Table of Links
A. Decentralization and Policymaking
C. Brief Evaluations per Layer
E. Fault Tolerance and Decentralization
11 Case Studies
We now apply our methodology to a number of case studies. First, we review Bitcoin’s status w.r.t. each identified layer, showcasing how a project can be analyzed across all strata. Second, we apply the Minimum Decentralization Test (MDT) (cf. Definition 1) on an array of projects and show that they fall short due to centralization some layer — specifically the governance layer.
Cross-layer Study: Bitcoin
Hardware. No concrete data could be found on the distribution of hashing power across PoW mining products. Although hundreds of ASICs are available, on top of generic hardware (e.g., GPUs), as of 2022 the market appears centralized around 4 ASIC manufacturers [185]. Interestingly, only some ASICs are profitable, so, unless the token’s price increases without an increase in PoW difficulty, mining should be expected to concentrate around these products.[17]
Software. An overwhelming majority (99.15%) of full nodes run Bitcoin Core,[18] so Bitcoin is completely centralized around (releases of) this product. Regarding the distribution of tokens across wallets, no data could be found.
Network. Regarding peer discovery, Bitcoin Core sets 8 outgoing and 125 incoming connections, chosen randomly from known and/or hardcoded peers. Most Bitcoin nodes communicate over Tor, making topology analyses particularly hard.[19] Nonetheless, it is estimated that the network is evenly spread across multiple Autonomous Systems, thus presenting high levels of decentralization [4].
Consensus. On the consensus layer, Bitcoin presents mixed results regarding decentralization (cf. Section 6). Hashing power is distributed across thousands of machines. Although no concrete data could be found, folklore evidence suggests that these machines are owned by a highly diverse set of users. However, Bitcoin also observes high levels of centralization around pools, i.e., w.r.t. block formation and the input to the PoW module; specifically, at the time of writing, 4 pools control more than 75% of the whole network’s mining power.[20]
Tokenomics. At its onset, no Bitcoin tokens existed. They were generated and allocated as the system progressed. Early participants were disproportionately favored, as half of all tokens were created within the first two years, when consensus participation was sparse and mining was conducted by only a few parties. As more transactions were issued, the tokens were distributed more widely, albeit wealth is still highly centralized, compared to real-world economies (cf. Table 2). Specifically, approx. 43M addresses own some amount of tokens, with the top 100 addresses controlling 14.01% of all wealth. Nonetheless, tokens are traded on more than 100 marketplaces at volumes of approx. $53B (cf. Table 3).
Client API. Most of the available Bitcoin wallet software is either SPV or explorer-based [100]. In the first case, the wallet downloads only the block headers, so it does not validate each block’s transactions, while in the second case the wallet relies entirely on a server. However, no data could be found on the ownership of Bitcoin tokens w.r.t. wallet types, therefore Bitcoin’s decentralization w.r.t. the client API layer is inconclusive.
Governance. Deciding on improvement proposals and conflict resolution in Bitcoin is somewhat centralized, but not entirely. Specifically, decisions, which are made by accepting suggestions via GitHub, are typically taken by a small set of developers, who are often the ones to comment during the relevant discussions [83,11]. In terms of development funding, Bitcoin makes no provisions. Therefore, the available data are inconclusive on how many sources of funding exist, e.g., companies and foundations, and how much influence each has.
MDT Studies. We now turn our attention to projects that fail the Minimum Decentralization Test (MDT), showcasing how the MDT can be used to identify points of centralization in blockchain systems.
Fiat-backed stablecoins. A prime example of projects that fail the MDT is fiatbacked stablecoins. Briefly, in a USD-backed stablecoin system, for each token that is live on the ledger there exists $1 which is held in escrow in a company’s bank account, s.t. at any point in time, a token holder can exchange their tokens for the equivalent number of USD. Therefore, the main selling point is that the token should always (in theory) be valued by the market at $1. Such projects include Tether (USDT), USDC, Gemini Dollar (GUSD), TrueUSD (TUSD), and Binance USD (BUSD).[21] In all these systems, there exists a single legal entity which is responsible for issuing tokens when receiving USD and redeeming tokens in exchange for the USD held in escrow. Therefore, these entities are single points of control within the governance layer of each system.
Wrapped tokens. Another family of systems for which the MDT often fails is bridges. A bridge enables transferring assets of one ledger to another, e.g., Bitcoin to Ethereum. This is achieved by creating a “wrapped” version of the original token on the destination, with each wrapped token corresponding to a (frozen) token on the source side. These systems can also be centralized in the governance layer, as a single custodian is typically responsible for the creation and destruction of the wrapped tokens. Such bridge examples include Wrapped Bitcoin (WBTC), with BitGo being solely responsible for minting new tokens, and Huobi Bitcoin (HBTC), with Huobi being the custodian.[22]
[17] A profitability calculator is available at nicehash.com.
[18] Source: blockchair.com
[19] 53.3% of Bitcoin’s nodes operate over Tor. [bitnodes; October 2022]
[20] Source: statista.com
[21] USDT: tether.to, USDC: circle.com, GUSD: gemini.com, TUSD: tusd.io, BUSD: binance.com
[22] WBTC: wbtc.network, HBTC: htokens.finance
