This chapter reviews a range of international evidence to address the challenging question of “what does higher education cost?”. The chapter highlights the complexity of answering this question given the range of variables that influence costs. These variables include the scope, nature and combination of activities undertaken by higher education providers, the “quality” at which teaching, research and service outputs are delivered and prevailing cost drivers in the locations where higher education institutions operate – most notably, differences in the cost of employing skilled labour in a labour-intensive sector.
The Financial Sustainability of Higher Education
Insights from Policy in OECD Countries
Abstract
Five main factors arguably have the greatest impact on the total institutional costs and revenue-raising ability of higher education institutions: a) the scope and type of teaching activity, b) research intensity, c) the scale of commercial and ancillary activity, d) the level of international student enrolment and e) institutional scale and location.
On average, two-thirds of expenditure on tertiary education institutions goes on staff costs, with variation in this share influenced by the scale of non-teaching activities, rates of outsourcing and labour costs in higher education relative to countries’ general wage levels and cost of living.
Disentangling expenditure – and thus costs – for different higher education activities is complex because staff time and space in buildings is often shared across multiple activities. Activity-based costing (ABC) systems make it possible to calculate the full economic cost of different activities by identifying direct costs and attributing overheads in a transparent manner, although such approaches have been implemented in relatively few OECD countries.
The main drivers of unit costs in instruction in higher education are student-to-staff ratios, staff teaching loads and staff employment status, although non-staff costs do contribute significantly to unit costs in fields using laboratories or other expensive infrastructure.
The limitations of existing metrics of the quality of the outputs produced by higher education (which mean it is hard to judge how well students have developed knowledge and skills or the real contribution of research to the stock of knowledge, for example) and the tendency for the cost of specific activities in the sector to reflect the level of resources available – rather than price levels established through market forces – complicate interpretation of data on costs.
Over the decade 2013-2022, inflation-adjusted spending per student on HEIs – including on research – increased by 13% in OECD countries, with the largest increases in Central and Eastern Europe and declines in 10 countries, including Canada, Finland and Mexico. The changes are frequently driven by public-spending decisions. In more market-based systems, research confirms the difficulty of increasing productivity in higher education (the “Baumol effect”) but finds cost increases can often be attributed to legitimate attempts to maintain quality – by offering competitive salaries, for example – rather than profligacy and inefficiency.
Further work is needed to improve the availability of reliable data on expenditure and costs in higher education across more higher education systems, refine methods and share practices on cost measurement to balance usefulness with administrative burden for institutions and staff and to understand the impact of artificial intelligence on teaching and learning and associated productivity.
Higher education institutions are complex, hybrid organisations
What higher education institutions spend their money on depends on their profile. Across OECD countries and beyond, higher education institutions are expected to educate students at an advanced level, undertake research and provide service to society. However, the extent to which individual institutions – and the staff who work in them – engage in these different activities varies depending on the mission and profile of the institution in question. The range and scale of activities in a comprehensive research university is greater than in a specialist teaching-focused college. Moreover, the mix of different institutional profiles within higher education systems varies between countries.
In some systems, such as the United Kingdom or Norway, most tertiary-level students are enrolled in universities with a research mission – even if the extent of research activities varies. In others, such as Canada, Korea or the United States, a substantial proportion of students are enrolled in professionally or teaching-oriented colleges with no or only limited engagement in research. This diversity within higher education increases the complexity of describing and comparing institutions and systems and of defining comparable measures of performance and efficiency.
The portfolio of activities that higher education institutions undertake has a substantial influence on the costs that they incur, the revenue they can raise and the way they allocate available financial resources. Five main dimensions of differentiation between higher education institutions arguably have the greatest implications for institutional costs, revenue and resource allocation:
1. The scope and type of teaching activity: institutions may focus on academic programmes – as in many research universities – or professionally oriented provision – as in many universities of applied sciences and colleges – or provide a broad mix of academic and professional programmes in the same type of institution, as in universities in Czechia or Slovenia, for example. Institutions may be comprehensive, providing education in a wide range of fields, focused in a more limited range of educational fields, or specialised in a single or several closely related fields, as in the case of many music and arts schools or business and management colleges (Wagner-Schuster, 2022[1]).
2. Research intensity: in higher education systems with a formal – often binary – distinction between universities and non-university institutions, research is typically concentrated in universities. However, research intensity varies between universities in the same system, with some evidence of increasing concentration of research activity in smaller numbers of institutions (Aagaard, Kladakis and Nielsen, 2020[2]). Additionally, several OECD countries, including the Netherlands, Portugal or Finland, have taken steps over the last decade to increase research activity in non-university higher education institutions, somewhat blurring traditional institutional distinctions related to research (OECD, 2020[3]).
3. Commercial and ancillary activity: alongside core teaching and research activities, some of which – contract training and research, for example – may be classed as “commercial”, higher education institutions may engage in other commercial and ancillary activities in areas such as knowledge transfer, consulting services, student housing, organised sports or events management. The extent of these activities impacts the scale of employment and infrastructure in individual institutions and overall patterns of income and expenditure.
4. International student enrolment: in many OECD countries, international students pay higher fees than domestic students (Golden, Troy and Weko, 2021[4]). While institutional- and system-level internationalisation strategies may seek to attract international students for a variety of reasons, including to pursue international education and development goals (OECD, 2025[5]), the number of international students that an institution enrols ultimately affects the scale of institutional activity and – in many countries – the balance sheet.
5. Institutional scale and location. The overall size of higher education institutions – measured in terms of student enrolment or operating budget – will influence internal organisation and might ultimately influence opportunities for economies of scale, although, as discussed below, the evidence on the latter is not conclusive. Institutional size may be a function of teaching and research focus – with specialised institutions likely to be smaller than comprehensive ones – or related to location, with institutions serving less densely populated regions often smaller than those in metropolitan areas (Daraio and Daraio, 2022[6]).
In some respects, higher education institutions are more complex than most commercial businesses (OECD, 2007[7]) and the range and complexity of activities undertaken increase the importance of sophisticated approaches to financial management. However, while the managers of individual higher education institutions may have access to detailed information to track expenditure and costs in their institutions, comparable, system-wide cost data are only collected in a few systems – and even in these cases remain limited. International data that would permit comparison of the costs of higher education activities across countries do not exist, making it necessary to rely on imperfect proxy indicators, such as per-student expenditure (see below).
In broad terms, data collated at the level of higher education systems – part of which is also submitted to international data collections – break down institutional expenditure in two main ways:
1. By distinguishing resources spent on the compensation of academic and non-academic staff (salaries, social security and pension contributions and other staff benefits) from spending on other day-to-day running costs (including utilities, supplies, routine maintenance or contracts for computing-related services) and capital expenditure (new buildings or large pieces of equipment).
2. By estimating expenditure on specific activity types, such as teaching, research, commercial or ancillary activities. As discussed below, some limited aggregated data on these variables are collected for international educational statistics. Gaining a more detailed understanding of activities and associated costs requires more fine-grained data, which is available only in the few OECD systems that use system-wide the activity-based costing (ABC) methods discussed later in this chapter.
Two-thirds of expenditure on tertiary education institutions is used to employ staff
International education statistics demonstrate that tertiary education – in common with educational activities more generally – is a labour-intensive sector, with a substantial proportion of resources in tertiary education institutions allocated to staff costs. The joint collection of international education data by the United Nations Educational, Scientific and Cultural Organisation (UNESCO) Institute for Statistics (UIS), the OECD and Eurostat (the UOE data collection) asks national educational statistics agencies to provide data on how expenditure transferred to tertiary education institutions is allocated to compensation of academic staff, compensation of non-academic staff, other operating expenditure and capital expenditure.
