Elie recently highlighted his doubts about our top charities, and a commenter responded:
Of course one ultimately never knows how much good a charity or any given donation will do. Bednets might end up saving the life of child who grows up to be the next Nelson Mandela – or the next Saddam Hussein. Everything we do is a gamble, but I’d like to make the best ones I can. These two charities look like good gambles.
I agree completely, but I’m still glad Elie emphasized these risks. Because in one specific way, supporting our top charities is riskier than supporting any other charity: if something goes wrong with one of our top charities, it will come out promptly and publicly.
All charities involve substantial risks, in terms of substantial possibilities that their activities won’t go as well as hoped. If there’s one thing that we’ve learned in over four years of researching charities, it’s that helping people isn’t easy, and there’s no such thing as a guarantee in aid. But with most charities, you’ll never have to face the embarrassment or cognitive dissonance that comes with acknowledging failure – simply because you’ll probably never really find out what happens to the money you give.
In picking top charities, GiveWell puts substantial weight on accountability: we want to be sure that we will find out what happens, particularly if things go poorly. We’ve been publishing regular updates on VillageReach, the first charity we directed significant funding to, and there have been some substantial bumps in the road that VillageReach (and GiveWell) has disclosed. We plan to do the same with our current two top charities.
Donors should also know that while we like to see strong evidence bases and track records, we don’t try to find the “safest bets” in giving – we try to find the “best bets.” Achieving maximal cost-effectiveness often means being relatively ambitious in terms of how far each dollar stretches. In addition, we believe that “upside” (the chance of doing a lot of good) usually means taking some risk.
For example, giving to GiveWell (directly) in 2007 was far more risky than giving to GiveWell in 2011. GiveWell in 2011 has more credibility and more evidence of impact. Yet a dollar of support for GiveWell in 2007 was far more helpful to us than a dollar of support today, for exactly the same reasons. Because there were only a small number of people who would go out on a limb for us, these people ended up being crucial for our startup and development.
Because we see a direct trade-off between donating early in an organization’s development (more risk, more upside) and late (less risk, less upside), we generally look for organizations that have good track records for their age. VillageReach, with just a single successful pilot project, has a substantial chance of failing to translate its model on a large scale – but huge potential for good if it succeeds. SCI, by contrast, has a well-established model with results from multiple countries, so donations to it are less likely to have either zero impact or outsized impact. AMF sits somewhere in the middle. All three charities are well past the startup stage, which we think poses too many due diligence challenges to be a good fit for our audience.
If you believe that GiveWell is committed to regular, fully honest reports on our top charities – with no sugar-coating of failures – this may be the best reason to have confidence in our research. It means our incentives are better-aligned than those of any other funder: if we overlook a problem we should have seen, we’ll be facing up to this publicly.
When I give to our top charities, I know there’s a chance I’ll find out a few years from now that they got disappointing results. For me, though, this fact makes my giving much more exciting and much more “real” feeling. Finally, I feel that giving – like other purchases and actions that are meaningful to me – has visible impact and provides learning experiences. I feel the anxiety that can only come with trying to accomplish something difficult and worthwhile. I hope our donors feel the same way.