Journal of Marketing Research Vol. XLII (February 2005), 67–82 67 *Min Ding is Assistant Professor of Marketing (e-mail: mqd9@psu. edu), Rajdeep Grewal is Assistant Professor of Marketing (e-mail: rug2@ psu.edu), and John Liechty is Assistant Professor of Marketing and Statis- tics (e-mail: jc112@psu.edu), Smeal College of Business Administration, Pennsylvania State University. The authors contributed equally to this manuscript. The authors thank Gary Bolton, Eric Bradlow, Wayne DeSarbo, Jehoshua Eliashberg, Tony Kwasnica, Gary Lilien, and Arvind Ramgaswamy for their constructive comments and two anonymous JMR reviewers for their feedback. This research was supported by the Market- ing Department at Pennsylvania State University. The authors greatly appreciate the cooperation of New Chinatown restaurant at State College, Pa., and the Laboratory for Economic Management and Auctions at Penn- sylvania State University. MIN DING, RAJDEEP GREWAL, and JOHN LIECHTY* Because most conjoint studies are conducted in hypothetical situations with no consumption consequences for the participants, the extent to which the studies are able to uncover “true” consumer preference struc- tures is questionable. Experimental economics literature, with its empha- sis on incentive alignment and hypothetical bias, suggests that more realistic incentive-aligned studies result in stronger out-of-sample predic- tive performance of actual purchase behaviors and provide better esti- mates of consumer preference structures than do hypothetical studies. To test this hypothesis, the authors design an experiment with conven- tional (hypothetical) conditions and parallel incentive-aligned counter- parts. Using Chinese dinner specials as the context, the authors conduct a field experiment in a Chinese restaurant during dinnertime. The results provide strong evidence in favor of incentive-aligned choice conjoint analysis, in that incentive-aligned choice conjoint outperforms hypotheti- cal choice conjoint in out-of-sample predictions. To determine the robust- ness of the results, the authors conduct a second study that uses snacks as the context and considers only the choice treatments. This study con- firms the results by providing strong evidence in favor of incentive- aligned choice analysis in out-of-sample predictions. The results provide a strong motivation for conjoint practitioners to consider conducting stud- ies in realistic settings using incentive structures that require participants to “live with” their decisions. Incentive-Aligned Conjoint Analysis Conjoint analysis, which has developed into a widely applied methodology for making inferences about con- sumer preferences and for uncovering empirical demand functions (Carrol and Green 1995), has many substantive applications in marketing, such as those for new product development (e.g., Kohli and Mahajan 1991), pricing (e.g., Mahajan, Green, and Goldberg 1982), segmentation (e.g., Green and Krieger 1991), and positioning (e.g., Green and Krieger 1992). Conjoint analysis also has been applied suc- cessfully in practice (Cattin and Wittink 1982; Wittink and Cattin 1989; Wittink, Vriens, and Burhenne 1994), and there is extensive literature on the subject (for reviews, see Green, Krieger, and Wind 2001; Green and Srinivasan 1978, 1990). As a result, there are many variants of conjoint analysis based on the way preference scores are elicited (e.g., ratings, rankings, self-explicated, constant sum, choice), the type of designs used (e.g., full factorial, frac- tional factorial, adaptive), the type of models estimated (e.g., regression, logit, probit, hierarchical Bayes), and the estimation procedures used to make inferences (e.g., maxi- mum likelihood, Markov chain Monte Carlo). Despite these differences, most methods have certain common elements. Data collection requires consumers to rate, rank, or select alternative products, and the goal of the data analysis is to find the set of partworths that, given a compositional rule, is most consistent with the respondent’s overall preferences (Green and Srinivasan 1978). Although early research on conjoint analysis rarely used out-of-sample predictions to assess model validity, scholars have suggested that such predictions are the strongest means to assess the validity of conjoint studies (Green and Srinivasan 1990). As a result, three types of validation or prediction tasks—aggregate-level market share predictions (e.g., Srinivasan et al. 1981), individual-level predictions of purchase intentions (e.g., Leigh, MacKay, and Summers 1984), and individual-level predictions of actual behaviors (e.g., Srinivasan 1988; see also Green and Srinivasan 1990)—have dominated the conjoint landscape. However, each method has limitations.