Why it's so hard to have confidence that charities are doing good (with Elie Hassenfeld)

GiveWell focuses on a small number of highly impactful charities, primarily in low and middle-income countries, that can effectively use additional funding, rather than attempting to rate a vast number of organizations. They prioritize evidence of impact and cost-effectiveness.

GiveWell's goal is to find where charitable dollars will have the highest return on investment, not to build a comprehensive database. They prioritize depth of impact and the capacity of a few programs (like malaria prevention) to absorb large amounts of funding effectively, even if it means fewer recommendations.

GiveWell uses a synthesized approach, drawing on academic literature (e.g., valuing a statistical life), surveys of people in low-income countries and donors, and empirical analysis of cash transfers to create an 'exchange rate' between outcomes, acknowledging there's no single 'right' answer.

GiveWell's shift towards valuing health more highly relative to income has come from a synthesis of all their inputs, including donor surveys, beneficiary preference studies, and empirical work, which consistently pointed in that direction over the last decade.

GiveWell heavily relies on DALYs for assessing the burden of disease and comparing health programs, generally adopting the disability weights for conditions. However, they adjust DALYs with age weights for the very young, reflecting an intuition that the moral value of a person increases rapidly in their first few years of life.

Many donors have motivations beyond maximizing impact (e.g., local giving, social signaling). Additionally, conducting credible cost-effectiveness analysis is extremely difficult due to complex factors like displacement of other funding, and some argue that such blunt instruments miss critical aspects of impact.

GiveWell operates within an expected value framework, meaning they consider programs with significant uncertainty but potential for huge impact (like deworming). While some recommendations offer high confidence, others, like deworming, are considered 'great expected value bets' despite a higher likelihood of having limited impact compared to their maximum potential.

Initial cost-effectiveness estimates tend to be optimistic, often implicitly assuming everything goes perfectly. As more investigation occurs, numerous potential failure points (e.g., implementation issues, specific contextual factors, simple calculation errors) are uncovered, which tend to reduce the estimated impact.

GiveWell aims to model second-order effects (e.g., developmental benefits of improved health, displacement of other funding) when they are substantial, but they acknowledge limits to how far they can go. They use the model as a practical tool for ordering opportunities, prioritizing clarity and usability over exhaustive, overly complex modeling.

GiveWell's experience suggests a power law distribution, where the very best organizations have a massively outsized effect. A donation to the best GiveWell-recommended charities can be 10 times more impactful than giving to GiveDirectly (which is itself about 100 times better than an average US charity for poverty reduction).

Information is often tracked at a project level rather than an aggregate organizational level, and publicly available reports are frequently more marketing than technical evaluation. Charities are often understaffed, and conducting rigorous, counterfactual-aware evaluations requires significant time, money, and participation from the charity, which many donors do not demand.

The biggest challenge is that, unlike venture capital where financial returns clearly indicate success, it's very difficult to know which charitable programs have truly succeeded after the fact without extensive, costly evaluation. This makes it hard to identify and scale up the 'winners' in a diversified portfolio.

Consider giving to organizations that align with GiveWell's moral values (e.g., helping the poorest people in the world). Also, leverage personal connections to support small, trusted funding needs where one has high confidence in the individuals involved, as GiveWell doesn't evaluate such small-scale opportunities.

While Elie Hassenfeld personally believes animals deserve significant moral weight, GiveWell does not focus on animal welfare. This is primarily because they believe other organizations, like Open Philanthropy's farm animal welfare team and Animal Charity Evaluators, have a comparative advantage in that area, and GiveWell focuses its efforts on international giving where it believes it adds the most value.

Yes, over-quantification can be problematic due to over-reliance on numbers from complex spreadsheets that may have wide uncertainty ranges or sensitive assumptions. GiveWell addresses this by always considering a simple, qualitative case for a grant as a gut check, and by simplifying models to avoid being overly reliant on a single set of numbers.