statistical and analytical capacity, and translat ing evidence into policy use. Progress is visible in each area, but initiatives remain dispersed an d their TVET-specific coverage is still limited. Figure 2. Interconnected pathways to improve TVET data usage

Improving TVET data usage Strengthening translation of evidence into policy Building statistical capacity Expanding data availability 7

MAKING COMPARABLE TVET DATA WORK FOR EVIDENCE-INFORMED POLICYMAKING © OECD 2026

Expanding the availability of comparable TVET data International organisations already provide important components of the TVET evidence base. The OECD’s “Spotlight on Vocational Education and Training” documents the diversity and outcomes of uppersecondary VET systems (OECD, 2023 [26]) The UNESCO Institute for Statistics monitors Sustainable Development Goal 4, the global commitment to ensure inclusive and equitable quali ty education and promote lifelong learning opportunities for all , including indicators on participation in technical and vocational programmes. The International Labour Organization ’s ILOSTAT database provides labourmarket and work-based-learning indicators, while the European Training Foundation produces evidence on partner-country TVET systems. These sources are substantial, but they do not yet form a unified TVET framework and often cannot isolate non-formal provision, apprenticeship quality, financing, governance or

market and work-based-learning indicators, while the European Training Foundation produces evidence on partner-country TVET systems. These sources are substantial, but they do not yet form a unified TVET framework and often cannot isolate non-formal provision, apprenticeship quality, financing, governance or comparable post-programme outcomes. Regional initiatives add depth, but their scope and functions differ. The Inter-American Development Bank’s CIMA platform (Centro de Información para la Mejora de Aprendizajes) brings together harmonised education indicators for Latin America and the Caribbean, and ILO/Cinterfor support knowledge exchange and south-south co-operation among more than 65 TVET institutions. The Association of Southeast Asian Nations established an ASEAN TVET Council with regional partners to supp ort co-ordination. These initiatives respond to regional priorities, but they were developed for different purpo ses and do not use a common set of TVET definitions, indicators, classifications or reporting practic es. They therefore support regional comparison and peer learning, but not yet consistent benchmarking across r egions. A crossregional framework c an complement them by linking regional data to agreed definitions and a minim um set of comparable TVET indicators. Several targeted initiatives are addressing specific gaps. TGTA, a collaboration established in 2025 between the OECD, the International Labour Organization, European Training Fou ndation (ETF), UNESCO-UNEVOC and the BIBB, funded by the German Federal Ministry for Educati on, Family Affairs, Senior Citizens, Women and Youth (BMBFSFJ ), is developing the forthcoming Global TVET Data Report and a feasible minimum set of priority indicators (see Box 2 ). The OECD’s Programme for International Student Assessment for Vocational Education and Training (PISA-VET) is being developed as the first international assessment of professional and employability skills in selected occupations (OECD, 2024[9]). UNESCO-UNEVOC’s Global Skills Tracker adds cross-country information on changing skill s demand across occupations and industries (UNESCO, 2025 [27]). Building statistical capacity for TVET data analysis (and production)

international assessment of professional and employability skills in selected occupations (OECD, 2024[9]). UNESCO-UNEVOC’s Global Skills Tracker adds cross-country information on changing skill s demand across occupations and industries (UNESCO, 2025 [27]). Building statistical capacity for TVET data analysis (and production) Expanding the availability of comparable TVET data is necessary but insufficient . The scale of overall education support is substantial: the UNESCO Institute for Statistics reported training more than 250 national statisticians from about 70 Member States in 2017 (UNESCO Institute for Statistics, 2017[28]). Yet the Global TVET Data mapping and stakeholder consultations identify few international initiatives dedicated primarily to TVET statistical capacity, particularly for linking educati on, apprenticeship and labour-market records. The Partnership in Statistics for Development in the 21st Century (PARIS21), whose Se cretariat is hosted by the OECD, provides guidance covering statistical planning, production, governance and evaluation. Its approach demonstrates how countries can combine locally defined priorities with common standards and peer learning, including through national strategies for the developmen t of statistics

(PARIS21, 2020[29]).

The UNESCO Institute for Statistics (UIS) supports national education data systems throug h training, technical assistance, methodological guidance and diagnostic tools . Its Education Data Quality Assessment Framework promotes National Strategies for the Development of Education Statistics8 

MAKING COMPARABLE TVET DATA WORK FOR EVIDENCE-INFORMED POLICYMAKING © OECD 2026

(NSDES): countries establish technical and steering structures, map data sources, i dentify gaps, assess quality and develop an action plan. UIS has applied this approach in the Paci fic and other country programmes, providing a documented model for building the foundations required for intern ational comparability (UNESCO Institute for Statistics, 2017[28]; Paris21, 2025[30]). Evidence from Africa highlights the size of the remaining gap. An ILO and African Development Bank report identifies weak TVET research and data as constraints on stronger links bet ween training and productive

comparability (UNESCO Institute for Statistics, 2017[28]; Paris21, 2025[30]). Evidence from Africa highlights the size of the remaining gap. An ILO and African Development Bank report identifies weak TVET research and data as constraints on stronger links bet ween training and productive sectors (ILO; African Development Bank, 2023 [25]). UIS guidance similarly notes shortages of skills to analyse, co-ordinate and communicate education data (UNESCO Instititute for Statist ics, 2025[31]). These sources document the need more clearly than a mature TVET-specific capacity programme: reforms such as expanding quality apprenticeships require systems a