The research system · living register
Structural Problems Register
The research system’s structural problems in one place, organised by the levers that can change them: what we know, how it looks in Denmark, what holds each problem in place and what is being tried.
Experiments and new institutions are useful. The problems below go deeper: they come from how funding, careers, power, assessment and publishing work together. This register keeps an ongoing list so they can be faced directly. It is organised by the seven levers used in A whole-system approach, and a problem appears under every lever that bears on it, so the counts below add up to more than the total. Each problem has a stable code. Feeds into and Fed by show how problems drive each other, and The other view gives the strongest counterargument. Held in place by links to the mechanisms that reproduce it, explained on Why problems persist. The status shows how well documented a problem is, especially in Denmark.
R01Project funding crowds out basic funding A growing share of university research runs on time-limited project grants while basic funding stays flat, which moves priority-setting from universities and researchers to external funders. Funding mixWell documented
Evidence
- Philanthropic money is also tied to specific projects, so independent research organisations must fundraise constantly and cannot redirect work when they learn something; Reinhardt proposes long-term public funding for high-performing research nonprofits as the equivalent of an “exit” (Spectech, October 2026).
- Sweden 2023: about 57% of university research revenue was external (UKÄ).
- Norway: basic funding’s share of university R&D costs fell from 65% in 2003 to just under 60% in 2023.
Denmark
- Basic research funding has stayed at about DKK 10bn a year in fixed prices since 2011 (DFiR).
- Private foundations’ share of university R&D rose from 7% (2007) to 18% (2021) (ministry answer, UFU 345).
- External funding has been the University of Copenhagen’s largest income source since 2020, rising from about DKK 3bn to 4.4bn in 2015–2025, or 38% of revenue (Uniavisen, 2026). The 2026–29 research agreement adds DKK 2bn in basic funding over four years.
- 79% of Danish researchers say they would change their research given two years of unconstrained funding (CFA survey reported by Forskerforum, September 2026).
Why it is structural
Basic funding is set politically with no link to the growth of external grants, foundation money grows with endowment returns, and universities cannot decline large grants without losing staff and standing.
Held in place by
The other view
Project funding brings competition, outside review and room for new priorities, and some argue that generous basic funding can shelter weak research environments.
Feeds into
Fed by
Being tried
Calls to raise basic funding (DFiR, Videnskabernes Selskab); the extra DKK 2bn in the 2026–29 agreement.
How we would know it is improving
Share of university research paid by basic funding, by university and field, each year.
R02External grants drain universities’ own money Funders have not paid the full indirect costs of what they fund, so each external grant draws on basic funding. Funding mixData and infrastructurePartly documented
Evidence
- UK universities recovered 65.8% of the full cost of research in 2023–24; the deficit in England and Northern Ireland was £5.4bn.
Denmark
- The University of Copenhagen estimates follow-on costs of DKK 65–75 per DKK 100 of external grants (board paper, 2022).
- In 2023 universities and six foundations agreed supplements of DKK 250,000 (wet) and 200,000 (dry) per project employee a year. It has not been evaluated.
- The ministry has no statistics on how much basic funding is tied up in external projects (2025).
Why it is structural
Overheads look to funders like money with no visible result, universities accept prestigious grants on any terms, and national statistics do not track the drain.
Held in place by
The other view
Foundations argue that paying full overheads would reduce the research they can fund, and universities are free to decline grants on poor terms.
Feeds into
Being tried
The 2023 overhead agreement; UK debate on full-cost funding.
How we would know it is improving
Share of basic funding tied up in co-financing, and the full-cost recovery rate.
R03Application burden and low success rates Low success rates mean researchers spend a large share of their time on proposals that are rejected. Funding mixAssessment and incentivesPartly documented
Evidence
- One Australian funding round of 3,727 proposals took an estimated 550 working years, and more time spent did not raise the chance of success (Herbert et al., BMJ Open 2013).
- Modelling suggests success rates below about 20% drive at least half of researchers away from federal funding (von Hippel & von Hippel 2015).
- An NSF proposal takes about 120 hours to write, and NSF’s three-year standard grants keep researchers on a fast “grant treadmill” (New Science, October 2026).
- Grant application volumes rose by an average of 57% across twelve funders from 2022 to 2025; Marie Skłodowska-Curie applications rose 64.6% from 2024 to 2025, and a fifth of proposals scored above 95% (Wilsdon, October 2026).
Denmark
- DFF 2024: 3,926 applications, 588 grants, a success rate of 14%. Applications have risen 87% since 2017.
- The latest national estimate of time spent on applications dates from 2007–2011.
Why it is structural
Each funder optimises its own call without bearing the system-wide cost, flat basic funding makes grants the only route to research time, and lower success rates drive more applications.
The other view
Competition selects strong proposals, and writing an application can sharpen a project even when it is rejected.
Being tried
Bill L 18 (2025) would allow caps on applications; partial lotteries at the Swiss NSF, the Volkswagen Foundation and the Villum Experiment.
How we would know it is improving
Success rates and estimated researcher time spent on applications, per funder.
R04Funding concentrates on a few Competitive funding concentrates on a small group of researchers and topics beyond what evidence on returns supports, and early luck compounds into lasting advantage. Funding mixAssessment and incentivesWell documented
Evidence
- Near-identical Dutch applicants either side of a funding line ended up €180,000 apart eight years later (Bol et al., PNAS 2018).
- A review of 92 studies finds flat or diminishing returns to grant size (Aagaard, Kladakis & Nielsen 2020).
- UK engineering researchers who just won an EPSRC grant went on to win about 0.8 more grants over five years than those who just missed (Baruffaldi, Jorzick & Lawson, Research Policy 2026).
Denmark
- About one in five researchers received nearly 90% of competitive funding from 15 Danish funders in 2004–2016 (Madsen & Aagaard 2020).
- 73% of Danish researchers prefer grants of DKK 3–10m (The Ideal Research Grant, 2019).
- CFA’s 2026 mapping covers about 18,000 grants worth DKK 83bn in 2018–2024.
Why it is structural
Excellence criteria reward track record, large foundations prefer large centre grants, and funders do not coordinate, so the same people win across funders.
The other view
Concentrating money builds strong environments that can compete internationally, and some fields need scale and expensive equipment.
Being tried
Smaller grants, caps on active grant holders (L 18), partial lotteries, the CFA mapping as a basis for coordination.
How we would know it is improving
Share of competitive funding received by the top 10% and 20% of researchers, across all funders combined.
R05Year-by-year politics and short horizons Much new public research money is decided year by year and earmarked to political themes, which makes long-term planning hard. Funding mixLaw and regulationPartly documented
Evidence
- Long, stable funding produces more novel work (Azoulay et al. 2011).
Denmark
- The research reserve (over DKK 6bn for 2026) was split into 37 one-year earmarks in 2025, which the Young Academy criticised.
- The Danish National Research Foundation, Denmark’s main source of ten-year funding, lacks capital: without a one-off state injection, the current round of new Centres of Excellence will be the last (Videnskabernes Selskab, October 2026).
