Consumer Response to Negative Publicity

s

ROHINI AHLUWALIA, ROBERT E. BURNKRANT, and H. RAO UNNAVA*

Even though negative information about brands and companies iswidely prevalent in the marketplace, except for case studies, there hasbeen no systematic investigation of how consumers process negative in 'formation about the brands they like and use. In the three studies in thisresearch, the authors attempt to bridge this gap. The findings of the firstand second studies provide a theoretical framework for understanding'how consumers process negative information in the marketplace.Commitment of the consumer toward the brand is identified as a moder-ator of negative information effects. In the third study, the authors use thistheoretical framework to derive and test response strategies that compa-nies can use to counter negative publicity for consumers who are high

and low in commitment toward the brand.

Consumer Response to Negative Publicity:The Moderating Role of Commitment

Incidents of negative publicity are widely prevalent in themarketplace, ranging from safety concerns with Valujet air-lines to tainted beef from Hudson foods. Such informationcan be devastating, resulting in major losses of revenue andmarket share. A study by DDB Needham Worldwide(Advertising Age 1995) finds that negative publicity andhow the company handles it are among the most importantfactors influencing consumers' buying decisions.

The potential impact of negative publicity is not surpris-ing. Publicity is considered a relatively credible source ofinfonnation and therefore is more influential than othermarketer-driven communications (Bond and Kirshenbaum1998). Furthermore, negative as opposed to positive infor-mation is known to be more attention getting (Fiske 1980).

Despite the potential impact of negative publicity in themarketplace, knowledge about its effects is limited. There islittle theoretical research dealing with how consumersprocess such information and how companies can developstrategies to combat its effects. The puhlic relations and

•Rohini Ahluwalia is Assistant Professor of Marketing, School ofBusiness, University of Kansas (e-mail: [email protected]). Robert E.Bumkrant is Professor (e-mail: [email protected]) and H. Rao Unnavais Associate Professor of Marketing (e-mail: [email protected]). FisherCollege of Business, Ohio State University. This article is based on the firstauthor's doctoral dissertation, which was awarded the 1997 John A.Howard Doctoral Dissertation Award by the American MarketingAssociation. The first author acknowledges the financial support of theGraduate Student Alumni Award and the William R. Davidson DoctoralFellowship in Marketing from the Ohio State University. The authors thankPunam Anand Keller, Zeynep Gurhan CanIi, Sanjay Mishra, Rich Petty,and Surendra Singh for their comments. To interact with colleagues on spe-cific articles in this issue, see "Feedback" on the JMR Web site atwww.ama.org/pubs/jmr.

publicity literature has examined this problem at jsomelength but has not converged on a unifying theoreticalframework. One underlying assumption in this literature isthat negative information is almost always devastating. Forexample, the Merriam formula used to determine the impactof media exposure gives negative news quadruple weightcompared with positive news (Kroloff 1988). Another as-sumption in this literature is that consumers respond io thenegative publicity in a homogeneous manner (Mgrconi1997; Pearson and Mitroff 1993).

In general, case studies have been used to arrive atj con-clusions about which strategies work and which do not seemto work in the marketplace (e.g., Chisholm 1998; Marconi1997; Pearson and Mitroff 1993; Weinberger, Romed, andPiracha 1991). This literature, however, provides little di-rection for understanding the problem from a theoreticalperspective. For example, a typical project in this area '(e.g.,Pearson and MiU-off 1993) presents a framework for crisismanagement based on the study of some companies in Crisissituations. Recommendations comprise general strategic di-rections (e.g., "Integrate crisis management into straitegicplanning process," "Provide training and workshops in cri-sis management") without any attempt at understandinghow consumers process this information and/or factor^ thatmoderate its effects on consumer response.

In the current research, we lay the foundation for a theo-retical framework of negative information processing by fo-cusing on consumer processing of negative publicity infor-mation about a company's products. Although negativepublicity may also relate to other aspects of a company'soperations (e.g., human resource issues), we focus on prod-uct-related publicity because of its preponderance (Dye1997; Irvine 1992). Specifically, the objective of out re-

203Journal of Marketing Research,Vol. XXXVII (May 2000), 203i-214

204 JOURNAL OF MARKETING RESEARCH, MAY 2000

search is to provide an understanding of how consumers re-act to negative product-related information about brandsthey like and use.

Departing from prior research, which has assumed a ho-mogeneous consumer response to negative publicity infor-mation, we argue that prior characteristics of the con-sumer—specifically, commitment to the target brand—moderate the processing and impact of negative publicity. Ina series of two studies, we test the differential responses tonegative publicity of consumers who are high and low incommitment to the publicized brand, and we delineate thepsychological processes responsible for this effect. The firstexperiment measures commitment for a leading brand in thetarget product category, and the second manipulates com-mitment for a relatively low-share brand in the same prod-uct category. On the basis of the findings from the first twostudies, we propose and test (in Experiment 3) two types ofresponse strategies to counter negative publicity, dependingon the level of commitment of the consumers. We then dis-cuss the implications of our findings for marketers.

CONCEPTUAL FOUNDATION

Negativity Effect

Although the literature is sparse on theoretical insightsabout consumer reaction to negative publicity about aknown brand, the issue of how consumers process and inte-grate negative information with positive information hasbeen studied in the impression formation literature in psy-chology (e.g., Fiske 1980; Klein 1996; Skowronski andCarlston 1989). A typical study in this literature involvesgiving subjects several pieces of information about a ficti-tious person, some positive and some negative, and then as-sessing the effects of negative information on the overall im-pression subjects form of that person.

A robust finding in the impression formation literature isthe negativity effect; that is, people place more weight onnegative than positive information in forming overall evalu-ations of a target (Fiske 1980; Klein 1996; Skowronski andCarlston 1989). This effect has been found in person per-ception as well as product evaluation contexts (e.g., Herr,Kardes, and Kim 1991; Wright 1974). For example,Maheswaran and Meyers-Levy (1990) find that when theprocessing is focused on message content, negative framingis more effective than positive framing.

One reason for the negativity effect may be that negativeinfonnation is considered more diagnostic or informativethan positive information (Maheswaran and Meyers-Levy1990; Skowronski and Carlston 1989). For example, whenconsumers are exposed to negative information about aproduct, they can categorize the product as low in quality.However, positive and neutral information about products isless useful in categorizing them, because such features arecommonly possessed by high-, average-, and low-qualityproducts (Herr, Kardes, and Kim 1991). Therefore, negativeinformation may simply be considered more useful or diag-nostic in making decisions and is given greater weight thanpositive information.

The negativity effect, however, has been reported in ex-perimental contexts in which subjects are unfamiliar withthe person whom they are evaluating and are forced to com-bine positive and negative attribute information to form anattitude toward the person. Negative publicity, in contrast.

deals with brands consumers are familiar with and towardwhich they may have a prior attitude. Whether and underwhat conditions the negativity effect applies to consumersevaluating negative publicity information are addressednext.

Attitude Strength

Recent research in consumer behavior indicates that con-sumers become attached to various brands and form "rela-tionships" with them (e.g., Fournier 1998), which results inequity for the brand (e.g., Keller 1993). The attitudes thatconsumers have for such brands are expected to vary instrength. Stronger attitudes are known to exhibit greater re-sistance to information that attacks them, that is, negativeinformation (e.g.. Petty and Krosnick 1995).

