How Government Research Funding Makes Science Worse

Getting your Trinity Audio player ready...

Government funding has enabled major scientific achievements, from space exploration and particle physics to vaccines, public-health systems, and fundamental mathematics. Many important research programs would be impossible without public money.

Yet government research funding also has a less visible effect: it changes the behavior of scientists and institutions competing for that money.

When grants are allocated through centralized agencies, elaborate applications, short funding cycles, institutional eligibility rules, and committee-based peer review, researchers gradually optimize for the funding system rather than for scientific truth. The result is not necessarily fraudulent science. More often, it is cautious, bureaucratic, fashionable, fragmented, and institutionally concentrated science.

The problem is therefore not simply that governments spend too little or too much. It is that conventional government grant systems frequently reward the wrong things.

Government Funding Can Turn Scientists Into Grant Writers

Researchers cannot conduct experiments, prove theorems, analyze data, or write software while they are preparing grant applications and compliance reports.

A study of federally funded US researchers found that principal investigators reported spending approximately 42% of their project time on administrative work rather than research. Another study of Australian health-research grants estimated that preparing one proposal required an average of 34 working days. Across one annual funding round, researchers collectively spent an estimated 550 working years preparing applications.

Most applications were unsuccessful. Therefore, much of this labor produced no research at all.

This creates a destructive cycle:

1. Low grant success rates encourage researchers to submit more applications.
2. More applications increase the workload of reviewers and agencies.
3. Review becomes less careful and more dependent on shortcuts.
4. Researchers spend even more time tailoring proposals to anticipated committee preferences.
5. The system funds grant-writing ability rather than scientific ability.

Administrative oversight is necessary when public money is involved. Governments must prevent fraud, conflicts of interest, misuse of funds, and unsafe research. But beyond a certain point, oversight consumes the activity it is intended to support.

Even the US National Institutes of Health has continued introducing administrative-burden reduction measures, indicating that the problem is recognized within the funding system itself. In December 2025, for example, the NIH removed certain preliminary requirements for large unsolicited applications.

Grant Committees Prefer Predictable Research

A grant proposal must describe a project before the project has been completed. Applicants are normally expected to state:

• what they will discover;
• why the discovery will matter;
• which methods they will use;
• how long each stage will take;
• what outputs they will produce;
• and why the project is likely to succeed.

This is reasonable from an accounting perspective, but often unreasonable from a scientific perspective.

Real discovery is uncertain. A genuinely original investigation may begin with a question whose importance is not yet widely understood. Its methods may change repeatedly. Its results may contradict the applicant’s initial assumptions. The project may fail in its original form while producing an unexpected discovery elsewhere.

Grant systems nevertheless require researchers to make uncertain exploration appear predictable. Scientists learn to present research as a managed production process with predetermined milestones.

This structurally favors:

• incremental extensions of accepted theories;
• projects with abundant preliminary results;
• established methodologies;
• fashionable subjects recognized by reviewers;
• and research that can promise visible outputs within a short grant period.

It disadvantages speculative, interdisciplinary, unconventional, or foundational work.

A recent analysis of competitive research funding identified concerns about unreliable allocation decisions, the costs of competition, and the effects of competitive funding on researchers’ willingness to conduct risky research.

A committee is rarely punished for funding a respectable project that produces modest results. It may, however, be criticized for funding an unusual project that fails. Rational reviewers therefore have an incentive to select proposals that are defensible rather than transformative.

Peer Review Can Become Science by Consensus

Peer review is indispensable for evaluating technical work, but pre-publication grant review asks peers to evaluate discoveries that do not yet exist.

That is much harder than assessing completed research.

Reviewers must predict whether:

• the central idea is correct;
• the applicant can execute it;
• the methods will work;
• the results will be significant;
• and the project will remain relevant several years later.

These predictions are inherently uncertain. Studies of grant-review practices show that reviewers use numerous formal and informal criteria, while the weighting of those criteria is not always transparent or consistent.

Consensus-based selection also creates an originality penalty. A proposal must usually appear innovative enough to justify funding but familiar enough to be understood and approved by several reviewers.

That produces a narrow “acceptable novelty” range:

The proposal should be new—but not so new that reviewers lack the conceptual framework needed to recognize its value.

This is especially harmful to research that challenges an established discipline’s terminology, assumptions, or division of subjects. A project may be rejected not because it is demonstrably wrong, but because it does not resemble the projects reviewers already consider legitimate.

Science then becomes partly governed by collective expectations. Ideas are funded because experts expect them to succeed, while ideas that could change expert expectations struggle to enter the system.

Funding Follows Prestige

Government grant systems are presented as competitions between proposals, but they are also competitions between reputations, institutions, and professional networks.

Reviewers may consciously or unconsciously use signals such as:

• the applicant’s university;
• previous grants;
• publication count;
• journal prestige;
• citation metrics;
• academic title;
• senior collaborators;
• or familiarity with the applicant’s research group.

