Designing Stated Choice Experiments

Abstract

Since major airports are reaching capacity, some solutions have to be found to solve the actual transportation problem. The government sees the development of secondary airports as a solution to this problem. First, this would result in air traffic diversion and secondly this would permit new economical opportunities to the concerned region. However, airlines and air passengers are the airport users, and in order to make secondary airports economically profitable, the attractiveness of these airports plays an important role. Therefore, getting more insight in the effects of airport location on air passengers' choice would be helpful for the decision process of implementing or not a secondary airport. Additionally, investigating air travellers' preferences may assist airports and airlines to adjust their strategy on air travellers' needs in order to gain a larger market share. In order to achieve this objective, using discrete choice modelling and more specifically stated choice (SC) experiments, is a proven way for estimating and forecasting travellers' behaviour. Since the reliability of the estimated parameters can be relying on the estimating models used or on the underlying experimental designs, more and more researches realise that so-called efficient design are able to produce more efficient data in the sense that more reliable parameter estimates can be obtained. Because this new knowledge has never been applied on practical cases, the aim of this study is to perform a SC experiment and to reach the following objectives: (i) determining which factors are influencing air passengers' choice in selecting flight itineraries, and (ii) proving that efficient designs result in more reliable parameter estimates than orthogonal designs. A SC experiment is defined as presenting a sample of respondents with a number of hypothetical scenarios, consisting of universal but finite number of alternatives that differ on a number of attribute dimensions. In this research the alternatives consist of trips from Amsterdam Airport Schiphol (AAS) to Barcelona city. Since the city of Barcelona has two airports, namely Barcelona airport and the further away located Girona airport, travellers may take into account the airport location in their itinerary choice. In each choice situation, respondents will have to choose one alternative between 5 proposed. Each alternative is composed of six attributes: airline, price, transfer time, departure time, egress time, and egress time. Generally, the chosen alternative is the one that maximizes the traveller's utility. In total each respondent will face 6 different choice situations. Obviously, there are many ways of assigning combinations of attribute levels to the attributes in each situation. The purpose of the underlying experimental design is to determine these combinations in the best possible way. The different methods for constructing experimental designs yield the generation of three designs, namely an orthogonal design with 108 choice situations, an efficient design with 18 choice situations, and another efficient design with 108 choice situations. Collecting data can be done by several ways, such as pen & paper surveys, CAPI surveys, or internet surveys. In this research, an internet survey has been performed because of its advantages compared to other methods, looking similarly to the real website of an online travel agent (e.g. ebookers). TeamVier, a Dutch market research company, provided a heterogeneous sample of respondents for this study. In the end, the collected data consist in 3,300 observations collected from 550 respondents. The collected data has been estimated with two different estimation models. The first model is the well-known multinomial logit (MNL) model and the second model is the more advanced panel mixed logit (ML) model. The results of both models yield better model fit of the panel ML model regarding to the MNL model, which leads in better estimations of the parameter estimates. With the outcomes of this model, statements can be done regrading to travellers' behaviour in itinerary choices. The outcomes of the analyses have shown that price and transfer time are the most important attributes influencing travellers' itinerary choice. The willingness to pay of travellers to avoid one hour of transfer time has been estimated on approximately €37. Additionally, the airport location plays an important role in travellers' itinerary choice. Due to the less egress time, travellers have a preference for arriving at the main airport instead of at the further away located secondary airport. In this study case, the egress time from Girona airport is about 40 to 60 minutes longer than from Barcelona airport, therefore people are willing to pay about €20 to travel to Barcelona instead of Girona. When comparing the data coming from the different experimental designs, the outcomes of the analyses show that efficient designs result in more reliable parameter estimates than orthogonal designs. Additionally, the results have shown that efficient designs permit the reduction of required sample size to produce a fixed level of reliability in the parameter estimates. This leads to the fact that, designs with a limited number of choice situations can be as efficient as larger designs. From now on, it is recommended to construct an efficient experimental design instead of an orthogonal design as it leads to more reliable parameter estimates and necessities a lower sample size, which is the most expensive part when performing a SC experiment. The statements on travellers' preferences in itinerary choice can be used by airports and airlines to adapt their strategies on the consumer needs. Although travellers have a preference for arriving at the main airport, this study has shown that the disadvantages of the location of secondary airports can easily be compensate with better price offers, which make them even more attractive than main airports. Developing regional airports could be a good alternative to solve the capacity problems of main airports. However, because this study focussed on the effects of the arriving airport location, further research is required on travellers having business proposes or on the influence of access time and access prices on local travellers' behaviour.