Part 2 of a 3 part blog post series on the Ethereum Merge. In this post, we dive deep into validator rewards, as evidenced on-chain in the 90 days around The Merge and supplemented by off-chain sources such as data from MEV relays.
Arguably the most important feature of the Merge is the fact that now validators get to process transactions by producing blocks from the execution layer of Ethereum and get rewarded for it. But exactly how much so would that be?
Daily (total) rewards for validators have increased by ~40% post-Merge
Priority fees on the network have increased by 8%, while at the same time transaction volume is down.
Home stakers (10 or less validators) are lagging behind pools and professional operators in terms of APR%, due to their slower uptake of mev-boost. That said, the average home staker did better than Vitalik in terms of APR%.
Using a daily average of all the rewards emitted to and fees collected by validators, from August 1, 2022 until October 30, 2022, we observed that average daily (total) rewards for validators have increased by 40% to 2,285 ETH since the Merge.
These priority fees and baseline maximal extractable value (MEV) have comprised around 27% of the validator rewards on average, reaching a share as high as 43% on days of high network activity.
Now, accurately tracking what validators earn in post-Merge Ethereum is not as trivial as one might think–especially as MEV Relays have introduced a variety of transaction patterns that might constitute payments from block builders, to block producers. We have come up with what we think is a robust methodology to separate baseline MEV from the bulk of execution layer rewards that validators receive—you can read more about it here.
On the whole, the ratio of CL-to-EL rewards in the post-Merge period we looked into racks up to approximately 3:1. Consensus rewards (net of penalties) have held mostly steady pre- and post- Merge showing little to no variance throughout, as one would expect.
Conversely, Execution rewards have been a lot more varied. Specifically looking at priority fees, which go to validators post-Merge—as opposed to miners pre-Merge, these have increased by around 8% so far compared to the previous period with the average now at 506 ETH per day. The variance in the daily priority fee print has also increased by 18% post-Merge compared to the period from August 1, 2022 until the Merge.
The increase in variance observed rhymes well with an increase in variance we have observed both in terms of basefee and in terms of transaction count on the Execution layer of the network. Looking at the time-series on both variables it is clear that patters and variance have significantly changed post-Merge. What’s worth noting here is the nominal trends on the overall period seem to be equally unchanged and relatively flat.
We believe that the most likely cause for the increase in variance across all EL rewards variables, is the introduction and rapid adoption of mev-boost, which quickly grew to 20% of mainnet blocks and has climbed all the way to 70%+ as of more recently.
Having kept track of trends on mev-boost blocks since the Merge, it has been abundantly clear that such blocks both pack significantly more transactions than locally built blocks (we also refer to those as “vanilla blocks”), and as a consequence of fuller blocks higher base fee and priority fee prints.
We’re diving deeper into MEV Relays post-Merge in part 3 of this series.
Adopting the same methodology as with part 1 of the series, when surveying rewards trends in post-Merge Ethereum, we grouped the validator set in 5 different cohorts in terms of size of the associated operation. The cohorts are shaped as follows:
To make the analysis of rewards easier to track and more relatable, we also translated rewards earned on the network in terms of Annual Percentage Return (APR%). You can dig deeper into the methodology that powers the APR% calculation via our documentation.
What we found is that gross APR% increased linearly with cohort size for the 3 smallest cohorts and then saturated on the last two cohorts. Worth noting that largely the behaviour the last cohort exhibited was due to Coinbase’s slow adoption of mev-boost compared to its comparables. Zooming in, operators with 10 or less validators mapped to them (home stakers) have exhibited the lowest returns in terms of both execution and consensus layer rewards, differing by a significant 40% in execution layer APR% compared to the highest cohort. To put things into perspective, this is just a difference of 0.74 in percentage points.
We believe the table above explains the majority of the variability between cohorts, with the sources tracking back to the differing mev-boost adoption rates, and (to a lesser extent) the allotment of proposal duties. The latter should track 1:1 with network penetration over longer time periods, but in the smaller sample we are working with here we can expect larger cohorts to map closer to their long-term average than smaller ones. Indeed, we have found that the ~17% of operators that belong in cohorts with 100 validators or less, have had materially smaller representation in the proposed block mass that makes up their sample than their overall network penetration (4% and 4.4% lower respectively); we think that this is in all likelihood the source of the delta they are exhibiting in terms of CL APR%, with the remainder attributable to differences in validator effectiveness (rewards and penalties accrued). Equivalently the different levels of mev-boost adoption explain the majority of the EV APR% variability.
In terms of the intra-cohort variance of rewards, by far the most volatile cohort in terms of execution layer rewards are the entities with 101 to 1,000 validators. This group has shown 3x more variance against all the other cohorts. October 22 had a spike in execution layer rewards due to block 15802413 that had 337 ETH go to validator 400385, which is over 2,000 times higher than the average for that day. This validator is part of an entity under the 101 to 1,000 cohort and the block was procured via mev-boost.
Changing the frame of reference from “cohort” to “identified entity” and more specifically in “staking pools”, we find that the trends observed (predictably) persist–the adoption of mev-boost and larger network penetration increase the gross APR% that a validator might enjoy on a relative basis.
At a glance, it looks like joining a staking pool does indeed lead to higher rewards, with all but one pool getting a higher APR than the average home staker. The difference can be as high as 47%, with the average being at 19%.
Note: the data we present here does not take into consideration the fees that some of these pools charge. In some cases these can be as high as 25% of the rewards earned, and in all likelihood tilt the change the results substantially, in certain cases.
That said, there are some really interesting outliers in the set, with Coinbase–most notably–raking in a lower EL APR% (0.78%) than the average home staker (1.08%); this is in direct correspondence with the uptake of mev-boost for the two groups. Over the sample, Coinbase produced only 17% of their blocks via mev-boost, while the home staker cohort did ~27% on average.
The relationship between APR% and mev-boost adoption becomes abundantly clear, when plotting out APR% and mev-boost penetration against each other.
The influence of stake-mass and and mev-boost penetration on overall rewards, looks similar at an operator level. The top ranking operators in terms of APR% were those that have been aggressive with their adoption of mev-boost from the earliest stages of post-Merge Ethereum.
Zooming out a bit further, it seems that the top 25 operators in terms of overall APR % are beating out the two smallest cohorts in terms of stake mass. However, there are cases in which home stakers have beat out some operators in terms of APR%, including client teams like Nimbus, Prysm and Nethermind, Lido-whitelisted operators like Consensys, and even Vitalik himself.
While pooling stake does seem to have an effect in terms of the level of rewards validators get, it looks like running mev-boost has a stronger relationship with overall validator returns. The linear relationship here is equally strong to the “pools” cohort, with one caveat; we have observed two distinct groups/tracks which are practically 2.9% in APR% apart. Upon closer inspection we realized that the lower APR% operator group is the “unlucky” group (annotated by the red dotted line above). We found that what’s distinctive about this group of operators, is that they have all been awarded significantly less proposal slots than their stake penetration would have one expect. Do note that these are not missed proposals, but rather proposals that statistically should have happened but never did.
In Part 3 of our Merge series, we will tackle what the increasing adoption of mev-boost means for the network and how post-Merge MEV and out-of-protocol proposer-builder-separation are developing. Stay tuned!