The data collection methodology considers expenditure on educational institutions from the system perspective, focusing on the amounts transferred – top down – to institutions from public authorities, households, other private organisations and international funding sources, rather than on the detailed revenue and expenditure of individual tertiary education institutions. Expenditure on institutions in a given financial year – which is essentially the equivalent of the revenue in institutions’ balance sheets – and expenditure by institutions – the expenditure side of institutional balance sheets – are largely considered to be identical in international statistics, even though this does not reflect reality (OECD, 2018[8]). In practice, institutions may spend more or less than they earn each year, accruing surpluses or deficits. As such, the available system-wide international data are best considered as estimations rather than accurate measurements of the real levels of expenditure by institutions on each expenditure category.
Notwithstanding these caveats, Figure 2.1 illustrates that over half of current expenditure in 20221 – i.e. excluding expenditure on capital – in public and government-dependent tertiary education institutions2 is allocated to staff compensation in all OECD countries. On average, two-thirds (66%) of current expenditure in public and government-dependent institutions is allocated to staff compensation in OECD higher education systems – a proportion that has remained largely constant in recent years (OECD, 2020[3]). The proportion of institutional expenditure allocated to staff compensation varies from over 75% in France, Spain, Bulgaria, Costa Rica, Poland and Portugal to less than 60% in the United Kingdom, Korea, Chile, the Slovak Republic, Italy, Japan and Ireland.
Disaggregated international data on compensation of academic staff and non-academic staff – which are available for only 25 countries – show that over 60% of current expenditure is allocated to compensation of academic staff in Romania, Denmark and Portugal, whereas this proportion is around 30% in Finland, the United States, Australia and the United Kingdom. On average, the costs of employing non-academic staff account for around 25% of institutions’ current expenditure, with the highest shares seen in Luxembourg, Estonia, Lithuania, the United States and France, and the lowest reported shares in Ireland, Austria and Mexico. Some of the variation likely results from differences in the categorisation of specific sets of staff in national statistical systems.
Current and capital expenditure on public and government-dependent private tertiary education institutions by major expenditure category, 2022
Note: (1) OECD (unweighted) average computed with countries that have disaggregated data for teaching and non-teaching staff. (2) No data are available for Colombia, Greece and Switzerland. (3) The reference year for the United States and Ireland is 2021.
Source: OECD (2025[9]), Full dataset - Indicators, source, destination and nature of expenditure on education (dataset), https://data-explorer.oecd.org/s/32z (accessed on 19 August 2025).
As discussed later in this chapter in more detail, the factors driving the reported differences in institutional expenditure on staff costs reflected in Figure 2.1 are complex. In addition to the differences in the composition of national higher education systems highlighted earlier, three factors are likely to be significant:
The scale of expenditure on ancillary activities and capital investment included within the scope of “other current expenditure”. Where tertiary education institutions operate large student housing and catering operations, events services or sporting facilities, these will generally increase the share of institutional expenditure on non-staff costs. Further, although earmarked expenditure on capital investment is recorded separately in international statistics, other institutional expenditure on infrastructure improvements paid for out of core operating budgets will not be captured accurately and be recorded as current expenditure (OECD, 2018[8]).
The scope of service outsourcing. When certain services – such as support for information and communication technologies (ICT) – are outsourced by institutions, the associated costs are reflected in general current expenditure rather than in in-house staff costs.
Labour costs relative to the overall cost of living in each country, staffing ratios and the dominant forms of contract used in the tertiary education system. In broad terms, higher social security contributions and lower student-to-staff ratios equate to increased costs, as do higher proportions of staff with tenured – as opposed to hourly paid or fixed-term – positions.
As noted, data on capital expenditure in the UOE data system simply reflect the earmarked cash sums invested in the year of reporting, rather than infrastructure expenditure accrued over multiple years, and do not include the costs of capital depreciation or borrowing for capital investment. As capital expenditure is intrinsically variable between years, depending on institutional building programmes and availability of capital investment funds, the values for this indicator for any given year should be interpreted with caution. Nevertheless, in 2022, international data suggest that capital expenditure as a proportion of total current expenditure ranged from nearly 18% in New Zealand to less than 4% in Sweden, Finland, Mexico, Iceland, Ireland and Luxembourg.
Disentangling expenditure – and thus costs – by higher education activity is complex
Although the breakdown between staff and non-staff expenditure provides some insights into the costs of higher education, institutional managers and policy makers are ultimately more interested in understanding how much it costs to deliver different types of higher education output and activity. What does it cost to deliver a bachelor’s or master’s degree programme in physics or dance, for example. International educational statistics cannot tell us this but do seek to estimate the proportion of expenditure on higher education that is allocated to research and supporting “ancillary services” and, by stripping out these costs, provide a better estimation of the amounts invested in teaching and learning.
Some aspects of estimating the costs of research in higher education are relatively straightforward. For example, expenditure used to employ research assistants or post-docs who are exclusively engaged in research projects, can be identified and quantified relatively easily at departmental and institutional level. However, the close connection between research and teaching in higher education means that many academics undertake both teaching and research and higher education premises – particularly in research universities – are generally shared between teaching and research activities. This makes it challenging to disentangle the activities and attribute costs and expenditure that are specific to research or specific to teaching. Although the OECD’s Frascati Manual (OECD, 2015[10]) suggests standard practices for breaking down research and development (R&D) expenditure embedded into institutional budgets and some higher education systems use internal costing models, practices vary between countries. This results in comparability issues with international data (OECD, 2018[8]).
Current expenditure of public and government-dependent private tertiary education institutions by “type” of expenditure, percentage, 2022.
Note: (1) OECD figures are the unweighted averages of the 32 OECD member countries with available data. OECD average calculated by treating missing values as zero to ensure internal consistency (the sum of all categories equals 100%). This approach may underestimate or overestimate some categories, as for some countries data are not available or are included in other categories rather than truly zero. (2) No data are available for Colombia and Switzerland. Japan, Iceland, Croatia, Romania, Costa Rica and Israel do not report disaggregated data. (3) The reference year for the United States is 2021.
Source: OECD (2025[9]), Full dataset - Indicators, source, destination and nature of expenditure on education (dataset), https://data-explorer.oecd.org/s/33q (accessed on 19 August 2025).
As shown in Figure 2.2, the share of expenditure on tertiary education institutions reported as allocated to research activities varies substantially across OECD systems – although the challenges of calculating this indicator mean that some systems – such as Australia and Canada – do not report data. Based on the systems that do report data, roughly one-third (31%) of expenditure on public and government-dependent tertiary education institutions is allocated to research. An average of 3% is allocated to non-core ancillary services (see Box 2.1 for further explanation of what this includes), leaving the remaining 66% assumed to be allocated to teaching and learning and other activities not explicitly identified as research or ancillary services.
The highest shares of institutional expenditure on research are observed in the Nordic countries, Estonia and Germany, all of which report research shares of around 50%, ranging from 56% in Denmark to 46% in Germany. Several other European systems report research expenditure shares between 35% and 45% (Belgium, Czechia, Norway, Portugal, Greece, Poland, the Netherlands, Austria and Luxembourg), while the share is less than one-third in the remaining countries. The lowest reported shares of expenditure on research are observed in Ireland (14%), the United States (13%), Chile (6%) and Bulgaria (5%).
