This article serves as the first of a multi-part series delving into the intricate world of DAOs. Throughout this series, we aim to explore various characteristics of DAOs utilising on-chain data and conducting empirical tests. Our aim is to provide the reader with a deeper understanding of the operational, financial, and governance structures of DAOs.
For decentralised autonomous organisations (DAOs), the mechanism of governance plays a pivotal role in democratic decision-making. Traditionally, the concept of governance has been rigorously studied within the domain of political science and corporate finance. One focus has been on analysing how the costs associated with participating in governance, particularly voting, affect stakeholder behaviour and system efficiency. Such studies are well-established in political science, where voter turnout and engagement are weighed against the costs incurred by voters, ranging from informational to logistical expenses. Similarly, in corporate finance, the study of shareholder voting unpacks the complexities faced by shareholders who must decide whether the potential benefits of influencing company decisions outweigh the costs involved.
With the advent of blockchain technology, organisations can now leverage smart contracts (SCs) to create unique governance frameworks, potentially lowering the costs faced by voters. The transparent and immutable nature of blockchain fundamentally alters the incentive landscape for organisational governance. In traditional systems, opacity can often lead to mistrust and inefficiency, with stakeholders lacking clear insight into decision-making processes or outcomes. Conversely, the inherent transparency of public blockchains allows every action and its corresponding outcome to be viewed publicly and scrutinised by all. This openness not only fosters trust among stakeholders but also encourages broader participation in the governance process, as the barriers to understanding and verifying organisational activities are significantly reduced. Furthermore, the use of SCs in DAOs automates the enforcement of rules and execution of decisions, minimising the need for any intermediary oversight. This automation extends to the voting processes within DAOs, where SCs are used to tally votes and execute outcomes based on pre-defined criteria and stakeholder input. This capacity for automation, combined with the potential for reduced transaction costs provided by blockchains, can potentially lower barriers to governance participation, democratising the governance process and enhancing the overall efficacy and responsiveness of the organisation.
However, unlike many political systems where each person's vote carries equal weight, in many DAOs, the weight of a vote is influenced by the number of tokens a wallet votes with. This design stems from the pseudonymous nature of blockchains, where attributing a single wallet to one individual is challenging. In this context, token-weighed voting is comparable to shareholder voting in corporate governance. Although no voting system — neither "one-person, one-vote" nor weighted-voting — can be definitively deemed superior, weighted systems incentivise the accumulation of voting power as a way to influence outcomes. This leads to a centralisation of vote power linked to economic means and has led many to question the appropriateness of the term “decentralised” to describe DAOs.
For instance, in a paper titled “The Hidden Shortcoming of (D)AOs – An Empirical Study of On-Chain Governance”,[1] the authors study the distribution of voting power and governance behaviour for 21 DAOs on Ethereum. The authors use various inequality measures, such as the Gini and Nakamoto coefficients, to show that voting power is highly skewed and concentrated in the hands of a few large entities. The on-chain analysis highlights that the centralisation of voting power is persistent over time, albeit with some DAOs showing slight improvements in decentralisation measures over the study period from March 2020 to January 2023.
One way that DAOs can facilitate improvements in decentralisation is by incentivising governance participation, especially from smaller token holders. Recognising that voter turnout is affected by associated voting costs, this insight provides a pivotal motivation for the current study. Utilising interdisciplinary perspectives from both traditional vote cost analysis in political science and corporate finance, along with contemporary evaluations of DAO decentralisation, this article leverages on-chain data to investigate whether smaller voters – those holding relatively fewer governance tokens – are disproportionately sensitive to the costs of voting compared to larger token holders.
Specifically, by utilising data on Ethereum, the aim of this article is to investigate the following hypotheses:
The belief that small voters are more deterred by transaction costs than large token holders (H1) can be motivated by the fact that small voters have a higher relative cost of voting on a per-token basis. This is a consequence of how transaction fees are calculated on most Layer 1 (L1) blockchains such as Ethereum. In this context, the cost of voting can be measured by the transaction cost incurred to cast a vote. This cost is influenced by exogenous factors such as network congestion and the price of Ethereum rather than by the number of tokens used to cast a vote with. Therefore, under the same network conditions, the relative cost of voting, on a per token basis is lower for a wallet voting with many tokens compared to a wallet voting with fewer tokens. An argument can be made that in the face of higher relative voting costs, small voters are more likely to delay their votes and wait for low-cost conditions to cast their votes. While large voters still have an economic incentive to vote when it is cheap to do so, they are less strongly motivated, financially, than smaller voters.
