Rejoinder for
Current issues and a 'wish list' for conjoint analysis
The detailed comments provided by Louviere, Magidson and Vermunt, Orme, and Swait, have brought to light a number of issues regarding my 'Wish List': (1) Some of my Wish-List issues have been (partially) addressed, but are not known to me, (2) Thought leaders in the area do not agree on a number of these issues, and (3) Should we try to address these issues under the current conjoint paradigm or do more radical changes need to be made (see Louviere comment)? Each of these issues though has, in my view, a very fundamentally different 'cause'.
The fact that a number of these issues have been addressed, unbeknownst to me, could of course reflect my deficiencies; but rather than admit that, I would rather attribute it to the vastness of literatures, just in the reviewers comments that are represented. Consider just the transportation literature, environmental economics literature, marketing literature, and organizational behaviour literature listed in Swait's comments, the sociology literature mentioned in Magidson and Vermunt, the statistics literature (and others) mentioned in Louviere, and the Advanced Research Techniques (ART) forum practitioner literature given in Orme. Could I, or anyone, even if I hoped to stay well-read, expect to see all of these different papers? How many of the comment writers are aware of many of these works? Thus, while conjoint analysis has benefited greatly from its widespread use, I believe that it has suffered academically as a theoretical research area (as per Louviere and Swait) because of the 'disjoint' variety of literatures in which its basic fundamental research has been published.
As an example of the difference of opinions, consider on the one hand the comments made by Louviere in that preference partworths are relatively stable (rather it is an error variance issue) with those by me (citing current research), Magidson and Vermunt, and Orme that have suggested within-task preference changes do happen. Further, consider the common folklore (and those recommended in most Marketing Research texts) on the need to consider a 'smallish' set of attributes with those of Louviere suggesting that conjoint can handle 'massive' numbers of attributes. Regardless, this discussion suggests that a difference of opinion does exist, there may be cases in which both sides are 'correct', and that we (academics and practitioners) should stop using simplifying statements like 'you can't have more than X attributes', as they may damage practice more so than providing useful prescriptive advice if they are based on anecdotal evidence.
Finally, can we 'coerce' conjoint analysis to handle many of the situations to which it was not psychologically or theoretically developed? While on one hand, and he is probably right, Louviere suggests that radical changes are necessary, on the other hand, such comments suggest that I will have a lot of fun over the next 20 years attempting to do just that, and to push the boundaries wherever possible.