PEOPLE (Constitution DAO): Analysis and Valuation of On-Chain Transaction Network
Introduction
Meme coins typically refer to cryptocurrencies that lack practical use cases or application value. These coins are mainly community-driven and exhibit high volatility compared to mainstream cryptocurrencies. When a meme coin gains community recognition and triggers FOMO (fear of missing out) sentiment, its price can skyrocket overnight, resulting in thousands of percentage gains. A few investors become wealthy as a result, but many investors suffer losses due to market fluctuations and scams.
Unlike most meme coins generated by internet meme culture (such as PEPE: Sad Frog, DOGE: Shiba Inu), PEOPLE has a specific use case. ConstitutionDAO launched a crowdfunding campaign in November 2021, attempting to bid on a manuscript of the United States Constitution. Participants donated Ethereum and received PEOPLE tokens for decision-making regarding the future use of the constitution. After the auction failed, all participants had the option to exchange PEOPLE back to Ethereum in the original ratio of 1,000,000:1. As a result, PEOPLE is deflationary, with its quantity continuously decreasing, and it has a theoretical minimum price. This sets it apart from other meme coins. This article starts with PEOPLE and explores the historical transactions of PEOPLE, analyzes trader behavior and transaction network characteristics. We also attempt to propose several factors to explain the price of meme coins.
Data
The raw data is sourced from transaction information and log data on the Ethereum blockchain. By capturing EVENTS and parsing and converting them, we obtained a total of 157,597 Ethereum transactions involving PEOPLE tokens from November 15, 2021, to May 22, 2023. The recorded transaction details include timestamps, senders, recipients, and transaction amounts. These transactions involve a total of 46,847 unique addresses. We have cleaned and filtered out duplicate nodes on transaction paths, along with their corresponding transaction data.
Transaction Types
By analyzing the transaction data fields, we are able to classify each transaction to differentiate their behaviors. The ‘transfer’ type refers to token transfers between user wallets, involving the transfer of PEOPLE coins. ‘DEX_xxx’ refers to PEOPLE coin transactions carried out by users through decentralized exchanges such as Uniswap, 1inch, and DODO. ‘mint’ and ‘burn’ represent the minting and burning of PEOPLE tokens by users. ‘uni_burn’ and ‘uni_mint’ denote transactions where users participate in providing liquidity. ‘batch_transfer’ represents batch token transfer transactions, where tokens are transferred to multiple addresses at once. Lastly, ‘MEVBot’ refers to transactions executed by robots or programs.
By analyzing transaction types, we observe that the majority of transactions in the network involve the transfer of PEOPLE tokens between user wallets (‘transfer’). Following that, there is a significant number of transactions conducted on Uniswap involving PEOPLE coins.
Address Types
To investigate the nature of PEOPLE transactions on the graph database, we aggregated the addresses involved in the transactions as nodes in the graphFrame. We differentiated between contract addresses and user addresses by examining whether an address has associated Solidity smart contract code. Contract addresses include Solidity smart contract code, whereas addresses without such code are considered externally owned accounts (EOA).
Address Returns
We aggregated the transactions in which each address participated to determine the changes in PEOPLE holdings for that address. We then calculated the ΔETH for each transaction. If 10 PEOPLE were minted, the corresponding ΔETH would be -10 * 1/1,000,000. On the other hand, if 10 PEOPLE were transferred, the corresponding ΔETH would be calculated based on the PEOPLE-ETH price at the end of the block in which the transaction occurred. We accumulated the ΔETH values in sequential order, and the lowest value in the accumulated ETH sequence represents the total cost. Next, we conducted a settlement at the last transaction of each address. We calculated the ETH and PEOPLE balances and converted PEOPLE to ETH based on the settlement price. The ratio of the remaining balance to the cost represents the address’s returns in PEOPLE transactions.
In this calculation method, we assume that users prioritize using the proceeds from selling PEOPLE to reinvest in PEOPLE. We did not consider fee income from liquidity mining. The returns are calculated based on the ETH standard. To provide a clear understanding of the returns, we did not convert them into annualized returns.
