The aim of the article is to analyse a couple of arguments for decentralisation of the common agricultural policy (CAP) in an abstract economic model. The CAP in the European Union has changed its focus towards supporting the multifunctional character of European agriculture. From an economic-theoretical point of view, this policy change will have to be founded in supporting positive external effects, but we show in a theoretical model that this is not straightforward. If the policy covers more than one country, the domestic value attached to an external effect can be different from one country to another. For landscape preservation it can be optimal to subsidise agriculture in countries where arable land is scarce and to tax agriculture (or support nature) in countries with less nature. In general, the most efficient level of policy provision depends on the nature of the external effect. If there are considerable differences in the evaluation of a certain multifunctional policy between countries, it can be more efficient to administer and implement the policy at a decentralised level of government where regional differentiation is possible.