Structural Properties of Consumer Information and Perceptions of Informativeness

ABSTRACT - This paper examines several structural properties of consumer information and assesses the impact these properties have on consumer perceptions of informativeness; how informative a particular set of consumer information is perceived to be when used in making a brand choice decision. The results of two experiments and replications of each show the importance that these structural properties have in contributing to perceptions of informativeness. In this study "information load" was a less important construct in contributing to perceptions of the informativeness of consumer information than other structural properties of consumer information introduced and evaluated.

Citation:

Roger J. Best and Daniel B. Williams (1980) ,"Structural Properties of Consumer Information and Perceptions of Informativeness", in NA - Advances in Consumer Research Volume 07, eds. Jerry C. Olson, Ann Abor, MI : Association for Consumer Research, Pages: 501-506.

Advances in Consumer Research Volume 7, 1980     Pages 501-506

STRUCTURAL PROPERTIES OF CONSUMER INFORMATION AND PERCEPTIONS OF INFORMATIVENESS

Roger J. Best, University of Arizona

Daniel B. Williams, University of Arizona

ABSTRACT -

This paper examines several structural properties of consumer information and assesses the impact these properties have on consumer perceptions of informativeness; how informative a particular set of consumer information is perceived to be when used in making a brand choice decision. The results of two experiments and replications of each show the importance that these structural properties have in contributing to perceptions of informativeness. In this study "information load" was a less important construct in contributing to perceptions of the informativeness of consumer information than other structural properties of consumer information introduced and evaluated.

INTRODUCTION

Consumer research, and in particular consumer information processing research have been greatly handicapped by the fact that it does not have a definition of "consumer information" (Wilkie, 1975). In the absence of a conceptual definition, consumer research has proceeded on the premise that the quantity of information is a valid and useful measure of "amount of information". However, we, and as well as others have taken the position that information load is only one property of consumer information that contributes to "informativeness in a consumer choice situation." The purpose of this paper is to introduce other quantifiable properties of consumer information and experimentally demonstrate their relative contribution to consumer perceptions of the informativeness of consumer information. Because "information load" plays an important role in consumer information processing research, it is also experimentally evaluated in terms of its contribution to the perceived informativeness of consumer information.

STRUCTURAL PROPERTIES OF CONSUMER INFORMATION

In general, "amount of information" has been treated as the quantity of potentially informative elements in a given set of consumer information. Jacoby, Speller, and Kohn (1974a, 1974b; Jacoby, 1977) have measured this property as the total number of elements in a given set of consumer information and called it "information load." The use of information load as a measure of "amount of information" resulted in a well-documented controversy. (Wilkie, 1974; Summers, 1974; Russo, 1974; Jacoby 1977). While the substantive issues of this controversy addressed questions of research design and method, the source of much of the disagreement can be traced to underlying differences at the conceptual level concerning what properly constitutes the "amount of information".

These criticisms raised two important points. First, "amount of information" depends on properties other than information load. Second, the "amount of information" present in any set of consumer information is determined not only by the data, but also by how consumers perceive the data and can be expected to use it in a brand choice situation.

Thus, while information load is perhaps the most apparent structural property of consumer information, a number of other distinct structural properties of consumer information can also be specified (Van Raajj, 1976). We have selected several specific properties which we believe might have an effect on consumer perceptions of informativeness. Each of these properties is described here along with a rationale for hypothesizing its effects on perceptions of informativeness in a brand choice decision.

Information Load

This term is used in this study to designate the total quantity of data provided a consumer in a brand choice decision. We have operationalized this construct as simply the quantity of information available in a brand choice decision. Furthermore we hypothesize that when all other things are equal (i.e. other properties of consumer information are held constant) consumers will perceive larger information loads to be more informative in brand choice than smaller information loads. It is important to keep in mind that we are not concerned with choice accuracy as a function of information load but with consumer perceptions of how informed they are as a function of information load. Thus, it seems plausible that consumers would perceive themselves to be more informed with greater quantities of information regardless of the information processing difficulties this might create.