To test this empirically, we examine the voting behaviour of wallets participating in Uniswap governance across 59 proposals. Specifically, we look at the number of tokens used for each vote, and the cost of transacting at that time of the vote. The cost of voting is estimated using the price of gas on Ethereum at the blocktime of the transaction. We treat each proposal as unique and rank wallets based on both the number of tokens they used to vote with and the cost of casting their vote. We create three equal-sized vote groups – small, medium and large – distinguishing between the size of token holders, and three cost groups – low, medium and high – distinguishing between the costs faced by voters. To test the hypothesis, we examine whether the proportion of small voters to large voters is relatively larger in the low-cost voting group compared to the high-cost group.
Before explaining the methodology and results of this study, in Section 1 we provide background on how voting in Uniswap works. In Section 2 we explain the data used and present descriptive measures of the datasets. Next, in Section 3 we present the methodology and results, rationalising both the expected and realised outcomes under the predicted hypotheses. Finally, in Section 4 we discuss why these findings are relevant to DAOs and provide insights into how DAOs may incorporate these findings to improve their own governance structure.
The Uniswap protocol is a decentralised exchange utilising the automated market maker (AMM) model, facilitating the trading of crypto-assets on various blockchains such as Ethereum, Polygon, Optimism, and several others.[2] According to its website, the protocol is a public good which is governed by holders of the protocol’s native ERC20 token - $UNI. $UNI holders engage in both off-chain and on-chain governance processes to oversee and implement changes to the protocol. This framework includes a proposal system where stakeholders can suggest alterations ranging from key protocol parameters like changing the trading fees on liquidity pools to the allocation of protocol treasury funds.
Uniswap’s formal governance process can be distinguished by three phases: [3]
In Uniswap’s governance process, both the Temperature Check and Consensus Check phases occur off-chain, sparing token holders from direct blockchain transaction costs. However, participating in these stages still involves economic costs, notably in the form of opportunity costs. Opportunity cost, a key concept in economics, refers to the potential benefits an individual misses out on when choosing one alternative over another. In the context of Uniswap governance, token holders spend time and effort to understand how to vote and the content of the proposals they are voting on, thereby incurring opportunity costs by not pursuing other activities. Though opportunity costs are subjective and challenging to measure, they are significant and should not be overlooked. For example, in general elections, a common barrier to voting is the time and effort required to reach polling stations. Democracies often seek to lower this opportunity cost to boost voter turnout, such as by increasing the number of polling stations, thereby reducing travel time for voters.
However, the third phase of Uniswap incorporates on-chain transactions, providing a measurable data point quantifying one element of the total cost faced by voters. This slice of the total vote cost faced by governance participants forms the foundation of this study.
Before discussing the data and methodology, a final feature of Uniswap’s governance architecture needs to be discussed – the idea of vote delegation.[7] In order for token holders to participate in Uniswap’s on-chain governance, they must first go through a delegation process. This entails delegating $UNI tokens to an address responsible for voting on the token holder’s behalf. $UNI holders can delegate either to an address controlled by themselves, or a third party which they entrust with their voting rights. For a vote on a proposal to be valid, the tokens must be delegated before both the voting period and the proposal period. In other words, $UNI holders can only participate in governance if they delegate their tokens in anticipation of any forthcoming proposals.
The data analysed in this study covers all 59 Uniswap proposals, from the first proposal launched in October 2020 to the latest proposal that went live in February this year. The dataset is obtained using The Graph, which is an indexing protocol used to query networks like Ethereum. It allows developers to build and publish various APIs, known as subgraphs, which can perform efficient and complex queries over blockchain data. The subgraph used in this study was open-sourced by the authors of “The Hidden Shortcoming of (D)AOs – An Empirical Study of On-Chain Governance”, who look at changes in governance participation and the distribution of voting power for a sample of DAOs overtime.
The following section provides an overview of the data points used in this study.
