Have an idea?

Visit Sawtooth Software Feedback to share your ideas on how we can improve our products.

DCE and CBC: Differences in underlying theory?

Louviere et. al. (2010) suggest that DCE, which is based on Random Utility Theory has a direct link to a theory of consumer behaviour, whereas Conjoint Analysis, which is based on Conjoint Measurement theory, has no basis in any theory of consumer behaviour. As I understand the paper, the key issue is the theory underpinning the two approaches, not the experimental design approach used.

I'd welcome your comments on the paper and on whether even CBC may be inappropriate to model consumer choice given the lack of support from consumer behaviour theory.

The paper I refer to is at (https://www.google.co.in/url?sa=t&rct=j&q=&esrc=s&source=web&cd=1&cad=rja&uact=8&ved=0CCMQFjAA&url=http%3A%2F%2Feconweb.ucsd.edu%2Fcee%2Fpapers%2FCarsonPapers%2FLouviere_Flynn_Carson_Oct2010.pdf&ei=Moh1VZ-wFaO8mAWp6oOwCw&usg=AFQjCNGpJyrkQ00CgGe_lmwoQO0fsDtLPw&bvm=bv.95039771,d.dGY)

asked Jun 8, 2015 by Manohar

3 Answers

+1 vote

I think the key here it to realize that the paper is NOT describing the differences between DCE and CBC.  This is clear from the statement that conjoint analysis was enhanced to use "rating scales instead of rankings to elicit preference orders and/or utility differences."  The conjoint analysis criticized in the paper is the more traditional conjoint analysis typically based on orthogonal main effects plans, preference rating scale evaluations and multiple regression analysis.  

CBC is choice-based conjoint analysis. It just IS a special case of designed choice experiment (there are others, like MaxDiff or availability experiments with just one attribute (brand) which may be available or unavailable).  CBC is the special case of DCE where we use choice sets composed of multi-attribute profiles as our stimuli.

Don't let the confusing nomenclature trip you up - it's been confusing from the start because DCE's have also been called choice experiments, designed choice experiments, choice-based conjoint analysis and discrete choice experiments, plus they've been given various proprietary names for marketing purposes.  Choice-based conjoint analysis is just one of those names.    Worse, they're sometimes called discrete choice models (DCM) which is especially confusing because that term would also describe non-experimentally designed choice models (e.g those based on cross-sectional data or even scanner panel data).   

CBCs are every bit as based on random utility as any other DCE because they are a type of DCE.
answered Jun 8, 2015 by Keith Chrzan Platinum Sawtooth Software, Inc. (63,000 points)
Thank you Keith for clarifying that CBC does have a basis in RUT.
+1 vote
Louviere's Discrete Choice Experiments are the same as Sawtooth Software's Choice-Based Conjoint (CBC).  Louviere argues that academics should stop calling Discrete Choice Experiments any term that uses "Conjoint" in the label, since he argues the term conjoint should be reserved for the ratings-based conjoint approaches of Paul Green (with his full-profile card-sort conjoint) and Rich Johnson with his Adaptive Conjoint Analysis (ACA) paired profiles ratings-based system.  So, he is not pleased when Sawtooth Software refers to our discrete choice experiments as Choice-Based Conjoint.  Sawtooth Software decided to refer to the methodology as Choice-Based Conjoint around 1992 or so to make the link in marketers' minds that the discrete choice experiments had some key things in common with what marketers already had fallen in love with: conjoint analysis.  So, it seemed natural to create that bridge with the name "Choice-Based Conjoint".

Sawtooth Software's CBC system follows closely Louviere's innovation of Discrete Choice Experiments as he and Woodworth laid out in the 1980s.  It uses RUT theory (logit estimation), choices of alternatives (profiles) from sets, the possibility of using alternative-specific designs.

It's mainly a branding thing.  Louviere and others who come out of the fine economics tradition (McFadden, Ben-Akiva), would prefer to see the theory and terms that were forwarded through the marketing tradition (Green, Rao) die and be fully recognized in the research and academic world as the victorious methodology and the proper terms to use.

Indeed, the economic stream (discrete choice experiments) with choices from sets, RUT theory, closely linked to the logit rule, is the dominant methodology today (about 90% of "conjoint-related" studies done by Sawtooth Software's customers are of the CBC variety rather than the ratings-based methods coming out of the marketing stream of literature).  They are generally superior both from a theoretical perspective and practical perspective.

We at Sawtooth Software recognize the superiority of the discrete choice experiments, but we alternatively refer to that methodology sometimes using the discrete choice terms and sometimes using the conjoint terms.  To us it's a way to be more inclusive of people whether they come from the econ tradition or the marketing tradition.
answered Jun 8, 2015 by Bryan Orme Platinum Sawtooth Software, Inc. (148,340 points)
Thank you Brian for the detailed answer
0 votes
Odd that Louviere ignores Luce & Tukey, 1964. Green in his development of conjoint analysis (CVA) referred to this. Luce & Tukey specifically refer to 'behaviour' in their 1964 paper. The paper was published in the journal of mathematical psychology; clearly they are not just interested in mathematics, but how human perception can be measured?
Yet Louviere claims "Thus, CA evolved out of the theory of “Conjoint Measurement” (CM), which is purely
mathematical and concerned with the behavior of number systems, not the behavior of humans or human preferences".
answered Sep 1, 2017 by Craig Kolb