Which is More Inclusive: Proof-of-Work or Proof-of-Stake?
Choosing a consensus algorithm is more than a technical design choice: it’s about determining who to involve, how to involve them, and where to delegate power.
Celo’s mission is to build a monetary system that creates the conditions for prosperity for all. We are working to create technology that can deliver significant improvements to financial inclusion, the opportunity for everyone to access suitable, affordable, and easy to use financial products and services.
We see a significant part of this opportunity arising from the increasing ubiquity of mobile phones and rapidly improving internet penetration. However for a financial system to deliver equity (i.e. be fair and impartial), we believe it has to be a public good, developed and operated by the community it serves. For us, this is where blockchain technology comes in. Blockchains resemble databases that secure system-wide agreement (consensus) using compute resources provided by individuals and organizations without a prior trust relationship, who often are rewarded for doing so, and who often also participate in their governance. But how do you design the technology itself to be inclusive?
As we have begun the development of the Celo platform, we’ve gathered some insights and thoughts on the two main approaches to consensus, proof-of-work (POW) and proof-of-stake (POS), and how the incentive structures implied by these algorithms affect participation in the protocols themselves and the services built on top of them.
We think about this in three main ways:
- Access — Give the ability to benefit from participating to as many as possible;
- Distribution — Equally (re)distribute tokens without favoritism or discrimination;
- Power — Divide governance power such that no single entity or small group have control over the platform.
Who has access to the network and its native assets?
Running and securing the Celo platform and getting rewarded for doing so, should be accessible to as many as possible.
Three types of network participants
Throughout our research we identified three types of network participants:
The Grassroots: They’re hobbyists motivated by access to passive income (with potential financial upside) and/or by a desire to learn and support projects they believe in (e.g. ideological alignment, strong team, interesting technology, etc.).
The Professionals: They represent a new type of company or fund designed specifically to extract value from these new cryptographic networks. They’re driven by financial returns and choose what to mine based on profitability or what to stake based on yield.
The Disenfranchised Citizens: They’re individuals who don’t have access to sound currency and are motivated by access to a trustworthy censorship-resistant store of wealth. They’re based in countries that have a poor monetary policy (e.g. high inflation) and in the case of mining, where electricity is affordable. They choose a token based on its brand/reputation and its liquidity.
The higher the market cap, the more professionalized a network is
No surprises, that’s how Satoshi Nakamoto designed Bitcoin: the more participants, the harder it is to participate. However, the subtlety here is not about participation difficulty but about the consequences of it.
As a network increases in value, it’ll attract more players, which increases the barrier to entry causing that ecosystem to become professionalized. The reverse also applies: as a network loses value it’ll attract fewer participants, which will lower the barrier to entry and attract hobbyists. This is true to a certain extent since networks with mature consensus algorithms will have a barrier to entry capped by the cost of specialized hardware (i.e. ASICs).
For example, Ethereum (POW) started with a low market cap and as its value and hashing power increased, the network professionalized, squeezing out the grassroot miners. Tezos (POS) on the other hand, launched with high market cap requiring self-stakers to invest at least ~$24,000 (price for 10,000 XTZ) and requiring anyone with less than that to delegate their stakes to professional services. However, as the network lost value the minimum required to self-stake fell to ~$3,500 opening up opportunities for grassroot stakers.
How are the network’s native assets distributed?
The minted assets and transaction fees should be equally distributed without favoritism or discrimination.
The “permissionless” debate
A fair consensus mechanism should allow anyone to join and participate in a network. When Bitcoin launched crypto-exchanges and other blockchain-based networks didn’t exist, which meant POW was the only way to distribute the cryptocurrency. Anyone could turn the cost of a computer and some electricity into Bitcoin. POS on the other hand, requires finding a way to distribute the staking tokens prior to or at network launch. Some argue that this represents a permissioned form of distribution since “gatekeepers” decide who can purchase pre-sale assets, receive airdrops, or get listed on exchanges, whilst others argue that the industry is now mature enough to consider exchange listings sufficient to be permissionless.
Fair distribution boils down to resource availability
POW requires access to cheap (or free) electricity and specialized hardware whilst POS requires access to token sales, exchanges (which in general also requires identity papers), or agents. In the case of POW, there are few places in the world where electricity is subsidized or cheap, which limits where people can mine from. In addition, with the increase of specialized mining hardware, few have access to the most competitive machines.
On the other hand, POS token sales are regulated to accredited investors, exchanges are still poorly represented in frontier markets, and agents only surface when there’s demand. Token distribution therefore depends on the types of resources a user has access too and determining what’s “fair” is, to a certain extent, subjective.
Confirmation bias is prevalent
It’s also interesting to observe that when asking people which is “better” POW or POS? They tend to support the consensus mechanism that’s most beneficial for them personally and that’s generally correlated with the resources they have access to. People’s perception of “fair,” if not driven by ideology, is biased by their experience and environment.
Who governs and has decision power over the network?
Celo governance should be decentralized such that no single entity or set of entities can control the network.
In the end, it’s all about power
The POW vs POS debate is ultimately about who gets controlling power. Some argue that separating power between POW miners and token purchasers creates a healthy tension. No one can move forward without the support of the other, which generates constructive debates and well thought through changes. Others contend that this leads to misaligned incentives and unnecessary debates, which cripples a network and stifles innovation. Better align incentives via a POS model where token purchasers are also those running the network. The risk, however, is that the POS incentive model hasn’t yet been proven. For example, these networks may be subject to insufficient checks and balances, collusion, and bribery. In addition, many argue POS enables “the rich to get richer”, closing out others and accentuating inequality.
In conclusion, staking is more accessible than mining for the majority of users because it doesn’t (yet) require access to specialized hardware or cheap electricity. POS however, faces token distribution challenges. Unlike POW, where asset distribution is automated, decentralized, and permissionless, POS token distribution involves at some point human decisions and is therefore permissioned. As for controlling power, the debate is more nuanced than painted here, having a variety of stakeholders (POW) seems more inclusive than aligning them into one role (POS) but with the caveat that in such situations stakeholder power may be unbalanced. In addition, it could be that, compared to POS, existing POW networks are missing a governance process skewing perceptions and assessments. Ultimately there’s no perfect solution. Choosing a consensus algorithm is more than a technical design choice, it’s about determining who to involve, how to involve them, and where to delegate power.
We’re still working through these questions ourselves. What’s your take on it? Which do you think is more inclusive? We’d love to hear your thoughts.