The different methods used to identify research expenditure – which vary between a simple identification of earmarked third-party funding for research to detailed estimates based on staff time-use surveys – mean that the accuracy of the reported shares and the extent to which they reflect the full extent of expenditure on research – including overheads – likely varies substantially. Although the reported 13% share of tertiary education expenditure on research in the United States partly reflects the high share or resources allocated to ancillary services and the country’s extensive community college sector that performs little or no research, it almost certainly fails to capture the true extent of expenditure on research across American public four-year institutions. Similarly, the reported 20% share of expenditure on research in UK universities fails to reflect the true extent of research activity. National data for 2022/23, drawn from the national cost-accounting system discussed below, show that 25% of the revenue of UK universities came from public and private research funders and research activity accounted for 34% of total costs in the sector (UKRI, 2024[11]). The difference between revenue for research and expenditure on research in UK universities reflects “cross-subsidy” of research, primarily with income from international student fees, as well as a misalignment between earmarked research funding and the real full economic costs of research activity (UKRI, 2025[12]).
The United States reports the highest share of spending on ancillary services, which is consistent with the on-campus accommodation model in many public four-year universities and the role these same institutions play in organised sports. The comparatively high rates (over 5%) reported in the United Kingdom, Estonia, Portugal, Bulgaria, Lithuania and Slovakia are also credible given the delivery of student services in these countries. However, differences in accounting approaches, student-service delivery models and the extent to which institutions provide services to the general public as well as just students will likely affect the scale of reported expenditure and complicate the interpretation of these data (OECD, 2018[8]).
The guidelines for the UOE data collection define ancillary services as services provided by educational institutions that are peripheral to institutions’ main educational and research missions. The two main components of ancillary services at tertiary level are:
Student welfare services: these include halls of residence (dormitories), dining halls and healthcare.
Services for the public: these include museums, radio and television broadcasting, sports, and recreational or cultural programmes.
All such ancillary services in educational institutions are included in the coverage of the expenditure data. The expenditure must have been explicitly designated or earmarked for ancillary services. This means that the amount spent for ancillary services may exceed the amount designated for these services in the form of fees charged to students or earmarked public fund allocations if other revenue streams are used to fund ancillary services.
The extent to which expenditure on ancillary services is reported to the UOE data collection varies from country to country. Many countries struggle to report this expenditure separately from expenditure on educational core services, which makes it harder to estimate the level of resources spent on core educational services.
Source: OECD (2018[8]) OECD Handbook for Internationally Comparative Education Statistics 2018: Concepts, Standards, Definitions and Classifications, https://doi.org/10.1787/9789264279889-en.
Understanding costs in higher education is important for decision making
As argued in Chapter 1, understanding the costs of different activities is important for the people who run higher education institutions and those who are responsible for supervising and steering higher education systems. For institutional leaders and department heads within higher education institutions, information on the cost of the different activities undertaken in their organisations is useful for effective planning and resource management (Anguiano et al., 2017[13]). Governments have an interest in understanding the costs of delivering a societally valuable volume and range of education and research in different fields to ensure there is a broad alignment between these costs and the income higher education institutions are able generate from public and private sources.
If, for example, income is too low to pay for well-qualified and able academic staff, appropriate facilities and equipment, and adequate guidance and support to students, there are serious risks for learning and research quality and student outcomes. Under-resourcing of fields of study that are important for national priorities could lead to undersupply of skilled graduates in these fields (Connew, Dickson and Smart, 2015[14]). Equally, however, public and private funders of higher education – like funders of other public services or investment projects – have an interest in ensuring scarce resources are used efficiently. Understanding what different activities and outputs cost to deliver and the factors influencing these costs is necessary to be able to make informed decisions about the efficiency of resource use.
Unfortunately, in addition to the complexity and heterogeneity of higher education institutions discussed earlier in the chapter, two further factors complicate the task of interpreting costs in higher education and making judgements about efficiency:
1. Information on the quality of outputs produced by higher education is, at best, imperfect and, at worst, entirely absent. This makes it difficult to judge whether, for example, higher costs in one institution compared to another are justified by higher quality activities and outputs. Objective and comparable measures of student learning outcomes are rarely available – and generally incomplete – and established research metrics are open to challenge. More diffuse or long-term outcomes from activities such as fundamental research cannot be captured by established quantitative measures. It is not impossible to make judgements about the quality of higher education – by combining indicators such as student satisfaction, graduate employment outcomes or research impact, for example – but it is harder than in many other sectors, including public services such as healthcare.
2. As most higher education institutions in OECD countries are non-profit organisations operating outside a market system, the costs of specific activities tend to reflect the level of resources available, rather than price levels established through market forces. Higher education institutions generally spend the income they receive (Usher and Balfour, 2024[15]) and studies have found a circularity between observed costs and institutional income levels, which is stronger than in market-based sectors (Deloitte Access Economics, 2016[16]). In sectors where market conditions predominate, consumers make judgements on whether the price for a given product or service is worth paying considering its characteristics. In higher education, this principle operates in the private higher education sector in some countries. However, in the publicly regulated and subsidised higher education institutions that account for most student enrolment in OECD countries, it is higher education funders that need to decide if they are paying the right price – a decision complicated by the limits of information on quality mentioned above.
In addition to these present-day challenges in identifying reasonable costs for given activities in higher education, technological developments offer a real possibility of disrupting current models of teaching and learning and research in higher education, with potentially significant impacts on costs and efficiency in the future. Artificial intelligence will not only impact demand for graduate skills, as discussed in Chapter 1, but also the opportunities to exploit technology to optimise educational processes. As illustrated in the example in Box 2.2, higher education institutions generally face two main options in the face of substantial increases enrolment. They can increase student-to-staff ratios at the risk of decreasing quality or deploy additional staff to maintain quality with attendant increases in costs. The question is whether AI will make it possible to achieve efficiency gains – delivering the same or better quality for the lower unit costs – in future.
Despite all these uncertainties, some OECD higher education systems have taken steps to collect and analyse more accurate information on the costs associated with different higher education activities and understand the factors influencing these costs. However, before turning to the results of these efforts, it is helpful to consider the limited internationally comparable information that exists on what it costs to run higher education institutions.
If enrolment in a higher education programme increases by 50%, but staffing remains constant and class sizes are also allowed to increase by 50%, the unit cost of educating each student will fall. Assuming broadly constant staff costs, the total cost of delivering the programme is likely to increase only slightly and the marginal cost of educating each additional student will likely be modest. Some cost increases driven by increased student numbers – such as registration activities and assessing a larger number of students – would be almost inevitable. However, in the absence of changes to pedagogical approach – the “production technologies” of the educational process – the impact of the increased class sizes on the student experience and quality of education will likely be negative. Teaching staff will have less time to devote to each student and may have to reduce labour-intensive activities, such as supervision of practical exercises or marking assignments.
If the increase in student enrolment is accompanied by deployment of additional teaching staff and teaching space to maintain student-to-staff ratios and learning conditions with the same educational processes, per-student unit costs are more likely to remain constant or increase, with the marginal cost of educating each additional student dependent on the efficiency with which additional staff and teaching space are deployed. In this second scenario, the total costs of delivering the programme will increase, but so will the likelihood that the student experience and educational quality can be maintained. A third scenario could see the increase in enrolment accompanied by a re-design of the pedagogical approach to exploit technology and optimise the use of instructors’ time to maintain educational quality and student experience, while reducing unit costs. As discussed later in this chapter, it is yet to be determined whether such productivity increases – with constant or improved educational quality – can be achieved in practice.
