Points clés :

• When national planners decide non-cooperatively, the spatial distribution of transport infrastructure is inefficiently biased towards central regions of adjacent countries.
• This bias is magnified by discrete border frictions caused by cultural or institutional differences or tariffs.
• Calibrating the model to European data, the mechanism explains about 20% of the so called border effect, i.e., the observation that intranational trade strongly dominates international one.
• Corroborating the main implications of the model, regression analysis shows that the estimated border effect shrinks when accounting for transport infrastructure.

 Résumé :
The distribution of transport infrastructure across space is the outcome of deliberate government planning that reflects a desire to unlock the welfare gains from regional economic integration. Yet, despite being one of the oldest government activities, the economic forces shaping the endogenous emergence of infrastructure have not been rigorously studied. This paper provides a stylized analytical framework of open economies in which planners decide non-cooperatively on transport infrastructure investments across continuous space. Allowing for intra- and international trade, the resulting equilibrium investment schedule features underinvestment that turns out particularly severe in border regions and that is amplified by the presence of discrete border costs. In European data, the mechanism explains about a fifth of the border effect identified in a conventionally specified gravity regression. The framework sheds light on the welfare costs of second best investment schedules, on the effects of intercontinental trade or of privatized infrastructure provision.

 Mots-clés : Economic Geography | International Trade | Infrastructure Investment | Border Effect Puzzle

 JEL : F11, R42, R13