PaperlayerView exact revision on arXiv

Source-linked AI summary

Empirical Analysis of EIP-1559: Transaction Fees, Waiting Time, and Consensus Security

Yulin Liu, Yuxuan Lu, Kartik Nayak, Fan Zhang, Luyao Zhang, Yinhong Zhao

arXiv:2201.05574v4econ.GNcs.CRcs.DCcs.GTq-fin.EC

TL;DR

Real-world evidence on transaction-fee mechanisms had been limited, despite their importance to blockchain incentives, user experience, and security. This study uses Ethereum blockchain, mempool, and exchange data with event-study and regression-discontinuity designs to evaluate EIP-1559. It reports lower intrablock gas-price volatility and waiting times, while emphasizing that a formal model explaining the waiting-time effect is absent.

  • Problem

    The real-world impact of EIP-1559 on fee dynamics, waiting times, and consensus security had not been systematically studied.

    Source: §1 · pp. 1–2

  • Method

    The authors combine Ethereum blockchain, mempool, and exchange data with event-study and regression-discontinuity designs around the London hard fork.

    Source: §4.3 · p. 6

  • Results

    EIP-1559 significantly reduces waiting times and latency, while intrablock gas-price volatility decreases as adoption of EIP-1559 transactions increases.

    Source: §6 · p. 13

  • Takeaways & Limitations

    Waiting times matter for user experience and blockchain security, but the literature lacks a formal model explaining EIP-1559’s waiting-time effect.

    Source: §6 · p. 13

    Exact transaction-submission times are difficult to obtain because they require monitoring traffic from all or most peer-to-peer nodes.

    Source: §4.3 · p. 5

Abstract

from arXiv · show

A transaction fee mechanism (TFM) is an essential component of a blockchain protocol. However, a systematic evaluation of the real-world impact of TFMs is still absent. Using rich data from the Ethereum blockchain, the mempool, and exchanges, we study the effect of EIP-1559, one of the earliest-deployed TFMs that depart from the traditional first-price auction paradigm. We conduct a rigorous and comprehensive empirical study to examine its causal effect on blockchain transaction fee dynamics, transaction waiting times, and consensus security. Our results show that EIP-1559 improves the user experience by mitigating intrablock differences in the gas price paid and reducing users' waiting times. However, EIP-1559 has only a small effect on gas fee levels and consensus security. In addition, we find that when Ether's price is more volatile, the waiting time is significantly higher. We also verify that a larger block size increases the presence of siblings. These findings suggest new directions for improving TFMs.

1 INTRODUCTION

Public blockchains must allocate scarce computation and storage, and Ethereum’s first-price-auction fees were replaced by EIP-1559 in 2021. Despite theoretical analysis and community expectations, its real-world effects had not been systematically studied; this paper evaluates fee dynamics, waiting times, and consensus security.

1 INTRODUCTION

  • Public blockchains have scarce computation and storage, so transaction fee mechanisms allocate those resources; Ethereum previously used first-price auctions.The paper frames fee mechanisms as affecting blockchain incentives, user experience, and security.Source: §1 · p. 1
  • EIP-1559 was presented as the first real attempt to depart from first-price auctions on a major blockchain.Earlier proposals for alternative mechanisms had not yet produced a real-world implementation, according to the paper.Source: §1 · p. 1
  • Ethereum activated EIP-1559 with the London hard fork on August 5, 2021, introducing a dynamically adjusted base fee while retaining backward compatibility.The base fee sets the minimum gas price for a block and adjusts according to gas used in the previous block.Source: §1 · p. 1
  • EIP-1559’s real-world impact had not been systematically studied, despite theoretical analyses and expectations that it could reduce fee volatility, overpayment, and waiting times.Prior work examined incentive compatibility, fee volatility, and possible mechanism modifications, but did not provide a systematic real-world evaluation.Source: §1 · p. 1
  • The study asks whether EIP-1559 changes transaction-fee dynamics, waiting times, and consensus security.The authors note limited empirical evidence on users’ rationality, how delays might change, and security implications of changed block sizes and incentives.Source: §1 · p. 2

Challenges and our approach

The study combines blockchain, mempool, and exchange data to examine EIP-1559’s effects. It uses distributed mempool monitoring to measure transaction waiting times and event-study and regression-discontinuity designs to estimate effects around the London hard fork.

