# Marginal Salience of Price in Brand Evaluations

##### Citation:

*
Vithala R. Rao (1972) ,"Marginal Salience of Price in Brand Evaluations", in SV - Proceedings of the Third Annual Conference of the Association for Consumer Research, eds. M. Venkatesan, Chicago, IL : Association for Consumer Research, Pages: 125-144.
*

[The author expresses his thanks to Messers. Michael Harper and Leonard Fertuck for their assistance in data collection and analysis and the Office of Sponsored Research of Cornell University for partial financial support.]

INTRODUCTION

The various brands in any product class can be described as a set of multi-attribute alternatives. These attributes may include both the significative and symbolic aspects (Coombs, Dawes, 6 Tversky, 1970). Consumer's decision to choose a brand can be thought of as a composite decision rule incorporating the various attributes of the brands. These rules obviously reflect the individual's idiosyncratic utility function.

One critical attribute of the brand that has been intensively studied by economists is brand price. It is only in the past decade or so that researchers of consumer behavior in marketing turned their attention to studying price as a perceptual dimension of evaluations with respect to brand quality and brand worth (defined as some measure of quality per unit price.) Past research (Gabor & Granger, 1966, Gardener, 1971, Jacoby, 1970, Leavitt, 1954, McConnell, 1968 a & b, Peterson, 1970, Rao, 1972, and Tull, Boring & Gonsoir, 1964) has indicated that price is used by consumers as a surrogate for quality in the absence of other brand information and that the importance of price in quality perceptions diminishes when a number of other brand cues are present.

Thus it appears that the salience of price is not independent of other brand information. This paper is a preliminary attempt to determine the nature and magnitude of the trade-off relations between importance of price and the number of non-price informational cues. The research question under study is whether there exists an optimal number of additional brand cues in the presence of which price is least salient. Obviously, the existence of such a number has pragmatic implications for promotional planning and public policy. The paper is organized into three broad section. (1) formulation of the hypothesis and discussion of the measurement model employed; (2) discussion of the experimental design and analysis procedure; and (3) results and possible implications.

HYPOTHESIS AND MEASUREMENT

The substantive hypothesis under examination is that consumer's reliance on price in making overall brand evaluations is nonlinearly related to the number of additional non-price informational cues on brands. As previous research indicates, not surprisingly when no other information is present consumer rely solely on price in their brand evaluations. Furthermore, as information on other attributes becomes available, they tend to utilize such information thereby decreasing the importance assigned to price. However, when the additional cues are too many (beyond the individual's cognitive limits with respect to information overload), the individual will tend to revert to relying upon generally accepted index such as price. That is to say, the importance assigned to price will increase after a certain number of additional non-price cues. This hypothesis may be shown as in Figure 1. The curve depicts the way in which importance on price is related to the number of additional cues. It reaches a minimum at the point k additional cues and rises thereafter.

HYPOTHESIZED RELATIONSHIP BETWEEN RELIANCE ON PRICE AND NUMBER OF NON-PRICE BRAND CUES

The model of additive conjoint measurement, developed by mathematical psychologists, (Coombs et. al., 1970, Luce, 1964, Roskam, 1968, Tversky, 1967 a & b) is employed in this paper for measuring the importance of price. This model is based on a paradigm where an ordering of a dependent variable (namely overall evaluation of preference) is obtained under different combinations of two (or more) independent variables (price and nonprice attributes). To make the discussion concrete, let us consider the product class of automobiles and a set of m.n hypothetically described auto stimuli on two attributes (A) size and (P) price respectively, of m ant n levels and identical in all other respects. Each stimulus has two coordinates (a., p.), i = 1, 2,...,m; i = 1, 2, ..., n. Assume that the mn stimuli have bean ordered by an individual with respect to his overall opinion of worth of the automobile to him. Let M(ai, p.) represent the ordinal measure of his evaluation. The additive model of conjoint measurement assumes that there exist functions f, g, and o defined on A = (ai, i = 1, 2, ...,m); P = (Pj, i = 1, 2, ...,n) and AXP respectively, such that

1. o(ai, pj) = f(ai) + g(pi)

2. o(ai, pj) > o(ak, p1) if and only if M(ai, pj) > M(ak, p1)