Figure 1 below shows the number of proposals submitted each year and their status. The statuses of Uniswap proposals—'Active,' 'Cancelled,' and 'Executed'—reflect the current state of these governance actions. 'Active' proposals are those still open for voting, 'Cancelled' proposals are those that have been withdrawn or nullified, and 'Executed' proposals are those that have successfully passed the voting process and been implemented.[8]
Figure 1: Uniswap Proposals per Year
The trend shows that governance activity as measured by proposal count has been growing year-on-year, with the exclusion of 2024 which hasn’t yet concluded and should therefore be excluded from the analysis. 2023 saw 18 executed proposals, marking a 60% increase in total executed proposals compared to 2022.
Transaction costs play a pivotal role in shaping voting behaviour and influencing decision-making outcomes. In the case of Uniswap and DAOs utilising similar governance structures, the majority of these costs can be broken down into three fundamental components. Firstly, there's the intrinsic cost of gas, an essential resource required to execute transactions on smart contract platforms. This cost is uniform for all participants and is contingent upon the complexity of the underlying operations to cast a vote. Secondly, the price of gas itself, denominated in either Wei or Gwei, reflects the network's demand-supply dynamics for transacting on Ethereum. The cost of gas prices is determined exogenously by Ethereum network conditions and is separate from Uniswap’s governance implementation. Lastly, the third factor completing the triad is the price of Ethereum which allows for the pricing of transaction costs in real-world monetary terms. Together, these three elements contribute to the economic cost voters face when participating in governance.
In this study, we focus specifically on the second component of transaction costs – gas prices. This deliberate focus stems from two key considerations. Firstly, the intrinsic cost of gas required to cast a vote is relatively uniform across all participants as it is based on the design of the Uniswap voting SCs themselves. Consequently, for the purposes of our analysis, we can safely overlook this component, as it does not significantly differentiate voter behaviour. Secondly, while examining the price of Ethereum itself can offer valuable insights, particularly when assessing transaction costs in dollar terms, we adopt a pragmatic approach by making certain assumptions. We assume that fluctuations in Ethereum's price follow a “random walk”, resulting in an equal average impact on all voters over time. Moreover, given the relatively short duration of voting periods (7 days), the price of Ethereum is expected to remain relatively stable throughout, albeit with some random fluctuations. By narrowing our focus to only examining the impact that gas price has on voting, we zero in on one element only and analyse whether the variance in the price of gas has a significant impact on voting behaviour.
To better understand the characteristics of gas prices, let us look at its distribution. Figure 2 below displays the gas prices faced by voters across all 59 Uniswap proposals.
Figure 2 - Distribution of Gas Prices across Proposals
Figure 2 above shows the notably skewed distribution of gas prices across Uniswap Proposals. Gas prices are denominated in Wei, which is the base unit of Ether (ETH) whereby 1 ETH equates to 1018 Wei. The lowest observed observation is 2.19 billion Wei while the largest observation is 339 billion Wei. The median gas price across the data set is 14.9 billion Wei. To provide some context to these numbers and the cost of voting more generally, let us calculate the average US Dollar cost of voting.
Across the sample, the average gas used (note the difference between gas used and gas prices distinguished in the earlier paragraphs) to vote on a Uniswap proposal is roughly uniform across the sample at 80,200 units of Gwei (where 1 Gwei is equal to 109 Wei). If we assume that the price of Ethereum is a modest $2,000 USD, we can calculate the average cost voting on a Uniswap proposal as follows:
We begin by multiplying the price of gas by the cost of voting to get the cost of voting denominated in Gwei (80,200 * 14.9 = 1,194,980 Gwei). This cost can be converted to units of ETH by dividing by 109 (1,194,980 / 1,000,000,000 = 0.00119498 ETH). To get the dollar value, we multiply the value in ETH by the assumed price of ETH of $2000 USD (0.00119498 ETH * $2000 = $2.39). Assuming a price of $2000 per ETH, the median transaction cost for voting on a Uniswap proposal during the sample period is roughly $2.39. It should be noted that this example is illustrative and is only provided to give some context to how voters realise transaction costs. In reality, with the high variance in gas prices, and changes in the price of ETH, many voters would face significantly higher transaction costs over the sample period.
Because of the extreme right skewness of the gas price data displayed in Figure 2, we take the natural logarithm of gas prices. The reason for this transformation is to normalise the distribution, reducing the extremity of the outliers. This is important as it directly impacts the “cutoffs” used to create the three gas price categories – a point that is expanded on in Section 3. The distribution for the logarithmically transformed gas price variable is displayed in Figure 3 below.
