Research
When Staking Rewards Compound: Measuring the Impact of Ethereum's Pectra Upgrade
Ethereum's beacon chain carries over 920,000 active validators, a count inflated by the legacy 32 ETH per-validator stake cap. The Pectra upgrade (May 2025) introduces 0x02 withdrawal credentials…

- arXiv
- 2606.23337
- Published
- 2026-06-22
- Authors
- Mohammed Benseddik, Benjamin Kraner, Claudio J. Tessone
AI summary
Overview
Research area: Blockchain economics and proof-of-stake consensus — specifically the incentive design of Ethereum validator rewards, withdrawal credentials, and validator consolidation following the Pectra upgrade (EIP-7251).
Technical level: Intermediate. The paper uses formal notation for effective-balance hysteresis, reward decomposition, and APR, and assumes the reader knows what validators, epochs, attestations, and withdrawal credentials are.
Scope: A combined Monte Carlo simulation and on-chain empirical study (913,225 active validators, snapshot 7 April 2026) of whether Ethereum's new 0x02 compounding validators actually raise staking returns, and how that effect differs between solo stakers and large staking providers.
What This Paper Is About
Ethereum's beacon chain carries over 920,000 active validators, a count inflated by the legacy 32 ETH per-validator stake cap. The Pectra upgrade (May 2025) introduces 0x02 withdrawal credentials, which raise the maximum effective balance per validator from 32 to 2,048 ETH and let consensus-layer rewards compound automatically inside the beacon balance instead of being swept out. The paper asks whether this compounding genuinely improves returns, how the benefit scales with stake size, why adoption has reached only 25% of staked ETH (9.7M ETH) eleven months after the upgrade, and which staker classes gain or lose.
Key Contributions
-
A formal model of 0x02 compounding under hysteresis. The paper specifies how actual balance (
B_act) converts into effective balance (B_eff) through the protocol's hysteresis band (downward threshold 0.25 ETH, upward threshold 1.25 ETH, in 1 ETH increments), and shows that compounding only boosts rewards onceB_actcrosses the upward threshold. -
Monte Carlo quantification of the compounding advantage by stake size. Simulations over a 1-year horizon across balances from 32 ETH to 10,240 ETH (200 runs per balance point, 96–99% participation) yield a CL APR uplift ranging from +4.7% for 32–2,048 ETH down to +0.3% for 8,192–10,240 ETH.
-
An empirical decomposition of where 0x02 stake comes from. The authors separate new deposits from migration via consolidation and show that consolidation contributes only 1.91M ETH (19.8%) of the 9.67M ETH 0x02 total, while new direct deposits contribute 7.76M ETH (80.2%) — a factor of four.
-
A measurement of realised APR by credential type and staker category. Using 0x01/0x02 subsets across two windows (30-day and 335-day, the latter covering the exact post-Pectra era from 7 May 2025 to 7 Apr 2026), the paper shows the apparent 0x02 advantage collapses from +12.3% at 30 days to +1.5% at 335 days, and attributes the short-window gap to proposer-reward variance rather than compounding.
Main Findings
-
Compounding gives a small, bounded consensus-layer uplift. The 2-year head-to-head simulation (one 0x02 validator versus two 0x01 validators, 64.5 ETH initial balance, 730 days, 99% participation) produced 2.2134% APR for 0x02 versus 2.1950% for 0x01 in Year 1 — an APR difference of 0.0184%, roughly 2 bps, or a 1.89% relative reward advantage. Final balances were 67.3848 ETH versus 67.3314 ETH, with final effective balances of 67 ETH versus 64 ETH.
-
The benefit is concentrated in small balances. Simulated CL APR uplift falls monotonically by balance bucket: +4.7% (32–2,048 ETH), +0.9% (2,048–4,096 ETH), +0.5% (4,096–6,144 ETH), +0.4% (6,144–8,192 ETH), and +0.3% (8,192–10,240 ETH). The mechanism is remainder ETH: 0x01 setups leave idle ETH below the 32 ETH activation threshold, whereas large providers can spread capital across many validators.
-
Adoption is 25% of staked ETH but only 1.3% of validators. Of 38.8M ETH staked across 922,721 active validators (7 April 2026), 9,496 validators (1.0%) remain on legacy 0x00, 901,650 (97.7%) on 0x01, and 11,575 (1.3%) on 0x02. Because each 0x02 validator can hold up to 2,048 ETH, a small validator count carries a large stake share.
