Behavioral Decision Making: a Comparison of Three Models

Citation:

J. Paul Peter and Lawrence X. Tarpey, Sr. (1975) ,"Behavioral Decision Making: a Comparison of Three Models", in NA - Advances in Consumer Research Volume 02, eds. Mary Jane Schlinger, Ann Abor, MI : Association for Consumer Research, Pages: 119-132.

Advances in Consumer Research Volume 2, 1975      Pages 119-132

BEHAVIORAL DECISION MAKING: A COMPARISON OF THREE MODELS

J. Paul Peter, Indiana State University

Lawrence X. Tarpey, Sr., University of Kentucky

[J. Paul Peter is an Assistant Professor of Marketing, Indiana State University, Terre Haute, Indiana, and Lawrence X. Tarpey is Professor of Business Administration, University of Kentucky, Lexington, Kentucky.]

This study compared three alternative decision-making strategies evidenced in the consumer behavior literature. The first strategy, minimization of expected negative utility (perceived risk) was formulated as the probability times importance of six facets of potential gain; the second strategy, maximization of expected positive utility (perceived return) was formulated as the probability times importance of six facets of potential gain; the third strategy, maximization of expected net utility (net perceived return) was formulated as the difference between perceived return and perceived risk. Using brand preference scores for six brands of automobiles as the dependent variable, multiple regression analysis indicated that the net perceived return model could explain more variance and produce more significantly related variables than the other two models.

INTRODUCTION

The literature dealing with the topic of consumer behavior is replete with theoretical constructions and models of varying degrees of complexity. A careful review of this literature leads one to the inexorable conclusion that the search for a "grand theory" of consumer behavior would be quixotic. A more reasonable task for the scholar-researcher is to rigorously test some of the existing theoretical formulations as a means of discovering which ones are more perspicacious explanations of how specific purchase decisions are made. In this paper the authors attempt to do just this.

One class of consumer decision-making models can be appropriately labeled as "cognitive-rational" because they focus on the key organizing variables of risk and utility (or pay-off). As decision making strategies these models postulate a consumer who operates in a manner similar to the rational man of economic theory in that behavior is (a) goal directed, (b) calculated and (c) predicated upon some knowledge of the costs and benefits of alternative choices. From an operational standpoint these models can be described as additive utility models and the literature reveals three basic formulations (or strategies).

First, as evidenced by the work begun at Harvard by Bauer in 1960 and carried on by others, there is the "perceived risk" strategy which generally assumes that consumers act to minimize (or at least reduce) any expected negative utility associated with purchase behavior; no serious consideration is given to expected positive utility. Second, there are the so-called attitude models which focus on the benefits of products which are positively evaluated and there is little consideration of expected negative utility. Third, there are studies which emphasize the "valence" concept. These were pioneered by Lewin (1943) and Bilkey (1953, 1955) where the research recognized the fact that consumers perceive products as having both desirable (positive valence) and undesirable features (negative valence). The implicit strategy in this research is that individuals attempt to maximize the "net valence" which is the arithmetic difference between expected positive and negative utility (i.e., "net perceived return").

Thus, in the context of a risk-return typology, there would appear to be three distinct strategies in terms of how consumers make decisions. These three strategies could be stated as:

(1) Select the brand that minimizes expected loss(perceived risk)

(2) Select the brand that maximizes expected gain(perceived return)

(3) Select the brand that maximizes net expected gain(net perceived return)

All three of these strategies have received some empirical support and two critical observations are relevant here: (1) to date no consumer behavior study has compared the three strategies in terms of their relative explanatory power and (2) few studies have investigated the various facets or dimensions of utility consumers consider in their selection process. With all this in mind the authors designed a study which formulated (based on previous literature) three models representative of each of the three strategies as a means of comparing them in terms of their relative explanatory power. Another purpose of the study was to investigate the facets of utility discussed in the consumer behavior literature which consumers consider in making certain types of purchase decisions.

