Innovations in decentralized governance include new ways to design voting, new forms of delegation, new institutions that move us beyond narrow token voting, and more. Yet each of these innovations relies on having sufficient participation in governance to succeed. Therefore, one of the biggest obstacles to decentralized governance is that not enough people participate.
When few people participate, projects are less decentralized, raising several interrelated concerns. Less participation:
- … makes it cheaper for adversaries to execute governance attacks;
- … raises regulatory and legal questions related to securities laws; and
- … erodes the democratic legitimacy of the organization, which may make it harder to recruit contributors, attract users, or achieve fundamental goals.
Motivated by concerns like these, people have been interested in designing incentives to induce people to vote in decentralized organizations. So in this piece, we show how to design a mechanism that is intended to get people to honestly report how costly they find it to vote — and that then pays the subset of voters that find voting least costly to actually vote.
This mechanism does so as inexpensively as possible and would be straightforward for a project to implement. As an added benefit, rewarding people for participation in this way increases decentralization, which improves the legal standing of the governance token. We conclude by discussing when projects might want to implement this mechanism.
The basic challenge to rewarding people for voting
Suppose a project for which there are N total tokens is holding a vote and wants to ensure that at least n of those tokens cast a vote. The problem is, the project doesn’t know how willing any given token holder is to vote. We can think of each token holder as being described by the price you’d have to pay them to get them to vote. Call that the token holder’s cost of voting.
The seemingly simplest solution would be for the project to offer a fixed price p for each token that casts a vote. (Paying per token avoids Sybil problems; for projects where identity is observed, the price could be per voter rather than per token and the logic would be the same.) Two things might happen.
First, the project might be overpaying. There might be more than n tokens whose cost of voting is less than p. In this case, some n‘>n tokens will cast a vote and the project will have to pay each of them p. So the total cost to the project will be p⋅n’.
Conversely, the project might be underpaying. There might be fewer than n tokens whose cost of voting is less than p. In this case, the project will fail to incentivize the n votes it was hoping to achieve.
In the case of underpaying, the natural thought might be that – having learned that costs of voting are higher than anticipated – the project could just increase the price to some p‘>p, paying the initial voters p and a next set of voters p‘. But this creates incentive problems. If token holders anticipate that the price is going to go up over time, they have a strategic incentive to wait – not accepting an offer of p even if their cost of voting is below p. As a consequence, the project will be unable to learn the true cost of voting and, even if it eventually ends up with n voters, it will have to pay them more than any of their actual costs of voting.
How the mechanism works
The key is to learn participants’ true cost of voting. There are many different mechanisms to achieve this. But in this setting, perhaps the simplest way is to consider a direct revelation mechanism: Token holders directly report their cost of voting to a smart contract; the smart contract tells token holders whether they will receive a reward for voting and what that reward is; and then the smart contract pays those voters their reward if they do in fact vote.
Call token holder i’s cost of voting ci and token holder i’s report of their cost of voting ri. We assume that there are not ntokens with negative costs of voting (i.e., where the token holders want to vote) because in that case there is no incentive problem.
Let’s see that we can design a direct revelation mechanism such that token holders want to report the truth (ri=ci ) and to vote if so instructed given the price they will be paid. Suppose the platform receives reports (r1, r2, …,rN) , in ascending order. The mechanism instructs the first n tokens to vote and offers them each a price p=rn+1. It offer all other token holders a price of p=0. If token holders report honestly, the first n token holders will in fact vote, since p=rn+1=cn+1≤ci for all i≤n. Moreover, token holders will want to report the truth about their costs of voting. Let’s see this in two steps.
First, suppose a token holder with ci<cn+1 lies. If they report ri‘<ci , they are still in the voting group and their payoff of p–ci=cn+1–ci is unchanged. If they report ri‘>ci, then one of two things happens: either ri‘≤rn and they still are in the voting group so their payoff is again unchanged or ri‘>rn in which case they are no longer offered a price for voting and their payoff goes from cn+1–ci>0 to 0. So they never benefit from lying and are sometimes hurt. That is, telling the truth is weakly dominant.
Second, suppose a token holder with ci≥cn+1 lies. If they report ri‘>ci , they are still in the non-voting group and their payoff of 0 is unchanged. If they report ri‘<ci, then one of two things happens: either ri‘>rn and they still are in the non-voting group so their payoff is again unchanged, or ri‘≤rn in which case they are offered a price for voting, but that price is insufficient to get them to vote since p–ci=cn+1–ci≤0. So they still don’t vote and their payoff remains 0.
This mechanism makes it weakly dominant for all token holders to report the truth about their costs of voting. Ultimately, n tokens vote at total cost to the platform of n⋅cn+1. No cheaper way exists to induce truth telling. If the mechanism ties any token holders’ price to their reported cost, that token holder has an incentive to lie. And if the mechanism pays different token holders different amounts, those paid less have an incentive to pretend to be the ones who are paid more. Thus, the mechanism must use a uniform price that is not tied to any of the voters’ costs of voting. This makes cn+1 the lowest price that will work.
The mechanism we just identified is an instance of the famous Vickery-Clarke-Groves (VCG) mechanism. In the VCG, all agents are paid an amount equal to their effect on the aggregate utility of all agents other than themselves. For each agent, we ask what would happen to the aggregate utility of all other agents were that agent to not participate, and that is the payment we offer to that agent for their assigned action. If any of the n lowest cost token holders didn’t participate, the n+1th token holder would need to vote to keep the total number of voters at n. This would reduce aggregate welfare by cn+1, which is that token holder’s cost of voting. Thus, the payment to each of the n voters is p=cn+1.
