Ethereum fee market reform: EIP-1559 as a question of fairness
Parts 1 to 4 of this article deal with the background to EIP-1559, its development and progress to date. These sections include links to some of the key articles and may be a useful reference. However for busy readers already familiar with EIP-1559’s background, skip to parts 5 to 7 for a discussion of the fairness of the proposal. If ‘gas’, ‘gas limit’, and ‘gas price’ are unfamiliar concepts, there is some useful grounding in these topics to be found on EthHub.
Reform of Ethereum’s fee market, and the question of how best to allocate income from transaction fees, has been discussed for a long time. Well before Ethereum’s July 2015 genesis block, founder Vitalik Buterin was sharing ideas about future changes to the the fee market, including the possibility of burning fees. The idea of fee burning in the context of ‘medium of exchange’ currencies was also included in a 2017 piece on the subject.
These ideas certainly generated interest in the Ethereum community, but without a concrete proposal to work with, they were confined to the realm of academic inquiry. Moreover fee market reform was not a pressing concern — transactions on the Ethereum blockchain were cheap. Indeed, throughout much of Ethereum’s early history, gas prices were well below miners’ estimated ‘rational’ minimum. Occasional spikes in demand caused congestion during the ‘ICO mania’ of 2017, but on the whole users had no difficulty getting their transactions included with almost negligible fees.
By early 2018 however, this situation had changed. Whereas previously blocks had only been partly full, instead they were routinely pushing up against the gas limit. High gas demand from applications including CryptoKitties, and incentivised crypto exchange FCoin, led to prolonged periods of congestion and extreme volatility in the gas price required to get a transaction included within a reasonable time frame.
Making sense of Ethereum’s gas price mechanism had always been a challenge for new users, but in conditions of congestion and volatile gas prices, this became much more severe. New tools like ethgasstation.info helped savvy users to judge appropriate gas prices, and wallets such as Metamask increasingly built in gas price estimation tools. Nonetheless evidence from the blockchain shows that some users massively overpay for gas, and endless support tickets for stuck transactions attest to the fact that these solutions only went part way to solving the problem.
2. A Proposal Emerges
This was the context for Vitalik’s July 2018 article which set out in more detail a proposal for fee market reform (it was subsequently included in a wider-ranging paper on resource pricing). It described a mechanism for setting a portion of the transaction fee in-protocol called the ‘basefee’, with miners to be compensated for transaction processing with a user-set commission on top of this, now often called the ‘tip’ or ‘bribe’. The mechanism targets half-full blocks (where a full block is double the size of a current Ethereum block), and achieves this target by adjusting the basefee up and down in response to demand. Since the amount of compensation miners need for including transactions varies very little with demand, the tip would be a small amount that most users won’t need to change. Therefore, users can be presented with a simple transaction price at which they can expect their transaction to be included, rather than having to figure out how to interact in an unfamiliar market for an intangible consumable (gas). A major improvement.
The question then is what should the protocol do with the basefee? On the surface there appear to be 4 possibilities:
- send the basefee to some fund (e.g. a developer fund);
- give the basefee to the miner of the current block;
- share the basefee amongst the miners of future blocks;
- burn the basefee.
Of these option 1 can be discounted immediately, as the creation of a centralised fund would create a governance nightmare. Option 2 is also not viable, since it would incentivise miners to fill their blocks with junk transactions at no cost, forcing the basefee upwards whether or not demand for blockspace was high. The choice was between 3 and 4.
Ultimately, burning the basefee was chosen, in part because this would result in another important benefit — that Ethereum’s native currency Ether would be cemented as the fuel for transactions on the network. This would avert the threat of economic abstraction which some had feared would lead to a death-spiral for the currency, with dire consequences on the security and overall viability of the network. There are other reasons for preferring fee burn, which will be discussed in parts 5 to 7.