In terms of the dimensions of perceived risk-return,-six facets have been selected for use in this study. They are: (1) financial risk-return, (2) performance risk-return, (3) psychological risk-return, (4) physical risk-return, (5) social risk-return, and (6) time risk-return. Those particular facets have been selected because they were the most prominent and widely discussed in the perceived risk literature. It should be noted, however, that only one study (Jacoby and Kaplan, 1972) has discussed all six of these facets and argued that they are conceptually independent. Most studies have only dealt with a subset or, in some cases, combinations of these six facets.

THE THREE MODELS

The Perceived Risk Model

Since the introduction of perceived risk to consumer behavior by Bauer (1960), the construct has been conceptualized as a dual-component, multi-faceted phenomenon. As defined by Kogan and Wallach (1964), the two components of perceived risk are " . . . 'a chance' aspect where the focus is on probability (of losing) and a 'danger' aspect where the emphasis is on severity of negative consequences." In terms of these components, Sieber, et al. (1964), Cunningham (1967), and Hansen (1972) have suggested a multiplicative model which could be described as:

PR = f (PL ^. IL)   [1]

where

PR = perceived risk

PL = probability of a loss

IL = importance of a loss

The above model is not a multi-faceted model and needs to be modified in order to accommodate the notion that consumer purchase risk is a multi-faceted concept. Thus, a multi-faceted model of perceived risk can be depicted as follows:

EQUATION   [2]

where

OPR = overall perceived risk

Ri = risk facets such as financial, performance, physical, psychological. social and time losses

Although discussed as a dual-component, multi-faceted phenomenon, the empirical research of perceived risk has been confined to measurements either of the facets or the components but seldom the two together. For example, the research done by Bauer, Cox, Cunningham, and others as Harvard, Cox (1967) as well as Copley and Callom (1971) and Hirsch, et al. (1972) employed general measures of both uncertainty and importance components but did not delineate the various facets of risk. On the other hand, Perry and Hamm (1969), Schiffman (1972), Roselius (1971), Jacoby and Kaplan (1972) and Zikmund and Scott (1973) ignored the components of perceived risk and used a general measure for each facet. Further, although Jacoby and Kaplan (1972) do not measure the two components, they suggested the following model for overall perceived risk: [They also mentioned "Since this study was conducted (Spring, 1970), Roselius (1971, p. 58) has identified a sixth variety of risk: Time loss. . ." (p. 393) and suggested that it should be included.]

OPR = f (UFR ^. CFR); (UPR1 ^. CPR1); (UPR2 ^. CPR2); (UPR3 ^. CPR3); (USR ^. CSR) + error   [3]

where

OPR = overall perceived risk

UFR = uncertainty of financial risk

CFR = consequences of financial risk

UPR1 = uncertainty of performance risk

CPR1 = consequences of performance risk

UPR2 = uncertainty of physical risk

CPR2 = consequences of physical risk

UPR3 = uncertainty of psychological risk

CPR3 = consequences of psychological risk

USR = uncertainty of social risk

CSR = consequences of social risk

Although they fail to hypothesize the relationship among the facets, in view of (1) the analogy drawn between perceived risk and the additive Fishbein Attitudinal Model (Zikmund, 1973), and (2) the consumer behavior additive utility models discussed by Sheth (1970) and Moinpour and MacLachlan (1971), an additive model can he formulated to accommodate the need to deal with multi-risk facets:

EQUATION    [4]

where

OPRj = overall perceived risk for brand j

PLij = probability of loss i from the purchase of brand j

ILij = importance of loss i from purchase of brand j

n = risk facets

In this extended model the concept of perceived risk is depicted not only as a multiplicative function of probability of loss and importance of loss as in equation (1), but also an additive model of the various facets of risk as in equation (2).

The Perceived Return Model

As pointed out by Wilkie and Pessemier (1973, p. 429), a basic formulation of the consumer behavior attitude model can be described as:

EQUATION    [5]

where

i = attribute or product characteristic,

j = brand,

k = consumer or respondent

such that

Ajk = consumer k's attitude score for brand j,

Iik = the importance weight given attribute i by consumer k, and

Bijk = consumer k's belief as to the extent to which attribute i is offered by brand j.

