Formal Models of Group Choice in Organizational Buying: Toward a Contingency Paradigm

Elizabeth J. Wilson, Louisiana State University
Gary L. Lilien, Pennsylvania State University
David T. Wilson, Pennsylvania State University
ABSTRACT - We configure seven formal models of group choice (Choffray and Lilien 1980) in a contingency paradigm composed of three situational factors identified by Sheth (1973). Our key research question is whether some models of group choice predict actual group decision making outcomes better than alternative models, given situational factors. We develop a general proposition concerning which model should fit best in each cell of the contingency paradigm. We then estimate the models and assess the appropriateness of the contingency paradigm. Results of a pilot test provide initial support for the contingency paradigm.
[ to cite ]:
Elizabeth J. Wilson, Gary L. Lilien, and David T. Wilson (1989) ,"Formal Models of Group Choice in Organizational Buying: Toward a Contingency Paradigm", in NA - Advances in Consumer Research Volume 16, eds. Thomas K. Srull, Provo, UT : Association for Consumer Research, Pages: 548-554.

Advances in Consumer Research Volume 16, 1989      Pages 548-554


Elizabeth J. Wilson, Louisiana State University

Gary L. Lilien, Pennsylvania State University

David T. Wilson, Pennsylvania State University


We configure seven formal models of group choice (Choffray and Lilien 1980) in a contingency paradigm composed of three situational factors identified by Sheth (1973). Our key research question is whether some models of group choice predict actual group decision making outcomes better than alternative models, given situational factors. We develop a general proposition concerning which model should fit best in each cell of the contingency paradigm. We then estimate the models and assess the appropriateness of the contingency paradigm. Results of a pilot test provide initial support for the contingency paradigm.

The purpose of this paper is to examine the goodness-of-fit of seven formal models of group choice (Choffray and Lilien 1980) within a contingency paradigm composed of three factors identified by Sheth (1973). The contingency factors are: the buying task, financial commitment, and product complexity. The key question of the research is how well the seven models predict group choice outcomes, given contingent factors.


We review seven formal models below. Although the models are intuitively simple representations of choice processes, they have not been empirically tested because of the difficulty of obtaining data from a relevant subject population.

Formal Models of Group Choice

Choffray and Lilien (1980) propose seven formal group choice models to describe how organizational buying centers might make supplier choice decisions (Choffray and Lilien 1980). Using constant-sum preference data, each model provides a distribution of preference scores.

Model 1: The Weighted Probability Model. The weighted probability model assumes that the buying center is likely to adopt a given vendor-product alternative in proportion to the relative importance of the buying center members. The weighted probability model is:



PG(ao) is the probability of the buying center choosing alternative ao,

wd is the weight (relative importance) of decision participant d, EQUATION

Pd(ao) is the preference that participant d has for alternative ao.

Model 2: The Equiprobability Model. The equiprobability model is a special case of the weighted probability model. Each decision participant is given equal weight--for a three person group, each participant would be assigned a weight of 0.33--and the weights sum to 1.0. "This is an appealing model, because it is a zero-information or naive model. The industrial marketing manager need only identify the decision participants and does not have to measure or provide subjective estimates of the importance coefficients" (Choffray and Lilien 1980, p. 241).

Model 3: The Autocracy Model. Another special case of the weighted probability model is the autocracy model. This is a "key-informant" model because the most important member of the buying center is given a weight of 1.0 and other members are given a weight of 0.0. This model represents the current norm for industrial marketing research studies, i.e., that one individual's preferences represent those of the buying organization.

Model 4: The Voting Model. The voting model does not include a weighting factor for buying center members. It states that the probability that the group will choose a vendor-product alternative [PG(aO)] is the likelihood that the alternative (aO) receives the highest preference score versus any other alternative in the set of vendor-product alternatives.