Many dimensions of attitude strength (e.g., prior knowl-edge, commitment, importance, extremity) have been iden-tified in the literature. Recent research, however, has con-cluded that there is no unitary construct of attitude strength(Krosnick et al. 1993; Petty and Krosnick 1995).Consequently, researchers in the area of attitude strength(Eagly and Chaiken 1995; Krosnick et al. 1993; Petty andKrosnick 1995) have issued a call for more research on theindividual dimensions of attitude strength and greater re-straint in generalizing the outcomes and processes obtainedfrom one dimension to another.

In this research, we examine the role of commitment, oneof the attitude strength dimensions, as a moderator of nega-tive information effects. We chose to focus on commitmentfor several reasons. First, commitment has been viewed asone of the two major dimensions of attitude strength(Krosnick et al. 1993; Pomerantz, Chaiken, and Tordesillas1995).' Second, commitment has been shown to play a crit-ical role in determining resistance to counterattitudinal in-formation. Specifically, the effect of several strength vari-ables, such as prior knowledge and importance, has beenshown to depend on the person's level of commitment to-ward the target (Wood, Rhodes, and Biek 1995). For exam-ple, although knowledge can enable objective processing(inducing resistance to counter- and proattitudinal informa-tion) for people who are not committed to a brand/issue, itcan lead to biased processing for committed subjects (in-ducing resistance to counter- but not proattitudinal informa-tion). Third, commitment is akin to brand loyalty, a conceptthat has recently engendered much research in marketing(e.g., Dick and Basu 1994; Fournier 1998). Commitmentprovides an essential basis for distinguishing between brandioyalty and other forms of repeat purchase behavior (Jacobyand Chestnut 1978). Although brand loyalty was viewedsimply as repeat buying in the past, it has become increas-ingly similar to the conceptualization of commitment as thefield of consumer behavior has matured (Morgan and Hunt1994). It has been defined as an emotional or psychologicalattachment to a brand within a product class (Lastovicka andGardner 1978).

HYPOTHESES

Prior research suggests that people who have positive at-titudes toward a target are likely to engage in biased as-similation, resisting counterattitudinal information more

'The other major dimension of attitude strength is embeddedness.

Consumer Response to Negative Publicity 205

than proattitudinal information (Ditto and Lopez 1992;Edwards and Smith 1996; Kunda 1990). It follows fromthis literature that consumers who have a positive attitudetoward a brand should counterargue the negative publicityrelated to it.

In contrast, the impression formation literature reviewedpreviously suggests that negative information receivesgreater weight than positive information and is more likelyto cause attitude shifts in its direction. We contend thatwhether negative publicity information about a positivelyevaluated brand will be discounted in a biased manner orwill be weighted heavily (as in the negativity effect) de-pends on the consumer's level of commitment toward thatbrand (see also Sawyer 1973). In other words, we suggestthat the commitment of the consumer toward the brand willmoderate these outcomes.

When commitment to a brand is lower, consumers are ex-pected to process negative publicity information in a rela-tively objective manner. This situation resembles the set-tings found in the impression formation literature in whichsubjects are not committed toward the person they are rat-ing. These people are likely to view negative publicity asmore diagnostic than positive publicity about the brand. Themore highly perceived diagnosticity of the negative infor-mation is expected to mediate the attitudinal changes expe-rienced by low-commitment consumers as they encounterand react to negative publicity information about a brand.

The high-commitment consumers, conversely, are likelyto engage in biased processing of the publicity infonnation.They are expected to counterargue the negative informationmore extensively than the positive information (Chaiken,Liberman, and Eagly 1989; Gross, Holtz, and Miller 1995)and therefore resist attitude change in response to negativeinformation. In addition, they are expected to show attitudi-nal shifts in the direction of the advocacy when the messageportrays their favored brand in a positive light. Because oftheir high level of attachment to the brand, they are lesslikely to accept negative information as more diagnostic(Feldman and Lynch 1988). Thus, it is not the perceived di-agnosticity but the counterargumentation that mediates atti-tude change for these consumers. Therefore, a negativity ef-fect is not expected for them.

H|: Low-commitment consumers will exhibit a greater amountof attitude change in response to negative as compared withpositive information (i.e., a negativity effect), but high-com-mitment consumers are not expected to exhibit a negativityeffect.

H2: Although the effect of valence of the information on atti-tude change is likely to be mediated by the perceived di-agnosticity of information for the low-commitment con-sumers, this effect is likely to be mediated bycounterarguments generated for the high-commitmentconsumers.

A related issue is attitudinal ambivalence. Attitudes thathave both positive and negative components associated withthem are termed "ambivalent" and tend to be unstable(Kaplan 1972). Cacioppo, Gardner, and Bemtson (1997) ar-gue that the impact of negative information can be betterjudged through attitudinal ambivalence measures instead ofthe standard measures of attitude valence. The integration ofnegative information into the consumer's attitude is likely tolead to increased attitudinal ambivalence.

As described previously, high-commitment consumersare expected to counterargue and discount negativd infor-mation. Therefore, this information is not likely to bestrongly associated with their post-negative message atti-tudes. Their attitudinal ambivalence is therefore not ex-pected to be affected by negative publicity. The low-com-mitment consumers, in contrast, are expected to integratethe negative publicity into their attitudes, which results inincreased ambivalence in the negative as opposed to thepositive information condition. |

H3: Low-commitment consumers are expected to have signifi-cantly more ambivalent attitudes when exposed to negativeas compared with positive publicity information, whe,reas nosuch differences in attitudinal ambivalence are expected forthe high-commitment consumers. ]

Summary

The level of commitment that the consumer has towardthe brand moderates the processing of negative infortiiationabout a well-liked brand. People highly committed to abrand are expected to counterargue the information aind re-sist attitude change. People whose commitment to thej brandis lower are expected to counterargue negative informationtoo, but less so than their high-commitment counterparts.They are expected to weight negative information more thanpositive information because of its higher perceived diag-nosticity. Therefore, whereas a negativity effect is expectedfor the low-commitment consumers, it is expected to be ab-sent for high-commitment consumers. In other wordsj com-mitment is hypothesized to moderate the occurrence of anegativity effect.

EXPERIMENT 1

To test the proposed hypotheses, a 2 (commitment |of theconsumer toward the target brand; high and low) x 2 (va-lence of the publicity information: positive and negative)between-subjects design was implemented. 1

Stimuli and Independent Variables j

Commitment was used as a measured variable in this ex-periment. The risk with measuring commitment is that| otherunmeasured factors associated with different levels of com-mitment may provide alternative explanations for the re-sults. To address this issue, a pretest was run to examijie thecorrelations between commitment and other potentially con-founding dimensions of attitude strength. Twenty-five sub-jects responded to questions assessing various dimetisionsof attitude strength (accessibility, commitment, extremity,importance, and prior knowledge) for brands in two prjoductcategories: athletic shoes and televisions. The measureswere adapted from Krosnick and colleagues (1993); Thecorrelations between commitment and the variables of priorknowledge (r - .20), importance (r = .22), and accessi^bility(r = .21) were low and nonsignificant (all ps >I.15).However, commitment was significantly correlated with at-titudinal extremity (r = .71, /7 < .001). To control for attitu-dinal extremity, high- and low-commitment subjects' withequivalent attitudes toward the target brand were recruitedto participate in the experiment. |

Subjects' commitment toward the target brand was meas-ured using a three-item brand commitment measure! pro-


posed and tested by Beatty, Kahle, and Homer (1988).2Consumers in the upper (lower) third of this scale were cat-egorized as high (low) in commitment.