- Education grants were cut 2% a year in 2016–2019. The 2026–29 agreement moves part of the money into a four-year frame.
Why it is structural
Appropriations are part of annual budget bargaining, visible themes win, and basic funding is the residual that can be held flat.
Held in place by
The other view
Annual reserves let parliament respond quickly to new priorities such as defence, health crises or green technology.
Feeds into
Fed by
Being tried
The multi-year 2026–29 research agreement; calls for a more open reserve process.
How we would know it is improving
Share of new public research money decided in multi-year frames.
R06Research waste Much research investment is lost through poorly chosen questions, weak design, non-publication and unusable reporting. Funding mixAssessment and incentivesDanish data thin
Evidence
- About 85% of biomedical research investment is avoidably wasted, by a compounding estimate (Chalmers & Glasziou, Lancet 2009).
Denmark
- No Danish estimate found.
Why it is structural
Funding and careers reward new studies over synthesis, and nobody bears the cost of unusable outputs.
Held in place by
The other view
Some loss is the normal cost of uncertainty in research, and the 85% estimate is a compounding calculation that is hard to verify.
Feeds into
Being tried
REWARD Alliance, reporting guidelines, evidence-based research.
How we would know it is improving
Share of studies and trials with results reported, and of new studies building on a systematic review.
Sources
Glasziou & Chalmers 2016R07A narrowing range of ideas As fields grow and incentives favour safe, citable work, attention concentrates on established ideas and the range of approaches may narrow. Funding mixAssessment and incentivesContested
Evidence
- Disruptiveness declined across 45 million papers and 3.9 million patents from 1945 to 2010 (Park, Leahey & Funk 2023); the finding is debated but held up in re-analysis.
- In large fields, canons ossify and new papers struggle for attention (Chu & Evans 2021).
Denmark
- No specific Danish evidence; linked to funding concentration (F4).
Why it is structural
Volume overloads attention, review favours consensus, and concentrated funding narrows approaches.
The other view
The decline in disruptiveness is debated, and part of it may reflect how fields grow and how citations are measured.
Feeds into
Fed by
Being tried
A wider range of funding instruments and institutions; plural indicators.
How we would know it is improving
Range of topics and approaches funded and published, read with care.
R40Business research rests on a few companies Research and development in Danish business is concentrated in a small number of large firms, which makes the whole system sensitive to their fortunes and leaves most firms outside research. Funding mixData and infrastructureDanish data thin
Evidence
- Concentrated business R&D is common in small open economies, where a handful of multinationals account for much of the spending.
Denmark
- The ten largest firms accounted for 46% of the growth in Danish business R&D, and only 18% of surveyed firms carry out R&D (Danmarks Statistik, 2016 figures).
- Leaders of Innovationsfonden, Danske Universiteter and DFiR say Denmark lacks a shared picture of how research becomes companies; DFiR is building linked data (2026).
Why it is structural
Tax rules, foundation ownership and global markets reward scale, while small firms lack the capacity to absorb research, and data on the paths from research to companies is missing.
The other view
Large firms create spillovers, partnerships and jobs that benefit universities and smaller firms.
Feeds into
Fed by
Being tried
Innovationsfonden schemes for small firms; DFiR’s linked-data project; industrial PhDs.
How we would know it is improving
Share of business R&D in the ten largest firms, and the share of firms doing R&D.
R41Grant formats are uniform Competitive funding comes almost entirely as one format, the project grant tied to a single lead researcher, a career stage, a host institution and a fixed timeline, which narrows how research can be organised. Funding mixAssessment and incentivesPartly documented
Evidence
- The US and other countries have “prematurely standardized” on a few ways to fund and organise research, chiefly committees allocating small grants to professors for incremental work, and rarely test alternatives (Tom Kalil, SCI PHI, October 2026).
- Funders that tried other formats, such as lotteries, anonymised review and golden tickets, found they reached different people and ideas (Volkswagen Foundation, Villum Experiment).
- US funding instruments are “often set by habit and comfort, rather than fit”; FAS design groups instead matched the instrument to the stage of a problem, from grants for discovery to investment for deployment (FAS, September 2026).
Denmark
- CFA’s mapping of 18,214 competitive grants worth DKK 83.1bn (2018–2024) finds diversity in topics and sums, while the format stays the same. 75% of surveyed researchers say current formats constrain how they would ideally work (Andersen & Schneider 2026).
- 79% would change their research given two years of unconstrained funding (Forskerforum, September 2026).
Why it is structural
Funders copy each other’s formats because they are familiar, easy to administer and accepted by review panels, and university careers are built around them.
The other view
A common format makes proposals comparable and review manageable, and researchers can combine grants to build what they need.
Being tried
Villum Experiment, partial lotteries, person-based and long-term grants, DFF’s small grants; a shared view of the format mix across funders.
How we would know it is improving
Share of competitive funding outside standard project grants, across all funders.
R08Precarity and the postdoc holding pattern A growing share of research is done on fixed-term contracts, and permanent posts have not kept pace with the number of PhDs and postdocs. Careers and employmentFunding mixLaw and regulationPartly documented
Evidence
- The OECD describes “a shrinking protected research elite and a large precarious research class that now represents the majority in most academic systems” (2021).
- Netherlands: 55% of academic staff on temporary contracts (Rathenau, 2024). Norway: 76% of grant-funded researchers temporary in 2024, against 5.5% of associate professors.
Denmark
- Postdocs at Danish universities rose 32% from 2018 to 2022 (Dansk Industri).
- Recruitment to fixed-term research posts rose from 2,865 (2008) to 4,004 (2022). The ministry says its registers cannot count how many researchers hold fixed-term posts today (2024).
- DM’s Forskerforum estimates two in three academic staff are on fixed-term contracts (union estimate).
Why it is structural
Project grants pay for staff only for the project’s duration, so fixed-term employment grows with external funding. PhD intake is not linked to permanent posts. Universities carry the cost of permanent hiring; researchers carry the cost of years in the queue.
Held in place by
The other view
Fixed-term posts give people a chance to try research before committing, support mobility, and let universities follow new fields.
Feeds into
Being tried
Tenure-track posts (job-structure orders 2019 and 2025); DTU’s policy of no fixed-term researchers; Norway’s binding plans against temporary employment.
How we would know it is improving
Number and share of researchers on fixed-term contracts, and median years from PhD to a permanent post.
R09The parenthood penalty and the gender gap at the top Women leave academic careers after having children far more often than men, and Denmark has the lowest share of women professors in the Nordics. Careers and employmentCulture and normsAssessment and incentivesLaw and regulationWell documented
Evidence
- A Danish register study of more than 13,000 researchers finds mothers 29% less likely to remain in academia eight years after becoming parents; mothers leaving explains about 70% of the gender gap in academic employment (LSE CEP DP 2160, as reported by CBS).
- Women are 24.7% of professors in Denmark, against 29.4% in the EU, 31% in Sweden, 31.9% in Finland and 34.2% in Norway (She Figures 2024).