Fishbein (1971, p. 315) refers to this model as an "attitude toward an object model; however, the perceived return model used in this study is couched more in terms of an "attitude toward behavior" model. The importance of this distinction is given by Fishbein (1971, p. 315):

Thus, the salient beliefs I want to consider are beliefs about (the consequences of) buying or using the product rather than beliefs about (the attributes of) the product per se. . . . I think this distinction between attitude toward behavior and attitude toward an object is a crucial one, and one that has been overlooked. . . . All I am saying is that whether I buy product X . . . will depend more on my beliefs about (the consequences of) buying the product . . . than on my beliefs about (the attributes of) product X.

In addition to this argument, there are two other reasons why the "attitude toward an object" model was not used in this study. First, of prime importance for the "attitude toward an object" model is the specification of salient attributes; as Wilkie and Pessemier (1973, p. 432) have observed, "attribute specification is the weakest part of composition models." Also, although Wilkie and Pessemier (1973, p. 433) stated that in terms of attribute specification, "theoretical development is preferable," there is ". . . a lack of theoretical concepts for attributes." This problem coupled with the theoretical background in the perceived risk literature dealing with expected behavior outcomes, led logically to a preference for a perceived return model based on "attitude toward behavior."

Second, since a primary purpose of this study was to compare decision making models, it was felt that these models should be as compatible as possible. Consequently, comparing negative expected outcomes with positive product attributes was felt to be conceptually undesirable. Thus, the perceived return model was based on positive expected outcomes rather than on product benefits and can be depicted as:

EQUATION    [6]

where

OPRej = overall perceived return for brand j

PGij = probability of gain i from purchase of brand j

IGij = importance of gain i from purchase of brand j

n = return facets

Although the perceived return model was formulated identically to the perceived risk model except for the focus on positive utility (or gains) instead of negative expected utility, the two models are conceptually independent. In other words, the level of perceived risk has no necessary bearing on the level of perceived return.

The Net Perceived Return Model

In essence, Lewin's vector hypothesis of consumer behavior states that (1) the net valence is the arithmetic difference between positive valences and negative valences and (2) if this remainder is positive, the purchase will tend to be made and vice versa. Based on this logic, the net perceived return model is formulated as:

EQUATION  [7]   and  [8]

where

NPRej = net perceived return for brand j

PGij = probability of gain i from purchase of brand j

IGij = importance of gain i from purchase of brand j

PLij = probability of loss i from purchase of brand j

ILij = importance of loss i from purchase of brand j

n = utility facets

This model is a combination of the perceived risk and perceived return models. Its purpose is to test the maximization of net utility hypothesis. Conceptually, since this model takes into account (explicitly) both positive and negative expectations, it is intuitively the superior model.

METHOD

Two questionnaires were employed in this study. Both questionnaires are identical except that questionnaire #1 contains items which deal with three brands of compact cars (Ford Pinto, Chevrolet Vega, Mazda RX3) while questionnaire #2 contains items dealing with three brands of intermediate cars (AMC Matador, Chevrolet Malibu, Volkswagen Dasher). Several different types of questions were asked. First there were questions relating to brand preference (the dependent variable). In addition, there were twenty-four items per brand intended to measure the independent variables: (a) six items intended to measure the probability of the six types of loss from purchase of the brand; (b) six items intended to measure the importance of the six types of loss from purchase of the brand; (c) six items intended to measure the probability of the six types of gain from the purchase of the brand; and (d) six items intended to measure the importance of the six types of gain from the purchase of the brand. All items were scored on a seven point semantic differential scale (anchored "probable-improbable" and "important-unimportant") in order to make the data more interval-like (Moinpour and Wiley, 1972). Relative to the "gain" and "loss" facets an important point should be noted. Since no previous study had attempted to measure (1) both expected positive and negative utility and (2) both the probability and importance components separately for each of the six facets, the authors had to formulate operational definitions for this particular study. However, the definitions generated were based on conceptual discussions in the consumer behavior literature and operational definitions used in similar types of studies, primarily those of Fishbein (1967, 1971), Roselius (1971) and Jacoby and Kaplan (1972). All of these conceptualizations have face validity.

