Staking yield and ratios
Commentary on the concept of staked ether as The Internet Bond
Author’s Note:
From a high level, many aspects of proof-of-stake seem intuitively simple, yet are misunderstood by even those who dedicate their lives to studying cryptoassets. This has led to an incredible quantity of misinformation in the space, with big Twitter accounts attracting engagement from the brainless masses while making fools of themselves to anyone with even a rudimentary understanding of the second-order effects of staking.
This piece hopes to clarify a couple misconceptions and inspire the reader to ask what else they might be misunderstanding.
Many people naïvely believe that everyone should stake their tokens, perhaps taking inspiration from alt-L1s that have developed high staking ratios. When you dive into the details, the situation reveals itself to be far more complex.
At an individual level, the choice of whether or not to stake can be approximated as a simple balance between yield, staking risk, liquidity drawbacks, and convenience1.
Adjusted_Yield = Yield × (1 - Discount_Risk) - Liquidity_Premium - Convenience
In shorthand:
A = Y × (1 - D) - L - C
A, the adjusted staking yield, is the return seen by a validator after adjusting for the other variables. If the adjusted staking yield is positive, then the optimal choice is for the ether holder to stake their tokens.
Y, the staking yield, is the average expected yield for the validator for the intended staking period. For a genesis staker yields may initially have been over 10%, but with their tokens locked for multiple years they may have seen on average a yield of 6%2.
D, the risk discount, is a direct modifier to the staking yield that accounts for validator downtime, possible slashing of rewards,
technical risk from a failed Merge, etc. It can also be used to incorporate pool fees or exchange fees, but I prefer to think of third party hosted staking as form of DeFi rather than staking.L, the liquidity premium, is the opportunity cost of being locked into staking. For example, if you’re locked into staking then you’re unable to use that same ether to flip NFTs or to instantly transfer funds or sell your tokens should the need arise.
C, is a convenience modifier, ie: I’m personally not going to go through the long process of setting up a node if my return on investment is going to be 0.1% annualized. I would personally require a couple percent per year to be encouraged to handle everything myself3.
It’s probably also appropriate to roll the cost of running your home validator into C. Or you can add another cost variable to the formula.
Each Ethereum holder must do this calculus and find their adjusted staking yield to determine whether it is positive or negative. If positive, the holder should choose to stake, but if negative then the risks and drawbacks outweigh the benefits and they should not lock away their tokens.
With this methodology in mind, it’s no surprise that liquid staking derivatives (LSDs) have become as popular as they are today:
The validators are run by generally reliable operators reducing the risk discount.
LSDs allow you to exit before withdrawals are enabled, perhaps below par-value but still better than alternative of being stuck holding indefinitely.
LSDs allow for increased liquidity as you can exit quickly through Curve rather than waiting for the validator queue.
LSDs are more convenient because you don’t need to manage the validators yourself.
To compensate, liquid staking providers take a cut of yield, but for the majority of holders LSDs are an effective tool to capture yield.
Rearranging the formula, we can define the positivity of the adjusted staking yield by the inequality:
Y × (1 - D) > L + C
Through this framing, the question of whether to stake evolves into whether the staking yield discounted for risk is larger than the opportunity cost of staking and effort required to run a validator.
When people from traditional finance come into the space and begin talking about staking as the risk-free rate, or as an internet bond, they frequently neglect the risk associated with the activity. A more accurate representation is:
RFR = Y × (1 - D)
With the introduction of a risk-free rate, a division forms between financial speculators and technologists. From a financial standpoint, to maximize the dollar value of ether a compelling strategy is to set a high risk-free rate in order to pull as many tokens as possible into the staking contract and create a supply crunch. This line of thinking leads to price models reliant on discounted cash flow (DCF) projections or similar, where analysts model staking yield like earnings from a company to suggest that higher yield should drive token price higher.
The under discussed ramification is that a high risk-free rate makes it nearly impossible for everything else to compete. If you have a risk-free rate of 10%, then everything else: DeFi, NFTs, blockchain gaming, etc. need to either generate yield or appreciate in excess of that rate (+ a spread for the additional risk). A high risk-free rate simply leads to a long-term stagnant ecosystem.
