We were curious about the voting behaviour in Polkadot OpenGov so we extracted the data from the chain and did some analysis over it.
The data is up to block 18_277_977, i.e. 2023-11-22 11:02:06 (+UTC), it only includes treasury tracks, and the last month is still open.
With the analysis we aim to address the following questions:
- Who is voting?
- Are voters engaged?
- How is conviction used?
- How is exercised voting power distributed?
A total of 1529 unique addresses have voted at least once in OpenGov for treasury tracks, and 298 unique addresses have delegated their votes. When we segment these accounts by the amount of DOT they hold (free + reserve balance), it becomes evident that more than half of voting and delegating accounts hold less than 1k DOT.
We aim to understand whether new voting accounts continue to vote on proposals over time. To explore this, we present the monthly retention cohorts of voting accounts.
From the charts we can see that typically less than half of new participants entering on a month continue to vote in the following months. The original cohort (Jun-23) is the most sticky and dedicated.
OpenGov utilizes conviction voting, allowing token holders to increase their voting power by declaring how long they are willing to lock up their tokens based on a conviction multiplier. To gain an overall understanding of how conviction is used, we present a plot of the mean and median conviction for casted votes.
The resulting plot reveals that the median conviction remains consistently at 1x, while the average fluctuates between 1.05 and 1.2, indicating that token holders typically use low conviction when voting.
For deeper insight into how conviction is used when voting, we segment the votes by the amount of DOT put into a vote and plot the percentage of votes per conviction in each segment.
Here, it is clear that voters putting less than 100k DOT in a vote tend to use 0x or 1x conviction the majority of the time. On the other hand, voters with 100k-1M DOT are more willing to use higher convictions, while voters with >1M DOT never vote with 0x conviction but are also cautious about using high convictions (4x-6x).
We define voting power as the amount of DOT put to vote multiplied by the conviction multiplier. The total voting power exercised in a vote includes both self-voting power and any delegated voting power, cumulative over referenda voted by the same account.
We observe a sharp increase in the centralization of exercised voting power in the top 1% in October and November, regardless of an increase in participation.
This could be correlated with the fact that votes with conviction do not stack.
While there is not enough data to draw hard conclusions, it can serve as a future indicator.
We are curious to know your thoughts