Unless I’m missing something, it seems like we all should be giving to EA advocacy groups until the amount of resources available to those groups reaches the threshold at which the donation is no longer leveraged. What’s the counter-argument? Has there been any analysis on what that threshold level is? In other words, if we take the total resources currently available to groups like Giving What We Can and The Life You Can Save, how many times those resources must we have before I should start donating to the Against Malaria Foundation?
The point at which you hit diminishing returns to funding an org may actually be pretty low. I’d be skeptical about the marginal value of budget increases of much more than a factor of 2 per year unless the org had demonstrated really impressive traction.
There are also signaling issues with only donating to metacharities, so if you’re public about your giving it might not be a great idea (“guys, look at how much I donate to these organizations that promote donating to themselves!”).
The error bars around these estimates are extremely high. The confidence interval for GWWC’s leverage ratio plausibly already includes numbers below 1 (in addition to numbers above 300). For instance, this is one reason GiveWell doesn’t recommend any metacharities—which, IIRC, someone said at one point that they would if they could find one with a robust enough case for impact. Donors with preferences similar to GiveWell’s may reject metacharities entirely for that reason.
The point at which you hit diminishing returns to funding an org may actually be pretty low. I’d be skeptical about the marginal value of budget increases of much more than a factor of 2 per year unless the org had demonstrated really impressive traction.
This isn’t incompatible with what you’re saying, but they may diminish well before that also. Taking the present example of Giving What We Can, the people who worked there or are involved with it thought that applied to it. They thought most of the value came from the existence of the organisation and a pledge people could sign if they wanted to commit to giving 10%, and other things which were done even before they started paying staff. So that would be diminish returns right at the $ 0 mark!
There are also signaling issues with only donating to metacharities, so if you’re public about your giving it might not be a great idea (“guys, look at how much I donate to these organizations that promote donating to themselves!”).
There are even less positive ways to frame that also, like giving to one another, and having organisations which heavily focus on promoting themselves (including by promoting the idea of metacharity, and making it a central concept in the movement). Even aside from signalling, we should see others’ discomfort with that as a reason to be wary of it ourselves.
The confidence interval for GWWC’s leverage ratio plausibly already includes numbers below 1
This is what those people I talked to from GWWC thought, due to their various experiences and observations. And GiveWell too as you say; they had had conversations with people at GiveWell who thought that GWWC’s future fundraising ratio was below 1.
Doesn’t every organization/social movement that efficiently allocates resources have diminishing returns beginning with the first dollar? One reason why this could theoretically not be true is if efficient use of capital requires upfront investment in infrastructure, but I don’t know if that applies here. The concept of diminishing returns seems distinct from leverage (though obviously not unrelated).
The signalling issue is complicated, and I’m open to suggestions. As I’m a consequentialist, I’m open simply to lying.
″ hits diminishing returns” is usually used as a shorthand for “investment in hits the point where returns have diminished enough that additional investment is no longer optimal.”
Doesn’t every organization/social movement that efficiently allocates resources have diminishing returns beginning with the first dollar?
That will be the case very often, except in cases like that which you have mentioned. In these comments Michelle Hutchinson came up with a few other possibilities, like economies of scale.
The signalling issue is complicated, and I’m open to suggestions. As I’m a consequentialist, I’m open simply to lying.
This wouldn’t address the non-signalling concern that I raised though (as I’m sure you’re aware of course).
I think you may be being too pessimistic about the confidence intervals. I’d be very surprised if GWWC turned out to have a leverage ratio less than 1. In fact, I’d be less surprised than if someone handed me a report making a completely knock down case that AMF has no impact, or that GiveDirectly turned out to be harmful.
To think the ratio is less than 1, you’d need basically all of GWWC’s pledgers to drop out immediately (or for it to be true that all of them would have donated otherwise), and you’d need a reason why over 80% of historical attributed donations weren’t in fact due to GWWC.
And the leverage ratio is a conservative measure of GWWC’s impact e.g. it doesn’t include additional donations to non-recommended charities, or the value of GWWC’s role growing the EA movement.
Disclosure: I work at CEA, the parent organisation of GWWC.
I agree that GWWC’s ratio is probably above 1 with a good deal of confidence (though I haven’t done the formal math to evaluate how extreme that statement is). But I think the more compelling argument is that expansion funding on the margin may not have a ratio above 1.
