The research system · approach
A Whole-System Approach
How to reform the research system as a whole: every actor, every lever, every level, and a way of working that faces the deep structural problems directly.
Experiments, small and large, are useful. So are new institutions and new funding formats. Yet the problems that shape researchers’ lives most are structural. Why do women lose career steps when they have a child? Why is power so unequally distributed? Why is almost all funding tied to projects?
These problems come from how funding, careers, power, assessment and publishing work together, and they persist because the parts hold each other in place. A whole-system approach names them, looks at how they connect, and changes the rules where they are made, with experiments and new institutions as part of the same effort.
This page builds on two others: the Structural Problems Register, an ongoing list of the problems, and Why problems persist, the mechanisms that reproduce them.
What “whole” means
An approach is whole-system when it covers four dimensions at once. Leaving one out is the most common reason reforms fade.
All actors
State, public and private funders, universities at every level, researchers at every stage, staff, unions, publishers, academies, users and international bodies.
All levers
Funding mix, careers and employment, governance and power, assessment, data, culture and law.
All levels
National (law, budgets, funders), institutional (universities, foundations’ programmes) and the research group and individual career.
Over time
Ongoing work with shared goals, monitoring and revision, outlasting single reforms and political cycles.
How the problems drive each other
Problems in the register feed into one another. A few of them drive many others, so change there spreads furthest. Others are mainly where the strain shows. The six drivers below feed into the most problems; the full set of 69 links is in the register, under each entry. The links are judgements based on the evidence, meant to be tested and revised.
R01 Project funding crowds out basic
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.
Feeds into
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.
Feeds into
Hiring, promotion and funding lean on journal-level proxies that measure individual work poorly and reinforce publisher prestige.
Feeds into
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.
Feeds into
Low success rates mean researchers spend a large share of their time on proposals that are rejected.
Feeds into
R14 Concentrated private funding
Danish research depends unusually heavily on a few very large foundations, which gives them agenda-setting power with little public overview.
Feeds into
Where the strain shows: problems fed by three or more others
The pattern matters for strategy. Outcomes such as unreliable findings, stress and the drift between research and teaching are where problems show up. Drivers such as the funding balance, precarity and journal prestige are where change spreads furthest. A whole-system approach measures outcomes and acts on drivers.
Where to start
Four questions set priorities: how many other problems a problem feeds, how severe it is, how tractable it is in the next few years, and who can move first. Severity and tractability below are my assessments, for discussion.
| Problem | Feeds | Severity | Tractability | Who can move first | Why now |
|---|---|---|---|---|---|
| R24 No one sees the whole | 4 | Medium | High | Ministry, DFiR, all funders | Low-conflict and makes every other problem visible. Start here. |
| R01 Project funding crowds out basic | 7 | High | Medium | Parliament, foundations, universities | Feeds precarity, application burden and the drift between research and teaching. |
| R08 Precarity | 6 | High | Medium | Universities, funders, unions | Feeds the parenthood penalty, stress, bullying and late permanence. |
| R21 Journal prestige as proxy | 6 | High | Medium | Universities, funders, CoARA chapter | Feeds publication bias, questionable practices and undervalued teaching and data work. |
| R09 Parenthood penalty | 1 | High | Medium | Funders, universities, unions | Severe and well documented; a package can show results within a few years. |
| R14 Concentrated private funding | 3 | Medium | Medium | Foundations, ministry | Shapes the funding balance; foundations are already partners on overheads and data. |
| R03 Application burden | 3 | Medium | High | Funders | Funders can change calls, caps and selection methods on their own. |
A sensible order: start with shared data (R24), which is low-conflict and makes the rest visible. Then take packages for the funding balance, careers and parenthood, and assessment, which together reach the main drivers.