Analysis
Network Structure
To investigate the network structure of PEOPLE token transactions on-chain, we examined the degree distribution of the PEOPLE network. By observing the frequency distribution of degrees in the undirected graph, we found that the majority of nodes have low degrees. Therefore, we used a power-law distribution fitting to explore the scale-free nature of the PEOPLE network. “Scale-free” refers to the absence of an intrinsic scale, which is the result of coexistence of nodes with significantly different degrees in the network. The degree distribution of the PEOPLE network can be approximated as:
Where k is the degree of a node and p_{k} is the frequency of nodes with degree k. By taking the logarithm and performing linear regression, we estimated the values of λ for the undirected graph to be 5.5563, for the network of user-to-user transactions to be 5.4054, and for the network of user-to-contract interactions to be 4.8768. According to the classification of scale-free networks, the PEOPLE network does not exhibit significant scale-free properties ( λ∈(2,3)) but rather demonstrates a small-world network structure ( λ>3).
The PEOPLE network exhibits high clustering, with many closely connected nodes. Through a small number of hub nodes (e.g., Uniswap, empty addresses), the average path length between user nodes is short. Despite not meeting the criteria for a scale-free network, although hub nodes exist, they are not large enough to significantly influence the average distance between nodes. In other words, the majority of users interact with specific addresses for the transfer of PEOPLE tokens, and there are no transaction addresses that have a particularly significant impact on user transaction choices.
We used Closeness Centrality to measure the centrality closeness of the PEOPLE network, with an average centrality value of 0.21 (centrality = 1 in a connected undirected network). Compared to other real-world networks, this indicates that in the PEOPLE network, most user addresses do not solely rely on a single address for transactions, indicating a strong network robustness.
We explored the dynamic changes in the network structure over time windows. We selected a time window of 216,000 time blocks, which roughly corresponds to one month on the Ethereum network. Within each time window, we calculated the network composed of transactions during that month. We then shifted the time window by 10,000 blocks and observed the changes in the outdegree and indegree of users. The following graph illustrates the trend of PEOPLE price and on-chain transactions, as well as the changes in the average degree (indegree + outdegree) and the number of nodes in the PEOPLE transaction network.
Address Characteristics and Returns
We first calculated the lifecycle and maximum holdings of each address in the PEOPLE transaction network. The start block represents the block number in which the user made their earliest PEOPLE transaction.
50% of addresses hold PEOPLE for less than 2 days. Only 7% of users hold PEOPLE for more than 200 days, and this includes both exchange addresses and wallets. Looking at the first purchase time, addresses that hold PEOPLE for a longer period are usually those that purchased it earlier. It is evident that for early adopters, PEOPLE serves as both an investment asset and a commemorative item. The holding duration of externally owned account (EOA) addresses does not show a significant correlation with the maximum holdings. However, the holding duration of contract addresses is positively correlated with the maximum holdings.
We define the average transaction interval as the EOA address lifespan divided by the total number of transactions. We classify addresses into “paper hand” addresses (the bottom 25th percentile) and “diamond hand” addresses (the top 25th percentile) based on the average transaction interval. Diamond hand addresses have longer transaction intervals. We also classify addresses based on their activity, participation in providing liquidity on Uniswap, and whether they are early participants (Minters) in PEOPLE. We then calculate the returns for different user categories:
The results show that early participants have obtained significant returns. Even without considering the transaction fees from providing liquidity (LP), addresses participating in LP have higher average returns. Users participating in ConstitutionDAO, as Minters, have also received substantial consensus rewards, with an average return rate of 14.01%. The long transaction intervals indicate that experienced and thoughtful investors have achieved higher returns.
Community Detection
We transformed the PEOPLE transaction network into a directed, multi-link graph to capture the characteristics of Meme coins spreading intensively through social networks. We used the label propagation algorithm to partition the community. Initially, each node in the network was randomly assigned a different label. Then, based on the transaction direction and amount between the current label and its neighboring nodes, weights were assigned to the neighboring nodes. Labels with higher weights would dominate over smaller labels. After multiple iterations of updating and iteration, nodes with the same label were grouped into a community.
After 10 iterations, we obtained 11,454 communities. Although the community results did not converge, there were two communities with a large number of members, consisting of 22,665 and 15,727 individuals, respectively. The average number of individuals in the remaining communities was around 1.2. We classified addresses outside these two large communities as the third category. In the visualization process using neo4j’s neovis.js, we merged the multiple links between two addresses into a single link and summed the transaction amounts. The size of the nodes represents the page rank value, the color represents the community type, and the thickness of the lines represents the transaction amount.