Yes, there’s a huge, huge difference between the impact of GWWC existing as a place where people who wanted to pledge 10% of their income to help those living in global poverty could join others in publicly doing so, and the impact of its marginal funded activities now. GWWC existed as that place before The Centre for Effective Altruism was founded around it as an organisation with donors and a budget supporting paid employees. If minimal resources were spent on creating the basic infrastructure, and I don’t know if that’s so, but if so, then it had a mega high impact ratio. But it seems wrong to use that to justify keeping on spending more money on more employees doing more marginal projects until the impact from the original resource gets “used up.”
Ok this is all fair. I think, however, that a big fraction of the historical impact is due to on-going activity, of the kind that could continue, rather than being all due to the ‘set up’ generating the stream. And that would mean the historical ratio is a reasonable guide to the future.
This can be hard to see from the outside, but if you look at where new pledgers are coming from, it’s often new press coverage or student group activity. Many also only take the pledge after being nudged by someone in person, even if they had heard about GWWC some time before, so there’s an important role just talking to lots of people about the pledge. These kinds of activities can be scaled much further.
Moreover, GWWC tries to monitor the number of new pledges generated by new activities and will quit on anything that’s not generating enough. You can see some estimates by Rob of marginal returns above.
Overall, you’d still have to be very pessimistic. Suppose GWWC’s historical leverage ratio is 60. For the future leverage ratio to be less than one, you’d (roughly) need to think there was a 98% chance that marginal activities produced zero returns rather than historical rates of return.
Finally, I think it’s helpful to bear in mind how tiny GWWC is. Suppose taking the GWWC pledge is as ethically demanding as vegetarianism, then it could one day reach 1% of the developed world. Since GWWC only has 1000 members, it has only reached 0.01% of its addressable market. Given such limited penetration, I think it would be surprising if there were no further returns to be had. It’s not as if students at Cambridge (where ~100 people have taken the pledge recently) are radically different from those at other prestigious universities, so we should expect similar efforts to work elsewhere, so GWWC could already be 10x bigger even if just expanded among prestigious universities. If anything you might expect marginal costs per pledge to be dropping at this stage in GWWC’s growth as you benefit from economies of scale and critical mass effects (and there’s some evidence this is happening). Finally, given the chance that 99.99% of the value still lies in the future, it seems well worth spending a sizeable amount of resources now figuring out whether further growth is possible and how best to do it. In some ways, GWWC’s marginal leverage ratio is besides the point: it’s like looking at Google in its early days and deciding whether to invest or not based on whether it’s already profitable.
Unless I’m missing something, it seems like we all should be giving to EA advocacy groups until the amount of resources available to those groups reaches the threshold at which the donation is no longer leveraged. What’s the counter-argument? Has there been any analysis on what that threshold level is? In other words, if we take the total resources currently available to groups like Giving What We Can and The Life You Can Save, how many times those resources must we have before I should start donating to the Against Malaria Foundation?
The point at which you hit diminishing returns to funding an org may actually be pretty low. I’d be skeptical about the marginal value of budget increases of much more than a factor of 2 per year unless the org had demonstrated really impressive traction.
There are also signaling issues with only donating to metacharities, so if you’re public about your giving it might not be a great idea (“guys, look at how much I donate to these organizations that promote donating to themselves!”).
The error bars around these estimates are extremely high. The confidence interval for GWWC’s leverage ratio plausibly already includes numbers below 1 (in addition to numbers above 300). For instance, this is one reason GiveWell doesn’t recommend any metacharities—which, IIRC, someone said at one point that they would if they could find one with a robust enough case for impact. Donors with preferences similar to GiveWell’s may reject metacharities entirely for that reason.
(Disclosure: I donate mostly to metacharities...)
This isn’t incompatible with what you’re saying, but they may diminish well before that also. Taking the present example of Giving What We Can, the people who worked there or are involved with it thought that applied to it. They thought most of the value came from the existence of the organisation and a pledge people could sign if they wanted to commit to giving 10%, and other things which were done even before they started paying staff. So that would be diminish returns right at the $ 0 mark!
There are even less positive ways to frame that also, like giving to one another, and having organisations which heavily focus on promoting themselves (including by promoting the idea of metacharity, and making it a central concept in the movement). Even aside from signalling, we should see others’ discomfort with that as a reason to be wary of it ourselves.
This is what those people I talked to from GWWC thought, due to their various experiences and observations. And GiveWell too as you say; they had had conversations with people at GiveWell who thought that GWWC’s future fundraising ratio was below 1.
