Two big critiques of GiveWell's charity evaluation + 6 lessons learned | Elie Hassenfeld (2023)

GiveWell Organizational Growth and Strategy

• GiveWell is raising approximately $600 million annually in 2022, managing a budget that has expanded from a focus on the Against Malaria Foundation to a broader portfolio.
• The organization now employs 60 staff members, with roughly 35 focused on research, 15 on outreach/fundraising, and the remainder on operations.
• GiveWell has shifted from a static list of recommendations to directing over $100 million annually to non-top charities in areas like malnutrition, water, and policy.
• The organization aims to direct marginal dollars to the highest cost-effective interventions, currently maintaining a threshold of 10x the cost-effectiveness of cash transfers (up from 6-7x in late 2021).
• GiveWell anticipates potential annual growth to $700–$900 million in the coming years, though financial market uncertainty presents a risk of stagnation.
• The organization is hiring senior researchers, program officers, operations staff, and entry-level researchers, with a specific focus on remote work capabilities across US Pacific to European time zones.
• GiveWell has adopted a "remote-first" model, with only one-third of staff based in the Bay Area, following a strategic shift after the pandemic.

Empirical Investigations and Specific Interventions

• Dispensers for Safe Water (Evidence Action): GiveWell reversed a 2019 decision to fund this program after a 2021 meta-analysis by Nobel laureate Michael Kramer revealed chlorination reduces all-cause child mortality more significantly than diarrheal deaths alone, raising its cost-effectiveness estimate from ~3x cash to >10x cash.
  • The organization plans to fund a large randomized control trial for water vouchers to better understand mortality impacts.
  • GiveWell identified that only ~50% of eligible users utilize chlorine dispensers, highlighting the implementation gap even with physical infrastructure.
• Syphilis Screening (Evidence Action): A $4 million grant was made to support technical assistance in Liberia, Cameroon, and Zambia to switch from single HIV tests to dual HIV/syphilis tests.
  • This program is estimated to be 10–30x as cost-effective as cash transfers due to reliance on government systems rather than commodity purchase.
  • GiveWell views this "technical assistance" model as high-leverage because it ensures existing government funds are used for more effective diagnostics without requiring ongoing external funding for commodities.
• Kangaroo Mother Care (Research Institute for Compassionate Economics): A grant supports training nurses and securing hospital space for skin-to-skin contact for low-birth-weight infants in Uttar Pradesh, India.
  • The program reduces neonatal mortality by 20–30% compared to hospitals with NICU warmers.
  • The estimated cost-effectiveness is ~13x cash ($400 per infant reached), driven by the high mortality risk of the target population and the relatively low cost of implementation.
  • Challenges to standard practice include high hospital staff turnover and a lack of available space, which the grant addresses through dedicated funding for staff and infrastructure.
• Miracle Feet (Clubfoot Treatment): GiveWell funded this organization to treat clubfoot using the Ponseti method, which has a cost of $100–$200 per child.
  • The intervention avoids significant "offsetting impacts" because it relies on casting and bracing rather than scarce specialist surgical capacity.
  • GiveWell's primary reservation is uncertainty regarding the marginal number of children reached due to the high fixed costs of setting up country offices.
• Malnutrition Treatment (Alliance for International Medical Action): A $1 million grant supports treatment for severe acute malnutrition in Niger.
  • GiveWell initially estimated high cost-effectiveness but corrected a modeling error in a meta-analysis regarding follow-up times, reducing the estimated benefit from "exceptional" to ~10x cash.
  • The analysis relied on observational data from the 1970s–90s, as recent randomized trials are difficult to power due to lower baseline mortality rates.
• Vitamin A Supplementation (Helen Keller International & Nutrition International): GiveWell continues to fund this program in high-mortality, high-deficiency contexts, despite evidence of near-zero effects in large trials in India.
  • The organization is considering a new large-scale RCT to assess current efficacy but cites the high cost and difficulty of powering trials against low modern mortality rates as barriers.

Theoretical Critiques and Philosophical Debates

• Subjective Well-being vs. Health/Income: GiveWell faces critique for prioritizing lives saved and income doublings over subjective well-being (SWB) metrics proposed by the Happier Lives Institute.
  • Ellie Hassenfeld acknowledges SWB is important but cites measurement difficulty and the "squishiness" of survey data compared to mortality.
  • GiveWell applies empirical adjustments to data, such as discounting Strong Minds' peer therapy results by 20% to account for social desirability bias in self-reported depression scores.
  • GiveWell also adjusts for the risk that program effects diminish at scale (fidelity loss), whereas some critics assume trial results hold at 100% implementation.
• Economic Growth vs. Direct Aid: GiveWell responds to the argument that funding economic policy and think tanks is superior to direct health interventions due to the long-term multiplier of GDP growth.
  • Hassenfeld argues that the consensus on specific growth policies is weak and that the risk of harm from unintended consequences is high.
  • The organization notes that many poor countries adopt bad policies due to political economy constraints (elite power) rather than ignorance, making external advocacy difficult.
  • GiveWell acknowledges a need for more research capacity to explore growth-focused opportunities but currently lacks the data to confidently fund specific policy organizations over direct health interventions.
• Value of Life vs. Well-being: GiveWell addresses the critique that saving lives is less valuable if the quality of life is low (e.g., a life scored 4/10 vs. 5/10).
  • Hassenfeld expresses discomfort with the implication that many people in poverty would prefer death, citing survey data where nearly everyone values continued existence.
  • GiveWell currently uses a "deprivation approach" to value death as the loss of current life satisfaction points rather than a philosophical zero, though this area is identified as a "mature" part of their work that requires further improvement.

Operational Capabilities and Culture

• GiveWell admits to "legibility" challenges, where their cost-effectiveness spreadsheets are difficult for outsiders to navigate, and is developing a "Change Our Mind" contest to identify errors.
• The organization values "truth-seeking" but explicitly fights against a culture of harshness, prioritizing kindness and consideration for grantees and colleagues despite a focus on rigorous quantification.
• GiveWell estimates that 2022 funding levels will remain similar to 2021 despite financial market volatility, a result they view as successful.
• The organization is moving toward hiring specialized researchers, such as a dedicated malaria researcher, while maintaining a generalist approach for most roles.
• GiveWell's "All Grants Fund" allows donors to support emerging opportunities like water and malnutrition, distinct from the "Top Charities Fund" which focuses on high-confidence interventions.