Discrimination Power

This construct is used to designate a consumer's perception of the relative difference in the over-all attractiveness of alternative brands. The consumer's judgment of overall brand attractiveness is based on some combination of judgments of individual attributes describing each brand. If the relative difference in over-all attractiveness among brands is large, discrimination power is high; if the relative difference in overall attractiveness is small, discrimination power is low. Discrimination power appears to be the concept that underlies Summers' (1974) concern for the relationship between differences in the relative attractiveness of brands and the relative "difficulty" of consumer decision making. We hypothesized that consumers would perceive consumer information high in discrimination power to be more informative when making a brand choice decision than a set of consumer information Low in discrimination power.

Attribute/Brand Ratio

The dimensions of any given set of consumer information are defined by the number of brands (n) and the number of attributes (m). The relationship between these parameters in any set of consumer information can be expressed as an attribute/brand ratio (m/n) (Van Raajj, 1976). Because the set of attributes describing a given brand can be considered a subset of potential brand attributes, as the number of attributes increases the accuracy of estimated expected utility associated with a given brand can be expected to improve. Empirical studies have shown that increasing the number of attributes used to describe a fixed set of alternatives (i.e. increasing the attribute/brand ratio) corresponds with increases in choice accuracy (Russo, 1974; Scammon, 1977) and confidence in choice accuracy (Oskamp, 1965; Slovic, 1973). Based on these findings we hypothesize that consumer information with a higher attribute/brand ratio would be perceived as more informative than consumer information with a lower attribute/brand ratio.

Within Brand Attribute Variance

For any brand in a particular set of consumer information, the consumer may perceive the set of individual attributes to be approximately equal in value (e.g., all very good, all fair, or all very poor), or the consumer may perceive the set of individual attributes to be more or less discrepant in value (e.g., some very good and some very poor). The variability among individual attributes describing a particular brand can require different levels of effort to process the information as well as result in different levels of confidence in choice accuracy. For example, assume that brand A and brand B are both evaluated by a consumer and given the same overall brand evaluation. However, the relevant attributes of brand A are each judged as "good" (i.e. low within brand attribute variance) while evaluations of the same attributes for brand B range from "very good" to "very poor" (i.e. high within brand attribute variance). The consumer can perform the task of aggregating several "good" attribute evaluations into an overall "good" brand evaluation judgment with relatively little effort, little chance of computational error, and high level of confidence. However, the task of combining a mixture "very good" evaluations with "good" evaluations and "very poor" evaluations requires potentially complex trade-offs, weighting considerations and is likely to result in more effort, a higher chance of computational error, and lower level of confidence in choice accuracy. Thus, we hypothesize that low within brand attribute variance is likely to be perceived as more informative than high within brand attribute variance.

Overall Favorableness

We assume that consumers can synthesize an evaluation of a set of consumer information into a global judgment of the "overall favorableness" of the information. This global judgment can range along a continuum from extremely favorable (the values of all the information are judged to be very good or excellent) to extremely unfavorable (the values of all the information are judged to be very poor or unsatisfactory). We hypothesize that consumer information judged to be generally more favorable overall would be perceived by consumers to be more informative than consumer information judged to be generally less favorable overall.

This hypothesis is based on the assumption that one of the consumer's objectives in making a brand choice decision is to avoid selecting a brand that will prove unsatisfactory. Thus, the higher the "overall favorable-ness" of the information, the lower the probability of making a brand selection that might prove to be unsatisfactory. On the other hand, as the "overall favorableness" of a particular set of consumer information declines, we would expect the consumer to perceive greater risk that one or more of the available brands might prove to be unsatisfactory.

Imputed Attribute Importance

Depending on the nature of the product, its intended use, and the consumer's knowledge of the product class, the consumer may assign different importance weights to the various attributes. These attribute importance weights must be applied to attribute evaluations in order to assess the expected utility associated with a given brand. Therefore, evaluations of the more important attributes have a greater impact on the level of expected benefits associated with a given brand than evaluations of less important attributes. We hypothesize that consumer information in which more important attributes were evaluated more favorably than less important attributes would be perceived by consumers to be more informative than consumer information in which more important attributes were evaluated less favorably than less important attributes.

EMPIRICAL EVIDENCE: METHOD AND RESULTS

Two separate experiments were designed to test the general hypothesis that variations in the structural properties of consumer information described, have a significant effect on consumer perceptions of informativeness. The first experiment was an exploratory study that examined the relationships between several properties of consumer information and perceptions of informativeness. The second experiment examined the relationship between consumer perceptions of informativeness and two specific properties that proved significant in the first study and information load.