Figure 3 - Distribution of Log Gas Prices across Proposals
Vote power represents the number of tokens each wallet uses to vote on a proposal. The amount of $UNI used to vote can be equal to or less than the amount held in the vote wallet (depending on how many tokens the holder chooses to commit to a particular vote). The distribution of vote power across voters can vary from one proposal to another, reflecting token holders’ dynamic participation in governance. To better understand the distribution of vote power across Uniswap proposals, Figure 4 below displays the descriptive statistics of vote power across the aggregate of all 59 proposals.
Figure 4 - Vote Power Descriptives
A total of 40,907 vote transactions were recorded across the 59 Uniswap proposals. The mean vote power per wallet stands at 48,226.57 $UNI, indicating a substantial average commitment per voter. However, this average is heavily influenced by outliers, as reflected in the high standard deviation of 510,041.5 and further underscored by the extreme kurtosis value of 334.55, suggesting a distribution with very fat tails and a significant concentration of values far from the mean. The median, another measure of central tendency which is less influenced by outliers stands at roughly 1. This median vote power suggests that contrary to the average, the typical voter votes with 1 $UNI token when participating in governance.
The distribution of vote power across percentiles provides additional insights into the varying levels of engagement. At the lower end, the 1st percentile of voters committed just 0.01 $UNI, and by the 25th percentile, this only rises to 0.4 $UNI, showing minimal engagement from a quarter of voting participants. Conversely, at the upper end, the 99th percentile of vote power reaches 2,500,000 $UNI, highlighting the substantial influence exerted by the top 1% of voters. The skewness of 15.76 further points to a distribution where the majority of values are clustered near the lower end, but with a long tail extending towards higher values, emphasising the disparity in vote power among participants.
While the point of the current analysis isn't to rationalise the sample distribution, a couple of points come to mind. The fact that 50% of wallets are voting with 1 $UNI or less suggests that many participants may be voting primarily to stay active in governance. Some voters might be motivated by potential future rewards, such as airdrops, which could incentivise them to maintain a presence in the governance process. This might even lead larger token holders to split their $UNI tokens across multiple wallets to vote with, thereby boosting their apparent governance activity by casting votes through several addresses.
Concerning the higher end of the distribution, specifically the 95th and 99th percentiles where very large vote powers are observed, it is likely that these represent big players in the ecosystem. These participants may be accumulating voting power deliberately to exert substantial influence over proposal outcomes. The economic incentive to control or sway decisions that could affect the protocol’s direction or financial outcomes is a strong motivator for these stakeholders to concentrate large amounts of voting power.
To analyse whether there is significant variation in voting behaviour across small and large token holders across Uniswap proposals we categorise voters into three groups based on the number of tokens they vote with. This categorisation is done for each voting event, marked by the voting start and end time for each proposal. The groups are created by ranking the size of voters for each proposal and splitting them into (roughly) equal groups containing 33% of the observations, therefore, each observation is contained either within the Small, Medium or Large Vote group for each proposal. We name this categorical variable Vote Group. We employ a similar but slightly different approach to categorise the size of transaction costs based on the Log Gas Price variable. Instead of creating three equal groups, we generate cutoffs based on the mean and standard deviation of observations within each proposal’s voting period. Specifically, we define the “Low” Gas Price group as observations falling below 1 standard deviation from the mean, and the “High” Gas Price group as observations above 1 standard deviation above the mean. The middle category, or the “Medium” Gas Price group is comprised of observations falling between the Low and High Gas Price groups, specifically, those observations greater than 1 standard deviation below the mean and below 1 standard deviation above the mean. Figure 5 below displays the cutoffs and grouping for the observations in each Gas Price group.
Figure 5 – Log Gas Price Groups
In Figure 5, the horizontal black lines, which are set to the mean +/- 1 standard deviation, display the cutoffs for the Low and High Gas Price groups respectively. The dashed red line shows the mean across all observations at 23.8.
While Figure 5 displays the cutoff and groups across all observations and proposals, we categorise these groups at a proposal level based on their own unique distribution of gas prices. In other words, each proposal has a unique mean and standard deviation of Log Gas Prices resulting in different Gas Price groups for each proposal analysed.