Denmark
- At about 0.8 percentage points a year, parity among professors comes around 2055, and after 2080 in science and technical faculties (UFU bilag 67, 2024/25).
- DFF research project 1 grants, 2024: success rate 10% for women and 13% for men.
- In the fully blinded Villum Experiment (2,041 applications), a small gender gap in success largely disappears once the shortage of senior women among applicants is accounted for, which points to careers more than review (Madsen, Mongeon & Schneider, QSS 2025).
- A ministry taskforce made 24 recommendations in 2015; the rectors’ taskforce made eight in 2023, including centrally funded parental leave.
Why it is structural
The years of short contracts, mobility and output pressure coincide with family formation. Assessment counts output per year since the PhD, grants tied to a named person make leave risky, and committees drawn from a mostly male professoriate reproduce it.
Held in place by
The other view
The share of women professors is rising, and some who leave choose well-paid careers outside academia that Denmark offers.
Feeds into
Being tried
DFF’s gender-equality policy; Horizon Europe gender equality plans; the rectors’ 2023 recommendations; Norway’s BALANSE programme, which aims at “fixing the system”.
How we would know it is improving
Retention of mothers and fathers eight years after a first child; women’s share of professors and of new professor hires.
R10Social background and missing ethnic diversity Researchers come disproportionately from academic homes, and ethnic minorities are almost absent among senior and public-facing researchers. Denmark measures neither. Careers and employmentCulture and normsData and infrastructureDanish data thin
Evidence
- Nearly a quarter of US tenure-track faculty have a parent with a PhD (Morgan et al., Nature Human Behaviour 2022).
Denmark
- Social self-selection at entry to Danish research careers is strongest among Young Elite Researcher award winners (Andersen 2012).
- A DM analysis (2024) found five women and nobody from an ethnic minority among the 50 most-cited experts in Danish media.
- No register statistics exist on researchers’ ethnic background.
Why it is structural
Unstable early-career years and mobility demands favour people with family backing, tacit knowledge of academia passes through families, and nobody collects the data.
Held in place by
The other view
Children of academics may simply know more about research careers, and fair selection can still produce skewed backgrounds if earlier education is unequal.
Feeds into
Being tried
First-generation networks abroad. No Danish policy found.
How we would know it is improving
Researchers by parental education and background at each career stage.
R11Barriers for international researchers Danish research depends on internationally recruited staff, who face language, residence and career barriers that push many to leave. Careers and employmentLaw and regulationPartly documented
Evidence
- Sweden: internationally recruited researchers rose from 4,460 to 6,140 in 2014–2024, mostly in short career-development posts (UKÄ 2025).
Denmark
- 43% of researchers at Danish universities held non-Danish citizenship in 2023, 65% at ITU (Forskerforum).
- Three in four newly hired postdocs in 2022 were non-Danish citizens (Dansk Industri).
- Only one in three foreign researchers at Aarhus University felt they had equal access to permanent posts (university survey).
Why it is structural
Fixed-term research capacity is recruited internationally while permanent posts carry teaching duties in Danish, and residence permits tied to contracts make a gap between contracts an immigration risk.
The other view
International recruitment brings talent, networks and fresh perspectives that a small country cannot supply alone.
Feeds into
Fed by
Being tried
Language courses, international staff offices, the researcher tax scheme.
How we would know it is improving
Retention of international staff after five years, and their share of permanent posts.
R12Invisible roles Research depends on technicians, research engineers, data stewards and research managers, but these roles have weak career paths and little recognition. Careers and employmentAssessment and incentivesDanish data thin
Evidence
- The UK Technician Commitment (2017) has about 100 signatories; the EU has a competence framework for research managers (2025).
Denmark
- No Danish data found on the numbers, careers or turnover of technical and research-support staff.
- The job structure has no permanent non-PI research track beyond researcher and senior researcher.
Why it is structural
Grants pay for these roles as project costs, evaluation counts PI outputs, and academic and technical-administrative job structures are separate.
Held in place by
The other view
Funding support roles through projects keeps flexibility when research needs change.
Feeds into
Being tried
Technician Commitment; ERA Action 17 on research management; CRediT contributor roles; CoARA’s commitment to recognise diverse contributions.
How we would know it is improving
Numbers, contract types and career paths of technical and research-support staff.
R13Late permanence and mobility demands Careers require years of short contracts, often abroad, before a permanent post, which filters out people with care duties or less money. Careers and employmentAssessment and incentivesPartly documented
Evidence
- A study of about 16,000 researchers finds that moving abroad raises output for both sexes, but men gain more (NIFU, 2026).
Denmark
- Going from PhD to full professor takes almost three years longer than a decade earlier, and the chance of becoming associate professor within six years of the PhD fell for the 2001–2009 cohorts (DFiR 2019).
- No verified figure exists for the average age at first permanent post.
Why it is structural
Mobility serves as a proxy for excellence in hiring and prestigious grants, and the supply of candidates far exceeds permanent posts, so the queue lengthens.
The other view
Mobility broadens networks and skills, and careful selection for permanent posts takes time.
Being tried
Formal tenure tracks; proposals for shorter stays abroad and dual affiliations.
How we would know it is improving
Average age at first permanent post.
R37Research and teaching drift apart Education is meant to be research-based, but a shrinking share of permanent staff who both research and teach puts the link under strain. Careers and employmentFunding mixAssessment and incentivesDanish data thin
Evidence
- Research-based teaching depends on staff who are active researchers; universities set standards for the ratio of permanent to part-time teaching staff to protect it (for example the University of Copenhagen’s Faculty of Humanities).
Denmark
- The associate professor share of academic staff has almost halved in 20 years (Forskerforum 2022), while postdoc posts, which carry little or no teaching, grew 32% in 2018–2022 (Dansk Industri).
- No national figures found on the share of teaching given by active researchers.
Why it is structural
Research money grows as project funding for fixed-term researchers, while teaching money follows student numbers, so the two are financed and staffed separately and careers reward research.
Held in place by
The other view
Specialised teaching staff can teach very well, and not every course needs an active researcher.
Feeds into
Fed by
Being tried
Faculty standards for research coverage in programmes; teaching portfolios in the job structure; teaching-focused posts.
How we would know it is improving
Share of teaching given by active researchers, by programme.
R14Concentrated private funding with limited accountability Danish research depends unusually heavily on a few very large foundations, which gives them agenda-setting power with little public overview. Governance and powerFunding mixData and infrastructurePartly documented
Evidence
- A large funder can use grant conditions to move culture (Wellcome), but as one funder it cannot change system-level incentives alone.
Denmark
- The Novo Nordisk Foundation granted up to about DKK 9bn for research in 2023, against about DKK 1.7bn from DFF.
- Foundations gave almost DKK 12bn to research in 2021, about a third of the Danish research budget.
- Magtudredningen researchers report “very limited knowledge” of the most influential actors; the Novo Nordisk Foundation has itself called for more transparency and coordination.
Why it is structural
Foundation ownership of industry and tax rules tie research funding to a few companies’ value, there is no national coordination between public and private funders, and foundations sit outside public transparency rules.