Athletic shoes were selected as the target product cate-gory, because students in the subject pool (introductory mar-keting class) were familiar with this product category andhad a fairly wide distribution of commitment toward it(mean = 6.11/9, standard deviation = 2.66, n = 456). Nikewas chosen as the target brand because of its wide distribu-tion of commitment scores and narrow distribution of atti-tude scores. This was important so that subjects with similarinitial attitudes but different levels of commitment towardthis brand could be recruited in the experiment.

Positive and negative target messages were developedthrough a series of pretests. Both versions of the messagedealt with information about the attribute of shock absorp-tion. In the final pretest, 35 subjects were exposed to eithera positive or a negative message about an unknown brand ofshoes. Subjects were asked to rate "how favorable or unfa-vorable was the presented article towards the target brand?"on an 11-point scale (-5 to +5). The messages were rated assignificantly different in their valence (meannegative = ~3.4,meanpositive = +3.8; F = 95.6, p < .001) but not in their ex-tremity (meannegative = 3.4, meanpositive = 3.8; p > .50). Theywere rated (on a 7-point scale) as equivalent in their believ-ability (meannegative = 5.3, meanpositive = 5.1; p > .80) andwere equated on their length. For the first experiment, thebrand name in the target messages was changed to Nike.

Six filler articles (three positive and three negative) weredeveloped. They were based on real articles published inmajor newspapers and dealt with products such as vitamins,computer software, automobiles, and orange juice. The fillerarticles served to reduce the likelihood of ceiling effects dueto excessive attention focused on the target message.

Procedure

At the beginning of the academic session, all the studentsin an introductory marketing class were administered amass-testing questionnaire that contained, embedded amongother questions, measures of their commitment and attitudetoward the target brand. Four weeks later, subjects were re-cruited over the telephone to participate in the experiment.To control for the extremity of the prior attitude toward thebrand, high- and low-commitment students with similar at-titudes (nine-point scale) toward the target brand were re-cruited (meanjow = 7.71, mean igh = 7.62; p > .80). A total of68 subjects (34 high- and 34 low-commitment) participatedin this study. On their arrival, subjects were told that theywere participating in a media study being conducted by thedepartment of marketing in collaboration with the school ofjournalism. Their task was to evaluate recent newspaper ar-ticles related to different products. The subjects were givena folder containing six newspaper articles—three of whichwere negative (N) and three of which were positive (P) to-ward the product^rand featured in them. To control for po-sition effects, the negative or positive publicity article about

three items were (1) "If brand X of athletic shoes were not avail-able at the store, it would make linle difference to me if I had to choose an-other brand"; (2) "I consider myself to be highly loyal to X brand of ath-letic shoes"; and (3) "When another brand is on sale, I will generallypurchase it rather than X brand." Subjects expressed their agreement withthe statements on a nine-point scale anchored by disagree/agree.

Nike was always in the fourth place. In the negative targetarticle condition, the sequence of articles was (N, P, P,Ntatget' P> N). In the positive target article condition, the se-quence of articles was (N, P, N, P,arget' P' N).

After reading the articles, subjects were given the depend-ent measures booklet. It started with a cognitive responsetask, followed by the attitude measures and the rest of the de-pendent measures. After they had completed the question-naire, subjects were debriefed. They were specifically di-rected to the target article and told that it was made up by theexperimenters and therefore should be discounted by them.Subjects were quizzed on whether they had guessed the pur-pose of the experiment. None of the subjects had.

Dependent Variables

Subjects were given 2.5 minutes to list all the thoughtsthey had while reading the target article (Petty and Cacioppo1977). The thoughts were coded by two judges into four cat-egories: counterarguments, support arguments, other mes-sage-related thoughts, and other thoughts (not related to themessage). Support (counter-) arguments in the negative arti-cle condition included negative (positive) thoughts aboutNike, whereas in the positive article condition they includedpositive (negative) thoughts about the target brand. Therewas 92% agreement between the judges. Disagreementswere resolved through discussion.

Attitude toward the target brand was measured using fournine-point semantic differential scales (good/bad, benefi-cial/harmful, desirable/undesirable, and nice/awful) (coeffi-cient alpha = .97). Using these measures, a mean attitudescore was computed. Mean attitude change was computed asthe difference^ between the premessage mean attitude (ob-tained during mass testing) and the postmessage mean atti-tude (measured during the study) for each subject.

Measures of ambivalence followed Kaplan's (1972) tech-nique. Respondents were asked to rate on a four-point scale(not at all, 0, to extremely, 3) the extent to which they hadpositive feelings toward the target brand (e.g., 3) and then,separately, the extent to which they had negative feelings to-ward it (e.g., 2). The subject's degree of ambivalence wascomputed by taking the sum of the positive and negative rat-ings (e.g., 3 + 2 = 5) of the attitude toward the object (thisrepresents the total amount of affect toward the object, re-gardless of valence) and then subtracting the absolute valueof the difference between the two scales (e.g., 3 - 2 = 1). Inour example, ambivalence would be calculated as 5 - 1 = 4.The ambivalence rating can range from 0 (no ambivalence)to 6 (very high ambivalence).

The perceived diagnosticity measures were similar tothose used by Skowronski and Carlston (1987) and Herr,Kardes, and Kim's (1991). Subjects were asked to estimatethe percentages of high- and low-quality shoes likely tohave a problem with the target attribute (shock absorption)and the percentages of high- and low-quality shoes likely to

^Measuring attitude change as a difference, however, raises the issue ofwhether these difference scores are reliable (Cronbach and Furby 1970;Harris 1967). Recent research has demonstrated that difference scores areunreliable only when the standard assumptions of classical theory thatpretest (x) and posttest (y) standard deviations are equal (i.e., X = a^/Oy =1) and the correlation between posttest and pretest scores is large (p,y ° I)hold (e.g., Collins 1996; Rogosa 1988). In our data, X = .52 and p , , = .34.Therefore, the reliability of difference scores is not a potential issue for ourresearch.


have a high (good) level of shock absorption. As in Herr,Kardes, and Kim's (1991) study, low quality was defined asaverage or low quality to ensure that mutually exclusive andexhaustive categories would be employed. The diagnosticityof negative information was computed by dividing the per-ceived probability of a low-quality shoe having a problemby the sum of the same probability and the probability of ahigh-quality brand having that problem. The diagnosticity ofpositive information was computed similarly.

Results

Attitudes. Low-commitment consumers were expected toexhibit greater attitude change and significantly more am-bivalent attitudes in response to negative (versus positive)publicity. However, high-commitment consumers were notexpected to exhibit greater attitude change or more ambiva-lent attitudes when exposed to negative (versus positive)publicity. These predictions called for an interaction be-tween commitment and valence.

The interaction between commitment and valence wassignificant for attitude change (F = 3.76, p < .05) and ap-proached significance for attitudinal ambivalence (F = 2.78,p < .10). Because attitude change could be either positive ornegative in our experiment, only comparisons using ab-solute values of attitude change could ensure an appropriatetest of the hypotheses.''

As predicted in H], low-commitment consumersexpressed significantly greater attitude change with negative(versus positive) information (meannega,ive = 1.69,meanpositive = 07; t - 5.26, p < .001). Furthermore, their at-titudes were significantly more ambivalent (meannegative =2.47, meanposjtive = 1-33; t = 2.74, p < .01) when they wereexposed to negative (versus positive) infonnation (H3).

