Measuring people’s preferences as a case study in shifting power

When a funder must prioritise between saving lives, reducing illness, or raising incomes, whose values decide? 

We think GiveWell is unique among global development funders in basing these trade-offs on explicitly stated moral values, rather than internal political debates. GiveWell directed $397m to global health and development programs in 2024, so the process for allocating this money is a question with a large impact on the world.

IDinsight supported GiveWell again this year as it continued considering how to accurately incorporate the preferences of the people its grants serve into funding allocations. 

We treat that methodological question as a practical power shift: building moral trade-offs around what people in poverty say matters most to them. We’ve come to view moral weights as a frontier experiment, imperfect and sensitive to framing, yet morally important precisely because it centres beneficiary preferences. This is in line with the Dignity Initiative’s 2025 focus on how organisations set cultures of dignity, turning principles into methods and decisions.

How GiveWell used to incorporate moral values

The first versions of GiveWell’s cost-effectiveness models relied heavily on moral weights suggested by their staff. No doubt staff thought hard about what the people they were serving would want, and many grounded their suggestions in an empirical literature on those preferences, but they were in many ways distant from the experiences of people living in poverty, and the literature they used had important limitations.

Therefore in 2017, they made an important change. They resolved to start basing their moral weights at least in part on their own, new empirical data about the views of the people their efforts serve. GiveWell commissioned IDinsight to determine the best way to measure this, because a decision of this consequence needs to be informed by precisely collected data. Since 2019, the organisation has used the survey results as one important input in its moral weights.

We think GiveWell’s unusual approach might make them among the world’s most democratic private philanthropic foundations, with the views of people living in poverty significantly shaping their funding decisions.

The story of the 2019 and 2025 projects

2019: First large-scale recipient preferences

In 2018–19, IDinsight surveyed 1,820 low-income respondents in Ghana and Kenya, using a mix of Value of a Statistical Life (VSL) questions and community-perspective choice experiments on lives saved vs. cash. We found higher valuations of averting death than prior extrapolations, and a consistent preference for saving under-5 lives over older lives. The preferences of people living in poverty pushed GiveWell to give more funds to health charities, especially those that save lives of young children, compared to programs that enhance livelihoods.

The work also exposed tough methods issues, notably a large share of “never-switchers” in the choice tasks – people who never moved off their first principle (e.g., always choosing to save more lives), which complicates estimation and interpretation.

2025: Pilots to strengthen credibility

In 2025, GiveWell commissioned IDinsight to revisit recipient preferences. We looked at three types of trade-offs: saving lives vs cash, reducing sickness vs cash, and accessing family planning vs cash. We ran three rounds of piloting in Kenya and India to make estimates more credible and decision-useful. We tested hypothetical-bias reductions (cheap talk, time-to-think), comprehension checks (teach-back, certainty scales), and incentive-compatible mechanisms (e.g., inferred valuation with rewards). These methods, which seemed promising, produced no gains in scope consistency and accuracy relative to 2019, and many respondents still prioritised life-saving over even large amounts of cash. In qualitative responses explaining this, many offered deontological moral commitments. 

For those using moral weights and guiding their work according to the preferences of the people their work serves, that leads us to conclude that the best way to do so is to use the methods we outlined in 2019.

Research on the trade-offs involved in saving lives and reducing sickness is not new. Work on how people value access to family planning is much less established. For family planning, a contingent valuation method paired with a guided visualisation showed strong engagement and appears more promising than standard choice experiments, which produced near-universal never-switching when the counterfactual was “no access to contraception.” We recommended further iteration, including more participatory/group valuation options.

Impact and what has changed

After the first large-scale survey in 2019, the conversation about moral weights at GiveWell has shifted significantly. Back then, skepticism was high: many decision-makers at GiveWell wondered whether the data truly reflected beneficiary preferences or just confusion in the survey process. While the results of the 2025 project mirror many of the findings from 2019 and there are remaining uncertainties, we think the 2025 project helped address some doubts and move the agenda forward.

Acting on recommendations. We understand that GiveWell is preparing to collect additional data on these preferences in upcoming surveys, using recommendations from these pilots. The 2025 Kenya results lined up fairly closely with the 2019 findings, boosting confidence in contingent valuation as a credible way to capture how people value life-saving interventions. These pilots have also helped rule out methods which don’t seem promising in low-income contexts – such as the Bayesian truth serum – which will help focus efforts in future data collection. GiveWell will also explore other approaches to collecting information on trade-offs individuals in these countries would make (e.g., by  investing in revealed-preference methods to learn from the choices people make in real life). 

Why this matters

Prioritisation is power. When models use only expert or staff judgments, they risk drifting away from the lived values of people affected by those decisions. The Dignity Initiative’s 2025 report asks how organisations set cultures of dignity. GiveWell has a unique method of setting program priorities, through its system of explicit moral weights. They therefore affirm dignity by building into their system deference to recipient preferences and by upgrading methods so those preferences are credible enough to steer funding. This aligns with our broader push to institutionalise program prioritisation as a dignity practice inside organisations, not just in one study, so choices reliably reflect the priorities of the populations served.

Looking forward, IDinsight believes the focus for funders emulating this method should be less on perfecting any single survey and more on building systems that sustain these voices over time. That means adding short modules to high-quality surveys, expanding into new geographies where funding decisions are most critical, and testing and refining tools that make these exercises more straightforward and reliable to understand.

Equally important is the mindset shift: moral weights should be seen as an ongoing democratic practice rather than a one-off measurement. The work ahead is about embedding the voices of those served by aid into the everyday machinery of decision-making, recognizing both the power and limitations of these methods.