Original source: Anders Elowsson, Ethereum researcher
Original compilation: Deep Chao TechFlow
I think it is very important to achieve the Minimum Viable Issuance (MVI, Minimum Viable Issuance), which is an important commitment to ordinary Ethereum users. Staking should be able to secure Ethereum, rather than become an inflation tax while reducing utility and liquidity and creating oligopoly risk.
Ethereum is constantly evolving and may drive the global financial system in the future. We must assume that the average user will have about the same understanding of the inner workings of Ethereum as the average person does of the current financial system.
Of course, we cannot assume that the average user will be driven by some ideology, as was the impetus for Ethereum’s initial creation. Our job is to ensure that the right incentives are in place so that Ethereum can grow unimpeded.
An important design principle that has existed since the birth of Ethereum is the Minimum Viable Issuance (MVI), that is, the amount of ETH issued by the protocol should not exceed the amount required for strict security. This principle is sound whether under Proof of Work (PoW) or Proof of Stake (PoS).
Under PoW, the role of MVI is to prevent miners from charging excessive inflation taxes to ordinary users. As a result, the block reward was reduced from 5 ETH to 3 ETH and finally to 2 ETH.
Under PoS, the MVI principle should also be adhered to, and excessive inflation taxes should not be charged to ordinary users. Ordinary users shouldn’t need to worry about the details of staking to avoid having their savings eroded, or supporting a potentially censorable set of validators, etc.
So, MVI is really about being able to keep the collateralization ratio (the proportion of all ETH that is pledged) high enough, but not higher. In this article, I will try to explain why issuance beyond the minimum feasible amount reduces the utility of Ethereum.
Benefits of MVI in user empowerment
For individuals, participating in staking has various opportunity costs. It requires resources, focus and technical knowledge, or trust in a third party, while also reducing liquidity. Liquidity Staked Tokens (LST) are not as reliable as native tokens, nor are they as suitable as currency or collateral.
Therefore, individuals hope to be able to earn rewards through staking. Define their minimum expected rate of return as the lowest rate of return they are willing to stake (using their best staking method). Ethereum’s (reverse) supply curve then derives from the minimum expected rate of return for future Ether holders.
The reserve yield for holders can be described as the “indifference point”, at which point the utility they receive from staking is equivalent to not staking. This means that lowering issuance can actually increase utility for everyone, even stakers, as long as Ethereum remains reliable and secure.
Consider a hypothetical supply curve (blue) with a yield elasticity of supply of 2. In this example, I set it so that when the mortgage amount D reaches 25 million ETH, the yield reaches 2%, that is, when 25 million ETH is mortgaged, the minimum expected yield of the marginal staker is 2%.
The supply curve in reality is a rather complex phenomenon, and we have not yet reached an equilibrium point where we can anchor its position, but we will start with this simple and fairly realistic scenario. We will also ignore the complexities of compound interest.
The burning rate b is set to 0.008. This is the amount of ETH that has been burned as a percentage of the total supply expressed as an annualized rate since the merger. But this is not the key point, as we are focusing on the shift in supply and demand between medium-term equilibrium points (circles), not the drift in the total supply of ETH.
Realized Extraction Value (REV) (just over 300,000 ETH per year) has been added to the protocol issuance to create the black demand curve (current policy) and the green demand curve (by reducing the base reward factor F from 64 to 32 , halve the issuance).
Halving the issuance reduces the yield y (red arrow). This reduces the issuance yield yi=y-yv (where yv is the yield from REV), thereby reducing the issuance i=yid and the circulating supply inflation rate s=ib (orange arrow).
Within one year, the proportion of circulating ETH holdings that someone can obtain changes P, depending on s and the rate of return y of each holder, according to the formula: P= 1+y/1+s-1
The current issuance policy gives P 1 , and halving the issuance gives P 2 . Then their proportional relationship is: P= 1+P 2/1+P 1-1
The relevant utility change is defined as u=P, but when calculating P 2 , for those who stop staking, their respective minimum expected returns are used. Below that yield, they would not have pledged in the first place, so they do not suffer additional utility losses as yields fall further.
According to this definition, everyone will gain higher utility at the new equilibrium point. Stakeholders see a decrease in returns, but a greater decrease in supply inflation, allowing them to receive a greater proportion of ETH.
Of course, non-stakers are obviously better off because the only change for them is that less ETH is issued to stakers. Those who stop staking are the only participants who receive a reduced proportion of circulating ETH at the new equilibrium point.
