Thanks for posting this. Some comments and questions:
I echo Habryka’s reluctance about cause-partial research, and I would have appreciated if you’d shared a little more context for EA’s, given that this is the EA Forum. (for example, why Founders Pledge decided to research this, how it’s relevant from an EA perspective and how it’s not, which assumptions it makes)
Some specific questions:
1. Why haven’t you compared charities on a comparably metric, such as the DALY, or Life-Satisfaction Points? Is this because the report assumes what is (intrinsically or robustly) important is empowerment, and that is hard to measure?
2. Why do you recommend both Village Enterprise and Bandhan’s ‘Targeting the Hardcore Poor’ programme? The latter appears 3.5 times as cost-effective (in terms of increasing purchasing power), and the organisations appear to have similar strength, room for more funding, and wider benefits.
3. Do you have a publicly available cost-effectiveness model for the 15 charities considered? By how much did these charities differ on comparable measures?
Thanks for these questions Siebe! And I take your point on sharing context; I’ll edit in some points in the main post.
1. We have internally compared these charities on something close to a DALY-equivalent to aid our decisions (similar to what GiveWell does in their cost-effectiveness analyses), but have not included this in the report. This is not because of any assumptions the report makes on empowerment (note that it defines empowerment simply as ‘improving lives’). It’s mainly because of time constraints: we didn’t think it was worth putting in the time to present our estimates in a polished way, given the aims we have with this report (making high-quality recommendations to our members). This is also because internally we are still in the process of developing our views on how to compare across causes and outcome metrics.
2. In terms of cost-effectiveness estimates both do better than GiveDirectly (which we also recommend), and there is obviously large uncertainty in such estimates. Furthermore, Bandhan only accepts donations over $320,000 at this point. Last but not least, the organisations differ in marked ways (where they work, programme focus, target group, type of evidence) and might appeal to different people in our community.
3. We do have rough cost-effectiveness models on almost all of the other charities, but unfortunately I cannot make those public. This is partially for reasons of information sharing (I’d have to check with the charities that provided extra info), but also because these models aren’t as worked out as the ones in the report, and a one-to-one comparison would in many cases be confusing rather than valuable. In fact, most initial cost-effectiveness estimates of the other charities are higher than the final estimates of the recommended charities, and we had to deprioritise them to a large extent because the evidence was weaker. Moreover, we find that as we do a more extensive cost-effectiveness analysis of a charity (as we did for our recommended charities), the numbers often go down rather than up, so it’s likely that our ‘final’ estimates of the other charities would be much lower than the initial, rough estimates we have now.
I remember Founders Pledge saying something about this before, they work with a lot of startup founders so they often take the existing priorities of people peripheral to EA as given. They have other cause reports like this.
Just to clarify, at FP we don’t take existing priorities/preferences as a given, but we of course do take them into account to some extent when making recommendations (if only because otherwise nobody would follow those recommendations!). We currently use something called the value-discovery approach, which is about asking members about the underlying values driving their preferences (e.g. do you care about people living in the future?), and then making cause/charity recommendations based on those rather than on cause/charity preferences themselves. We also spend quite some time on educating our community on EA/effective giving principles, e.g. this is a main focus of our Programmes team.
Thanks for posting this. Some comments and questions:
I echo Habryka’s reluctance about cause-partial research, and I would have appreciated if you’d shared a little more context for EA’s, given that this is the EA Forum. (for example, why Founders Pledge decided to research this, how it’s relevant from an EA perspective and how it’s not, which assumptions it makes)
Some specific questions:
1. Why haven’t you compared charities on a comparably metric, such as the DALY, or Life-Satisfaction Points? Is this because the report assumes what is (intrinsically or robustly) important is empowerment, and that is hard to measure?
2. Why do you recommend both Village Enterprise and Bandhan’s ‘Targeting the Hardcore Poor’ programme? The latter appears 3.5 times as cost-effective (in terms of increasing purchasing power), and the organisations appear to have similar strength, room for more funding, and wider benefits.
3. Do you have a publicly available cost-effectiveness model for the 15 charities considered? By how much did these charities differ on comparable measures?
Thanks for these questions Siebe! And I take your point on sharing context; I’ll edit in some points in the main post.
1. We have internally compared these charities on something close to a DALY-equivalent to aid our decisions (similar to what GiveWell does in their cost-effectiveness analyses), but have not included this in the report. This is not because of any assumptions the report makes on empowerment (note that it defines empowerment simply as ‘improving lives’). It’s mainly because of time constraints: we didn’t think it was worth putting in the time to present our estimates in a polished way, given the aims we have with this report (making high-quality recommendations to our members). This is also because internally we are still in the process of developing our views on how to compare across causes and outcome metrics.
2. In terms of cost-effectiveness estimates both do better than GiveDirectly (which we also recommend), and there is obviously large uncertainty in such estimates. Furthermore, Bandhan only accepts donations over $320,000 at this point. Last but not least, the organisations differ in marked ways (where they work, programme focus, target group, type of evidence) and might appeal to different people in our community.
3. We do have rough cost-effectiveness models on almost all of the other charities, but unfortunately I cannot make those public. This is partially for reasons of information sharing (I’d have to check with the charities that provided extra info), but also because these models aren’t as worked out as the ones in the report, and a one-to-one comparison would in many cases be confusing rather than valuable. In fact, most initial cost-effectiveness estimates of the other charities are higher than the final estimates of the recommended charities, and we had to deprioritise them to a large extent because the evidence was weaker. Moreover, we find that as we do a more extensive cost-effectiveness analysis of a charity (as we did for our recommended charities), the numbers often go down rather than up, so it’s likely that our ‘final’ estimates of the other charities would be much lower than the initial, rough estimates we have now.
I remember Founders Pledge saying something about this before, they work with a lot of startup founders so they often take the existing priorities of people peripheral to EA as given. They have other cause reports like this.
Just to clarify, at FP we don’t take existing priorities/preferences as a given, but we of course do take them into account to some extent when making recommendations (if only because otherwise nobody would follow those recommendations!). We currently use something called the value-discovery approach, which is about asking members about the underlying values driving their preferences (e.g. do you care about people living in the future?), and then making cause/charity recommendations based on those rather than on cause/charity preferences themselves. We also spend quite some time on educating our community on EA/effective giving principles, e.g. this is a main focus of our Programmes team.