International data provide an imperfect measure of how much each student costs
The most used metric for how much higher education “costs” is expenditure per full-time equivalent student. While this is a metric frequently used in international analyses of school education, there are two main limitations to using this metric for analysis of higher education:
1. First, as discussed above, higher education institutions, unlike schools, do not just educate students but typically undertake research and perform various forms of service to society. While there is a clear logic in analysing the amount of money spent per student on educational activities to provide a measure of the resources invested for each person educated, the rationale for considering research or service spending per student is less obvious. In practice, student numbers provide a useful proxy for the size of tertiary education institutions, so research or service spending per student provide an indicator of the intensity of investment in these areas relative to the size of the institution or the higher education system.
2. Second, the difficulty of reliably disentangling time and money spent on the different activities within higher education institutions makes it hard to isolate the resources invested in educating students – and thus the cost of educating them.
Notwithstanding these challenges, Figure 2.3 shows expenditure per full-time-equivalent student on publicly funded tertiary education institutions for 39 OECD member and accession countries, where possible broken down by the expenditure categories (research, ancillary services and core spending) already illustrated in Figure 2.2. In 2022, the average level of spending per FTE student on tertiary education institutions, after spending in each country had been corrected for purchasing power parity (PPP) was USD 20 204. Adjusted rates of per-student expenditure were over USD 25 000 in Luxembourg, Japan, the United States, the United Kingdom, Sweden, Norway, Austria and Denmark.
A second group of comparatively high-spending systems clusters 12 countries around the OECD average, which spend between USD 18 000 (Slovenia) and USD 24 600 (Australia) per FTE student. A further 11 countries spend between USD 12 000 (Croatia) and USD 18 000 (Estonia) and seven countries spend below USD 12 000 per FTE student in PPP terms.
Expenditure on public and government-dependent tertiary education institutions per full-time-equivalent (FTE) student by type of activity, 2022
Note: (1) Values are given in United States dollars converted for Purchasing power Parity (PPP) (2) OECD figures are the unweighted averages of the 30 OECD member countries that report, at a minimum, data on direct expenditure and R&D. (3) No data are available for Colombia and Switzerland. (4) The reference year for the United States is 2021.
Source: OECD (2025[9]), Full dataset - Indicators, source, destination and nature of expenditure on education (dataset), https://data-explorer.oecd.org/s/33r (accessed on 19 August 2025).
Of the high-spending countries, Japan and the United States both have substantial independent private higher education sectors, meaning that the data for public and government-dependent institutions presented only provides partial picture of per-student expenditure in the higher education sector overall. The high level of spending per student in Luxembourg is partly explained by the unusual composition of the enrolment at the University of Luxembourg, the dominant provider in the country. Almost half of all tertiary education students in Luxembourg are enrolled at a master’s or doctoral level, reflecting institutions’ relative focus on research and cost-intensive postgraduate education.
The UOE data collection only provides a breakdown of expenditure on higher education institutions and enrolment by the legal and financial status of higher education institutions (public, government-dependent, independent). For the purposes of higher education policymaking, it is often more important to understand differences in costs and expenditure between types or profiles of higher education provider. The internationally comparable data that come nearest to this comes from the European Commission-sponsored European Tertiary Education Register (ETER), which provides both revenue and expenditure data for individual higher education institutions. In contrast to the UOE data collection, ETER data focuses on the institutional perspective and incorporates revenue and expenditure data separately. Where such data are available for a sufficient proportion of institutions, it makes it possible to compare expenditure between different categories of institution in each system.
Figure 2.4 shows that per-student expenditure by universities is systematically higher than that of non-university institutions in the six example systems covered, all of which have formal distinctions between universities and non-university institutions. In the Flemish Community of Belgium, Denmark, Finland, the Netherlands and Lithuania, expenditure per FTE student is more than twice as high in universities as in universities of applied sciences. Although expenditure data by activity type are not available from ETER, these patterns are largely driven by substantially higher research activity in universities and confirm previous OECD analysis based on national data (OECD, 2019[17]). The smaller difference between expenditure per student between universities and polytechnic institutes in Portugal is likely explained by the absence of dedicated research grants for universities in Portugal’s public funding model, which contrasts with the situation in the other systems shown here (see Chapter 3). Additionally, higher education research centres in Portugal are organised as separate legal entities, outside of universities, meaning a proportion of research expenditure is not captured in university balance sheets (OECD, 2022[18]).
Total expenditure per full-time equivalent student in euros converted for purchasing power parity, 2022
Note: (1) Values are given in euros converted for Purchasing power Parity (PPP) (2) Student Full-Time Equivalent (FTE) estimates were derived using country-specific multipliers calculated from 2022 UOE data. In cases where part-time/full-time breakdowns were not reported (e.g. ISCED 8 in Belgium and the Netherlands; all levels in Denmark), all students were assumed to be full-time. (3) Only institutions with available data were included meaning data coverage varies by country and by institutional type. A total of 175 institutions are included across the six countries covered (Belgium, Denmark, Finland, Lithuania, Netherlands, and Portugal). (4) Ireland has not been included here because of limited availability of financial data for the Irish institutions included in ETER and changes to the binary system in Ireland since the last ETER data collection, which make comparison of sub-sectors less relevant.
Source: European Commission / EACEA (2025[19]), The European Higher Education Sector Observatory - European Tertiary Education Register data collection (dataset), https://eter-project.com/data/data-for-download-and-visualisations/database/ (accessed on 25 August 2025).
More fine-grained data on the cost of activities exist in some OECD systems
Over the last decade, some OECD higher education systems have adopted system-wide protocols for activity-based costing (ABC). Such protocols make it possible to identify the direct costs of different types of activity in operational units – such as departments, faculties or research centres – and to attribute indirect (overhead) costs to these activities. In some European OECD jurisdictions, the adoption and development of ABC accounting systems was driven largely by the financial reporting requirements of competitive research funding programmes, in particular the European Union’s research and development framework programmes – most recently branded as Horizon Europe. Such funding programmes require beneficiaries to maintain records of direct project costs, such as the salaries of staff engaged in funded projects and be able to demonstrate and justify eligible overhead costs related to premises and central services.
The United Kingdom was the first European and OECD country to introduce a system-wide cost-accounting model for universities – the Transparent Approach to Costing (TRAC) – in 1999. The basic process for attributing costs to activities through the TRAC system is illustrated in Figure 2.5. The process involves taking institutional expenditure information from consolidated financial statements, adding “sustainability adjustments” to these reported costs to give a better idea of the funds need to cover the full sustainable cost of delivery and then allocating costs to activities. Costs are allocated to activities directly in a bottom-up process where this is possible – for example when a staff member or specific item is deployed exclusively for one activity – or in a more indirect, top-down fashion by applying “cost drivers” that make assumptions about the allocation of staff time allocation and floor space between different activities (TRAC Development Group, 2009[20]).
The Transparent Approach to Costing (TRAC) used in the United Kingdom’s higher education sector
Source: TRAC Development Group (2015[21]) TRAC - A Guide for Senior Managers and Governing Body Members: The Transparent Approach to Costing for UK Higher Education Institutions, https://www.trac.ac.uk/wp-content/uploads/2018/07/TRAC-A-guide-for-Senior-Managers-and-Governing-Body-members.pdf (accessed on 8 July 2025).
