EIP-1559 consists of three basic mechanism:
- Set a target gas usage, which is 50% of block gas maximum usage
- When sending transactions, users will be charged a fee, as known as ‘base fee’, which will be burned after execution. The ‘base fee’, as a price, will automatically changed based on utilization rate of older blocks，which means, if gas actual usage greater than target usage, base fee (price) will increase; if gas actual usage smaller than target usage, base fee will decrease
- When sending transactions, users need to pay ‘tips’ to incentive miners to include their transactions into blocks.
In the rest of this post, I will always use ‘base fee’ and ‘tip’ as a price. So
base fee * gas used = fee burnedandtip * gas used = miner's reward
I. Introduction
”EIP-1559“ is a proposal born in March. 2019. Now it is popular as it introduces a transaction fee burning mechanism, which was highlighted by David Hoffman as ‘the final puzzle piece to ETH’s momentary police’, even as a key to have ETH ‘sustain a monetary premium’.
One and a half year ago, after reading explainer from Eric Conner, the co-writer of this proposal, I wrote a sternly worded post to prove that this proposal cannot ‘fix’ those issues, such as unpredictable gas price, expensive transaction cost. It cannot provide a better user experience, rather, it will introduce more friction in transaction processing.
Nowadays, I find that some thinkers are supporting EIP-1559. They argue that it can not only provide better user experience, but also introduce other benefits to Ethereum, e.g. better security, resistance to economic abstract, etc. They are:
- Analysis of EIP-1559 (by Hasu, Georgios Konstantopoulos) (marked as “#1”)
- EIP-1559 51% Attacks: Should you live in fear (by Micah Zoltu) (marked as “#2”)
- Ethereum fee market reform: EIP-1559 as a question of fairness (by Pintail) (marked as “#3”)
I really respect those who contributes intellectual resource into this topic. But I am going to say that these conclusions are kind of rough and lack of evidences.
I will begin from some easy economics reasoning, and then discuss some issues in these posts.
II. A similar example
(If you have read my previous post on this topic, or agree that EIP-1559 can not provide better user experience, you can skip this section.)
Supposed that there are some expensive goods. Do you believe that depriving providers of revenue getting from this kind of goods would make the cost of getting this kind of goods cheaper?
Apparently, yes. But actually, no.
Because market price is determined by both supply and demand. It seems that price regulation can help consumers to get goods and services at a cheaper price. But actually it will decrease providers’ incentive, and will also obstruct information transfer function of price (profit rate), which decrease potential provider’s willing to join in this industry. Finally, it will decrease supply amount of relevant good, or at least limit the increase of supply amount. When supply amount is more limited, consumer will have to pay more for it (note that obvious money cost may not be total cost).
But EIP-1559 is actually supposed that depriving miners of revenue getting from Gas can make gas cheaper. How is it possible if miner do not will to provide more Gas?
Let’s make it intuitive. Given that you are hurt by expensive hospital fees (which is direct revenue of doctors). Which one can help you gain health services with better cost performance (the same cost, better services; or, the same services, lower cost)?
A. Lobbying the government to regulate hospital fees, by rating doctors and setting price ceiling for different classes of doctors
B. Accusing doctors in Twitter or Facebook platform, judging that all of doctors are immoral, that they don’t care whether patients lives or dies
C. Lobbying the government to tax on hospital fee, or increase tax rate on hospital fees
D. On the one hand, the government tax on hospital fee, on the other hand, the government promises that they will never spend this tax income, it will go to an observable account
E. The government provide subsidy for doctors, and tax on hospital fee, and never spend the tax income
Which one will works? No one.
As option A, Price regulation on hospital fee (of course, regulated price should be lower than before), will increase demand amount and decrease supply amount. The symbol is waiting in line, which means, patients not only need to pay regulated hospital fee to doctors, but also need to cost time in waiting, as too much patients want health services at lower price. The lower regulated hospital fee than unregulated level, the longer the line than normal. Other words, high quality health services are DoS attacked.
As option B, it is equal to put a negative price on doctor’s income. It will influence students’ willing to become doctors, which decrease supply in the long term. Hope no one chooses this option.
As option C, doctors will get less per consultation, so service quality will be worse.
As option D, you may realize that it is actually EIP-1559. It is policy C (thus tax effect) plus monetary deflation. As someone correctly argued, a reduction in the amount of the money will increase the money’s value. But, if burning a part of property belong to them can make themselves better, why can’t we see that people burning their money regularly? In this “tax + deflation” combination, doctors’ hospital fee become lower in terms of money unit, but the value of money unit increase, so, even we do not dig into it, we should say ’we cannot determine whether doctors’ income will increase or not‘. And, if this policy combination can benefit providers, why can’t we see that industries being charged consumption tax go to lobby the government to burn the tax income?
