Most structural problems in research have been known for decades. They persist because the system reproduces them. This page names the mechanisms that do it, with evidence, Danish examples and what weakens each one.
David Hilmer Rex · 14 mechanisms · updated 23 October 2026
Each mechanism links to the problems it holds in place in the Structural Problems Register. No mechanism works alone. They form feedback loops that run through the whole system, which is why changing one part rarely changes the outcome and why a whole-system approach is needed.
Cumulative advantage
Early success brings money, visibility and credibility, and these make the next success more likely. Small or even random early differences grow into large, lasting gaps.
Origin: Robert K. Merton, “The Matthew Effect in Science”, Science, 1968.
Evidence from research systems
Applicants just above the funding line for a Dutch early-career grant later received about €180,000 more and had a 47% higher chance of becoming full professor than those just below, with no sign that the first grant’s results explained the gap (Bol, de Vaan & van de Rijt, PNAS 2018). A 2026 replication across 14 programmes finds the same pattern alongside a smaller setback effect for near-misses.
In the US, 20% of PhD-granting universities supply 80% of faculty (Wapman et al., Nature 2022).
In Denmark
About one in five Danish researchers received nearly 90% of competitive funding from 15 major funders in 2004–2016 (Madsen & Aagaard 2020). DFF’s success rate was 14% in 2024.
What weakens it
Damp the loop early: stop using past funding as a merit criterion, spread smaller grants, use partial lotteries among fundable proposals, create second-chance schemes, and keep several funders with different selection logics so one verdict does not become the system’s verdict.
Prestigious departments train most future faculty, who carry their norms and networks into hiring and review. Those who already hold academic capital define what counts as excellence.
Origin: Pierre Bourdieu, Homo Academicus, 1984; Clauset, Arbesman & Larremore on faculty hiring networks, 2015.
Evidence from research systems
Faculty hiring follows a steep prestige hierarchy, and few faculty move to an institution more prestigious than the one that trained them (Clauset et al., Science Advances 2015).
Faculty with PhDs from outside the top fifth of US institutions leave academia at higher rates after being hired (Wapman et al. 2022).
In Denmark
Heine Andersen (2012) finds social self-selection at entry to Danish research careers, strongest among Young Elite Researcher award winners. No current Danish data on hiring one’s own graduates.
What weakens it
Change who judges: external, cross-sector and early-career members on committees, structured criteria-based hiring, recognition of more kinds of contribution, and support for institutions outside the prestige core so there is more than one ladder of esteem.
Once a system takes a path, every step makes switching more costly. Budgets, skills, expectations and coalitions build up around the existing arrangement, and alternatives start at a disadvantage simply because they are new.
Origin: Paul David (1985), W. Brian Arthur (1989); Paul Pierson, “Increasing Returns, Path Dependence, and the Study of Politics”, 2000.
Evidence from research systems
Five publishers published more than half of all papers in 2013, up from about 20% in 1973 (Larivière, Haustein & Mongeon 2015). Journal brands carry accumulated prestige new outlets cannot quickly match.
In Denmark
In 2019 only 13% of universities’ basic research funding was redistributed by the results-based model; the rest was carried forward from earlier allocations (ministry answer, UFU 29, 2019). Every university has built its budget on its historical share.
What weakens it
Change new money and new entrants more than the stock: protected space for alternatives long enough to build their own returns, and coordinated switching moments, such as a new funding agreement or contract renewal, when many actors move at once.
Organisations in a field grow alike because the shared model looks legitimate, is required by those they depend on, or is what professionals were trained to expect. Universities chase the same world-class research university template.
Origin: Paul DiMaggio & Walter Powell, “The Iron Cage Revisited”, 1983 (coercive, mimetic and normative isomorphism).
Evidence from research systems
Rankings reshape university strategy, recruitment and policy (Hazelkorn 2015) and produce gaming and resource shifts (Espeland & Sauder 2007).
Performance-based research funding exists in at least 14 countries; the prestige competition it creates matters more than the money it moves (Hicks 2012).
