You are on page 1of 3

POLICY Why governments

COMMENT EPIGENETICS New tools to MUSIC Icelandic singer FUNDING Universities should
should measure well-being address old genomic Björk combines music spend less on bureaucracy
and promote the good life p.532 questions p.534 and nature in an app p.537 and more on researchers p.538
ILLUSTRATIONS BY JONATHAN BURTON

Fund people not projects


John P. A. Ioannidis proposes ways to save scientists from
spending all their time writing grants.

T
he research funding system is broken: and hopes for a solution. Although detailed system overhauls, which are likely to make
scientists don’t have time for science proposals may be indispensable for some some scientists justifiably nervous. But
any more. Because they are judged projects, such as rigorous clinical trials and smaller, pilot efforts that enable us to evalu-
on the amount of money they bring to their large-scale collaborative research, ideas ate what works could begin right away.
institutions, writing, reviewing and admin- abound for more efficient ways to fund
istering grants absorb their efforts1. The general research. Some organizations are FUND EVERYBODY (OR A LUCKY FEW)
requirement that they promise taxpayers already experimenting. Multiple options Some — or all — of the research budget
specific results to justify research tends to could co-exist, with portions of the budget could be allocated to eligible scientists in
invite either exaggeration or boringly pre- earmarked for different schemes. equal shares, or given to a few lucky ones
dictable projects. Yet the research behind Here are some of the most promis- at random. With egalitarian sharing, each
30% of the pivotal papers from Nobel lau- ing proposals to reduce the amount of scientist would receive only a small amount,
reates in medicine, physics and chemistry time scientists spend trying to fund their which could quickly evaporate without
was done without direct funding2. research, and the pros and cons of each (see returns when research costs are high. But
Every scientist recognizes this problem table). Definitive fixes would require major scientists in some fields — mathematics,

2 9 S E P T E M B E R 2 0 1 1 | VO L 4 7 7 | NAT U R E | 5 2 9
© 2011 Macmillan Publishers Limited. All rights reserved
COMMENT

say — could achieve much on a small In the MacArthur Fellows Program, will rely even more on indices when it comes
share; and in some settings the shares could for example, meticulous peer assessment into effect in 2014. Metrics also underlie
be substantial. For instance, if half of the is used to select 20–30 individuals a year many hiring decisions in Italy, a country
US$31.2 billion that the US National Insti- on the basis of exceptional creativity and that is struggling to remedy widespread
tutes of Health spends on research each year promise for important advances. Recipi- nepotism, and in Germany’s Max Planck
was shared equally among 300,000 research- ents do not have to justify what they do institutes. However, most of these assess-
ers, each would get more than $50,000 a year. with the $500,000 award, which is spread ments are simplistic, focusing on number
Lottery distribution, too, flies against the over 5 years. However, close scrutiny of an of peer-reviewed publications, or inappro-
principle that science funding should be individual’s career may become prohibitive priate — looking at the impact factor of the
meritocratic. Still, some agencies are trying for systems that award thousands of grants journal rather than of an individual article.
it — the Foundational Questions Institute — it might save grant recipients time, but it More sophisticated formulae are needed if
in New York, which tackles key questions adds to the administrative load of reviewers. a scientist’s merit is to be captured.
in physics and cosmology, uses a lottery sys- The approach is also vulnerable to favour- Furthermore, indices are open to
tem to award its mini-grants, which range itism, in which only elite individuals and gaming, although some are more difficult
in value from $1,000 to $15,000. Such an lines of research to influence than others. To counter this,
approach may not be as radical as it sounds: “It is a scandal are selected and the system could use indices that exclude
the imperfections of peer review mean that thousands of sci- self-citations and capture quality rather
that billions of
as many as one-third of current grants are entists doing qual- than quantity (such as average citations
effectively being awarded at random3. This
dollars are spent ity, smaller-scale per paper instead of number of papers),
situation will only worsen as falling accept- on research science are left out. discourage gift authorship by adjusting for
ance rates encourage investigators to bom- without To avoid t he co-authors and penalize quantity that is not
bard agencies with proposals, leaving fewer knowing the subjectivity and accompanied by quality. Several metrics
qualified reviewers to judge each one. The best way to burden of evaluat- could be combined.
downside of using aleatoric allocation is that distribute that ing thousands of Funding systems could reward good
not every deserving scientist will be funded. money.” careers, an auto- scientific citizenship practices, such as
mated system to data sharing4, high-quality methods, care-
FUND ACCORDING TO MERIT evaluate relative merit would have to be ful study design and meticulous reporting
Leading thinkers and experimenters worthy devised. Such a system would depend on of scientific work5. Openness to collabora-
of unconditional support could be identi- objective indices. The share of the annual tion, non-selective publication of ‘negative’
fied through peer assessment of their work funding budget scientists receive would be findings, balanced discussion of limitations
and credentials. Appraisals of project-based based on their value, calculated with a pre- in articles and high-quality contributions
proposals already take a scientist’s merit into specified formula. to peer-review, mentoring, blogging or
account, but they typically give less weight to Metric-based appraisals are familiar to database curation could also be encour-
it than to the project plan. Peer assessment many scientists already, particularly those aged. Researchers might be rewarded for
does not work well for early-career scientists, in European countries. The UK Research publishing reproducible data, protocols
who have a short track record. But for those Assessment Exercise, for example, relies on and algorithms. However, some citizenship
more established in their field, a career trajec- them. It is a much hated and debated system practices are difficult to capture in auto-
tory offers a wealth of information. By con- for evaluating departments, but its replace- mated databases, so would be subject to the
trast, an isolated project is only a snapshot. ment, the Research Excellence Framework, disadvantages of peer assessment.