This mechanism has some attractive practical properties — most importantly, token holders are incentivized to be honest and only have to pay attention twice: when they make their report and when they vote. But it may also be somewhat difficult to understand, which could be an impediment to its adoption and/or use. So it is worth noting that this same outcome can be implemented with a simple ascending price mechanism. Here’s how it works.
The project breaks up the window in which the vote takes place into discrete time periods. The offered price for voting increases over these periods. In period t, the offered price is pt. Any token holder who votes at moment t is guaranteed a price of at least pt. The price continues to increase until n tokens have voted. Call the price on offer at that moment π. All token holders who voted at or before the moment that π was offered are paid a price π. Any token holder who votes after that receives no payment (but they are still free to vote until the voting window closes). So the total amount paid by the project is n⋅π.
Once the price is above a token holder’s cost of voting, they prefer voting and getting paid at least that price rather than not voting. And because the final price they are paid will be whatever turns out to be the highest price paid to any token holder, accepting as soon as the price exceeds their personal costs has no effect on the final price they are paid – meaning there is no incentive for delay. In fact, there is some downside to delay: Not accepting immediately creates some risk that enough other token holders will vote and they’ll be frozen out of the rewards.
The one moment where this logic doesn’t quite hold is for a token holder whose vote would put the number of tokens voting over the n voter requirement. For such a token holder, there is some incentive for delay, since by delaying they could affect the price they are paid. In particular, they should delay until just when the next token holder is indifferent between voting or not. Thus, we get π=cn+1.
From the project’s perspective, they are in some sense overpaying for votes under these equivalent mechanisms. In particular, they are paying all n tokens whatever it costs to get the n+1th voter. But this is inevitable, as we saw above — the project can’t price discriminate because it doesn’t know the token holders’ costs of voting. And under this mechanism the project is getting truthful revelation of voting costs, so it is paying the least it possibly can given no price discrimination.
Directly rewarding people to participate in decentralized governance makes the project more robust to governance attacks, creates stronger democratic legitimacy, and increases decentralization, which improves the legal standing of the governance token.
Rewarding people to participate in a budget-efficient manner is somewhat straightforward, and we would encourage projects to try our direct revelation mechanism because it is simple and clear-cut.
If projects do try this simple mechanism, the resulting data would be valuable for understanding governance more generally because, as a byproduct of its design, the mechanism produces data on the distribution of voting costs. Thanks to the openness of the blockchain, researchers would be able to study these costs and understand how they vary with key voter attributes like the size of holdings, holdings of other tokens, and so on, and would be able to see whether people whose costs of voting are higher vote in different directions on proposals than people whose costs are lower. (If you want advice on how to implement this mechanism or want to share your data, please be in touch.)
But, if projects experiment with rewarding participation, the clear problem will be ensuring that the subsequent increase in participation leads to well-informed and sincere voting – and not blind voting or even bot voting to harvest rewards – which we will discuss here next. Feel free to share your thoughts in meantime with @ethanbdm and @ahall_research.
Ethan Bueno de Mesquita is the Sydney Stein Professor at the Harris School of Public Policy at the University of Chicago. His research focuses on applications of game theoretic models to a variety of political phenomena. He advises tech companies and others on governance and related issues.
Andrew Hall is a Professor of Political Economy in the Graduate School of Business at Stanford University and a Professor of Political Science. He works with the a16z research lab and is an advisor to tech companies, startups, and blockchain protocols on issues at the intersection of technology, governance, and society.
Editor: Tim Sullivan
The views expressed here are those of the individual AH Capital Management, L.L.C. (“a16z”) personnel quoted and are not the views of a16z or its affiliates. Certain information contained in here has been obtained from third-party sources, including from portfolio companies of funds managed by a16z. While taken from sources believed to be reliable, a16z has not independently verified such information and makes no representations about the enduring accuracy of the information or its appropriateness for a given situation. In addition, this content may include third-party advertisements; a16z has not reviewed such advertisements and does not endorse any advertising content contained therein.
This content is provided for informational purposes only, and should not be relied upon as legal, business, investment, or tax advice. You should consult your own advisers as to those matters. References to any securities or digital assets are for illustrative purposes only, and do not constitute an investment recommendation or offer to provide investment advisory services. Furthermore, this content is not directed at nor intended for use by any investors or prospective investors, and may not under any circumstances be relied upon when making a decision to invest in any fund managed by a16z. (An offering to invest in an a16z fund will be made only by the private placement memorandum, subscription agreement, and other relevant documentation of any such fund and should be read in their entirety.) Any investments or portfolio companies mentioned, referred to, or described are not representative of all investments in vehicles managed by a16z, and there can be no assurance that the investments will be profitable or that other investments made in the future will have similar characteristics or results. A list of investments made by funds managed by Andreessen Horowitz (excluding investments for which the issuer has not provided permission for a16z to disclose publicly as well as unannounced investments in publicly traded digital assets) is available at https://a16z.com/investments/.
Charts and graphs provided within are for informational purposes solely and should not be relied upon when making any investment decision. Past performance is not indicative of future results. The content speaks only as of the date indicated. Any projections, estimates, forecasts, targets, prospects, and/or opinions expressed in these materials are subject to change without notice and may differ or be contrary to opinions expressed by others. Please see https://a16z.com/disclosures for additional important information.