Challenges and our approach

  • Blockchain, mempool, and exchange data provide measures of transaction activity, waiting time, and intraday Ether prices.Measuring waiting time requires accurate mempool observations.Source: Challenges and our approach · p. 2
  • A distributed system records when transactions enter Ethereum’s mempool, providing a more precise waiting-time measure than prior studies used.The authors describe this as a more precise measure, not an exact observation of every transaction’s submission time.Source: Challenges and our approach · p. 2
  • Event-study and regression-discontinuity designs compare observations around the London hard fork to estimate EIP-1559’s local average treatment effect.The framework is intended to control for confounders such as price volatility, network instability, and time trends.Source: Challenges and our approach · p. 2

Our Findings

The paper reports that EIP-1559 made fee estimation easier and reduced waiting times, without lowering transaction-fee levels in the observed period. Its consensus-related findings include higher fork rates, no significant sustained increase in network load, and a larger MEV share of miner revenue.

Our Findings

  • EIP-1559 did not lower transaction-fee levels in the observed period, but its bidding style made fee estimation easier and reduced intrablock gas-price variance.The paper reports that users adopting EIP-1559 bidding paid less than legacy bidders and that variance fell as adoption increased.Source: Our Findings · p. 2
  • The paper defines transaction waiting time as the interval from first mempool observation to mining, and notes that dependent transactions can make delay costly.Users may be unable to submit a new dependent transaction until the earlier one is included in a block or canceled.Source: Our Findings · p. 2
  • Transaction waiting times declined after the London hard fork, possibly because of easier bidding and variable-sized blocks.The reported improvement applied to both transactions using the new bid style and those still using legacy bidding.Source: Our Findings · p. 2
  • The London hard fork increased average block size and raised fork rates by approximately 3%.The paper identifies larger blocks’ propagation time as a possible reason for more forks.Source: Our Findings · p. 2
  • EIP-1559 did not significantly increase network load for an extended period, while MEV became a larger share of miner revenue, mainly because base fees are burned.The study also reports no significant difference in load-spike frequency before and after the fork; the larger MEV share might incentivize more MEV extraction.Source: Our Findings · p. 3

2 RELATED WORKS

Related work spans transaction-fee mechanism design, waiting-time modeling, and consensus security. The paper situates its empirical study among theoretical mechanism proposals, approaches to measuring blockchain delay, and research on forks and MEV.

2.1 Transaction Fee Mechanism Design

  • Transaction-fee mechanism research analyzes first-price auctions and proposes alternatives, including second-price-style mechanisms, StableFees, and posted-price designs.Prior EIP-1559 work examined incentive compatibility, fee and waiting-time properties, system stability, and concerns about block-size oscillations and fee estimation [55] [75].Source: §2.1 · p. 3

2.2 Waiting Time Modeling in Market Design

  • Waiting-time research connects congestion and delay to user welfare, while direct measurement in Ethereum has been limited by mempool-observation challenges.Prior work used external Bitcoin data or congestion proxies; Azevedo Sousa et al. directly observed Ethereum’s mempool but encountered negative waiting times associated with network latency.Source: §2.2 · p. 3

2.3 Consensus Security

  • Prior blockchain-security research examines incentive incompatibilities and attacks across proof-of-work and proof-of-stake systems.The literature also considers how larger blocks can lengthen propagation and increase uncle blocks, potentially weakening resilience to double-spend attacks and selfish mining.Source: §2.3 · pp. 3–4
  • MEV research studies extraction across blockchain infrastructures, including Flashbots’ rapid growth and other extraction tools.Flashbots made MEV easier to extract and observe; its adoption was reported above 95% when the paper was written.Source: §2.3 · p. 4

3 BACKGROUND

Ethereum transaction fees price scarce computation and storage, while the legacy first-price auction required users to guess competitors’ bids. EIP-1559 changes block capacity, fee adjustment, bidding, and miner compensation.

