Empowering The Community: The Positive Impact of Decentralized Voices on OpenGov
Introduction
In this report, we analyze the impact of the first wave of Web3 Foundation’s Decentralized Voices (DV) program on the OpenGov landscape of Polkadot (main text) and Kusama (Appendix). In addition to already available metrics (Dune), we make novel contributions in investigating the degree to which delegate’s votes overlapped as well as how voting power (defined in terms of the Banzhaf power index) distribution has changed with the DV program. In the first part of the report, we analyze the turnout, voting behavior and agreement of DV delegates. After that, we zoom out and take a look about the impact of the DV program on the voting power distribution in OpenGov.
Our analyses reveal that the DV program has had a very positive impact on OpenGov by amplifying the voices of delegates with diverse opinions and significantly contributing to a more equitable distribution of voting power.
What is the Decentralized Voices program?
The Decentralized Voices program, initiated by the Web3 Foundation, aims to enhance the decentralization of Polkadot by decentralizing its governance. In the first cohort, this program delegated 7x 1M DOT at 6x conviction on Polkadot and 6x 5k KSM at 6x conviction on Kusama, although the number of delegates has increased to 10 on both Polkadot and Kusama for the second cohort. In the first cohort, votes were delegated on the two Tipper and three Spender tracks, although additional tracks were added for Cohort 2.
The key reason for the Decentralized Voices is to allow passionate and knowledgeable participants within the ecosystem to expand their delegations and influence governance decisions. The overarching goal of the DV program is to foster a more inclusive and representative decision-making process within the Polkadot and Kusama networks, thereby strengthening the overall decentralization of the platforms.
Who are the delegates?
Delegates in the DV program are selected based on criteria such as community involvement, past voting activity, and their potential to contribute positively to the ecosystem. The roster of delegates rotates every three months to ensure fresh perspectives and continued engagement. The first roster included ChaosDAO, Jimmy Tudeski, Kukabi, Polkadotters, Polkaworld, Saxemberg, and William. Delegates are entrusted with significant voting power and are expected to represent a diverse range of views and interests within the Polkadot and Kusama communities. This structured approach to delegate selection and rotation helps maintain a dynamic and responsive governance framework.
Voting Behavior
Given the transparent nature of the blockchain, we are able to closely observe the voting behavior of the DV delegates. For this analysis, we incorporate data of Referenda 500 to 837 on Polkadot.
Turnout
In total, there were 326 referenda up for vote (of which 273 were finished as of writing). ChaosDAO voted on almost every referendum, while Kukabi and Saxemberg participated in >75% of referenda. On average, delegates voted in 67% of all finished referenda.
Voting Behavior
Polkaworld showed the highest Aye share at 78.3% (on a relatively small sample size compared to other DVs), while ChaosDAO had the highest Nay share at 45.5%. Overall, DVs voted in a fairly balanced manner, with average Aye/Nay shares at 53.9% and 35.4%, respectively. Kukabi and William have cast the highest share of Abstain votes, at 18.6% and 26.8% of the time, respectively.
Agreement between DV Delegates
In this paragraph, we analyze how many delegates agreed with other delegates in their votes. This metric compares each delegate pairwise and calculates a score of agreement. In essence, if two voters always voted the same way (and voted), then their score would be 100%. If two voters always voted differently (and voted), their score would be 0%. Non-votes are not counted in the comparison. Since every delegate voted in a significant amount of referenda, each comparison includes enough data points to be meaningful. The diagonal (the agreement within a delegate) is by definition 100% and, to illustrate that point, marked with “NA”.
The average agreement between delegates is ~69%.
Controversial Referenda
In this paragraph, we delve into a subset of referenda. Previously, we considered all referenda, including those that were obviously approved (such as non-controversial runtime upgrades) or obviously rejected (like referenda with faulty pre-images). These cases didn’t offer much insight into delegate opinions because the outcomes were clear-cut. However, for a more compelling analysis, we now focus on the controversial referenda — those where opinions varied significantly. Specifically, we examine referenda with over 20% approval despite being rejected or under 80% approval despite being accepted. While there is no agreed-upon definition of a controversial referendum, these numbers strike us as a substantial portion of voters disagreed with the final outcome. According to this criteria, we identified 85 controversial referenda in total. The following table provides the total votes of DV delegates in these referenda.
Turnout
By focusing on controversial referenda only, the overall turnout increases compared to the previous numbers, indicating that some delegates considered not voting in referenda that were already clearly accepted or rejected.
Vote directions
While the general voting behavior of most delegates aligns with their patterns in all referenda, ChaosDAO stands out as more critical in these controversial cases. Their share of Nay votes increases significantly, reaching 71.1%.