In Denmark
Danish universities were merged in 2007 into fewer, broader institutions with similar strategies and ranking ambitions. That this produced convergence is an interpretation, not a measured finding.
What weakens it
Reward different missions (regional, teaching-led, applied, independent institutes), avoid tying funding to ranking positions, and have public bodies legitimise alternative models so copying is not the only safe choice.
When a measure becomes a target, people optimise the measure instead of what it was meant to capture. Indicators reshape the system they describe, and the more decisions depend on a number, the more it gets gamed.
Origin: Charles Goodhart (1975); Donald T. Campbell (1979); Jerry Z. Muller, The Tyranny of Metrics, 2018.
Evidence from research systems
A model of scientific careers shows that selecting for output selects for weaker methods and more false discoveries, without any individual cheating (Smaldino & McElreath 2016).
Norway’s publication indicator moves about 2% of funds but changed publication behaviour (Aagaard, Bloch & Schneider 2015). Australian evidence (Butler 2003) is contested.
In Denmark
The Danish BFI (2009–2021) changed publication behaviour in several fields (Deutz et al. 2021), and its points travelled from a national funding tool into local evaluation of individuals. No study yet of whether those local habits outlived it.
What weakens it
Keep indicators low-stakes and plural, use them to inform judgement, retire them before they ossify, and watch for national indicators being reused on individuals.
Many agree the game is harmful, but each actor loses by quitting first. A researcher who stops chasing high-impact journals, or a university that stops counting, pays the cost while competitors gain.
Origin: Mancur Olson, The Logic of Collective Action, 1965; Garrett Hardin, 1968; Elinor Ostrom, Governing the Commons, 1990.
Evidence from research systems
Plan S was designed as funders moving together so individual researchers would not bear the first-mover cost.
The journal impact factor still appears in promotion documents at 40% of research-intensive US and Canadian universities (McKiernan et al. 2019).
In Denmark
Joint national licence negotiations and the Danish CoARA chapter (January 2026, co-chaired by DFF, Villum Fonden and Aalborg University) are coordination devices of this kind.
What weakens it
Coordination: joint commitments with thresholds, funder alliances, national chapters, shared timetables, and protection for first movers.
Reviewers, committees, deans and council members are by definition those who succeeded under the current system. It was fair to them, so they tend to see it as fair, and they select successors who resemble them. Those who left are not in the room.
Across six research systems after COVID-19, plans to “build back better” returned to normal; the authors call it “the gravity of the status quo” (Derrick et al., Higher Education 2024).
78% of UK researchers say competition creates unkind and aggressive conditions; only 29% feel secure in a research career (Wellcome 2020).
In Denmark
At about 0.8 percentage points a year, women will make up half of Danish professors around 2055 (Folketing UFU bilag 67, 2024/25). Committees and councils are drawn from this pool.
What weakens it
Bring in those outside the survivor pool (early-career researchers, those who left, other sectors), collect exit data, and keep the evaluation of reforms independent of those whose success the system confirmed.
Ministry, public and private funders, universities, publishers and rankers each optimise their own part, and costs are pushed onto someone else: universities’ basic funds, researchers’ careers, the next government. Each can rightly say it is not their decision.
Origin: Principal–agent theory (Jensen & Meckling 1976; Guston 2000); externalities (Pigou 1920); the problem of many hands (Thompson 1980).
Evidence from research systems
The OECD (2021) recommends linking research funding to HR policy because no single actor owns career outcomes.
In Denmark
Foundations long left indirect costs to universities’ basic funds until a 2023 agreement. Project funders pay for project posts; universities carry the career risk; researchers carry the precarity.
What weakens it
Make costs visible and allocate them (full-cost agreements, career conditions attached to grants), create arenas where the system is discussed as a whole, and name who is accountable for system-level outcomes.
Budgets run one to four years, grants three to five, careers and research programmes ten to thirty. Slow costs such as lost talent lose out to visible short-term priorities, and short cycles push researchers toward safe, quick projects.
