This guide, which uses examples from recent reforms for adult Medicaid beneficiaries, is intended to support demonstration states by describing best practices in causal inference. In this context, “causal inference” is the process of determining whether a demonstration policy (also called the treatment) is responsible for an observed outcome. Establishing an association between treatment and outcome variables is relatively straightforward, requiring only that they move reliably in the same or opposite directions. Establishing causation—that is, confidence that a change in treatment caused observed changes in outcomes—is much more difficult. Yet a primary goal of demonstration evaluations is to determine whether particular state Medicaid policies cause changes in outcomes such as health care access, utilization, and costs, and—in the case of some eligibility and coverage policies—the uptake of commercial coverage.