Thus, if the model holds, the cell entries of M-matrix can be rescaled such that their order is preserved and such that every rescaled entry is expressed as the sum of the components of the two attributes, A(size) and P(price). Such a model exists if the axioms of cancellation and solvability are satisfied (Coombs et. al., 1970). Under these conditions, the two attributes A and P can be regarded as independent in the sense that they contribute independently (additively) to produce the joint effect. This model is quite similar to that of the analysis of variance with no significant interactions. The variation in rescaled values of M-matrix can be written as the sum of the variations contained in the component functions (hereafter called partworth functions) of the attributes A and P. The operational measure of the importance of price in overall judgments is then the percent of the total variation in rescaled values of M due to attribute, P:

Importance of price = __Sum of squares attributable to price__

Total sum of squares of rescaled M-matrix

Computer programs exist to perform the above analysis on ranked judgments of stimuli derived according to a factorial design. The program developed by Kruskal (Kruskal, 19653, known as MONANOVA which does monotone analysis of variance is employed in this study. This program computes [MONANOVA program analyses data from a factorial experiment (fractional or otherwise) and finds the transformation which reduces the interaction as much as possible. It finds a monotone transformation of the data so as to achieve the highest possible percentage of variance accounted for by main effects. The program can handle any number of missing observations.] a measure of goodness to fit, known as stress (S) defined by:

where Zi denotes monotonically transformed values of the dependent variable (i.e., ranked judgments); Zi(B) denotes the parameters of the fitted model and Zi(B) denotes the mean of Zi(B). The lower limit of this measure (representing perfect fit) is zero.

EXPERIMENTAL DESIGN AND ANALYSIS

To understand the dynamics of the trade-off between price and nonprice attributes, hypothetically stated brands were used as stimuli. Such a decision avoided the confounding, if any, due to the image conjured up by brand names. The stimuli were automobiles. Each stimulus was described as a vector of attributes.

Based on a preliminary investigation, six attributes (including price) of the automobile considered most important in brand choice were selected. In addition to price, these were: (A) size; (B) horsepower; (C) miles per gallon; (D) repair record; and (E) origin of manufacture. Except for the origin of manufacture, which was described at 2 levels, four levels were selected for the attributes. These are shown in Table 1.

LEVELS SELECTED FOR ATTRIBUTES

In addition to price (P), four combinations of attributes, namely A and B; A, B, and C; A, B, C, and D; and A, B, C, D, and E were used in developing stimulus descriptions. For each combination, 16 hypothetical stimuli were generated using the principles of a Graeco-Latin Square design. For the first three sets, the attributes were combined using each of the four levels making sure that the successive designs were orthogonal to each other. For the fourth set, 1/4 factorial was selected employing the two extreme levels for the five attributes other than origin of manufacture for which the two levels (American and Foreign) were used. For each combination, the stimuli were described without and with price. The design thus enabled comparison of the price influence, not only within each combination but across the various combinations. The layout of the stimuli is presented in Table 2. Each stimulus was described on a 4" x 5" card as a profile description of an automobile. Subjects were presented two decks of 16 stimulus cards each in the experiment.

Four different groups of subjects (residents of the Ithaca area including some graduate students of Cornell University) participated in this experiment. [Some description of the sample characteristics may be in order here. Although not reported in the main paper, data were collected on two background characteristics, namely, annual income and years of driving experience. The means and standard deviations of these for the experimental groups are shown below:

In addition, data on a set of 12 attribute-interest-opinion questions tapping the individual's attitude toward cars, pollution and car maintenance were collected. These data (after transforming into factor scores) were employed to discriminate the four groups of subjects with the following results:

The rate of correct classification was 34 percent.]

Respectively 24, 36, 27 and 31 subjects (totalling 118) responded to the 2, 3, 4 and 5 attribute stimuli. The experimental tasks included the following:

(a) Ranking of the 16 automobile descriptions without price data from 'excellent' to 'poor' buy for the money assuming the cars were identical on all other characteristics including price.

(b) Ranking of the 16 automobile descriptions with price information including from 'excellent' to 'poor' buy for money assuming that cars were all identical on all other characteristics.

The ranked judgments of worth of the car excluding and including price information were analyzed for each individual using the Kruskal's method of monotone analysis of variance, known as MONANOVA (Kruskal, 1965). The analysis yielded partworth (utility) functions for each attribute included in the set. For example in the 2-attribute group, analysis of ranks yielded partworth functions for the attributes of size, horsepower, using the first set of ranks and the partworth functions for price also, using the second set of ranks. From these functions, the proportion of total variation due to each attribute before and after price was computed. The measure of reliance on price was simply the proportion of variation attributable to price ranked preferences obtained after price information.