As option E, someone may say it is TRUE EIP-1559, because miners have both Gas fee income and block rewards, just like doctors have subsidy. But it is missing the point. Change the proportion of different income will also change relevant providers’ incentive. If the subsidy become the main part of doctors’ income, they would be happy to go to the office, but not providing services, just writing reports to send you to another doctor. Similarly, miners can submit empty block to the network.
(Under this kind of policies, Doctors choose these strategy is not because of they are unethical, but they need to take risks when working.)
Logically, the key to associate EIP-1559 with the above example is a concept that EIP-1559 is a consumption tax at first. Why?
One of the main mechanism of EIP-1559 is base fee, which will make a difference between how much users pay and how much miners receive. Assume that User pay 10 Gwei per gas, there is just 5 Gwei per gas going to miners’ package. Where does the 5 Gwei (base fee) go? It doesn’t matter! No matter where it goes, base fee mechanism is a consumption tax!
Once you realize that it is a consumption tax with a specific amount, you can associate it with two basic concepts in tax economics: (1) consumption tax not only extracts interests from providers, but also extracts interests from consumers. Look at this diagram. Consumption tax will eat both consumers’ gain from exchange (called ‘consumer surplus’) and providers’ gain (called ‘provider surplus’); (2) consumption tax must bring ‘deadweight loss’, because at those quantity where sum of consumer surplus and provider surplus is smaller than tax amount, exchange just doesn’t happen. That is also why consumption (G2) will be less than G1, which is a free market level. Without tax policy, providers and consumers can get more welfare from more exchange.
Here is a simple explainer for those readers who are not familiar with economics reasoning. If you can understand last paragraph, just skip it.
In the above diagram, where blue line is demand curve that D1(G2) represents highest price for purchasing extra one gas when consumers already have G2 gas and orange line is supply curve that S1(G2) represents lowest price for providing extra one gas when they already provide G2 gas. If there is a free market (even without block size limit), gas usage will be G1, which is a win-win result, and gas price will be P1. With every one gas used, consumers can get some surplus (called ‘consumer surplus’) because market price for every one gas is lower than the highest price for which they would like to pay. Geometrically, consumer surplus is an area below demand curve but above price flat line. Likely, provider surplus is an area below price flat line but above supply curve. Under tax policy, consumption will be the level that consumers’ bid price minus tax amount equals providers’ income. Tax amount multiplies by consumption is the government’s Tax income, is how much the government extracts from providers and consumers. Beside this, consumers and providers also loss something which doesn’t turn into the government’s income, because those exchange don’t happen, and the government can not charge it. That’s ‘deadweight loss’.
I am wondering whether those EIP-1559 supporters realize that it is a tax policy on gas or not. At least, in some cases, consumer will have to pay higher fee (base fee + tip) than in current fee market. Although there is a case that consumer can consume more gas than in current fee market, but it sacrifice security. (In next section, I will provide a more detail model to analysis how EIP-1559 impact consumption and security.)
As predictability of gas fee, EIP-1559 can not help, either. In current fee market, transaction senders just need to pay gas price. But in EIP-1559 fee mechanism, transaction senders need to pay base fee and tip. The algorithm just makes base fee predictable, doesn’t make tip predictable. What makes gas price unpredictable in current market is continuously changing demand and gas bid war between consumers. It remains in EIP-1559, where change in demand is still unpredictable, and consumers still need to incentive miners to include their transaction. Only one thing changed is the name. Yesterday it is ‘gas price’, and tomorrow it is ‘tip’.
(Gas bid war is not caused by miners, but competition between users. Users must find a way to incentive miners as they have transaction order right, which is the core function of PoW in distributed systems. Current fee market is the best way in terms of ETH’s moneyness and to serve users.)
I hope this reasoning can help you understand that EIP-1559 can not provide better user experience.
III. A more complete analysis
In this section, I will response to #2 and #3 at first, and do my best to explore an argument of #1: “EIP-1559 can bring better security”.
(I). #2 and ‘EIP-1559 is unmanipulable’
I agree this conclusion that EIP-1559 is unmanipulable, that miners can not cheat the mechanisms to make base fee stay at zero. But, it can not be a good reason to support EIP-1559, as current fee market is also unmanipulable, in my opinion. Although Micah says that current PoW and fee market may suffer from 51% attack and selfish mining, we didn’t see that any 51% attack is for grabbing transaction fees.
