Thoughts on GiveWell's 2013 recommendations

Summary: Some observations regarding GiveWell's 2013 charity recommendations.

Charity evaluator GiveWell has just announced its charity recommendations for 2013. The highlights:

• Against Malaria Foundation (AMF), the previous #1 recommendation, has been removed because it has had difficulty setting up net distributions to spend its cash reserve fast enough
• The charities are no longer explicitly ranked from first to last, but cash transfer charity GiveDirectly (previously #2) is mentioned first when the charities are listed, is the one with donations matched by Good Ventures and "which most [GiveWell] staff are favoring for their personal donations this year"
• The remaining two slots are taken up by SCI (previously #3) and DtWI (new), deworming charities recommended by Giving What We Can since 2010

Unless GiveDirectly's impact per dollar has improved significantly from last year, it looks like GiveWell is saying that donations to its top charities this year will deliver less impact per dollar than donations to its top charities less year. It also places the extremely simple and scalable intervention of cash transfers near the top. Are the picks worse than 2012, and if so what conclusions might that lend support to?

Is GiveDirectly still a worse prospect, in 2013, than AMF was in 2012?

Since 2012 GiveDirectly has scaled up its cash transfers by some sixfold, has received extensive media attention, has attracted "8 potential donors who were considering gifts of $1-$5+ million," and has had an evaluation study completed.

GiveWell employees mostly estimated that the expected object-level effectiveness of cash transfers is a few times worse than deworming efforts, and worse still than AMF's bednet distributions. They suggest that this is compensated for by the upside potential for GiveDirectly's model to catch on and spread, either displacing even less effective interventions or increasing total aid funding. It seems to me that, with respect to this criterion, the return per dollar is worse in 2013 than 2012 in light of GiveDirectly's success. Each additional dollar makes a smaller proportional difference to GiveDirectly's resources, and its future as an organization seems far more secure.

The additional evaluation study's results were favorable (although this also reduces dependence on GiveWell intervention), but I think that on balance the expected impact of a dollar given to GiveDirectly is not stronger than it was last year.

A temporary dip in recommendation quality favors delaying donations to GiveWell's top picks

GiveWell suggests that AMF may or may not soon be able to spend its cash reserve and offer room for more funding (RFMF), whereupon GiveWell could recommend it again. Since AMF was substantially preferred by GiveWell over the charities it is recommending this year, holding money in reserve for that possibility looks good unless the value of other opportunities is declining very rapidly.

AMF itself is more optimistic than GiveWell about its RFMF situation, and suggests donating to it to increase its cash reserve for later net distributions. But if one was less confident than that, it is still possible to donate now to a fund for later distribution, claiming tax benefits and keeping annual giving pledges. Then, later in 2014 money could be disbursed if conditions improve with AMF, or some other high-value opportunity presents itself.

Logistically, this could involve setting up one's own account with a donor-advised fund, or just donating to the Giving What We Can charitable trust. Or one could donate to a trusted foundation, as some have donated to the Gates Foundation, almost against its will. I don't know whether Good Ventures would be willing to accept donations as well, but it is both closely partnered with GiveWell, and is saving the overwhelming majority of its donations until better charitable opportunities have been found.

One reason a GiveWell user might want to wait for better opportunities is GiveWell Labs, GiveWell's effort to find better options in new areas. See this interview with Holden Karnofsky:

Rob: Do you think there is a 50% or higher chance that GiveWell will find something twice as high impact as malaria within the next, say, five years?Holden: I do, though again I’m worried about being crazy.  I don’t think we can make a really quantified or irrefutable case, so we expect people to use their own judgement calls.  GiveWell will move to just putting up information about different areas and explaining what Holden thinks and where he gives his money, but we’ll tell people they’ll have to decide for themselves because it will come down to very intangible judgements.
I personally think this highly measurable stuff is just not as high impact as this “blue sky” stuff, as long as the “blue sky” stuff is done right.  I assume it’s easy to do badly and therefore gets a bad reputation, but it could be done well and be at least an order of magnitude better than global health interventions.  But that’s just my guess, and I don’t know if Elie Hassenfeld would agree with me.

Note that Holden has not recommended delaying donation for small donors, citing learning and marketing value from increasing GiveWell's money moved. And while I am also optimistic about GiveWell Labs improving GiveWell's best picks, others may disagree, as do GiveWell employies such as Elie.

But the combination of factors suggests to me that at least the expected direct impacts of donating to GiveWell's top charities will be better in 2015 than in 2014 or 2013, and perhaps the all-things-considered impact.

Charity research looks more important relative to GiveWell's money moved

If one limited one's giving universe to GiveWell's top picks, I think that recent news would favor delayed donations until further charity research or developments provide a top pick with greater returns. But it is also possible to pay for charity research rather than just donating based on its results. Declining recommendation quality makes it more urgent to find better replacements: perhaps GiveWell supporters should shift funding towards that?

GiveWell's 2012 money moved was over $9,500,000. Data for the first three quarters of 2013 showed money moved from small donors more than doubling, which has historically been well correlated with total money moved. Doubling money moved should double the value of research that delivers a 1% improvement in expected average impact per dollar.

GiveWell is currently trying to raise funds to continue and expand its own research operations while retaining independence from Good Ventures. It projects funding needs of $1.2 million for next year, up from $850,000, and would like to build some reserve capacity.

The standard counterargument would be that higher money moved increases GiveWell's capabilities, either by making charities more cooperative with it (in hopes of influencing a larger spending stream), by allowing observation of the impact of grants, or by attracting more donors to use GiveWell. However, I think this consideration has limited weight at the present margin, for several reasons.

First, total money moved of $10 million, $20 million, or more looks large relative to the cost of staff time spent by charities working with GiveWell. Second, since research activity is so much smaller than money moved, the proportional impact of a dollar is greater for research. Third, media attention for GiveWell and effective altruism are influenced by many other factors aside from money moved, including quality of research.

For example, a major factor in growth in 2013 was Peter Singer's TED talk on effective altruism, but this was not particularly dependent on GiveWell's money moved. Instead, the spread of ideas from person to person, cumulative communication efforts, and the quality of research look like they mattered more. I suspect that increased spending on charity research would do more to bolster money moved over time than spending the same funds on 2013's recommended charities.

This is not to say that research spending should rise to, say, 50% of money moved (which could cause serious credibility problems), but if money moved exceeds $20 million and research stands at 5% of that, I would rather see 10%, especially until GiveWell Labs can finish its tour of the broader landscape of charitable opportunities. So I would be happy to see more donations to GiveWell vs GiveDirectly, and those donations would be better sooner rather than later.

For myself, I take extra interest in the results from GiveWell Labs because I expect that the results will not only be helpful in the pursuit of short-term QALYs and income, but are also more likely to shed light on the best interventions from a long-run perspective.