When and where is cross-national diffusion an important determinant of policy innovation? I posit that the characteristics of a policy innovation-whether it imposes high or low "sunk" costs on adopters-and country attributes such as wealth, mediate the importance of diffusion in domestic policy choices. Competing risks analysis of two structural pension reform models in 71 developing and industrialized countries supports these hypotheses. Peer diffusion weighs heavily in the adoption of the costly "funded" defined-contribution pension reform model, and does so principally among middle-income nations, while the less-costly "notional" defined-contribution pension reform is not governed by diffusion.