3.1 Transaction Fees in Ethereum

  • Ethereum fees are GasUsed × GasPrice, and the block gas limit rose from 15 million to 30 million after EIP-1559.Users may set a gas limit; unused gas is refunded.Source: §3.1 · p. 4

3.2 Pre-EIP-1559 Transaction Fee Mechanism

  • Legacy first-price auctions reward the highest bids but lack a dominant strategy, making bid optimization impractical and contributing to overpaid, volatile fees and long waits.Source: §3.2 · p. 4

3.3 The new TFM in EIP-1559

  • EIP-1559 doubled the block gas limit to 30 million while retaining a 15-million target, and its base fee adjusts with prior-block gas use.The base fee is the minimum gas price required for inclusion; it rises above target usage and falls below it.Source: §3.3 · p. 4
  • EIP-1559 lets users specify a priority-fee tip and a maximum fee; the gas price is the lesser of base fee plus tip or the fee cap.Legacy-style gas-price bids remain allowed.Source: §3.3 · pp. 4–5
  • EIP-1559 burns the base fee and pays priority fees to miners, replacing their pre-upgrade receipt of all transaction gas fees.Miner revenue also includes static rewards, uncle rewards, and MEV.Source: §3.3 · p. 5

4 DATA

The study combines blockchain, mempool, price, and miner-revenue data to measure fees, waiting times, and forks around EIP-1559. It describes measurement choices and preliminary patterns in base fees, supply, and adoption.

4 DATA

  • The study combines Ethereum chain data, four geographically distributed mempool probes, minute-level Bloomberg ETH prices, and Flashbots revenue; final data records are published on Dataverse.Source: §4.1 · p. 5, §4 · p. 5

4.1 Data Sources and Metadata

  • The pre-upgrade sample covers blocks 12,895,000–12,965,000 and the post-upgrade sample covers 13,035,000–13,105,000, excluding the adoption transition.Source: §4.1 · p. 5

4.2 Transaction Fee Data

  • Block-level fee measures use the median gas price and standardized IQR to capture representative prices and intrablock variation robustly to outliers.Transaction and block data include gas prices, EIP-1559 fee bids, base fees, gas use, and block size.Source: §4.2 · p. 5

4.3 Waiting Time Data

  • Four geographically distributed Geth probes estimate submission time from the earliest mempool observation; negative waits were set to zero (0.4% pre, 0.3% post, 0.8% afterwards).Source: §4.3 · p. 6

4.4 Miners’ Revenue Data

  • Miner-revenue estimates combine observable rewards and fees with Flashbots revenue as a proxy for MEV, divided into five revenue categories.Source: §4.4 · p. 6

4.5 Fork Rate Data

  • Sibling counts derived from uncle-block pointers indicate how many competing blocks appeared at a given height.Source: §4.5 · p. 6

4.6 Preliminary Visualizations

  • After London, the base fee oscillated between 30 and 200 Gwei and Ether supply was sometimes negative; EIP-1559 adoption rose sharply near block 13.05 million.Source: §4.6 · pp. 6–7

5 EMPIRICAL RESULTS

This section examines EIP-1559’s effects on transaction fees, waiting times, consensus security, network conditions, and miner revenue. It reports lower intrablock fee dispersion and waiting times, little overall change in fee levels or consensus security, more sibling blocks with larger blocks, and post-London changes in MEV revenue share and non-FBB gas-fee distribution, while short-term MEV revenue distribution and measured network-load indicators were not significantly affected.

