Escaping the Monopolist: An Economic Model of Censorship Resistance in Multi-Proposer Protocols
Session summary
The presenter introduces a formal model for studying censorship resistance in multi‑proposer blockchain protocols with finite block capacity. The core metric, termed “social loss,” measures how much consumer surplus is lost for a chosen set of target users, taking into account their values, the protocol outcome, and the presence of a censor under strategic behavior assumptions. Social loss decomposes into a censorship premium, capturing additional harm from an active censor, and an intrinsic inefficiency, reflecting unavoidable loss from throughput limits even without censorship. This avoids classifying obviously bad mechanisms, such as always‑empty blocks, as desirable simply because they treat all users equally poorly.
Within a three‑stage full‑information game between users, a single‑minded censor, and block proposers, the authors compare fee‑distribution rules. In a single‑proposer winner‑pays‑bid benchmark, they identify censor value thresholds below which the censor is ineffective and above which users cannot avoid exclusion. Extending to multiple proposers, they find that an Equishare rule, which splits fees evenly across all proposers, can worsen censorship resistance by making it cheaper to bribe proposers. A Shapley‑style mechanism, which only pays proposers who actually include a transaction, matches the benchmark’s intrinsic inefficiency but, in the authors’ analysis, offers a more favorable censorship premium that improves with the number of proposers.
Key takeaways
- The authors propose a “social loss” metric to assess censorship resistance under finite capacity, focusing on consumer surplus for a specified target set of users.
- Social loss is decomposed into censorship premium and intrinsic inefficiency, preventing degenerate mechanisms like always empty blocks from appearing attractive.
- In this model, naive multi‑proposer fee splitting (Equishare) can increase social loss relative to a single‑proposer auction, while a Shapley‑inspired distribution preserves baseline inefficiency and improves censorship premium as the number of proposers grows.