The Medicare Shared Savings Program (MSSP) is a voluntary program that provides incentive payments to healthcare providers that form integrated healthcare organizations, known as Accountable Care Organizations (ACOs). The MSSP rewards ACOs for keeping average per capita spending below a benchmark based on historical spending by ACO providers under the fee-for-service (FFS) system. I provide reduced-form and model-based evidence that benchmarks affect ACOs’ participation and performance in the MSSP. I show that blending the ACOs’ historical spending with a regional adjustment factor induces adverse selection, enabling some ACOs to earn incentive payments without reducing spending below the FFS level. Moreover, the benchmark rebasement, which updates the benchmark over time to reflect observed healthcare spending among ACOs’ providers, incentivizes ACOs to delay spending reductions to avoid lowering future benchmarks. I propose a revised benchmarking methodology to address these inefficiencies. Counterfactual analyses demonstrate that this alternative policy mitigates adverse selection and significantly increases Medicare savings.