There’s a certain cluster of ideas that seem really cool but haven’t had much success in the real world. Well, there are lots of those. They’re practically half of my blog at this point. This post is about one: in abstract terms, using distributed knowledge and incentive design to solve resource allocation problems that traditional markets can’t. Concretely, it includes things like prediction markets, impact markets, futarchy, prizes, bounties, assurance contracts, and quadratic funding. I’ll focus on impact markets but other mechanisms face similar challenges.

Why haven’t impact markets created much value in practice?

Multi-sided marketplaces

The most apparent problem is that impact markets require three parties to reach their full potential: projects, investors, and retrofunders. A three-sided marketplace is pretty challenging to set up in general, and especially when it’s inventing new roles that will need to be filled. By itself this isn’t an insurmountable problem; it might even be the most straightforward one. You just need to find lots of attention and talent and money and get it all in the same place. Some approaches are discussed here on Astral Codex Ten.

Adverse selection

One of the advantages of an impact market is that it can fund projects that established institutions wouldn’t. But that’s very much a double-edged sword! As much as we love to romanticize contrarianism, and as memorable as stories like Katalin Karikó’s are, a lot of the ideas turned down by established institutions are just… mediocre. These are the projects that are going to be overrepresented in an impact market, whereas the obviously good projects are going to take funding from traditional sources—why would they deal with this complex and risky new system if they don’t have to? It’ll take a lot of smart money to filter through the remaining projects and overcome this disadvantage.

Competing for talent

If you want a market to make the best possible decisions, you want it to get consistent full-time attention from smart and coordinated groups of people, i.e., investment companies. The problem is that starting a company is hard. Relatively few people can do it well, and those people already have some very attractive opportunities.

So a small financial incentive isn’t going to cut it. Sure, most people would pick up a free $20 on the sidewalk, but we’re talking about people that see free $20 bills everywhere. Impact markets would have to offer enough of an upside to outweigh the opportunity cost of a more traditional venture. And if the market is hoping to attract profit-interested founders too, it’ll face a similar challenge there. The necessary incentives are a significant fraction of any realistically achievable volume, big enough to make funders take a second look at the alternatives.

The baseline is actually pretty strong

Suppose an impact market is wildly successful. It’s doing millions in daily volume, and investment firms are hiring the best and brightest to optimize their trading strategies. How much counterfactual value does the market provide? It’s hard to say because the quality of decision-making in the traditional system is going to vary a lot. Sure, you can find examples of nonprofits that are inefficient or even downright corrupt, which will make your impact market look pretty good. I’d argue that’s the wrong baseline for systems developed and used by serious impact-oriented people. The correct baseline would look more like Open Philanthropy. Good foundations already have plenty of things to fund, and unlike Wall Street they don’t have to divert any of their resources to arbitrage strategies of questionable social value.

It seems like the combined brainpower of the finance-adjacent should have something to offer here. And yeah, it probably would. On the other hand: if I were watching NVIDIA’s stock price for the last 10 years, and you were reading the Open Phil blog, who would have been first to see what was coming with AI?

None of this is to say that impact markets are entirely doomed. In principle most of their problems could be solved with large enough sums of money. But they aren’t a slam dunk. The ultimate goal is to make good, cost-effective decisions, and when incentive design gets in the way of that, it’s probably not worth the tradeoff.