Doesn’t every organization/social movement that efficiently allocates resources have diminishing returns beginning with the first dollar? One reason why this could theoretically not be true is if efficient use of capital requires upfront investment in infrastructure, but I don’t know if that applies here. The concept of diminishing returns seems distinct from leverage (though obviously not unrelated).
The signalling issue is complicated, and I’m open to suggestions. As I’m a consequentialist, I’m open simply to lying.
″ hits diminishing returns” is usually used as a shorthand for “investment in hits the point where returns have diminished enough that additional investment is no longer optimal.”
duplicate comment
That will be the case very often, except in cases like that which you have mentioned. In these comments Michelle Hutchinson came up with a few other possibilities, like economies of scale.
This wouldn’t address the non-signalling concern that I raised though (as I’m sure you’re aware of course).
I think you may be being too pessimistic about the confidence intervals. I’d be very surprised if GWWC turned out to have a leverage ratio less than 1. In fact, I’d be less surprised than if someone handed me a report making a completely knock down case that AMF has no impact, or that GiveDirectly turned out to be harmful.
To think the ratio is less than 1, you’d need basically all of GWWC’s pledgers to drop out immediately (or for it to be true that all of them would have donated otherwise), and you’d need a reason why over 80% of historical attributed donations weren’t in fact due to GWWC.
And the leverage ratio is a conservative measure of GWWC’s impact e.g. it doesn’t include additional donations to non-recommended charities, or the value of GWWC’s role growing the EA movement.
Disclosure: I work at CEA, the parent organisation of GWWC.
I agree that GWWC’s ratio is probably above 1 with a good deal of confidence (though I haven’t done the formal math to evaluate how extreme that statement is). But I think the more compelling argument is that expansion funding on the margin may not have a ratio above 1.
Yes, there’s a huge, huge difference between the impact of GWWC existing as a place where people who wanted to pledge 10% of their income to help those living in global poverty could join others in publicly doing so, and the impact of its marginal funded activities now. GWWC existed as that place before The Centre for Effective Altruism was founded around it as an organisation with donors and a budget supporting paid employees. If minimal resources were spent on creating the basic infrastructure, and I don’t know if that’s so, but if so, then it had a mega high impact ratio. But it seems wrong to use that to justify keeping on spending more money on more employees doing more marginal projects until the impact from the original resource gets “used up.”
Ok this is all fair. I think, however, that a big fraction of the historical impact is due to on-going activity, of the kind that could continue, rather than being all due to the ‘set up’ generating the stream. And that would mean the historical ratio is a reasonable guide to the future.
This can be hard to see from the outside, but if you look at where new pledgers are coming from, it’s often new press coverage or student group activity. Many also only take the pledge after being nudged by someone in person, even if they had heard about GWWC some time before, so there’s an important role just talking to lots of people about the pledge. These kinds of activities can be scaled much further.
Moreover, GWWC tries to monitor the number of new pledges generated by new activities and will quit on anything that’s not generating enough. You can see some estimates by Rob of marginal returns above.
Overall, you’d still have to be very pessimistic. Suppose GWWC’s historical leverage ratio is 60. For the future leverage ratio to be less than one, you’d (roughly) need to think there was a 98% chance that marginal activities produced zero returns rather than historical rates of return.
Finally, I think it’s helpful to bear in mind how tiny GWWC is. Suppose taking the GWWC pledge is as ethically demanding as vegetarianism, then it could one day reach 1% of the developed world. Since GWWC only has 1000 members, it has only reached 0.01% of its addressable market. Given such limited penetration, I think it would be surprising if there were no further returns to be had. It’s not as if students at Cambridge (where ~100 people have taken the pledge recently) are radically different from those at other prestigious universities, so we should expect similar efforts to work elsewhere, so GWWC could already be 10x bigger even if just expanded among prestigious universities. If anything you might expect marginal costs per pledge to be dropping at this stage in GWWC’s growth as you benefit from economies of scale and critical mass effects (and there’s some evidence this is happening). Finally, given the chance that 99.99% of the value still lies in the future, it seems well worth spending a sizeable amount of resources now figuring out whether further growth is possible and how best to do it. In some ways, GWWC’s marginal leverage ratio is besides the point: it’s like looking at Google in its early days and deciding whether to invest or not based on whether it’s already profitable.