Model 5: The Minimum Endorsement Model-Majority Rule. The minimum endorsement model assumes that in order to be accepted by a firm, a product alternative has to be the choice of a prespecified number (quota) of participants involved in the decision. For a majority rule, at least two members of a three person buying center must be in favor of a vendor-product before it can be assigned a choice probability. The majority rule model is:



PG(ao) is the probability of the group choosing alternative ao,

EQUATION is the conditional probability of the buying center selecting ao, given that the majority of group members are in agreement on their judgments of the aj alternatives.

Model 6: The Minimum Endorsement Model-Unanimity Rule. The unanimity model is a second version of the minimum endorsement model. The assignment of choice probabilities to vendor-product alternatives requires that all members of the buying center agree in their judgments of the preferred alternatives.

Model 7: The Preference Perturbation Model. The preference perturbation model assumes that if a group does not reach unanimity, it is most likely to choose the alternative that perturbs individual preferences least. "The probability that a given product would be chosen by a firm's buying center is inversely proportional to the number of preference shifts that would be needed to make that alternative the first choice of every decision participant" (Choffray and Lilien 1980, p. 140). The preference perturbation model is:



PG(ao | gw) is the probability that the group chooses ao given the preference structure gw,

Pr [ gw ] is the probability of obtaining that preference structure.


The formal group choice models reviewed represent several forms of the choice process. A priori, no one model should be expected to dominate in terms of predictive ability. We hypothesize that use of a particular choice process is dependent on the situation surrounding the decision.

Three factors are included in our contingency paradigm for two reasons. First, these factors appear often and consistently in the organizational buying literature (Moller and Laaksonen 1986). Second, three factors are a manageable number for an empirical study--to test all factors identified in past studies would not be parsimonious nor offer generalizable conclusions for theory development.

Buying Task. Our first factor is the type of buying task (Robinson, Faris, and Wind 1967). Modified rebuy and new task decisions are two levels of this factor. The straight rebuy situation is not included in the paradigm because straight rebuys tend to be characterized by routinized response behavior (Moller and Laaksonen 1986) and are not likely to require group interaction in order for a decision to be made.

Financial Commitment. Our second factor is the level of financial commitment required in the purchase decision. Financial commitment is similar to the "perceived risk" component of Sheth's (1973) model. We use ''financial commitment" instead of "perceived risk" because there is likely to be less measurement error in assessing price (a dollar amount) versus measuring perceived risk. Financial commitment reflects the magnitude of adverse consequences if a wrong choice is made. We assume that loss of dollars is an adverse consequence because it may lead to negative sanctions for buying center members. We specify two levels of financial commitment (high and low/ moderate).

Technical Complexity. Finally, we include technical complexity of the product in the paradigm at two levels (simple and complex). This is another "product specific" factor;

In sum, we identify each contingency factor at two levels that represent a range of consequences for a bad decision. At one end of the range, consequences are high for new tasks, high financial commitment, and technically complex products. At the opposite end, consequences are low for modified rebuys, low financial commitment, technically simple products.


Proposition: Our main proposition is that the degree of predictive accuracy of the seven group choice models depends on the buying task, financial commitment required, and technological complexity of the product.

Rationale: Organizational buying decisions are diverse and it is unlikely that one type of choice model will fit best all the time; rather, we expect contingent factors to affect the predictive ability of the models. This idea is reflected in Exhibit 1, with the seven group choice models configured within the paradigm.

With a small sample of data from the pilot study, a statistical test of the individual cells in the paradigm cannot be conducted. However, an overall test of the paradigm is possible. Therefore, only an overall proposition for the paradigm is offered at this point in the research. A larger scale study (currently in progress) will allow a sufficient number of observations so that individual cells can be examined empirically.


We assessed the goodness-of-fit of each of the seven models using a small sample of buying centers. Buying center members first worked individually to develop preferences for a range of product alternatives. Buying center members then worked together to form preference scores as a group. The preference scores predicted by each model using individual preference data as input were then compared to the actual preference score distribution of the group. A chi-square goodness-of-fit statistic was calculated to see which model had the best fit to the actual group preference data.