Two essential realizations for non-speculators:
Low staking yield is better than high staking yield for enabling a robust ecosystem.
A high staking ratio is a sign of a failing blockchain.
Polynya suggests that their preferred staking ratio could be as low as 10% (assuming sufficient network security).
The comparison between staking ether and an internet bond carries additional nuance: buying a bond makes you short yields because you’re locked into a rate, unless the bond is floating rate, but usually the rate on floaters is designed to hold the spread constant as an underlying quantity like LIBOR or the Federal funds rate changes. By contrast, when staking ether the holder remains long yields. As a reminder, staking yield is currently inflated and will certainly be falling in the coming months, therefore being long yields is not ideal.
Staking ether also comes with massive risk if the holder is not hedging changes in spot price. In this sense, staked ether is similar to a foreign bond with large currency risk that most institutions would be required to hedge by shorting spot or futures4.
Case Study: EIP-1559
Burning transaction fees is the most prominent example of a misunderstood upgrade. Often criticized as financial engineering designed to inflate the price of tokens, in a proof-of-stake environment EIP-1559 lowers the risk-free rate and makes it easier for dApps to compete with staking.
The main community narrative around EIP-1559 has always been deflationary supply, but for infinite holders (who should gravitate towards staking to guarantee non-debasement) token burning is actually a net-negative!
A mental model that can help clarify this is to reframe token burning from a reduction in supply into a distribution to all ether holders. Through this lens, the transaction fees that normally would have gone solely to stakers are fed back to all community members—when implemented with a proof-of-stake blockchain, EIP-1559 is actually a net-decentralizing mechanism, not the centralizing mechanism that bitcoin holders often claim.
A recent criticism of EIP-1559 is that it’s a form pro-cyclical monetary policy. During an episode of the On The Brink podcast, Nic Carter pointed out that the correlation between token price and activity leads to less burn during a bear market and more during a bull market. This generates a feedback loop that in theory will lead to higher highs but less attenuated bottoms.
One of the big issues with reasoning about first and second order effects with crypto is that we have so many correlated and inversely correlated metrics, so it gets really easy to extrapolate inaccurately. In this context, what the gas market is able to control isn’t the price of the cryptoasset, but instead the utilization of the network—by decreasing gas prices there will be lower burn, but the lower gas prices help to spur usage of the network. This effect is similar to the Federal Reserve printing money during a recession: in theory every piece in circulation becomes less valuable, but with more money going around the economy can recover more quickly than through austerity.
Meanwhile, the decreased income from stakers due to the burning of transaction fees (rather than using fees to pay stakers) also keeps yields depressed, further incentivizing network activity.
In DMs on Twitter, Nic (unprompted) walked back the pro-cyclical comment and acknowledged the complexity of the price vs network activity relationship, pointing out that there are both pro-cyclical and counter-cyclical aspects to the network upgrade. I agree with his more nuanced position that would have been impossible to properly communicate on the spot during a podcast.
Real vs Nominal Staking Yield
The conversation around real staking yield, as well as fully diluted valuations, has substantially evolved over the past few months, with yield farming coming under particular scrutiny. However, many alt-L1s continue to champion high staking yields, but offer little return once adjusted for inflation due to their high staking ratios.
In the figure below we see that the majority of the biggest proof-of-stake chains offer essentially no return for staking one’s tokens. For example, delegating tokens on Cardano is often seen as an easy way to generate nearly 4% income, but after adjusting for debasement and for fees, the yield actually drops negative. It’s a ridiculous situation: you’re actually expected to lose money, yet you’re still expected to pay taxes on the tokens you’re earning.
The key point to remember is:
If there’s no one left to extract yield from, then no one is getting yield.
Staking yield, framed as a liquidity tax, is only valuable if there are people who want liquidity.
The situation is currently much better for Ethereum, but that’s largely attributable to the low staking ratio and token burning. As the transition to proof-of-stake stabilizes and withdrawals are enabled (derisking staking) the staking ratio will rise dragging down the real staking yield.
