There has been a lot of buzz recently about ways to fund public goods more democratically. I made an overview of proposals in a previous post.
Recently a new proposal appeared on the interwebs. Though it sounds promising, it is also quite futuristic and complex. In order to understand better how it works, I'm fleshing out the details and putting the proposal in a broader perspective below. But first, let's go back to basics.
Building block 1: what are public goods?
A quick refresher: public goods are non-excludable and non-rivalrous, which results in them being typically underprovisioned by markets. Examples include: clean air, free education and open source software. Individual sellers are not optimally incentivised to produce them because they cannot capture all of the gains from creating the goods: there is a free-rider problem. Historically, governments and nonprofits have been the institutions mainly taking care of public goods provision: whatever society as a whole wants, these organizations provide - despite it not making immediate financial sense for them.
It is important to realise that whether or not something qualifies as a public good is not a binary yes/no decision. All goods have positive and negative externalities which are not factored in the price, the longer the causal chain you look at, the more externalities which may arise. That's why the government does not just invest in clean air initiatives (the positive results of which are directly experienced by people), but also subsidizes education (the results of which may only be experienced a decade from now).
Building block 2: how can voting rights be allocated?
Voting is used as a technique for a group to coordinate on what they want. There are many, many, many different voting systems, and for good reason: they all try to resolve some fundamental questions like how do you ensure everyone can freely participate? How can voters indicate the strength of their preferences?
An important feature of voting systems is how voting rights are allocated. Over the years, many different methods have been introduced, tying the right to vote to:
- Identity (e.g. 1-person-1-vote)
- Wealth (e.g. Proof of Stake)
- Wealth and identity (e.g. Quadratic Voting)
- Computing resources (e.g. Proof of Work, Proof of Space and Time)
Thanks to some of these innovations, you no longer have to belong to a particular family or region in order to participate in your group's decisionmaking process, you can just show a passport or run your computer's mining software. The way votes are allocated has a strong influence on how fair the outcome of the group's decisionmaking process is, as well as how well they, for example, fund public goods!
Identity-based public goods funding
In systems where everyone's identity is known, there are various known systems which can be used to determine which public goods to create: from complex nation state governments, to targeted schemes like Quadratic Funding and Mutual Matching.
As outlined in this excellent blog post by Vitalik, a fundamental challenge which identity-based public goods funding mechanisms deal with is that they are prone to collusion. Individual voters can be bribed with money, priviliges or sacks of potatoes to vote for proposals which are not good for the general public. Less dramatically sounding but equally dangerous are instances of legitimate vote trading or the power which large investment firms are amassing. Privacy remains a core feature to make such systems work: an important property of voting systems is that you should not be able to prove what you voted for.
Identity-free public goods funding
It is not always possible or desired to know the identity of every participant in a system. In the case of a decentralized cryptocurrency, users may be anonymous by default, and establishing the unique identity of participants which live all around the world and which speak different languages is very costly. Even within a single jurisdiction it is still an arduous task to verify unique identities: do you use passports, a digital ID service or something else? How are you verifying correctness of the attestations? Do your users have to bring documentation or remember a password?
For these instances, you can imagine people, organizations or software directly voting and indicating their preferences with their fiat or cryptocurrency coins. However, here bribing is also an issue. Someone could offer a small bribe to rational voters, who may prefer the sure gain of a bribe over the uncertain cost that this might cause for the entire ecosystem. You can check this original post for a more detailed description, but the following summarizes the issue well:
"In all of these cases, the key reason for the failure is that while voters are collectively accountable for their votes, they are not individually accountable. If a vote leads to a bad outcome, there is no way in which someone who voted for that outcome suffers more than someone who voted against."
So, how do you let voters internalize the costs of all of the other voters? Make the cost of votes a direction function of the revealed preferences of other users:
"in order to vote on a proposal with N coins, the voter needs to put up a buy order: if the current price at the start of the vote is P, then they need to be willing to purchase an additional N coins at 0.8 * P for a period of 1 week if the vote succeeds [...] These orders can be claimed by anyone who votes against the decision."
This will have the consequence that every vote transfers funds from those who won a vote to those who lost a vote. Which does not sound like such a bad idea from an income redistribution perspective, and indeed it exactly serves the purpose of letting voters incur the total cost of a proposal.
A lot of testing and tweaking will have to happen to see if the proposal can hold up in practice. First and foremost, there is a major problem with proposals that are very (un)likely to succeed:
"if some vote has a too-close-to-zero chance of passing, there’s no incentive to commit capital to voting against it, and so in equilibrium it has to succeed at least some of the time."
Proposed mitigations (at the time of writing) include:
1. Adding a “yes bias”: making the requirement that no votes have to provide a buy order weaker, or conditional on there being a sufficient number of yes votes
2. Requiring a proposal to have a fee equal to 1% of the funds distributed, so only proposals that have at least some minimum probability of success actually get proposed
3. Using a prediction market to filter proposals: anyone can make a bet that the proposal will fail with >95% probability, and bringing the proposal to a vote requires someone making counter-bets against all open bets
4. Using a ladder of yes/no markets with varying payout ratios appropriate to a set of increasing probabilities of passing, an auxiliary market that just estimated the probability of passing, and if the estimate reached the top of the ladder, passing it stochastically so it didn’t need to deal with a singularity at 100% probability.
5. Increasing the cost to more than P to propose a vote
To conclude: a lot of exciting proposals, but don't throw away your passport just yet!