Airline overbooking models with misspecification

Abstract

This paper looks at static overbooking models. In theory, a random show demand follows a binomial distribution with each reservation showing up independently and with the same probability. However, in practice, some overbooking models assume that the show demand is the product of the overbooking level and the random show-up rate. The decision model embedded in a commercial revenue management system is misspecified. We explore the consequences of the modeling error and find that the performance of the model with misspecification decreases as show-up probability decreases. Among our three choices of show-up rate distributions, normal, beta and deterministic, the beta model performs best. We also identify situations in which an airline might prefer the deterministic model to the normal model.