Also as predicted in H|, the high-commitment consumersdid not exhibit more attitude change with negative than pos-itive publicity information; instead, the negativity effect wasnearly reversed for these consumers. They had marginallymore attitude change with positive than negative informa-tion (meanpositive - -69, meannegative = -24; t = 1.37, p < .10).Consistent with H3, there were no significant differences inthe attitudinal ambivalence of high-commitment consumers(meannegative = 1-53, meanpositive = ' -38; p < .25). Therefore,the results supported the hypothesis that high-commitmentconsumers resist negative information.5

Cognitive responses. Low-commitment consumers wereexpected to process the publicity information objectively.Because the messages were equally strong, no differenceswere expected in the number of counterarguments generatedin the positive and negative conditions. High-commitmentconsumers, conversely, were expected to engage in defen-sive processing of the messages, generating more counterar-guments in response to the attitude-inconsistent negativemessage as compared with the attitude-consistent positivemessage.

••For example, if positive publicity results in +2 units of attitude changeand negative publicity leads to -2 units of attitude change, the amount ofattitude change is not significantly different in the two conditions, eventhough the difference between -2 and +2 is likely to be significant.

'Even though the data provide clear evidence of resistance to negativeinformation by high-commitment consumers, we cannot infer that high-commitment consumers were not at all influenced by the negative infor-mation, because the power of the attitude change measure was too low todetect differences (power = .31).

The commitment by valence interaction was significant(F = 26.41, p < .001). The main effect for valence whs alsosignificant (F = 42.94, p < .001), which suggests mor4 coun-terarguments in the negative than the positive condition. Asexpected, high-commitment consumers had significantlymore counterarguments in the negative (versus positive)message condition (meannegative = 3.59, meanpositive F -75;t = 5.85, p < .001), whereas low-commitment consumers didnot exhibit a significant difference between the two condi-tions (meannegative = 2.18, meanpositive = 1.83; t < 1).

The pattern of results obtained with support argumentswas consistent with the attitude results. High-commitmentconsumers had more support arguments in the positive thanthe negative condition (meannegative = -94, meanpojlitive =2.63; t = 3.63, p < .01), whereas low-commitment con-sumers had more support arguments in the negative than thepositive condition (meannegative = '-94, meanpositive - 1 " ;t= 1.78,p<.05).

Perceived diagnosticity. The negativity effect is ba$ed onthe assumption that negative information is perceived to bemore diagnostic than positive infonnation. Consistet^t withthis rationale, low-commitment consumers perceived nega-tive information to be more diagnostic^ than positive infor-mation (meannegative - -70, meanpositive = -66; t = 2.3p, p<.05). A reversal, however, was found for the high-commit-ment consumers. They perceived positive (versus negative)information as more diagnostic (meannegative ^ -67,meanpositive = -72; t = 3.47, p < .001). Although we hjid notpredicted this reversal, it suggests that negative infortiiationis not viewed as diagnostic under all conditions and the levelof commitment affects perceived diagnosticity.

Mediational analysis. Ho predicts that perceived diagnos-ticity of information is likely to mediate attitude change forlow-commitment consumers, whereas counterargumenta-tion is likely to mediate the attitude change for highS-com-mitment consumers. To test these predictions, for each, com-mitment condition, the two variables (perceiveddiagnosticity^ and counterarguments) were tested as (jioten-tial mediators of the impact of information valence oh atti-tude change. Following Baron and Kenny (1986), we ran aseries of regressions. Mediation would be implied if |all ofthe following three conditions were met: (1) Information va-lence significantly predicted the variable (i.e., perceivfcd di-agnosticity or counterarguments), (2) valence significantlypredicted attitude change, and (3) when both the variableand valence were regressed on attitude change, the impact ofvalence was attenuated, but the variable remained signifi-cant. Therefore, we estimated five regression equation^ (de-

6A11 the consumers (irrespective of whether they saw negative or posi-tive target information) rated the perceived diagnosticity of negative hs wellas positive information relating to the target attribute. Therefore, the diag-nosticity scores were averaged for each level of commitment across the dif-ferent levels of valence. That is, we calculated the diagnosticity index ofnegative information for low-commitment consumers by averaging the di-agnosticity ratings of all the low-commitment consumers, irrespective ofwhether they saw the negative or positive information. We performed t-tests on the composite means obtained.

''To minimize within-subject variance, the relative perceived diagnostic-ity term was computed as the standardized diagnosticity of the publicity in-formation to which the subject was exposed. For subjects who were ex-posed to negative (positive) publicity, the relative diagnoslicity wascomputed by dividing their perceived diagnosticity of negative (positive)information by their perceived diagnosticity of positive (negative) infor-mation about the target attribute. ,


scrihed in Tables 1 and 2) for each group of consumers. Theresults are described next.

For the low-conimitment consumers, perceived diagnos-ticity emerged as a significant mediator of attitude change.Valence was a significant predictor of attitude change (p <.001). Furthermore, not only did information valence predictperceived diagnosticity (p < .001), but including the latter inthe regression equation also attenuated the impact of va-lence on attitude change (from p< .001 top< . 10), whereasthe effect of perceived diagnosticity remained significant(p < .05). See Table 1. Counterarguments, however, did notmediate attitude change for these consumers, because va-lence was not a significant predictor of counterarguments.

As predicted, counterarguments significantly mediatedattitude change for the high-commitment consumers.Information valence predicted attitude change (p < .10) aswell as counterarguments elicited (p < .001). Furthermore,including counterarguments in the regression equation at-tenuated the impact of valence on attitude change (from p <. 10 to p > .60). As expected, diagnosticity did not emerge asa significant mediator of attitude change for these con-sumers. It failed to predict attitude change when perceiveddiagnosticity and information valence were regressed on at-titude change (Equation 3). See Table 2.

Therefore, low-commitment consumers exhibited attitudechange in response to negative information because theyperceived it to be highly diagnostic, whereas the high-com-

mitment consumers resisted negative information becausethey effectively counterargued it.

Discussion

Commitment emerged as a moderator of negative infor-mation effects on attitude change. The data indicate that low-commitment consumers give more weight to negative thanpositive information, because they perceive it to be more di-agnostic. However, they do not appear to counterargue it anyless than the positive information. In other words, theyprocess the information objectively but are influenced by thehigher perceived diagnosticity of negative information andtherefore give it more weight than positive information.

In contrast, the level of defense offered to negative infor-mation by the high-commitment consumers was remark-able. They extensively counterargued the negative informa-tion while supporting the positive information. Thecounterarguments generated emerged as a significant medi-ator of attitude change for this group of consumers.

The high-commitment consumers perceived positive in-formation to be more diagnostic than negative information.Prior literature (Feldman and Lynch 1988; Herr, Kardes, andKim 1991) suggests that a person's goals are likely to deter-mine the perceived diagnosticity of a piece of information.Consumers may exhibit inferential biases to be consistentwith their goals, over- or underestimating the diagnosticvalue of a piece of information (Herr, Kardes, and Kim

**p < .05.***/)< 001.

Table 1MEDIATIONAL ANALYSIS: LOW-COMMITMENT CONSUMERS

EquationNumber

(1)(2)(3)

(4)(5)

DependentVariable

Attitude changePerceived diagnosticity

Attitude change

CounterargumentsAttitude change

IndependentVariable(s)

ValenceValenceValence

Perceived diagnosticityValenceValence

Counteraiguments

StandardizedRegressionCoefficient

-.55-.52-.32-.38-.12-.55-.24

t-Value

-3.74***-3.53***-1.91*

2.25**-.69

-3.79***-1.63

Table 2MEDIATIONAL ANALYSIS: HIGH-COMMITMENT CONSUMERS

EquationNumber

(1)(2)(3)

(4)(5)

DependentVariable

Attitude changePerceived diagnosticity

Attitude change

CounterargumentsAttitude change

IndependentVariabte(s)

ValenceValenceValence

Perceived diagnosticityValenceValence

Counterarguments

StandardizedRegressionCoefficient

.31

.51

.20

.20

.73-.08-.54

t-Vatue

1.80*3.27**1.041.025.90***-.36

-2.31**

•*p < .05.•**p<.001.