The TRAC system in the United Kingdom resulted in part from pressure from within the university sector itself to demonstrate the costs of delivering project-based funding and to make the case for higher funding rates from the national research councils. Debates continue in many OECD jurisdictions, not just in Europe, over the level of indirect costs (overhead) to which external research funders should contribute. There is widespread evidence from the United States, European countries and Australia that external research funding tends to cover only a fraction of the full costs associated with research projects, leading to research being cross-subsidised by revenue from teaching activities (UKRI, 2025[12]; Norton, 2015[22]). The higher education sector in Europe has argued for European and national research funding programmes to contribute a higher proportion of the full economic cost of research activity (EUA, 2018[23]).
In addition to demonstrating the cost of research activity, the UK cost-accounting system was further developed to identify the cost of publicly funded teaching overall and in 45 different subject fields. The TRAC for Teaching (TRAC(T)) methodology was developed to inform the level of subsidy and regulated tuition fees for publicly funded study places for UK and European Union students (DfE, 2019[24]), but concerns about the design of TRAC(T) (KPMG, 2021[25]) led to the system being officially discontinued in 2021.
More broadly, as summarised in Table 2.1, research undertaken for the OECD Resourcing Higher Education Project in 2020, found that system-wide approaches to cost accounting in higher education had been introduced or tested primarily in English-speaking and Nordic countries. Confirming this pattern, Denmark also introduced a system-wide, activity-based cost-accounting system – the fælles kontoplan or “common chart of accounts” – in 2020 (OECD, 2021[26]).
Note: 1. Australia conducted pilot exercises with fewer institutions in 2011 and 2016.
Source: OECD (2022[27]) Resourcing higher education in Ireland: Funding higher education institutions, https://doi.org/10.1787/67dd76e0-en.
Evidence on the impact of new activity-based costing models on internal resource allocation and accounting practices within HEIs and on public policy is patchy. In some cases, introduction of new ABC models is reported to have had an impact on the way higher education institutions operate. In Sweden for example, the introduction of the SUHF model led to most institutional income being transferred to departments, which then pay a transparent unit overhead surcharge for each unit of their direct salary and operating costs. In Finland, the use of the ABC model – which is designed primarily as a tool to support research funding – is reported to have increased cost awareness among staff and made the cost implications of engaging in externally funded projects more transparent (EUA, 2018[23]).
In Ireland, the Full Economic Cost system, which originally emerged from a partnership between the Irish University Association (IUA) and the Higher Education Authority (HEA), is reported to have increased planning and management of capacity and also to have led to greater cost awareness among staff (Estermann, Kupriyanova and Casey, 2018[28]). As in the United Kingdom, but to a far greater extent than in many OECD systems, the existing cost models in Ireland have also been used to place information on the costs at the centre of debates and policy making concerning the future of higher education financing (Box 2.3).
Drawing inspiration from the Transparent Approach to Costing (TRAC) used in UK universities, Ireland’s universities launched a joint project in 2007 to develop and implement their own system of Full Economic Costing (FEC). As a core part of the system, academic staff in universities report annually on their time use using a set of nine commonly defined types of activity referred to as Academic Activity Profiles (APPs)a. Following “reasonableness” checks by management, this allows direct staff costs (salaries etc.) to be assigned to each of the nine activity categories in each academic unit. The main steps in the cost allocation process for universities are then:
Overhead (indirect) costs, such as central ICT services, premises costs and central administration costs, identified in accounting systems and subject to some initial adjustments, are allocated to each academic unit using common cost drivers (allocation factors). These cost drivers include the surface area of different categories of space (laboratory vs. classroom or office, etc.) and numbers of FTE staff and students for shared ICT costs.
Within each academic unit, these overhead costs are then apportioned to cost pools for each of the nine activity types, using different cost drivers to differentiate appropriately between overhead used for teaching, research or other activities (or two or more of these).
The direct and indirect costs for administration and management (AAP 9) are then redistributed among the other eight activity types, based on staff costs in each activity type.
The direct and indirect costs of internally funded research and clinical services (AAP 5, 6 & 8) are allocated to the three levels of teaching (AAP 1, 2 & 3), based on FTE students.
Full economic costs are calculated for a) student FTEs by subject category and level of instruction; b) the overhead for externally funded research; c) overhead for other income-generating activity.
Note: a. These are: 1. Teaching (Undergraduate); 2. Teaching (Postgraduate); 3. Teaching (Postgraduate Research); 4. Research with External Sponsor; 5. Research No External Sponsor but with an output; 6. Other Research & Scholarly Activity; 7. Other Income-generating Activities; 8. Clinical Services; 9. Administration and Management.
Source: HEA (2017[29]) Working Paper 6: Cost Drivers and the Costing System Underpinning Higher Education, https://hea.ie/assets/uploads/2017/06/HEA-RFAM-Working-Paper-6-Costs-of-Higher-Education-Provision-06217.pdf (accessed on 27 July 2025).
More detailed data on costs has permitted analysis of cost drivers in higher education
The availability of more detailed data on costs in some higher education systems has made it possible to analyse costs in the sector and seek to better understand the drivers of costs and cost differentials between different higher education providers. In the broadest terms, “cost drivers” are factors that cause a change in the cost of a particular activity. In higher education, the number of students or the volume of research projects are key examples of drivers of the total cost of operating higher education institutions, as a higher volume of activity requires typically more staff and more space.
A growing body of international research has also examined the factors that influence unit costs in higher education. These analyses typically take observed unit cost – such as the cost by full-time-equivalent student per year, per credit or per module in a specific field of education – and use statistical techniques to assess the influence of different factors on these unit costs. Research has tended to focus on explaining differences in the cost of instruction between fields of study and between higher education institutions, and on changes in unit costs over time.
International research consistently finds that the ratio of students to teaching staff is the primary driver of unit costs in instruction in higher education. Based on analysis of data on the direct costs of instruction in different disciplines in US universities, Hemelt et al. (2021[30]) identify student-to-staff ratios as the largest determinant of unit cost. This is followed in importance by two further staff-related factors: the number of hours each staff member teaches (teaching load) and the proportion of tenured (as opposed to temporary or hourly paid) faculty involved in programmes. They also find that non-personnel expenses drive costs for sciences with laboratory components, albeit with less influence than staff-related factors, but in other fields explain relatively little of the cost differences observed. The study also identifies different trade-offs applied by institutions and departments in different fields. For example, in US universities, some fields, like economics, offset high faculty wages with large classes, resulting in unit costs that are comparable to English, despite higher faculty pay. Others, like physics, partially offset higher faculty salaries with heavier faculty workloads.
The first in a series of studies into higher education costs in Australia on behalf of the Department of Education, Skills and Employment by Deloitte Access Economics (2016[16]) and a 2019 study by KMPG LLP (2019[31]) in England also identified student-to-staff ratios and other staff-related factors as the main drivers of unit cost differences between fields of study and institutions. The Australian study also found the size of institutions to be weakly correlated to lower costs, suggesting some scale efficiencies in larger institutions. However, the authors note that smaller institutions also tend to have smaller class sizes, making it difficult to establish the specific contribution of institutional size to costs. The study also identified a weak positive correlation between universities located outside major urban centres and higher unit costs, which it attributes in part to higher proportions on non-traditional students in such institutions. A 2022 report on costs in Australian higher education by the same authors found that greater research intensity or focus within a field or institution may drive higher costs in teaching, as more research intensive institutions and departments tend to employ more senior – and thus expensive – faculty with both teaching and research roles (Deloitte Access Economics, 2022[32]).
Table 2.2 provides an overview of the main factors that available international literature finds to influence unit costs in instruction in higher education. For each unit cost driver, the table provides a general indication of its influence on total costs and summarises the main mechanisms through which the driver is assumed to influence direct departmental costs and indirect costs (overhead) in the institution. In cost-accounting systems, the term “cost driver” is frequently used to describe the factors used to allocate indirect costs to activities.