The other view
Foundations have built much of Danish research, can make long-term and risky bets, and move faster than public funders. Their independence is a strength of the Danish system, and several have joined efforts on overheads and shared data.
Feeds into
Being tried
The 2023 overhead agreement; the CFA mapping co-funded by public and private funders.
How we would know it is improving
Each funder’s share of research funding, and how much grant data all funders publish.
R15Concentrated power and weak collegial governance Leaders are appointed and answer upwards to boards, and grant money flows through PIs who control junior careers, so power in universities is steeply concentrated. Governance and powerCareers and employmentLaw and regulationPartly documented
Evidence
- 21% of 6,300 PhD students in a Nature survey reported bullying, mostly by supervisors.
Denmark
- The 2003 University Act replaced elected leadership with appointed rectors, deans and heads of department, and boards with external majorities.
- Critics, including Forskerbevægelsen, call for an international evaluation of the Act and its effects.
- Former rector Børge Obel traces the shift over three reforms: civil-servant status ended for new professors and external members joined governing bodies (1993), boards with external majorities hired leaders top-down (2003), and the minister came to approve board chairs (2017). Each step was defensible alone; together they moved power from academic authority to management (Politiken, October 2026).
Why it is structural
Leaders are accountable to boards and the ministry, juniors depend on PIs for contracts and references, and fixed-term contracts make dissent costly.
Held in place by
The other view
Appointed leadership can take decisions that collegial bodies avoid, and boards bring outside perspectives and accountability.
Being tried
Proposals for an evaluation of the University Act; Børge Obel’s five-part package (2026): protected employment after external assessment, boards chosen by a broad representative body without ministerial approval of the chair, more long-term basic funding, a collective professorial veto on permanent appointments, and academic councils with a suspensive veto.
How we would know it is improving
Staff voice in governance, and reported bullying by supervisors.
R16Pressure on academic freedom Political pressure and fear of public attacks narrow what researchers feel free to study and say. Governance and powerLaw and regulationCulture and normsPartly documented
Evidence
- The White House science adviser argues that political appointees should oversee whether grant review is “meritorious”, while not deciding individual grants; critics see a concentration of power in the executive branch (Cowen & Kratsios, New York Times, October 2026).
- Governments in several countries rewrite grant language toward favoured themes and reduce access to public data (de Rijcke, Nature 2026).
Denmark
- Parliamentary resolution V 137 (2021) on “excessive activism in certain research environments” was called a direct attack on research freedom by the chair of the rectors’ conference.
- 13% of researchers had experienced pressure on their research freedom in five years, and four in ten had refrained from commenting in the media within their expertise (DM surveys).
- Roskilde University’s de facto dismissal of professor Michelle Pace in 2026 set off a national debate on academic freedom; Børge Obel argues the case is “the tip of the iceberg” of a systemic problem (Politiken, October 2026).
- 49.4% of surveyed Danish university researchers report fearing, being threatened with or experiencing reprisals for speaking up (Bjerre et al., British Journal of Sociology 2026, drawing on Forskerbevægelsen’s #SafeToSpeak testimonies).
Why it is structural
No regular independent measurement of academic freedom, dependence on short contracts, and no clear duty for universities to defend staff under attack.
The other view
Elected politicians have a legitimate interest in the quality and balance of publicly funded research, and criticism is part of open debate.
Feeds into
Fed by
Being tried
The revised Danish Code of Conduct for Research Integrity (January 2026) gives academic freedom its own section; proposals for an annual survey of research freedom; de Rijcke’s six principles as a test for any measure that affects universities.
How we would know it is improving
Share of researchers reporting pressure on their research freedom or holding back from public comment, in a regular survey.
R38Geopolitics reshapes collaboration and funding Rivalry between great powers and political swings in the US narrow international collaboration and make funding less predictable, while small countries depend heavily on both. Governance and powerFunding mixLaw and regulationPartly documented
Evidence
- Only a quarter of China’s research collaborations now involve US researchers, down from over half a decade earlier (ASPI analysis reported in 2025); US–China co-authorship in nanotechnology has fallen since the late 2010s.
- The US administration proposed cutting NSF’s budget by more than half; Congress rejected the cut for 2026, and the House has proposed a 20% cut for 2027 against the 55% requested (AAAS). The UK limited its science agreement with China to non-sensitive areas in November 2025.
Denmark
- 43% of researchers at Danish universities hold non-Danish citizenship, and EU programmes are a major source of competitive funding. The Irish presidency proposed FP10 at €158bn in October 2026, about 10% below the Commission.
- From 29 October 2026, lead applicants for Wellcome Discovery Awards must be based in the UK or eligible low- and middle-income countries, so Danish institutions can no longer lead.
- No Danish analysis found of how geopolitical shifts affect Danish collaboration patterns.
Why it is structural
Security, foreign and budget policy are made outside the research system, while researchers and institutions absorb the consequences one collaboration at a time.
The other view
Shifting partnerships can strengthen European and Nordic collaboration.
Fed by
Being tried
National research security guidelines (2022); EU debates on research security and FP10; diversifying partners and funding sources.
How we would know it is improving
International co-authorship by partner country, and the share of funding from EU programmes.
R39Weakened research for public policy Research that serves ministries and agencies moved into universities in 2007, where it competes with incentives that reward publications and grants. Governance and powerAssessment and incentivesFunding mixPartly documented
Evidence
- Countries rely on research-based advice to agencies for regulation, environment, food and health; the Academy of Technical Sciences describes the Danish model as a long-standing strength (2014).
Denmark
- Most sector research institutes were merged into universities in 2007, with stronger research freedom as a stated aim.
- Five years later, former sector researchers reported less freedom, and analysts warned that public-service work gets little recognition in university careers (Danish Centre for Studies in Research and Research Policy, 2012).
- Universities’ income from research-based public service has fallen 26% since 2014 (Danske Universiteter, 2026).
Why it is structural
Contract work for ministries is valued by the state but not by academic merit systems, and nobody follows whether the capacity is maintained.
The other view
Moving sector research into universities raised research quality and gave some institutes more independence.
Feeds into
Being tried
Framework contracts between ministries and universities; calls to evaluate the 2007 mergers.
How we would know it is improving
Volume and funding of research-based public service, and how fast advice is delivered.
R17Teaching counts for less than research Hiring and promotion reward research far more than teaching, so good teaching does little for a career. Assessment and incentivesCareers and employmentDanish data thin
Evidence
- A review of 104 UK universities found all used research in promotion criteria, while 31 left teaching out (2009).
Denmark
- The 2019 job-structure order aimed to put research and research-based teaching on an equal footing and introduced teaching portfolios.
- The associate professor share of academic staff has almost halved in 20 years (Forskerforum 2022). No data on how teaching weighs in actual hiring.
Why it is structural
External grants, citations and rankings measure research only, and teaching money follows student numbers without being tied to teaching quality.
The other view
Research excellence is what distinguishes universities from other teaching institutions, and good researchers often make good teachers.
Feeds into
Being tried
Teaching portfolios and academies; CoARA; the Dutch Recognition & Rewards programme.