In both experiments semantic content of the consumer information was controlled by identifying alternative brands alphabetically and the various product attributes numerically. The purpose of this control was to reduce the variability in response that could result from interindividual differences in knowledge of or interest in a product class.

As an additional means of controlling for interindividual differences in information processing, values of individual brand attributes were presented in a "preprocessed" mode. That is, each attribute value was presented in the form of the output of an evaluative judgment process (i.e., very good, good, fair, poor, very poor) rather than in the form of a descriptive input (e.g., $200, 30 pounds, all-alloy construction, etc.). This "preprocessed" mode of presentation was intended to reduce the variability in the interpretation of descriptive data that could result from interindividual differences in processing criteria.

In the first experiment a set of eight information displays were constructed using three structural properties of consumer information. Each property was set at two levels, one presumed to be high in perceived informativeness, the other low in perceived informativeness. In the second experiment, a set of 12 information displays was constructed using as treatment variables information load and the two other properties whose effects on perceived informativeness were significant in the first experiment. The values of the two other properties were each set at two levels, one high and the other low in perceived informativeness. Information load was arbitrarily set at three levels which were assumed to be substantially different from one another in perceived quantity of information. Table 1 lists the treatment variables employed in both experiments, provides a brief definition of each, and describes the procedure by which each variable was measured.

In both experiments, subjects were presented with a sequence of pairs of information displays. The subject was instructed to examine each of the two information displays and asked to judge which of the two information displays was "more informative for the purpose of making a brand choice decision." This comparative judgment of consumer information displays was performed for all possible pairs of displays constructed in each experiment.

Experiment I

In the first experiment three treatment variables were selected for manipulation; discrimination power (DP), attribute/brand ratio (AB), and within brand attribute variance (AV). High and low levels were specified for each treatment variable and eight information displays were constructed in accordance with the design shown in Table 2. After each information display was constructed, measures of "overall favorableness" and "imputed attribute importance" were also made using the operational definitions shown in Table 1.

TABLE 1

CONCEPTUAL DEFINITIONS AND OPERATIONAL MEASURES OF INDEPENDENT VARIABLES

TABLE 2

RESEARCH DESIGN USED IN EXPERIMENT I

TABLE 3

RESEARCH DESIGN USED IN EXPERIMENT II

All possible pairs of these information displays (28) were randomized according to position and order. These displays were then presented to 25 subjects who were instructed to evaluate each pair of information displays and indicate which display provided them "more information for the purpose of making a brand choice decision." Therefore, each subject made 28 comparative judgments involving all combinations of the experimental treatments shown in Table 2.

The pairwise judgments of information displays made in terms of perceived informativeness were used to construct a Thurstone type scale. The Thurstone scale shown in Figure la illustrates the position of each information display used in experiment I along a continuum that varies in perceived informativeness. In this case information displays 6 and 7 were perceived to be considerably more informative than displays 1 and 2. To evaluate the effect each treatment variable had on the perceived informativeness a univariate F-ratio was computed for each of the three treatment variables shown in Table 2 and two other structural properties that were measured for each information display but not systematically manipulated. This analysis revealed that discrimination power (DP), overall favorableness (F), and imputed attribute importance (AI) each explained significant portion of variance in perceived informativeness (p < .10) while attribute/brand ratio (AB) and within brand attribute variance (AV) were not significant (p = .20) in explaining variance in perceived informativeness. A stepwise regression analysis of these variables produced the following multivariate relationship.

PI = -3.206 + .16 DP + .9F + 33AI; Adjusted R² = .92    (1)

         (5.41)     (6.43)   (4.72) (3.62)

In this case the same variables that were significant as univariate explanatory variables combined to produce a multivariate equation which explained 92% of the variance in perceive informativeness shown in Figure 1a. Each coefficient was statistically significant at the .01 level (t-statistics shown in parentheses). Thus, two of the experimental treatments, attribute/brand ratio (AB) and within brand attribute variance (AV) were not significant (p = .20) either as univariate or multivariate predictors of perceived informativeness.

