Better skills data for smarter financing of education and training

2. Current data gaps and their consequences

Recent decades have seen remarkable progress in data collection across OECD countries. Governments now routinely gather information on enrolment, graduation rates, and basic labour market outcomes. Administrative systems have become more sophisticated, and statistical agencies have expanded their capacity to monitor education and training. Yet despite these advances, significant gaps remain in the evidence available to guide skills policy.

Many countries still lack a unified and comprehensive view of their skills systems. Information on education, training, and labour market outcomes is often fragmented across multiple agencies and databases. Where data do exist, they are frequently incomplete, inconsistent, or incomparable across time and place.

The problem goes beyond the availability of basic statistics. One critical gap concerns financial information. While governments track overall education budgets, they often lack detailed data on how much is spent on specific programmes, and how spending is distributed across different population groups and stages of the life course. Understanding the flow of public resources is essential, but it represents only one piece of the puzzle. This challenge is closely linked to the absence or weakness of quality assurance mechanisms. Evidence from OECD work on adult education and training shows that increased spending, in the absence of robust quality assurance and outcome tracking, can lead to the expansion of low-quality provision and weak returns on investment (OECD, 2024[3]).

Policymakers need a broader set of indicators to understand whether investments are achieving their goals. They need data on participation and completion rates, on how graduates perform in the labour market, on the quality and relevance of training, and on equity across different groups and regions. With these complementary indicators, policymakers can better see the full picture. They can track how skills are developed and used across the life course, or judge whether public resources are being deployed effectively.

This section identifies the main dimensions of these data gaps, highlights which policy areas are most affected, and explains how they hinder effective decision making.

2.1. Fragmentation of data systems

In many countries, skills-related data are dispersed across siloed systems managed by different ministries, agencies and levels of government (OECD, 2019[4]). This fragmentation typically reflects the division of responsibilities across government portfolios and across levels of administration. In some systems, national governments set policy while subnational authorities manage delivery. In others, particularly federal systems, key functions are allocated across levels. For example, education is largely a state responsibility in the United States and a provincial responsibility in Canada.

Data systems are typically developed by different public bodies, each aligned with their specific mandate and regulatory requirements. As a result, information is collected and managed in separate systems reflecting sectoral responsibilities – for example, education authorities focus on enrolment and graduation, labour ministries on employment and training programmes, and tax and social security institutions on earnings and employment histories. While these systems serve important administrative functions, they are rarely designed to operate together. Fragmentation is not only institutional or technical but also semantic: systems may be connected yet remain difficult to use jointly if they rely on different definitions, nomenclatures, metadata or exchange standards (see Section 2.4).

This limits their comparability and integration, making it difficult to link information across education, training and labour market outcomes. As a result, policymakers lack a consolidated view of the skills system and face constraints in assessing how individuals move through it and how public spending translates into results.

The consequences for policy are severe (see Section 3). Fragmentation limits the ability to trace how investments in education and training translate into labour market outcomes. Without linking school or training records to employment and earnings data, policymakers cannot answer basic questions. How do graduates fare in the job market? Which qualifications lead to sustainable employment? Which training programmes deliver the best returns? Similarly, budget data on skills programmes may not be connected to outcome indicators, making it difficult to assess value for money.

A further complication is timescale: tracing the link between education investments and sustainable employment outcomes can take years or even decades, by which point the economic and occupational landscape may have shifted substantially. The result is that governments often work with partial and disconnected information. This hampers their capacity to design coherent strategies or to allocate resources where they are most effective. Policy decisions are inevitably made with incomplete evidence, but the degree of incompleteness matters. Better data can narrow the margin of uncertainty, even if it cannot eliminate it. Without integration, however, opportunities to learn from past investments are lost.

