Using Single Informants to Study Group Choice: An Examination of Research Practice in Organizational Buying

Purchasing agents are often employed as individual informants in studies of organizational buying. This practice occurs in spite of the fact that several researchers have identified problems with using individual informants to study group buying behavior. The purpose of this study is to examine the appropriateness of using either single or multiple informants both in non-new task and new task group buying situations. We compare the predictive accuracy of a single-informant. autocracy model with that of a multiperson, majority rule model. The study includes 104 group decisions made by buying centers in organizations. Overall, we find that data from multiple informants significantly outperforms single informant data for both new task and non-new task situations. and that singlefkey informant data should be used in organizational buying research only with care and caution. Organizational buying behavior researchers often identify and use buyers as individual informants to gain quantitative information about processes of group decision making and supplier choice (Moriarty and Bateson 1982). This practice is controversial given empirical evidence that key informant data has questionable reliability and validity (Phillips 1981; Silk and KaIwani 1982). Campbell (1955) suggests that key informants in quantitative research should be we11 informed about the phenomenon of interest and able to communicate effectively with the researcher. When a buyer is chosen as the key informant based on his or her knowledge of and participation in the decision, this approximates the original purpose of informants (Mead 1953, Campbell 1955). When the buyer is chosen be*We thank the anonymous Marketing Letters reviewers, Donald R. Lehrnann, Kim P. Corfman, and Joel H. Steckel for comments on earlier drafts of this manuscript. We are grateful for the financial support for this research provided by Pennsylvania State University's Institute for the Study of Business Markets, the National Association of Purchasing Management, and the College of Business Administration of Louisiana State University. 298 ELIZABETH J . WILSON AND GARY L. LILIEN cause he or she is convenient to the researcher, we can only say that buyer is a single informant, not necessarily a key informant. Several questions motivate our research. First, in what situations can single informants provide an accurate report on organizational buying behavior? Second, in what situations should researchers collect data from multiple informants about buying center choice? We hypothesize that collecting buying center choice information from a group of individuals is likely to provide better information than that provided by a single individual. Therefore, we make the following conjecture: the less experience the firm has with the product in question, the worse a single informant is likely to do at representing the outcome of a group decision. In support of our conjecture, we maintain that while the informant may be willing to answer questions about the buying process and choice decision, it is possible that he or she is inexperienced or not very knowledgeable about the product. This possibility could be reduced by gathering data from others involved in the buying process. In using single informants to study group choice in organizational buying, researchers risk generating biased data since inexperienced respondents may "tell more than they can know" (Nisbett and Wilson 1977) to appear credible. Finally, our third research question centers around a common practice in organizational buying. Should the buyer automatically be designated the single informant? This happens in some cases because buyers have specialized knowledge but in other cases they are chosen because they are convenient to researchers (e.g., Jackson, Keith, and Burdick 1984). To address the research questions above, we categorize buying decision data as new task or non-new task decisions. Generally, we expect information from multiple informants to allow better choice predictions than information from single informants across decision situations. Specifically, we expect single informant data to improve on choice predictions more in non-new situations than in new task situations (Patton, Puto, and King 1986). 1. Modeling single informant and multiple informant choice To explore group choice prediction using one versus several informants, we draw on two models proposed by Choffray and Lilien (1980) -the autocracy model to predict choice based on single informant reports and the majority rule model to predict choice based on multiple informant information' (see Steckel et al. 1991 for a discussion of both modeling and measurement issues in the group choice area). The autocracy model implies that an organization designates a single member of the buying center as the decision-maker in supplier choice. By having a single decision-maker, his or her preferences become the preferences of the entire buying center and, thus, he or she acts autocratically (Choffray and Lilien 1980). Identification of the autocrat is based on estimated expertise weights of buying center members. Findings of Patton et al. (1986) support the existence of autoRESEARCH PRACTICE I N ORGANIZATIONAL BUYING 299 cratic decision making in organizational buying; in many cases, individuals, not groups, tend to make low-risk, modified rebuy decisions. We use the majority rule model to incorporate information from multiple informants in buying center choice. In a test of Choffray and Lilien's (1980) group choice models, the majority-rule model is relatively robust compared to other multiperson choice schemes (Wilson, Lilien, and Wilson 1991). While this may or may not be the best model for representing group choice, it is a sensible one with theoretical and empirical support (see Choffray and Lilien 1980 and Wilson et al. 1991). The majority rule model assumes that to be acceptable, a supplier alternative must be the choice of at least a prespecified number of buying center members. More formally, the model states that the probability that the buying group chooses some alternative from a choice set is equal to the probability that the alternative is the first choice of at least a simple majority of buying group members. In a three member group, that alternative must be chosen by either two or three members; in a two member group, both individuals must agree on a choice. When alternatives are equally preferred (tied), a secondary decision scheme is invoked to eliminate ties.'