Price Adjustment When Updating Requires Execution: Evidence from Automated Market Makers (Research Square)

Summary

• The study draws on a high-frequency analytic panel of 2.52 million block-level observations across two matched 5 bp ETH–USDC pools on Ethereum and Base.p.2
• After reference and finiteness screening, the analytic sample comprises 354,506 block-level observations for ETH5 and 2,164,574 for BASE5.p.4
• For ETH5, the pre-specified holdout diagnostic produces a slope change estimate β̂2 of -3.653 bp per unit x, with a 95% confidence interval of [-4.146, -3.026].p.9
• For BASE5, the holdout diagnostic slope change estimate β̂2 is -1.478 bp per unit x, with a 95% confidence interval of [-1.775, -1.107].p.9

Key insights

1. Passive concentrated-liquidity automated market maker pools mechanically eliminate non-execution price updates because native pool price changes only when a reserve-changing swap occurs.p.3
2. Raw aggregate correction increases in Confirmation relative to Discovery in both primary pools, whereas comparing the windows under a standardized common-state distribution yields a negative difference.p.4
3. The pre-specified fixed-hinge diagnostic yields a negative slope change estimate β̂2 across both primary pools, contradicting the post-discovery directional prediction of a positive value.p.9
4. Applying Discovery incidence and magnitude maps to Confirmation state frequencies overpredicts realized mean correction, resulting in negative transport errors in both primary pools.p.10
5. Symmetric Shapley allocations reveal that positive state-composition contributions (+1.256 bp for ETH5) are offset by negative conditional-magnitude contributions (-1.171 bp for ETH5).p.12
6. The qualitative pattern of positive composition and negative magnitude contributions persists when excluding the open-ended tail bin down to x ≤ 5, but is not established on x ≤ 2.p.12