Causal inference from a single treated state: a peer review of Lee et al. (2024) and the exchange that followed

Abstract

Lee et al. (2024) report that state-level anti-transgender laws caused increases of 7–72% in past-year suicide attempts among transgender and non-binary (TGNB) young people. A Matters Arising (Cohn et al., 2026) showed that the paper’s largest estimates derive from a single treated state, Idaho; the authors’ reply (Lee et al., 2026) defended the design, and the format permits no further response. This peer review shows that the reply’s principal statistical defenses are incorrect; that the single-treated-cluster setting renders the paper’s reported uncertainty invalid rather than merely fragile, and that the decisive analysis — randomization inference with a correction for group-size heteroskedasticity — was never conducted; and that several design problems independent of the single-state issue remain unaddressed by the main study or the follow-ups. The remedies are specified and are implementable, quite possibly with the data the authors possess.