These signals can contain useful information. A researcher who has successfully completed several projects may genuinely be more likely to complete another one. But prestige also creates cumulative advantage.

Scientists who receive early funding gain staff, equipment, time, publications, and institutional visibility. Those advantages make them stronger applicants in the next round. Researchers who narrowly miss early opportunities must compete without the resources that would have made their later proposals competitive.

This is a version of the Matthew effect: recognition and resources accumulate around researchers who already possess recognition and resources. The National Academies has also noted that bibliometric indicators influence funding, hiring, promotion, and researcher incentives.

The system can consequently confuse three different questions:

1. Has this person succeeded within the academic system?
2. Can this person write a persuasive grant proposal?
3. Is this person’s current research useful and correct?

These questions overlap, but they are not identical.

Institutions Receive Funding, Not Just Scientists

Many government grants are effectively accessible only through universities, recognized research institutes, or other approved organizations.

This excludes or disadvantages:

• independent researchers;
• amateur scientists;
• researchers without conventional degrees;
• scientists between academic positions;
• authors working on unconventional long-term projects;
• and developers building scientific infrastructure outside universities.

Institutional administration can provide laboratories, ethics review, legal support, financial controls, and collaboration networks. But institutional eligibility also gives universities a gatekeeping role.

An idea may need approval from several layers before it can receive public support:

• a department;
• a host institution;
• internal grant administrators;
• external reviewers;
• a funding agency;
• and sometimes political or strategic authorities.

Every layer may be individually defensible. Together, they create a narrow institutional channel through which an unconventional idea must pass.

Consider a researcher writing a 400-page mathematical monograph on a new subject without a PhD or university affiliation. Traditional funding committees may see several warning signals: the work is too large to review easily, the author lacks conventional credentials, the topic has no established panel, and the expected impact cannot be expressed through standard short-term metrics.

Yet these facts do not prove that the work is wrong.

A system designed to minimize institutional risk can systematically reject work whose value is difficult to classify.

Governments Can Politicize Research Priorities

Governments must choose priorities because resources are finite. Public agencies may reasonably concentrate funding on public health, energy, climate, infrastructure, defense, agriculture, or strategically important technologies.

The danger appears when priority setting becomes excessively centralized.

When large funding programs target a limited number of officially favored themes, universities reorganize around those themes. Researchers rewrite proposals using the preferred vocabulary. Departments hire people who can attract the designated funding. Graduate students choose topics with visible financial support.

The result can be a scientific monoculture.

Important questions outside the current policy agenda receive less attention—not necessarily because they are less valuable, but because their value is more difficult to connect to a government program.

Large, coordinated research programs can solve problems that scattered individuals cannot. However, the National Academies has also documented concerns that public funding of large scientific teams can transfer excessive control over research questions, methods, and goals from scientists to administrative authorities.

Political influence does not always mean direct censorship. It may operate through the quieter mechanism of agenda selection: governments do not need to prohibit a question when they can simply decline to fund it.

Short Grant Cycles Distort Long-Term Research

Some discoveries require decades of work. Government grants usually operate on much shorter cycles.

A scientist may receive funding for two, three, or five years and then need to apply again. This encourages researchers to divide ambitious programs into grant-sized units with frequent measurable outputs.

Consequently, scientists may prioritize:

• publishable fragments over comprehensive theories;
• rapid data collection over long-term observation;
• multiple small papers over one definitive work;
• fashionable applications over foundational infrastructure;
• and positive results over careful negative findings.

Long-term intellectual work becomes difficult because the researcher must repeatedly stop to justify its continuation.

The funding cycle can therefore change the unit of scientific thought. Instead of asking, “What problem should be solved?” researchers are encouraged to ask, “What part of this problem can produce fundable deliverables during the next grant period?”

Competitive Funding Can Contribute to Publication Bias

Grant recipients are expected to demonstrate productivity. Institutions want papers, citations, follow-on funding, patents, media attention, and measurable impact.

These expectations can make null results, failed experiments, replications, and corrections professionally unattractive.

Publication bias occurs when published research does not represent all the studies that were conducted—for example, when positive or statistically significant results are more likely to appear than negative findings. The National Academies describes publication bias as a situation in which the published literature becomes an unreliable summary of the actual evidence.

Government funding does not create publication bias by itself. Journals, universities, commercial incentives, and academic promotion systems all contribute. But project-based funding can reinforce it when continued financial support depends on demonstrating visible success.

A failed hypothesis can be scientifically valuable. It eliminates an incorrect possibility and prevents others from repeating the same error. In a competitive grant system, however, a failed hypothesis may look like poor project performance.

The funding system thus risks rewarding the appearance of continuous success rather than the honest elimination of error.

More Government Money Does Not Automatically Solve the Problem

It is tempting to conclude that all these problems result from insufficient budgets. More money would increase success rates and reduce competition.