This time, with a concrete proposal on offer, strong interest from the Ethereum community followed quickly, and the idea was specified as Ethereum Improvement Proposal (EIP) 1559. Predictably, Ether holders were excited about the potential economic consequences of fee burn, which is equivalent to proportional payouts for all Ether holders, as in a stock buyback. This enthusiasm was demonstrated in the emergence of the crowdfunded EIP-1559 community fund, which received by far the most contributions in round 6 of the Gitcoin Grants programme.
3. Developer Scepticism
Nevertheless, if implemented, EIP-1559 would be the most dramatic change to Ethereum’s economics since genesis. It would also create significant work for wallet devs, who would need to adapt their user interfaces to the new paradigm. Accordingly, the proposal has been received with caution by experienced Ethereum developers. The most often quoted of these is Dan Finlay of Metamask, who notes that since EIP-1559 places on miners the task of fixing ‘basefee’ parameter which has the effect of diminishing their income, there appears to be a perverse incentive for miners to attempt to minimise the basefee. If miners were to succeed in doing this, EIP-1559 would effectively degenerate to the current first-price auction fee market, with users varying the tip parameter to bid for inclusion of their transactions.
Dan responded with his own ‘Escalator’ proposal which also confronts the UX problem, enabling efficient price discovery without ceding control over fee parameters to miners. This solution is also elegant, but has so far not received the same level of backing as EIP-1559. This may be in part because it doesn’t deal with the other problems which EIP-1559 addresses, such as economic abstraction. And of course, the Escalator proposal has not been received as warmly by much of the Ethereum community, who tend to hold a position in Ether, and are particularly attracted to the prospect of ETH burn. As a side note, a hybrid of both EIP-1559 and Escalator combining the advantages of both is one possible outcome, but has yet to be specified as an EIP.
Other long-standing Ethereum developers, such as Nick Johnson from Ethereum Name Service, have expressed concern about the amount of momentum which has grown behind EIP-1559 before the impact and possible risks of fundamentally altering the game theoretic basis for Ethereum’s fee market are fully known. Clearly, it would be foolish to implement such a sweeping change to Ethereum’s core economics until the implications are very well understood.
4. Renewed Impetus
During 2020, demand for gas on Ethereum has continued to grow, as stablecoins, decentralised exchanges and a range of other decentralised finance (defi) applications have exploded in popularity. The result is that from June until August this year the total value of fees paid for Ethereum transactions has been greater than total Bitcoin fees on all except 8 days. Counter-intuitively, such strong fee income directed at miners may in fact cause problems for the blockchain. One group of researchers has argued that as miner income from fees overtakes the income from the block subsidy, the blockchain becomes unstable, and security is dramatically impacted as certain attack strategies become viable with much less that 51% of hashpower. Meanwhile, the extreme volatility of the gas market and corresponding UX problems remain unaddressed.
By directing excess fee income away from miners, and maintaining the primacy of the block subsidy in rewarding them, EIP-1559 would sidestep this problem too. Accordingly, work on evaluating and implementing EIP-1559 for potential inclusion in a future Ethereum hard fork continues apace. This includes modelling and simulation work carried out by Barnabé Monnot of the Ethereum Foundation’s Robust Incentives Group, as well as the recently announced analysis to be conducted by algorithmic game theorist Tim Roughgarden, which aim to perform a rigorous evaluation of the proposal’s dynamics.
Meanwhile, work on client implementations of the EIP is also progressing. As of late August 2020, two Ethereum clients have implementations of EIP-1559, and are running an early testnet to iron out specification and implementation issues. Outside of the core Ethereum client space, implementations are also being tested by the Filecoin and Celo networks.
With significant resources now focused on EIP-1559, no viable attack vectors have yet been identified. Attack scenarios involving collusion from a majority of hashpower can be imagined, but, as noted in an article by Micah Zoltu, even if a 51% cartel were to form and attempt to manipulate the basefee, the strong profit incentive on miners would be to defect from the cartel. This contrasts with the well-known and much more profitable 51% block censorship attack which is already possible in Proof of Work blockchains, and allows cartel members to double their income by ignoring blocks from non-members.
As research into EIP-1559 continues, we might expect that any catastrophic defects in its economics will be identified. In the absence of such defects, perhaps cautious optimism that it can be implemented is warranted.