Who holds which lever
No actor holds every lever, and every lever is held by several actors. This is why problems persist when one actor acts alone (the mechanism No one owns the whole) and why reforms need several actors moving together.
| Actor | Funding mix 18 problems | Careers and employment 11 problems | Governance and power 9 problems | Assessment and incentives 22 problems | Data and infrastructure 13 problems | Culture and norms 13 problems | Law and regulation 11 problems |
|---|---|---|---|---|---|---|---|
| Parliament and government | ● | ○ | ● | ○ | ○ | · | ● |
| Ministry and its agencies | ● | ● | ○ | ○ | ● | · | ● |
| Public funders (DFF, Grundforskningsfonden, Innovationsfonden) | ● | ○ | · | ● | ○ | ○ | · |
| Private foundations | ● | ○ | ○ | ● | ○ | ○ | · |
| University boards and leadership | ○ | ● | ● | ● | ○ | ● | · |
| Departments and PIs | · | ● | ○ | ● | · | ● | · |
| Researchers at every career stage | · | ○ | ○ | ○ | · | ● | · |
| Unions and collective agreements | · | ● | ○ | · | · | ○ | ○ |
| Technical and administrative staff | · | ○ | · | · | ○ | ○ | · |
| Publishers, databases and rankings | ○ | · | · | ● | ● | ○ | · |
| Academies and learned societies | · | · | · | ○ | · | ● | · |
| Industry, public-sector users and citizens | ○ | ○ | ○ | ○ | · | ○ | · |
| EU and international bodies | ● | ○ | · | ● | ○ | ○ | ○ |
● holds the lever directly ○ shapes it · little role. A first mapping for Denmark, to be refined.
How the levels undo each other
Reform at one level is often undone at another. Examples from recent years:
- Dutch universities broadened how they recognise researchers, then national budget cuts from 2024 removed money for the extra permanent posts.
- Funders introduce narrative CVs while departments keep hiring on journal metrics (R21).
- Gender schemes fund individual women while the parenthood penalty builds up through short contracts and output-per-year assessment (R09, R08).
- Denmark closed the BFI in 2021, while publication counting lives on in local hiring and international rankings (R20).
What earlier reforms covered
Most system-wide reforms elsewhere worked on one or two levers, usually assessment. Funding mix, employment, power and data were often left untouched, and the untouched levers pulled practice back.
| Reform | Funding mix | Careers and employment | Governance and power | Assessment and incentives | Data and infrastructure | Culture and norms | Law and regulation | What happened |
|---|---|---|---|---|---|---|---|---|
| Netherlands: Recognition & Rewards (2019–) | · | ○ | · | ● | ○ | ● | · | Broad joint commitment and a national culture barometer; budget cuts from 2024 undercut its career side. |
| UK: People and Culture Strategy (2021) and REF 2029 | ○ | ○ | · | ○ | · | ● | · | Culture made a policy object; it stuck once culture became 20% of institutional scores in REF 2029. |
| CoARA (2022–) | · | · | · | ● | ○ | · | · | 908 signatories but 256 action plans by the end of 2025. |
| Norway: NOR-CAM and BALANSE | · | ○ | · | ● | · | ● | · | A shared assessment matrix with local adaptation; gender work framed as fixing the system. |
| Finland: researcher evaluation recommendation (2020) | · | · | · | ● | ○ | · | · | Learned societies, funder and universities committed together, with monitoring. |
| Germany: fixed-term contracts law reform (2021–) | · | ● | · | · | · | · | ● | A mass movement (#IchBinHanna) produced minimum contract lengths; the six-year cap remains. |
| Australia: Universities Accord and ATEC (2024–26) | ○ | · | ● | · | ○ | · | ● | Independent review, wide consultation and a permanent steward institution; mostly about education. |
| Denmark: 2026–29 research agreement | ● | · | · | · | · | · | · | Multi-year frame and DKK 2bn more basic funding. |
| Denmark: rectors’ equality taskforce (2023) | · | ○ | · | ○ | · | ○ | · | Eight recommendations adopted university by university. |
● main lever ○ partly addressed · not addressed.
How to work: a learning cycle
A whole-system approach is a way of working more than a single plan. It borrows from experimentalist governance (Sabel and Zeitlin): shared provisional goals, freedom to act locally, regular joint review and revision.
- Map the system togetherKeep a shared register of problems and the mechanisms behind them, built with all actors.
- Agree shared goalsProvisional and measurable, such as a lower fixed-term share or a closed parenthood penalty by a set year.
- Act in packagesCombine reforms of rules with experiments and new institutions, across actors and levers at once.
- Measure practiceTrack what happens in hiring, careers and funding flows, beyond signatures and strategies.