Despite the friction, they are still implicitly better off due to increased utility. For example, a marginal staker at the old equilibrium point didnt care about staking at all, so could stop staking and get the full utility improvement from supply deflation.
This is when people who stop staking find themselves somewhere in between, still benefiting from the drop in inflation but suffering some loss in earnings, until they become indifferent to staking and de-staking. We have shown that, from a utility perspective, issuance policy is not a zero-sum game.
In addition, any increase in utility gained by any group will generally benefit all token holders.
As long as they own the underlying ETH, everyone can benefit from MVI. This excludes CEX and other staking service providers (SSPs) that profit from staking fees. They will not benefit from a reduction in supply inflation and want to keep yields high to keep production cuts high.
But issuance above MVI forces unwilling stakers to suffer reduced utility when staking, or reduced economic consequences when not staking. Under a realistic supply curve, things get worse even for willing stakers. Note that this example doesnt even take into account tax implications.
For a PoS cryptocurrency with a 5% yield, everyone stakes, and the average tax on the staking yield is 20%, with 1% of its market value used for tax every year. This is higher than the amount that Bitcoin will dissipate to miners after the next halving.
The debate does not necessarily hinge on how users feel about tax levels or how to interpret staking benefits. We can still conclude that by implementing MVI, Ethereum remains more neutral regarding differences in tax policies between nation-states.
Arguably, Proof of Stake requires lower rewards to achieve the same level of security as Proof of Work, and it is important to take full advantage of this to maximize user utility. For example, if the yield is 2% and the total reward for staking 25 million is Y= 0.022500 = 500,000 ETH.
The return rate for maintaining this solid security is approximately r=Y/S= 0.4%, which is an astonishingly low number. We take full advantage of this to maximize utility for our users. Potential balances with current issuance policy are represented by black circles.
The rate of return is about 3%, and 50 million ETH are pledged, that is, Y = 1.5 million ETH/year. A reward difference of 1 million ETH per year (over $1 billion at current token prices) could be awarded to Ethereum users in a non-dilutive way for token holders.
For MVI, withdrawing an average staking fee of 15% will provide CEX and SSP with approximately $250 million in excess profits per year. Some will be passed on to the companys shareholders, and some may be used to lobby to keep yields above the MVI forever.
From a macro perspective, the benefits of MVI
I often think it is desirable for ether to permeate the ecosystem. In the case of L2, Bridge Ether binds L1 and L2 together and provides users on L2 with external funds, thereby increasing their financial security.
If you create a system where users have to rely on some opinionated ETH derivative as funding to avoid inflation taxes, then the entire ecosystem is more susceptible to disruption.
For example, consider the following scenario: users who cannot stake give their ETH to an organization (SSP) that runs a validator for them. These organizations can issue LST as collateral and use it on Ethereum.
If the protocol did not run under MVI, but ran with a higher deposit ratio, one or a few LSTs might replace the currency in the Ethereum ecosystem, embedding it in every layer and application. What impact will this have?
First, the positive network externalities brought about by the currency function may allow LST to remain dominant while its SSPs provide worse services than competitors (e.g., charge higher fees or only offer poorer risk-adjusted rewards).
Secondly, and most importantly, LST holders and any application or user that needs LST to maintain its value will form a common destiny with LST and the eventual LST issuing organization (SSP).
This would require Ethereum to destroy a significant portion of itself. Affected users may prefer to reinterpret the error or inappropriate behavior as something completely different. Once you become the currency of Ethereum, you become the social layer in a way. We no longer only care about the proportion of ETH pledged under LST, but the proportion of the total amount of ETH under LST. Corruption institutions are accordingly located one layer above the consensus mechanism.
It is clear from The DAO that if the proportion of the total circulating supply affected by the outcome becomes large enough, the social layer may waver in its commitment to the underlying intended consensus process.
If the community can no longer effectively intervene in events such as 51% activity attacks, then risk mitigation in the form of early warning systems discussed by Buterin may not be effective.
In this case, the consensus mechanism has become so large and interconnected through derivatives that its ultimate arbiter - the social consensus mechanism - is overloaded.
Now consider a different situation under MVI. First, each LST will face tougher competition from non-collateralized ETH. As a result, the ability to monopolize monetary functions and then charge high fees or offer riskier products is diminished.
