How cost-effective is cost-effective?
We at SoGive argue that a good starting point for assessing a charity is to ask two questions:
- how much does the charity's intervention cost
- and what do the beneficiaries get for your money?
But once you've found this out for a charity, what should you benchmark that against in your quest to find a high bang for your buck charity? This post will set out some examples of "good" cost per outcome figures.
TL;DR you can achieve an outcome about as good as saving a life for roughly $1,000 to $4,000; call it about $2,000. (Or about £1,500 if you think in sterling). However there may be some opportunities which are better than that.
-----------------------------------------------------
More detail:
Note that this is an update of a post last published late in 2019.
The below is taken from GiveWell's cost-effectiveness model as at 25th Nov 2019.
Why use GiveWell's figures? We consider the GiveWell-recommended charities a good set of independently assessed benchmarks of charity performance. In particular, these charities have been assessed to a substantially higher level of rigour than most charities, which means that you can have a relatively higher level of confidence that the outcomes are being achieved.
Note that including these figures to the nearest dollar risks insinuating a spurious level of accuracy. The figures are in fact based on a number of value judgements and have several uncertainties and assumptions baked in, and should not be taken too literally.
However they are useful for giving an idea of the sort of cost we can expect to pay for a good outcome like saving a life. In particular, something in the region of $2,000 should do it.
You could argue that we should go lower still and say that $1,000 (and a bit) is enough; after all Deworm the World has a costing as low as that in the table.
However this risks taking these numbers too literally. For example, let's compare Deworm the World and (say) Malaria Consortium. The Malaria Consortium work of Seasonal Malaria Chemoprevention (or SMC) is a fairly clear direct intervention. Deworm the World, however works with government staff to implement deworming programmes. One might argue, therefore, that as the Theory of Change for SMC is less complex and has fewer steps, we can have that little bit of extra confidence in the work. (For brevity, I'm not providing a lengthy comparison of those two charities here)
There may be other charities which are not on this list which may be more cost-effective still.
Note further that the table shows costs for an outcome *as good as* saving the life of an under-5 year old. This is because the charity might do things other than save lives.
This table splits that out. I.e. it shows what proportion (out of 100%) of the modelled benefits of the charity's work come from each of increases in consumption / development effects, mortality reduction in under-5s and mortality reduction in over-5s. (The last two being, essentially, saving lives)
** This shows GiveWell's split of the impact of development effects and lives saved, although we think that GiveWell may have slightly overstated the development effects (more below)
The deworming charities (Deworm the World, Sightsavers, Schistosomiasis Control initiative, END fund) all get credit only for "increases in ln(consumption)" and not for any health benefits. "Increases in ln(consumption)" is fancy economist-speak for people being financially better off and therefore better able to spend money on things they need.
Development effects refers to the fact that people who have suffered less malaria in their youth or have received better vitamin A intake receive more income in later life (presumably because they have "developed" better).
GiveWell's spreadsheet also allows you to take your copy and enter your own values. You can also look at outcomes as good as (e.g.) doubling consumption instead of an outcome as good as saving a life. You can find this here. The version used for the above analysis was that dated 25th Nov 2019. My copy is here, where you can see how I put together the above table.
Helen Keller International figures
Note that the split between development and mortality (23% vs 78%) shown in the table is based on the figures in the GiveWell model, which GiveWell estimated as set out in this document. We note that the process involved taking the approximate proportion of impact attributable to development (c30%) and applying a 75% factor to that, which gets to 22.5%.
Since then, GiveWell has incorporated research conducted by IDinsight about moral values held by people in Kenya and Ghana. In light of this, the proportion of the impact of malaria interventions attributable to development effects is probably no longer close to 30%, and more like 20%, or maybe even less. This means that the figures used in GiveWell's model are likely overstated, and instead of being 23%, should probably be in the range 10%-15%. We have not reviewed the underlying carefully enough to be confident of this conclusion.
-----------------------------------------------------
There are some other important but thorny questions which are not covered in this post, such as
- if we know about the outputs achieved by the charity (those being the immediate effects of the charity's work) what do we know about the ultimate outcomes?
- are we even confident that the outcomes are happening at all?
- when looking at charities that do quite different things, how do we go about comparing them?
- if you look at a charity which spends way more than $2,000 hoping to achieve an outcome that you think is obviously less good than saving a life, that's easy; what if the charity spends less than $2,000 and the outcomes, if achieved, are also less good than a life saved?
The aim of this short post is not to answer all of these questions, it is simply to point you towards some benchmarks you can use to help you judge what a high-impact charity might look like.
Note that this post is an update of a post published in late 2019.