Our view: altruists in general, and effective altruists in particular, should more often pay for results rather than effort, based on evaluations rather than projections, with an explicit accounting of causal responsibility.
These changes are complementary, and the impact purchase makes all of them. We hope that broader use of certificates could (1) provide useful price signals for altruistic decision-making, (2) provide stronger and better-targeted incentives to do good effectively, (3) facilitate lower-overhead philanthropy, and (4) more effectively give money to people who can use it well, without tying their hands or wasting their time.
Certificates create a “market for impact,” and the theoretical argument for certificates is essentially the same, and nearly as strong, as the theoretical argument for markets in other domains. In practice, markets are by far the most successful resource allocation mechanism that has been seriously tested. Note that certificates are more like a product market—where you buy a car—than like an equity market—where you buy a share of a car manufacturer.
This impact purchase will create an extremely thin market, and it’s much harder to predict how these purchases will play out. But it does look like even a thin market could address some real problems. The links above describe some of the advantages we see as most important. We’ve focused on the advantages which apply even for relatively thin markets. If you are interested in more information, you can read about another angle (and post questions or comments) here.
We are purchasing certificates not because we think it is probably a better way to distribute philanthropic funds, but because we think it might be. There are a number of problems with starting out at a small scale, but it still seems like the most realistic way to try it out. We’re excited to see how the experiment goes!
Fortunately for you, most of the possible concerns are on the buyer’s side. Selling a certificate is similar to normal consequentialist reasoning for accepting a grant, though simpler and with less zero-sum conflict.