2.2. Lack of longitudinal linkages

A second major gap concerns the limited availability of longitudinal data that track individuals over time. Understanding the impact of skills investments requires following people through their education, training, and careers. Yet in many countries, this proves difficult. Education, employment and earnings data are collected separately, often with no unique identifiers or data-sharing arrangements that allow them to be matched. Individuals can be observed at discrete moments – when they enrol in a course, when they graduate, when they register as employed – but the connections between these moments are often difficult to establish. Their subsequent trajectories disappear from view. These linkages matter not only between education and work, but also within education systems themselves: tracking trajectories such as dropout, grade repetition, transitions between levels and access to higher education is essential for understanding where and why skills development falters. Some countries, notably the Nordic nations, have made significant progress in this area by linking education, employment and earnings records through unique personal identifiers – demonstrating that integrated longitudinal data systems are achievable. Australia’s Person-Level Integrated Data Asset (PLIDA) similarly enables longitudinal tracking of individuals before and after training, combining multiple administrative sources to assess outcomes such as post-completion earnings. France’s InserJeunes platform links school, apprenticeship and employment records to monitor, within a single framework, both further study and the labour market insertion of young people leaving vocational programmes.

The consequences are profound. Without longitudinal linkages, policymakers cannot assess the long-term effects of different education and training pathways. Critical questions go unanswered. Do graduates of vocational programmes secure stable jobs? How do their earnings evolve over the course of their careers? Does mid-career training improve employment resilience during economic downturns? Evaluating the effectiveness of a training programme requires knowing not only who participated, but also how their employment outcomes changed compared to similar individuals who did not take part. In the absence of linked data, such analysis is rarely possible. Notable exceptions exist, particularly among countries that have invested in integrated administrative data infrastructures.

The absence of longitudinal evidence also prevents governments from identifying which groups benefit most from training and which may need additional support. Without longitudinal data, such disparities remain hidden. Policies risk overlooking those who are less well served, and resources may flow to groups who would have succeeded regardless of intervention.

Ultimately, without these linkages, skills policies are designed on short-term snapshots rather than a fuller picture of outcomes. At the same time, even comprehensive life‑course data must be interpreted with caution, as the labour market conditions under which outcomes were observed may differ from those facing current cohorts.

2.3. Coverage gaps across the skills system

Data collection is generally strongest for formal education systems. Primary, secondary, and tertiary education are well documented in most OECD countries, with regular reporting on enrolment, completion, and basic outcomes. Yet coverage weakens at both ends of the learning continuum. Early childhood education and adult training remain poorly monitored, despite their growing importance for skills development across the life course.

In early childhood, data collection is particularly fragmented and patchy. Early childhood services often operate under multiple regulatory frameworks (Leseman and Slot, 2025[5]; OECD, 2019[6]). Some providers fall under the purview of education ministries, while others are overseen by social services, family affairs, or health ministries. Private providers, community organisations, and informal care arrangements further complicate the landscape. This institutional fragmentation makes comprehensive data collection challenging. Many countries lack consistent information on participation rates, quality of provision, staff qualifications, and developmental outcomes across these diverse settings. Data collection often begins only when children enter primary school, leaving the crucial first years largely undocumented. This creates difficulties in evaluating the impact of early investments on later educational performance and labour market success. The gap is particularly problematic given the well-documented importance of strong foundations in the early years and the long-term returns that early interventions can generate.

Adult learning represents an even larger blind spot in skills data. Many adults acquire new competences through non-formal courses, employer-provided training, apprenticeships, or online platforms. Yet these activities are rarely captured in official statistics. Because much of this learning occurs outside formal institutions and across diverse settings, systematic monitoring of participation, investment, and outcomes is limited (OECD, 2025[7]). Employer-sponsored training, which represents a substantial share of adult learning activity, is particularly difficult to track. One practical reason is that firms rarely structure their accounts in ways that allow training-related costs to be readily identified, making accurate reporting burdensome and potentially unreliable. As a result, governments have little visibility on how much employers invest in training, which workers participate, what skills are being developed, and what benefits these efforts generate over time. In many cases, employers address evolving skills needs directly through in-house training, often without reference to government-collected data or public policy frameworks. This underscores the importance of understanding, rather than supplanting, private training activity.

The consequences are significant. Policies to expand early childhood education or lifelong learning are often introduced with incomplete evidence on who participates, the quality of provision, and the outcomes achieved. Without such data, it is difficult to know whether disadvantaged families have equitable access to high-quality childcare, or whether mid-career reskilling programmes genuinely help workers transition into stable employment. This lack of visibility means that vulnerable groups may be underserved without policymakers realising it. Children from low-income households may miss out on early interventions that could transform their trajectories. Adults in low-skilled jobs, older workers, and those in rural or remote areas may have limited access to training opportunities. As a result, investments risk being directed towards those already better served, reinforcing rather than reducing inequalities.

