Authors: @jackiepoo, @alexwdong (AWG)
Written: 2021-12-14, Data pulled as of 2021-12-09
Abstract
While it is well known that the INDEX token distribution is heavily weighted towards investors (VCs, OTC trades) and founders (Set, DFP) over contributors (Jack has written about this here ), in practice, most token voters do not participate in most proposals, and as a result the largest holders do not actually dominate the IIP process the way the token distribution would suggest.
Alex and I decided to study what actually happened historically for Index Coop’s Snapshot proposals, and objectively measure participation and centralization in Index Coop’s governance process. We used Snapshot’s GraphQL API and downloaded all of Index Coop’s voting activity, and did some number crunching. This is best summarized in the following table:
Table 1: This counts the number of proposals in each category. “No Quorum” indicates when proposals don’t reach quorum. The “Has Deciding Voter” column indicates how many proposals were decided by a single vote.
First, Index Coop has a participation problem, with a large number of proposals unable to reach quorum. This problem is especially bad with Metagovernance.
Second, of the proposals that do reach quorum, half the time they are “decided” by a single vote. In our terminology, a “deciding vote” is a single vote that is on the winning coalition that, if it had decided to vote the other way, would cause the other side to win. In other words, a deciding vote is a vote that wins no matter which way it votes on that proposal.
Third, there are only nine voters who ever got to be the “deciding vote”. Most of them we actually already know and have informal relationships with, for example, 1kx. In fact, when there are important IIPs that we need passed, since we (the contributors) don’t actually have enough votes to reach quorum, we will often reach out to these voters to help us avoid deadlock!
What’s on display here is a dysfunction in the IIP governing system, where the people who are executing (the contributors) are consistently unsure if they will have the mandate to execute on anything unless they have the backing of specific activist whales who have the voting power to swing entire proposals. This has not been a problem large enough to threaten the legitimacy of the voting process so far, but is a significant vulnerability in our system.
Our recommendations:
- Move decision making power out of the IIP process and delegate it to formal bodies of contributors. This is already happening - for example the Metagovernance Committee was formed precisely to allow Index Coop to exploit its metagovernance power when the proposal does not reach quorum. We also recently formed a Wise Owls committee that was selected entirely by the contributors, and thus we formed an alternative body that can create mandates without needing the input of token holders. Index 2.0 was conceptualized precisely to advance this objective.
- Formalize and make transparent our relationships with activist whales. A prime example here is 1kx. They are an activist whale who consistently swing the results of our proposals, but they have no formal position in our org structure. My proposal is to create something like a board of directors, with whom we can have dialogues with, and represent directly the views of token holders who dominate the IIP process regardless.
Theory
The key thing we are looking for in the data are what we call “deciding voters”, which in voting theory are called “critical voters”. A voter is considered a deciding voter for a proposal if:
- They were on the winning side of a proposal
- Assuming everyone else’s votes stayed the same, if they were to change sides, they would cause the proposal’s result to change with them, meaning that they would still win
In voting theory, deciding voters are contrasted with “dummies”, who are non-critical voters. A “dummy” is a voter whose contribution is individually non-essential to the outcome of the vote. This is not necessarily a bad thing - a large scale democratic election, ideally, consists of 100% dummies, wherein there is no person who single handedly makes the difference between victory and defeat. Therefore, it’s less that having dummies is bad, but more that the presence of deciding voters is indicative of the centralization of power.
An existing way to summarize the power differential between voters is the Banzhaf power index, which estimates the probability that a given voter will be the deciding vote based on voting power. See: Wikipedia, Jake Brukhman applied this concept to MolochDAO
There’s a few problems with this model for our use case. One is that it equally weighs all permutations of voting preferences, which does not reflect reality (46% of our proposals are unanimous). More importantly, however, the Banzhaf power index assumes everyone participates, whereas in Index Coop specifically and across DAOs generally, most voters (by token weight) do not participate. Quorum for Index Coop is 5% of circulating INDEX, and most proposals struggle to reach even that.
Thus, metrics like the Nakamoto Coefficient that imagine a supermajority of the largest whales colluding with one another are completely theoretical and don’t capture the reality on the ground. In truth, one could dominate the IIP process with only 2.5% of the INDEX, assuming you reach quorum and are the biggest fish in the voting pond. More importantly, it’s relatively predictable whether or not voters participate at all, with the vast majority of potential voters not participating.
It’s best to look at the historical votes and actually measure who the deciding voters were and how often they decided the outcomes.
