SoK: Measuring Blockchain Decentralization

Christina Ovezik, Dimitris Karakostas, Mary Milad, Aggelos Kiayias, Daniel W. Woods · 2025

Read SoK: Measuring Blockchain Decentralization by Christina Ovezik, Dimitris Karakostas, Mary Milad, Aggelos Kiayias, Daniel W. Woods (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2501-18279.txt.

Introduction

Christina Ovezik, Dimitris Karakostas, Mary Milad, Aggelos Kiayias, and Daniel Woods sort out how to measure decentralization in blockchains, in work from 2025. Everyone praises the trait, but no shared ruler exists. Studies pick their own windows, cutoffs, and math, then compare poorly.

The team frames every past method by what resource it counts, how the split is pulled from data, and what function turns the split into a score. They then test whether the choices change the answer. Small data choices, like window size, swing scores a lot. Factor tests show work-based chains split into two signals, not one. The paper urges care in what a score really names.

Catalog record

CREATORChristina Ovezik, Dimitris Karakostas, Mary Milad, Aggelos Kiayias, Daniel W. Woods
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE75 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2501-18279

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

SoK: Measuring Blockchain Decentralization Christina Ovezik, Dimitris Karakostas, Mary Milad, Aggelos Kiayias, Daniel W. Woods Abstract In the context of blockchain systems, the importance of decentralization is undermined by the lack of a widely accepted methodology to measure it. To address this gap, we set out a systematization effort targeting the decentralization measurement workflow. To facilitate our systematization, we put forth a framework that categorizes all measurement techniques used in previous work based on the resource they target, the methods they use to extract resource allocation, and the functions they apply to produce the final measurements. We complement this framework with an empirical analysis designed to evaluate whether the various pre-processing steps and metrics used in prior work capture the same underlying concept of decentralization. Our analysis brings about a number of novel insights and observations. First, the seemingly innocuous choices performed during data extraction, such as the size of estimation windows or the application of thresholds that affect the resource distribution, have important repercussions when calculating the level of decentralization. Second, exploratory factor analysis suggests that in Proof-of-Work (PoW) blockchains, participation on the consensus layer is not correlated with decentralization, but rather captures a distinct signal, unlike in Proof-of-Stake (PoS) systems, where the different metrics align under a single factor. These findings challenge the long-held assumption within the blockchain community that higher participation drives higher decentralization. Finally, we combine the results of our empirical analysis with first-principles reasoning to derive practical recommendations for researchers that set out to measure blockchain decentralization, and we further systematize the existing literature in line with these recommendations. 1 Introduction The security, reliability, fairness, and even legal status of blockchain systems is determined in part by their level of decentralization. In practice, decentralization concerns how critical resources, such as tokens or mining power, are distributed across participants [ 49 ] . A system is considered centralized if a small number of parties — or, in the extreme case, a single party — controls a large share of the resources that secure the blockchain protocol. Prospective users or developers of blockchain systems may want to monitor centralization in block production, as it increases the risk of attacks, like double spending or censorship. Similarly, centralization of token ownership can increase the potential for market manipulation. Regulators also evaluate decentralization levels to treat blockchain-based assets and classify them as securities or commodities [ 58 , 19 , 16 ] . Thus, decentralization metrics can be used by a wide range of participants across the blockchain ecosystem, to make quantitative assessments and comparisons of blockchain systems across different layers and over time. Despite the fundamental importance of decentralization in blockchain systems, there is no widely accepted method for how to measure it in the blockchain and distributed systems literature. 1 1 1 An illustratory example can be found in the proceedings of the SEC vs. Ripple case, where the SEC argued that the XRP token was centralized and hence a security, while Ripple argued the opposite. Ultimately, the lack of a widely accepted methodology for assessing decentralization led to Ripple being exonerated from most accusations [ 1 ] . (NB: this ruling is currently under appeal by the SEC). This gap in the literature motivated us to undertake a systematization of knowledge on the topic of measuring blockchain decentralization. To conduct our systematization, we put forward a framework for standardizing decentralization measurements. This involves addressing three critical questions: (1) What is the resource of interest, whose allocation one should measure? (2) How can its distribution be extracted from the available information? (3) Which functions should be applied to the distribution to produce the final decentralization measurement? Regarding question (1), the existing literature focuses predominantly on the distribution of protocol participating resources (e.g., hashing power or stake) or the distribution of tokens to different participants. In terms of question (2), there is a wide variety of approaches that demand careful systematization. One key issue is that the distribution of resources may not lend itself immediately to extraction. For instance, the amount of resources, such as computational power, held by consensus participants cannot be observed directly. Instead, block production is typically used as a proxy for it, but this approach requires caution — for instance, inferring hashing power from the number of produced blocks involves fixing an estimation window, the choice of which can significantly influence the outcome of the analysis. Other considerations include how to attribute blocks or tokens to entities and whether to prune the data set of token holders to approximate the most useful segment of the token ownership distribution, given the pseudonymity of addresses. Finally, regarding question (3), a plethora of functions have been put to the test in the literature. Given the overlapping nature of many of these functions, the interesting part from a systematization perspective is to identify what exactly they measure and to what extent their estimation truly captures decentralization. Contributions. Our first contribution is the development of a framework that describes the core workflow of blockchain decentralization measurements and facilitates our systematization. This includes: (1) choice of resource; (2) data pre-processing, and in particular choices on (2a) clustering, (2b) resource estimation, (2c) population estimation, (2d) measurement frequency, and (2e) inclusion thresholds; and (3) choice of decentralization metric. We then systematize the blockchain decentralization literature using the above framework, to showcase how prior work has approached the various steps of the measurement workflow. We observe that the literature lacks a unified methodology for addressing these concerns, with studies making a wide variety of choices, frequently without complete justification. We complement our systematization with an empirical analysis using data from live blockchain systems, demonstrating that different methodological choices taken in the literature have a significant impact on decentralization estimates. In particular, our analysis reveals that the results can be manipulated via arbitrary decisions on how to pre-process data, such as removing wallets with balance below a certain threshold, or aggregating block data over an arbitrary time period. We review all decentralization metrics from prior work and perform exploratory factor analysis to uncover the underlying constructs they capture. A key finding is that participation, namely the number of entities that actively command resources, and decentralization, namely how such entities share these resources, do not always align, with distinct factors emerging to represent the two in some cases. This challenges the common assumption that high participation inherently leads to high decentralization in blockchain systems. Based on our systematization and empirical analysis, we derive a range of recommendations for researchers who set out to measure decentralization in blockchain systems and organize the existing literature along these recommendations. Scope. Our systematization framework applies to fungible resources distributed across entities that participate in a blockchain system. The canonical applications are tokens distributed across users, and computational power or stake distributed across block producers. The framework can also extend to layer-2 blockchains or DeFi applications, such as those involving on-chain governance tokens. However, it is not suitable for blockchain layers involving non-fungible resources, such as the network layer, where the graph structure must be incorporated into any analysis of decentralization. We further note that our pragmatic approach differs from prior systematizations that have focused on conceptualizing decentralization [ 49 ,

Read the complete text