OPTIONS FOR REVAMPING THE FUNDING SYSTEM


Option Pros Cons Example Who would be funded?
Egalitarian Avoids peer-review biases Does not support large research efforts Some universities All
(fund everybody) Gives sufficient amounts to Does not recognize exceptional scientists fund the salaries of
scientists doing low-cost research all their faculty
Small administrative burden
Aleatoric Avoids peer-review biases Will not capture all deserving scientists Foundational Flexible
(fund at random) Small administrative burden Questions Institute
Assessment of Captures career trajectory Is vulnerable to favouritism MacArthur Fellows Few elite scientists
career Has gold-standard status Inappropriate for young researchers Program (or else administratively
Is labour-intensive burdensome)
Automated impact Eliminates favouritism There are many indices, all with flaws; no consensus UK Research Flexible
indices Evaluates many applicants with about best one to use Excellence
ease Indices can be gamed Framework
Approaches objectivity Databases have shortcomings (such as
imperfect citation coverage, entry errors, name
disambiguation problems)
Scientific citizenship May improve science, if good Automation is not yet possible for data gathering, Financial incentives Could be extended to
practices are rewarded and bad and is difficult for some citizenship practices to peer reviewers many scientists only
ones penalized Has peer-review biases for aspects that can be
automated
Projects with Proposals are easy to write and Does not eliminate project proposals NIH Director’s Few elite scientists
broad goals review Is vulnerable to favouritism Pioneer Awards
Formulating work can be flexible Holds potential for exaggerated promises and Howard Hughes
Permits targeted innovation claims Medical Institute
Two or more options can also be combined (for example, automated impact indices plus evaluation of scientific citizenship).

5 3 0 | NAT U R E | VO L 4 7 7 | 2 9 S E P T E M B E R 2 0 1 1
© 2011 Macmillan Publishers Limited. All rights reserved
COMMENT

