The Effects of Unfavorable Product Rating Information

ABSTRACT - A problem of both practical and theoretical importance to marketers is the impact of the increased amount of unfavorable product information to which consumers are exposed. To study the effects of unfavorable information 240 women were presented with rating information describing the rating source, unbranded goods and services and the actual favorable or unfavorable rating evaluations. Results indicate significant main effects for rating type (favorable/unfavorable) and product type (good/service). Significant interactions for rating source x rating type as well as rating source x product type were also found.

Citation:

Marc G. Weinberger and William R. Dillon (1980) ,"The Effects of Unfavorable Product Rating Information", in NA - Advances in Consumer Research Volume 07, eds. Jerry C. Olson, Ann Abor, MI : Association for Consumer Research, Pages: 528-532.

Advances in Consumer Research Volume 7, 1980     Pages 528-532

THE EFFECTS OF UNFAVORABLE PRODUCT RATING INFORMATION

Marc G. Weinberger, University of Massachusetts/Amherst

William R. Dillon, University of Massachusetts/Amherst

ABSTRACT -

A problem of both practical and theoretical importance to marketers is the impact of the increased amount of unfavorable product information to which consumers are exposed. To study the effects of unfavorable information 240 women were presented with rating information describing the rating source, unbranded goods and services and the actual favorable or unfavorable rating evaluations. Results indicate significant main effects for rating type (favorable/unfavorable) and product type (good/service). Significant interactions for rating source x rating type as well as rating source x product type were also found.

INTRODUCTION

The reality of unfavorable product or company information issued for public consumption is a phenomenon that has become increasingly problematic for marketers. Seemingly on a daily basis one hears of safety or health deficiencies in automobiles, soft drinks, fast foods, hamburgers, candy and numerous other products. At times the information is traceable to consumer groups, government testing agencies, regulatory bodies, competitors or environmental groups, while in other instances the negative information might just be a matter of rumor. It appears, therefore, that the impact of such favorable information and strategies to deal with it are particularly important issues facing the marketer. In the remainder of this section we briefly review the literature and discuss several possible theoretical justifications for the greater impact of negative information.

Background Literature

Psychologists have examined the impact of negative and positive information in the context of forming inferences or impressions about people and objects. In the typical impression-formation paradigm subjects are presented with a set of trait descriptors regarding various characteristics about a person or object, and asked to combine information about isolated traits into an overall evaluation of the stimulus object.

The differing effects of negative and positive information is well documented in the psychology literature. For example, Goodman (1950) found that trait words such as "cold" were more powerful in influencing impressions than positive trait words such as "warm" Similarly, Osgood, Succi and Tannenbaum (1957) found that, contrary to their proposed congruity principle, equally polarized positive and negative information did not have a balancing effect on impression formation; rather, in every instance of a reported error the direction of influence favored the negative information. The relatively greater impact of negative vis-a-vis positive information has also been shown to be unaffected by the subject's familiarity with the target object (Richey, Richey and Thieman, 1972) and, in addition, to affect a subject's overall emotions (Anderson, 1965) as well as behavioral intention (Posavac, 1974) toward the target persons described. Finally, in a more recent study, Richey, Koenigs, Richey and Forgin (1975) found that a single negative behavior could neutralize as many as five positive behaviors.

Consistent with the psychology literature, several marketing studies (Arndt 1968; Wright 1974) have found a greater reliance on negative information. The major focus of these studies, however, was not on assessing the impact of negative information and, therefore, they do not adequately address the issue of the differential impact of negative versus positive product information on the likelihood of purchase.

Theoretical Explanations

Several theoretical justifications for the greater impact of negative information have been offered. Feldman (1966), for example, argued that the greater impact of negative information is the result of its greater surpriseness. Zajonc (1968) theorized that negative information is less frequently utilized than positive information and as a result of this lower exposure rate, an extreme negative weighting of the information would result. The rationale for the Zajonc and Feldman arguments is that, because negative information runs counter to expectation it carries greater information value for subjects.

