Modeling the competitive dynamic among air-travel itineraries with generalized extreme value models

This dissertation develops models that forecast air-travel itinerary passenger (market) shares between airport-pairs. This is the first study to model aviation demand at the itinerary level and air-carriers can use these models to assist their strategic decision-making. The motivation for developing these models is both to understand the impact of different air-carrier service attributes on itinerary share, and to understand the underlying competitive dynamic between itineraries.; Aggregate multinomial logit models with independent variables measuring air-carrier and itinerary service attributes such as level-of-service, connection quality, carrier attributes, aircraft size and type, and departure time are estimated. These model estimations describe the importance of these attributes on itinerary share. The results of these models are intuitive and offer new perspectives on the impacts of changing various service attributes on itinerary and carrier market share. Additionally, validation results from the implementation of these models by a major U.S. air-carrier are presented (the logit-based models improved the carrier's forecasting accuracy compared to its previous QSI-based model).; These multinomial logit models cannot measure the underlying competitive dynamic (if any) between itineraries due to constraints inherent in the model derivation. This competition is hypothesized to be differentiated by proximity in departure time, carrier and/or level-of-service. This hypothesis is tested by estimation of incrementally more complex generalized extreme value models. Results indicate that inter-itinerary competition is indeed differentiated by proximity in departure time, by carrier, by carrier within departure time periods and (to a lesser extent) by level-of-service within departure time periods. The advanced models estimated (extending to multi-level generalized nested logit and ordered generalized extreme value models) are shown to outperform the more basic specifications with regard to statistical tests and behavioral interpretations. Additionally, these models offer clear insights into air-traveler behavior and capture the underlying competitive dynamic among air-travel itineraries.