Methodology and Findings
What follows is a more detailed description of our methodology, assumptions, and findings. You can line this up with the code that’s in the Python Notebook in the Analytics repo.
Using GraphQL, we first pulled a list of all proposals from Snapshot. Then, we iterated over the proposals and pulled each individual vote for each proposal and collated it into a large data frame. From there, we can group and pivot by either proposal or by voter.
Data Processing steps:
- Proposals with total 0 votes are ignored as either test proposals or Snapshot data not fully propagated
- All proposals pass with 50% majority except for DG2, which requires 60% majority
- Quorum is set at a flat 100k INDEX across all proposals (we know this is not strictly true, but I think it’s a close enough approximation)
- Multi-choice poll style proposals were excluded
- For binary proposals, the first choice is FOR, the second choice is AGAINST. Specifying this is important - for all historical proposals in our analysis this is true, but future proposals may swap this order if we’re not careful.
- Proposals were categorized as IIP, DG1, DG2, or Metagovernance based off the presence of specific substrings in their titles
Findings
We’ve made all of the processed and pivoted data available in a spreadsheet here, that you’re free to copy and pivot to your heart’s content. I’m sure there are more ways to slice the data up that we haven’t done in this report.
Proposal Level Findings
Table 1 replicated from summary for convenience.
Participation
There are quite a few proposals that fail to reach quorum. The No Quorum rates are as follows:
- DG1, 5 out of 19 = 26% No Quorum
- DG2, 2 out of 12 = 17% No Quorum
- IIP, 8 out of 73 = 11% No Quorum
- Metagov, 81 out of 102 = 79% No Quorum
We talked to the Governance Working Group (GWG) about this. Anecdotally, everyone already knows this is an issue. The GWG pointed out that the Metagovernance Committee was formed precisely to allow the Index Coop to act when the voters are apathetic about the outcomes of the Metagov proposals, and furthermore when it came to other critical votes, we would proactively reach out to activist whales that we trust and have them push proposals through to reach quorum.
Deciding Voters / Decision Centralization
Among the votes that do reach quorum, about half the time the votes are dominated by at least one deciding voter, which we term “centralized”. The Deciding Vote Rates are as follows:
- DG1, 8 out of 14 = 57% centralized
- DG2, 5 out of 10 = 50% centralized
- IIP, 30 out of 65 = 46% centralized
- Metagov, 10 out of 21 = 47% centralized
Of the 53 proposals that were centralized, only 51 of them had one deciding vote, and 2 had more than one deciding vote. One proposal had 2 deciding votes, and one had 3 deciding votes. If a proposal has more than one deciding vote, that means a vote was close enough to the margin that multiple whales were needed in collaboration to make the decision go in that direction, and if any of them defected, the defector would have changed the outcome.
We decided to smooth all that out so that each proposal could be categorized as either having a deciding vote or not.
Voter Level Findings
When we group the proposals by voter and whether or not they were deciding voters or not, we find that only 9 addresses ever manage to be the deciding voter, but the distribution of decisions among those 9 addresses is also highly unequal, both along voting power lines as well as participation lines.
Table 2: This is a partial table showing all of the users who were deciding voters and some of the users who were not. “vp” here stands for “Voting Power”, i.e. the IndexPowah that they had. This amount can fluctuate across the various proposals each user votes on, so we summarized it as mean, minimum, and maximum across all the proposals they voted on.
There’s a clear difference between, say, 0x8627 that voted 33 times and was the deciding voter 100% of the time, vs. 0xaAa6, who voted 40 times, a quarter of those proposals failing to reach quorum, and only was the deciding voter once. There’s obviously a threshold minimum voting power required to even have a chance to be the deciding voter (empirically, about 24k INDEX).
Furthermore, there’s a clear difference between 0x3650, with 196k INDEX but only voting 6 times and was the deciding vote each time, and 0x4a3e, with 71k INDEX but voting 20 times and was only the deciding vote 8 times.
Evidently, included in this list of 9 is a band of “dolphin” type voters who are highly active, contribute to many votes, and as a result end up being the deciding vote sometimes or another, especially on closer margin votes. I think that’s less of a concentration of power than the voters with nearly 200k INDEX who essentially single handedly make decisions.
