Modelling the asymmetric probabilistic delay of aircraft arrival

The main purpose of this paper is to present an asymmetric logit probability model to estimate and predict the daily probabilities of delay in aircraft arrivals. The proposed model takes into account statistical regularity, noting that more arrivals are on time than delayed, thus reflecting an asymmetric pattern of behaviour. The data analysed were obtained from the BTS and IATA databases for December 2014, corresponding to delays within the US airspace system for each carrier, measured at various US airports. The model was evaluated by analysing both in-sample and out-of-sample data, for main and control samples. The performance of the proposed asymmetric Bayesian logit model was compared with that of two others: frequentist logit and symmetric Bayesian logit. The main conclusion drawn is that the model we propose obtains the best fit, according to the statistics considered, and identifies a novel delaying factor, namely distance, which is not identified by the other models analysed.