1991). Our finding of higher perceived diagnosticity of thegoal-consistent positive information as compared with thegoal-inconsistent negative infonnation for high-commit-ment consumers is consistent with this literature. However,diagnosticity did not mediate attitude change for them.

Although this experiment provides support for our hy-potheses, further evidence of both internal and external va-lidity of the findings would be desirable. Commitment wasmeasured for a leading brand in the product category in thisexperiment. Commitment levels for such brands should bestrong and capable of generating the biased processingfound in Experiment 1. A question that follows is whetherthese effects would be replicable for a low-share brand, forwhich commitment levels may not be as high. In otherwords, subjects with extremely high levels of commitmentwere recruited. It would be desirable to test whether theseresults are generalizable to less extreme levels of commit-ment. Another major concern is that commitment was meas-ured in this experiment. Although many precautions weretaken to control for confounding variables, it may still be de-sirable for commitment to be manipulated.

We conducted a second experiment to address the previ-ous issues. We manipulated commitment for a low-sharebrand. The selection of a low-share, lesser-known brand en-ables us to equate the high- and low-commitment groups onother attitude strength dimensions and ensure that the effectswe observe are driven by commitment. Therefore, in thesecond experiment, we attempt to extend the external andthe internal validity of the framework tested in Experiment1. Because all of these issues can be addressed within thecontext of negative information, we focused on only thiscondition in Experiment 2.

EXPERIMENT!

Design

Commitment (high versus low) was manipulated. Twoconditions (experimental and control), described in the pro-cedure section, were run for each commitment group.

Stimuli

A pretest was conducted to identify the low-share targetbrand. Three hundred ninety students from an introductorybusiness class filled out a questionnaire assessing their atti-tudes about, familiarity with, and commitment toward vari-ous brands. On the basis of the results, Mizuno athletic shoeswere identified as the target brand. On a nine-point scale,students were relatively unfamiliar with the brand (mean =3.07), had low levels of prior commitment toward the brand(mean = 2.18), and had moderately positive attitudes towardit (mean = 5.13). Cameras were chosen as a filler product forreasons to be described in the next section. Materials, in-cluding background information on the products/companies.Consumer Reports articles for the brands, and advertise-ments, were developed for the two product categories.

Procedure

Seventy-one students from an introductory business classparticipated individually in this experiment. On arrival, sub-jects were informed that they were participating in a con-sumer survey being conducted by a market research com-pany in collaboration with the business school. The subjectswere told that the study would pertain to two products that

were going to be introduced in their local area. They werethen handed a folder containing materials related to the twoproducts (a camera and an athletic shoe). The materials in-cluded background information, a Consumer Reports arti-cle, and draft copies of advertisements for both products.After subjects finished reviewing the materials, they wereasked to record their thoughts related to the two products onan audio tape. They were asked to point out specifically thepositive qualities of the brand that the company could use inits advertising and were encouraged to suggest a pbtentialendorsement or slogan for each product.

The manipulation for commitment was administered afterthe subjects had tape-recorded their thoughts. The sbbjectsin the high-commitment condition were asked if the Mizunocorporation could use their taped thoughts about the brandalong with their photographs in the company's advertisingand publicity campaigns. The subject was photographed andasked to sign a release statement to this effect. This induc-tion follows the procedure used in prior commitment stud-ies, which have shown that public attachment of self to thetarget results in increased commitment toward it (e.g.,Halverson and Pallak 1978; Keisler 1971). This proceduredirectly follows from the definition of commitment as thepledging or binding of the individual to behavioral acts( Keisler 1971) and refers to the associations between peo-ple's attitudes and their overt, often public behaviors in sup-port of that attitude. The subjects in the low-comniitmentcondition underwent the same procedure but were asked torelease their thoughts related to the filler camera brand.Therefore, subjects in both the high- and the low-cpmmit-ment conditions went through exactly the same set pf pro-cedures and provided thoughts for both products. The onlydifference was the brand for which they signed the releaseand were photographed.

To examine the effects of commitment on negative infor-mation processing, the subjects were exposed to the nega-tive brand-related information after the commitment ilrianip-ulation. When the experimenter went to his desk to get thequestionnaire, he acted surprised to find a loose-leaf piage onthe desk and inquired whether the subjects had read thepage. When the subjects confirmed that they had not, ihe ex-perimenter apologized and told the subjects that the pagehad apparently slipped out of the folder unnoticed and re-quested that the subjects read it before they filled ciut thequestionnaire. The binder holes in the "missing page" weredeliberately torn, so that the mishap could be explained eas-ily. The missing page was the newspaper article used inExperiment 1, but it featured Mizuno instead of Nike as thebrand name. After reading the article, subjects filled but thedependent measures questionnaire.

Two control groups (high- and low-commitment) wefe usedfor assessing the manipulation checks (commitment, attitude,product involvement, and ambivalence). These manipulationchecks were not administered to the experimental groups be-cause of the potential for demand artifacts. Subjects in thje con-trol group performed the same tasks as did those in the Exper-imental group but did not see the target article. That is, theydid not go through the mishap of the missing page. All st bjectswere thoroughly debriefed and quizzed for potential hypothe-ses guessing. One subject reported being suspicious abOut thearticle slipping out and was dropped from the analysis.]


Dependent Variables

Attitudes, cognitive responses, commitment, ambiva-lence, and perceived diagnosticity were measured as inExperiment 1. Thoughts were also coded in a similar man-ner. The brand-level involvement measure was based onZaichkowsky's (1985).

Results

Manipulation checks. In the control groups, as expected,high-commitment subjects reported a significantly higherlevel of commitment toward the target brand than the low-commitment subjects (meanlo , = 2.14, meanhjgi, = 3.71;F(i,32) = 7.37, p < .01) but had equivalent attitudes(meanio v = 5.74, mean jgi, = 5.65), associated with equiva-lent levels of ambivalence (mean|o^ , = 1.53, meanhjgh =1.53), and were not significantly different in their involve-ment with the target brand (meanig^y = 4.00, meanhjgt, = 4.60;

Attitudes. As anticipated, low- (versus high-) commitmentsubjects reported significantly less positive (mean|(J , = 4.23,meanhigh = 5.21; F(i 34) = 4.99, p < .05) and more ambiva-lent attitudes (meani ^^ = 3.44, meanhigh = 2.44; F(i 34) -5.20, p < .05) after exposure to negative information.

Process evidence. Consistent with the results of the firstexperiment, low-commitment subjects perceived negativeinformation to be more diagnostic than the high-commit-ment subjects (meanio,,, = .68, mean jgh = .61, F(i 34) = 3.89,p < .06). Conversely, high-commitment subjects generatedsignificantly more counterarguments in response to the neg-ative information than the low-commitment subjects (mean-high = 4.06, mean,o^ = 2.11; F(,34) = 9.13, /? < .005).