Source: OECD (2022[27]) Resourcing higher education in Ireland: Funding higher education institutions, https://doi.org/10.1787/67dd76e0-en.
Comparing cost levels internationally is challenging
Data on the observed costs of different activities and information on the main drivers of cost in higher education make it possible to identify the costs of different activities and explain and understand the main factors affecting these differences. Table 2.3 presents the results of cost analysis by field of study for Ireland, the United Kingdom and Australia using historical values presented in a report prepared as part of the OECD Resourcing Higher Education Project (OECD, 2022[27]). These data cover the same academic years in the three countries and shows the average unit cost for full-time equivalent undergraduate students in selected subject fields and the ratio between these costs for each system.
Unit costs per full-time equivalent undergraduate student in 2016 USD adjusted for purchasing power parity (PPP)
Notes: (1) Cost indicated are averages for the subject groupings to which the indicated field was assigned in each costing study. These subject groupings are not consistent across the three studies, meaning these average figures should be interpreted with caution.
(2) The “weights” values show the ratio of costs for each field to the field to the lowest-cost field in each system.
Source: OECD (2022[27]) Resourcing higher education in Ireland: Funding higher education institutions, https://doi.org/10.1787/67dd76e0-en.
The data in Table 2.3 are not fully comparable but do illustrate the cost differentials discussed above between classroom-based subjects, such as history, with the lowest costs, programmes with some laboratory component, including biological science and engineering and the most expensive fields: medicine, dental studies and veterinary science. Attempts such as this to compare costs and cost differentials across higher education systems are, however, hampered greatly by differing classifications and clustering of subject fields and differences in the underlying cost-accounting models, even though the models used in the three systems shown here all work by attributing overhead costs to individual departments and programmes. Moreover, the data used in Table 2.3 are not regularly updated and published, even for the three countries shown, making it challenging to examine costs across fields over time.
Economic theory posits that the relative cost of higher education will increase over time
The first report produced as part of the OECD Resourcing Higher Education Project argued that higher education has a “chronic difficulty in identifying ways to boost the productivity with which it carries out its missions” (OECD, 2020[3]). Education is one of the labour-intensive sectors that is assumed by economists to suffer from a tendency for costs to rise faster than general inflation due to stagnant productivity (see Box 2.4). In essence, the theory goes that the core activities of the educational process – including in higher education – cannot easily be accelerated or automated through the deployment of technology, without sacrificing quality. The presence of teachers and their interaction with students are fundamental to the educational process and difficult to replace. As noted in Box 2.4, similar patterns are observed in the parts of the health sector which depend on human interaction, such as long-term care, although Baumol effects are less pronounced in areas such as surgery, where automation can more easily be applied to enhance productivity.
Baumol posited that some labour-intensive sectors of the economy are “non-progressive”, meaning that they do not benefit from productivity improvements driven by technological advancements as much as other, “progressive”, sectors. In “non-progressive” sectors, including health and education, new technologies cannot easily replace human labour. The Baumol effect states that as productivity and wages rise together in progressive sectors of the economy, the health and education sectors – being non-progressive – will experience only wage increases to keep up with the rest of the economy, thus driving increases in unit costs. Recent literature has shown that the Baumol effect – as captured by the excess wage growth over productivity growth in the total economy – is significant in explaining health spending trends some parts of the health care system, such as long‐term care.
Source: Baumol (1967[33]) Macroeconomics of Unbalanced Growth: The Anatomy of Urban Crisis, http://www.jstor.org/stable/1812111 (accessed on 21 August 2025); Lorenzoni et al (2019[34]) Health Spending Projections to 2030: New results based on a revised OECD methodology, https://doi.org/10.1787/5667f23d-en
The implication of the difficulty of enhancing productivity in the higher education sector is that costs rise relative to other sectors of the economy and consumers and societies must pay proportionally more to secure higher education services.
Real-world evidence shows a nuanced picture
Data on spending per full-time-equivalent student over time can provide a broad indication of whether and how the Baumol effect has influenced costs in higher education. At the time of writing, international data on spending on tertiary education institutions per FTE student are available from 2013, when the new ISCED classification was introduced, to 2022, the most recent date for which comparable financial data have been reported. Figure 2.6 shows the change in expenditure per FTE student in constant prices (using as the reference year 2020) between 2013 and 2022 for the 36 OECD member and candidate countries with available data.
Average total expenditure per FTE student on tertiary education institutions overall increased by just under 13% in real terms in the decade between 2013 and 2022, while public expenditure increased by 13.4%. In public higher education institutions, average total spending per FTE student increased by almost 15% and public spending by almost 16% in real terms over the same period. The largest real-terms increases in spending per student were observed in Romania, Costa Rica, Slovenia, the Slovak Republic, Hungary and Bulgaria. All these countries also saw real terms increases of over 50% in public expenditure per student, as did Estonia and Poland. In contrast, per-student expenditure declined in real terms by over 5% in eight countries: Portugal, Italy, Sweden, Canada, Türkiye, Israel, Finland and Mexico.
The evidence from international per-student spending data provides a more nuanced picture of increasing unit costs in higher education. Although the increases in average spending per students in OECD and accession countries over the decade to 2022 tend to lend weight to the theory of a Baumol effect in higher education, it is notable that this increase is driven to a significant extent by substantial increases in Central and Eastern European countries with historically low levels of spending per student. It is plausible that a proportion of the changes in spending observed thus related to a gradual convergence of spending levels in these countries with those in other OECD countries, as economic growth made this possible, rather than the influence of stagnating productivity.
Percentage change in total and public spending per FTE student in constant prices (2020)
Note: (1) OECD and EU25 figures are calculated as unweighted averages of the countries shown from the respective organisations with available data. (2) Comparable data are not available for Colombia, Korea, Switzerland, Greece, and the Netherlands. (3) The reference year for the United States is 2021.
Source: OECD (2025[35]), Expenditure on educational institutions per full-time equivalent student (dataset), https://data-explorer.oecd.org/s/333 (accessed on 06 August 2025).
Moreover, the real-terms decline in spending per student observed in a total of 10 of the 35 countries illustrates that the Baumol effect is not inevitable or universal. It is not possible to judge based on the high-level data whether these declines in spending reflect improvements in productivity – whereby students are educated, and other outputs are delivered to the same level of quality for lower cost – or cost-cutting potentially associated with a decrease in quality, such as larger class sizes or reduced student-teacher interaction. As highlighted in the discussion of cost drivers above, higher education institutions can influence costs across their institutions or in specific programmes or departments by increasing the workload of teaching staff and by deploying lower cost temporary or hourly paid teaching staff, rather than tenured faculty. There is widespread evidence of the increased use of part-time and hourly paid contingent faculty in many OECD higher education systems (OECD, 2024[36]), although more detailed research on each system would be required to understand the extent to which this is one of the drivers of the falling per-student expenditure seen in the data presented in Figure 2.6.
The United States is among the countries that have seen a real-terms increase in per-student spending over the decade to 2022, albeit at the same rate as the average of OECD countries. Previous research in the United States suggests that the observed cost increases in per-student spending observed in US higher education institutions have resulted from both Baumol effects – a combination of higher staff costs and stable productivity – as well as increases in spending on student services (Archibald and Feldman, 2018[37]; Hemelt et al., 2021[30]). Institutions in the United States – and particularly more prestigious public and private four-year colleges – have considerable flexibility to increase student fees to raise additional revenue. This contextual and regulatory factor also influences the pattern in per-student funding and likely also partly explains why per-student spending in the United States has on average increased at a faster rate than in many European systems where institutions depend to a greater extent on public funding and tuition fees tend to be more strictly regulated.
