Retroactive public goods funding: How to improve Karl and Vitalik’s really good idea

Ellie Rennie
8 min readAug 4, 2021

Retroactive public goods funding uses tokens and incentives to encourage developers to work on projects that benefit the ecosystem. In this post, I look back at a similar approach from the field of welfare policy called ‘social impact bonds’. I propose that we commence trials of retroactive public goods funding with projects that involve complex social dynamics.

“My Dad works for Ethereum and he says scalability is coming”. The tweet stuck in my head long after it disappeared from my feed. It still makes me wince at the idea of Ethereum being a regular corporation rather than a distributed infrastructure. But the second part is harsh because it’s true; Ethereum has been talking about addressing the cost and speed of the network (scalability) since forever.

At ETHCC in July, Karl Floersch admitted that he spends a lot of time thinking about why Ethereum’s roadmap is so slow to execute [1]. Floersch now works at Optimism, a company building scaling solutions for Ethereum. Prior to that he was a researcher at the Ethereum Foundation, which makes him the closest thing to the Dad in that tweet as you can get. Floersch’s ETHCC presentation introduced a new approach to dealing with the speed of change in Ethereum called ‘retroactive public goods funding’, which was first proposed by Ethereum’s founder, Vitalik Buterin (see this post and this talk).

What is retroactive public goods funding?

The concept, as Floersch and Buterin describe it, changes the ‘metagame’ of what gets built by using Ethereum’s distinct sociotechnical mechanisms to encourage investment in things that benefit the system as a whole. In very basic terms it would work like this:

· An open source project develops a mission and strategy for achieving it, and perhaps a prototype.

· The project founders create a token and sell tokens to people who believe in the project in order to raise funds to develop it.

· If the project succeeds, a ‘results oracle’ buys the tokens for a higher sum, effectively providing the same kind of ‘founder exit’ incentive that drives private good tech innovations.

A lot of details are missing at this point, including how the results oracle would work. Optimism is proposing to fund the results oracle through earnings generated from sequencing (processing transactions), sparking an interesting debate on whether this would create a virtuous cycle for Optimism. Floersch and Buterin both emphasise that such a system would depend on workable governance for the results oracle, ensuring it cannot be gamed. The mechanics of governance design for this hypothetical DAO or prediction market would be one of the harder problems to solve. They also acknowledge that it is important to recognise specialisation and suggest that we would need ways to reward earlier innovations that the funded project builds on.

Some open source software projects may be straightforward under this model, particularly where the product can be measured by integration into decentralised applications (dapps). However, software innovations that interface with users directly are likely to confront three main difficulties:

1. Technology adoption is unpredictable. Can a ‘systems approach’ to public goods that have complex dynamics be incorporated into this mechanism?

2. Measuring outcomes is hard. Who decides what success looks like and how do we ensure they are independent and incorruptible?

3. Outcomes-based funding can lead to perverse incentives if done poorly. Will organisations work on the hard stuff, or just the things that are easy to measure?

Getting retroactive public goods funding to work in cases that rely on a mix of technical and social innovation could prove important for accelerating some of the critical needs in the blockchain ecosystem, including identity. I have provided an example of how I think it could work below.

Before I get to that, it is useful to reflect on what we have learnt from similar experiments with retroactive funding in welfare policy — what’s sometimes called ‘social innovation’. A retroactive funding mechanism called ‘social impact bonds’ has been tried in various contexts and countries, revealing benefits and drawbacks.

What is a social impact bond?

A social impact bond typically involves private companies funding projects that address social problems [2]. So far it has been used in areas such as reducing recidivism and homelessness. If the project meets predefined targets, the government pays the private investor a sum greater than the original investment [3]. If the project fails, the investor receives nothing. Social impact bonds are intended to encourage innovative approaches to service delivery and efficiency in government spending.

The main difference between social impact bonds and the retroactive public goods model described by Floersch and Buterin is that the results oracle in social impact bonds is a combination of independent research (such as this company) and the government, which pays out if outcomes are achieved. With social impact bonds, those undertaking the work do get paid at the time they do the work but there is no founder exit incentive [4].

Social impact bonds can encourage more holistic approaches to solving a problem than typical public funding programs might. I observed this with an Indigenous organisation I worked with that attempted to a social impact bond off the ground. Those involved found they had to develop partnerships with other service providers as it became clear that addressing one part of a problem was unlikely to achieve the desired outcomes.

A major critique of social impact bonds is that, unless the outcomes are clearly defined, organisations undertaking the work are incentivised to focus on the easy cases. For instance, if the outcome is reducing recidivism, organisations might become selective, working only with prisoners who stand the best chance of avoiding the justice system. Those who are most marginal could lose out on services they would have otherwise received.

Implications for Retroactive Public Goods Funding

Getting blockchain-based public goods projects off the ground may also require looking beyond a single provider and considering the different actors that make an innovation work. For example, I have been working closely with a group called the Trust Alliance, which was established to develop practices and standards that promote the use of verifiable credentials in humanitarian work.