Second, the social layer will continue to be natively tied to Ethereum and ETH, rather than to external organizations and the ETH derivatives they issue. Keeping the collateralization rate low enough through MVI therefore changes the risk calculation of the participants.
Under MVI, when the collateralization rate is low enough to prevent moral hazard from developing, the agency problem (PAP) that delegates collateral to the principal that dominates LST can be priced more accurately. No LST will grow to the point of being too big to fail in the eyes of the Ethereum social layer.
This pricing will reflect the fact that the larger the share of the pledge controlled by an agent acting on behalf of the principal (or any party capable of intervening in the relationship), the better its chances of obtaining consensus that is degraded for its own benefit.
A delegated staker must always consider what security guarantees it has (e.g. the risk to the value of the staking agent or intervening party itself), knowing that it could lose everything if the worst happens.
Removing the direct dominance of the Ethereum currency, and assuming that the deposit ratio has grown to a utility-maximizing size under MVI, larger SSPs are likely to find the non-monopoly strategy more profitable (i.e., fees increase).
This is just a comment relevant to now. But importantly, it reflects the fact that the value proposition of a secure and value-aligned SSP increases in relative terms for every cartel class we can eliminate.
An important step towards MVI is MEV burning, which may also have the potential to eliminate the cartel class that is more important than the monetary function. MEV burning helps reduce the reward variance for independent stakers, which would increase if issuance yields were reduced.
It also brings greater precision to targeting MVI because it eliminates a revenue source that may change over time in ways that cannot be predicted in advance.
It is worth noting that various methods may be adopted in the future to deal with the principal-agent problem of certain aspects of delegated staking (i.e. one-time signatures). But fundamental issues of building trust, monopoly incentives and the ability to scrutinize may be difficult to escape.
Another benefit of MVI is that it improves the conditions for (independent) staking, which is related to the direct relationship between staking size, number of validators, and validator size. If the stake size changes, the validator size or number of validators (network load) will also change.
This effect spreads throughout the protocol design space and affects any goals that might be substituted for higher or lower network load, such as parameters related to variable validator balances.
This is a basic attribute of the current consensus mechanism. If the issuance policy results in d= 0.6 at the mid-term equilibrium point instead of d= 0.2, independent staking would require three times as much ETH to maintain the same network load, ceteris paribus.
Getting back to the basics, I think the most important benefit of MVI is its ability to provide utility to the average user. Ethereum is in a unique position to enable the native cryptocurrency to become a global currency, and I think this is an opportunity worth pursuing.
When countries implement price inflation by increasing the monetary base, they control the time choices of ordinary people, assuming that such control is still feasible in a digital and globalized world.
Ethereum should not control ordinary people nor force them to conserve energy for liquidity. We should allow them to maximize the ease and utility of using Ethereum currency. The risk-free rate in Ethereum is simply holding (and trading) ETH.
Solving potential problems with MVI
After elaborating on the potential benefits of MVI, Part II will address some of the proposed shortcomings. These include reduced economic security and the notion that if we reduce returns, delegated staking will replace all independent staking.
When it comes to the first point, this is indeed true, as a higher deposit ratio does force the attacker to spend more resources on, for example, restoring finality. This is not something to be taken lightly.
Our goal is not minimum issuance. We must always make sure it is doable. Buterin provided some intuitive explanations on how expensive a 51% attack on Ethereum should be.
We can also consider the nearly 14 million ETH that secured Ethereum at the time of the merge to be the collateral that the ecosystem deems to be secure enough under the current consensus mechanism (in terms of resistance to female ledger attacks, not just super committee accountability) The preference of scale.
At the same time, having a sizable margin is indeed good, and the current collateralization rate (d 0.2) relative to the collateralization rate at the time of the merger (d 0.1) may also provide a meaningful improvement from the perspective of resisting false accounting.
The slope of the reward curve cannot be too steep, which is why we may want to operate at some distance from the preference point and can ultimately determine d from a probabilistic analysis of staking supply and demand.
Some may argue that delegated staking somehow makes it easy to attack resource allocation and that this is only apparently secure. But by making all staking subject to penalties and removing moral hazard (via MVI), delegators have to be very careful when delegating staking, as mentioned previously.
In this setup, the market determines the appropriate capitalization ratio for staking operators and prices the risk of staking. Instead, Ethereum is responsible for punishing misbehavior and maintaining the value of ETH relative to the value it secures.
