Impact Markets: A Funding Mechanism for Speculative Work

A quick primer on our work at GoodX

What are impact markets?

“Impact markets” are an alternative way to fund socially impactful projects. They are best explained by examples. Imagine the following scenario:  

It’s 2025. A funder who is interested in AI safety announces a prize of $1 million for any contributions to an important problem in the field. A team of researchers at a startup nonprofit have loads of ideas for relevant experiments, but they need money for compute, rent, and coffee. A shrewd impact investor – who cares about impact, but also wants to make money – knows that they have potential from her connections, so she gives them $150,000 in exchange for an 80% share of the prize. She estimates that they have a 25% chance of winning the prize, meaning that she’ll make a $50,000 (~ 33%) profit on her investment in expectation.

The funder, who has no private information about the various teams, thinks that only 1 in 10 teams can win. So it would cost them $1.5 million to solve the problem with prospective funding.

The project succeeds – the researchers make a substantive contribution to AI safety. The funder pays out $800k to the investor, who makes a big profit, and $200k to the researchers.

With impact markets, philanthropic funders offer prizes for specific, impactful discoveries or achievements. These can be broad or narrow. For example, they might offer $3 million to whoever:

• solves a specific research problem in machine learning

• creates a new effective medicine

• develops an intervention to help people in poverty

• finds a way to produce cultured meat much more cheaply

• does research into a crucial question in longtermism.

For some examples of possible prize questions, see Dawn’s list of ideas for EA Forum research projects, Holden Karnofsky’s “Important, actionable research questions for the most important century”, and these lists of AI safety research ideas. 

We call the organizations that offer prizes “funders” or “retrofunders” (because they fund things retroactively).

Profit-motivated investors then fund researchers or entrepreneurs to work on these problems, in exchange for a share of the prize if the project leaders succeed. When projects succeed, the retrofunder pays out the prize and the investors make a profit. If a project fails, the investors lose their investment. Overall, altruists get paid; investors are incentivized to identify and fund impactful projects; and useful work gets done. 

The aim of impact markets is to bring some of the advantages of the VC investment ecosystem into the nonprofit world. Further evolutions of impact markets might include other features of conventional markets, for example stocks and derivatives.

What are the advantages?

Better access to funding for researchers doing impactful work

There are very few EA funders, and getting funding for EA projects is hard. Under the status quo, if a researcher wants to get funded to attempt an altruistic project, they must persuade one of only a handful of organizations to fund them. With impact markets, the same researcher would have access to a wide network of venture capitalists and other for-profit investors,  including personal connections like friends or family. This would be especially beneficial for EAs who are less well-networked within EA, who don’t live in major hubs, or who want to try out more speculative research directions. 

Better access to infrastructure

With impact markets, investors could buy infrastructure that’s needed for the sort of projects that they want to invest in, saving funders and researchers money through economies of scale. For example, an investor interested in cultured meat research might buy a lab and rent it to researchers they’re funding in return for a greater share in the impact certificate. Similarly, investors interested in seed-funding AI safety work might buy a whole GPU server (rather than researchers having to individually rent compute from Amazon), or provide large clean datasets.

Easier for funders to scale

Currently, the staff working at big EA grantmaking organizations have two main jobs: global priorities research, and start-up picking. They need to be able to figure out both which broad types of project are likely to be most impactful and cost-effective, and which teams are most likely to accomplish those projects successfully. With impact markets, the start-up picking would be outsourced to for-profit investors, meaning that the funders could focus on creating impactful prizes and evaluating successful contenders. Because funders would only have to evaluate the successful projects, this would greatly increase their capacity: if (for example) 1 in 10 projects in a certain area tend to succeed, impact markets would enable a funder to evaluate and fund ten times more projects in that area.

More impact and profit for investors

Impact investors we’ve spoken to say that they’re keen to invest in things that are highly impactful and at least moderately profitable, but they struggle to find these opportunities. Impact markets would bring impact and profitability closer together. However, we don’t think that impact markets will be limited to impact investors – ordinary, non-impact-focused investors are likely to be interested as well, since the profit from these prizes could be just as high as the profit from alternative, more standard investments.

What are the risks (and how might we mitigate them)?

The biggest general risk of impact markets is that profit can only be positive, but impact can be both positive and negative. We call this risk the “distribution mismatch” between profit and impact. If an investor funds a bad project, the worst that can happen to them personally is that they lose their investment, whereas the worst that can happen overall is (potentially) extremely bad. 

To use a simple example, imagine that there’s a project developing a vaccine against viral pandemics. This project has a 90% chance of succeeding and saving many lives; but the vaccine can only be developed by engineering a deadly virus in a lab, and there’s a 10% chance of that pandemic leaking out and killing millions. Under the status quo, no grantmaker would fund such a risky project. But with impact markets, an investor could be bamboozled into funding this by an unscrupulous scientist who downplays or underestimates the risk. (Here’s another example.)

Projects like this will be retroactively funded when they succeed – thus incentivizing project leaders to start them and investors to fund them – and where disasters happen, the loss to the investor is not proportional to the harm caused. Over time, we might expect many risky projects of this kind to happen, meaning that eventually we’ll get unlucky and there’ll be a disaster. 

We think that we can mitigate this risk (and others like it) in the following ways:

Curating access to the market

We could curate access to the impact market for any of the actors involved: investors, retrofunders, and project leaders (in different configurations). This would mitigate the risks that researchers would embark on dangerous or harmful projects, that investors could be persuaded to fund them, and that rogue retrofunders would propose harmful bounties.

Encouraging cooperation

We want to anchor impact markets on gains from moral trade. This might involve retrofunders cooperating so as not to fund projects that cancel each other out. Retrofunder cooperation can also help people avoid the retrofunding equivalent of the Giver’s Dilemma, where people hold off on donating and wait to see if others will fill funding gaps. 

Internalizing externalities

We do this by rewarding helpful exposés of certificates that violate our market norms. In the future, there might be more mechanisms to do this, for example:

• forcing project leads to buy insurance against the certificate’s removal from the market,

• introducing shorting to the market to allow people to profit from exposing bad certificates,

• introducing collateral.

To read more about the possible risks of impact markets and how we might mitigate them, see here and (in summary form) here.  

Sounds great, sign me up!

Interested to learn more? We (GoodX) are currently building infrastructure for impact markets. We’re looking for the following types of people to help us. 

• Funders interested in using our platform

  • We want to talk to you and find out exactly how we can make our impact markets maximally useful for you!

• Software development (TypeScript + React + Prisma, fully typed)

You can also find out more by reading this more detailed post, joining our Discord, or checking out our website.