5. Who owns Ethereum’s blockspace?
So we might tentatively conclude that we probably can safely implement EIP-1559 in Ethereum. But should we? This is the question of fairness raised in this article’s title. Do miners have any right to continue to receive all income from transaction fees since they do the work of processing transactions?
Ethereum’s 2013 whitepaper pointed out that this is not the case. In contrast with Bitcoin, miners in the Ethereum network are paid a block subsidy in perpetuity for the task of securing the network — this is currently fixed at 2 ETH per block, plus additional income from uncle blocks (i.e. blocks not included in the canonical chain, but which receive a partial block reward). Accordingly there is no requirement that income from transaction fees should ever meet the costs of securing the network. This is a strength, since transaction fee income is highly variable, and there can be no guarantee that it will be adequate to meet the network’s security needs.
The other reason miners don’t “do all the work of processing transactions” is that transactions impose a cost on the whole network, in the form of state bloat. While all full nodes bear this cost, only miners are paid for it. Users also pay an opportunity cost when gas prices are so high that they can’t execute transactions which would be profitable to them. So rather than being a resource that is in any sense ‘owned’ by miners, Ethereum’s transaction capacity is better viewed as a commons. Miners who extract full market value from this shared resource through high transaction fees are really extracting economic rent, since the cost to the miner of including the transaction is far less than the payment they will receive for it.
6. Fair Compensation
To understand which part of the transaction fee is fair compensation for the miner’s task, and which part is unearned rent, we need to know how much transaction processing actually costs miners. As mentioned in part 1, there is a “rational minimum” transaction fee, which is the marginal cost to miners of including a transaction. This is the opportunity cost arising from the increased risk of creating a less profitable uncle block. This concept is an extension of a methodology originally applied to Bitcoin. (1)(2)
Arriving at an accurate figure for the miner cost of including transactions is difficult, and this figure changes with improvements in client implementations and changes in amount of gas required to carry out certain operations on the Ethereum Virtual Machine (EVM). The figures from Vitalik’s work on the subject referenced above are therefore no longer up to date, and obtaining data on which to base new estimates is tricky. Nonetheless, improvements in client performance mean that miner costs are likely cheaper than they were in the past whereas the fees they are paid have dramatically increased.
Fortunately, the new fee market in EIP-1559 can work this out for us. Rational miners will accept only those transactions for which the tip/bribe is sufficient compensation. Users who have a very low “time preference” for having their transaction included may try a lower tip, and those desperate to be included quickly for time-sensitive arbitrage transactions might increase it. But in general the tip is not expected to be as volatile as gas prices themselves. Most users should be able to accept wallet defaults to pay the market rate, rather than having to worry about the gas price they pay.
Anything above the fair miner compensation may be thought of as a ‘fee surplus’ or perhaps ‘transaction dividend’. This is EIP-1559’s basefee. If it doesn’t belong to miners, where should it go?
7. Allocating the Fee Surplus
The only option remaining of the 4 possibilities mentioned in part 2, is for the basefee to be burned. By burning the basefee, it is returned to Ether holders. Are they worthy recipients? For two reasons, I think the answer is yes.
Firstly, Ether holders pay for the Ethereum’s security through issuance, which amounts to a gradual dilution of the currency. Therefore, any surplus generated by the operation of the network might reasonably be returned to them as a rebate on this payment. It may be that the fee surplus is at times equal to or even greater than Ethereum’s issuance, but in the end it is Ether holders who are on the hook for funding security, whether or not fees are sufficient for this.
Secondly, Ether is not the only asset that benefits from the security of the network. A multitude of assets are secured by Ethereum — stablecoins, synthetic assets, prediction market outcomes, liquidity shares, equity tokens and a host of ICO tokens — all piggyback on the security funded by Ether issuance without having to pay this cost themselves. Ether is thus uniquely disadvantaged by the current fee market. EIP-1559 helps to fix this problem, and by minimising the net issuance of the token, perhaps eventually even reducing it below zero, maximises the value of the currency which ultimately secures the network.