How we would know it is improving
Weight of teaching in hiring and promotion decisions.
R18Review penalises novelty Panel review and short-term citation measures systematically disadvantage novel, high-risk research. Assessment and incentivesFunding mixWell documented
Evidence
- In a 2025 study of 100,000 NIH applications, projects taking out-of-consensus approaches were less likely to have their funding renewed (Azoulay & Greenblatt, reported in The Economist, October 2026)
- One standard deviation more novelty lowered a proposal’s rank by 4.5 percentile points in a randomised evaluation (Boudreau et al. 2016).
- Novel papers are less likely to be top-cited in the short run and more likely in the long run (Wang, Veugelers & Stephan 2017).
- Long, failure-tolerant HHMI funding produced more high-impact work than comparable NIH funding (Azoulay et al. 2011).
- A grant raises output and future funding but not novelty or disruption, in a causal study of EPSRC applicants (Baruffaldi, Jorzick & Lawson 2026).
- In an FAS design exercise, peer review that “drives funding decisions toward safe, incremental research” was the constraint no group managed to fix, though every other reform depends on it (2026).
Denmark
- Dedicated risk schemes exist, such as the Villum Experiment.
- DNRF Centers of Excellence report that the system penalises risk-taking, especially for early-career researchers on short contracts, and that the DNRF’s own year-five mid-term evaluation makes centres more cautious in their first period (DNRF, September 2026).
- A CFA brief for the DNRF finds that funders rarely take more risk just by asking reviewers to; risk is governed through mandate, autonomy, instrument design and portfolios (Degn, September 2026).
Why it is structural
Low success rates push panels toward safe choices, and short contracts make failure unaffordable for researchers.
The other view
Feasibility matters, many novel proposals fail for good reasons, and panels must spend public money responsibly.
Feeds into
Fed by
Being tried
High-risk schemes, anonymised review, partial lotteries, long person-based funding.
How we would know it is improving
Share of funding in exploratory or high-risk schemes, and outcomes of those schemes over time.
R19Interdisciplinary proposals are funded less Proposals that span disciplines are funded less often, despite broad support for interdisciplinarity in principle. Assessment and incentivesFunding mixDanish data thin
Evidence
- Across about 19,000 Australian proposals, more interdisciplinarity meant a lower chance of funding (Bromham, Dinnage & Hua, Nature 2016).
Denmark
- No Danish data found.
Why it is structural
Panels, funding instruments and careers are organised by discipline.
Held in place by
The other view
Interdisciplinary proposals can be harder to judge and sometimes trade depth for breadth.
Feeds into
Being tried
Interdisciplinary calls and cross-disciplinary panels.
How we would know it is improving
Success rates of interdisciplinary against disciplinary proposals.
Sources
Bromham et al. 2016R20Counting publications When careers and funding follow publication counts, researchers produce more, shorter and safer papers. Assessment and incentivesPartly documented
Evidence
- China launched a major evaluation reform in 2020, yet paper counts are still widely used in hiring and promotion, a sign of how policy can change without practice following (Nature, October 2026)
- Indexed articles grew about 47% from 2016 to 2022, far faster than the number of scientists (Hanson et al. 2024).
- Australia’s count-based funding was followed by more output in lower-impact journals (Butler 2003), a finding later contested.
Denmark
- The Bibliometric Research Indicator (BFI) allocated part of basic funding by publication points from about 2009 until December 2021, and changed publication behaviour in several fields (Deutz et al. 2021).
- Its point lists were also used locally to judge individuals. No study of what changed after 2021.
Why it is structural
Counts are cheap, comparable and legible to managers and rankings, and a university that stops counting fears disadvantaging its staff.
The other view
Counting made output visible and raised productivity in some systems, and judgement without any numbers can be more biased.
Fed by
Being tried
BFI closed; DORA, CoARA, narrative CVs.
How we would know it is improving
Whether publication counts are used in local hiring and allocation.
R21Journal prestige stands in for quality Hiring, promotion and funding lean on journal-level proxies that measure individual work poorly and reinforce publisher prestige. Assessment and incentivesCulture and normsPartly documented
Evidence
- Even a funder that bars all spending on journals and fees finds researchers working to the letter of the rule, so the load of journal thinking moves to policing the policy (Avasthi on the Astera Institute, October 2026)
- The impact factor appears in promotion documents at 40% of research-intensive US and Canadian universities (McKiernan et al. 2019).
- DORA has about 26,000 signers; CoARA had 908 signatories but 256 action plans at the end of 2025.
Denmark
- Denmark set up a CoARA national chapter in January 2026. A full list of Danish signers and evidence of changed hiring practice are missing.
Why it is structural
Metrics save committees time, rankings use citations, and signing is cheap while changing panels is slow.
Held in place by
The other view
Journal standing carries real information about the rigour of review, and committees have limited time to read everything.
Feeds into
Being tried
DORA, CoARA, the Leiden Manifesto, narrative CVs.
How we would know it is improving
Share of hiring and promotion criteria using journal metrics, and CoARA action plans carried out.
R22Data, software and engagement go unrewarded Assessment rewards papers and grants, so data curation, software, team science and public engagement are under-credited. Assessment and incentivesCareers and employmentDanish data thin
Evidence
- Missing merit frameworks for data stewardship have held back recognition and careers (OS-CAM study).
Denmark
- Merit criteria at Copenhagen and SDU formally include teaching and societal impact; their actual weight in hiring is undocumented.
- Data stewardship emerged as a profession around 2019–20 without clear career paths.
Why it is structural
These contributions are local, hard to compare internationally and invisible to rankings.
Held in place by
The other view
Papers remain the main way knowledge is shared and checked, and other outputs are harder to assess fairly.
Being tried
CoARA commitment 1; research software engineer associations; UKRI’s 2026 data policy, which requires code sharing and contributor credit.
How we would know it is improving
Recognition of data, software and engagement in merit criteria and actual hires.
R23Language and global inequity English-language journals from rich countries define visibility, so research in national languages and from the Global South is undervalued. Assessment and incentivesCulture and normsDanish data thin
Evidence
- In Norway, the share of humanities papers in Norwegian fell from about 65% in 2005 to under 40% in 2014.
- Some 4.6 million Chinese-language articles sit outside the global datasets on which international indicators rely, so research-information infrastructure decides what a system can see and reward (James Wilsdon on the China Fundamental Research Report, October 2026).
Denmark
- No measured trend for Danish-language publishing exists.
Why it is structural
International journals carry more prestige, international reviewers cannot read Danish, and rankings count indexed output.
The other view
English gives Danish research international reach, scrutiny and collaboration.
Feeds into
Being tried
The Helsinki Initiative on Multilingualism (2019); Norway’s indicator including national-language outlets.
How we would know it is improving
Share of publications and teaching in Danish, by field.
R24No one sees the whole system Denmark lacks shared data and a forum that looks at the research system as a whole, so structural problems stay anecdotal and no actor answers for them. Data and infrastructureGovernance and powerWell documented
Evidence
- Even in the US, assessments of the research system are infrequent and field by field; Cerf and Córdova propose an annual forum to check whether the whole architecture of government, universities, industry and philanthropy keeps pace (Physics Today, October 2026).