Regression analysis. Mean attitude was regressed on per-ceived diagnosticity and counterarguments for both high-and low-commitment consumers. Although perceived diag-nosticity emerged as a marginally significant predictor of at-titudes for the low-commitment consumers (beta = .40, t =1.72, p < .10), it failed to predict attitudes for high-commit-ment consumers (beta = .04, t = .15, p > .80).Counterarguments emerged as a significant predictor of atti-tudes for both high- (beta = .50, t = 2.28, p < .05) and low-commitment (beta = .61, t = 3.07, p < .01) consumers.

Thus, the results of Experiment 2 increased our confi-dence in the findings of the first experiment. High-commit-ment consumers counterargued the negative information toa greater extent than did low-commitment consumers.Furthermore, counterargumentation, and not the perceiveddiagnosticity of negative information, predicted attitudechange for the high-commitment consumers. Conversely,perceived diagnosticity of negative information emerged asa significant predictor of attitudes for low-commitment con-sumers. This replication was achieved in the context of alow-share brand and when commitment was manipulated.To that extent, the confidence in the generalizability of ourfindings is enhanced. How our findings can affect the com-munication strategy for a company facing negative publicityis the focus of Experiment 3.

EXPERIMENTS

The processing differences between low- and high-com-mitment consumers, observed in Experiments 1 and 2, sug-gest that a common strategy as suggested by several case-

based studies (e.g., Johnson 1993) may not be equally ef-fective in combating the ill effects of negative infonnationfor consumers with different levels of commitment. Prior re-search indicates that strategies that provide consumers withadditional information that they did not spontaneously con-sider are likely to be more effective in persuasion than arestrategies that duplicate or match these processes (e.g.,Kardes 1988).

Specifically, because high-commitment consumers en-gage in extensive counterargumentation of the negative in-formation, a company response that counterargues the pub-licity may just replicate their spontaneous processes.However, a response focusing on the diagnosticity of the in-formation would provide them with information not consid-ered spontaneously upon exposure to the negative message.Therefore, this response would be more likely to enhancepersuasion than one that focused on counterarguments.

The low-commitment consumers, in contrast, appear toconsider spontaneously the perceived diagnosticity of the at-tribute information. However, the number of counterargu-ments was not significantly different in the negative andpositive conditions. Therefore, a response strategy that pro-vides these consumers with counterarguments is likely to bemore effective in changing attitudes than one that focuses onthe perceived diagnosticity of the information. Thus,

H4: For the high-commitment consumers, a response strategyfocusing on the perceived diagnosticity of the negative in-formation is likely to be more persuasive than one that coun-terargues it, whereas for the low-commitment consumers,the counterargumentation response is likely to be more per-suasive than the diagnosticity response.

Methodology

To test the proposed hypotheses, a 2 (commitment: highand low) x 2 (response strategy: counterargumentation anddiagnosticity) between-subjects design was implemented.

The strategies. Counterargumentation is one of the mostpopular strategies currently used by companies in dealingwith negative publicity. It involves arguing against the neg-ative publicity by questioning its validity. The diagnosticityresponse strategy, in contrast, focuses on reducing the valueof the negative information for discriminating between al-ternative brands in the product category. For example, giventhe negative publicity message used in Experiments 1 and 2,the counterargumentation response was developed to focuson the reliability of the data, the completeness of the data,and the validity of the sample. The diagnosticity response,in contrast, argued that the target brand was not differentfrom the others on the focal attribute, because all brandsused the same shock absorption technology.

The two response strategies were operationalized asnewspaper advertisements, which is a common mediumthrough which companies respond to negative publicity. In apretest, both versions of the advertisements were rated asequivalent in believability (mean(.oun,er = 4.7, meandjagngs =4-9; Fi_i8 = .12, p> .70) and strength (meancounter =5 .1 ,meanjiagnos = 5.0; Fj jg = .06, p > .80). The advertisementswere also approximately equal in length.

A series of pretests was conducted to establish the valid-ity of the strategy manipulations. In the first pretest, 30 sub-jects were given definitions of the two strategies, shown oneof the two advertisements, and asked to classify it into one


of the following categories: diagnosticity, counterargumen-tation, a combination of the two, or neither of the two. Allsubjects who saw the counterargumentation advertisementand two-thirds of those exposed to the diagnosticity adver-tisement classified them appropriately.

In the next pretest, 20 subjects rated the diagnosticity ofthe negative publicity information after seeing a companyresponse advertisement (counterargumentation or diagnos-ticity). Supporting the validity of our operationalization,subjects who saw the diagnosticity advertisement rated thenegative information as not at all diagnostic (mean = .50),whereas subjects who saw the counterargumentation adver-tisement perceived it as significantly more diagnostic(mean^ounter = -66, mean iagnos = -50; F,,i8 = 4.92, p < .05).Therefore, the diagnosticity response advertisement reducedthe perceived diagnosticity of the attribute information inthe negative publicity article.

The counterargumentation strategy, conversely, was ex-pected to result in thoughts questioning the validity of theinformation contained in the article. In a third pretest, 30subjects were exposed to the target newspaper article fol-lowed by one of the company response advertisements.Subjects were asked to list the thoughts they had while read-ing the company response advertisement. As expected, sub-jects in the counterargumentation condition had signifi-cantly more thoughts questioning the validity of thenewspaper article than subjects in the diagnosticity condi-tion (meancounter - 2.07, meandjagnos = -67; Fi_28 = 8.92, p <.006). Therefore, the advertisements were viewed as effec-tive operationalizations of the two response strategies.

Commitment. Brand commitment was measured. Nikewas the target brand.

Procedure and dependent variables. Using the same pro-cedure as in Experiment 1, we recruited subjects (35 high-and 35 low-commitment) over the telephone on the basis oftheir commitment scores. Subjects were told that they wereparticipating in a two-part media study: In the first part, theirtask was to evaluate recent newspaper articles related to dif-ferent products; in the second part, they would evaluatenewspaper advertisements. All subjects read four articles—two negative and two positive toward the product/brand fea-tured in them. One of them was the negative target articlefrom Experiment 1. After reading the articles, subjects weregiven a questionnaire, which asked them to rate the qualityof the articles and indicate their attitude toward the products.In the second part, they read advertisements and were in-formed that the companies ran these advertisements in re-sponse to the publicity received in the newspaper articlesfeatured in the first part of the study. Subjects were asked torate each advertisement on how believable, strong, and ap-propriate it was. After completing the second questionnaire,subjects reported their brand evaluations again.

Two levels of attitude change were computed: attitudechange in response to negative infonnation (post-negativemessage attitude minus prestudy attitude) and attitudechange after exposure to the response strategy (postadver-tisement attitude minus post-negative message attitude).

Results and Discussion

Consistent with the results of Experiments 1 and 2, low-commitment consumers exhibited a significant decline intheir attitude toward Nike after being exposed to negative

publicity about the brand (mean = -1.53; p •< .01).Surprisingly, a small but significant attitude change Was ob-served in the high-commitment subjects as well (mean =-.68; p < .05). However, the attitude change was signifi-cantly greater in the low- than the high-commitmejnt con-sumers (F = 21.27, p < .001). Thus, the pattern of data ob-tained in the first two experiments was replicated. ;

In their reactions to the company response advertise-ments, high-commitment consumers were expected to ex-hibit greater attitude change and less ambivalent attitudeswith the diagnosticity than with the counterargumentationresponse strategy. However, low-commitment consumerswere hypothesized to demonstrate greater attitude 'changeand less ambivalent attitudes when exposed to the counter-argumentation as compared with the diagnosticity m'essage.

