Why Foundations Make Worse Decisions Than They Should, and What Behavioral Science Can Do About It

Last updated October 8, 2026. Reviewed quarterly.

Exponent Philanthropy's 2026 Foundation Operations and Management Report found that 4 in 5 participating foundations made single-year grants of general operating support, the flexible funding a nonprofit can spend wherever it is most needed. About 1 in 3 made that same kind of grant for more than one year, even though foundation leaders surveyed by the Center for Effective Philanthropy overwhelmingly endorsed multiyear general operating support in principle.

The gap between what funders believe and how they allocate is a decision problem, and behavioral science is the discipline built to study decisions. At The Decision Lab, we study how the people inside foundations decide where money goes: program officers, review panels, investment committees and boards. In our impact strategy and governance work with funders including the Skoll Foundation, the Open Society Foundations, the Gates Foundation and the World Bank, the largest gains have come from redesigning the decision process at the specific points where it departs from a foundation's own strategy.

Behavioral science for evidence-based philanthropy is the use of research on how people actually decide to design grantmaking processes that hold up against predictable bias. Most behavioral writing on philanthropy covers donors and how to persuade them to give. This guide covers the other side of the table: the five blind spots we see most often in institutional grantmaking, the evidence behind each one, what to change, and a self-assessment your team can run on its next grant cycle.

1. Choice overload. Long dockets of proposals that are hard to compare push reviewers toward deferral and familiar renewals.
2. Present bias and habit. A board feels the cost of a multiyear commitment now, while the benefit arrives later and lands with the grantee.
3. Social proof in co-funding. A respected co-funder's commitment starts to stand in for a foundation's own due diligence.
4. Identity and familiarity. Referral networks and the language an application is written in shape who reaches the shortlist before anyone assesses merit.
5. Reviewer noise. Reviewers scoring the same proposal disagree far more than most foundations assume, and highly novel proposals score lower than conventional ones.

Choice overload is the drop in confidence and follow-through that can come from facing too many options at once. The evidence for it is conditional. An earlier meta-analysis led by Benjamin Scheibehenne found no average effect of the number of options. A 2015 meta-analysis by Alexander Chernev and colleagues, covering 99 results and 7,202 people, found that overload appears reliably under four conditions: the options are complex and hard to compare, the decision itself is difficult, the chooser is unsure what they prefer, and the goal is to minimize effort.

A grant review docket meets all four conditions. Proposals differ in geography and theory of change, and the weight each criterion carries often exists only in reviewers' heads. In that setting we would expect the easiest options to win, such as renewing a known grantee or tabling a hard call until the next meeting. Neither is visible as a bias in the minutes.

Give every proposal the same one-page summary on the same criteria, and agree the weights before anyone reads the docket. Then compare proposals within a cause area before comparing across areas, and cap the number of proposals the panel discusses in one sitting. Each year the Skoll Foundation sorts through hundreds of applications for the Skoll Award for Social Entrepreneurship, and when it asked us to identify bias in that selection process, we started by separating it into four parts: sourcing, screening, selection and general considerations. A bias that is hard to see across a whole process is often obvious inside one stage.

Present bias is the tendency to give more weight to costs and benefits that arrive now than to those that arrive later. For a foundation board, the cost of a multiyear commitment is immediate, a loss of flexibility in next year's budget. The benefit arrives over several years and mostly shows up in the grantee's ability to plan and hire.

The Center for Effective Philanthropy's 2020 report New Attitudes, Old Practices found that only about 2 in 5 grantee leaders surveyed had received multiyear general operating support in the year before the pandemic, despite broad support for it among foundation leaders. Researchers at the center found little evidence that the usual objections were holding funders back, and pointed to organizational inertia instead. Exponent Philanthropy reached a similar conclusion in 2026, describing annual cycles and project grants as familiar habits more than positions anyone defends. We read that as status quo bias layered on present bias: the annual grant is the default, and every multiyear grant needs someone willing to argue against it.