I believe that competition between miners limits their capacity to cheat EIP-1559, but it also limits their capacity to manipulate current fee market. So it is not an improvement.
(II). #3 and ‘commons’
In the ‘Who owns Ethereum’s blockspace’ section, Pintail says that ‘miners in the Ethereum network are paid a block subsidy in perpetuity for the task of securing the network’, and that miners are not the only one group bearing the cost caused by transactions, so blockspace is not a resourse ‘that is in any sense “owned” by miners’. (It implies that miners don’t have right to claim transaction fees.)
Finally, Pintail concludes that Etherum’s blockspace is ‘commons’, and miners are extracting ‘economic rent’ from this commons.
To be honest, it really shocks me. It seems that Pintail misunderstand what does PoW means. In Etherum-like systems (permissionless distributed system), the core function of PoW is to order the transactions. It is why we need PoW. PoW is not wasting energy to produce useless hashes, but generating an ordered ledger and making it hard to break. This function (ordering transaction) can not be split from ‘securing the network’, otherwise (no matter banning it in protocol level or eliminating miners’ incentive to do it) the system will become a centralized ordering system in terms of transaction processing , or issued money will be nothing as it cannot transfer.
On the other hand, even though we agree with that Ethereum’s blockspace is a commons, we should know that the first principle of commons governance is setting property right, which means making rules to determine who have right to utilize the commons and who can share the profit. In current fee market, the competitive advantage of collecting and processing transaction determines which mining pools can utilize blockspace to a certain extent and get how much profit. If you prohibit them from extracting rent from this commons, the commons will go out of cultivation. Everyone loses.
P.S. In economics, every one who is pursuing supply surplus is, by definition, rent-seeker.
(III). #1 and “better security”
In my opinion, ‘EIP-1559 can bring about better security’ is the most complicated and ultimate argument from EIP-1559-supporters.
Here is what they say: when the transaction fees become the main part of miners’ income, miners’ income will be volatile, so miners’ security input will be volatile, leaved Ethereum unsecure; EIP-1599 will both reduce miners dependency on transaction fees and increase value of block rewards, ultimately maintain the stability of security input.
However, I believe that it will break Etherum’s security, as in most cases, it will reduce miners’ income.
I want to provide a formal analysis to compare miners’ income in current mechanism with that in EIP-1559, based on the same demand and supply condition.
Comparability also base on these two assumption:
- the block Gas limit in current mechanism is also target usage in EIP-1559;
- Those value extract from consumers will completely turn into block rewards (an obviously unrealistic assumption, which is in favor of EIP-1559 supporters)
In current mechanism, there are two market conditions, basically:
We use ‘equilibrium’ to refer to where how much consumers willing to use meets how much providers willing to supply, in a free market. In the left condition, demand isn’t hot, so the equilibrium usage is U, which is lower than block Gas Limit, and gas price will be P. In the right one, equilibrium gas usage beyond Gas Limit, so actual gas usage is equal to Gas Limit, and the gas price will be P. It explains that why gas price will go up in FOMO times. (So, as you can see, if we control supply curve unchanged, ‘equilibrium’ will a useful concept to describe market conditions.)
Add EIP-1559, we can find that there are five conditions to which we need to apply analysis:
A. The equilibrium usage is higher than target gas usage, but lower than max gas usage (2X target gas usage); base fee is equal to zero;
B. The equilibrium usage is high than target gas usage, but lower than max gas usage; base fee is greater than zero, but not high enough to make actual usage equal to max gas usage;
C. The equilibrium usage is high than target gas usage, but smaller than max gas usage; base fee is high enough to make actual usage equal to max gas usage;
D. The equilibrium usage is lower than target gas usage; and base fee is greater than zero;
E. The equilibrium usage is lower than target gas usage; and base fee is equal zero;
You should realize that:
#A describes conditions when demand surges; #B describes conditions when base fee mechanism begins to work (base fee is going up), given that demand are increasing; #C describes conditions when base fee finish its adjustment, limiting actual usage equal to target usage;
#D describes conditions when demand are going down, but base fee is still changing (not equal to zero); and #E describes conditions when demand remains depressed and base fee is equal to zero.
But how can we determine whether miners’ income will increase or decrease after EIP-1559 deployed? We need those two assumptions listed above and concepts like ‘supply surplus’ and ‘consumer surplus’.