Based on the design of the Gas Price group cutoffs, if voters are insensitive to gas prices, then for each Vote group (Small, Medium, Large), we would expect roughly 16% of the observations to fall within each of the Low and High Gas Price groups and roughly 68% of observations to fall within the Medium Gas Price group. This is the rationale that we use to test whether the null hypothesis holds. If there is a significant departure in the count of observations for each Vote group from these expected percentages, then the analysis provides support for H1. Specifically, we are interested in whether there is a difference in the count of observations for the Small and Large Vote groups within the High Gas Price group from the expected value of 16%.
While the statistical test is relatively straightforward it has powerful insights. By counting the number of observations (where each observation represents a distinct wallet’s vote for a certain proposal) and analysing which Gas Price group the vote falls under, we can see whether the distribution of votes departs from the expectation. This provides insights not only into the behaviour of different-sized voters but also into whether voters are sensitive to the relative cost of voting.
We first H0 for the aggregate of all Uniswap proposals. The main result of this analysis is presented in Figure 6 below.
Figure 6 – Vote Contingency Table
The main result of the study displayed in the above contingency table categorises votes based on the Gas Price groups (Low, Medium, High) and the size of the Vote group (Small, Medium, Large). The data shows the distribution of votes across these categories, with the total number of votes at 40,907. A significant Pearson chi-squared statistic of 553.42 with 4 degrees of freedom indicates a strong association between the gas price level and the voting behaviour across the different-sized vote groups. This statistic measures the likelihood that observed differences in voting behaviour across different-sized voter groups occur by chance, with a high value indicating a significant association between the two categorical variables of interest.
Each cell within the table represents the count of observations that fall into the intersection of the respective column and row. For instance, the cell value 1,611 represents the number of Small group votes cast during periods characterised by Low gas prices. Conversely, a value of 1,039 corresponds to Large group votes within the Low Gas Price group. The percentages presented in the Low/High group cells represent the percentage of Small and Large group voters out of the total voters for the associated Gas Price group.
The analysis of voting during high gas price conditions (the highlighted row) reveals a significant discrepancy in voter behaviour: the Large Vote group, contributing 2,448 votes, exhibits less sensitivity to high transaction costs compared to the smaller voters, who posted 1,405 and 1,568 votes for the Small and Medium groups, respectively. This trend indicates that higher gas prices disproportionately deter smaller voters from participating. Recall, that if voters are price insensitive, we would expect a roughly equal distribution (around 33%) of votes for each gas price category. Instead, the analysis reveals that during periods of high gas prices, out of the 5,421 votes casted, roughly 26% of the total votes come from small voters compared to roughly 45% of votes coming from large voters. This represents nearly a doubling in voter turnout for the Large group compared to the Small group. In other words, on average, during periods of high transaction costs, large voters are nearly twice as likely to vote compared to small voters.
We can also interpret Figure 6 vertically, analysing how the distribution of votes for both the Low and High Gas Price groups departs from expectations under the null hypothesis that voters are insensitive to gas prices. If voters were truly indifferent to gas costs, we would expect a roughly normal distribution of votes across the categories—approximately 16% of votes in both the low and high categories and the majority, around 68%, in the medium category, reflecting the statistical expectation set by defining these categories within one standard deviation from the mean.
For both the Small and Large Vote groups, there is an observed higher concentration of votes in the Medium Gas Price group. Specifically, nearly 80% (12,006 votes) for the Small group occur during the Medium Gas Price periods, while for the Large Vote group, about 74% (9,862 votes) of the votes are cast during the same cost bracket. However, the patterns of voting become particularly insightful when examining the high gas price periods.
For the Small Vote group, only about 9% of the votes occur during high gas price periods, significantly below the expected value of 16%. This indicates a heightened sensitivity to higher transaction costs among smaller stakeholders. In contrast, the Large Vote group shows a higher engagement during high gas periods, with approximately 18% of their votes occurring when gas prices are high, exceeding the expected 16%. This suggests that larger voters are either less affected by or more willing to absorb higher transaction costs, potentially due to experiencing a lower relative cost of voting (on a per token basis).
To see whether these trends hold over time, we conduct the same analysis presented above for each of the last three years. Specifically, we create the contingency table over the aggregate of proposals for each year and calculate the ratio of Small to Large group voters in the High Gas Price group. Figure 7 below displays these results.
