Do Layer-2s Decongest Ethereum? Regime-Aware Causal Evidence from 2021-2024

Ethereum’s rollup-centric roadmap assumes that adoption of layer-2 (L2) rollups will relieve congestion on the layer-1 (L1) mainnet, but there is little causal evidence on how large that relief is across major protocol changes. We construct a daily panel for Ethereum from August 2021 to December 2024 and estimate the total effect of L2 adoption on L1 congestion using a regime-aware interrupted time-series design that spans the London, Merge, and Dencun upgrades and adjusts for macro demand. In the pre-Dencun regime, a 10 percentage point increase in L2 adoption reduces median L1 base fees by about 11%—roughly 4–5 Gwei per 21k-gas transfer—and yields similar declines in a harmonized congestion index, with only modest changes in block utilization.

Combining these elasticities with a Merge-era counterfactual path for L2 adoption, we estimate that observed adoption avoided about $80–$90 million in base and priority fees over 137 days, or roughly $0.6–$0.7 million per day. These effects are precise while L2 adoption is still ramping up but become statistically local once adoption exceeds 85% after Dencun, implying that congestion relief is economically meaningful yet regime-specific. Our findings support continued investment in L2 infrastructure alongside L1 mechanism design and illustrate how to benchmark congestion relief in other multi-layer digital platforms.