Data Collection

Data were collected from a convenience sample of 217 juniors and seniors enrolled in Business Administration curriculum at the University of Kentucky. A total of 210 usable questionnaires were obtained. Of this total, 108 dealt with the compact brands and 102 dealt with the intermediate brands. Since this was an exploratory study to test the efficacy of three related models the authors felt that the problem of external validity was not critical; nevertheless replication of this study should use a population which would permit the results to be generalized.

The perceived risk, perceived return and net perceived return indices for each of the six brands were formed in the following manner. The perceived risk indices were formed by multiplying the appropriate responses to the probability and importance of loss questions; the perceived return indices were formed by multiplying the appropriate probability and importance-of gain questions; and net perceived return indices were formed by subtracting each of the perceived risk indices from each of the perceived return indices.

The Analysis

Eighteen (3 conceptual models x 6 automobile brands) multiple regression runs [K. J. Johnson, BIGSTEP Regression Program Version DVT4, University of Kentucky. Lexington. Kentucky. 1970.] were performed using the brand preference scores as dependent variables. Each computer run involved two procedures. First, each set of six indices were forced into the regression equation and regressed against the appropriate brand preference score. This procedure was employed to determine the total amount of variance explained (r2) by the six facets. Second, a stepwise procedure was employed which brought into the regression equation only those independent variables (facets) significantly (p'.05) related to brand preference. This procedure was necessary to determine those facets of perceived risk, perceived return, and net perceived return which are significantly related to brand preference for each of the brands. In addition, examination of the beta coefficients for significant facets provides further insight into the relationships. Beta coefficients estimate the amount of change in brand preference which can be explained when a unit change in a facet is made while the T-ratio indicates whether this amount of change is significantly different than zero. Thus, the beta coefficients show the relative importance of each facet for each brand and the sign of the beta coefficient indicates whether the facet is directly or inversely related to brand preference. Conceptually, perceived risk should be inversely related to brand preference and perceived return and net perceived return should be directly related to brand preference; examination of the signs of the beta coefficients is a test of this notion.

RESULTS

Comparison of the Three Conceptual Models

Table 1 below shows all six brands studied along with the coefficients of determination for each of the three models. These r2's were generated by forcing all facets into the regression equation.

TABLE 1

BRAND PREFERENCE AS A FUNCTION OF PERCEIVED RISK, PERCEIVED RETURN, AND NET PERCEIVED RETURN: A COMPARISON OF R2'S FORCING ALL FACETS INTO THE REGRESSION EQUATION

In analyzing these data three points are noteworthy. First, for four of the six automobile brands (Pinto, Dasher, Malibu, and Matador) the net perceived return model explained more variance in brand preference than either the perceived risk or perceived return models. Second, for the two brands which the net model did not explain the most variance, the other models accounted for only one brand each and the net perceived return model in both cases explained the second largest amount of variance. Third, the average unadjusted and adjusted r2's were appreciably higher for the net perceived return model than for the other two models. Thus, if the amount of variance explained were the only available criterion for determining the validity of a model, based on this portion of the analysis, the net perceived return model appeared to be the preferred formulation.

Analysis of Significantly Related Variables- Perceived Risk Model

Table 2 below presents the perceived risk facets which were found to be significantly related to brand preference in the stepwise regression equation.

TABLE 2

BETA COEFFICIENTS, T-RATIOS, AND SIGNIFICANCE LEVELS FOR PERCEIVED RISK FACETS SIGNIFICANTLY RELATED TO BRAND REFERENCE FOR SIX AUTOMOBILE BRANDS

Relative to the data in the above table three points should be noted. First, it is important to recognize that every brand had at least one risk facet which was significantly related to it. Second, all beta coefficients for the significantly related perceived risk facets were negative as they logically should be. Third, no single facet of risk dominated all six brands; this indicates that perceived risk may not be only product specific as Cunningham (1967) has shown, but also brand specific which is reasonable and consistent with the literature.

