Transport costs and new economic geography

This paper uncovers some little-known properties of the iceberg transport cost functions which are employed in some new economic geography models. In particular the behaviour of the delivered prices generated by the Krugman iceberg specification is very different to those generated by the original Samuelson iceberg model, and these differences require careful interpretation of the outcomes of the explicitly spatial versions of new economic geography models. If the iceberg model is viewed simply as transport costs, then these properties can be shown to be largely implausible when compared with a wide range of empirical evidence. At the same time, even if the iceberg model is viewed as capturing a range of different distance costs, there are still grounds for cautiousness in the way that inferences are made from these models when moving from theory to reality or policy analysis.