Interpretation of Consensys Paper: Is Ethereum Becoming More Centralized?

Author: Deep Tide TechFlow

In the cryptocurrency space, Ethereum has become a focal point.

Decentralization is not only the core promise of Ethereum but also the main source of its appeal. However, with the upgrades and developments of Ethereum, the community has begun to question its true level of decentralization.

These doubts are not unfounded, as we often see scattered posts discussing "Ethereum becoming centralized" from various perspectives.

Decentralization involves not only the dispersion of power and control but also the network's resistance to censorship, security, and transparency. Therefore, understanding the level of decentralization in Ethereum is crucial for investors, developers, and ordinary users.

To address these concerns, Simon Brown, a researcher at Consensys, the parent company of MetaMask, decided to conduct an in-depth study using a more rational, rigorous, and systematic approach, employing various metrics and models for academic analysis, aiming to reveal the true state of decentralization on Ethereum.

The original paper is titled "Measuring the Concentration of Control in Contemporary Ethereum."

I. Methodology: How to Measure the Level of Decentralization in Ethereum?

Decentralization is a complex concept that involves multiple dimensions and perspectives. To accurately measure the level of decentralization in Ethereum, the author Simon Brown employed various statistical methods and metrics for comprehensive measurement.

First, he introduced the Gini index, a statistical measure of inequality commonly used in wealth distribution statistics.

Here, it is used to measure the distribution of resources or power within the network. The Gini index ranges from 0 to 1, where 0 indicates complete equality and 1 indicates complete inequality. For example, a Gini index value of 0.9 means that resources or power in the network are highly concentrated.

In addition to the Gini index, the researchers also considered: HHI (Herfindahl-Hirschman Index): commonly used to measure the level of competition in specific industry sectors;

Shannon Index: sensitive to the size of distributions and the diversity of different values within the data, thus highlighting differences in distributions that the Gini index may not capture.

Atkinson Index: based on social welfare methods, inferring how much resource redistribution is needed to achieve a certain level of equality.

Jensen-Shannon Divergence: since we are interested in the changing levels of decentralization over time, this method was used to measure the similarity or dissimilarity between two probability distributions (note: in simple terms, whether the level of centralization in Ethereum is similar or significantly different before and after a certain point in time).

Nakamoto Coefficient: a measure of the resources or power controlled by the largest n entities in the system. For example, if the top 5 entities control over 50% of the resources, the Nakamoto coefficient is 5. This provides an intuitive sense of the influence of major entities in the network.

The selection of these metrics and models is not arbitrary. They have been carefully chosen to ensure that various aspects of decentralization are captured from different angles. This multi-faceted approach ensures the comprehensiveness and accuracy of the results, helping researchers better understand the true state of decentralization in Ethereum.

II. Research Subjects: Which Metrics to Represent the Level of Decentralization in Ethereum?

To conduct an in-depth study of Ethereum's level of decentralization, it is first necessary to select appropriate data and subjects that can represent various important aspects of the Ethereum network and provide meaningful metrics for decentralization.

With Ethereum's transition to POS and the emergence of new technologies such as MEV and account abstraction, this paper considers various aspects and processes involved in an Ethereum exchange and summarizes the following suitable metrics to measure the level of decentralization in Ethereum:

Based on the Original Nakamoto Coefficient:

Consensus nodes by client

Consensus nodes by country

Execution nodes by client

Execution nodes by country

Distribution of native asset by amount

Amount staked by pool/staking service provider

Metrics Related to PBS:

PBS refers to blocks proposed by builders and relayers, which play a key role in the Ethereum network.

Blocks proposed by builder

Blocks proposed by relay

Metrics Related to Account Abstraction:

Number of user operations per bundler

Number of wallets per deployer

Other Metrics:

Effective inflation rate adjusted for burn

Percentage of total supply staked

Layer 2 rollups by relative TVL

Additionally, different weights were assigned to different metrics.

III. Results and Interpretation: How Decentralized is Ethereum?

The researchers conducted a 90-day observation of Ethereum from May 23, 2023, to August 23, 2023. The overall research goal was to explore how the dynamic characteristic of decentralization changes over time within Ethereum, rather than being a static quality that remains balanced.