FIGURE 1

"PERCEIVED INFORMATIVENESS" OF INFORMATION ARRAYS CREATED IN EXPERIMENT I

Since the degrees of freedom used to estimate this regression were quite small for estimating the parameters of a three variable equation, the reliability and validity of this equation can be questioned. Therefore, a manipulation and replication of this experiment was performed. Information display A₂ was modified such that a new display, A₂^' was created. Display A₂^' was constructed se that it had greater discrimination power, greater overall favorableness, and greater imputed attribute importance than array A₂. With these changes regression equation (1) was used to predict the perceived informativeness of display A₂^'. By this procedure display A₂^' was predicted to be between arrays A₆ and A₇ on the perceived informativeness scale shown in Figure la.

To evaluate this prediction, the original instrument was altered by substituting the modified display A₂^' in place of original array A₂. Approximately one month after the initial treatment the same 25 subjects performed the experimental task again using the modified instrument. The data generated in this manipulation and replication phase of Experiment I were used to construct a second Thurstone scale of "perceived informativeness" which is shown in Figure lb. As figure 1b shows, changes in these three structural properties repositioned the modified display A₂^' precisely between displays A₆ and A₇ as predicted. In this case the predicted and actual scale positions for display A₂^' were .53 and .42, respectively.

In terms of aggregate reliability, the correlation between the respective scale positions of each of the unchanged displays shown in Figure 1a and 1b was .95. With respect to individual reliability, the outcomes of each judgment task in the first treatment agreed with the corresponding outcome in the second treatment in 76% of the cases. Finally, the scale values of perceived informativeness shown in Figure 1b were regressed on the same three predictor variables and the following equation was derived.

PI = -2.485 + .16 DP + .62F + .40IA; Adjusted R² = .905    (2)

         (4.05)    (6.10)    (3.16)   (4.38)

As in regression equation (1) each of the three variables in equation (2) were significant (t-statistics shown in parentheses) and the three variables collectively explained a large portion of the variance in perceived informativeness. Also, in terms of coefficients' stability, the coefficients of the two regression equations were numerically very similar (differences nonsignificant).

Experiment II.

A second experiment was designed to evaluate the influence of the two most significant properties from Experiment I while information load was varied systematically. The principal purpose of this experiment was to examine the robustness of these two properties of consumer information in the face of changes in information load (IL).

As shown in Table 3, information load (IL) was varied across three levels while discrimination power (DP) and overall favorableness (F) were each varied across two levels of treatment. This design produced the 12 consumer information displays.

As in the first experiment, an instrument was created by randomizing the sequence of all possible pairs of these 12 information arrays. Twenty-six new subjects selected the information array of each pair that provided them with "more information for the purpose of making a brand choice decision."

The data generated from these evaluations and comparative judgments were used to construct the Thurstone scale shown in Figure 2a. The scale positions of each information display depicted in Figure 2a were then regressed on the three treatment variables used to construct the information arrays. The results of this analysis are shown below:

PI = 4.737 + .375 DP + 1.12F + .054 IL; Adjusted R² = .84    (3)

        (5.15)     (4.92)      (3.74)     (1.53)

In this case, the three treatment variables accounted for 84% of the variation in overall subject "perceptions of informativeness". However, information load (IL) had the weakest relationship (p = .167) and contributed least to perceptions of informativeness.

As in Experiment I, a replication and manipulation was performed. However, this time two arrays were modified, information display A₁ was replaced with display A₁^', to enhance the informativeness of this array by improving its "discrimination power" and "overall favorableness" while not changing its information load. Display A₁₂ was replaced with display A₁₂^' in an attempt to depress the perceived informativeness of this consumer information. In both of these manipulations the values of the discrimination power (DP) and overall favorableness (F) variables were changed while information load was held constant. On the basis of these changes in structural properties, regression equation (3) was used to predict "perceived informativeness" for the modified displays A₁^' and A₁₂^'. The predicted scale positions for arrays A₁^' and A₁₂^' were 1.75 and -2.12, respectively.

FIGURE 2

"PERCEIVED INFORMATIVENESS" OF INFORMATION ARRAYS CREATED IN EXPERIMENT II

A second instrument was created to incorporate these changes, and the same 26 subjects were asked again to evaluate all possible pairs of the information arrays and, in each case, indicate which information array provided them "more information for the purpose of making a brand choice decision." The data obtained from this judgment task were used to construct a second Thurstone scale shown in Figure 2b. The relative scale positions of the modified arrays derived from the experimental data were close to positions predicted by the regression equation. The observed position of A₁₂^' was below A₂ as predicted. However, the observed position of A₁^' fell short of exceeding the perceived informativeness of display A₁₁, as predicted. Regressing the scale positions for arrays shown in Figure 2b on the same predictor variables produced the following regression equation.