Kinder (1971) and Kanouse and Hanson (1971) discounted surpriseness and frequency of usage as the determinants of greater negative informational impact. Rather, Kanouse and Hanson suggested that an attributional framework is a more appropriate explanation. It suggests that negative information about a stimulus object stands out more than positive information simply because there are more positive cues in the individual's social environment. As a result, negative cues tend to attract more attention and, therefore, are more heavily attributed to the stimulus object (Kanouse and Hanson 1971).

In their correspondent inference theory Jones and Davis (1965) looked at an observer making inferences based upon perceived actual or situational causes. If an actor behaves in an expected fashion then it is hard for an observer to make a correspondent inference about the disposition of the actor. When the actor departs from the norm of expected behavior, however, the action provides better dispositional information to the observer. According to correspondence theory this occurs when the observed behavior was unusual or unexpected. The unusual information about the actor provides the observer with meaningful information about the actual disposition of the actor.

Essentially, Feldman (1966), Zajonc (1968) and Kanouse and Hanson (1971) all argue that negative information is counternormative. According to correspondence theory it should therefore have more dispositional value to an observer to receive such information as opposed to more expected positive information. It might be expected that such dispositions or attributions would be reflected in attitudes, intentions or behaviors toward the actor or target object (Kelley 1973).

The study reported here sought to extend the research on the effects of negative and positive information on purchase intention by focusing attention on alternative types of products (goods vs. services) and sources of information (neutral vs. marketer-dominated vs. consumer-dominated). Due to the wide range of products receiving negative information and the numerous sources releasing such information the inclusion of rating source and product type treatments was considered particularly important. The study sought to determine if the impact of negative and positive product information on purchase intention was mediated by the type of product and/or source of product information. Aside from the prediction that negative product information would have a greater impact on purchase intentions than positive information, it was also hypothesized that rating information of whatever type or source would have greater impact on services than goods. This latter prediction was based on findings that consumers typically place greater reliance on outside information sources when selecting services (Johnson 1969) and that there is a greater degree of risk associated with services (Lewis 1976). Though it was expected that each rating source would have differing relative effects on subjects, our primary interest was not in predicting main effects due to source but in examining its interaction with rating type and product type. Thus, no specific predictions were made about the impact of rating source on subjects.

METHOD

Procedure

The experimental design was a 3 x 2 x 2 factorial with repeated measures on one factor. Subjects were given descriptions of two different brands for each of two products, information on the source of the product rating, and the nature of the ratings themselves. The source and rating type represented the two between group factors while the type of product rated (good versus service) was the repeated factor. Each subject was randomly assigned to one of the source, rating, and product set conditions with thirty (30) subjects per cell (see Figure 1).

FIGURE 1

DESIGN FOR STATISTICAL ANALYSIS

The experimental treatments were administered in booklets containing all four brand ratings. No brand names, only neutral letters, were attached to the brand ratings. Although the ratings were fabricated, they were represented in the experiment as being authentic and the result of extensive product testing over the past six months to one year by the source designated.

After reading the rating for each brand, the respondents were instructed to indicate their intention to purchase by answering the question: How likely is it that you would select this item (brand) for use? Responses were recorded on annotated seven point scales ranging from extremely unlikely to extremely likely with a neutral midpoint. In all instances subjects in a favorable or unfavorable treatment condition were influenced in the scale direction corresponding to the treatment received. Thus, the means displayed in the Results section of this study truly reflect the magnitude of rating influence for the respective polar rating treatments. [Direct attribution measures were elicited. To date this data has not been properly analyzed.] After experimental exposure, subjects were debriefed. The results confirmed their focus on the listed rating sources and their belief in the authenticity of the actual ratings.