Voter Alignment
Compared to the voter alignment in all referenda, where the average was higher, the alignment in controversial referenda has dropped to around 50%. This decrease makes sense, given that controversial referenda inherently involve more disagreement. It is positive to see this anticipated divergence play out among the delegates. This vibrant mix of opinions highlights how delegates are actively enriching the OpenGov landscape with their diverse perspectives, making the process more dynamic and representative.
Voting Power
Voting Weight vs. Voting Power
The simplest way to analyze voter influence is to compare the number of (DOT) tokens or total voting weights, with voting weights being the number of tokens times the conviction multiplier. This does not, however, accurately represent the power of a voter — consider, for example, the following situation: total token supply of 100, Alice has 49, Bob has 45, and Charlie has 6. A simple majority is required to win. Since each set of two voters achieves >51 of tokens to vote, everybody’s power is equal, despite their percentage holdings of the total token supply varying greatly.
In this analysis, we use the Banzhaf power index to analyze the power structures on the highly active Medium and Big Spender tracks on Polkadot and compare the situation before versus after the introduction of the Decentralized Voices program. Snapshots were taken on January 8, 2024 (before DV) and June 3, 2024 (after DV) with a 90-day lookback period for voting extrinsics and aggregated to define an electorate, which encompasses all currently active OpenGov voters.
Medium Spender Track
Before the DV program was introduced, the 10 most powerful voters were:
The top 3 most powerful voters on the medium spender origin were 14DN… with a voting weight of 33.75M DOT and a voting power of 0.208; 16DG… with a voting weight of 30M DOT and a voting power of 0.173; and 14zP… with a voting weight of 18M DOT and a voting power of 0.104. This means that, for example, 14DN… was required in 20.8% of all winning coalitions within the Medium Spender electorate of voting DOT holders.
The best case for small voters (<0.1% of total voting weight each) of forming a coalition together would lead to a total voting weight of 8.53M DOT and a resulting voting power of 0.044, occupying spot no. 7 in the top 10 voters. Overall, the top 10 voters would then hold 77.7% of the total voting power, with the rest, meaning medium-sized voters outside the top 10, at 22.3%.
With the introduction of Decentralized Voices, the power structures changed considerably:
The total voting weight on this track increased from 175M DOT to 274M DOT, highlighting strongly increased participation by DOT holders beyond the 42M DOT in voting weight added through the DV program. 15Qu… is now the largest voter, with 15.5% of the total weight and 17.4% voting power. 15Qu… is separated from 16DG… by a single hop on-chain and thus possibly the same entity. Overall, the power structure is more balanced: While the top 3 voters held almost 50% of the voting power before the DV program, they now hold about 37%. The Decentralized Voices are sitting at ca. 2.2% voting power per DV delegate, or ca. 15.5% in total.
Voting Power Distribution (Gini Coefficient and Lorenz Curve)
To gain a clearer understanding of the distribution of voting power, we can use the Gini coefficient, a well-established measure of equality. The Gini coefficient, along with its graphical representation through the Lorenz Curve, allows us to visualize how evenly voting power is distributed among participants. Essentially, a lower Gini coefficient indicates a more equitable distribution of voting power, suggesting that it is spread more evenly across individuals. Conversely, a higher Gini coefficient signifies greater concentration of voting power in the hands of fewer individuals, highlighting inequality in the distribution. Historically, the gini coefficient is often used to track income or wealth inequality among the population of countries. Given a recent report of Credit Suisse (2022), empirical observations range from a minimum (best-case) of 0.51 to a maximum (worst-case) of 0.89. The world (weighted) average in 2022 was around 0.89, which is highly unequal. With these numbers in mind, we can look at the Gini Coefficient of voting power and its visual representation, the Lorenz Curve. Here, the cumulative share of voters is plotted against the cumulative share of voting power. If voting power was perfectly equally distributed to every voter, we’d observe a diagonal line. The larger the “belly” is, i.e., deviates from the perfectly diagonal, the larger the inequality.
In the figure above, we can clearly see that the Decentralized Voices program has had a significant and positive effect on the distribution of voting power. This is because it created a counter-weight to previously very strong voters, which meant that their relative share of voting power decreased and was redistributed to previously (almost) non-existent voters. Additionally, it motivated more DOT holders to participate in OpenGov, or to increase their conviction.
Big Spender Track
On the Big Spender track, before the introduction of DV, the following power structure was present:
In the absence of 14DN…, the next largest voter 16DG… held a lot more power. On the Big Spender track, 16DG… held 34.7% of the total voting weight, but 40.7% of the voting power. The next largest voters, 14zP… and 16Zw…, had voting weights of 18M DOT and 12.6M DOT, respectively, but both had 14.6% of the voting power — each of them was enough to achieve ca. 50% of all voting weight together with 16DG…. The top 5 voters held 83% of all voting power.
