In the neoclassical growth model of Barro et al. [Am. Econ. Rev. 85 (1) (1995) 103-115], partial capital mobility across economies generates implausibly large growth effects under a standard parameterization of preferences and technology. Reasonable growth effects only occur if substantially less than the share of physical capital in factor income can serve as collateral for external borrowing. This finding confines the empirical relevance of the open-economy neoclassical growth model to the case of international capital flows, where market imperfections are likely to prevail. But for partial capital mobility across economies such as US states, where market imperfections appear less relevant, the model cannot produce plausible long-run growth effects.