- Review and reviseMeet regularly, compare results across institutions and funders, and change goals and tools.
Two conditions run through every step: the people outside the current winners (early-career researchers, those who left, technical staff, other sectors) take part, and every measure passes a legitimacy test for academic freedom (below).
Packages, illustrated
Packages make complementarity concrete. Each problem gets a set of changes held by different actors, so no single change has to carry it alone. Four sketches, covering the main drivers:
Closing the parenthood penalty
Problems R09 R08 R13 · mechanisms: the ideal academic, the winners become the judges, mismatched time horizons
- Funders
- Automatic extension of grants and contracts for parental leave, the same across public and private funders.
- Universities
- A central parental-leave fund, as the rectors’ taskforce recommended in 2023, and earlier permanent posts.
- Assessment
- Output judged relative to time available; narrative CVs; leave never counted as lost productivity.
- Agreements and law
- Minimum contract lengths and limits on chains of fixed-term contracts.
- Data
- Register-based monitoring of who leaves and when; the Danish register study shows it can be done.
- Culture
- Part-time leadership, visible role models, leaders accountable for retention.
Rebalancing project and basic funding
Problems R01 R02 R03 R04 R14 · mechanisms: lock-in, no one owns the whole, cumulative advantage
- Parliament
- Multi-year growth in basic funding, linked to the growth of external grants.
- Foundations and councils
- Full-cost funding, a shared calendar and coordination of priorities, and joint data on funding flows.
- Funders
- A wider range of instruments: smaller grants, longer person-based grants, partial lotteries, caps on concentration.
- Universities
- Internal allocation that protects time for research without a grant.
- New institutions
- Organisations with long horizons and different selection logics alongside universities.
- Data
- National figures on co-financing and on the share of research funded by projects, by field.
Opening up assessment
Problems R21 R20 R22 R17 R28 · mechanisms: measures become targets, nobody can exit alone, reform on paper
- Universities
- Hiring and promotion criteria without journal metrics, with teaching, data, software and engagement counted, and with the change checked in actual hires.
- Funders
- Narrative CVs and the same principles across public and private funders, so researchers meet one set of expectations.
- Coordination
- The Danish CoARA chapter sets a shared timetable, so no institution moves alone.
- Infrastructure
- Open research information (OpenAlex, Crossref) in national and institutional assessment, as NWO and CNRS have chosen.
- Data
- A survey of criteria used in Danish hiring, repeated every few years.
Seeing the whole system
Problems R24 R02 R08 R10 R14 · mechanisms: what is not measured stays invisible, no one owns the whole
- Ministry
- National statistics on fixed-term staff (the stock, as well as hires), exits from research, and co-financing.
- Funders
- All public and private funders publish grant data in a shared format, building on CFA’s 2026 mapping.
- DFiR
- Linked register data on careers and on paths from research to companies, extending its current project.
- Shared report
- An annual state-of-the-system report built on this register and the indicators below.
What to measure
Each problem in the register has an indicator for how we would know it is improving. For the priority problems, here is what Denmark can measure today:
| Problem | Indicator | Available? | Today |
|---|---|---|---|
| R24 No one sees the whole | Whether national statistics exist for each gap named in this register. | Partly | The ministry cannot yet count fixed-term staff or co-financing. |
| R01 Project funding crowds out basic | Share of university research paid by basic funding, by university and field, each year. | Yes | Basic funding about DKK 10bn a year since 2011; private share of university R&D 18% (2021). |
| R08 Precarity | Number and share of researchers on fixed-term contracts, and median years from PhD to a permanent post. | No | Only recruitment flows are counted, not the stock. |
| R21 Journal prestige as proxy | Share of hiring and promotion criteria using journal metrics, and CoARA action plans carried out. | No | No survey of metrics in Danish hiring criteria. |
| R09 Parenthood penalty | Retention of mothers and fathers eight years after a first child; women’s share of professors and of new professor hires. | Partly | She Figures and a register study exist; no regular national monitoring. |
| R14 Concentrated private funding | Each funder’s share of research funding, and how much grant data all funders publish. | Partly | Foundations’ totals are known; combined public and private flows by field are new (CFA 2026). |
| R03 Application burden | Success rates and estimated researcher time spent on applications, per funder. | Partly | Success rates are published; time spent on applications was last estimated in 2007–2011. |
Legitimacy: steering and academic freedom
A whole-system approach steers, and steering can slide into interference. Sarah de Rijcke’s six principles (Nature, 21 September 2026) give a test that applies to governments, funders and university leaders alike:
- Ground academic freedom in the duty to produce reliable knowledge.
