Statistical Modelling 24 (1) (2024), 82–106
Interpretable modelling of retail demand and price
elasticity for passenger flights using booking data
Jan Felix Meyer,
Department of Statistics,
Ludwigs-Maximilian-Universität München,
Munich,
Germany.
Göran Kauermann,
Department of Statistics,
Ludwigs-Maximilian-Universität Mü,nchen,
Munich,
Germany.
e-mail: goeran.kauermann@stat.uni-muenchen.de
Michael Stanley Smith,
Melbourne Business School,
University of Melbourne,
Australia.
Abstract:
We propose a model of retail demand for air travel and ticket price elasticity at the daily booking and individual flight level. Daily bookings are modelled as a non-homogeneous Poisson process with respect to the time to departure. The booking intensity is a function of booking and flight level covariates, including non-linear effects modelled semi-parametrically using penalized splines. Customer heterogeneity is incorporated using a finite mixture model, where the latent segments have covariate-dependent probabilities. We fit the model to a unique dataset of over one million daily counts of bookings for 9 602 scheduled flights on a short-haul route over two years. A control variate approach with a strong instrument corrects for a substantial level of price endogeneity. A rich latent segmentation is uncovered, along with strong covariate effects. The calibrated model can be used to quantify demand and price elasticity for different flights booked on different days prior to departure and is a step towards continuous pricing; something that is a major objective of airlines. As our model is interpretable, forecasts can be created under different scenarios. For instance, while our model is calibrated on data collected prior to COVID-19, many of the empirical insights are likely to remain valid as air travel recovers in the post-COVID-19 period.
Keywords:
flight price elasticity,
mixed non-homogeneous Poisson model,
P-splines,
price endogeneity.
Downloads:
Supplementary material in PDF
back