Each year, GiveWell identifies more great giving opportunities than we are able to fully fund. As a result, in our charity recommendation decisions, we necessarily face very challenging questions, such as: How much funding should we recommend for programs that reduce poverty versus programs that reduce deaths from malaria? How should we prioritize programs that primarily benefit children versus adults? And, how do we compare funding those programs with others that have different good outcomes, such as reducing suffering from chronic health issues like anemia?
We recently received results from research we supported to help us answer these questions from the perspective of communities similar to those our top charities operate in. This blog post provides a brief summary of the project and results. Additional details are available on this page.
Background on the project
We assess charities based on their overall impact per dollar. In order to compare the impact per dollar across programs, we assign quantitative “moral weights” to each good outcome. We have invested a significant amount of time to arrive at these weights, but we still find our conclusions unsatisfying, in large part because of the fundamental difficulty of these questions. We have worked to improve our process for valuing different outcomes over the years, but we believe our current process is far from ideal.
Moral weights seems to be a highly neglected research topic. Limited information exists on how people value different outcomes. In particular, very few researchers have asked people living in low-income countries how they would make these tradeoffs. We see this as a potentially important input into our weights but have been unable to incorporate this information because it largely did not exist.
We recently supported a project intended to help address this gap in the literature. We provided funding and guidance to IDinsight, a data analytics, research, and advisory organization, to survey about 2,000 people living in extreme poverty in Kenya and Ghana in 2019 about how they value different outcomes.
The results from this research are now available here. Among other findings, they suggest that survey respondents have higher values for saving lives (relative to reducing poverty) and higher values for averting deaths of children under 5 years old (relative to averting deaths of individuals over 5 years old) than we had previously been using in our decision-making.
Although we see these study results as adding to our understanding, we would caution against putting too much weight on them. Research methods like those used in the survey have major limitations, discussed here. This study is one that should be put in the context of a larger literature about these questions and represents one approach to moral weights among many.
Nevertheless, we see this research as a valuable contribution to the literature on preferences and moral views in communities with high rates of extreme poverty. It seems to be the first study of its kind conducted in sub-Saharan Africa, and the people surveyed for this study had a substantially lower average consumption level than other studies using similar methods.
Preliminary conclusions and updates
We have provisionally updated our moral weights to place more emphasis on programs that avert deaths (relative to those that reduce poverty) and to value programs averting deaths at all ages more equally (relative to our previous assumption of valuing programs that avert deaths of individuals over 5 years old more highly). The direction of these updates was driven by this study and other, independent arguments for putting more weight on health relative to income. However, we have not yet thoroughly debated how to revise our framework for moral weights or fully completed our analysis of these results, so we see our current, provisionally-updated moral weights as a work in progress. We plan to revisit our framework for moral weights in the future.
These updates did not have a major impact on our recommended funding allocation to charities in 2019.
We share additional details on the survey and our early interpretation on this page.
Josh Rosenberg and Catherine Hollander co-authored this post.
Very glad to see work in this area. Thanks for commissioning and releasing the study.
Seems interesting. I wonder how much weight we should place on the new survey. It seems intuitive for most people to think that saving children under 5 is morally preferable to saving adults. However, I dont think most people pause to really reflect on that (regardless of if they are in extreme poverty or not). If you do pause to think of it (I didn’t really until reading GiveWell’s research on moral weights), I think you realize that adults tend to be caregivers on which children and others typically rely. As such, we should give more weight to adults rather than the intuitive thought of “innocent children”.
I’m not sure how the survey was constructed, but I would be careful about how much moral weight to place on it. This seems to be a classic System 1 (thinking fast and intuitively) vs. System 2 (thinking slow and deliberately) issue.
Givewell’s primary metric of cost-effectiveness is “Cost per outcome as good as: averting the death of an individual under 5”. Since most of the benefit of the top charities is not from reduced child mortality, putting a lower value on under-5 mortality makes this metric misleading (perhaps even dishonest). If a donation of $5000 to AMF saves one adult life, but Givewell considers an adult life 2x as valuable as an under-5 life, they will report that a donation of $2500 does “as much good as saving the life of a child”. This new survey data supports the intuition that most people would disagree with such a methodology if they knew about it.
For the subset of the weights questions that consider health, you may be interested in the June 2020 Brocher Foundation Summer Academy, “Healthy, Wealthy, and Wise: The Ethics of Health Valuation” (https://www.brocher.ch/en/events/406/brocher-summer-academy-in-global-population-health-2020-healthy-wealthy-and-wise-the-ethics-of-health-valuation/).
“survey about… how they value different outcomes”
Study FLAWED, because METHOD inherently is flawed. Money wasted.
What such method EVER gets is the reponses of persons in the situation of having been approached with a question. NOT to be equated with “how they value”.
We agree with your question about how much weight to place on the study. It’s not something that we have a definitive conclusion on, but we discuss some of the inputs you could use to inform an answer in our report here.
Mike, thanks for the heads up about that event!
Thanks for your feedback. We agree with you that using the unit “cost per equivalent under-5 death averted” (which incorporates different good outcomes, such as averting a death and increasing income, as well as other health benefits) would have counterintuitive implications for some. For this and other reasons, in most of our public communications we now tend to emphasize more intuitive results such as “cost per death averted” or “cost per treatment provided,” rather than “cost per equivalent under-5 death averted.” For example, our homepage references calculations here that are based on deaths averted and do not include the expected income benefits of top charities that primarily avert deaths, which would require a broader metric.
That said, it is difficult to avoid having some kind of all-in-one metric somewhere in our cost-effectiveness models, since we ultimately want to compare all of the benefits of the charities we consider alongside each other (for reasons discussed in the post above). How to present that comparison is a difficult question for which we have not found any answer without issues like the one you described.
In addition, we have updated our moral weights due to this survey and other arguments, and our moral weights today value averting a death at all ages equally. We think many would find this a reasonable assumption, and by sharing this research report and on our blog, we hope to make the rationale behind our approach transparent.
p, we discuss the limitations of the survey methodology in this section of our report.
Comments are closed.