Subjects

The experiment utilized 240 women from church and community groups in a large metropolitan area in the west. Ages of the respondents ranged from eighteen to sixty-six years with 85% of the total falling in the twenty-four to forty-five age range. Subjects were not paid but participated in the study under the sanction of the community or church groups to which they belonged. The women were given the experimental manipulations in groups of twenty where treatment conditions were randomly assigned.

Rating Treatments

Two variations in rating treatments were created after a pilot study. The primary function of this initial study was to develop a group of positive and negative descriptive adjectives which were similarly polarized but in opposite directions. This was necessary to reduce systematic bias by virtue of the selection of more potent favorable or unfavorable descriptive rating adjectives. A pretest form was administered to twelve women similar to the group in the main experiment. The form presented lists of adjective pairs but the adjectives were not associated in any way with specific objects or persons. The women were asked to assign a number from 1 (very worst description possible) to 50 (neither good nor bad description) to 100 (the very best description possible). One extremely negative and one extremely positive descriptor was needed as well as one fairly neutral descriptor. The three descriptors which emerged were Excellent/Recommended, Very Poor/Not Recommended and Fair/Marginally Recommended. The first two pairs resulted in average scores of 95.5 and 5 on the 1-100 scale, respectively. If measured from the scale midpoint of 50, these two adjectives are extremely close in relative polarity. Fair/Marginally Recommended emerged with a value of 49.2.

Respondents in the main experiment received the rating information for two brands for each of one good and one service. Thus, each woman received ratings about a total of four brands. The favorable treatment condition involved a rating of Excellent/Recommended for a good and service and conversely the unfavorable treatment provided a rating of Very Poor/Not Recommended (see Figure 2). For both treatment conditions the remaining two brands were always rated with the neutral Fair/ Marginally Recommended. The rating treatments are displayed in Figure 2. This neutral treatment was solely for the purpose of reducing subject suspicion of receiving all favorable or unfavorable ratings and it remained as a common anchor for all subjects.

Source and Product Treatments

The source of the rating information received in the booklets varied along three levels: (1) Independent consumer testing agency, (2) Local housewives, and (3) Local trade or professional group. These three information sources correspond to a scheme developed by Cox (1963) designed to classify information sources in a marketing setting. Cox's first information source is labeled neutral and is one that would generally be perceived as unbiased. The example that Cox utilized is an agency such as Consumers Union. This neutral label would correspond most closely with the label independent testing agency used in this study. A second label assigned by Cox was marketer-dominated. This reflects information sources that originate from commercial institutions, and according to Cox, are perceived by the buyer as informative though not necessarily unbiased. This marketer-dominated label corresponds most closely with the local trade and professional groups listed as a source in this study. The third source classification that Cox developed was titled consumer-dominated and flows from the social-organizational setting. This would consist largely of friends, acquaintances, peers, and family. The consumer-dominated label corresponds to the local housewives utilized here.

FIGURE 2

FAVORABLE AND UNFAVORABLE RATING TREATMENTS

In order to focus respondent attention on the source of the ratings received in their individual booklets, the women were instructed to read carefully a separate page describing what the source of information was. In addition, at the top of each page the rating source was highlighted. It should be noted that each respondent received ratings from only one of the three sources.

As noted earlier, it was believed that the three information sources might have different influences on the subjects, however the interaction of sources with the rating and product type and not the main effect of source was of prime interest.

Rating Object

The target object in most of the negative information literature has been unknown persons; however, here unknown (new) goods or services were substituted as the objects of rating information. Following a pretest determining products with which the women had a better than average degree of personal past purchase experience, a group of four consumer goods and four services were randomly chosen. Unfamiliar product categories were eliminated so that the degree of experience variance between subjects would be reduced. The four goods selected were spray deodorant, sunglasses, simple snapshot camera, and dry breakfast cereal; and the four services were dry cleaner, physician, health spa, and beauty salon. Four pairs of one good and one service were randomly created such that any one subject received ratings about just one product pair.