Buying Decision Stimuli

We chose twenty products as stimuli for buying decision tasks. Many of the products we selected are central to the operation of manufacturing and service organizations across industries. We conducted background research to learn realistic levels of price, quality, delivery, and service/maintenance for each product. We personally interviewed purchasing agents involved in buying each product to construct buying tasks with a high degree of realism. Products included eleven manufacturing and MRO-related items (e.g., industrial valves, solvent reclaimers, warehouse vehicles) and nine office products (e.g., copiers, facsimile equipment, furniture).



We described products on four attributes: price, quality, delivery, and service/maintenance. We selected these attributes because they have been reported as criteria used in industrial buying (Lehmann and O'Shaughnessy 1974; Lilien and Wong 1984). We determined three realistic levels of each attribute from the personal interviews. A total of 43 = 81 possible configurations resulted for each product.

The buying choice task presents nine vendor-product alternatives in a balanced, orthogonal, incomplete block design with no one vendor-product alternative being best or worst on all attributes. The nine alternatives for each product can be vendors or products and are referred to as vendor-products for simplicity.


Two manufacturing firms in the Northeast US participated in the pilot study. The purchasing manager at each firm identified members of naturally occurring buying centers that had worked together in the past year. The purchasing manager then identified three to five products (from the list of 20) that each buying center had procured during the past year or was likely to procure during the coming year. We then assembled questionnaire sets for each buying center based on the relevant products.

A total of four buying centers participated in the pilot test. Three buying centers were composed of three persons each (a buyer, an engineer, and a user). One buying center was composed of a buyer and an engineer.




All data were collected on-site (by the first author) to preserve a realistic task environment. For each decision, subjects were instructed to work individually and examine the nine alternatives, cross out those they would not consider, and award a "share-of-available requirements" (a constant sum scale) to the vendor-products they preferred.

Measurement by a constant sum scale is realistic because buying centers have the option of using single or multiple sources of supply in reality. The preference data from the constant sum measure were used to calculate preference scores relative to the seven group choice models. Each model had nine preference scores (choice probabilities), one for each of the nine vendor-product alternatives.

Subjects then worked as a group to evaluate the same products. They discussed the nine alternatives to reach agreement on which vendor-products would be crossed out and how points would be allocated. This task yielded nine actual group preference scores (choice probabilities) that we compared to the model predictions. We completed fifteen buying decisions. At the end of the group task, each buying center was asked to classify each decision into a cell in the contingency paradigm based on how they evaluated the purchase within the context of their organization. The 15 decisions and fit rankings are configured in the contingency paradigm in Exhibit 2.


A small x2 value (15.5, 8 degrees of freedom) indicates that a model fits adequately (p < .05); that is, the preference responses for the actual group decision are close to those responses predicted by a model. Support for the contingency paradigm proposition may be evidenced if the models predicted

to do best in each cell have both a low x2 value and a good fit ranking versus the other models. The goodness-of-fit and ranking results for four decisions are presented in Exhibit 3 to illustrate the model scores and ranking results.

To read Exhibit 3, consider the decision to buy corrugated boxes, where the autocracy model should fit best. For Buying Center 1 (BC 1) the autocracy model is ranked fourth in terms of absolute fit and for BC 2 the autocracy model is ranked first in fit.



To examine the appropriateness of the overall contingency paradigm, a meta-analysis of the rankings was conducted. In seven decisions, the models specified by the contingency paradigm were the best fitting (ranked number 1 or tied for first). In two decisions the models predicted to fit best were in second place. In one decision, the predicted model was ranked third and so on as noted in Exhibit 4. This distribution of the rankings is substantially different from that expected by chance (X2= 15.40, 6 d.f., p < .02). Therefore, the contingency paradigm is supported based on these initial pilot-test findings. See Exhibit 4.


The study has several limitations and strengths. The first limitation is the small sample of decisions (n = 15). The results reported here were based on a pilot test and a larger scale study is currently in progress. We plan to obtain a sample of at least one hundred buying decisions in order to have enough data in each cell of the contingency paradigm to conduct a statistically meaningful meta-analysis.