Make the multiyear grant the default for grantees who have passed a first renewal, with an annual check-in in place of a new application. At the moment of the grant, write down what evidence would lead the foundation to reduce or end support. A pre-agreed exit rule answers the accountability concern that keeps many boards on annual cycles, and it is easier to apply than a renewal debate held under time pressure. When the Open Society Foundations' Strategy Unit wanted to tighten the link between strategy and spending after a 2017 endowment of $18 billion, we studied the patterns in how its strategies had been set and built skill-building resources to tighten the link between strategy and action.

Social proof is the tendency to treat other people's choices as evidence of what is right, and it grows stronger as a decision gets more uncertain. Between funders it is partly rational, because a co-funder that has done careful due diligence is passing on real information. The risk starts when another funder's decision becomes a substitute for a foundation's own assessment.

In a 2019 Bridgespan Group study of funder collaboratives, published in Stanford Social Innovation Review by Alison Powell and colleagues, more than 3 in 5 grantees of collaboratives built around backing strong organizations said that attracting money from other funders set those collaboratives apart. One grantee described funders investing after seeing that collaborative members already had. The authors also warned that concentrating capital in collaboratives can magnify blind spots and implicit biases. When the same few judgments steer more money, an error in any one of them travels further.

The same dynamic runs inside a single review panel, where in our experience the first confident opinion often sets the frame for the rest. Among the nine behavioral barriers we found in the Skoll Foundation's selection process was referee influence, the pull exerted by the people who recommend or vouch for an applicant. When Health Canada asked us to help its organ donation and transplantation steering committee agree on a governance framework, we collected each member's priorities separately through a discrete choice experiment, a survey in which people repeatedly choose between options that differ on several features at once, and only then brought the committee together to adjust the plan in a structured workshop.

Record each funder's or reviewer's independent assessment before anyone shares a position. When a co-funder's commitment comes up in discussion, ask what that funder learned in due diligence that your team has not.

Funders tend to back leaders who resemble the leaders they already know: people who arrive through trusted referral networks and describe their work in the funder's own terms. Each step feels like ordinary judgment, and the pattern only shows up when outcomes are counted across a whole portfolio.

Echoing Green and The Bridgespan Group counted those outcomes in their 2020 report Racial Equity and Philanthropy. Among about 160 applicants who reached the semifinalist stage of Echoing Green's fellowship or beyond, Black-led organizations had revenues about a quarter smaller than white-led organizations, and unrestricted net assets about three quarters smaller. Within the fellowship focused on Black men and boys in the United States, Black-led organizations had unrestricted net assets less than a tenth the size of white-led ones doing the same work. These organizations had already cleared a selective screen, so the gap is hard to explain by quality alone.

In the Skoll Foundation's selection process we found two structural barriers alongside the nine behavioral ones, including language bias, the advantage that goes to applicants who write fluently in the funder's working language. Our 14 design recommendations ranged from region-specific applications in local languages to structured feedback after each application round to keep selection panels objective. Two checks apply to almost any foundation: track pass-through rates by region and by leader background at every stage, and widen sourcing beyond referrals before the shortlist forms. Our guide to choice architecture in grantmaking covers the application side in more detail.

Noise is the variation in judgments that should be identical, such as two qualified reviewers scoring the same proposal. In our experience, few foundations measure it. When Elizabeth Pier and colleagues replicated the US National Institutes of Health grant review process in a 2018 study in the Proceedings of the National Academy of Sciences, 43 reviewers rated the same 25 applications and showed no agreement on their quality. The authors concluded that the outcome depended more on which reviewer an application was assigned to than on the research it proposed. In Pier's comments on the study for UW–Madison, she added that collaboration among reviewers can make agreement worse.

Noise also has a direction. In a study at a leading research university, Kevin Boudreau, Eva Guinan, Karim Lakhani and Christoph Riedl randomly assigned evaluators to grant proposals, creating 2,130 evaluator and proposal pairs. Evaluators gave systematically lower scores to highly novel proposals, and also to proposals close to their own field. A foundation that says it funds bold ideas can run a review process that scores them down without anyone intending it.