A preliminary ANOVA model revealed no significant main effect or interactions with the product sets. Further analysis of the means confirmed the absence of differences between product sets. In the main analysis the four goods and then the four services were treated as homogeneous entities; that is, we collapsed across the product sets.

RESULTS

Tables 1 and 2 show the mean purchase intention scores for each cell of the 3 x 2 x 2 repeated measure design and the corresponding analysis of variance table, respectively. As noted earlier, the dependent measure asked: How likely is it that you would select this item (brand) for use?, and the results were recorded on seven point scales ranging from extremely likely to extremely unlikely with a neutral midpoint. Response differences from this midpoint are reported here. Considering first the significant main effect of Rating Type (p<.001), the cell means shown in Table 1 indicate that unfavorable information generally had a more pronounced effect on purchase intention scores than did favorable information; the only reversal of this pattern occurred for services when product information was received from a marketer-dominated source.

TABLE 1

NEW MEAN PURCHASE INTENTION SCORES FOR RATING TYPE, RATING SOURCE AND PRODUCT TYPE FACTORS

TABLE 2

ANALYSIS OF VARIANCE RESULTS

However, the significant rating source x rating type interaction (p<.001) indicates that group differences exist across corresponding two-factor combinations. To investigate these differences consider Table 3a which contains cell means for the collapsed table, and Figure 3. The cell means within brackets do not differ significantly, while those means not bracketed differ at the p<.05 level of significance. The significant cell means indicate that unfavorable information had a greater impact on purchase intentions than favorable information if received from either a neutral or consumer-dominated source, whereas no statistically significant difference between unfavorable and favorable information was evident with a marketer-dominated source, although the directional effect of information type was consistent with the other two rating sources. Stated somewhat differently, the relatively strong effect on purchase intention of favorable information when received from a marketer-dominated source was reversed and became relatively weaker as compared to the other two sources with unfavorable product information (see Figure 3). An opposite reversal was exhibited for the neutral source. It seems reasonable to suggest that the "opposite effects'' of these two rating sources on purchase intention contributed to the overall nonsignificant main effect of the Rating Source treatment.

TABLE 3

MEAN PURCHASE INTENTION SCORES FOR SIGNIFICANT INTERACTION EFFECTS

FIGURE 3

MEAN INTENTION SCORES OF SOURCE AT EACH LEVEL OF RATING TYPE

The main effect of Product Type was significant (p<.001) and the cell means shown in Table 1 indicate that purchase intention scores for services were consistently stronger than for goods. In addition, the nonsignificant Product Type x Rating Type interaction implies that product information, whether unfavorable or favorable, had a relatively greater impact on purchase intention for services than goods. This result is consistent with prior literature on services and lends support to the belief that consumers choosing products utilize outside information about services to a greater extent than when choosing goods.

The Product Type x Rating Source interaction was also significant (p<.02) which means that profile differences exist in purchase intention scores across corresponding two-factor combinations. Inspection of Table 3b and Figure 4 suggests that this statistically significant interaction was due largely to the fact that the mean purchase intention scores for information received from a consumer-dominated source showed no significant decrease when the target object was a good, whereas the mean purchase intention scores for the other two sources did show a statistically significant (p<.05) decrease.

FIGURE 4

MEAN INTENTION SCORES OF SOURCE AT EACH LEVEL OF PRODUCT TYPE

DISCUSSION

The results of this study provided answers to several of the questions raised in the introduction. First, the type of information did have an impact on purchase intention scores. In general, unfavorable product ratings tended to have a greater impact on purchase intention than did favorable ratings which suggests that impressions formed about unfamiliar products are not unlike those formed about unknown persons. Correspondence theory would suggest that the more unexpected negative ratings provided more valuable dispositional information to the subjects. Kelley (1973) has suggested that the differing inferences or attributions would be expected to reflect in attitudes or behavior toward the actor or, in this instance, object. The differential intentions manifested here appear to correspond to the predictions of the theory.

