I think the biggest improvement would be correcting the fact that this model (accidentally, I think) assumes that improving any arbitrary high budget charity by 5% is equally as impactful as improving a Givewell equivalent charity by 5%. Most charity’s impact is an order of magnitude smaller.
You could solve this with a multiplier for the charity’s impact at baseline.
If I understand correctly, you figure that if you become a trustee of a £419668/year budget charity, if only you can improve the cost effectiveness by 5%, then you can divide that by 42 hours a year, to get £419668*5%/42 hours=£500/hour in the value of your donated time. (A style tip—it would be helpful to put the key equation describing roughly what you’ve done in the description, to make it all legible without having to go into the spreadsheet.)
I think it is fair to say that, were you to successfully perform this feat, you have indeed done something roughly as impactful as providing a £500/hour value to the charity you are trustee-ing for. So, if you improved a Givewell-top-charity-equivalent’s cost effectiveness by 5% for a year, then maybe you could fairly take 5% of that charity’s yearly budget and divide it by your hours for that year, as you’ve done, to calculate your Givewell-top-charity-equivalent impact in terms of how it would compare to donated dollars.
But if you improve a £419668/yr budget charity which is only 1% as cost-effective as a Givewell-top-charity-equivalent by 5%, then that makes your hourly impact 1%*£419668*5%/42 hours = £5/hour of Givewell-top-charity-equivalent impact—you’d be better served working a bit extra and donating 5 dollars to Givewell.
I don’t think this model has credence even after these adjustments as I’m skeptical of the structure, but you did make those assumptions explicitly which is good. If you think the effecttakes ~42 hours/year then this hypothesis is potentially cheap to just test in practice, and then revise your model with more information. Have you joined any boards and tried this in practice, if yes how did it go?
edit—ah, you’re using the term “5% increase” very differently.
Instead it assumes a 5% increase, perhaps from £0 of impact to 5% of the annual income or perhaps from 100% of annual income to 105%
So just to be clear, this implies that making 100% of your annual income in impact would mean that you are the most cost effective charity in the world (or whatever other benchmark you want to set at “100%”). Used in this sense, the word “5% increase” doesn’t mean “the shelter saves 5% more kittens” but that charity as a whole has gone from being part of the long tail of negligible impact to being 1/20th as cost effective as the most cost effective charity in the world. This isn’t the way percents are usually expressed / this seems like a confused way to express this concept since the 100% benchmark is arbitrary/unknown—it would be better in that case to express it on an absolute scale rather than a percentage.
Thanks for your time in writing this. You’ve clearly thought about it a lot. I’ll have a look at your comments.
I guess we want to encourage people to do more pieces like this and I think that were I not quite robust I would find this response intimidating. Maybe you could have started with a hello or a thank you to lighten the tone?
I agree that more people trying to do cost effectiveness analyses is good! I regret that the tone seemed otherwise and will consider it more in the future. I engaged with it primarily because I too often wonder about how one might improve impact outside of impact-focused environments, and I generally find it an interesting direction to explore. I also applaud that you made the core claim clearly and boldly and I would like to see more of that as well—all models suffer these flaws to some degree and it’s a great virtue to make clear claims that are designed such that any mistakes will be caught (as described here). Thanks for doing the piece and I hope you can use these comments to continue to create models of this and other courses of action :)
I think the biggest improvement would be correcting the fact that this model (accidentally, I think) assumes that improving any arbitrary high budget charity by 5% is equally as impactful as improving a Givewell equivalent charity by 5%. Most charity’s impact is an order of magnitude smaller.
You could solve this with a multiplier for the charity’s impact at baseline.
If I understand correctly, you figure that if you become a trustee of a £419668/year budget charity, if only you can improve the cost effectiveness by 5%, then you can divide that by 42 hours a year, to get £419668*5%/42 hours=£500/hour in the value of your donated time. (A style tip—it would be helpful to put the key equation describing roughly what you’ve done in the description, to make it all legible without having to go into the spreadsheet.)
I think it is fair to say that, were you to successfully perform this feat, you have indeed done something roughly as impactful as providing a £500/hour value to the charity you are trustee-ing for. So, if you improved a Givewell-top-charity-equivalent’s cost effectiveness by 5% for a year, then maybe you could fairly take 5% of that charity’s yearly budget and divide it by your hours for that year, as you’ve done, to calculate your Givewell-top-charity-equivalent impact in terms of how it would compare to donated dollars.
But if you improve a £419668/yr budget charity which is only 1% as cost-effective as a Givewell-top-charity-equivalent by 5%, then that makes your hourly impact 1%*£419668*5%/42 hours = £5/hour of Givewell-top-charity-equivalent impact—you’d be better served working a bit extra and donating 5 dollars to Givewell.
I don’t think this model has credence even after these adjustments as I’m skeptical of the structure, but you did make those assumptions explicitly which is good. If you think the effect takes ~42 hours/year then this hypothesis is potentially cheap to just test in practice, and then revise your model with more information. Have you joined any boards and tried this in practice, if yes how did it go?
edit—ah, you’re using the term “5% increase” very differently.
So just to be clear, this implies that making 100% of your annual income in impact would mean that you are the most cost effective charity in the world (or whatever other benchmark you want to set at “100%”). Used in this sense, the word “5% increase” doesn’t mean “the shelter saves 5% more kittens” but that charity as a whole has gone from being part of the long tail of negligible impact to being 1/20th as cost effective as the most cost effective charity in the world. This isn’t the way percents are usually expressed / this seems like a confused way to express this concept since the 100% benchmark is arbitrary/unknown—it would be better in that case to express it on an absolute scale rather than a percentage.
Thanks for your time in writing this. You’ve clearly thought about it a lot. I’ll have a look at your comments.
I guess we want to encourage people to do more pieces like this and I think that were I not quite robust I would find this response intimidating. Maybe you could have started with a hello or a thank you to lighten the tone?
I agree that more people trying to do cost effectiveness analyses is good! I regret that the tone seemed otherwise and will consider it more in the future. I engaged with it primarily because I too often wonder about how one might improve impact outside of impact-focused environments, and I generally find it an interesting direction to explore. I also applaud that you made the core claim clearly and boldly and I would like to see more of that as well—all models suffer these flaws to some degree and it’s a great virtue to make clear claims that are designed such that any mistakes will be caught (as described here). Thanks for doing the piece and I hope you can use these comments to continue to create models of this and other courses of action :)
It’s tricky isn’t it. It’s a poor model :P but it’s better than my poorly informed intuition. Not quite sure what to do about that. Have a good one :)