Origin: Political business cycle (Nordhaus 1975); Azoulay, Graff Zivin & Manso on incentives and creativity, 2011.
Evidence from research systems
Long, failure-tolerant HHMI appointments produced more high-impact and more novel work than comparable short-cycle NIH funding (Azoulay et al. 2011).
In Denmark
The 2025 research reserve was split into 37 one-year earmarks, which the Young Academy criticised; the 2026–29 agreement moves part of the money to a four-year frame.
What weakens it
Match funding horizons to the problem: longer, stable instruments for people and programmes, multi-year political agreements, and funders whose mandate deliberately takes the long view.
Parts of the system fit together. Change hiring criteria but leave funders, rankings and journal prestige as they are, and the old behaviour returns. Each part is also defended by an actor who can block change.
Origin: Hall & Soskice, Varieties of Capitalism, 2001 (institutional complementarities); Tsebelis, 2002, and Immergut, 1992 (veto players).
Evidence from research systems
A review of 224 studies finds inequalities in the research system “overlapping and mutually reinforcing” (Bak, Quantitative Science Studies 2026). Twenty expert interviews show competition, prestige, geography and publishing becoming “structurally intertwined” (Bak, Zenodo preprint 2026).
In Denmark
Denmark closed the BFI in 2021, but publication counts and journal tiers persist in hiring, rankings and foundation review, so the incentive is held in place by the parts around it (interpretation).
What weakens it
Coordinated packages across several parts at once (funders, universities and assessment together), or deliberately building alternative clusters where hiring, funding and evaluation are aligned differently.
Many structural problems lack system-level data, so they stay anecdotal and easy to dismiss, while what is easy to count (papers, citations, grants) gets the attention.
Origin: Agenda-setting and problem definition (Kingdon 1984; Stone 1989); Donella Meadows on information flows as a leverage point.
Evidence from research systems
Comparable international data on research staff in non-standard employment is scarce (OECD 2021). Research on inequality in science relies heavily on bibliometrics and rarely studies structural drivers (Bak 2026).
Evidence on what works also suffers from survivorship: we know the DARPA programmes that produced GPS and the internet, and much less about the bets that failed or the agencies that copied the model (FAS, September 2026).
In Denmark
The ministry cannot count how many researchers hold fixed-term posts (2024) or how much basic funding is tied up in external projects (2025). Magtudredningen researchers report “very limited knowledge” about the most influential actors in research and foundations.
What weakens it
Invest in system data: career tracking, exits, full costs, and funding flows across public and private funders. This is high-leverage and low-conflict, because it changes the information everyone acts on.
Much of what researchers and universities compete for is relative: rank, journal tier, elite grants. If everyone improves, nobody moves up, so effort escalates without improving the whole.
Origin: Fred Hirsch, Social Limits to Growth, 1976 (positional goods); Robert Frank on positional arms races.
Evidence from research systems
Wellcome (2020): 78% say competition creates unkind and aggressive conditions and 75% say creativity is stifled.
In Denmark
Applications to DFF rose 87% after 2017 while success rates fell to about 14%.
What weakens it
Reduce reliance on single rank orders, create shared non-positional goods (infrastructure, open data, collaboration funding), and let institutions excel at different things so prestige has more than one axis.
New initiatives are layered on top of old ones, so incentives pile up and contradict each other. Organisations adopt the form of reform (a signature, a strategy) while daily practice stays the same.
Origin: Meyer & Rowan, “Institutionalized Organizations: Formal Structure as Myth and Ceremony”, 1977; Streeck & Thelen on policy layering, 2005.
Evidence from research systems
CoARA had 908 signatories but 256 action plans at the end of 2025. The UK’s Athena SWAN review found box-ticking, with 77% of respondents in one survey calling the workload excessive.
In Denmark
Denmark has a CoARA national chapter from 2026; nobody yet measures whether local hiring and promotion practice follows.
What weakens it
Require implementation plans and follow-up, measure practice instead of pledges, and remove old incentives when adding new ones.