A. Result Discussion

From the data below, it can be seen that when the research model is applied to the same data, different indices yield different results. Taking the Gini coefficient as an example:

Execution nodes by country: Gini index of 0.85

Execution nodes by client: Gini index of 0.74

Consensus nodes by country: Gini index of 0.79

Consensus nodes by client: Gini index of 0.57

Distribution of native asset: Gini index of 0.91

Amount staked by pool/staking service provider: Gini index of 0.76

Blocks proposed by builder: Gini index of 0.78

Blocks proposed by relay: Gini index of 0.54

Number of user operations per bundler: Gini index of 0.13

Number of wallets per deployer: Gini index of 0.03

Layer 2 rollups by relative TVL: Gini index of 0.87

The color coding of these values is for ease of reading, with colors closer to red indicating higher concentration and colors closer to green indicating higher decentralization.

It is evident that the levels of centralization in Ethereum are very high for the metrics of stablecoins and staking pools. This seems to correspond with the high market shares of USDC and USDT in reality, as well as the leading position of LDO in the liquid staking sector.

B. Results Based on the Original Nakamoto Coefficient Subsystem Selection

Network nodes: When the researchers examined the data related to the measurement dimensions based on the original Nakamoto coefficient subsystem, they first applied the Gini index to metrics related to network nodes.

From these analyses, we can observe that both "execution nodes by country" and "consensus nodes by country" maintain relatively high Gini index values of 0.85 and 0.79, respectively. This indicates that these consensus nodes and execution nodes are highly concentrated in certain countries.

C. Metrics Related to PBS

PBS refers to blocks proposed by builders and relayers. The data shows that between the first and last 24-hour intervals of the dataset, the number of blocks proposed by a single builder and relay, as well as the overall concentration in the market, changed significantly. This indicates that the level of decentralization in the Ethereum network increased during this period.

D. Other Metrics

When we examine rollups by TVL, we observe a noticeable level of centralization, with a 90-day average Gini index value of 0.87. These values are meaningful when observing the underlying data, where a single rollup (Arbitrum One) accounts for 54.3% of the total TVL (down from 64.5% at the start of the sample). The second is Optimism, with a TVL of 25.9%. This means that although multiple rollups exist, two of them dominate the majority of the market share.

E. Metrics Related to Account Abstraction

This section of the data mainly focuses on the impact of ERC-4337 bundlers. The data shows:

The 90-day Gini index for User Operations by Bundler ranges from 0 to 1, with a median of 0.6 and a standard deviation of 0.36.

A specific bundler holds 76% of the market share among 13 active bundlers.

These data indicate that the account abstraction space is still in its early stages, and the infrastructure has not seen significant maturity or adoption. The impact of ERC-4337 bundlers on the network's level of decentralization is relatively limited, so a moderate level of centralization is acceptable.

F. Main Index Reveals Overall Increase in Centralization

The main index provides us with a high-level metric to measure how the level of decentralization in the entire network changes over time (from 2023/05/23 to 08/17).

The Gini index decreased from 21% to 14% across all measurement dimensions, while the HHI index slightly increased from 7.5% to 10%.

Significance:

The decrease in the Gini index indicates an increase in decentralization or a decrease in concentration.

The increase in the HHI index indicates an increase in concentration.

Conclusion: By synthesizing these two indices, we can say that overall, the level of decentralization in the Ethereum network increased during this period.

IV. Overall Conclusion and Future Outlook

Through in-depth research and data analysis, the researchers have drawn the following conclusions about the level of decentralization in Ethereum:

The level of decentralization in Ethereum is dynamic: rather than being a static quality that remains balanced, decentralization is a dynamic characteristic that changes over time. This means that as the Ethereum ecosystem develops and evolves, its level of decentralization will also change accordingly.

Different measurement dimensions show different levels of concentration: for example, the classification of execution nodes and consensus nodes by country and client shows different levels of concentration. This emphasizes the importance of employing multiple methods to measure decentralization.

Certain areas exhibit high levels of centralization: for instance, in rollups by TVL, Arbitrum One and Optimism hold a significant portion of the market share, indicating clear centralization.

Continuous efforts are needed to maintain decentralization: the research results clearly indicate that the entire Ethereum ecosystem shows elements of concentrated control, which may be much less than the community expects. This means that continuous efforts are required to maintain a healthy level of decentralization.

Future Outlook

As the Ethereum ecosystem continues to evolve, its components and overall infrastructure will also change. This means that to ensure Ethereum's decentralization, the community needs to continuously monitor and assess its level of decentralization and take necessary measures to ensure its health and stability.

Additionally, with the emergence of new technologies and solutions, new measurement dimensions and methods may be added to the evaluation of decentralization. This requires researchers to continually update and refine their models to ensure their accuracy and relevance.

Overall, decentralization is not only a technical challenge but also a shared responsibility of the community and ecosystem. Only through continuous efforts and collaboration can we ensure the decentralization of Ethereum and its long-term success.