Based on these individual analyses, subjects were divided into two groups: consistent and others. The consistent subjects were those satisfying the two criteria: (a) satisfactory fit of the model to their data with stress levels of 0.10 or below; and (b) the monotonicity of their partworth function of price (i.e.,high values of partworth for low values of price and decreases in the expected direction). The number of such consistent subjects in the 4 groups were respectively 14 out of 24; 21 out of 36; 21 out of 27 and 20 out of 31. Further summarization of data was done separately for consistent subjects and all subjects, discussion being confined mainly to the former group.

STIMULUS DESCRIPTIONS FOR THE 4 GROUPS

RESULTS

The average partworth functions derived from the preference judgments excluding and including price for the consistent and all subjects are presented in Tables 3-6 for the 4 groups responding to 2, 3, 4, and 5 nonprice attributes. The same data are also graphically represented for the consistent subjects in Figures 2-5.

The following descriptive observations based on the spread and direction of the partworth functions of the various attributes for the consistent subjects may be noted:

(a) In all cases, size is the least salient factor in the total preference. In general, the subjects evince low preference (utility) for full and mini sizes in their judgments excluding price. The pattern after price is monotonically decreasing with the size level.

(b) Subjects' utility for horsepower is highest in the 2 and 3 attribute group in their judgments excluding price. In the 4 and 5 attribute case, the horsepower continues to be high, but is dominated by repair record. Furthermore, the pattern is about the same in judgments including price. The partworth function is nonlinear and is increasing with horsepower becoming asymptotic after 250 h.p. or so.

(c) The function for gas mileage is monotonic with mpg. as expected.

(d) The attribute of repair record dominates others in 4 and 5 attribute cases. Its specific function is monotonically decreasing with the quality of repair record, as should be expected.

(e) The origin of manufacture does not turn out to be a salient factor. However, subjects express higher utility for foreign made cars than domestic ones.

(f) Not surprisingly, in all cases the utility expressed for price attribute diminishes with price.

In order to obtain a quantitative measure of importance assigned to each attribute, the percent contributions of each attribute, the percent contributions of each were computed. As noted earlier these are the percent total sum of squares due to each factor. The means and standard errors of these statistics applicable to the consistent subjects are presented in Table 7. Figure 6 shows the percent reliance on (contribution of) price as a function of number of non-price attributes. While no formal statistical tests are possible, the figure does provide empirical support for the hypothesis under study. The percent reliance on price does diminish monotonically up to the 4 non-price attributes and rises at 5.

Fortunately, owing to the nature of the design, similar analysis is possible for two of the non-price attributes of size and horsepower. In the case of size, the percent reliance decreases steadily from 11.5 percent for the 2 attribute group to 5.5 percent for the 5 attribute case. The pattern is more pronounced for the horsepower attribute the range being 42.7 percent to 15.0 percent. Thus, the importance given to these non-price variables drops steadily with the number of other attributes not showing any signs of recovery as noted for the price variable. This finding enhances the face validity of the hypothesis under examination.

PARTWORTH FUNCTIONS FOR JUDGMENTS OF CONSISTENT SUBJECTS IN 2 ATTRIBUTE GROUP

PARTWORTH FUNCTIONS FOR JUDGMENTS OF CONSISTENT SUBJECTS IN 3 ATTRIBUTE GROUP

PARTWORTH FUNCTIONS FOR JUDGMENTS OF CONSISTENT SUBJECTS IN 6 ATTRIBUTE GROUP

PERCENT CONTRIBUTIONS OF VARIOUS NON-PRICE AND PRICE ATTRIBUTES BY GROUP FOR CONSISTENT SUBJECTS

PERCENT RELIANCE ON PRICE IN JUDGEMENTS OF CONSISTENT SUBJECTS BY NUMBER OF NON-PRICE ATTRIBUTES

CONCLUSIONS

While the above analysis is indicative of ways in which individuals process attribute information on alternative stimuli, the results need to be replicated with more representative samples and with real brand names to be of use in marketing planning and policy formulation. However, two very tentative implications may be noted. First, from the public policy point of view, this study points out some directions as to the number of non-price attributes on which information should be disseminated to consumers enabling a better purchase decision. Furthermore, more attention ant research seems necessary as to which non-price attributes should be emphasized. A similar implication follows for promotional planning by marketers of brands.