- Countries that made progress built shared monitoring: the Dutch culture barometer, Finland’s open-science monitoring, the EU study of CoARA action plans.
- In an FAS design exercise (July 2026), seven groups each built a new R&D institution from scratch; those that tackled coordination all concluded that convening is not enough without authority to make decisions stick, and that existing coordinating bodies are under-resourced.
Denmark
- The ministry cannot count fixed-term researchers (2024) or the basic funding tied up in external projects (2025).
- The chairs of Innovationsfonden, the rectors’ conference and DFiR say the debate on research becoming companies runs on anecdotes; DFiR is building linked data over three years (October 2026).
- CFA’s 2026 mapping is the first to link competitive grants to the size of research fields.
- “Our system is very disorganized. There is no strategy. There’s not a forum for a strategic dialogue,” said Thomas Trøst Hansen, head of DFiR’s secretariat, who proposes shared system targets such as the share of women professors and the balance of basic and competitive funding (interview, March 2025).
Why it is structural
Each actor collects data for its own needs, and no body is responsible for system-level outcomes.
Held in place by
The other view
Collecting more data costs money and time, and data can become another instrument of control.
Feeds into
Being tried
DFiR’s linked-data project; the CFA mapping; the Danish CoARA chapter.
How we would know it is improving
Whether national statistics exist for each gap named in this register.
R25Unpublished null results and trials Null and negative results are less likely to be written up and published, which distorts the evidence and repeats dead ends. Data and infrastructureCulture and normsLaw and regulationPartly documented
Evidence
- About 65% of null social science experiments in one large set were never written up (Franco et al. 2014).
- 475 Nordic trials completed in 2016–19 had never made results public (2022).
Denmark
- In 2020 over half of trials run by major Danish universities and hospitals lacked results on the EU registry (Rigshospitalet 74, Aarhus University 50), against 5 for all drug companies combined. Current status unknown.
Why it is structural
Journals filter for novelty, there is no credit for null results, and enforcement for academic sponsors is weak.
Held in place by
The other view
Journals have limited space and readers limited attention, and trial registries now catch far more unreported results than before.
Being tried
Trial registration rules, Danish sanctions, registered reports.
How we would know it is improving
Share of Danish trials with results posted within twelve months.
R26Overloaded peer review Peer review rests on unpaid, uncredited work from a shrinking willing pool, while submissions grow fast, now accelerated by generative AI. Data and infrastructureAssessment and incentivesPartly documented
Evidence
- Reviewers spent more than 100 million hours in 2020 (Aczel et al. 2021).
- arXiv submissions doubled in two years to 40,363 in September 2026, and arXiv now caps submitters at two papers a month.
- About 21% of ICLR 2026 reviews were flagged as fully AI-written (detector accuracy contested).
- Manuscript-only review cannot catch the methodological flaws of AI-generated research; the authors of a 2026 PNAS study ask journals and conferences to require full workflow logs and code.
- When applicants to a Volkswagen Foundation call reviewed each other, each proposal got nine or ten reviews and each reviewer spent less than half the time of a panel member, but distributed review and the panel agreed on only three and then one of 18 funded proposals, and only about 56% of its selections would survive a rerun (RoRI Working Paper 23, October 2026).
Denmark
- No Danish data on reviewing load.
Why it is structural
Reviewing is a commons: costs fall on individuals, benefits accrue elsewhere, and it does not count in assessment.
Held in place by
The other view
Peer review remains the best quality check available, and AI tools may also help reviewers.
Being tried
Reviewer credit, paid review trials, publish-review-curate models, distributed review in funding.
How we would know it is improving
Reviewer acceptance rates and time to review.
R27Publisher oligopoly and costs A few commercial publishers capture large public spending, and the shift to author fees moves the barrier from readers to authors. Data and infrastructureFunding mixPartly documented
Evidence
- Five publishers published more than half of all papers in 2013, 70% in the social sciences (Larivière et al. 2015).
- Elsevier’s adjusted operating margin was about 38% in 2024.
Denmark
- Danish research libraries pay about DKK 300m a year in licences, about DKK 80m of it to Elsevier (SDU, around 2022). The 2025–28 Elsevier price is not public.
Why it is structural
Prestige attaches to journal brands that publishers own, each journal is a monopoly good, and researchers choose venues while libraries pay.
The other view
Publishers run infrastructure and services, and national deals have slowed price rises and expanded open access.
Feeds into
Being tried
Plan S, diamond open access, rights retention, read-and-publish deals.
How we would know it is improving
Total national spending on licences and author fees, and the share of diamond open access.
R28Closed research information Evaluation, rankings and policy depend on proprietary citation databases whose coverage, costs and opacity shape what counts. Data and infrastructurePartly documented
Evidence
- The Barcelona Declaration on Open Research Information (2024) has about 146 signatories. Sorbonne left Web of Science, CNRS ended Scopus, and NWO drops both from 2027.
Denmark
- Only DTU’s National Open Research Analytics was found among Danish signatories.
Why it is structural
Rankings are built on commercial data, switching is costly, and databases come bundled in publisher deals.
The other view
Commercial databases are curated and comparable over time, and open alternatives still have quality gaps.
Feeds into
Fed by
Being tried
OpenAlex, Crossref, the Barcelona Declaration.
How we would know it is improving
Use of open research information in national and institutional assessment.
R29AI is changing research work faster than institutions adapt AI tools are moving bottlenecks in research to verification, review and judgement, while assessment, review and training are still built for the old pace. Data and infrastructureAssessment and incentivesCulture and normsDanish data thin
Evidence
- Journals have begun judging submissions by their suspected AI origin, which cannot be verified: AI detectors rated the same excerpts anywhere from very likely human to 100% AI (Stanger, Chronicle of Higher Education, October 2026).
- As AI removes effort and polishes proposals, criteria of “excellence” stop discriminating, and an LLM rewrite of a paper with unchanged content reliably raises AI-reviewer scores; AI is already used informally in assessment with limited oversight (Wilsdon, October 2026).
- Nearly half of 637 surveyed scientists use AI daily, the bottleneck is moving to lab testing and verification, and almost half say AI steers them toward safer, incremental questions (Google DeepMind, 2026).
- arXiv submissions doubled in two years and arXiv now caps submitters at two papers a month (2026).
- The UK Metascience Unit is funding about £5m of research on AI-driven indicators of research quality (October 2026).
- Two open-source “AI scientist” systems choose which results to report using test performance, a built-in form of p-hacking, and silently use synthetic or subsampled data. Reviewing only the paper detects these flaws barely better than chance, 53.7% against 81.5% with logs and code (Luo, Kasirzadeh & Shah, PNAS 2026).
Denmark
- No Danish data on AI use in research. Christian Hendriksen (CBS) warns that without institutional access to frontier models, good practice becomes a privilege of the well-funded (Altinget, June 2026).