Although the commitment x response strategy interactionwas significant for the attitudinal ambivalence measure (F =7.37, p < .01), it approached significance for the postadver-tisement attitude change variable (F = 2.81, p < . 10). Furtheranalyses revealed that this interaction was generated by apattern of mean attitude scores, which were consistent withour predictions. The high-commitment consumers exjhibitedgreater attitude recovery and lower ambivalence when con-fronted with a diagnosticity advertisement than a counterar-gumentation advertisement (means = .76 versus means = .35for attitude, p < .07; means =1.1 versus means = 2.0 for am-bivalence, p < .05). In contrast, a reverse pattern Wa§ foundfor the low-commitment consumers. Although attitude re-sults were in the expected direction, with greater chiknge inthe counterargumentation condition (mean = .46) than in thediagnosticity condition (mean = .24), this difference \yas notsignificant. However, low-commitment consumers j exhib-ited lower ambivalence after reading the counterargumenta-tion advertisement (mean = 1.8) versus the diagnosticity ad-vertisement (mean = 2.67; p < .05). Thus, the pattern ofresults generally supported H4.

GENERAL DISCUSSION AND CONCLUSlOl^

Negative information about brands and companies iswidely prevalent in the marketplace. Yet except fpr casestudies on this topic, there has been no systematic invbstiga-tion of how consumers process negative publicity informa-tion about the brands they like and use and which strategiesto counter the negative publicity are appropriate. It» threestudies reported in this article, we attempt to bridge this gapin the literature.

In the first two studies, we found that commitment is animportant moderator of consumer response to negatjve in-formation. Specifically, the response patterns of high- andlow-commitment consumers are very different. Consumerswho are committed to a brand instinctively counterarguenegative information about that brand. These deffensiveprocesses mitigate the ill effects of that information in thatthey reduce the likelihood of attitude degradation. Low-commitment consumers, conversely, counterargue the nega-tive information to a lesser degree. Furthermore^ eventhough low-commitment consumers seem to like the brandas much as the high-commitment consumers do, they Exhibitgreater attitude change and increased attitude ambivalenceupon exposure to negative information about it. This attitudedegradation is driven by their perceptions of the diagnostic-ity of negative information. Therefore, an important contri-


bution of our work emerges from our melding of the litera-ture on biased assimilation and impression formation anddemonstrating that a person's level of commitment dictateswhich literature is more applicable in determining responseto negative communications.

In addition, whereas most research has focused on thenegativity effect and the reasons for that effect, our researchuncovers its limitations. That consumers focus on negativeinformation because it is more diagnostic is now well ac-cepted in the impression formation literature. We show thatthe consumer's level of commitment qualifies this effect.Specifically, the negativity effect appears to be more likelywhen a consumer's commitment to the target object or issueis low. At high levels of commitment, not only is the nega-tivity effect absent, but also consumers actually consideredpositive information about that object more diagnostic thannegative information about it. This reversal of the negativityeffect attests to the powerful nature of commitment as amoderator of communication effects.

Our results suggest that a committed consumer can resistinformation effectively that is likely to induce switching be-havior: negative information about the target and positiveabout the competition. This information processing biasmay therefore lead to repeat purchase behavior observed forcommitted consumers. In other words, our research providesan insight into why and how committed consumers engagein repeat purchase behavior.

The current research attests to the value of customer com-mitment or loyalty for a company. Not only does it enableconsumers to resist negative information, but it also en-hances the impact of positive publicity. This finding comesin the wake of market data that shows brand commitment ison the decline (e.g., Schriver 1997). Further theoretical re-search examining how companies can enhance customers'commitment toward their products would be desirable.

Currently, most companies use a "mass approach" in re-sponding to negative publicity (see, e.g., Pearson andMitroff 1993; Weinberger, Romeo, and Piracha 1991). Ourresearch argues that they should consider using a targetedapproach. Different response strategies are likely to be moreeffective for high- and low-commitment consumers.Furthermore, an important contribution of our work lies inits delineation of the two types of response strategies—counterargumentation and diagnosticity—and our ability topredict a priori the effectiveness of a strategy given the com-mitment level of consumers. Thus, marketers worried aboutthe effects of negative publicity on their loyal consumersshould attempt to use the diagnosticity strategy, because theconsumers might already have generated the necessarycounterarguments. In contrast, when addressing a consumersegment that likes the brand but is not committed to it (e.g.,is part of the consumer's consideration set), marketers areadvised to focus more on the counterargumentation strategy.

Our research also underscores the importance of measur-ing dimensions of attitude other than bipolar valence. Wemeasured the effects of negative information on both attitu-dinal ambivalence and bipolar valence measures. Prior re-search suggests that in the context of negative information,measures of ambivalence may be more appropriate thanbipolar measures of valence (see Cacioppo, Gardner, andBerntson 1997; Haugtvedt et al. 1994). Our results (espe-cially Experiment 3) demonstrate that the effects of negative

information emerge more clearly with the ambivalencemeasures. Furthermore, the ambivalence of a brand attitudeconveys to the marketer important information relating tothe ability of this attitude to predict the consumer's futurepurchase behavior. The more ambivalent an attitude, the lesslikely is it to predict behavior.

The external validity of our findings can be examinedwithin the context of background factors controlled in ourresearch setting (Lynch 1982). Although some of these fac-tors, such as age of the respondent, are not expected to in-teract with the variables of interest, others may be likely to.First, our research focuses on product attribute-related in-formation. However, previous literature suggests that theperceived diagnosticity of negative information is likely tobe higher in the morality (e.g., company values) versus theability (e.g., product attributes) domain, which leads togreater weighting of negative information in the former do-main (Skowronski and Carlston 1987). Thus, the findingsobtained here may not be generalizable to negative informa-tion reflecting a company's values (e.g., its managementpractices such as bribery and sexual harassment). Further re-search could examine this issue in an experimental settingby manipulating the type of negative information (values-re-lated versus attributes-related).

Second, although the messages in our experiments werehighly positive (negative), they did not imply extreme orlife-threatening consequences. However, prior research(Fiske 1980) indicates that extreme information is perceivedas more diagnostic than moderate information and thereforeis weighted more in overall evaluations. It could be arguedthat extremely negative information (e.g., relating to the"transgressions" that Fournier [1998] discusses) might pro-vide a boundary condition to our findings. It may be diffi-cult for even the committed consumers to discount thishighly diagnostic information. Moderate information, con-versely, because of its lower diagnosticity, may lead to an at-tenuation of the negativity effect for the low-commitmentconsumers. These predictions could be tested in an experi-mental setting by manipulating the extremity of the negativeinformation.

Finally, our research focuses on negativity in the context ofpublicity. However, the findings could be extended to othertypes of negative information in the marketplace, for exam-ple, negative advertising, comparative advertising, and nega-tive word of mouth. Given that most types of negative infor-mation differ primarily on the basis of their source and itscredibility (e.g., advertising versus publicity), factoring in theeffect of the source of information in the framework wouldfurther enrich its generalizability. For example, Sternthal,Dholakia, and Leavitt (1978) suggest that counterattitudinalnegative information conveyed by a credible source, such aspublicity, is likely to be more damaging than the same infor-mation presented by a less credible source, such as advertis-ing. Furthermore, Shiv, Edell, and Payne (1997) suggest thatin the advertising context, negative framing may, under cer-tain conditions, generate tactics-related cognitions, therebydiminishing the impact of negative information.

