 |  |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
2005 Annual Forum
View Abstract
A Qualitative Choice Model for the US Airline Network
(Formerly - Network Evolution: A Markov Chain Analysis of US Air Transportation)
Dipasis Bhadra, Brendan Hogan, Center for Advanced System Development (CAASD), The MITRE Corporation
What are the factors that determine the types of network? Is there any pattern to how the air traffic network is set up with respect to the size of the markets, types of competition, geospatial features, and other structural factors? Have the networks changed empirically meaningfully following the events of 9/11? Are these changes transitory or permanent in nature? Can we lay out the trajectory possibilities of the network and determine factors influencing them?
The National Airspace System (NAS) in the United States is structured primarily around a web of air transportation markets linked to each other through a network of 465 commercial airports located in and around 363 metropolitan statistical areas (MSAs). The total number of origin-destination (O&D) markets in the NAS ranges somewhere between 36,000 ? 40,000 pairs depending upon seasons and economic cycles. In its present structure, these markets are hierarchical; a small number of markets account for the largest number of passengers and, hence, air traffic flows. For example, there were approximately 105 markets (0.3% of the total) which had 1,000 or more passengers a day (i.e., thick markets), but these accounted for almost 17% of the total passengers. On the other hand, there were almost 28,000 markets (78% of the total) with 10 or fewer passengers a day that accounted for only 6% of the total passengers in 2003. These O&D market pairs have been served, generally speaking, by 52,000-56,000 segments depending upon the extent and intensity of network. In recent years, however, the network segments have increased sizably to an average of 67,000-72,000 segments leading to increased fragmentation [Bhadra and Texter (2004)].
Using a multi-nomial logit model, we attempt to answer the following two questions: What are the fundamental factors that determine and drive evolutions in the US airline network? Second, how can the changes in the network - either temporary or permanent and empirically meaningful ? be adequately explained? Answering these questions may provide us with some important insight into understanding the formation and evolution of the US airline network. Furthermore, this may lead us to improved policy-making in the context of the industry that appears to be ever changing.
Data for this exercise comes from the Bureau of Transportation Statistics/Department of Transportation's (BTS/DOT) Origin (O) and Destination (D) survey (DB1B). DB1B is a 10% sample of airline tickets from reporting carriers. It includes such items as the reporting carrier, number of passengers, ticket fare, and total miles flown for each itinerary, as well as information about whether the itinerary was domestic and round-trip. This file is related to both the O&D segment and market files by the unique Itinerary ID on each record [see http://www.transtats.bts.gov/ for more details]. For demonstrating empirical analysis reported in this paper, we use data for the second quarter of 2003. During this period, there were 493,100 unique itineraries that were reported under 10% sample. This was reduced to 359,837 observations when summed over by unique itineraries. When these itineraries were aggregated over distinct O&D pair markets, there were 68,442 observations. Using this dataset primarily, we seek to answer the above-mentioned empirical questions.
Related Links
- Download Presentation (PDF, 3877K)
- View the 2005 proceedings