REFERENCES

Coombs, Clyde H., Dawes, Robyn M. 6 Tversky, Amos. Mathematical Psychology An Elementary Introduction. Englewood Cliffs, N. J.: Prentice-Hall, 1970.

Coombs, C. H. A Theory of Data, New York: John Wiley and Sons, 1964.

Gabor, A. 4 Granger, C. W. J. Prices as an Indicator of Quality: Report on an Inquiry. Economica, 1966, 33, 43-70.

Gardener, David W. Is There a Generalized Price-Quality Relationship? Journal of Marketing Research, 1971, 8, 241-3.

Green, Paul E., Carmone, Frank J., X Wind, Yoram. Consumer Evaluation of Discount Cards: A Conceptual Model and Experimental Test. Working Paper, University of Pennsylvania, February, 1971.

Green, Paul E. X Rao, Vithala R. Conjoint Measurement for Quantifying Judgmental Data. Journal of Marketing Research, 1971, 8, 355-63.

Howard John A. 4 Sheth, Jagdish N. The Theory of Buyer Behavior. New York: John Wiley and Sons, 1969.

Jacoby, Jacob, Olson, Jerry C. 4 Haddock, Rafael A. Price, Brand Name, and Product Composition Characteristics as Determinants of Perceived Quality. Purdue Papers in Consumer Psychology, Paper No. 111, 1970.

Kruskal, Joseph B. Analysis of Factorial Experiments by Estimating Monotone Transformations of the Data. Journal of the Royal Statistical Society, Series B, 1965, 27, 251-63.

Leavitt, Harold J. A Note on Some Experimental Findings About The Meaning of Price. Journal of Business, 1954, 28, 205-10.

Luce, R. Duncan. Simultaneous ConJoint Measurement: A New Type of Fundamental Measurement. Journal of Mathematical Psychology, 1964, 1, 1-27.

McConnell, J. Douglas. An Experimental Examination of the Price-Quality Relationship. Journal of Business, 1968a, 40, 439-44.

McConnell, J. Douglas. The Price-Quality Relationship in an Experimental Setting. Journal of Marketing Research, 1968b, 5, 300-3.

Peterson, Robert A. The Price-Perceived Quality Relationship: Experimental Evidence. Journal of Marketing Research, 1970, 7, 528-8.

Rao, V. R. Salience of Price in the Perception of Product Quality- A Multidimensional Measurement Approach. In Niel H. Bordon, Jr. (ed.), Proceedings of the 1971 Fall Conference of American Marketing Association, Chicago, Illinois (1972).

Roskam, E. E. Metric Analysis of Ordinal Data in Psychology. Van Voorschoten, 1968.

Scitovsky, Tibor. Some Consequences of the Habit of Judging Quality by Price. Review of Economic Studies, 1944-45, 12, 100-05.

Stafford, James E. 6 Enis, Ben M. The Price-Quality Relationship: An Extension. Journal of Marketing Research, 1969, 6, 456-8.

Tull, S. S., Boring R. A. 4 Gonsoir M. H. A Note on the Relationship of Price and Imputed Quality. Journal of Business, 1964, 38, 186-91.

Tversky, Amos. A General Theory of Polynomial Conjoint Measurement. Journal of Mathematical Psychology, 1967a, 4, 1-20.

Tversky, Amos. Additivity, Utility and Subjective Probability. Journal of Mathematical Psychology, 1967b, 4, 175-201.

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

##### Authors

Vithala R. Rao, Cornell University

##### Volume

SV - Proceedings of the Third Annual Conference of the Association for Consumer Research | 1972

##### Share Proceeding

## Featured papers

See More#### Featured

### P2. The Upside of Myopic Loss Aversion

Daniel Wall, Carnegie Mellon University, USA

Gretchen Chapman, Carnegie Mellon University, USA

#### Featured

### Product Transparency in Online Selling Mechanisms: Consumer Preference for Opaque Products

Lucas Stich, Ludwig-Maximilians-University Munich

Martin Spann, Ludwig-Maximilians-University Munich

Gerald Häubl, University of Alberta, Canada

#### Featured

### The Upside of Incompetence: How Discounting Luxury Affects Retailer Price Image

Karen Wallach, Emory University, USA

Ryan Hamilton, Emory University, USA

morgan k ward, Emory University, USA