5.1 Methodology

  • To isolate EIP-1559’s effects from confounders, the study combines an event-study framework with regression discontinuity, using gradual adoption after the London hard fork.Controls include time trend, price volatility, median gas price, ROI, block size, and hour fixed effects.Source: §5.1 · pp. 7–8

5.2.1 Overview.

  • EIP-1559 adopters paid a 45 Gwei median block gas price versus 54 Gwei for legacy users, while overall fee levels changed little and priority bids remained low.Max-fee bids were usually higher than prices paid; the authors say EIP-1559 made fee estimation easier, and the predicted priority-fee auction degradation did not occur much in practice.Source: §5.2.1 · pp. 8–9, Figure 6 · p. 9

5.2.2 Intrablock interquartile range.

  • At full EIP-1559 adoption, standardized intrablock IQR is predicted to fall 24 percentage points, from 0.34 to 0.10, despite an immediate 8-point increase.The authors conclude that intrablock gas-price inequality decreases as adoption grows, and that concerns about base-fee volatility making fee estimation more difficult do not hold in practice.Source: §5.2.2 · p. 9

5.3 Waiting Time

  • After the London hard fork, median block waiting-time quartiles fell from 10.7 to 5.5 seconds, 16.9 to 10.4 seconds, and 34.0 to 18.6 seconds.The authors report that waiting times fell for both legacy and EIP-1559 transactions; higher 90-block Ether-price volatility was associated with longer waits.Source: §5.3 · pp. 9–10, Figure 7 · p. 10

5.4.1 Fork rates.

  • Block size rose from 64.05 kbytes before the London hard fork to 78.01 kbytes after it, and larger blocks significantly increased the likelihood of siblings.The authors conclude EIP-1559 may raise sibling likelihood mainly through block size; adoption itself had a significantly negative association with sibling appearance after controlling for size.Source: §5.4.1 · pp. 10–11, Table 3 · p. 11

5.4.2 Network load.

  • EIP-1559 did not significantly affect integral network load or load-spike frequency, especially for spikes lasting 40 seconds or three blocks or more.The authors calculate average gas consumed per second over 20–120-second intervals and compare load distributions before and after the London hard fork.Source: §5.4.2 · p. 11, Figure 8 · p. 11

5.4.3 Miner Extractable Value (MEV).

  • Miner revenue from MEV dipped temporarily after the London hard fork, recovered within 50,000 blocks, and eventually became a larger share of miner revenue.Overall miner revenue decreased after the upgrade, primarily because base fees are burned.Source: §5.4.3 · p. 12, Figure 9 · p. 12
  • Approximately 4% of total revenue and 30% of nonstatic revenue came from MEV after the London hard fork, up from approximately 3% and 15%, respectively, before it.Source: §5.4.3 · p. 12
  • The distribution of revenue from Flashbots bundles changed little between the periods before and after the London hard fork.Source: §5.4.3 · p. 12
  • EIP-1559 did not change the short-term distribution of MEV revenue, but significantly changed non-FBB gas-fee revenue, the main reason the nonstatic-revenue distribution changed.Source: §5.4.3 · p. 12, Figure 10 · p. 12

6 CONCLUSION AND DISCUSSION

The study finds that EIP-1559 adoption reduces intrablock gas-price volatility and waiting times, while price volatility is associated with higher fees and waits. It also identifies a negligible EIP-1559 effect on the presence of siblings and describes broader applications for its empirical methods.

6 CONCLUSION AND DISCUSSION

  • Intrablock gas-price volatility decreases significantly as more users adopt EIP-1559 transactions, and the authors report that priority-fee degradation to a first-price auction rarely occurs.The reduced volatility implies easier fee estimation and a better user experience; users’ bidding patterns are consistent with the obvious optimal bid described by Roughgarden (2020).Source: §6 · pp. 12–13
  • Waiting times and latency decline significantly after EIP-1559, while higher Ether-price volatility is associated with higher transaction fees and waiting times.The authors note that formal modeling of the waiting-time effect is absent and suggest future work on queuing factors and layer-1 TFM design.Source: §6 · p. 13
  • Larger block size increases the presence of siblings, but the effect size through EIP-1559 is negligible.Source: §6 · p. 13
  • The empirical method could be applied to study EIP-1559 or other Ethereum Improvement Proposals across decentralization, network features, attacks, dark-venue adoption, and financial inclusion.The authors leave these applications for future research.Source: §6 · p. 13

A DATA DICTIONARY

The appendix points readers to a data dictionary covering the datasets and variables used in the study.