- Mogens Fosgerau (University of Copenhagen) warns that if AI removes the struggle through which researchers build expertise and earn recognition, it may erode the human knowledge that makes AI useful, and asks who will own AI-generated knowledge (October 2026).
Why it is structural
Volume incentives meet cheap production, review capacity stays fixed, and no actor is responsible for adapting the system as a whole.
The other view
AI can speed up routine work and free time for judgement, and it is early to judge its effects.
Fed by
Being tried
arXiv caps and bans; the UK Metascience Unit’s AI indicator projects; reproducibility and data policies that make checking easier.
How we would know it is improving
Share of researchers using AI in their work, and review capacity per submission.
R43Funders rarely test their own practices Funders and universities change how they select and support research without testing whether the changes work, so the system learns slowly about itself. Data and infrastructureGovernance and powerPartly documented
Evidence
- “We invest in the most cutting-edge science and innovation, but we’ve been far less scientific about our own funding practices,” said UKRI’s Ian Chapman (Global Research Council, May 2026).
- The UK, the US and Ireland have created metascience units to run experiments on their own funding; RoRI’s META-MOMENT study covers 21 countries.
- NSF’s pilots have been small, decentralised and rarely evaluated: of eight peer-review pilots in the early to mid-2010s, only one was agency-wide, and its review committees rarely look back at outcomes. NSF announced an Office of Metascience on 10 September 2026 (New Science, October 2026).
- Government innovation units tend to follow “cycles of responsibility”: innovation is everyone’s job, then centralised, then cut as overhead. The US units 18F and the OPM Lab were shut down in 2025. Metascience units therefore need their own budget, career leadership and the right to publish findings without leadership review, and about ten years to show results (Gerard & Gustetic, October 2026).
- Much R&D institutional design is “part inherited guesswork”: decisions made by instinct and analogy are later called evidence-based once they work out, and few designs are documented or compared (FAS, September 2026).
- A counter-example: the Volkswagen Foundation ran distributed peer review in parallel with its usual panel for two rounds and commissioned RoRI to evaluate it, with results published (RoRI Working Paper 23, October 2026).
- The Federation of American Scientists is placing three metascience fellows in US agencies such as NSF and DOE to design funding experiments with built-in evaluation (October 2026).
Denmark
- Denmark has no government metascience unit and is not among the 21 countries in META-MOMENT.
- In October 2026 the Villum Foundation gave CBS (Hans Christian Kongsted) DKK 5m to study whether the anonymous Villum Experiment attracts and selects riskier ideas, what they lead to, and which applicants dare to propose them.
Why it is structural
Funders answer for spending and results, not for learning; experiments carry visible risk and evaluation costs money.
The other view
Experiments on real funding decisions raise questions of fairness to applicants, and evaluation diverts money from research itself.
Feeds into
Fed by
Being tried
National metascience units (UK, US, Ireland); funder-run trials such as the Villum Experiment and partial lotteries.
How we would know it is improving
Number of funder and university practices tested with a published evaluation each year.
R30Stress and poor mental health early in careers PhD students and fixed-term researchers report stress, anxiety and depression well above other highly educated groups. Culture and normsCareers and employmentWell documented
Evidence
- 32% of Flemish PhD students are at risk of a common psychiatric disorder (Levecque et al., Research Policy 2017).
- A meta-analysis finds depression in 24% and anxiety in 17% of PhD students (Satinsky et al. 2021).
Denmark
- PAND/DM 2023 (1,501 PhD students): more than one in four consider dropping out; one in three often or always feel burnt out.
- DM 2024: 56% of researchers had stress symptoms in the past month, against 45% of other public-sector members, and researchers work 3.9 hours more a week.
Why it is structural
Grant deadlines and a tight job market make overwork rational, and dependence on one supervisor who is also employer and referee makes complaints costly. Wellbeing is not part of how research environments are judged.
The other view
Research is demanding work that many find highly meaningful, and wellbeing surveys often rely on self-selected samples.
Feeds into
Being tried
Co-supervision requirements, wellbeing programmes, regular DM and PAND surveys.
How we would know it is improving
Stress symptoms and dropout among PhD students and fixed-term staff.
R31Bullying, harassment and attacks on researchers Steep hierarchies make bullying and sexual harassment common and underreported, and researchers on contested topics face harassment from outside. Culture and normsGovernance and powerPartly documented
Evidence
- 61% of UK researchers had witnessed bullying or harassment and 43% had experienced it, 49% of women against 34% of men (Wellcome 2020, self-selected sample).
Denmark
- 32% of female PhD students had experienced malicious sexual comments, gestures or exposure (VIVE, 2024).
- In 2021, 262 gender and migration researchers warned that colleagues had gone on sick leave after intimidation. There is no systematic Danish survey of external harassment.
Why it is structural
One PI often controls funding, renewals, authorship and references; fixed-term and foreign staff have most to lose by reporting; complaints run through the same management line.
The other view
Most supervision works well, and self-selected surveys may overstate how common bullying is.
Feeds into
Being tried
Harassment policies and contact points since 2018; university guidelines on freedom of expression.
How we would know it is improving
Reported bullying and harassment, and cases handled.
R32Findings that do not hold up A substantial share of published findings, especially in fields with small samples and flexible analyses, fail to replicate or replicate with much smaller effects. Culture and normsAssessment and incentivesData and infrastructureWell documented
Evidence
- Replication alone does not reveal error: artifacts can be “very, very reproducible”, so a successful replication may echo the original mistake (Allchin, SHPS 2026)
- 36% of 100 psychology replications were significant, against 97% of the originals (Open Science Collaboration 2015).
- In cancer biology only 50 of 193 planned experiments could be repeated, and effects were a median 85% smaller (Errington et al. 2021).
- 13 of 21 social science experiments from Nature and Science replicated (Camerer et al. 2018).
Denmark
- The Danish Reproducibility Network was founded in 2023 with researchers from all eight universities. No Danish field-wide estimate exists.
Why it is structural
Novel positive results are rewarded, replications are hard to fund and publish, and no institution owns the reliability of the literature.
The other view
Failed replications can reflect real differences between settings, and some fields replicate well.
Feeds into
Fed by
Being tried
Preregistration, registered reports, reproducibility networks, data and code mandates such as UKRI’s 2026 data policy.
How we would know it is improving
Share of studies sharing data and code, and results of replication studies.
R33Questionable research practices Practices short of fraud, such as selective reporting and flexible analysis, are widespread and weaken reliability. Culture and normsAssessment and incentivesWell documented
Evidence
- Preregistration combined with selective reporting or vague plans can leave the published record worse than no plan, because the plan lends unearned credibility (Little, Montgomery & Schnakenberg, SocArXiv, October 2026)
- 51% of Dutch researchers frequently used at least one of 11 questionable practices (National Survey on Research Integrity, 2021).
Denmark
- About nine in ten of 3,402 Danish researchers admitted at least one questionable practice, with patterns very similar to other countries (Schneider et al., PRINT, 2024).
Why it is structural
Pressure to publish, publication bias, uneven methods training and low risk of detection.