Academia carries an implicit picture of the ideal researcher: always available, mobile, single-minded, free of care duties. It feels neutral but fits some lives better than others, and the belief in pure meritocracy makes the outcomes look deserved.
Origin: Joan Acker, “Hierarchies, Jobs, Bodies”, 1990 (the unencumbered worker); Michael Young on meritocracy, 1958.
Evidence from research systems
A third of full-time UK researchers work more than 50 hours a week (Wellcome 2020). Women leave disproportionately at the early-to-mid career transition (OECD 2021).
In Denmark
Mothers in Danish academia are 29% less likely than they would otherwise be to remain eight years after becoming parents (Danish register study, LSE CEP DP 2160).
What weakens it
Visible alternative career patterns and leadership behaviour, care-compatible timelines, part-time leadership, and recognition of team science. High leverage, slow to move.
Seen together, most mechanisms are feedback loops. Reinforcing loops amplify (success breeds success). Balancing loops pull the system back to where it was. The loop below links several register problems into one cycle, showing how a funding decision ends up shaping who becomes a professor, and back again.
flowchart LR
A["Basic funding held flat"] --> B["More reliance on project grants"]
B --> C["More fixed-term posts"]
B --> H["Co-financing drains basic funding"]
H --> A
C --> D["Parents and people without backing leave"]
D --> E["A narrow professoriate"]
E --> F["Committees judge by track record and the ideal academic"]
F --> G["Funding concentrates on past winners"]
G --> B
Systems thinking names recurring loop patterns, or archetypes. Each appears in the research system:
Archetype
In the research system
Mechanisms
Problems
Success to the successful
Grant winners get more grants; prestigious departments place more faculty.
Cumulative advantage, Reproduction of academic elites
R04, R09
Shifting the burden
Project money stands in for basic funding, and co-financing then drains the basic funding that pays for free research and careers.
No one owns the whole, Path dependence and lock-in
R01, R02, R08
Fixes that fail
Publication-count funding raises output and invites gaming; one-year earmarks solve this year’s priority and fragment the next.
Measures become targets, Mismatched time horizons
R20, R05
Tragedy of the commons
Reviewer time and trust are shared resources depleted by each actor’s rational submission volume.
Nobody can exit alone
R26, R03
Escalation
Positional races for rank, journal tier and elite grants.
The race for relative position
R03, R30
The status quo pulls back
Reforms are absorbed and the system returns to its earlier state.
The winners become the judges, Reform on paper, The parts hold each other in place
R21, R15
Where change has leverage
Donella Meadows ranked the places to intervene in a system from weakest to strongest. Most research reforms work at the first level: a new grant scheme, a pool of money, a lottery. These are useful, and they work best when combined with change at the higher levels, which alter the loops themselves.
1
Numbers and parametersWhere most reforms sitGrant sizes, success rates, a new pool of money, a new call.
2
Stocks and buffersThe number of permanent posts; the size of basic funding.
3
Structure of flowsWho funds what: the balance of basic and project money, public and private.
4
DelaysGrant length, time to a permanent post, budget cycles.
5
Balancing loopsOverhead agreements, caps on applications, minimum contract lengths.
6
Reinforcing loopsPartial lotteries, limits on concentration, second-chance schemes.
7
Information flowsSystem data on careers, exits, costs and funding flows; a shared register of problems.
8
RulesAssessment criteria, employment law, the University Act, grant conditions.
9
Self-organisationRoom for new institutions, funders and formats to emerge and grow.
10
GoalsWhat the system is for, and what counts as excellence.
11
ParadigmThe shared belief that research advances through individual competition.
Higher leverage is slower and more contested, and it involves more actors. That is the argument for working on several levels at once, with experiments and new institutions (level 9) complementing changes to information, rules and goals (levels 7, 8 and 10).
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, “Why Problems Persist”, The research system, Metascience, 23 October 2026, https://davidhilmerrex.nu/metascience/why-problems-persist.html.