That would help, but it would not eliminate the structural defects.

A larger version of the same grant system could still:

• impose extensive administrative requirements;
• concentrate resources at prestigious institutions;
• favor predictable proposals;
• exclude independent researchers;
• prioritize fashionable topics;
• and evaluate promised impact rather than demonstrated usefulness.

The central question is not only how much money science receives, but how scientific value is identified.

Why Abolishing Public Research Funding Would Also Be a Mistake

Criticism of government grant systems should not be confused with an argument for eliminating public science funding.

Private markets systematically underfund research whose benefits are:

• uncertain;
• long-term;
• difficult to patent;
• widely shared;
• or impossible for one company to capture.

Fundamental mathematics, replication studies, public datasets, rare-disease research, open-source scientific software, environmental monitoring, and basic physics may generate enormous social value without generating immediate private revenue.

Removing government funding could make science even more dependent on wealthy corporations, private donors, and short-term commercial priorities.

The better objective is therefore not less public support, but more pluralistic, transparent, and outcome-sensitive funding.

How Research Funding Could Be Improved

No single mechanism should control all science. A resilient funding ecosystem should use several complementary models.

1. Distribute More Small Grants

Smaller grants can support more researchers, reduce the cost of individual mistakes, and allow unusual ideas to receive initial testing.

Instead of attempting to predict the few projects that will become major successes, funders can support a larger portfolio and increase support when evidence of value appears.

2. Use Partial Lotteries Among Qualified Proposals

When expert reviewers cannot reliably distinguish between many similarly rated proposals, a lottery among proposals above a quality threshold may be fairer than pretending that committee rankings are exact.

A partial lottery can also reduce strategic grant writing and weaken prestige bias while retaining technical screening.

3. Fund Researchers for Longer Periods

Longer and more stable support would allow scientists to pursue difficult questions without constantly rewriting their plans to satisfy the next application cycle.

Evaluation could focus on a body of work rather than whether each project followed its original timetable.

4. Support Independent Researchers

Funding eligibility should depend more on the proposed or completed work and less on institutional affiliation.

Financial controls can be provided through fiscal sponsors, transparent payment systems, milestone contracts, or specialized nonprofit organizations without requiring every serious researcher to become a university employee.

5. Reward Replication, Negative Results, and Infrastructure

Datasets, proofs, replications, corrections, research software, and well-documented failures are scientific outputs even when they do not produce spectacular headlines.

Funding mechanisms should explicitly value these contributions.

6. Allocate Some Funding After Results Exist

Conventional grants attempt to predict value before the work is completed. A complementary system can reward useful work after it becomes inspectable.

Post-publication funding does not eliminate uncertainty, but it changes its location. Instead of asking committees to imagine whether a proposal will become valuable, evaluators can examine code, data, proofs, experiments, adoption, citations, replications, and dependencies that already exist.

AIIM: Funding Demonstrated Scientific Contribution

AI Internet Meritocracy (AIIM) proposes an alternative model in which artificial intelligence helps evaluate scientific and open-source contributions and distribute funding according to their demonstrated usefulness.

Rather than relying entirely on a small pre-publication grant committee, such a system could examine a broader range of evidence:

• completed articles and books;
• mathematical dependencies;
• software packages and their downstream use;
• corrections and replications;
• datasets and research tools;
• expert evaluations;
• and relationships between one contribution and later work.

This approach could make funding more accessible to researchers who are excluded by credentials, geography, institutional status, or unconventional subject matter.

AI assessment would not automatically be objective. Algorithms can reproduce bias, misunderstand unusual work, or be manipulated. Therefore, an AI-based funding system must be transparent, auditable, contestable, and governed by people rather than treated as an unquestionable authority.

AIIM proposes combining automated analysis with decentralized governance and human voting. More information is available in the project’s explanation of merit-based scientific funding and its proposal for government adoption.

The important innovation is not merely the use of AI. It is the shift from funding promises primarily through institutional prediction toward rewarding visible contributions through continuously updated evidence.

Governments Should Finance Science Without Controlling Its Shape

Government research funding makes science worse when it tries to define, predict, administer, and rank too much of the scientific process.

Public funding should provide researchers with resources while preserving intellectual diversity. It should tolerate failure, support long time horizons, admit outsiders, reward corrections, and avoid treating committee consensus as a substitute for truth.

The current system often asks:

Which applicants can persuade us that their proposed projects will succeed?

A better system would also ask:

Which contributions have proved useful, which neglected ideas deserve testing, and how can resources reach capable researchers without forcing all science through the same institutional gate?

Governments should continue financing science. But they should stop assuming that traditional grant committees are the only legitimate mechanism for doing so.

A scientific system becomes stronger when no university, agency, journal, algorithm, or committee has a monopoly on deciding what knowledge deserves to exist. Supporting scientific discovery through meaningful contributions.