The other view
Many practices labelled questionable are legitimate in context, and the PRINT study itself estimates real prevalence well below self-reported use.
Feeds into
Being tried
The Danish Code of Conduct for Research Integrity, revised in January 2026 for the first time since 2014; integrity training for PhD students.
How we would know it is improving
Self-reported questionable practices in repeated national surveys.
R34Society wants more engagement than the system rewards The public trusts research and wants researchers to take part in public debate, while careers give little credit for it and engaging carries personal risk. Culture and normsAssessment and incentivesPartly documented
Evidence
- In a 68-country survey of about 72,000 people, most trust scientists and want them to engage more with society and policymaking (Cologna et al., Nature Human Behaviour 2025).
- Practitioners in an FAS design exercise gave current public-input mechanisms a “failing grade” as box-checking that comes too late, and wanted the public to shape which problems are chosen (2026).
Denmark
- About three in four Danes trust research and roughly two-thirds want researchers to take a bigger part in public debate, yet nearly half say they do not understand what researchers say in the media (V&D Barometer 2025).
- Four in ten Danish researchers have refrained from commenting in the media within their expertise (DM).
- No Danish data found on how much engagement counts in hiring and promotion.
Why it is structural
Assessment counts papers and grants, engagement is local and hard to compare, and researchers who speak publicly carry the risk of attacks alone.
The other view
Not every researcher should engage publicly, and expertise in public debate needs its own quality control.
Feeds into
Being tried
CoARA’s recognition of diverse contributions; university guidelines on freedom of expression; the V&D Barometer as a recurring measure.
How we would know it is improving
Weight of public engagement in merit, and public understanding of research (V&D Barometer).
R35Fraud, paper mills and a thin integrity system Industrial-scale fraud and individual misconduct grow faster than journals’ and institutions’ capacity to detect and correct them. Law and regulationData and infrastructureCulture and normsPartly documented
Evidence
- Nearly 60% of retracted papers in 2014–2024 included authors with China affiliations, which researchers link to pressure to publish (Nature, October 2026)
- Retractions passed 10,000 in 2023, more than 8,000 of them from Hindawi journals. The Retraction Watch Database passed 65,000 retractions in July 2026.
Denmark
- The national board (Nævnet for Videnskabelig Uredelighed) has handled a small caseload since 2017; 13 cases were brought in 2023.
- No count of Danish retractions found.
Why it is structural
Volume-based business models, reputational costs for institutions investigating their own staff, and a narrow national definition that leaves most cases to local processes.
The other view
Retractions show the system correcting itself, and Danish misconduct cases are few.
Feeds into
Being tried
STM Integrity Hub, United2Act, the 2017 Danish Act on Research Misconduct.
How we would know it is improving
Misconduct cases and retractions involving Danish institutions each year.
R36Research security with unclear rules Research security rules are growing, but they are vague enough to leave hard judgements to individual researchers and can narrow open collaboration. Law and regulationGovernance and powerPartly documented
Evidence
- The Dutch academy warned in 2025 that a proposed screening law would limit researchers’ freedom to choose collaboration partners (KNAW, cited by de Rijcke 2026).
- De Rijcke argues security and openness must be held as a single responsibility, since security measures designed without academic freedom in mind place control with those who manage access.
Denmark
- National guidelines for international research collaboration (URIS) were issued in May 2022, covering critical research, knowing your partners and protecting staff. PET names China, Iran and Russia as risk countries.
- A 2024 analysis describes the guidelines’ “strategic ambiguity”, which leaves much interpretation to individual researchers (Internasjonal Politikk).
Why it is structural
Security is set by ministries and intelligence services, academic freedom by universities, and the space between them is left to individuals with little support.
The other view
The threats are real, and some flexibility lets institutions adapt rules to different fields.
Feeds into
Fed by
Being tried
University security units and guidelines (for example Aarhus University); de Rijcke’s principles as a test for proportionate measures.
How we would know it is improving
Collaborations stopped or delayed for security reasons, and researchers’ clarity about the rules.
R42Funding rules fit only universities Eligibility rules and requirements for host institutions and lead investigators route public money to universities, which keeps other kinds of research organisation out. Law and regulationFunding mixDanish data thin
Evidence
- New private institutions such as Convergent Research’s focused research organisations and Episteme now take on work universities fit poorly, and US policy is starting to fund organisations outside universities (NSF X-labs, $1.5bn), though some, such as Arena BioWorks, closed within two years (The Economist, October 2026)
- An open letter from Science Works (2026) argues that UKRI’s eligibility rules, such as three years of audited accounts, keep new research organisations from public funding.
- Newcomers are 55% of organisations in Horizon Europe but win only 14% of the money (European Commission, October 2026).
Denmark
- No Danish review found of who is eligible for public research funding and on what terms.
Why it is structural
Rules written for one type of institution become the default, and those already eligible have little reason to change them.
Held in place by
The other view
Eligibility rules protect public money by requiring financial stability, governance and accountability that new organisations may not yet have.
Fed by
Being tried
Calls to reform UKRI eligibility; the Commission’s call for “alternative funding methods” in the next Horizon programme.
How we would know it is improving
Share of public research funding going to organisations other than universities, and eligibility rules reviewed.
No problems match. Clear the search or choose another area.
Biggest gaps in the evidence
Several problems stay hard to discuss because nobody measures them. These are the data Denmark most lacks:
- How many researchers hold fixed-term posts today, and who leaves academia. The ministry cannot count the stock (R08, R13).
- How much basic funding is tied up in co-financing external grants (R02).
- Funding flows across public and private funders combined, by field and institution (R04, R14).
- Researchers’ social and ethnic background (R10) and the careers of technical and support staff (R12).
- What happened to local publication counting after the BFI ended in 2021 (R20), and whether CoARA changes hiring practice (R21).
- Time spent writing applications (R03; latest national estimate 2007–2011) and the reviewing load on Danish researchers (R26).
- The trend in Danish-language publishing (R23).
How the register works
- Problems are structural when they persist across institutions and over time and come from how the system is built. Individual misconduct and one-off scandals are left out unless they reveal a pattern.
- Codes were set on 17 October 2026, when the register moved to levers, and stay stable from now on, so they can be cited. New problems get the next free number.
- Links between problems are judgements based on the evidence in each entry, meant to be tested and revised. They are mapped on A whole-system approach.
- A problem’s first lever is the one where change would bite hardest; its other levers also bear on it.
- Figures come from primary sources or reports quoting them. Where a number could not be confirmed, it is left out or the gap is named.
- The register is revised as new evidence arrives. Suggestions and corrections go to davidhilmerrex@gmail.com.
Part of The research system series on Metascience. Continue with Why problems persist and A whole-system approach.
Updated 23 October 2026. Updates are drafted with an AI agent from linked sources and reviewed by David Hilmer Rex. Corrections and new sources are welcome: davidhilmerrex@gmail.com. Cite as: David Hilmer Rex, “Structural Problems Register”, The research system, Metascience, 23 October 2026, https://davidhilmerrex.nu/metascience/structural-problems.html.