Compared to Medium Spender, the changes in the power structure after the introduction of Decentralized Voices were even larger on Big Spender:
The total voting weight increased remarkably from 86M DOT to 362M DOT. Big Spender has been a very active track attracting high turnouts, such as on referendum 684 (Chainalysis) or 714 (Interlay). Overall, the power structure on Big Spender now seems well distributed, with the top 3 voters holding 28.6% of the voting power (versus 70% before the introduction of DVs). Decentralized Voices hold about 1.7% voting power each, or ca. 11.5% in total.
The following figure plots the Lorenz Curve and states the respective Gini Coefficients.
Even more pronounced than on the medium spender track, here we see a huge improvement in terms of voting power distribution and voter participation.
Wish for Change
Since April 2024, Polkadot’s OpenGov has introduced a new track labeled “Wish for Change.” This track enables proper aggregation of DOT holders’ opinions using OpenGov’s real voting mechanism. Although the outcomes of these referenda do not trigger any immediate on-chain actions, they serve as valuable signals for the community.
Because this track was introduced after the launch of the Decentralized Voices (DV) program, there has been no delegation on this track in wave 1. Thus, the analysis does not fully align with our focus on the impact of DV. However, we still include this analysis to understand the current distribution on the track. This serves as a useful baseline for comparison with wave 2 of DV, when the program will delegate to this track as well.
The following graph plots the Lorenz Curve, illustrating the current distribution of voting power on the “Wish for Change” track. This preliminary analysis will help us later evaluate how wave 2 impacts the distribution.
Conclusion
The Decentralized Voices (DV) program has had a notably positive impact on the OpenGov landscape of Polkadot and Kusama. The program has seen significant voting turnout among delegates, with an average participation rate of 67% across all finished referenda, and some delegates like ChaosDAO participating in nearly every referendum. The presence of DV delegates has been especially influential in controversial referenda, often swaying outcomes in critical votes. The voting behavior among DV delegates is diverse, showcasing a balanced mix of Aye, Nay, and Abstain votes, which helps represent a broad spectrum of community opinions. Importantly, the introduction of the DV program has led to a more equitable distribution of voting power. Prior to the program, the top three voters on the Medium Spender track held nearly 50% of the voting power, a figure that has since decreased to about 37%. Similarly, on the Big Spender track, the power held by the top three voters dropped from 70% to 28.6%. The total voting weight on both tracks has also increased significantly, highlighting greater participation by DOT holders beyond the additional weight brought in by the DV program.
Overall, the initiative has enhanced the inclusivity and representativeness of the decision-making process within the Polkadot and Kusama networks. The power distribution has become more balanced, as indicated by a substantially lower Gini coefficient. While the DV program has markedly improved the on-chain metrics for Polkadot’s OpenGov process, it is important to acknowledge that off-chain discourse remains a vital component of governance. Even smaller community members can influence larger voters through thoughtful discussions and thorough analyses, underscoring the multifaceted nature of decentralized governance.
Special Thanks
We’d like to thank the Parity Data team for providing part of the data for this report leveraging their infrastructure DotLake. The Parity Data team, now part of the Parity Infra/Data department, is happy to support any ecosystem initiative, analysis or research with data access or exports. For more details, please refer to the following Polkadot Wiki page: https://wiki.polkadot.network/docs/parity-data-dashboards.
Appendix: Results for Kusama
Kusama, Polkadot’s “canary network”, also had its own version of the Decentralized Voices program. As the number of Referenda is much smaller on Kusama, analysis is much more difficult and values will be more variable. For the sake of completeness, the data and limited analysis is included in this appendix.
Voting Behavior
Turnout
Total number of referenda was 51.
Vote Direction
Vote Alignment (only controversial referenda)
Voting Power
Medium Spender: PreDV
Medium Spender: PostDV
Medium Spender: Gini & Lorenz
On the Medium Spender track on Kusama, voting power has already been well distributed between voters, as illustrated by the low Gini coefficient of 0.51. This remains true also after the introduction of Decentralized Voices, with a Gini coefficient of 0.55. Each DV had a voting power of ca. 4%.
Big Spender: PreDV
Big Spender: PostDV
Big Spender: Gini & Lorenz
Noteworthy for the Big Spender track on Kusama is especially the much increased voting weight, from ~1.0M KSM pre-DV to ~1.8M KSM post-DV, exceeding the 240k KSM voting weight added directly through the DV program. Responsible for this increase was mostly the high turnout for referendum 377 (Subscan). As seen on the Medium Spender track, the Gini coefficient is low both before and after the introduction of DVs. Each DV had a voting power of ca. 1.7%, very similar to the voting power of DVs on Polkadot’s (highly contested) Big Spender track.
