My working assumption is that medical research is the most cost-effective domestic charity. My toy model is:
- Disease X kills N people per year
- In expectation, we’ll need M researcher-years to find a cure, costing salaryM dollars
- Currently disease X receives F dollars in funding each year, so in expectation we’ll find the cure in salaryM/F years
- With an extra donation of D dollars, the expectation date of the cure gets brought forward to (salaryM—D)/F years, i.e., bringing it forward by D/F years.
- The earlier cure means that ND/F people who would have died will now live.
Orders of magnitude: 10^7 cancer deaths worldwide annually, 10^10 dollars in funding --> 10^3 dollars per future statistical life saved.
There are enough problems with this toy model not to take it too seriously. I know that GiveWell have thought a lot about medical research, but it’s a complicated thing with commercial interests getting involved in some stages, the question of how important the marginal researcher is, different diseases might have a different number of research “leads”, and so on. The numbers will also change depending on the discount rate.
Still, given the orders of magnitude involved, I think medical research is pretty good impact-per-dollar-wise. Choosing a particular charity then comes down to looking at which diseases are most underfunded relative to their DALY burden, and which charity puts the money into research rather than “awareness raising” or whatever. And the latter issue is definitely one which most people can appreciate.
I would change “kills X people” to “reduces X years of quality life”. If there is a late onset disease that affects people who might not live much longer or much happier then that is different to a disease which kills those with many years of quality life left.
Obviously the proof is in the pudding and you would have to be able to measure all that.
Combined with the effects of diminishing returns this is the reason I don’t think much of cancer research warrants a lot more funds from an EA because each individual extra dollar makes little impact on quality life years across the population.
The cancers I would focus on are ones that affect children, but then again, the cost of research is high and the number of children affected is low.
Sure. But if you change 10^3 into 10^4, we’re still talking the same order of magnitude of cost-effectiveness as GiveDirectly’s cash transfers (depending on how highly consumption should be valued against DALY’s averted, etc.). Even if we assume that a full accounting would show the cost-effectiveness of donations to medical research to be worse than that, what other domestic charities would have a “first-guess cost-effectiveness estimate” even in the same ballpark?
My working assumption is that medical research is the most cost-effective domestic charity. My toy model is:
- Disease X kills N people per year
- In expectation, we’ll need M researcher-years to find a cure, costing salaryM dollars
- Currently disease X receives F dollars in funding each year, so in expectation we’ll find the cure in salaryM/F years
- With an extra donation of D dollars, the expectation date of the cure gets brought forward to (salaryM—D)/F years, i.e., bringing it forward by D/F years.
- The earlier cure means that ND/F people who would have died will now live.
Orders of magnitude: 10^7 cancer deaths worldwide annually, 10^10 dollars in funding --> 10^3 dollars per future statistical life saved.
There are enough problems with this toy model not to take it too seriously. I know that GiveWell have thought a lot about medical research, but it’s a complicated thing with commercial interests getting involved in some stages, the question of how important the marginal researcher is, different diseases might have a different number of research “leads”, and so on. The numbers will also change depending on the discount rate.
Still, given the orders of magnitude involved, I think medical research is pretty good impact-per-dollar-wise. Choosing a particular charity then comes down to looking at which diseases are most underfunded relative to their DALY burden, and which charity puts the money into research rather than “awareness raising” or whatever. And the latter issue is definitely one which most people can appreciate.
I would change “kills X people” to “reduces X years of quality life”. If there is a late onset disease that affects people who might not live much longer or much happier then that is different to a disease which kills those with many years of quality life left.
Obviously the proof is in the pudding and you would have to be able to measure all that.
Combined with the effects of diminishing returns this is the reason I don’t think much of cancer research warrants a lot more funds from an EA because each individual extra dollar makes little impact on quality life years across the population.
The cancers I would focus on are ones that affect children, but then again, the cost of research is high and the number of children affected is low.
Sure. But if you change 10^3 into 10^4, we’re still talking the same order of magnitude of cost-effectiveness as GiveDirectly’s cash transfers (depending on how highly consumption should be valued against DALY’s averted, etc.). Even if we assume that a full accounting would show the cost-effectiveness of donations to medical research to be worse than that, what other domestic charities would have a “first-guess cost-effectiveness estimate” even in the same ballpark?
Agreed :)