Things I learned from Toby Ord


Toby Ord recently gave a talk about foreign aid to Harvard High-Impact Philanthropy, titled “Aid Works (On Average)”. It was really enlightening in many ways. Here are a few juicy points from the talk itself:

  • Eradicating smallpox saved so many lives that, assuming all other foreign aid interventions were neutral, it still averages out to well below the amount the US is willing to pay to save a life.

  • The eradication also paid for itself, as in, there’s a direct and obvious causal path through which it paid for itself (most health interventions probably pay for themselves through increased GDP, but that’s always hard to be sure about): smallpox inoculation was no longer necessary, saving billions in vaccination costs each year.

  • Also, there were several other public health foreign-aid efforts with effects on the same order of magnitude, including fighting cholera and malaria. Each of these efforts may have yielded more utility than anything else humanity has done to date.

  • These efforts were largely “vertical”, that is, focused on eliminating a single disease. Despite their massive successes, vertical interventions are no longer in vogue in public health circles, possibly because nobody has run the numbers and realized how effective they are. Dr. Ord suggested overspecialization as a potential reason why nobody else did this relatively simple calculation.

During the Q&A afterwards, Dr. Ord brought up a few other really interesting points.

  • According to him, at least, the impressive-sounding things that he does—hobnobbing with special advisers to the Prime Minister, helping set disease control priorities, attracting tons of donors—don’t actually require superpowers, just a willingness to be bold and network/make connections. He claims they’re the kind of thing that any sufficiently motivated and reasonably bright person could do, and could be a good fit for many of the students listening to him. I think this may have been the part of the talk that was best for the undergrads to hear—certainly heartening for me!

  • (The big one, in my opinion:) some preliminary modeling results show that the impact differential between meta-research (on which interventions are most effective) and direct interventions could be a factor of 1,000 or higher. Unsurprisingly, this is quite sensitive to the amount of money you assume will be redirected to more cost-effective interventions, so I’m guessing they used pretty optimistic figures for this (I think Toby mentioned something in the billions). Nevertheless, if you assume it’s linear in money moved and you go down from a billion to just GiveWell’s money moved ($10 million this year), you still have at least a factor of 10.

    I assume that this figure is very tentative because otherwise I would surely have heard it before, but even rumors of that kind of differential are enough to suggest that I might want to wait before making a donation.

2 commentscomment


I think the 1000x more effective figure is coming from the Disease Control Priorities Project. E.g. they give the cost of malaria prevention at $5-$15/DALY, while treating osteoporosis is $34,166-$835,622/QALY. (The low-cost treatments mostly seem to be measured in DALYs and high-cost ones in QALYs - not sure how they compare exactly, but assuming they’re somewhat comparable, wow.)


Anonymous, it’s true that the best currently-known interventions are a thousand times more effective than average, but Toby was talking about a different comparison: spending money on future research into find interventions that are even more cost-effective than the best currently known ones, versus spending that money on the current best-known interventions. That is to say, spending money on research might be 1000 times as effective as malaria prevention, not as osteoporosis.

The claim of a factor of 1000 in this scenario has a couple distinct parts:

  1. Funding future effectiveness research has a good chance of uncovering other cost-effective interventions
  2. Such interventions are going to be a lot more cost-effective than the current best-known ones
  3. This will cause a lot of money to be diverted to said new interventions

These effects, when combined, are what will produce the additional factor of 1000. I was just surprised at the magnitude of potential difference.

comment