-
Consolidation activity is highly non-uniform and recent. Daily consolidation volume averaged 186 requests per day, but roughly half of the 60,405 total requests occurred in a single month (March 2026), with a peak of 3,343 requests in mid-March and the three busiest days (3,343, 3,062, 2,971) all in that month. The cumulative total through December 2025 was 28,195. Multi-validator consolidation dominates at 89.5% (54,049 requests); self-consolidation is 10.5% (6,356 requests).
-
Migration is overwhelmingly new capital. Consolidation flows break down into 197k ETH from 6,165 self-consolidations and 1.71M ETH from 53,561 multi-validator merges into 2,150 targets, totalling 1.91M ETH (anchored to 59,726 unique source validators × 32 ETH). New direct deposits of 7.76M ETH across 3,260 validators make up the remaining 80.2%.
-
Solo stakers adopt more, relative to their stake. Solo stakers hold 6.37M ETH of 0x02 stake (16.4% of the 38.8M ETH beacon stake), while staking providers hold 3.29M ETH (8.5%), summing to the 25% total.
-
The 0x02 effective-balance distribution is bimodal. Approximately 30% of 0x02 validators cluster at 32–64 ETH (recent credential upgrades without merging), while 40% sit above 1,024 ETH (aggressive consolidation). Solo stakers show a mean balance of 842 ETH and providers 815 ETH, indicating two distinct migration strategies rather than one.
-
Empirically, 0x02 lead is modest at long horizons. Over 30 days, aggregate median CL APR was 2.52% ± 0.92 for 0x02 versus 2.25% ± 1.09 for 0x01 (+12.3%), covering 10,763 validators and 8.65M ETH. Over 335 days, medians were 2.65% ± 0.22 versus 2.61% ± 0.27 (+1.5%), covering 11,395 validators and 9.37M ETH.
-
The short-window "advantage" is proposer luck, not compounding. Solo stakers showed +13.9% at 30 days (2.561% vs 2.248%) while staking providers sat at parity (+1.1%; 2.274% vs 2.249%). The solo cohort's median effective balance was 480 ETH, giving roughly 15× the proposer selection probability of a 32 ETH validator. Because that lead collapses to +1.5% at 335 days, the authors conclude it is variance that averages out rather than a compounding gap that would persist or grow.
-
Reward distributions are heavy-tailed. The raw 0x01 distribution has excess kurtosis of +10.9, an order of magnitude above a Normal, driven by proposer-reward and sync-committee excursions. An "average" 0x01 validator earns approximately 2.61% when annualised over 365 days, against an attestation baseline of about 2.24% per year at 38.8M ETH staked.
-
Execution-layer rewards confer no credential advantage. Proposer selection probability is proportional to effective balance (
B_eff,v / B_total), so a single 0x02 validator with 2,048 ETH has the same expected proposals as 64 0x01 validators at 32 ETH each. EL rewards also do not compound: they are paid to the withdrawal address. -
Operational frictions cut against providers. Under 0x01, excess balance above 32 ETH is swept automatically at no cost; under 0x02, accessing compounded rewards requires explicit partial withdrawal requests that incur gas fees and queue delays. Queue data show peak entry waits of 45 days (June 2023) and an exit-queue peak of 46 days (September 2025), when Kiln withdrew 1.6M ETH.
Methodology in Plain English
The authors work on two fronts. First, they build a simulation. They model how a validator's actual balance grows from consensus-layer rewards, how the protocol's hysteresis rule converts that into effective balance, and how effective balance in turn drives rewards (rewards scale proportionally with effective balance). They run a direct 2-year head-to-head — one compounding 0x02 validator against two non-compounding 0x01 validators holding the same 64.5 ETH — and then a Monte Carlo sweep over balances from 32 to 10,240 ETH with 200 runs per balance point, holding total network stake fixed at 38.8M ETH and excluding execution-layer priority fees as a conservative baseline. Crucially, they compute APR against total portfolio ETH (staked balance plus swept rewards), so 0x01 APR erodes as idle ETH piles up waiting to reach 32 ETH.
