Part 2: Objections about evidence
These questions, several of which come from hanniabu’s list [13] compiled from discussions with SSPs, stakers and DeFi protocols. Those questions share a structure: we lack evidence on X, therefore wait. Two things are worth saying about that structure up front, with all due respect because the questions themselves are good ones. Several have answers in the literature and the rest shaped the EIP’s design.
- “Do nothing” is not the null hypothesis. It is a policy choice like any other, and it happens to be the only policy whose consequences we have direct evidence about: the entry queue is saturated at maximum churn, the ratio passed 1/3 in April 2026, and the worst-case trajectory reaches >70M ETH (>55% of supply) by January 1, 2028 [7].
Demanding proof before changing the curve while demanding none for keeping it places the entire burden of uncertainty on one side.
- Several of these unknowns are unknowable which is an argument for the taper’s design, not against acting. The supply curve (the distribution of stakers’ reservation yields) cannot be observed by the protocol, only revealed by the market over time. That is precisely why the EIP does not tune a parameter to a point estimate: it removes the yield floor and lets the market find the equilibrium, wherever the unobservable supply curve puts it [1]. Elowsson’s framing: because “the future level of the supply curve will always be unknown,” a reward curve must be designed to produce acceptable outcomes under any reasonable scenario [2].
A mechanism robust to ignorance is the correct response to ignorance.
6. Where will the staking equilibrium settle under the current curve and the proposed curve?
The steelman: without knowing the likely equilibrium ratios and yields under both curves, we can’t even establish that the current approach is a problem.
The answer: the precise location is unknowable in advance under either curve but the two curves differ in the one property that matters: whether an equilibrium is guaranteed to exist at all.
-
Under the taper, an equilibrium below 50% exists for any positive risk premium. Net yield declines to zero at the 50% saturation ratio, so it crosses whatever premium the marginal staker demands at some ratio strictly below 50% [1]. We don’t need to know the supply curve to know the market has a stable resting point.
-
Under the current curve, an equilibrium is not guaranteed. Yield never falls below ~1.5% even at full staking, so growth stops only if the marginal staker’s premium stays above that floor. This is a premium the protocol neither observes nor controls, and which keeps falling as staking matures [1]. If it falls below the floor, there is no equilibrium short of near-total staking.
-
What can be estimated has been estimated (including by people who reach the opposite conclusion). On the location of the equilibrium, Elowsson’s supply-curve scenarios put the yield difference between the current curve and a tempered curve at roughly half a percentage point, with equilibria near ~30–36M ETH under moderate assumptions [2]. On the related question of composition, three cyber·Fund-supported studies reach a conclusion adverse to this proposal, and we would rather point at them than around them: Eloranta & Helminen measure a 12% mean-return advantage for large pools today, widening to 13–15% under the reductions they model absent MEV burn [9]; Arnold et al.'s calibrated game-theoretic model predicts that “a reduction in issuance is likely to crowd out solo stakers,” with preliminary empirical support [10]; and Zhu, Korinek & Duckworth’s interviews across eight staking cohorts find solo stakers “most vulnerable to changes in the issuance curve,” while institutional and retail delegators are comparatively inelastic [12]. Two observations that we present as disagreement rather than dismissal.
First, the mechanism all three identify is the same: as consensus issuance shrinks, execution-layer rewards (where large operators have a measurable edge) become a larger share of staker income. [9] puts that edge at 12% today and 13–15% under the reductions it models. This EIP leaves execution-layer income untouched, so it neither creates nor closes that gap, but it does make it relatively more important. [9] separately warns that economically capped curves could leave large pools profitable while smaller ones run at a loss; this proposal is not an economic cap ie. net yield reaches zero only at a 50% ratio the market is not expected to reach, but that is a distinction about regime, not a refutation of the trend. Second, the competitive channel [9] and [10] identify (delegated staking’s access to MEV and stacked DeFi yield) is real, and this EIP does not address it; that is precisely why [9] recommends MEV burn alongside gradual issuance adjustment. Our proposal anticipates the sequencing rather than contradicts it. Overall, the supply curve is unobservable and only revealed as the market moves.
And is the status quo acceptable?
The demand for equilibrium evidence is difficult to sustain as a defence of the current curve, because the current curve is the one generating the evidence: a saturated entry queue adding ~1.75M ETH per month, with no crossing point in sight [7]. Waiting for the current curve’s equilibrium to reveal itself is the experiment and this experiment runs at ~1.5 extra percentage points of supply per month.
7. Don’t reservation yields and operating costs differ across cohorts (solo stakers, LSTs, exchanges, custodians, ETFs) in ways that must be analysed separately?
The steelman: aggregate analysis hides the distributional story; each cohort has distinct costs and incentives. Two sharpening demands come with it: differentiate the operator’s costs and reservation yield from the beneficial ETH owner’s, and use an average developed-country tax rate rather than dismissing tax.
The answer: agreed on all three counts and the analysis exists.
-
The cohort literature (accurately) is split. Elowsson’s reservation-yield distributions model solo and delegating stakers separately and conclude that “risks could very well price out delegating stakers earlier than solo stakers as the yield falls” [2]. Three cyber·Fund-supported studies conclude the reverse: Eloranta & Helminen measure a 12% mean-return advantage for large pools today, widening to 13–15% under the reductions they model absent MEV burn [9]; Arnold et al.'s calibrated model predicts that “a reduction in issuance is likely to crowd out solo stakers” [10]; and Zhu, Korinek & Duckworth’s interviews across eight cohorts find solo stakers “most vulnerable to changes in the issuance curve” [12]. Julian Ma’s stake-distribution model ties issuance level to who stakes, not just how much [14]. So, to put it bluntly: has separate cohort analysis been done? Yes and it is genuinely contested.
