In cryptonetworks where a token provides some kind of voting power (e.g. a DAO or proof-of-stake), we might determine the cost of each vote by calculating how much interest it would cost to borrow that token in secondary lending markets for the duration of the vote. This idea highlights the important role of time as a variable in the governance process, because the longer the period one needs to borrow the token to vote, the more expensive it is in terms of interest paid. If this is true, we can use this insight to design stronger governance systems. Protocols can’t control second-market interest rates, but they can influence the “cost of governance” by manipulating how much time it takes to complete the voting process.
There’s a growing ecosystem of on-and-offchain lending markets where anyone can lend and borrow tokens. They’re useful for what you’d expect: you can lend to earn interest, and you can borrow (paying interest) to get some leverage on your collateral or go short. But when a token has some kind of governance power over its network, it is also possible to borrow influence in a governance decision: whether it’s a protocol upgrade, a community funding proposal, or a deeper part of consensus.
To explore this back-of-the-envelope style (i.e. imperfectly), let’s consider ZRX which is used to vote on 0x governance. Some ~6.5 million ZRX tends to vote on each decision, which at ~$0.20 per token implies there’s $1.3 million backing each decision. However that number is not the actual cost of the decision. On Compound, you can borrow ZRX at ~3.30% APR, which means it would cost about 590 ZRX in interest to borrow that much ZRX for a day in order to vote. This means that, under similar conditions, total cost of the next decision would be about $120 (at $0.2/ZRX) and marginal cost per vote just about $0.000018 [1]. This number applies to both borrowing and non-borrowing voters alike, since non-borrowing voters are incurring an equivalent opportunity cost by voting instead of lending.
These can be striking numbers because they suggest it can be very cheap for a borrower with sufficient collateral (or credit) to influence a vote. And I’m not the first to make this kind of observation. Tarun Chitra highlighted how lending could threaten the security of proof-of-stake systems should the yields from lending a token ever exceed the returns from using it (however, it is possible that the security of the system may be worth more to most holders to me than the extra yield). But it could be a feature, not a bug: first because it allows us to measure the cost of voting with more accuracy and precision, and second because it means protocols and DAOs can influence the cost of their governance by changing how long it takes to vote. This insight gives us an important variable to work with in the design of these systems.
Secondary markets influence the cost of voting through token prices and interest rates. Cryptonetworks themselves cannot control either of those factors, but they can control the duration a token must be held by the user in order for their vote to count. They can make voting more or less expensive by influencing the total amount of interest vote-borrowers would have to pay to participate (which, again, also applies to non-borrowers in the form of an opportunity cost). Part of the reason 0x governance can be so cheap is that votes are tallied at a narrow moment in time (a specific block height), but are otherwise transferable immediately before and after that point. But if it took longer to vote it would be more expensive. In Decred’s proof-of-stake layer, for example, you acquire a “voting ticket” by locking a certain amount of DCR for some ~28 days before it is called to vote. This means borrowing DCR to influence the system is a lot more expensive because you’d have to pay a month’s worth of interest (plus the cost of the collateral and volatility risk).
Manipulating the cost of governance by changing the duration a token must be locked in order to vote may be an elegant solution to many problems. Systems like 0x and MakerDAO, which use tokens for voting, could emulate Decred and increase their cost of governance by locking tokens for N days past the completion of the vote before returning them to voters. Maybe there’s the possibility of an appeal process within that time frame. Or maybe you get more creative and make it variable, for example by requiring longer lock-times if a decision is particularly important or controversial.
There’s also the notion that increasing the cost of a vote must increase its value, because, in an economic sense, at equilibrium cost = price = value. So the more expensive the governance, the more valuable the decisions have to be to maintain balance. But simply making it more expensive to vote won’t necessarily create additional governance security. If it’s too expensive to govern (i.e. the cost of the decision is higher than the value of the decision), you may experience low turnout. If it’s too cheap, the system may be easily attacked.
There’s a bunch of practical issues with this hypothesis that makes it difficult to use as a reliable and precise measurement of the cost of governance in a network. For example, interest rates vary based on demand and supply side liquidity, and on a relative basis few tokens have active secondary markets we can analyze. Also, none of this addresses Tarun’s concern. I’m less worried in part because markets generally react well to change and I suspect successful networks find a new equilibrium to satisfy Tarun’s security requirements – even if we break some of these systems in the process (another part of me dares to say it may actually be good for people to be able to lend and borrow voting power, but I’ll reserve that controversy for another day). For now, I just want to highlight the idea that we can manipulate the cost of governance simply by changing how much time it takes to complete a vote. It may be difficult, while the market remains immature, to be very precise with this calculation. But we can get away with it if we’re measuring the right thing. At a minimum, we can look at time differently.
—
[1] The big hole in this calculation is that, in the current state of crypto lending (where loans are over-collateralized), as an individual you’d have to have well over $1.3 million in crypto collateralize in order to obtain the loan, which is a high and expensive hurdle for an individual. We’d have to do some additional math to add the cost of that collateral to the calculation; further, this caveat assumes a single person seeks to influence the vote in this way, to this magnitude. Still, if you have the means, the cost can be quite low. This can be problematic as it severely overpowers the wealthy: the more you have, the cheaper it is. So do take these numbers as imperfect: but my point is to highlight the effect of time, not to be precise about the cost of 0x governance.
One last thing – I do want to point out that a “low” cost of governance is not necessarily a bad thing. Just because it’s cheap doesn’t necessarily mean it’s insecure. The numbers in the 0x example may seem uncomfortably low on the surface, but it could be that $120 is the exact right cost for the kinds of decisions being debated. My take is that 0x decisions so far simply aren’t controversial enough to demand a higher cost (and again, this is a crude and imperfect calculation). Remember there’s so much we can’t predict about social systems, in or outside the lab.