Experiments with SP and CA Approaches to Mode Choice

Experiments with SP and CA Approaches to Mode Choice


AXHAUSEN K, University of Innsbruck, and Koell H, Ingenieurbfiro Koell and BADER M, Ampass, Austria


Stated Preference-based survey techniques have become an accepted part of the transport planning tool kit. in spite of their known limitations and problems. In general these surveys are implemented as Stated ('hoice-experiments. in which respondents are a


Stated Preference-based survey techniques have become an accepted part of the transport planning tool kit. in spite of their known limitations and problems. In general these surveys are implemented as Stated ('hoice-experiments. in which respondents are asked to choose between the two or more alternatives described to them. The analysis of Stated-Preference data is decompositionai in the sense, that one derives the part-worths of the different variables deseribing the alternatives from the one, joint judgement'0fthe alternative as a whole (A was chosen, B not; A was ranked 5th and B 10th; A received a scale value of 6 out of 10). Logit or related utility-maximising models are used to derive those part- worths in the case of choice data.

The class of survey approaches based on such hypothetical markets/goods is caned Conjoint Analysis in marketing. The approaches used. for example the very popular ACA-So~ware (Adaptive Conjoint Analysis) (Sawtooth. 1996). are hybrid s of compositional and deeompositional analysis methods. Compositional utility, estimation is based O n the isolated, independent ranking of the different variables for their importance and of their levels for desirability with th e utility of a composite good calculated as the importance-weighted sum of the desirabilities.

Conjoint Analysis (CA) and Stated Preferences (SP) belong therefore to the same general class of techniques, where the respondents are offered hypothetical goods, mostly in the form of a written description, either individually or in sets, which they are asked to rate. or rank or choose between (Hensher. 1994 or Axhausen, 1996). Both approaehes are in the general tradition of the social sciences. in particular psychology and microeconomics (Green and Rao. 1971: Louviere. Meyer. Stetzer and Beavers. 1971 etc.). They know that the true complexity of the decision is larger then the level presented ' to the respondent in the survey. This simplification is accepted as the price to obtain results of predictive value for those factors under, the control of the authority or firm undertaking the study within an acceptable time frame (see Broeg, 1997. for an opposing view).

The Stated Preference or hybrid Conjoint-Analysis approach encompasses a wide variety of specific methodologies, which at! share the aim of

* obtaining holistic statements of preference in a specified format

* for a series of (hypothetical) goods described by varying levels of a small number of attributes

* within a specified behavioural frame (overall context)

The decision problem is reduced for the respondent, as he has only the given response format, but he has both to imagine the stimuli and to adjust to the specified behavioural frame. The analyst looses the full behavioural context detail, but can focus on those aspens under management control. The control over the description of the hypothetical goods permits, in addition, the generation of well behaved statistical data for the estimation of appropriate decision models.

The development of Conjoint Analysis is well documented in a series of scholarly reviews in the marketing research literature (Green and Srinlvasan. 1978: Bocker. 1986: Huber. 1987; Louviere. 1988). The growing usage of the methodology in marketing over the last two decades is equally well documented by three surveys of market research firms (Cattin and Wittink- 1982: Wittink and Cattirk 1989 and Wittink. Vriens and Burhenne. 1994). The development of the methodology in transport plamaing can only be reconstructed from a series of How-to-manuals. which have been published over the years (Kocur. Adler. Hyman and Aunet. 1982: Pearmain. Swanson. Kroes and Bradley. 1991: Axhausen. 1996 or Pearmain. Swanson and Ampt (forthcoming). but see also Bates. 1988 or Hensher. 1994). The usage of the methodology, in transport planning has not been surveyed yet.

The current return of SP to the US. the coming together of professional market researchers and transport.planners in the commercialized punic transport firms (bus. rail and air) and the growing interest in market research in choice-based conjoint formats (Louviere and Woodworth, 1983 or Sawtooth. 1995) opens up new opportunities for the further development of both methodologies. The purpose of this study is to compare two particular SP and CA methods with regards to the comparability of their results and with regards to the response behaviour of the respondents using the same issue, mode choice behaviour in the City of Innsbruck. reflecting the interests of the co-funders of the study.

The remainder of the paper is structured as follows. The next section describes the two approaches implemented in the survey, which are then described in detail. The survey administration and the response behaviour is analyzed in the following section, while the resuhs from the modelling of the responses is the topic of the final substantive section. A summary and a discussion of further work concludes the paper.


Association for European Transport