Accurate data on spending and costs alone do not allow users to make judgements about whether the observed costs are appropriate to achieve a societally desirable level and quality of outputs. In the United States, for example, some degree of consensus emerges from the literature that a large proportion of the average cost increases observed in US universities to date can plausibly be attributed to legitimate attempts to maintain quality – by offering competitive salaries to talented academics, for example – rather than profligacy and inefficiency (Archibald and Feldman, 2018[37]; Hemelt et al., 2021[30]).
Further work is required to improve understanding of costs and cost management
The review of existing data on costs and spending on higher education undertaken for the OECD Resourcing Higher Education Project highlighted several areas where more progress is required to improve understanding of costs and cost management in higher education. In broad terms, three areas for further work emerge:
1. Improving the availability of reliable data on expenditure and costs in higher education across more higher education systems.
2. Refining methods and sharing practices on cost measurement to balance usefulness with administrative burden for institutions and staff.
3. Understanding the impact of the artificial intelligence on teaching and learning and the associated costs.
The experience of higher education systems that have implemented cost-accounting systems to improvement measurement and understanding of the costs of different activities in higher education is broadly positive. Although the implementation of cost-accounting approaches creates an administrative burden for higher education institutions, staff and officials at a system level, the data collected increase the transparency of the system, understanding of costs and cost drivers and aid decision making (KPMG, 2021[25]). Such cost-accounting models are most useful in mature systems with medium-to-high levels of per-student investment. In higher education systems facing acute funding challenges, implementing cost-accounting models will inevitably be a second-order concern and will in any case not generate meaningful results. It is notably that efforts to measure costs in higher education to date have been concentrated primarily in English-speaking and Nordic countries. There is scope for OECD countries to explore collectively if and how data on costs in higher education could be improved across more higher education systems.
The most recent data collection and analysis in the Transparency in Higher Education Expenditure project in Australia highlighted persistent issues in allocating staff time to activities, appropriately accounting for depreciation of assets and investment needs and finding the optimal level of disaggregation for costs (Deloitte Access Economics, 2022[32]). These challenges are common to other systems (KPMG, 2021[25]). The technical challenges come on top the fundamental issue highlighted earlier, that observed costs in higher education reflect historical patterns of resource allocation rather than, necessarily, actual resourcing requirements. A key issue in collecting data on costs is achieving the right balance between accuracy and disaggregation – which typically require bottom-up data on staff time use and space allocation – and administrative burden for staff and institutions. Again, there appears to be scope for higher education systems to learn from the experience of others in these areas.
As discussed in Chapter 1, artificial intelligence will likely have a profound impact on the operations of higher education institutions and the nature of many graduate jobs. However, the extent to which AI can and should be deployed in teaching and learning and the impact of this on the productivity of teaching staff are unclear at the time of writing. Some existing research has analysed the effects of deploying technology in instruction in higher education, although evidence on the relationship between use of online and distance learning and unit costs is mixed (Xu and Xu, 2019[38]). Hemelt et al. (Hemelt et al., 2021[30]) find online and blended programmes are associated with a modest reduction in unit cost, while Chirikov et al. (2020[39]) use a randomised experiment to show how blended undergraduate science programmes can be designed to achieve acceptable student learning outcomes at substantially lower costs than in-person instruction. Some commentators are sceptical about the scope for AI to substantially displace human labour in the teaching and learning process, given the inherent importance of human interaction in learning (Appoldt and Gong, 2025[40]). The development of AI provides an important area for further research, not only on the effectiveness of deploying AI in the teaching and learning process, but also the cost implications of doing so.
References
[2] Aagaard, K., A. Kladakis and M. Nielsen (2020), “Concentration or dispersal of research funding?”, Quantitative Science Studies, Vol. 1/1, pp. 117-149, https://doi.org/10.1162/QSS_A_00002.
[13] Anguiano, M. et al. (2017), Optimizing Resource Allocation for Teaching: An experiment in activity-based costing in higher education, University of California Riverside, https://s22553.pcdn.co/wp-content/uploads/2017/05/UCR_ABC_Whitepaper.pdf (accessed on 27 July 2025).
[40] Appoldt, G. and Z. Gong (2025), Baumol’s cost disease: long-term economic implications where machines, UNESCO, https://www.unesco.org/en/articles/baumols-cost-disease-long-term-economic-implications-where-machines-cannot-replace-humans (accessed on 22 August 2025).
[37] Archibald, R. and D. Feldman (2018), Drivers of the Rising Price of a College Education, Midwestern Higher Education Compact, https://mhec.org/mhec_reports/drivers-of-the-rising-price-of-a-college-education/ (accessed on 27 July 2025).
[33] Baumol, W. (1967), “Macroeconomics of Unbalanced Growth: The Anatomy of Urban Crisis”, The American Economic Review, Vol. 57/3, pp. 415-426, https://www.jstor.org/stable/1812111 (accessed on 21 August 2025).
[39] Chirikov, I. et al. (2020), “Online education platforms scale college STEM instruction with equivalent learning outcomes at lower cost”, Science Advances, Vol. 6/15, https://doi.org/10.1126/sciadv.aay5324.
[14] Connew, S., M. Dickson and W. Smart (2015), A Comparison of Delivery Costs and Tertiary Education Funding by Field of Study - Results and Methodology, New Zealand Government, Wellington, https://www.educationcounts.govt.nz/__data/assets/pdf_file/0020/164054/A-Comparison-of-Delivery-Costs-and-Tertiary-Education-by-Field-of-Study.pdf (accessed on 27 July 2025).
[6] Daraio, A. and C. Daraio (2022), What ETER tells us about size distribution of Higher Education Institutions in Europe, European Tertiary Education Register, https://eter-project.com/wp-content/uploads/2022/02/ETER-brief-5.-What-ETER-tells-us-about-the-size-distribution-of-Higher-Education-Institutions-in-Europe.pdf (accessed on 26 July 2025).
[32] Deloitte Access Economics (2022), 2022 Transparency in Higher Education expenditure, Department of Education, Australian Government, https://www.education.gov.au/higher-education-publications/resources/2022-transparency-higher-education-expenditure-publication (accessed on 11 July 2025).
[16] Deloitte Access Economics (2016), “Cost of delivery of higher education”, No. ED16/034517, Australian Department of Education and Training, Canberra, https://www.education.gov.au/higher-education-reviews-and-consultations/resources/cost-delivery-higher-education (accessed on 27 July 2025).
[24] DfE (2019), Measuring the cost of provision using Transparent Approach to Costing data, Department for Education, London, https://assets.publishing.service.gov.uk/government/uploads/system/uploads/attachment_data/file/909417/Measuring_the_cost_of_provision_using_transparent_approach_to_costing_data.pdf (accessed on 14 July 2025).
[28] Estermann, T., V. Kupriyanova and M. Casey (2018), Efficiency, Effectiveness and Value for Money: Insights from Ireland and Other Countries, European University Association, Brussels, https://www.eua.eu/publications/reports/efficiency-effectiveness-and-value-for-money-insights-from-ireland-and-other-countries.html (accessed on 20 August 2025).
[23] EUA (2018), Accepting University Accounting Practices under Horizon Europe A compendium of national and institutional cases, European University Association, Brussels, https://eua.eu/downloads/publications/compendium%20of%20accounting%20practices_web.pdf (accessed on 27 July 2025).