Give each proposal more independent reviewers and average their scores before discussion, so that one reviewer's leaning carries less weight. Score each criterion separately instead of asking for one overall rating. If novelty is part of the strategy, ring-fence a share of the budget for high-novelty proposals and judge them against criteria written for that purpose.

We work with foundations and their grantees to identify behavioral barriers, run research that combines qualitative and quantitative methods, test interventions and build systems that help people use evidence during implementation. The qualitative work shows how a decision is actually made, and the quantitative work measures how often a bias appears and whether a change reduces it. A typical engagement maps the decision journey through interviews and document review, then tests the weakest points with surveys, psychometric tools or experiments run with the real decision-makers.

Education foundations often need better decisions well beyond their own grantmaking. Their grantees, district partners, curriculum providers and research organizations all make choices about evidence use, implementation, procurement, family engagement, educator adoption and program participation, and those choices decide whether a funded program reaches students. We design behavioral research and interventions that help foundations and their grantees understand those decisions and test practical ways to improve them.

• For the Gates Foundation, we tested interventions on how districts choose instructional materials with 13 California school districts in a year-long cohort, run with CalCurriculum, a partnership between Pivot Learning and EdReports.
• For EdReports, we researched how more than 500 district purchasers use evidence on curricula through hundreds of interviews and surveys, alongside a map of how education technology vendors develop products and engage with evidence.
• From that research we built the Evidence Uptake Framework, which sets out five drivers of evidence use: information should be easy to find, easy to obtain, presented with implementation in mind, framed as rewarding to use and grounded in real users and outcomes. It is written for anyone producing evidence meant to guide purchasing decisions.
• With the Collaborative for Academic, Social, and Emotional Learning, we identified the barriers keeping educators and policy leaders from acting on its recommendations and redesigned its digital tools to reach them. For the Gates Foundation, we also mapped the barriers that push teaching candidates from marginalized backgrounds out of the teacher pipeline.

This diagnostic gives a foundation a quick read on how well its grantmaking process is protected against the five blind spots above. Score each statement 0 if it is not true of your last grant cycle, 1 if it is partly true and 2 if it is fully true, then add the scores for a total out of 20.

A diagnostic like this measures how a process is designed, and it cannot show whether reviewers follow the design in practice. In our engagements, we pair it with interviews and with a review of a sample of past decisions, because the largest gaps usually sit between the written process and the one people actually run.

What is behavioral science for evidence-based philanthropy?

It is the use of research on human judgment to improve how funders allocate money. Work on donors asks how to raise more. Work on foundations asks whether program officers and boards are choosing the grants their own strategy calls for, and redesigns the process where they are not, from application formats to renewal rules.

Which biases affect foundation grantmaking most?

In our work, the most consequential are choice overload in large dockets, present bias and habit in renewal decisions, social proof between co-funders, familiarity in sourcing and noise between reviewers. How much damage each does depends on process design more than on individual judgment, so the same reviewers can decide well in one structure and poorly in another.

How is this different from donor psychology research?

Donor psychology studies individuals deciding whether to give and how much. Foundation decision research studies professionals allocating a fixed budget under institutional rules, with committees and co-funders, over multi-year horizons. The biases overlap, but the fixes differ: a donor responds to how an appeal is framed, while a foundation changes through its procedures and defaults.

Does a foundation need a randomized controlled trial to improve its grantmaking?

No. A randomized trial is the strongest test of whether a change works, but many process changes can be judged by comparing a grant cycle before and after the change, or by measuring reviewer agreement directly. A trial earns its cost when a decision rule will govern large sums over many years.

Can The Decision Lab work with a foundation's education grantees?

Yes. We work with foundations and their grantees directly, as well as with the schools and districts they serve, to study how people decide and to test interventions that change those decisions. In education, that has included district purchasing and curriculum adoption, and it extends to program participation, family engagement, educator behavior and the use of research in policy and practice.

Where should a foundation start?

Run the diagnostic above on your most recent grant cycle, pick the blind spot with the lowest score, and change one step of the process before the next cycle. We usually begin an engagement by reviewing a sample of past decisions with the people who made them, looking for cases where the written criteria and the final choice disagree.