Second, they use on-chain data. Dune Analytics supplies staking metrics and the community-maintained validator labels that split validators into "Staking Providers" (any validator appearing in a labelled entity set such as Lido, Coinbase, Binance, Kraken, Rocket Pool, or EigenLayer operators) and "Solo Stakers" (everything unlabelled — a deliberately conservative definition that absorbs untagged professional operators). Beaconcha.in supplies per-validator reward accumulators for 1-, 7-, 31- and 365-day rolling windows, and Etherscan supplies execution-layer reward attribution. They use a 30-day window from the 31-day accumulator and a 335-day window spanning the exact post-Pectra era, annualising the 365-day accumulator as (cl_365d / B_eff) × 365 / min(d, 365) with a minimum of 7 days of history, winsorised at 5/95 percentiles within each group. Comparisons use median APR with IQR-based outlier removal and a formal Mann-Whitney test on the long-horizon cohort.
Why This Matters
Impact on research. The paper supplies a clean template for evaluating protocol upgrades that change incentive structure rather than consensus rules: simulate the mechanism, then test whether the predicted effect survives contact with on-chain data at horizons long enough to average out heavy-tailed reward luck. Its central methodological point — that a 30-day window inflates the 0x02 advantage to +12.3% while a 335-day window reduces it to +1.5% — is a cautionary result for anyone measuring validator performance on short windows.
Real-world applications:
- Solo staker decisions: quantifying whether migrating to 0x02 is worth the operational effort at a given balance, given the +4.7% simulated CL APR uplift in the 32–2,048 ETH range.
- Staking provider operations: informing whether to consolidate validator fleets for key-management savings when the APR gain is only +0.3% at 8,192–10,240 ETH and withdrawal requests incur gas and queue delays.
- Protocol governance: the finding that adoption is driven 80.2% by new deposits rather than migration of existing 0x01 stake is direct evidence for debates about whether reward accessibility needs adjustment (the paper points to EIP-8068 on hysteresis thresholds and the proposed 0x03 credential in EIP-7804).
- Staking yield analytics: the bimodal 0x02 effective-balance distribution (30% at 32–64 ETH, 40% above 1,024 ETH) gives analysts a way to distinguish credential upgrades from genuine consolidation when interpreting fleet restructuring.
Industry relevance. With 38.8M ETH staked (32% of a 121.6M supply) and 75% of staked ETH still on 0x01 eleven months post-upgrade, the paper documents a live, incomplete migration. Its conclusion is that without improved reward accessibility and stronger economic incentives, 0x02 migration will remain gradual despite the network efficiency benefits of fewer validators.
Future Directions
-
Re-run the empirical comparison at longer horizons. The 335-day window is the longest the data allow because Pectra is only about 11 months old at the snapshot. A genuine compounding gap would persist or grow with time, so a multi-year window would separate compounding from proposer variance more decisively.
-
Evaluate the proposed hysteresis revision. The paper notes that EIP-8068 proposes revising the 0.25 ETH and 1.25 ETH thresholds to address yield disparities; the same simulation framework could test whether relaxing hysteresis materially changes the +4.7%-to-+0.3% uplift curve.
-
Model the operational cost side explicitly. The simulations assume zero gas fees, no queue delays, and 96–99% participation. Since the central trade-off for providers is gas cost plus withdrawal delay against a marginal APR gain, incorporating these frictions would establish whether the +0.3% bucket is net positive or negative.
-
Track the March 2026 consolidation burst. Roughly half of all 60,405 consolidation requests occurred in a single month, and the migration component of 0x02 stake was nearly flat from mid-2025 through February 2026 before that uptick. Whether this is a one-off fleet restructuring or the start of a durable migration trend is unresolved, as is the effect of the proposed 0x03 credential type.
Target Audience
Protocol researchers and mechanism designers working on proof-of-stake incentive structures; quantitative analysts and data scientists at staking providers, liquid staking protocols, and exchanges who need to model validator economics; Ethereum core developers evaluating EIP-7251, EIP-8068, and 0x03; and academic researchers in blockchain economics interested in a worked example of combining simulation with on-chain causal inference. Readers should be comfortable with validator terminology and percentage-based yield metrics; the mathematics is presented but not the main barrier.
Authors’ abstract
Ethereum's beacon chain hosts over 920,000 active validators, a number inflated by the legacy 32 ETH stake cap. The Pectra upgrade (May 2025) addresses this by introducing 0x02 compounding validators, raising the maximum stake per validator from 32 to 2,048 ETH and enabling automatic reward reinvestment. This paper examines how compounding affects consensus-layer rewards, whether higher balances provide execution-layer advantages, and whether the APR uplift justifies…
The full abstract on arXiv