There is a thriving debate on the usefulness of economic models for policy analysis which this paper explores. It argues that there is a need for models that incorporate both the modern intertemporal approach to macroeconomics and short-run ad-hoc behaviour. This need cannot be met by the simple models that permeate the macroeconomics literature, but requires large-scale simulation models such as the MSG2 and G-Cubed multi-country models. In particular, it is shown that the approach of these models gives insights into the adjustment into a number of historical episodes that are not well explained by simple macroeconomic frameworks: Reaganomics in the 1980s; German reunification in the early 1990s; fiscal consolidation in Europe in the 1990s; the formation of NAFTA; the Asian Crisis; and the current productivity boom in the USA. The paper also argues that using a well-defined theoretical specification but introducing real-world rigidities can also go some way towards explaining the so-called "six major puzzles in international macroeconomics" highlighted in a recent paper by Obstfeld and Rogoff (2000). Copyright 2000 by Oxford University Press.