International Choice Modelling Conference, International Choice Modelling Conference 2015

Font Size: 
A Novel Dual-Choice Experiment in the Evaluation of Air Traveler Itinerary and Seat Preferences
Jeff Keller, Daniel Weinstein

Last modified: 11 May 2015


International choice modelling conference 2015

Austin, Texas - USA, May 10-12, 2015

A Novel Dual-Choice Experiment in the Evaluation of Air Traveler Itinerary and Seat Preferences

Jeff S. Keller1 and Daniel J. Weinstein1

1 RSG, USA. (E-mail:,


There has been growing interest in a deeper understanding of preferences of air travelers from both the airline and aircraft manufacturing industries in recent time. Space aboard aircraft is limited and subject to numerous engineering constraints resulting in tradeoffs between configurations of fewer luxurious seats at a price premium and more standard seats at an economical price. Meanwhile, air travelers themselves have varying preferences and are faced with tradeoffs between comfort, cost, and any number of itinerary or seat choice-related features. Airlines and manufacturers seek to meet the needs of this heterogeneous population while simultaneously improving their financial prospects.

The basis for this work is data collected as part of an online stated choice survey for a major player in the air travel industry. Some of the specific attributes and levels tested in the stated choice experiments may be masked for confidentiality and proprietary reasons.

The sample is composed of over 2,000 air travelers across three countries. Respondents of the survey were tasked with reporting a recent air trip which was to be used as a point of reference for a series of stated choice experiments. Details of this reference trip were collected, including airline, origin/destination, ticket fare, and number of connections. These data were used to construct realistic flight itinerary options by varying the itinerary attributes around the reference trip. Furthermore, secondary seat choice experiments were constructed to follow each itinerary choice. There, respondents were tasked with choosing seats on the flights associated with their chosen itinerary, subject to availability constraints, different types of seating configurations, and prices. The itinerary and seat options were presented in a fashion to mimic the actual purchase experience of an online travel booking website.

An efficient design of 144 experiments grouped into twelve blocks was used in combination with a respondent's reference flight details to construct the twelve itinerary choice experiments. Each choice was made between three itinerary alternatives. The details of subsequent seat choice experiments were determined by the chosen itinerary in conjunction with a random design indicating seat availability. Groups of seats were randomly selected to receive an additional upcharge as a function of the ticket fare.

The seat choice experiments made use of realistic and interactive graphical depictions of the aircraft interior (Figure 1). This allowed for the concise conveyance of potentially dozens of cabin layouts. A graphical preview of the cabin details was dynamically made available as respondents reviewed the available options. This additional visual aid provided the necessary context for making decisions that are hypothesized to be at least partially influenced by the physical proximity of seats to other seats or to other areas of the aircraft (e.g., exits, lavatories). Respondents had the choice of choosing seats for each individual in their travel party or indicating that they would choose another flight instead.

Figure 1: Seat Choice Exercise

The primary objective of this work is to identify the effects of aircraft cabin features on itinerary choice and seat choice. To accurately quantify these effects it is important to capture as many other effects as possible, such as flight length, flight purpose, party size, and possible interactions of these with the cabin details - the result being a holistic and realistic model of itinerary and seat choice behavior. The inclusion of price in both choices allows for the quantification of marginal rates of substitution for itinerary and cabin details and price (willingness-to-pay). This is a convenient measure for quantifying the value that air travelers place on these items.

Though data collection is still underway, the data will be modeled under the multinomial logit or nested logit frameworks. A variety of model specifications will be tested to identify non-linearities, interactions, and other systematic sources of preference heterogeneity. For instance, the value of a certain level of comfort may diminish as flight lengths decrease, or rise dramatically for overnight flights. Testing combinations of nesting structures will identify seat or itinerary types that are more substitutable. Finally, mixed logit and/or Hierarchical Bayesian methods may be employed to capture any remaining unexplained heterogeneity.

Keywords: air travel, flight choice, seat choice

Acknowledgment: The authors would like to express appreciation for the support of the aircraft industry sponsor, for their permission to use the itinerary and seat choice data for this work, and to our unnamed collaborating author.

Conference registration is required in order to view papers.