Quadratic Voting is a concept created in 2013 by Glen Weyl and Eric Posner.
The basic idea is that you pay for votes based on the square of the amount of votes you’re buying. 1 vote is $1, 2 is $4, 3 is $9, and so on. Some people vote for, some vote against, and just like in normal voting some majority or super-majority wins.
Another important part of quadratic voting is that at the end, the funds are returned to the participants split equally among them. This helps them make up some of their lost utility if they don’t get their way.
What Glen and Eric showed was that as long as everyone follows the rules, QV approaches ideal preference aggregation as more voters participate.
That “as long as everyone follows the rules” part is also important, because one large drawback of QV is that it heavily rewards collusion – and in a decentralized system, Sybil attacks.
But let’s pretend that we had solved the collusion and Sybil attack problems. For what else could you use a preference aggregation algorithm?
Quadratic Utility Functions
When I first read the paper, one idea immediately jumped out at me – if you vote on values, instead of on policies, you can use the amount of for votes minus the amount of against votes to create a weighted linear payoff function. You could then use another system, such as forecasting tournaments or prediction markets to figure out how some policy or organization will effect those different values. Essentially, a Futarchy based system where the “Vote on values” part is handled by QV.
Quadratic Share Pricing
QV has the neat property of where the marginal cost o a vote is always the same as you payed for all previous votes, but the price increases so fast that wealth disparities have to be extreme for wealth to be a decisive factor in the majority.
One place where this seems like a desirable trait is in selling shares of the company (I recognized this this weekend when somebody misinterpreted the above idea as this idea.) You want people to be able to buy into the company more if they believe more in it, but you want to make it really hard for any one shareholder to get a controlling share of the company. If you priced shares quadratically, you get both of these desirable properties.
Random Thoughts on Quadratic Voting
- Still haven’t seen Eric or Glen address the collusion issue, which is the biggest issue right now with QV.
- One issue I see with it is that if you have a floor price of a dollar, you’re always guaranteed to get back at least your floor price. This means that voters who care LESS than a dollar are still incentivized to vote the minimum amount. This leads to a strong effect where if there’s people who care only a little on one side of an issue, they’re always incentivized to vote more than they care about, which could overwhelm the people who really care on the other side of the issue (the exact issue that QV is trying to avoid). One solution if you use cryptocurrency is to use infinitely divisible payments and infinitely divisible votes. This way, you have no floor price, and aren’t guaranteed to get back what you put in – you’re then incentivized to vote only what you care about (in practice, all cryptocurrencies have a floor price, but hopefully it’s so low that you care about the issue at least as much as the smallest denomination is worth). Of course, then you’ll have a lot of people putting in very small amounts in hopes of getting large amounts… but this will quickly become unprofitable from an other resources (time/electricity/gas) standpoint as more and more people do it.
- The third issue is that QV encourages coalitions. I haven’t seen Eric and Glen’s models, but my guess is that they don’t allow for coalitions. If you modeled it out, it’s possible that QV could ultimately end up encouraging a two party system, which is exactly what we want to avoid.
- The mechanism above shows that QV can allow wealth redistribution towards people who care less, but another interesting aspect of it is that it encourages wealth redistribution over time to people who HAVE less. This is a really cool property of QV, and also something I haven’t seen modeled anywhere. How long would it take for the effects of wealthy people buying extra votes to create total wealth redistribution.