By ensuring that ETH tokens penetrate the real economy and that all consensus participants have a real stake, we set a price for attacks that is harder to circumvent through financial engineering.
I mention this because there are indeed some interesting alternatives being discussed where Ethereum steps into the delegation process without any risk to the delegator. Then the risk is much lower for delegators who contribute to the deterioration of consensus.
Or at least thats what it seems. When Ethereum forks and/or must be saved by social intervention, if the worst happens, risk-free delegators may be surprised by how the social layer evaluates their delegation and the damage they are perceived to have caused.
Here I return to Buterins request not to overload the consensus. My point and a theme of this article is that when the proportion of ETH involved in the consensus process is very high, everyone will be involved and a neutral outcome may not be achieved.
The conclusion to the first question is that d under MVI must remain large enough to ensure security. Delegation does reduce security to a certain extent, but as long as their pledge is risky, parties will try to assess the risk and delegate wisely. .
Retaining independent pledgers is indeed a complex puzzle. Economies of scale are difficult to design out, and we don’t pay enough attention to liquidity in staking. However, there are some nuances in the current argument that are more favorable to MVI that I hope to be able to bring up.
Independent home stakers of Ethereum will incur certain costs when staking. They pay a large portion of the cost up front, including access to knowledge. They also incur variable costs such as bandwidth, troubleshooting time, and outage risk.
Many of Ethereum’s SSPs also incur significant costs when designing their services and bear other types of operating costs that independent stakers don’t have to worry about. However, they rely on economies of scale to lower the average cost of operating a validator.
We must assume that SSPs seek to maximize profits and can consider what their fees might be under different equilibria. What are the differences in economies of scale between d= 0.2 and d= 0.6? It seems reasonable to assume that SSP has a much lower average cost at d = 0.6.
Remember that at d= 0.2, an individual staker might be able to run a validator three times smaller than at d= 0.6. In terms of the proportion of individual stakers we can attract, there may be a difference between the minimum number of validators being 32 ETH and 96 ETH (or 11 ETH - 32 ETH).
So not only does a higher d force independent stakers to have more ETH for the same network load, they also have to compete with SSPs who are able to charge lower fees. While fees will be set based on market strategy, average cost should ultimately matter.
If we reduce earnings, SSP expects fee increases to be required to properly cover and amortize costs. The cost of delegating a staker is variable and includes PAP and fees. They can easily get away with the added fees.
The argument that lowering returns will cause independent stakers to leave (earlier than delegated stakers) is one that should be taken seriously. But because current household pledgers have already incurred fixed costs, their current personal income supply elasticity may not be high.
However, their lower elasticity in the short term will not help if we reduce returns to the point where independent staking by households becomes unfeasible (including for new entrants). If we want to maintain independent staking, there is a lower limit on the total staking return that we cannot go below.
Assume that the total cost of staking an independent family (denominated in ETH) is C, and consider other factors such as the annual risk to funds when staking, R. Then, the income must be higher than y>C/32+R. Even if re-staking brings liquidity, a reasonable margin is required.
Here, I also want to discuss the impact of DeFi returns. All stakers will receive the return y endogenous to the pledge. This “endogenous revenue” comes from issuance, MEV and priority fees. Some people may also obtain exogenous benefits yc outside of the consensus mechanism.
There is no way to simply sum y+yc for LST holders and conclude that no matter how y falls, LST holders always make a profit relative to independent stakers. It can be expected that ETH tokens will bring higher utility relative to LST (when their endogenous returns are not taken into account).
Delegated stakers must weigh y(1-f), where f is the percentage fee, against the risks/costs including the inherent disadvantages of PAP and LST relative to native ETH, and only if y(1-f) (instead of When y+yc) exceeds these costs, the decision is made to pledge.
When y=0, the agent will not delegate pledges. They can get better liquidity or higher yc with native ETH, and face serious disadvantages by delegating staking to an SSP that is operating at a loss. Independent stakers may not pledge either.
For someone who wants to hold ETH anyway, the decision may not depend on whether yc is 1% or 5%. At 5%, one can expect ETH to offer +5%. Of course, that 5% carries risk and is not free money (neither should our returns, hence the MVI).
As y rises, potential independent stakers and delegated stakers will gradually find the staking claim worthwhile, starting with the most ambitious/adventurous ones. Here we are forming a supply schedule where each agent makes decisions based on its specific situation.