There are issues still to be resolved. All


funding options face a tension over how
many scientists should receive awards, and
there is no good evidence on whether it is
better to give fewer scientists more money
or to distribute smaller amounts between
more researchers. Some funding schemes
are well suited to funding numerous scien-
tists; others favour elitism (see table).
We will need to find ways to figure out
which approach works best. It is a scandal
that billions of dollars are spent on research
without knowing the best way to distribute
that money. Retrospective assessments are
easy, but subject to confounding when com-
paring groups that were funded through
different schemes. For example, one study
has found that HHMI-funded investiga-
tors publish more high-impact papers
and get more recognition than matched
NIH-funded peers6. But it’s impossible to
match scientists perfectly: the prestige of
the HHMI name alone may lead to more
peer-based recognition. Prospective com-
parisons are more reliable, but require long
follow-up. For example, controlled trials
could randomize consenting scientists to
different funding schemes, then compare
surrogate metrics and long-term successes.
Ultimately, funding schemes should sup-
port the long-term goals of science. Few
isolated research efforts have an immediate,
substantial and durable impact; successful
translation of basic research to practical
applications occurs sparingly and with
STATE BROAD GOALS funding to test. But going after many and average delays of almost three decades7.
Another way to save time would be to sim- expensive grants costs institutions money, The aim of science is to expand our knowl-
plify the application. Researchers could be because both scientists and administrators edge base, which, eventually, yields useful
asked, for example, to submit short summa- spend an inordinate amount of time evalu- applications. This is what scientists entered
ries of their intended research, describing ating proposals, supplements, timesheets their profession to do, so requiring them
broad goals only. Such applications require and justifications, as well as progress and to spend most of their time writing grants
less effort to write, review and administer final reports. is irrational. It’s time to seriously consider
and would allow flexibility in carrying out Many large projects never result in a another approach. ■
the work, if funded. Examples include the scientific achievement, so even if the strategy
NIH Director’s Pioneer Awards and the brings in short-term grant funding, it may John P. A. Ioannidis is the C. F. Rehnborg
Howard Hughes Medical Institute (HHMI) not pay off in the long term. The size of a chair in disease prevention, professor of
grants. The HHMI selects 300 established portfolio should therefore not be a criterion medicine and of health research and policy,
investigators and 50 young investigators for promotion; committees should focus and director of the Stanford Prevention
through peer assessment of their creden- instead on real work and achievements. Research Center at Stanford University
tials and of proposals for high-risk, uncer- Judging scientists by the size of their portfo- School of Medicine, and professor of statistics
tain prospects of innovation. Its awards are lio is equivalent to judging art by how much at Stanford University School of Humanities
usually renewed after 5 years, requiring money was spent on paint and brushes, and Sciences, Stanford CA 94305.
simple documentation of effort rather than rather than the quality of the paintings. e-mail: jioannid@stanford.edu
demonstrated results, although results are
1. Editorial. Sci. Am. 304, 10 (2011).
needed for renewal after 10 years. Never- WHAT WE CAN DO NOW 2. Tatsioni, A., Vavva, E. & Ioannidis, J. P. FASEB J.
theless, any system that demands high-risk All of the options above could be achieved 24, 1335–1339 (2010).
innovative goals, and requires results, gen- — either through small, progressive steps, 3. Graves, N., Barnett, A. G. & Clarke, P. Nature 469,
299 (2011).
erates potential for exaggeration. or through more extensive changes to the 4. Baggerly, K. Nature 467, 401 (2010).
system. A major overhaul is likely to take 5. Simera, I., Moher, D., Hoey, J., Schulz, K. F. &
IGNORE GRANT PORTFOLIOS IN PROMOTIONS years to implement, and will meet with wide Altman, D. G. Eur. J. Clin. Invest. 40, 35–53
(2010).
Many institutions use the size of a scientist’s resistance and debate. Smaller steps, such as 6. Azoulay, P., Graff Zivin, J.S., & Manso, G.
grant portfolio as a basis for tenure and pro- changing promotion and tenure criteria, Incentives and Creativity: Evidence from the
motion. This practice may prompt scientists are easier to make. Pilot programmes of Academic Life Sciences (NBER Working Paper
No. 15466. 2009).
to prepare multiple grant applications focus- proposal-free or broad-goal-based funding 7. Contopoulos-Ioannidis, D. G., Alexiou, G. A.,
ing on expensive, even if dull, projects and can be incorporated into existing funding Gouvias, T. C. & Ionnadis, J. P. A. Science 321,
to abandon brilliant ideas that need limited structures. 1298–1299 (2008).

2 9 S E P T E M B E R 2 0 1 1 | VO L 4 7 7 | NAT U R E | 5 3 1
© 2011 Macmillan Publishers Limited. All rights reserved

You might also like