2.4. Inconsistent definitions and quality issues

Even where data exist, differences in definitions, methodologies, and quality standards often limit their usefulness. Education and training data are not always collected using consistent classifications, making it difficult to compare across programmes, institutions, or regions. What one country records as vocational training may differ substantially from another, reflecting variations in programme duration, entry requirements, certification standards, and the balance between classroom instruction and workplace learning. Yet these differences may not be clearly reflected in the data, creating apparent comparability where little actually exists. These challenges are further magnified in international comparisons, where differences in definitions and data collection practices can lead to misleading conclusions if not carefully accounted for. The problem extends within countries as well. Regional authorities may apply different criteria for categorising participants or measuring completion, while institutions may report outcomes at different intervals or use different definitions of employment success.

These definitional inconsistencies are compounded by weak data quality, which further undermines usability. Errors in administrative records, outdated information, and missing variables all reduce the reliability of analyses. Consider earnings data that lack details on occupation, industry, or hours worked. Without this context, it becomes impossible to distinguish between full-time stable employment and precarious part-time work. Similarly, training records may note participation but not completion, or record completion without capturing the credential obtained. Demographic information is often incomplete, preventing analysis of equity across different groups. When datasets from different sources cannot be meaningfully combined due to incompatible standards and quality controls, the result is a fragmented system that resists integration despite the abundance of available information. Improving data quality requires practical instruments rather than general principles alone: data dictionaries, common schemas, harmonised reference values, clear data‑entry rules and quality indicators can make data quality visible, measurable and improvable over time.

The consequence is a troubling paradox. Governments may find themselves unable to answer basic questions despite sitting on vast administrative databases, simply because the information they hold is too fragmented, inconsistent, or unreliable to yield meaningful insights.

2.5. Timeliness and accessibility

Another common limitation is that data are not available when they are most needed, creating a temporal mismatch between information collection and policy decisions. In many countries, there is a substantial lag between the collection of information and its release, sometimes stretching across multiple years. Administrative data on training programme participation, completion and employment outcomes, for instance, may take one or two years to be processed, validated, and published, even though the raw information exists in government systems much earlier. By the time these data become available, budget allocations have been set and programme adjustments have been made based on older evidence or anecdotal information.

Accessibility poses an additional challenge. Even when relevant data are available, legal restrictions, confidentiality rules, or technical limitations often prevent policymakers and analysts from accessing them. While these safeguards serve essential purposes in protecting individual privacy (see Section 5.3), their cumulative effect can limit the analytical use of data for policy purposes. In many cases, policymakers must rely on outdated or less relevant indicators simply because more timely and detailed data are not accessible.

The combined effect of these delays and access restrictions is profound. Decisions are frequently made on the basis of incomplete or outdated evidence, undermining governments’ ability to respond swiftly to economic shocks or adapt training systems to evolving skills demands. Emerging inequities may go undetected until they have already widened significantly.

2.6. Granularity: Equity and regional blind spots

Insufficient granularity in data can obscure important disparities across population groups and regions. In many countries, data are collected at a highly aggregated level, with limited disaggregation by gender, age, region or rurality, socio‑economic background, migration status, disability status, or ethnic, cultural or linguistic background. Without this level of detail, inequalities in access to and outcomes from training may go undetected. National averages may show strong overall employment outcomes after vocational training, while concealing that women or disadvantaged minorities benefit far less.

Regional differences often remain similarly hidden. Subnational data on participation, investment, and outcomes are frequently incomplete or inconsistent, making it difficult to assess how national policies perform in different local contexts. A region with low training participation and high unemployment may be masked by national figures, and varying regional data standards can further hinder comparison across jurisdictions. More granular data would also support territorial planning processes, helping to identify differentiated skills needs across regions, productive sectors and local demographic dynamics.

The lack of granular information has direct implications for policy. Governments may inadvertently allocate resources to groups or areas that are already well served, while those most in need remain overlooked. Workers in remote areas, for example, may have limited access to upskilling programmes, and individuals from low-income households may face barriers to participation that go unaddressed. Limited granularity therefore undermines both the efficiency and the equity of skills policies.