It is unclear how the minimum expected yield is distributed between potential independent pledgers and delegated pledgers. At the mid-term equilibrium point of d= 0.2, the proportion of independent stakers may be lower than d= 0.6, but the alternative is also very likely.
Higher d may allow SSPs to be more diversified, but the cartel class of monetary functions puts pressure on this. The proportion of individuals with enough ETH to stake independently is also limited, which sets a soft cap on the total number of independent stakers.
This is indeed a topic worthy of further research. The key is that the opportunity cost of staking must always be fully taken into account, and economies of scale and monopoly can affect the underlying equilibrium analysis in quite complex ways.
Finally, restaking has the potential to make independent stakers more competitive. It enables them to re-mortgage their stake when they wish (but they may also run into principal-agent problems themselves if they want to provide financial security).
One benefit of restaking is that if the Active Verification Service (AVS) can quantify decentralization, it can also give economic residual value to decentralization. This is something Ethereum cannot do as an open protocol.
The previous argument also applies to the functional re-pledge of EigenLayer outside the regulations. At very low yields, users are better off using non-collateralized ETH directly (free staking). For many use cases, it seems reasonable that AVS would prefer a token that will not evaporate easily.
Also note that if PEPC expands its scope beyond the block production use case, the benefits generated may become more endogenous, depending on the residual utility provided.
Looking to the future
This concludes the discussion of the advantages and disadvantages of MVI. While there are some worrisome issues with staking alone, MVI is a fundamentally sound design policy that gives Ethereum a real chance of delivering the best digital currency ever to its users.
Every argument has its nuances, and some discussions cannot be succinctly expressed in a tweet. But I think taking all factors into consideration, it should be possible to accept that MVI is also a favorable design principle under PoS.
We must always focus on the “average user” first, which requires looking at the micro-foundations and evaluating how we can maximize utility for the average person when Ethereum (hopefully) becomes their new financial system.
The question then is how do we implement MVI, which is something Ive been digging into. Dietrichs mentioned the importance of communicating current release policy research during a recent developer call, and my process started with this tweet.
Changing issuance policy is a sensitive issue. What we desire is an issuance policy that maximizes utility without requiring further developer intervention, so that it can always proportionally distribute utility-maximizing MVIs.
However, the current reward curve does not allow the protocol to affect the staking rate (security), but the staking size. In the medium term, the results are closely related to the two, but in the long term equilibrium, there may be clear divergences, as the circulating supply drifts.
This is the topic of my article in Ethresearch and speech at Devconnect in 2021: Define how the circulating supply S drifts toward equilibrium (i=b) so that we can improve the reward curve and achieve the minimum feasible issuance amount under proof of stake.
Since according to the current reward curve, the issuance amount i can be expressed as i=cFd/S, it will change with changes in circulating supply (the mortgage rate d provides a certain adjustment space). The graph shows Ethereum’s issuance rate diagonally and its average b since the merger.
The burn rate b will not depend on the circulating supply - the demand for block space will not change due to changes in the currency denomination unit. If i>b,S will rise and pull i down until it equals b. if i
In 2021, stakers do not have REV yet, so I directly used the minimum expected rate of return y-, and concluded that the security of Ethereum is d=b/y.
Today, we simply add the “REV rate” v to the equation and get d=(b+v)/y. The point is that we have no control over the collateralization rate and security in the long term unless we are prepared to change F from time to time.
We can cut F as a temporary solution to avoid paying too much for security (this will be discussed in the next tweet). However, Ethereum will eventually return to the same long-term equilibrium collateralization rate at a lower circulating supply (all else being equal).
This is why we ultimately want to change the reward curve to be related to d rather than D. It then looks tempting to simply replace D with S 0 d (where S 0 is the current circulating supply). This brings us one step closer to an autonomous issuance policy, but there is still no guarantee of achieving it.
Assuming MEV burning, the protocol is fully adaptable to changes in income, but still cannot adapt to permanent shifts in expected yields, i.e., the supply curve. This can be handled by allowing the entire reward curve (demand curve) to drift slowly.
The ultimate goal is a dynamic equilibrium where circulating supply can change at a constant rate without external influences. Whether it is inflationary or deflationary depends on the supply curve and how the block space value is reflected in the ETH market cap.
As a result, we achieve what Polynya calls constant security, which I think aptly describes our ultimate goal, to ultimately wrest issuance control from developers and make Ethereum autonomous under MVI.