PI = -3.266 + .190 DP + .810F +  .051IL; Adjusted R² = .68    (4)

         (3.95)     (3.42)       (3.58)       (1.76)

As in the first experiment each of the predictor variables representing the other structural properties of consumer information were statistically more significant than information load (which was significant in the regression analysis at p = .12). Also, as in the other regression analyses, this equation explains a significant portion of the variance in aggregate perceptions of informativeness.

DISCUSSION

The principal finding of these experiments is that other structural properties of consumer information can have a significant and predictable effect on the perceived informativeness of consumer information. In addition the manipulation of these structural properties of consumer information can be made to overwhelm completely the effects of information load as demonstrated in Experiment II.

Our findings, although positive, are preliminary and subject to significant limitations. First, the use of information without semantic content presented in a preprocessed mode is artificial and raises questions concerning external validity. Second, the validity of the dependent variable "perceived informativeness" requires further study as a construct both in terms of identifying its components and the relationships among them, and in terms of clarifying the relationship between the decision making process and the perception of informativeness. Third, the Thurstone scaling procedure employed for measuring the dependable variable aggregates across individuals. The use of alternative measurement approaches sensitive to individual variations are needed to make the perceived informativeness construct more credible. Finally, the relationship between variations in the structural properties examined in this study and perceived informativeness were rather crudely measured. This is partly the result of using only two levels, "high" and "low," for each independent variable, but it is also a function of a certain arbitrariness that is unavoidable in the initial operationalization of structural properties such as those examined in this study. Each of these limitations presents an opportunity for further testing and clarification of the constructs and hypotheses presented in this paper.

CONCLUSIONS

The two experiments described in this study demonstrate that other structural properties of consumer information can have a significant effect on perceptions of informativeness. Of particular importance is that in the second experiment two other structural properties were manipulated to overwhelm the effect information load had on perceptions of informativeness derived from a set of information to be used in making a brand choice decision.

It is important to keep in mind that in this study the criterion variable of interest was perceptions of informativeness. Therefore, while these results suggest that information load may be a weaker variable in affecting the perceived informativeness of consumer information, this result does not mean that information load may not strongly influence other criterions such as choice accuracy. However, perceptions of the informativeness of consumer information and knowledge of those structural properties of information that influence these perceptions offer considerable potential in the study of consumer information seeking behavior and consumer decision making.

REFERENCES

Jacoby, Jacob (1977), "Information Load and Decision Quality: Some Contested Issues," Journal of Marketing Research, 14, 569-573.

Jacoby, Jacob, Speller, Donald E., and Kohn, Carol A. (1974a), "Brand Choice Behavior as a Function of Information Load," Journal of Marketing Research, 11, 63-69.

Jacoby, Jacob, (1974b), "Brand Choice Behavior as a Function of Information Load: Replication and Extension,'' Journal of Consumer Research, 1, 33-42.

Oskamp, S. (1965)., "Overconfidence in Case-Study Judgments," Journal of Consulting Psychology, 29, 261-265.

Russo, J. Edward (1974), "More Information Is Better: Re-evaluation of Jacoby, Speller, and Kohn," Journal of Consumer Research, 1, 68-72.

Scammon, Debra L. (1977), "Information Load and its Effects on Consumers," Journal of Consumer Research, 4 148-155.

Slovic, Paul (1973), "Behavioral Problems of Adhering to a Decision Policy," paper presented at the Institute of Quantitative Research in Finance, Napa, California.

Summers, J. "Less Information is Better," Journal of Marketing Research, (November 1974), 467-68.

Van Raajj, W. Fred (1976), "Consumer Information Processing for Different Information Structures and Formats," in Advances in Consumer Research Vol. 4, ed. William D. Perreault, Jr., Atlanta: Association for Consumer Research, 1976-184.

Wilkie, William L. (1974), "Analysis of Effects of Information Load," Journal of Marketing Research, 11, 462-466.

Wilkie, William L. (1975), How Consumers Use Product Information: An Assessment of Research in Relation to Public Policy Needs, report prepared for National Science Foundation, Washington, D.C.

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