For example, as #A, we can describe it via this diagram:
In current fee market, as block Gas Limit constrains miners’ supply, so gas usage will be equal to Gas Limit, and gas price will be P1. And miners’ income will be area B + area C + area D + block rewards, consumers get area A. But if we implement EIP-1559, Gas Limit level is just target usage level, not the maximum we can used. Assuming that base fee rate is still at zero, actual gas usage will be U, and tip rate will be P2. In this condition, consumers get area A + area B + area E. It is definitely greater than area A. It is consumers' gain from 'slack mechanism' (named by Hasu). But how is miners' income? Now miners get area C + area D + area F + block rewards. If it is greater than area B + area C + area D + block rewards, in current market, miners will collectively increase Gas Limit. So we can conclude that miners' income decrease.
Similarly, we can use the same method to discuss #B and #C:
When we are in #B, with EIP-1559, miners’ income will be area B + area C + area D + area E + block rewards. With EIP-1559, as consumers still need to pay tip to incentive miners and have to pay base fee, actual gas usage will be U, and consumers' actual transaction cost will be TIP+BF (still lower than current market level!). Consumers get area A + area B + area F(still better than current market level). But miners get area D + area E + area H + block rewards*, as area C + area G is base fee burning. According our second assumption, burning value will completely turn into block rewards, then we can determine miners' income is area C + area D + area E + area G + area H + block rewards.
Whether miners’ income will increase or not depends whether area G + area H is bigger than area B.(In my opinion, the answer is 'no'. As GL must be a level that miners can maximize their income.)
When it goes to #C, base fee will be high enough to constrain gas usage to be target usage (equal to GL). Consumers can get area A(equal to current market level). The tax, area B + area C + area D will be burning. As it will go into block rewards (as we assumed), miners income will be area B + area C + area D + area E + block rewards. Yes, it is the same as current level. So about #C, we can conclusion that it can not bring about better security. It will remain the same level.
As #D and #E, look at this:
When equilibrium usage is lower than Gas Limit, in current market, consumers can get area A + area B + area F. With EIP-1559, if base fee rate is not at zero, consumers will have to pay TIP+BF instead of P, so they can only get area A. Miners get area D + block rewards* which equals to area B + area C + area D + block rewards. (Guess whether it is greater than area C + area D + area E + block reward? )
If base fee rete is at zero, the welfare of consumers will be the same as in current market. Miners, too.
Let’s get it together:
- When demand increase rapidly, as EIP-1559 allow miners to produce 2x bigger block (‘slack mechanism’), consumers can get more benefits. But when demand decrease and base fee is still greater than zero, consumers have to pay more costs which is higher than current mechanism level.
- EIP-1559 does not necessarily bring about more security. At least, in #C and #E, security level will remain as in current mechanism.
- But, #C and #E is two static equilibrium that we cannot reach, or at least can not keep it for a long time. So other conditions are what we should care.
- It seems that in #B, we can not assert that EIP-1559 will reduce miners’ income. But actually, in current mechanism, when miners find that exciting Gas Limit can not maximize their transaction fee income, they can increase Gas Limit. This fact implies that EIP-1559 (bigger block) cannot increase miners income. The only one case we cannot sure is #D. We really don’t know whether
area Bis greater thanarea E, as miners cannot limit gas supply to U1 (due to competition between miners). But, anyhow, consumers and miners losearea E + area F. Lose-Lose. - If we relax the second assumption, burning value will completely turn into block rewards, (we should relax it, as theoretically, deflation will increase every unit’s value, rather than just increasing new issued unit’s value), we have strong confidence that, EIP-1559 will decrease Ethereum’s security.
If we really want to apply ‘minimum necessary issuance’ policy, we should not implement EIP-1559.
IV. Final words
I have provided a most completed analysis to discuss every important argument relevant to EIP-1559. To summarize up, EIP-1559 can not make transaction cost predictable as demand changing is unpredictable; can not provide better user experience. Although slack mechanism can increase gas supply, it sacrifices security.
Although there are no core developers saying that it is going to deploy in Eth1.0 chain, I really worry about that. The worst is not that people supports a policy that they don’t understand, but that they know that the policy is injustice but they are beneficiary so they support it.
Half a year ago, block utilization of Ethereum is around 80%. But recently, it goes to 95%. Why? When average gas price is low, many mining pools choose produce empty block. But when average gas price is very high, they have strong incentive to produce full block. Otherwise, miners will give up them and join in pools with higher revenue.
This fact shows the correct way to solve this problem.
Incentive to miners is not evil, but the precondition to solve this problem. The rest should load to client implements.
