Centralization in Attester-Proposer Separation
Mallesh Pai, Max Resnick
TL;DR
This paper studies Attester-Proposer Separation through Execution Auctions and Execution Tickets and demonstrates that ahead-of-time auctions intensify centralization among block proposers even without multi-block MEV. It employs a two-period, incomplete-information model where valuations originate from distributions with non-decreasing hazard rates, and the Period 1 auction winner can resale in Period 2 using Myersonian optimal mechanisms. The main finding is a unique subgame-perfect equilibrium in which the ex-ante strongest builder wins the execution auction and secures the right to propose or to resell at a markup, leading to a larger share of blocks than under MEV-Boost and thus increasing centralization. Numerical examples corroborate the centralization effect, showing substantial shifts in proposer shares when the ex-ante leader wins, and the paper discusses significant implications for Ethereum governance and competition, proposing alternatives such as Multiple Concurrent Proposers and Inclusion Lists to mitigate these effects. Overall, the work highlights a fundamental tension between revenue-optimizing proposer-utility arrangements and healthy, competitive builder participation in the network.
Abstract
We show that Execution Tickets and Execution Auctions dramatically increase centralization in the market for block proposals, even without multi-block MEV concerns. Previous analyses have insufficiently or incorrectly modeled the interaction between ahead-of-time auctions and just-in-time (JIT) auctions. We study a model where bidders compete in an execution auction ahead of time, and then the winner holds a JIT auction to resell the proposal rights when the slot arrives. During the execution auction, bidders only know the distribution of their valuations. Bidders then draw values from their distributions and compete in the JIT auction. We show that a bidder who wins the execution auction is substantially advantaged in the JIT auction since they can set a reserve price higher than their own realized value for the slot to increase their revenue. As a result, there is a strong centralizing force in the execution auction, which allows the ex-ante strongest bidder to win the execution auction every time, and similarly gives them the strongest incentive to buy up all the tickets. Similar results trivially apply if the resale market is imperfect, since that only reinforces the advantages of the ex-ante strong buyer. To reiterate, these results do not require the bidders to employ multi-block MEV strategies, although if they did, it would likely amplify the centralizing effects.