This paper aims to investigate and model airport choice behaviour in a multi-airport region in order to analyze some facets of the phenomenon scantily treated in the literature: the effects of type of flight connection, trip duration and departure date. The analysis made use of discrete choice random utility models estimated on an specific stated preferences survey. Homoscedastic and heteroscedastic formulations were estimated, different correlation structures were tested, non-linear effects of level of service attributes were investigated, and sensitivity analyses were performed. All was carried out in a hitherto uninvestigated Italian multi-airport region (Campania, southern Italy). Major findings were: (i) access time, airfare, age, experience and income proved to be the most significant variables; (ii) non-linear transformation of access time and frequency (for direct flights) and of in-flight travel time (for connecting flights) appreciably improved models goodness-of-fit; (iii) correlation criteria based on geographical considerations or on operating airlines should be investigated; (iv) airport choices if only connecting flights are available cannot effectively be simulated through models estimated for direct flights; (v) airport choice probabilities may be significantly different as travel plans change; (vi) though Mixed Multinomial Logit and Cross-Nested Logit models statistically outperformed closed form formulations, Multinomial Logit model continue to be an effective modelling solution.
Modelling airport choice behaviour for direct flights, connecting flights and different travel plans
DE LUCA, STEFANO
2012-01-01
Abstract
This paper aims to investigate and model airport choice behaviour in a multi-airport region in order to analyze some facets of the phenomenon scantily treated in the literature: the effects of type of flight connection, trip duration and departure date. The analysis made use of discrete choice random utility models estimated on an specific stated preferences survey. Homoscedastic and heteroscedastic formulations were estimated, different correlation structures were tested, non-linear effects of level of service attributes were investigated, and sensitivity analyses were performed. All was carried out in a hitherto uninvestigated Italian multi-airport region (Campania, southern Italy). Major findings were: (i) access time, airfare, age, experience and income proved to be the most significant variables; (ii) non-linear transformation of access time and frequency (for direct flights) and of in-flight travel time (for connecting flights) appreciably improved models goodness-of-fit; (iii) correlation criteria based on geographical considerations or on operating airlines should be investigated; (iv) airport choices if only connecting flights are available cannot effectively be simulated through models estimated for direct flights; (v) airport choice probabilities may be significantly different as travel plans change; (vi) though Mixed Multinomial Logit and Cross-Nested Logit models statistically outperformed closed form formulations, Multinomial Logit model continue to be an effective modelling solution.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.