The task of airline network management is to develop new flight schedule variants and evaluate them in terms of expected passenger demand and revenue. Given the industry's trend towards global cooperation, this is especially important when evaluating the potential synergies with alliance partners. From the econometric point of view, this task represents a discrete choice modelling problem in which one has to account for a large number of dependent alternatives. In this paper we discuss the applicability of recently proposed approaches and introduce a new multinomial probit specification designed for the airline network management task. The superior performance of the new model is demonstrated in a real-world application using airline bookings data.