Figure 7 – Proportion of Small to Large Group Voters – High Gas Group
Figure 7 shows that the ratio between Small-to-Large group voters within the High Gas Price group has trended down slightly over the last three years. A decreasing ratio implies that the participation of Small is diminishing over time. This could mean that Small voters are becoming less active or less willing to participate in governance compared to Large voters when gas prices are high. While this observation may be somewhat concerning for DAOs, further analysis is necessary to assess the strength of this relationship and to determine whether voting power is becoming less distributed during high gas price environments over time.
In this study, we show that small voters, those voting on proposals with a relatively small number of $UNI tokens, are significantly more sensitive to transaction costs (measured by the price of gas) than large voters. During times of high gas prices, (defined by one standard deviation above the mean of the log natural gas price distribution), large token holders are nearly twice as likely to vote than small token holders. Specifically, we find that during high gas price periods, 45% of the total votes come from large voters, while only 25% of votes come from small voters, and the remaining 30% of the total votes come from “medium-sized” voters. This finding deviates from our expectation under the null hypothesis, whereby we expected a roughly equal voting participation rate amongst small, medium, and large voters.
We argue that this finding can be rationalised by the fact that on a per-token basis, large token holders have a relatively lower cost of voting compared to small token holders. This is because the fixed cost of submitting a transaction is spread over a larger number of tokens, reducing the cost that can be attributed to each token.
Based on prior research, it is apparent that a significant issue facing many DAOs is maintaining a sufficiently decentralised structure. As voting power becomes concentrated in the hands of a few DAO members, its structure may start resembling more of a centralised organisation rather than a fully decentralised one. For DAOs, the issue of decentralisation isn’t only problematic from an organisational perspective, but also increasingly from a legal perspective. For instance, the soon-applicable Markets in Crypto-Assets (MiCA) regulation will affect crypto-asset services providers (CASPs). However, the scope of MiCA excludes entities which are deemed to be sufficiently decentralised. While the legislation provides no clear definition on the concept of decentralisation or on how to measure it, it is likely that for DAOs the distribution of voting power, measured by token concentration could be a defining factor.
One way that DAOs can attempt to become more decentralised is by incentivising small token holders, who face higher relative voting costs, to participate in governance. Because small token holders are more sensitive to high gas costs, it is likely that they delay voting and wait for periods when gas prices are relatively low. Because the price of gas fluctuates and is determined by exogenous network conditions, it is out of the control of the DAO. However, by changing governance parameters such as extending the voting period, the probability of entering “low gas periods” during the voting window increases. At the margin, this can motivate small token holders to vote on a proposal.
Knowing that small voters are more price-sensitive to voting than large voters, DAOs should ensure that their voting SCs are optimised to minimise the gas required to broadcast a vote to the network. While this concerns the amount of Gwei needed to submit a vote transaction, rather than the price of gas (as analysed in this study), the two are intertwined because they collectively determine the cost of voting for DAO participants. By streamlining the code within smart contracts, DAOs can reduce the cost of executable actions, which in turn reduces the fixed cost of voting.
Optimising smart contracts to use less gas for transactions reduces the expense of participating in governance for every voter. However, since small voters exhibit a higher degree of price sensitivity, such improvements will disproportionately benefit them, likely resulting in a greater increase in their participation rates. This means that even a small reduction in gas costs could have a significant effect on democratising voting outcomes, encouraging a broader and more equitable engagement across the entire spectrum of voters.
DAOs could also implement more creative solutions to tackle this problem directly. One idea is to compensate token holders for the costs they incur. The funding of such initiatives could come from a DAO’s treasury or inflationary token rewards. Furthermore, such systems can be designed to “nudge” the participation of certain groups more than others. For example, these systems could be designed to disproportionately incentivise small over large token holders.
These are only some ideas that could serve as a starting point for DAOs to consider. The fact is that DAOs can have incredibly complex structures, with many stakeholders, and the impact of transaction costs on voting behaviour is only one aspect of this complexity.
---
[1] https://link.springer.com/chapter/10.1007/978-3-031-48806-1_11
[2] https://support.uniswap.org/hc/en-us/articles/14569415293325-Networks-on-Uniswap
[3] https://uniswap.org/governance
[5] https://snapshot.org/#/uniswapgovernance.eth
[6] https://uniswap.org/governance
[7] https://docs.uniswap.org/concepts/governance/guide-to-voting
[8] https://docs.uniswap.org/contracts/v3/reference/governance/overview