To allocate a fleet of vehicles to a given number of locationsand to redistribute free vehicles are important control problems.In this paper we use the results for multi-location inventorymodels to develop an approximate solution for the combinedfleet sizing and redistribution problem. For a defined rewardstructure we investigate:a) some properties of the optimal redistribution and allocationdecision respectively;b) the concavity of the expected one-period reward;c) the advantages gained from co-operation of locations overindependent locations.Since a model with discrete time is used, the proposed solutionis an approximation. The two-location model is investigatedin more detail.