Abstract:

In this paper， the dynamic pricing problem of multi-class considering passengers’ choice behavior is studied by empirical research. The passengers’ choice behavior is modelled using a Multinomial Logit model. This passengers’ choice model characterizes observable attributes of passengers， including price， the number of tickets， transfer limits， refund limits and advance booking days. Then Markov Chain Monte Carlo method is used to estimate the parameters of passengers’ choice model and passengers’ arrival rate based on real operational data of airline industry. On this basis， a new dynamic pricing model applying the arrival rate and the parameters of passengers’ choice model is proposed to present the optimal price decision making process. The choice model estimates are then used to access the revenue performance of the current pricing strategy by the airline relative to pricing strategy optimized to account for passenger’s choice behavior. Empirical results show that， the optimal price decision according to the available seats of each class in each time period can be obtained. When the decision time is approaching the departure date， the optimal price of each class will increase accordingly. The price of each class will increase with the reduction of available seats at the same decision period. Our results show 22.32% average revenue improvements using the pricing strategy considering choice behavior. Especially， expected revenue of the flights with low revenue in actual will significantly increase by the pricing strategy considering passengers’ choice behavior. Overall， it is suggested that pricing strategy based on passengers’ choice behavior is both feasible to execute and economically significant in real-world airline environments.

Key words: passengers’ choice behavior, airline revenue management, dynamic pricing, multinomial logit model