SoK: Measuring Blockchain Decentralization
Christina Ovezik, Dimitris Karakostas, Mary Milad, Aggelos Kiayias, Daniel W. Woods
TL;DR
This SoK addresses the lack of a standardized methodology for measuring blockchain decentralization and proposes a framework that classifies measurement work by resource (layer), preprocessing (clustering, estimation windows, population, frequency, thresholds), and final metrics. Through an empirical analysis of five ledgers across PoW and PoS systems, the paper shows that pre-processing choices such as window size and clustering can substantially alter decentralization estimates, and that participation does not always map to decentralization in PoW but tends to align in PoS. The authors find that no single metric consistently captures decentralization, with PoS and PoW exhibiting different factor structures, and they advocate for principled metric selection driven by system architecture. The work provides concrete recommendations for researchers (e.g., use blocks for consensus, tagging, ≥7-day windows, avoid arbitrary thresholds) and calls for standardization, improved data quality, and future research into token clustering, population estimation, and governance/network layers to enable robust, cross-study comparisons and governance insights.
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.