A DATA DICTIONARY

B STATISTICS TESTS

The appendix reports checks on control-variable correlations, variable stationarity, and autocorrelation in the regression outcomes.

B STATISTICS TESTS

  • Figure 11 presents a triangle correlation heatmap for the control variables used in the analysis.Source: Correlation Tests · p. 15, Figure 11 · p. 15
  • The Augmented Dickey-Fuller test indicates that all variables used are stationary.Source: Stationary Tests · p. 15
  • Gas prices have significant autocorrelation, whereas standardized IQR and waiting time have trivial autocorrelation.Figure 12 presents the autocorrelation tests for the regression dependent variables.Source: Auto-correlation Tests · p. 15

C ROBUSTNESS CHECKS

Robustness checks report that fee-level conclusions depend on controls, while intrablock price-variance results persist across specifications and volatility measures. Waiting-time results are similar under LAD regression, and transaction type has trivial effects on median waiting time.

C ROBUSTNESS CHECKS

  • Median gas prices increased after the London hard fork without controls but dropped when controlling for time trend and price volatility.Source: Fee Dynamics · p. 15
  • With controls, London hard fork coefficients for standardized IQR are significantly positive, while EIP-1559 adoption coefficients are significantly negative.Table 8 reports that IQR rose immediately after the fork but fell significantly more in blocks with greater EIP-1559 bidding adoption; Table 9 reports robustness to different volatility measures.Source: Fee Dynamics · p. 15
  • Least absolute deviation regression yields waiting-time results similar to OLS, with coefficient estimates smaller overall.Source: Fee Dynamics · p. 17
  • Transaction type has trivial effects on median waiting time.Source: Fee Dynamics · p. 17

TABLES AND VISUALIZATIONS

The visualized regressions associate greater EIP-1559 adoption with lower waiting times and smaller intrablock gas-price differences. Gas-price-level estimates are mixed, while priority-fee bids cluster around two peaks.

TABLES AND VISUALIZATIONS

  • Priority-fee bids have peaks at 2 and 5 Gwei, with a majority of median bids below 10 Gwei.Figure 15 shows the priority-fee distribution.Source: TABLES AND VISUALIZATIONS · p. 17, Figure 15 · p. 17
  • Controlling for other factors in a representative scenario (median gas price 39 Gwei; volatility 0.005), predicted mean waiting time falls from 19.93 to 8.79 seconds as EIP-1559 adoption rises from 0 to 1, versus approximately 28.5 seconds before the London hard fork.Separately, Figure 18 reports that the waiting-time result remains robust when the analysis time range is extended.Source: TABLES AND VISUALIZATIONS · p. 17, Figure 18 · p. 19
  • 0.34 to 0.09: predicted standardized interquartile difference falls as adoption rises from 0 to 1, outweighing a small short-term increase after the hard fork.The IQR time series oscillates with the base fee; the standardized-IQR regression is robust to different volatility measures.Source: TABLES AND VISUALIZATIONS · pp. 17–19, Figure 14 · p. 17, Figure 16 · p. 19
  • Median gas-price estimates are mixed: prices rose after the hard fork without controls but fell when time trend and price volatility were controlled.The cited table reports median gas price as the dependent variable.Source: TABLES AND VISUALIZATIONS · p. 19, Table 6 · p. 19
  • LAD waiting-time regressions produce results similar to OLS, but their coefficient estimates are smaller overall.Table 10 reports the LAD regression.Source: TABLES AND VISUALIZATIONS · p. 20, Table 10 · p. 20
  • Transaction type has trivial effects on median waiting time, and legacy and EIP-1559 waiting-time distributions are generally similar.Figure 17 compares the two transaction types after the hard fork.Source: Figure 17 · p. 18, TABLES AND VISUALIZATIONS · pp. 19–20, Table 11 · p. 20

Related papers

Loading 2201.05574v4…