- Protect questions that have value that cannot be shown in advance.
- Treat universities as democratic infrastructure.
- Hold security and openness as a single responsibility.
- Protect the people who carry the risk of independent work.
- Return the trust on which institutional autonomy rests.
Her core test: does a measure strengthen or weaken independent scrutiny? A whole-system approach that passes it is also more likely to win the support of researchers themselves.
Denmark
What exists
- A policy council (DFiR), now building linked data on research and companies.
- A rectors’ conference (Danske Universiteter) with sector-wide recommendations on equality.
- Large foundations that agreed a shared overhead model with universities in 2023.
- A multi-year research agreement for 2026–29 with DKK 2bn more basic funding.
- A CoARA national chapter (2026), the Danish Reproducibility Network, and CFA’s 2026 mapping of competitive funding.
- Register data that few countries can match.
- A debate that has started to think in systems: former rector Børge Obel proposes five changes to university governance that “work together”, because “the problem is systemic, and so the solution must be too” (Politiken, October 2026).
What is missing
- A shared map of the system’s problems and how they connect.
- System data on fixed-term posts, exits, co-financing and combined public and private funding flows (R24).
- A forum where the state, public and private funders, universities, unions and early-career researchers look at the whole together.
- Someone responsible for system-level outcomes such as careers and the balance of funding.
- Evaluation of reforms once they are in place.
Six first steps would put a whole-system approach in place:
- An annual state of the research system report built on a shared problem register.
- A national research-system forum that includes the private foundations, unions and early-career researchers, with a mandate and the authority to make agreed decisions stick. In a 2026 FAS design exercise, every group that tackled coordination reached the same conclusion: convening alone is not enough (FAS).
- Data commitments from the ministry and funders on the gaps listed in the register.
- Packages across actors for two or three problems at a time, starting with the parenthood penalty and the funding balance.
- A portfolio of experiments and new institutions with shared evaluation, so what works can spread.
- The legitimacy test above applied to every measure.
Pitfalls
- Reform on paper. Signatures and strategies without changed practice: CoARA’s gap between 908 signatories and 256 action plans; box-ticking in Athena SWAN.
- Reform fatigue. Response to the Dutch culture barometer fell from 24.6% to 14.1% at Groningen.
- Backlash. 171 Dutch professors called Recognition & Rewards harmful in 2021; 113 younger researchers answered.
- Fiscal undercutting. Budget decisions elsewhere in government can undo a reform’s base.
- Narrow scope. One lever at a time, while the others pull practice back.
- Uneven side effects. Extra equality work often falls on the people the reform should help.
- Capture. The strongest actors shape the agenda, especially when private funders’ influence is poorly mapped.
- Overreach. System steering that weakens academic freedom loses legitimacy.
Frameworks behind this approach
- Donella Meadows, leverage points, and Peter Senge’s system archetypes.
- Frank Geels’ multi-level perspective: new institutions and formats as niches; universities, councils and assessment conventions as the regime; geopolitics, AI and budgets as the landscape. Niches and regime change together.
- Schot and Steinmueller, transformative innovation policy (2018).
- Sabel and Zeitlin, experimentalist governance.
- OECD, Systems Approaches to Public Sector Challenges (2017).
- Mariana Mazzucato on mission-oriented policy and UNDP’s portfolio approach.
Sources for the cases
Dutch Recognition & Rewards · Culture barometer (Groningen) · UK People and Culture Strategy · CoARA figures · EU study of CoARA action plans · NOR-CAM · BALANSE · Finland · Germany, Bundestag · Australia, ATEC · Danish rectors’ taskforce · Danish 2026–29 agreement · Athena SWAN review
Part of The research system series on Metascience. See the Structural Problems Register and Why problems persist.
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, “A Whole-System Approach”, The research system, Metascience, 23 October 2026, https://davidhilmerrex.nu/metascience/whole-system.html.