Footnotes for "GiveWell’s 2023 recommendations to donors"

[1] $5,000 is our conservative estimate of the impact of the average donation to one of our recommended life-saving programs (via top charities or the All Grants Fund). The estimated marginal cost-effectiveness (i.e., the cost-effectiveness of the "last" grant we make) is lower, around $7,500 to save a life, but we think it generally makes sense for donors to think of the average in this context. We reached $5,000 by looking first at top charities, where the latest annual estimates range from $4,000 to $5,500 (rounded up to the nearest $500) by charity and more recent grants are in a similar range, as seen here. We then checked that $5,000 estimate against the life-saving grants we've recommended to non-top charities so far this year, and it seemed fair as a reasonable overall estimate.

[2] This example is of a top charity. The majority of the funding we direct goes to top charities, but we also fund programs beyond top charities, including research projects. Our top charities are all programs that prevent deaths primarily in young children, while some other programs we recommend primarily improve other health outcomes or increase incomes.

[3] This $6 figure doesn't account for other considerations, like whether another donor would have funded this program if GiveWell didn't. Those leverage and fungibility considerations mean we think that fewer than one additional child is reached per $6 donated, as some children may have received the medication in a world in which we didn't fund this distribution.

[4] In practice, it's unlikely that every child in a village would be reached with a full course of SMC. But, the $6 figure above refers to cost per child fully reached (setting aside fungibility), not just the cost per child targeted by the program.

[5] We calculate this by multiplying 270 children by a malaria prevalence of 37%, which results in 99.9 children testing positive for malaria.

[6] We calculate this by multiplying 270 children by the 22 percentage point reduction in malaria prevalence in the treated population, resulting in 59.4 children who would be protected by SMC, then subtracting that from the 99.9 likely to test positive for malaria: 99.9 - (270 * 22%) = 40.5.

[7] An estimate of 3.7 deaths averted per 1000 kids treated implies one death averted per (1000 / 3.7) = 270 kids treated.

[8] Estimated impact figures are based on our latest cost-effectiveness estimates, which are uncertain and could either overstate or understate actual impact.

[9] In our latest cost-effectiveness analysis, we estimate that New Incentives' program would be 6x cash in Yobe State, Nigeria, and cost $9,000 per life saved, and that the Against Malaria Foundation's long-lasting insecticide-treated nets program would be 5.5x cash in Chad and cost $14,000 per life saved.

[10] This selection of programs is designed to be illustrative rather than representative of all our research. We caution against taking these estimates too literally, especially for the programs we aren't funding (where we've done limited work). Note also that 10x cash isn't a hard limit; we consider qualitative factors in our recommendations, as discussed here.

More information about the programs in this chart, from top to bottom:

• Malaria Consortium's seasonal malaria chemoprevention program in Burkina Faso, estimated at 19x cash.
• New Incentives extension and expansion in northern Nigeria; grant recommended in May 2023. The cost-effectiveness of the grant will depend on how much funding goes to each state, with a lower-bound estimate of 17x cash.
• Helen Keller International's vitamin A supplementation program in Madagascar. Our estimate is that this program is about 12x cash, and we recommended funding it earlier this year.
• Evidence Action's in-line chlorination program in Malawi. We recommended a grant to this program in July 2022 and estimated it was around 11x cash.
• Clean cookstoves. We're very uncertain but estimate this to be very roughly 6x cash (though it could vary substantially, depending on the specific program and context).
• Programs to increase the supply of midwives. We're very uncertain but estimate this to be very roughly 6x cash (though it could vary substantially, depending on the specific program and context).
• Unconditional cash transfers, estimated (by definition) to be 1x cash. While we don't represent anything below 1x cash here, 1x cash isn't meant to represent the floor; we just tend to focus our research on programs that we think are likely to be more cost-effective than cash.

[11] After transaction fees, 100% of donations to the All Grants Fund are distributed to the recipient organizations. GiveWell does not use any of these funds for our own operating support.