Analysis of Significantly Related Variables--Perceived Return Model

Table 3 is similar to the previous table. It presents the perceived return facets which were found to be significantly related to brand preference in the stePwise regression equation.

Relative to the data presented in Table 3 four points should be noted. First, it should be noted that every brand had at least one facet of return significantly related to it. Second, all beta coefficients for the significantly related perceived return facets were Positive as one would logically expect.

TABLE 3

BETA COEFFICIENTS, T-RATIOS, AND SIGNIFICANCE LEVELS FOR PERCEIVED RETURN FACETS SIGNIFICANTLY RELATED TO BRAND PREFERENCE FOR SIX AUTOMOBILE BRANDS

Third, the fact that no single facet of return was significant for every brand suggests that perceived return may well be brand specific also which is not inconsistent with the theory. Fourth, none of the perceived return facets found significant were the same as the perceived risk facets found significant in Table 2. This suggests that consumers probably associate different dimensions of positive and negative utility with their brand preferences.

Analysis of Significantly Related Variables-- Net Perceived Return Model

Table 4 below is similar to the two previous tables. It presents the net perceived return facets which were found to be significantly related to brand preference in the stepwise regression equation.

Two points should be noted. First, the sign of all beta coefficients for the significantly related net perceived return facets were positive as one would expect if net perceived return and brand preference are truly directly related. Second, more net perceived return facets were found to be significantly related to brand preference than for either of the other two models. Twelve as opposed to ten for perceived return, and nine for perceived risk. Thus, if the validity criterion of the three models were based on the number of risk-return facets which were shown to be significantly related to brand preference, the net perceived return model would again be preferred.

TABLE 4

BETA COEFFICIENTS, T-RATIOS, AND SIGNIFICANCE LEVELS FOR NET PERCEIVED RETURN FACETS SIGNIFICANTLY RELATED TO BRAND PREFERENCE FOR SIX AUTOMOBILE BRANDS

CONCLUSIONS AND IMPLICATIONS

The basic purpose of this study was to investigate three alternative decision making strategies in terms of their relative ability to explain brand preference. The three strategies, minimization of perceived risk, maximization of perceived return, and maximization of net perceived return were conceptualized, measured, and analyzed as multiple objective models and the results of these analyses were compared. Our findings have clearly indicated that the net perceived return model could explain more of the variance in automobile brand preference than the other two models and that perceived risk was a more potent explainer than perceived return. In addition, three other conclusions can be drawn from the results of this study. First, consumers consider expectations of both positive and negative utility in their automobile brand preference decisions. Second, perceived risk, perceived return, and net perceived return are brand specific rather than product specific. Third, automobile brand preference varies inversely with perceived risk and directly with perceived return and net perceived return.

Although there was both conceptual and empirical support for the variables and models employed in this study, more research is needed. The authors offer three suggestions for further verification and improvement of the net perceived return model. First, as Edwards (1961) has previously noted in a discussion of decision making models, the major problem is that the rules of combination for the variables are not known, i.e., it is not known how consumers weight and combine variables to make decisions. Thus, although it was assumed in this study that probabilities and importances were combined in a multiplicative fashion, and that risk was subtracted from return to determine the net perceived return model, future conceptualization and research should investigate other methods of variable combination.

Second, although six facets of utility were determined for use in this study, there may be other important facets consumers consider. For example, expectations of gain and loss because of product obsolescence could conceivably be an important consideration. In addition, the social facet could perhaps be further disaggregated into expectations concerning friends, relatives, or other relevant reference groups.

Third, although this study attempted to replicate results using different samples, and different classes and brands of automobiles, much fruitful research could be done in further replication. In particular, different brands of automobiles should be studied. Different brands of other product categories should be considered. Wherever possible more reliable samples should be employed. Finally, replications should be done for different products and brands in different stages of the product life cycle although the present study attempted to do this by including two relatively new brands in the product sample, viz., the Mazda and the Dasher.