Second, the buying centers chosen for participation may not have included all the persons involved. In reality, others may be involved in a peripheral manner; i.e., financial executives that provide budget approval. Third, we studied only the choice phase of the decision. There was no opportunity for activities such as information search, meetings with vendors, or buying center meetings over time. Finally, the decision tasks were simulated rather than real purchases.

A major strength of the study is that it is the first reported empirical examination of Choffray and Lilien's (1980) models of group decision making. The study addresses a need in the literature for theoretical and empirical work in the area of group choice. Such research is scant because of the difficulty of collecting data from relevant subjects about relevant problems. The method used here overcomes many of the difficulties.

A second strength is that it employs a relevant subject population, i.e., real buying centers. Third, we match members of buying centers to products that they would work with in the course of their jobs. Fourth, we have collected all data in a naturalistic environment within the subjects' organization.

Finally, in post-hoc reactions to the study, subjects reported that the perceived the decision tasks to be highly realistic in terms of the attribute descriptions of vendor-products. They maintained that the decision processes used to evaluate each product were similar to those that occur in real buying decisions.



Implications and Future Research

Results from this research may provide insight to buying groups in organizations about decision making styles (Wilson 1972). Decision making strategies in buying may be improved if the underlying processes and effects of situational factors are understood.

Sales managers and sales representatives may benefit from knowledge of decision processes of buying centers. Sales strategies may be more efficiently and effectively designed by understanding the type of decision making process that is likely to be used by buying centers. For example, if the seller's product is a technically simple, low cost, rebuy item, the seller may find it beneficial to call on the buying center member with the most expertise who is likely to make an autonomous decision.

Results of this study have implications for industrial marketing researchers. A "key informant" research methodology may be valid and acceptable if the type of decision being studied is a rebuy for technically simple, low cost products. The autocracy model tended to fit best in these situations and conducting interviews with all buying center members may not be necessary if one person is actually making the decision. However, for other types of decisions, the key informant method will likely not yield results with a high degree of validity (Phillips 1981). This research suggests what combining rule for individual respondents to an industrial market research study is most appropriate.

These models could also be applied in a family decision making setting with spouses plus other family members (children, grandparents, other relatives) as group members. An application of the seven formal models and contingency paradigm to the family setting is possible because Sheth's (1974) model of family decision making has many of the same contingency factors as those used for organizational buying. Theory development in the family decision making area could benefit from such research.


Choffray, Jean-Marie and Gary L. Lilien (1980), Market Planning for New Industrial Products, New York: Wiley.

Lehmann, Don and John O'Shaughnessy (1974), "Difference in Attribute Importance for Different Industrial Products," Journal of Marketing, 38 (April), 36-42.

Lilien, Gary L. and M. Anthony Wong (1984), "An Exploratory Investigation of Structure of the Buying Center in the Metalworking Industry," Journal of Marketing Research, 21 (February), 1-11.

Moller, Kristian and Martti Laaksonen (1986), "Situational Dimensions and Decision Criteria in Industrial Buying," in Advances in Business Marketing, vol. 1, Arch G. Woodside, ed., Greenwich, CT: JAI Press, 163-208.

Phillips, Lynn W. (1981), "Assessing Measurement Error in Key Informant Reports," Journal of Marketing Research, 18 (November), 395-415.

Robinson, P., C. Faris, and Yoram Wind (1976), Industrial Buying and Creative Marketing, Boston, MA: Allyn and Bacon.

Sheth, Jagdish (1973), "A Model of Industrial Buyer Behavior," Journal of Marketing, 37 (October), 50-56.

Sheth, Jagdish (1974), "A Theory of Family Buying Decisions," Models of Buyer Behavior, New York: Harper and Row, 17-33.

Wilson, David T. (1972), "Industrial Buyers' Decision Making Styles," Journal of Marketing Research, 8 (November), 433-436.