We calculated the PageRank values of nodes in each community based on transaction frequency and amount, and identified the top ten central nodes in each community. In the first community, the central nodes are primarily cross-chain transaction addresses of several exchanges, such as Binance, OKEx, and gate.io. In the second community, the central nodes are mainly on-chain transaction addresses, such as Uniswap, 0x, and 1inch. The third category consists of more scattered user addresses.
The central node in the first community stands out, with the Binance 14th address having an interaction frequency of 13,130 times. From the scatter plot of daily transactions, it can be observed that this address remains consistently active and does not exhibit the same level of volatility as the overall transactions. Among the top three exchanges, gate.io has the fewest number of transactions but the largest transaction amount. The average return rate is 2.2546, the Minter percentage is 9.1687%, and liquidity providers account for 0.3353%.
In the second community, on-chain DEX transactions dominate, with the top-ranking addresses being trading bots active on Uniswap v3, Uniswap v2, and other DEX platforms. The first address in this community has a transaction frequency of over 50,000 times. The average return rate is 9.0938, the Minter percentage is 64.8076%, and liquidity providers account for 2.9062%.
From the statistical observations, the label propagation algorithm has successfully partitioned the two communities that are biased towards interacting with exchange-related wallets and on-chain interactions, respectively. Furthermore, the second community has a higher Minter content and a higher return rate compared to the first community, which aligns with our previous calculations and observations. Therefore, these two communities likely have significant differences in their network structures and exhibit observable differences in trading patterns.
Valuation
The value of Meme coins mainly comes from their social properties, speculative opportunities, potential commercial premiums (such as DOGE), and value storage. For Meme coins that have other business operations, their value also includes commercial premiums.
value = social properties + speculative opportunities (+ commercial premiums) (+ value storage)
We introduce Zipf’s Law to measure the social properties of PEOPLE, using the number of addresses holding PEOPLE coins as an indicator of the number of active users. The more users holding the coin, the stronger the consensus, and the more valuable the transaction network becomes. Speculative opportunities are reflected in changes in market FOMO sentiment. In this analysis, we measure market FOMO sentiment using on-chain transactions and the PEOPLE-USDT trading pair on Binance.
valuation = NlogN(1+emo)
emo = (daily trading volume — average trading volume of the past five days) / total supply
The results show that this model can predict the general trend of market prices and capture the timing of price increases. However, the model has limitations as it mainly relies on on-chain data. Once PEOPLE is listed on exchanges, trading moves from the chain to the exchanges, introducing derivatives. Therefore, the information obtained by the model is incomplete. The model does not consider the value of holding PEOPLE as a collectible, i.e., the impact of addresses holding PEOPLE for the long term on the price. The model relies heavily on PEOPLE trading volume, making it difficult to estimate its value when trading activity is low and liquidity is low.
Conclusion
The on-chain trading network of PEOPLE does not exhibit a clear scale-free property but still conforms to a small-world network structure: highly clustered, but important central nodes (such as exchanges and pools) have minimal influence on transfer users. Additionally, the community segmentation results indicate behavioral differences in the network structure between users who interact with PEOPLE addresses associated with exchanges and those who solely transact with on-chain addresses.
Users participating as Minters in PEOPLE have achieved significant average returns. Addresses that did not participate in Minter only gained profits if they were early participants. Half of the addresses hold PEOPLE for only two days. PEOPLE trading activity was active only in the early stages and has shown a downward trend since. It is evident that PEOPLE has passed the stage of active trading and high speculative value and has transitioned to its later stage as a meme coin. Most current holders consider PEOPLE as a collectible and are not seeking profits.
PEOPLE is a relatively pure M project, where initial participants were not solely driven by speculation, and the project’s development cycle is complete. Based on the analysis of PEOPLE, we will analyze other meme projects and improve the valuation model.
Note:
The data used in this analysis are publicly available on-chain data. This analysis is for educational purposes only, and all information provided is for discussion and reference only. It should not be used as a direct basis for any decision-making.
