CGE modelling has dominated analysis of the impact of external liberalisation on poverty. This article provides a structuralist critique of standard neo-classical CGE models. It highlights five sets of gaps and partial achievements in the modelling of issues affecting the poverty impact of macroeconomic policies: duality and structural rigidities; efficiency gains and quota rents; the investment and savings specification; the nature of public expenditures; and the modelling of financial fragility, risk premia and issues of credibility. It outlines a model that makes it possible to analyse more plausible stories about the impact of both current and capital account liberalisation and questions the realism of existing approaches to ex-ante poverty impact assessment.