Secondly, it was found that unfavorable product information received from an independent testing agency or local women as sources had a relatively stronger effect on purchase intentions than did similar information communicated by the trade and professional association source. Furthermore, under the unfavorable information treatment condition the independent testing agency source had the strongest effect on purchase intentions.

Perhaps the most significant result from a marketing viewpoint was the effect of the Product Type treatment.

Regardless of the source or type of information received, purchase intention scores for services were uniformly higher than those for goods. The result lends support to speculations of Parker (1960) and Johnson (1969) that services might be harder to judge than goods and, therefore, consumers may well place more reliance on outside sources of information when available. In addition, the greater impact of information on services is consistent with Lewis's (1976) finding that risk levels associated with services are higher than those associated with goods.

It appears that, on the one hand, the mere threat of negative information about one's product might serve as a warning to marketers to be more vigilant about their products. Conversely, if negative product information does have the impact suggested in this study, some degree of restraint in releasing negative information might be necessary by the growing legion of negative information sources. If the goal of public policy is educational and not punitive then perhaps restraint about negative information should be exercised. No stronger case for restraint could be made than the Bon Vivant botulism reports which caused the demise of the entire label within six months.

Finally, future research efforts should be directed at gradually removing the sterility of the studies conducted to date and injecting more varied and realistic treatment and exposure conditions. Such efforts combined with some direct measurement of the attribution process thought to underlie the informational effects should help uncover the true impact which unfavorable information disclosures have upon consumers. Eventually, by examining rebuttal strategies to negative informational effects, a more solid basis for strategy formulation by marketers will be possible.

REFERENCES

Anderson, Norman H. (1965), "Averaging Versus Adding as a Stimulus Combination Role in Impression Formation," Journal of Personality and Social Psychology, February, 1-9.

Arndt, Johan (1968), 'Word-of-Mouth Advertising and Perceived Risk," in Perspectives in Consumer Behavior, eds. Harold Kassarjian and Thomas Robertson, Coleview, Illinois: Scott Foresman and Company.

Cox, Donald (1963), "The Audience vs. the Communicators," in Proceedings of the American Marketing Association, 58-72.

Goodman, S. M. (1950), "Forming Impressions of Persons from Verbal Reports," unpublished Ph.D. dissertation, Columbia University.

Johnson, Eugene M. (1968), "Are Goods and Services Different? An Exercise in Marketing Theory," unpublished Ph.D. dissertation, Washington University.

Jones, E. E. and Davis, K. (1965), "From Acts to Dispositions,'' in Advances in Experimental Psychology, ed., Leonard Berkowitz, New York: Academic Press, 219-266.

Lewis, William (1976), "An Empirical Investigation of the Conceptual Relationship Between Services and Products," unpublished Ph.D. dissertation, University of Cincinnati.

Osgood, C. E., Suci, G. J., and Tannenbaum, P. H. (1957), The Measurement of Meaning. Urbana: University of Illinois Press.

Parker, Donald (1960), The Marketing of Consumer Services, Seattle: Bureau of Business and Research, University of Washington.

Posavac, E. J. (1974), "Relative Weighting of Positive and Negative Information and Confidence in Reports of Behavioral Intentions," Bulletin of the Psychonomic Society, 4(5A), 481-483.

Richey, M., Koegnigs, R. J., Richey, H. W., and Fortin, Richard (1975), "Negative Salience in Impressions of Character: Effects of Unequal Proportions of Positive and Negative Information," Journal of Social Psychology, 97, 233-241.

Richey, M. H., Richey, H. W., and Thieman, G. (1972), "Negative Salience in Impressions of Character: Effects of New Information on Established Relationships," Psychonomic Science, 28, 65-66.

Wright, Peter (1974), "The Harassed Decision Maker: Time Pressures, Distractions, and the Use of Evidence," Journal of Applied Psychology, 59, No. 5, 555-561.

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