The Trust Alliance is not a specific software product; it is a network of organisations that are utilising these credentials because they meet ‘humanity first’ principals in technology design. The first tech vendor the Trust Alliance worked with is called Traverse, which was developed by the Australian Red Cross with technology company Type Human [5]. As I have written elsewhere, the value of the Trust Alliance is that it brings organisations together to identify common needs that can be addressed through the use of verifiable credentials. This extends beyond training to credentialising sector wide standards in areas such as workplace behaviour as a way to prevent perpetrators of misconduct being re-employed in the sector.

Where is the public good in this example? A public good is something that is non-rival and non-excludable. uPort’s open source software that enables other groups to develop out an ecosystem of dapps that are crafted to meet specific needs meets this definition. It could be argued that Traverse and the Trust Alliance are working on something more akin to a club good (developed by a group for their own purposes). However, the innovations coming out of this alliance could make their way into broader use as humanity first approaches to technology become the norm rather than the exception.

Tools for specific humanitarian purposes are unlikely to attract private investment. While organisations like Australian Red Cross can provide initial funding to build these technologies, unless there is a business model in place the technology will fold when the organisation’s funding runs out. Traverse, for instance has had to consider offering products to non-humanitarian customers that fall outside of the Trust Alliance privacy expectations in order to sustain itself. The market for Web3 verifiable credentials is simply not mature yet.

How could retroactive public goods funding work in this scenario? First, Traverse could launch a token and sell it to those who see a need for verifiable credentials, which could include humanitarian organisations that wish to use it now. During its token launch, Traverse might allocate some portion of its tokens to those who worked on the code libraries that their product builds on (uPort, for instance). They might also allocate some tokens to the Trust Alliance for its work in coordinating organisations, educating the sector about the technology, and surfacing credentials through field experiments. A portion of the proceeds (not tokens) would also be directed at independent research for the evaluation of the work.

On a long enough horizon, the results oracle is simply committing to purchasing Traverse tokens when the wider market is ready for Web3. For instance, the oracle would commit to purchasing the tokens when a critical threshold of customers for the app or a component of its software has been reached. The results oracle effectively compensates the humanitarian organisations that supported it for their original investment if not more by buying their tokens. Those who provided essential software components (uPort), as well as those who coordinated the necessary ‘off chain’ relationships (the Trust Alliance) would also be rewarded through this mechanism, which they could direct at future projects (assuming they are non-profit entities). The only component that would not have this ability is the research team; they would need to be funded from the initial investment so that there is no risk of them distorting findings for their own profit.

As this example shows, multi-partner projects may be easier to administer through token allocations than typical service delivery contracting processes. Open source projects can also receive rewards for work they have done once it is in use. However, as with social impact bonds, results oracles will need to find ways to incentivise developers to work on the hard stuff and not just the low hanging fruit.


Retroactive public goods funding is one of the more underdeveloped ideas in blockchain right now. It relies on a combination of good token design, governance integrity and a future-oriented mindset. Secondary markets may make this mechanism work even better, facilitating price discovery and liquidity, and providing a decentralised solution to the public goods problem. While that might seem too hard to contemplate, remember that Ethereum has achieved far more in six years than anyone expected. If you don’t believe me, ask your Dad who works for Ethereum.

About me: I am a Professor at RMIT, working across the RMIT Blockchain Innovation Hub and the Digital Ethnography Research Centre. I am also an Associate Investigator (AI 🤖) of the ARC Centre of Excellence for Automated Decision-Making and Society. I acknowledge the support of the Australian Research Council, FT19010372.


[1] Ethereum generates fees from use, but a disproportionate amount of this is directed to securing the network by rewarding those who, in good faith, run the software that keeps the ledger functional and distributed. In Buterin’s view, the social contract that underpins Ethereum makes it highly unlikely that public goods will ever be funded from this source. Grants from the Ethereum Foundation or donor funding paired with mechanisms such as quadratic funding (Gitcoin grants), has kept the protocol moving forward. These systems have funded important developments but they are slow compared to other parts of the system where tokenomics have been a rocket booster to market-based innovation.

[2] Social impact bonds were invented by New Zealand economist and public servant Ronnie Horash in 1988, and have been extensively trialled in the UK. This paper by Edmiston and Nicholls provides a good overview of learnings so far. I first looked into it when doing research for Goolarri Media Enterprises, an Indigenous organisation in Broome. Jason Potts described our interest in it here.

[3] In reality, investors in such bonds have included philanthropists who are not seeking a return but who are interested in supporting innovative and multi-stakeholder approaches to addressing social problems.

[4] Like other ‘outcomes-based funding’ models, any remaining funds can be kept rather that returned to the funder.

[5] Sempo and Mattr are also members and working on dapps with for-purpose organisations in ways that are responsive to on-the-ground challenges such as digital inclusion.



Ellie Rennie

Professor at RMIT University, Melbourne. Australian Research Council Future Fellow 2020–2025: “Cooperation Through Code” (FT190100372) Twitter: @elinorrennie