REFERENCES

Bauer, R. A. Consumer behavior as risk taking. In R. S. Hancock (Ed.), Dynamic marketing for a changing world. Chicago: American Marketing Association, 1960, 389-398.

Bilkey, W. J. A psychological approach to consumer behavior analysis. Journal of Marketing,1953, 17, 18-25.

Bilkey, W. J. Psychic tensions and purchasing behavior. Journal of social Psychology, 1955, 43, 247-257.

Copley, T. P. and F. L. Callom. Industrial search behavior and perceived risk. In D. M. Gardner (Ed.), Proceedings. College Park, Maryland: Association for Consumer Research, 1971, pp. 208-231

Cox, D. F. (Ed.). Risk taking and information handling in consumer behavior. Boston: Harvard University Press, 1967.

Cunningham, S. M. The major dimensions of perceived risk. In D. F. Cox (Ed.), Risk taking and information handling in consumer behavior. Boston Harvard University Press. 1967. 82-108.

Edwards, W. Behavioral decision theory. Annual Review of Psychology, 1961, 12, 473-498.

Fishbein, M. A. (Ed.) Readings in attitude theory and measurement. New York: Wiley, 1967.

Fishbein, M. A. Some comments on the use of 'models' in advertising research. In Proceedings: Seminar on translating advanced advertising theories into research reality. Amsterdam, The Netherlands: European Society of Market Research, 1971.

Hansen, F. Consumer choice behavior: A cognitive theory. New York: Free Press, 1972.

Hirsch, R. D., R. S. Dornoff and J. B. Kernan. Perceived risk and store selection. Journal of Marketing Research, 1972, 9, 434-439.

Jacoby, J. and L. B. Kaplan. The components of perceived risk. In M. Venkatesan (Ed.) Proceedings. Iowa City, Iowa: Association for Consumer Research, 1972, pp 382-393.

Johnson, K. BIGSTEP Regression Program Version DVT4, University of Kentucky, 1970.

Kogan, N. and M. A. Wallach. Risk taking: A study in cognition and personality. New York: Holt-Rinehart, 1964.

Lewin, K. Forces behind food habits and methods of change. Bulletin of the National Research Council, 1943, 108.

Moinpour, R. and D. L. Maclachlan. The relations among attribute and importance components of rosenberg-fishbein type attitude models: An empirical investigation. In D- Gardner (Ed.), Proceedings. College Park, Maryland: Association for Consumer Research, 1971, pp. 365-375.

Moinpour, R. and J. B. Wiley. An approach to the resolution of multicollinearity in the attribute structure of attitudes. In M- Venkatesan (Ed.) Proceedings. Iowa City, Iowa: Association for Consumer Research, 1972, pp. 341-348

Perry, M. and B. C. Hamm. Canonical analysis of relations between socioeconomic risk ant personal influence in purchase decisions. Journal of Marketing Research, 1969, 6, 351-354.

Roselius, T. Consumer rankings of risk reduction methods. Journal of Marketing 1971, 35 (1), 56-61.

Schiffman, L. G. Perceived risk in new product trial by elderly consumers. Journal of Marketing Research, 1972, 9, 106-108.

Sheth, J. An investigation of relationships among beliefs, affect, behavioral intentions and behavior. Unpublished manuscript, University of Illinois, 1970.

Sieber, J. E. and J. T. Lanzetta. Conflict and conceptual structure as determinants of decision making behavior. Journal of Personality, 1964, 32, 622-641.

Wilkie, W. L. and E. A. Pessemier. Issues in marketing's use of multiattribute attitude models. Journal of Marketing Research, 1973, 10, 428441.

Zikmund, W. G. An empirical investigation of the multidimensional nature of perceived risk and related variables. Unpublished doctoral dissertation, University of Colorado, 1970.

Zikmund, W. G. and J. E. Scott. A multivariate analysis of perceived risk, self-confidence and information sources. In M. J. Schlinger (Ed.), Proceedings. Chicago: Association for Consumer Research,1974.

----------------------------------------