Anonymous. (2026). Individual-level randomization does not systematically underestimate population-level effects when treatment and control groups share network connections, as spillovers would contaminate the control group and bias results toward null. Isonomia. https://www.isonomia.app/a/rPNTHCC5@7ca2e4cc1ab5c8610311887666e7febe6becea2d2d096f2c95b9cd92b0f954db