We calculate various indicators of the utilisation of preferences granted to developing countries by the EU and the US in the agricultural, food and fisheries sector. We conclude that only a very small proportion of the imports eligible to these preferences is actually exported outside a preferential regime. The rate of utilisation is therefore high. However, the flow of imports from poorest countries remains very limited in spite of rather generous tariff preferences, which leads to question the overall impact of the preferential agreements. In addition, preferential regimes overlap, and in such cases some regimes are systematically preferred to others. We use econometric estimates of the (latent) cost of using a given preference in order to explain why particular regimes are used. We focus on possible explanations, such as the cumulation rules (that restrict the use of materials originating from other countries), fixed administrative costs, and differences in the preferential margin.