Measuring the Concentration of Control in Ethereum

TL;DR

• This presentation speaks to how we can measure the effective level of decentralization in Ethereum.
• There have been many studies on measuring decentralization in cryptocurrencies, which is a complex topic.
• Ethereum has changed significantly in the past 12 months, and will continue to change significantly going forward.
• We need a way to measure decentralization in a continually evolving and dynamic ecosystem.

Motivation

• Regulatory��Regulators are looking to try to define and quantify the level of decentralization in cryptocurrencies and public blockchains.

• Credibly Neutrality��Introspectively measuring the effective level of decentralization is something we should be proactively doing ourselves as a space if we want to maintain credible neutrality

Unpicking Decentralization

And what does decentralization even mean?!

Unpicking the term “Decentralization”....

2019

Discourse that is Informing Regulatory Opinion….

​

• Angela Walch is one of the most cited experts in the intersection of law, cryptocurrency / blockchain, and financial stability.
• This is what she later said about her testimony to the U.S. Senate Banking, Housing, & Urban Affairs Committee…

“I further testified that mainstream beliefs about basic characteristics of crypto systems are flawed because they are based on idealized views of these systems rather than realistic ones. (To give this a bit more color, if you have heard that crypto systems are immutable, allow the direct transfer of value online without intermediaries and don’t require you to trust anyone because they are decentralized, then you need to look for more realistic sources of information.)”

2021

Regulatory Discussion Paper - Banque de France…

“Decentralised” or “disintermediated” finance: what regulatory response? - Fliche, Uri, Vileyn, April 2023

2023

Are we Decentralized?

And how do we measure it?

Main Points

1. Decentralization is an overloaded term! For our purposes let's talk about decentralization as a resistance to control by a small number of powerful actors or intermediaries.
2. The overall effective level of decentralization in the ecosystem can only be measured by considering all vectors of centralization, including those that are exogenous to the core protocol.
3. The overall level of decentralization in an ecosystem is not static, it’s dynamic, and it changes over time. This is a feature, not a bug!

How do we measure effective decentralization?

1. Decide what the dimensions of measurement are.
2. Decide what indices to apply and how to interpret them.
3. Rinse . . . repeat . . .

Any model needs to adapt to disruptive innovation e.g.: zkEVMs, ePBS, OFAs, 4337, SUAVE, DVT, EigenLayer, and all the other stuff that’s slowly cooking in the ecosystem!

Criteria for Selecting our Dimensions of Measurement

Infrastructure components that are considered for inclusion as a dimension of measurement based on two criteria:

1. The size of the infrastructure compared to the base layer, as measured in either percentage of base layer transactions that flow through the infrastructure and/or the Total Value Locked compared to the base layer.

• The potential effect on the base layer should the infrastructure be compromised or develop misaligned incentives, whether this effect is a level of effective degraded performance of the network, or an increased level of censorship.

Initial Subsystem Selection

• Consensus nodes by country
• 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

​

• Blocks proposed by builder
• Blocks proposed by relay

​

• Number of user operations per bundler
• Number of transactions per bundler

​

• Effective inflation rate adjusted for burn
• Percentage of total supply staked
• Layer 2 rollups by relative TVL
• Stablecoins by relative TVL

​

Using the original Nakamoto Coefficient’s subsystem selection as a base

Accounting for PBS / mev-boost

Accounting for ERC-4337

Other important miscellaneous metrics

Now that we’ve decide what to measure�How do we measure it?

Centralization is usually easy to spot

But how do we describe it succinctly, and communicate changes in centralization over time?

“A bit centralized”

“A bit less centralized”

“(sort of)”

Clearly we need some sort of summary statistics!

Gini Coefficient

The Gini coefficient is based on the comparison of cumulative proportions of the population against cumulative proportions of income they receive, and it ranges between 0 in the case of perfect equality and 1 in the case of perfect inequality.��It’s a very widely used well understood index, making it ideal for communicating decentralization . . .

The Gini index is not enough by itself

The following are two unequal populations of the same size, both have a Gini index of ~0.5

The Gini index is not enough by itself

The following are two unequal populations of the same size, both have a Gini index of ~0.5

The Herfindahl–Hirschman Index

The Herfindahl–Hirschman Index, HHI, or HHI-score, is a measure of the size of companies in relation to the industry sector they are in and is an indicator of the amount of competition in that sector.

​

It is widely used in competition law, antitrust regulation, including the US DoJ and FTC. These agencies generally classify markets into three categories:

The HHI is not enough

HHI = 2,714

HHI = 2,726

Both of these populations have the same HHI, despite one having a majority of members with nothing

Measuring (de)centralization through Entropy

• The Shannon index (or Shannon’s diversity index) is a popular measurement index in ecology, commonly used to measure the diversity of species in a community.
• It was originally developed by Claude Shannon in 1948 to measure entropy in strings of text.
• Various forms of measuring entropy measurement have found applications in economics in terms of measuring inequality, and the Shannon index is not the only entropy-based measurement available.

The Shannon Index by itself is not enough

• Like other measurements, it is possible to arrive at the same Shannon index value for different distributions.
• The Shannon index is influenced by the number of categories in the distribution. Adding or removing categories can impact the entropy value, even if the underlying inequality remains unchanged.
• The Shannon index is not as widely used, or as easily interpreted as other indices, but . . . can be a useful cross-reference when interpreting the Gini index, or HHI.

​

Results?

Average Index Values Over 30 Day Period

other

PBS

Nakamoto Coefficient

not established enough

Average Index Values Over 30 Day Period

other

PBS

Nakamoto Coefficient

not established enough

Geographical distribution is quite concentrated

Client diversity is too concentrated, especial exec nodes

This is where the red flags are!

Stablecoins are a red flag

• Over 80% of all stablecoins are controlled by private companies
• These companies have on a number of occasions seized funds at the request of authorities
• They are used to interact with basically every DeFi protocol we have

HHI of Block Builders and Relays

• Interestingly, the Block Builder and Relay market is only “moderately concentrated”

according to the definitions used by the US DoJ and FTC.

​

• As you can see, it also changes quite a lot, even within a 30 day period.

Project Sunshine!

https://ethsunshine.com/

Future Work and Open Questions

• What are the correct dimensions of measurement to include in our model?
• How we do we correctly adapt our model over time?
• What are the correct indices to incorporate and how do we interpret them?
• What is the best way to communicate the results?

​

Please get in contact if you’d like to discuss this research:

@orbmis

Beware the Decentralization Paradox…

Decentralization: An Incomplete Ambition - Nathan Schneider, 2019