[19] European Commission / EACEA (2025), The European Higher Education Sector Observatory - European Tertiary Education Register data collection (dataset), https://eter-project.com/data/data-for-download-and-visualisations/database/ (accessed on 25 August 2025).
[4] Golden, G., L. Troy and T. Weko (2021), “How are higher education systems in OECD countries resourced?: Evidence from an OECD Policy Survey”, OECD Education Working Papers, No. 259, OECD Publishing, Paris, https://doi.org/10.1787/0ac1fbad-en.
[29] HEA (2017), Review of the Allocation Model for Funding Higher Education Institutions: Working Paper 6: Cost Drivers and the Costing System Underpinning Higher Education Contents, Higher Education Authority, Dublin, https://hea.ie/assets/uploads/2017/06/HEA-RFAM-Working-Paper-6-Costs-of-Higher-Education-Provision-06217.pdf (accessed on 27 July 2025).
[30] Hemelt, S. et al. (2021), “Why is math cheaper than english? Understanding cost differences in higher education”, Journal of Labor Economics, Vol. 39/2, pp. 397-435, https://doi.org/10.1086/709535.
[25] KPMG (2021), Review of TRAC - Final Report, KPMG, London, https://www.officeforstudents.org.uk/media/78f19d2c-3363-46ba-be30-612c1de1a749/review-of-trac.pdf (accessed on 14 July 2025).
[31] KPMG LLP (2019), “Understanding costs of undergraduate provision in higher education provision”, Costing Study Report, UK Department for Education, London, https://www.gov.uk/government/publications/cost-of-undergraduate-higher-education-provision (accessed on 27 July 2025).
[34] Lorenzoni, L. et al. (2019), “Health Spending Projections to 2030: New results based on a revised OECD methodology”, OECD Health Working Papers, No. 110, OECD Publishing, Paris, https://doi.org/10.1787/5667f23d-en.
[22] Norton, A. (2015), The cash nexus: how teaching funds research in Australian universities, Grattan Institute, https://grattan.edu.au/wp-content/uploads/2015/10/831-Cash-nexus-report.pdf (accessed on 20 August 2025).
[35] OECD (2025), Expenditure on educational institutions per full-time equivalent student (dataset), https://data-explorer.oecd.org/s/333 (accessed on 6 August 2025).
[9] OECD (2025), Full dataset - Indicators, source, destination and nature of expenditure on education (dataset), https://data-explorer.oecd.org/s/33r (accessed on 19 August 2025).
[5] OECD (2025), “Policies for transnational collaboration in higher education”, OECD Education Policy Perspectives, No. 121, OECD Publishing, Paris, https://doi.org/10.1787/e07d8c0f-en.
[36] OECD (2024), “The state of academic careers in OECD countries: An evidence review”, OECD Education Policy Perspectives, No. 91, OECD Publishing, Paris, https://doi.org/10.1787/ea9d3108-en.
[27] OECD (2022), “Resourcing higher education in Ireland : Funding higher education institutions”, OECD Education Policy Perspectives, No. 51, OECD Publishing, Paris, https://doi.org/10.1787/67dd76e0-en. (accessed on 29 June 2022).
[18] OECD (2022), Resourcing Higher Education in Portugal, Higher Education, OECD Publishing, Paris, https://doi.org/10.1787/a91a175e-en.
[26] OECD (2021), “Resourcing higher education in Denmark: Thematic policy brief”, OECD Education Policy Perspectives, No. 49, OECD Publishing, Paris, https://doi.org/10.1787/5cc2d673-en.
[3] OECD (2020), Resourcing Higher Education: Challenges, Choices and Consequences, Higher Education, OECD Publishing, Paris, https://doi.org/10.1787/735e1f44-en.
[17] OECD (2019), Benchmarking Higher Education System Performance, Higher Education, OECD Publishing, Paris, https://doi.org/10.1787/be5514d7-en.
[8] OECD (2018), OECD Handbook for Internationally Comparative Education Statistics 2018: Concepts, Standards, Definitions and Classifications, OECD Publishing, Paris, https://doi.org/10.1787/9789264279889-en.
[10] OECD (2015), Frascati Manual 2015: Guidelines for Collecting and Reporting Data on Research and Experimental Development, The Measurement of Scientific, Technological and Innovation Activities, OECD Publishing, Paris, https://doi.org/10.1787/9789264239012-en.
[7] OECD (2007), “On the Edge: Securing a Sustainable Future for Higher Education”, OECD Education Working Papers, No. 7, OECD Publishing, Paris, https://doi.org/10.1787/220180871707.
[21] TRAC Development Group (2015), TRAC - A Guide for Senior Managers and Governing Body Members: The Transparent Approach to Costing for UK Higher Education Institutions, https://www.trac.ac.uk/wp-content/uploads/2018/07/TRAC-A-guide-for-Senior-Managers-and-Governing-Body-members.pdf (accessed on 8 July 2025).
[20] TRAC Development Group (2009), Policy overview of the financial management information needs of higher education, and the role of TRAC, https://www.trac.ac.uk/wp-content/uploads/2018/07/FSSG-TRAC-policy-overview.pdf (accessed on 8 July 2025).
[12] UKRI (2025), Research financial sustainability: insights paper 2025, UK Research and Innovation, https://www.ukri.org/publications/research-financial-sustainability-insights-2025/research-financial-sustainability-insights-paper-2025/ (accessed on 27 July 2025).
[11] UKRI (2024), UKRI Data Pack on Research Financial Sustainability, UK Research and Innovation, Swindon, https://www.ukri.org/wp-content/uploads/2024/09/UKRI-050924-RFS-FinancialFundingFlowsSankeybriefingmaterials-commsversionv3.pdf (accessed on 27 July 2025).
[15] Usher, A. and J. Balfour (2024), The State of Postsecondary Education in Canada 2024, Higher Education Strategy Associates, Toronto, http://www.higheredstrategy.com (accessed on 6 July 2025).
[1] Wagner-Schuster, D. (2022), What ETER tells us about subject specialisation in European higher education, European Tertiary Education Register, https://eter-project.com/wp-content/uploads/2022/02/ETER-brief-1.-What-ETER-tells-us-about-subject-specialization-in-European-higher-education.pdf (accessed on 23 July 2025).
[38] Xu, D. and Y. Xu (2019), The Promises and Limits of Online Higher Education - Understanding how distance education affects access, cost and quality, American Enterprise Institute (AEI), Washington, D.C., https://www.aei.org/wp-content/uploads/2019/03/The-Promises-and-Limits-of-Online-Higher-Education.pdf (accessed on 8 April 2020).
← 1. International data on educational expenditure become available with an approximately two-year time lag compared to the most up-to-date data in the national systems with the most advanced system-level financial data collections. This reflects the varying speed at which financial data become available after institutional accounts are closed in different systems, differences in the timing of financial years and the time taken to collate and process data in an internationally comparable format.
← 2. Chapters 2 and 3 of this report primarily use data on public and government-dependent private tertiary education institutions to reflect the situation in the publicly subsidised institutions most directly affected by public funding decisions and public funding policies. In most countries, these categories of institution account for most student enrolment. However, in Poland, Mexico, Colombia, Hungary, Chile, Japan and Korea independent private institutions, which typically do not receive public subsidies for the core operating budget, account for between 30% and 80% of total student enrolment at tertiary level. In these countries, public and government-dependent private institutions are less representative of the tertiary education system as a whole.