Information overload in conjoint experiments.

This paper explores the relationship between individual level variables, stimulus variables and the experience of information overload in conjoint experiments. Drawing on theories of contingent information processing, it develops a set of hypotheses linking product class involvement and product class knowledge to the level of information overload experienced by individuals when performing a conjoint task. It also investigates the effects on overload of the total amount of information. The paper also explores to what extent the amount of information to be processed prompts respondents to change their information-processing strategy in order to avoid the unpleasant effects of information overload.

**********

Introduction

Researchers have long been concerned about the impact of designing tasks that are too demanding for respondents (e.g. Milford & Perry 1977). Highly demanding tasks can result in careless responding and lower response rates, both of which threaten the validity of empirical results. Conjoint researchers have feared that the information load created by asking respondents to evaluate large numbers of product profiles, each containing a substantial amount of information, could create a condition similar to that described as 'information overload' in the literature on consumer information processing (Jacoby et al. 1974a, 1974b). If respondents' information-processing systems become overloaded during performance of conjoint tasks, the benefits of having persons respond to realistic (e.g. full profile) product descriptions may be offset by reactions to an overly demanding task (Huber et al. 1993). A substantial amount of work has been done in the area of adapting the basic conjoint methodology in order to make conjoint tasks easier for respondents to perform. In particular, much development work has focused on modifying the methodology so that the likelihood of experiencing information overload is reduced. This has led to a new set of techniques, termed hybrid conjoint measurement, that have in common the reliance on self-explicated information (attribute importance and attribute level preferences) that is used in combination with decompositional information derived from the performance of traditional conjoint tasks (Green et al. 1981; Green & Krieger 1996). However, traditional conjoint models continue to be widely used by academic as well as applied researchers (Green et al. 2002), and the focus of this paper is on information overload in traditional conjoint, rather than hybrid conjoint, measurement.

The first issue to be explored in this paper is the extent to which respondents' experience of information overload is influenced by the amount of information contained in conjoint stimulus materials. Second, it investigates how the experience of information overload is affected by the individual level variables product class involvement and product class knowledge. This issue is important because it is concerned with the applicability of conjoint research in product classes for which consumer knowledge and involvement are known to be high or low. Third, research on contingent information processing (Bettman et al. 1991) indicates that consumers may react to high levels of information load by changing their information-processing strategy. This issue is related to the ecological validity of conjoint results: if the information-processing demands in a conjoint experiment instigate a change in information-processing strategy away from that used in a realistic setting, results may not carry over to that setting. The third issue of interest in this paper, then, is whether and how respondents react to the amount of information in conjoint tasks by adapting the way in which they process this information.

In order to explore the above issues, we designed a quasi-experimental study where information load was varied by using product profiles (descriptions of hypothetical mobile telephones) including three, five, seven and nine attributes. The paper proceeds as follows: first, it develops a set of hypotheses derived from an integration of literature on information overload and contingent information processing; next, the methodology used for testing these hypotheses is explained in more detail; this is followed by a presentation of the findings; finally, the results are discussed, and the implications of the study developed and presented.

Theoretical background and hypotheses