A comparison of flight routes in a dual-airport region using overlapping error components and a cross-nested structure in GEV models. (January 2017)

• Record Type: Journal Article
• Authors: Yang, Chih-Wen; Wang, Hsiao-Chun
• Is Part Of: Transportation research. Volume 95(2017)
• Page Start: 85
• Page End: 95
• Publication Date: 2017-01
• Subjects: Flying-route choice -- Cross grouping -- Cross nested logit -- Error components logit -- Model validation; Transportation -- Research -- Periodicals; 388.011
• DOI: 10.1016/j.tra.2016.11.006
• Languages: English
• ISSNs: 0965-8564

Abstract

Highlights: Investigating the similarity among flight routes with GNL and ECL models in a dual-airport region. Model comparisons with three criteria of interpretation, reproduction, and generalization. ECL with overlapping EC can be effective in dealing with cross correlations among alternatives. The flying route with two international airports could suffer competition from multiple sources. Airfare and access time are proven as significant variables to increase the share of flying routes. Abstract: This study investigates flying-route choice behavior with cross-grouping relationships. When there are different combinations of origin and destination airports for a city pair, it is important to identify the cross relations among the various flying routes. Here, this study does so using the allocation parameter of the cross-nested logit model and the overlapping error components of the error components logit model. A survey based on the stated preference method was designed to collect travelers' preferences in order to examine the distribution of choices for Taiwan–Shanghai direct flights. The empirical results reveal that airfare and access time are significant variables, and could be used as operating strategies to increase the market share of flying routes. The flying route between the two international airports existed simultaneously within two groups, namely the leading group and the PVG group. Furthermore, comparing the models based on the criteria of interpretation, reproduction, and generalization indicated that the error components logit model can be effective in dealing with cross correlations among alternatives.