-
Where we stand, and on what. Our reading rests on the wedge structure. A delegator must clear a fee (20–80 bps) plus an issuer-risk premium; a solo staker post-EIP-7251 has low per-ETH opex, no fee, and arguably the lowest risk premium of any cohort [11]. In delegated staking two different people each have a minimum they need cleared: the operator, who is paid through the fee, and the owner, who pays that fee and carries the counterparty, smart-contract and governance risk. Both of them come out of the same yield. A solo staker is both people at once, so there is only one minimum to clear.
-
Where the adverse studies bite, and where we think they don’t. [9] and [10] model curves that drive yield toward zero or impose an economic cap, and identify delegated staking’s access to MEV and stacked DeFi yield as the channel that squeezes solo stakers. Both observations are fair. Our answer is that this proposal’s equilibrium sits at a positive market yield rather than in the regime they model, and that [9]'s own recommendation, ie. MEV burn alongside gradual issuance adjustment, is (again) the sequencing this EIP anticipates. What we do not claim is that the MEV and DeFi-stacking asymmetry is solved here; it is untouched by this proposal, and it remains the strongest open objection in this section.
-
On tax, accepting the methodology lands the same way. Take an average developed-economy treatment: staking rewards are typically taxed as income on the nominal amount at receipt (with capital gains on later disposal). Applied across cohorts, this taxes the dilution component of yield as if it were real income which is a burden that scales with issuance and falls hardest on the non-pooled staker who can’t structure around it. A lower-nominal/higher-real policy shrinks precisely that taxable phantom component [1, 4].
-
Where genuine uncertainty remains, the EIP absorbs it. The dispersion of delegator risk premia versus solo cost floors (the “valley” hypothesis [2]) is real open territory. This is why the taper lands at a positive market yield rather than the near-zero regime where that valley would bite, and why the 18-month transition plus fork lead time gives every cohort ~2 years to reveal its reservation yield gradually rather than at a cliff [1].
And is the status quo acceptable?
Cohort dynamics don’t pause while we study them. Under the current curve, the differences compound in one direction: dominant SSPs’ scale advantages grow with quantity staked, the LST money-function externality strengthens, and delegation becomes relatively more attractive the higher the ratio climbs [2].
8. Don’t decentralization effects depend on who exits, who remains, who switches to delegation and not on total staked ETH?
The steelman: an issuance cut could leave total stake looking healthy while gutting exactly the validators that matter.
The answer: his is the correct frame and the frame the EIP was designed in.
-
Exit ordering is the design criterion, not an afterthought. The burn is sized from idealised per-duty rewards precisely so that no correctly-attesting validator is pushed into negative epochs while waiting for rare duties, ie the failure mode that would push small validators out first [1, 3]. The wedge analysis then predicts the ordering at the macro level: highest-wedge cohorts (fee-charging delegation) hit their exit point before lowest-wedge cohorts (consolidated solo) [2].
-
Switching to delegation is a yield-chasing behaviour and the taper reduces the pressure that drives it. Under the current curve, dilution-avoidance pressure pushes holders into whichever wrapper is most convenient (usually an LST or ETF). At a market-clearing yield, the “stake or be diluted” coercion weakens because issuance is bounded [1], so the marginal holder’s rational choice can again be not staking (the only outcome that adds no delegation at all).
-
What’s uncertain: sticky delegators (ETF flows that never rebalance) blunt the ordering, and solo-staker behaviour at low yields has real dispersion [2].
And is the status quo acceptable?
By the objection’s own metric (who exits, who remains, who delegates) the current curve is not performing well at all: solo stakers are a single-digit share and falling; marginal inflows arrive through ETFs, exchanges, and LSTs; and the dilution-plus-nominal-tax mechanism pressures the most independent validators first, with no mechanism to bring them back [1].
9. What level of staked economic security is actually viable or sufficient? There’s no definition or evidence.
Note: compared to 1. we are looking at what is the minimum viable here.
The steelman: we’re re-engineering the security budget without ever defining what security requires.
The answer: the absence of a derived optimum is real and it cuts against the status quo at least as hard as against the taper. The EIP does now defend the 50% demarcation on principle: past majority-staked, any rescue fork depends on the stakers themselves and derivatives drawn from the majority pool dominate collateral; half the supply is “the only value in the range defensible without appeal to preference”; and it is bounded below by today’s ~33%, where a lower saturation point would zero net yield at activation [1].
-
The status quo picks a security level too. The current curve’s implicit target is “as much as the market will stake at ≥1.5% yield” therefore a priori trending toward everything that can go at stake. That number was never derived from a security requirement either. Choosing no bound is also choosing a level: ~100%, without evidence it buys anything.
-
What we do know bounds the problem. The cost of attack is set by the slashable stock an attacker must acquire and forfeit; at any ratio in the tens of percent that stock is vast relative to plausible attack rewards [1]. Named reference points exist (Drake’s ~1/4 of supply, Vitalik’s note that ~1/8 would be fine [2]) and Elowsson’s 2016 comparison shows the binding term is ETH’s value, not the staked fraction [2].
-
Under uncertainty, choose the recoverable error. If the equilibrium under the taper ever proved genuinely too low, raising issuance by fork is straightforward. The reverse correction, clawing stake back from an entrenched, custodial, fee-earning majority, is the politically hard one.
And is the status quo acceptable?
It answers the undefined-security question with the most expensive available guess: pay for all of it. On the do-nothing trajectory, ~1.4M ETH per year is issued at 70M staked [2], ie. more than a billion dollars a year in unnecessary costs by Elowsson’s welfare accounting [2] to purchase security increments no one has shown to be needed, while degrading the social-layer backstop [1, 2].