Even if we think it would be reasonable to sell a certificate, why is it reasonable to buy one?
In the long run, the existence of a secondary market for certificates could function as an effective philanthropic funding mechanism, with marginal funding converted straightforwardly into additional altruistic effort. For now, the direct effect is more nebulous.
We still think there is a reasonable chance that purchasing certificates will be an effective funding mechanism even in the short term.
We can separate out two kinds of motivations for recipients:
- They did the project because they care about its impact. In this case, (a) we expect that giving them money is reasonable, (b) we expect them to care about us using our money well, and to take the certificates system more seriously, (c) we expect the recognition and information to contribute to further do-gooding.
- They did the project because they wanted to sell the certificate. In this case, the incentive effect is operating appropriately.
For now we expect participants to mostly be motivated by altruism without the expectation of receiving a payment; if the system is successful we eventually expect incentives to play a larger role.
1a. Putting money in good hands
Completing one good project is correlated with having the expertise, information and inclination to do another. So simply handing money to people who have done good work may already be defensible. It may be less effective than targeting individuals who have concrete proposals for follow-up work, and we may end up giving a significant fraction of our money to people who are unlikely to do anything good in the future.
But these costs must be compared to the costs of administering conventional grants, and to the benefits of providing funding for projects and to individuals who would tend to be overlooked by the existing system (assuming that there are diminishing returns for different kinds of work).
1b. Altruism budgets
If people’s behavior is very sensitive to consequentialist considerations, then small incentives can be relevant even for people who care about the direct impact of projects.
In practice, people seem to often adopt “budgets” for altruistic behavior, such as 10% of their income or 15 hours a week.
If people use budgets rather than prices, then it matters whether they think about selling a certificate as undoing the associated impact. If so, then the counterfactual effect of the sale is comparable to a direct contribution to their altruism budget, rather than to their normal budget. Based on my experience with effective altruists, I think this is a likely outcome of a certificate purchase.
For example, suppose that Alice sets aside $5000 a year for philanthropy. This year, she spent $3000 on an EA project and donated $2000 to AMF. If we pay Alice $4000 in exchange for the certificate for that EA project, Alice’s altruistic spending for the year is only at $2000. To bring it back up to $5000, she can donate another $3000. (She may use the $1000 profit in the same way as marginal income, or she may treat it as philanthropic income and plough it back into philanthropy.)
(Under the anticipated tax treatment of the impact purchase, this will actually be more tax-efficient than if we had simply made a $4000 donation to AMF ourselves.)
1c. Reinforcement and information
In general, implementers may not be aware of funders’ willingness-to-pay for projects. Purchasing certificates sends a price signal which may be useful, even if a recipient never sells a certificate again.
Moreover, many people are more likely to pursue activities which are rewarded or recognized. For homo economicus this may not be a big consideration, but for real people it may be an important complement to incentives, which operates even when the possibility of future payoffs is very uncertain.
2. Incentives
If a commitment to purchasing certificates were sufficiently widely known in advance, and if the funding was significant compared to other sources of funding in the area, then the certificate purchase seems to have a straightforward incentive effect.
We will be administering the impact purchase over the course of the year, and we think it is reasonably likely that we or other funders will purchase certificates for similar activities in the future. This is a far cry from a long-lasting, credible commitment. And the amounts of money involved are very small. But it may still be a strong enough force to have a (correspondingly small) direct effect on activities within the year.
Even unreliable and small incentives are still incentives. For homo economicus the unreliability actually makes little difference: if there is a 10% probability of an award in one year, the incentive effect is 1/10th as large, but so is the cost. Similarly, awards which are a small fraction of the field will have a small effect, but also a small cost.
For large projects, risk-aversion will reduce the effectiveness of uncertain incentives—though for small awards, people may also err in the opposite direction, overvaluing the potential of receiving a small award. On top of that, probabilistic rewards take the teeth out of the provision: “treat selling a certificate as if you were undoing the impact.” But these considerations do not seem large enough to completely erode the incentive effect.
The impact purchase 