Recommendations
My (Jack’s) recommendations come from a place of minimizing governance dysfunction and trying to empower contributors to have the latitude to grow Index Coop without legislative deadlock. Put another way, I do not think the centralization of power is inherently bad - it’s only bad if it leads to perverse outcomes. Given the structure I see right now, there’s two main dysfunctions in the IIP governance system:
- Contributors do not have the ability to make decisions on their own, as evidenced by the high rate of proposals that do not hit quorum
- De facto nexuses of power can make single handedly make decisions, but there’s no mechanism to make these decisions transparent or accountable
These dysfunctions are a side effect of the one token one vote (1t1v) system that we’re using today. 1t1v has a lot of advantages, in that it’s a very simple way to create a legitimate group consensus, but as we’re growing as a DAO we have the opportunity to amend it to make something better.
Correspondingly, my recommendations are to rely less on 1t1v for operations, and formalize the positions of powerful activists to make their voices and decision making process clear, transparent, and accountable.
To be clear, as we get into recommendations we are moving away from concrete data and more into Just My Opinion, so obviously all of this stuff is up for debate and discussion and I welcome everyone’s thoughts on the matter.
Recommendation 1 - Rely less on token voting for operations, delegate more decision making power out of the IIP process
This is something that everyone already recognizes and we are actively taking steps towards - I just wanted to reiterate that this is the path forward, and the stakes are avoiding centralized IIP processes and proposals that fail to reach quorum. The Wise Owl council just started and we’re already seeing pods reorganize around them - for example the Liquidity Pod intends on using the Wise Owls as our legitimacy anchor rather than having to go back to IIPs all the time. Furthermore, the Metagovernance Committee was formed precisely to deal with the lack of quorum and interest in metagovernance votes.
This is great. We should keep doing this, and continue to experiment with delegating authority to accountable groups of contributors. There’s a risk we might overdelegate, but we can deal with that when that happens.
Recommendation 2 - Create a board for activist whales
This specific idea I haven’t seen floating around anywhere, so I’m interested in seeing what other people think about it. Essentially, my thought process is, if everyone knows we’re already counting on 1kx to push proposals through, and whenever they vote, they decide the result, why don’t we make that decision-making position transparent and official? They seem to have the willingness and the interest to consistently pay attention to our governing process, so I think we should actively cultivate a dialogue with them.
What I want to avoid is having backchannel relationships with 1kx, where specific individuals have a privileged relationship with the 1kx representatives. If 1kx is going to be a core part of our voting system, their thought process should be public, in exchange for prioritizing their perspectives. This would be a real benefit for the legitimacy of IIP proposal outcomes. Furthermore, for crucial proposals such as a compensation proposal, we could be explicit about gathering feedback and buy-in from the board.
The board should not simply be whales. It should have activist whales who have the interest to be active in our governing system. Right now, it seems like that’s just 1kx. But if other investors start to show more interest, I think it makes sense to start including them in the board as well.
Not recommended - an INDEX airdrop does not fix the IIP process
To be clear, I am fully in support of Pepperoni Joe’s super detailed and workshopped Community Ownership & Compensation plan here: Community Ownership & Compensation | V2 Proposal
Not only is it a staggering work of incredible cat herding, I think it’s a broadly good idea, because it enables additional compensation for additional labor and rewards the contributors properly for the work that they’ve been doing building the Coop. If the question is, “should we pay contributors more to incentivize more high quality labor,” the answer is a resounding yes.
However, “paying contributors more” is emphatically not a solution to concentrated voting power in a 1t1v system. I would go so far as to say that plutocratic vote concentration in a 1t1v system is capitalistically inevitable.
In any decision making system, what we want to ideally do is to gather power with people that we trust to do a good job, and provide mechanisms to control the damage if we end up trusting the wrong people. Providing more voting power to people that we trust does make intuitive sense, but raises a whole bunch of other hairy questions. Who gets the airdrop? How much? How do we make sure they’re not incentivized to just dump their tokens as soon as they get it? PJ has put a ton of work into building the mechanisms in the above proposal to tackle these questions fairly, and that’s purely from just a contribution reward perspective. He’s not considering at all whether or not the amount being distributed is enough to solve our governance deadlock and centralization problems.
Final thoughts
Overall, the IIP system has been working mostly okay and we have not yet seen any governance failures that are threatening to the legitimacy of the system. When we look at the voting patterns, we see consistent weaknesses, like consistent failure to reach quorum and centralized decision making from specific large activist whales. This points to a potential vulnerability in our governance, but doesn’t delegitimize any of the decisions we’ve made so far - we’ve been really lucky that the whales involved have been generally constructive and helpful for us. We have time to put into place structures that can help us grow as a DAO.
Would love to hear your thoughts on this below. Also, if anyone has connections to other DAOs that could use this kind of analysis, it turns out that the code is pretty reusable and I’d be happy to re-run this analysis or help other people fork the analysis for their own governing systems.