Diving deeper into Ethereum’s yield structure, we can create a naïve yield diagram neglecting MEV and token burning to set a floor for yield and a roof for debasement. In the figure below, we see that although nominal yields remain above 2% until 60% of tokens are staked, real yields collapse below that mark once the staking ratio hits 30%—if 60% of ether becomes staked, baseline real yields fall to below 1%.
Staking ratios that high are simply not a good network dynamic. Locking tokens up and taking on slashing + liquidity risk for a 1% gain isn’t an efficient state for a blockchain designed to function as a world computer. The only people happy at these yields should be infinite duration speculators who want to safeguard their market share but don’t want to use the network.
We can also frame the diminishing returns of staking by calculating the total real yield extracted by validators. Once the validator count crosses 30% of the total ether supply, extracted yield begins to decrease and any additional validator additions begin to resemble pure inflation rather than yield extraction. Essentially the declining rewards from more ether being staked are larger than the increase in total real staking yield from the additional of a marginal validator. It would actually be better for the validator pool to pay the marginal validator their potential staking yield to not stake!
Extracted Real Yield = Real Yield × Staking Ratio
Creating a more realistic model that accounts for token burning and MEV in the current low gas environment, by a staking ratio of 30% the real yield is strikingly low at only about 3%. Most of the DCF or yield models that you see today claiming yields as high as 8% are only valid in the immediate short-term while the number of stakers remains near 10%. It’s not unless we get back into a high fee environment that higher staking ratios allow for meaningful real returns—yet a lot of this real return comes from token burning, which non-stakers would also benefit from.
In terms of real yield extraction, after accounting for MEV stakers run into the same maximum extraction barrier around a 30-35% staking ratio. MEV could also possibly push this boundary even lower since a higher staking ratio suggests a smaller amount of ether being actively transacted resulting in lower gas prices.
Note: many people in the community expect The Merge to drive a lot of demand into Ethereum, which could lead to an increase in on-chain activity and to higher burn rates and MEV. None of my models have ever relied upon or attempted to characterize a demand surge from The Merge.
Burn is neglected for the extracted real yield curve as its inclusion would make the relationship linearly increasing with the percent of supply current staked. This would be misleading because non-stakers also extract that supply deflation from the network. The relationship that is more interesting is the relative extraction compared to non-stakers, not the absolute value.
A subtlety worth addressing, that would at first appear in ways contradictory to my writing Proof-of-Stake and the Cantillon Effect, which encourages widespread staking, is that it’s not important for everyone to stake, but instead for everyone to have the option to stake. If a blockchain provides actual utility, one hopes that the alternatives available to staking are able to generate sufficient yield that you don’t end up at 70% or 80% of tokens stationary indefinitely.
Fetishizing the non-use of a cryptoasset, whether billed as a decentralized virtual machine or a p2p cash system, is the height of chasing financial performance over utility, and will lead to the eventual stagnation and death of communities that seek it. High staking yields encourage return chasing, and in my opinion, should be avoided.
Ironically, the tendency of bitcoin holders to tell people to never spend their coins fits extremely well with proof-of-stake: any staker by definition keeps their market share and will not be debased, therefore proof-of-stake allows a store-of-value cryptoasset to instantly be locked up forever, without having to worry about future issuance or potential endgame instability that could lead to changes in monetary policy.
Recommended Read:
The darker side of high yield protocols and dApps was covered well by Anthony Lee Zhang. In the context of this piece, if a blockchain were to increase staking rewards it forces dApps to increase their yield to remain competitive on a risk-adjusted basis. Paired with high growth aspirations and generally declining (not increasing like many tech companies) economies of scale for higher AUM this has the effect of forcing creators to either take on more risk or subsidize returns.
This is just one method to approximate the dynamic, I’m sure there are more intricate and better ways to model it.
From a fixed income standpoint, it would be more accurate to measure this as the expected cashflows during the staking period.
The convenience modifier could also in theory be negative if one is altruistically interested in supporting the network and willing to stake at an adjusted loss.
I haven’t spent time thinking about it, but there might be an interesting corollary to YCC and the current JGB situation + Yen implosion. I think the dynamic would be inverted though?
Is the takeaway essentially that if the primary use case for ETH is earning yield from staking, it is destined to die? Is there a timescale on which other usecases need to ramp up for the network to become sustainable?