REFERENCES

Advertising Age (\995), "Numbers for the 9O's," 66 (October 23), 3.Baron, Reuben M. and David A. Kenny (1986), 'The Moderator-


Mediator Variable Distinction in Social Psychological Research:Conceptual, Strategic, and Statistical Considerations," Journalof Personality and Social Psychology, 3\ (6), 1173-82.

Beatty, Sharon E., Lynn R. Kahle, and Pamela Homer (1988), 'TheInvolvement-Commitment Model: Theory and Implications,"Joumal of Business Research, 16 (2), 149-^7.

Bond, Jonathan and Richard Kirshenbaum (1998), Under theRadar: Talking to Today's Cynical Consumer. New York: JohnWiley & Sons.

Cacioppo, John T., Wendi L. Gardner, and Gary G. Bemtson(1997), "Beyond Bipolar Conceptualizations and Measures: TheCase of Attitudes and Evaluative Space," Personality and SocialPsychology Review, 1 (January), 3-22.

Chaiken, Shelly, Akiva Liberman, and Alice H. Eagly (1989),"Heuristic And Systematic Processing Within and Beyond thePersuasion Context," in Unintended Thought, J.S. Uleman andJohn A. Bargh, eds. New York: Guilford Press, 212-52.

Chisholm, Samuel J. (1998), "Coping vk-ith Minority Crises,"Advertising Age, 69 (April 6), 24.

Collins, Linda M. (1996), "Is Reliability Obsolete? A Commentaryon 'Are Simple Gain Scores Obsolete,'" Applied PsychologicalMeasurement, 20 (September), 289-92.

Cronbach, Lee J. and Lita Furby (1970), "How We ShouldMeasure Change—Or Should We?" Psychological Bulletin, 74(July), 68-80

Dick, Alan S. and Kunal Basu (1994), "Customer Loyalty: Towardan Integrated Conceptual Framework," Journal of the Academyof Marketing Science, 22 (2), 99-113.

Ditto, Peter H. and David F. Lopez (1992), "Motivated Skepticism:Use of Differential Decision Criteria for Preferred andNonpreferred Conclusions," Joumal of Personality and SocialPsychology, 63 (April 6), 568-84.

Dye, Patrick (1997), "Emergency Services," Marketing Week, 20(April), 73-74.

Eagly, Alice H. and Shelly Chaiken (1995), "Attitude Strength,Attitude Structure, and Resistance to Change," in AttitudeStrength: Antecedents and Consequences, Richard E. Petty andJon A. Krosnick, eds. Mahwah, NJ: Lawrence EribaumAssociates, 413-32.

Edwards, Kari and Edward E. Smith (1996), "A DisconfirmationBias in the Evaluation of Arguments," Joumal of Personalityand Social Psychology, 11 (July), 5-24.

Feldman, Jack M. and John G. Lynch (1988), "Self-GeneratedValidity and Other Effects of Measurement on Belief, Attitude,Intention, and Behavior," Journal of Applied Psychology, 73(August), 421-35.

Fiske, Susan T. (1980), "Attention and Weight in PersonPerception: The Impact of Negative and Extreme Behavior,"Journal of Personality and Social Psychology, 38 (6), 889-906.

Fournier, Susan (1998), "Consumers and Their Brands:Developing Relationship Theory in Consumer Research,"Joumal of Consumer Research, 24 (March), 343-73.

Gross, Sharon Ruth, Rolf Holtz, and Norman Miller (1995),"Attitude Certainty," in Attitude Strength: Antecedents andConsequences, Richard E. Petty and Jon A. Krosnick, eds.Mahwah, NJ: Lawrence Eribaum Associates, 215-45.

Halverson, Richard R. and Michael S. Pallak (1978),"Commitment, Ego-Involvement, and Resistance to Attack,"Journal of Experimental Social Psychology, 14 (January), 1-12.

Harris, C.W. (1967), Problems in Measuring Change. Madison:University of Wisconsin Press.

Haugtvedt, Curtis P., David W. Schumann, Wendy L. Schneier, andWendy L. Warren (1994), "Advertising Repetition and VariationStrategies: Implications for Understanding Attitude Strength,"Joumal of Consumer Research, 2\ (June), 176-89.

Herr, Paul M., Frank R. Kardes, and John Kim (1991), "Effects ofWord-of-Mouth and Product-Attribute Information onPersuasion: An Accessibility-Diagnosticity Perspective,"

Joumal of Consumer Research, 17 (March), 454—62.Irvine, Robert B. (1993), "The Crisis Outlook for |'92 and

Beyond," Public Relations Joumal, 48 (October), 8+.Jacoby, Jacob and Robert W. Chestnut (1978), Brand\ Loyalty

Measurement and Management. New York: John Wiley & Sons.Johnson, David G. (1993), "Crisis Management: Forewarned Is

Forearmed," Joumal of Business Strategy, 13 (2), 58-62.Kaplan, Kalman J. (1972), "On the Ambivalence-Indifference

Problem in Attitude Theory and Measurement: A SuggestedModification of the Semantic Differential Technique,"Psychological Bulletin, 77 (5), 361 -72. '

Kardes, Frank R. (1988), "Spontaneous Inference Processes inAdvertising: The Effects of Conclusion Omission andInvolvement on Persuasion," Journal of Consumer Research, 15(September), 225-33. '

Keller, Kevin Lane (1993), "Conceptualizing, Measuring, andManaging Customer-Based Brand Equity," Journal ofMarketing, 51 {\), \-ll.

Kiesler, Charles A. (1971), The Psychology Of Commitment:Experiments Linking Behavior to Belief. San Diego: AcademicPress. I

Klein, Jill G. (1996), "Negativity in Impressions of PresidentialCandidates Revisited: The 1992 Election," Personality andSocial Psychology Bulletin, 22 (March), 289-96.

Kroloff, George (1988), "At Home and Abroad: Weighing In,"Public Relations Joumal, 44 (October), 8.

Krosnick, Jon A., David S. Boninger, Yao C. Chuang, Matthew K.Berent, and Catherine G. Camot (1993), "Attitude Strength: OneConstruct or Many Related Constructs?" Joumal of Personalityand Social Psychology, 65 (December), 1132-51.

Kunda, Ziva (1990), "The Case for Motivated Reasoning,"Psychological Bulletin, 108 (November), 480-98. '

Lastovicka, John L. and David M. Gardner (1978), "Componentsof Involvement," in Attitude Research Plays for High Stakes,J.L. Maloney and B. Silverman, eds. Clhicago: AmericanMarketing Association, 53-73. I

Lynch, John G. (1982), "On the External Validity of Experimentsin Consumer Research," Joumal of Consumer Research, 9 (3),225-39.

Maheswaran, Durairaj and Joan Meyers-Levy (1990), "TheInfluence of Message Framing and Issue Involvement," Journalof Marketing Research, 27 (August), 361-67. j

Marconi, Joe (1997), Crisis Marketing: When Bad Things Happento Good Companies. Chicago: American Marketing Association.

Morgan, Robert M. and Shelby D. Hunt (1994), "ACommitment-Trust Theory of Relationship Marketing," .Journalof Marketing, 58 (July), 20-38. '

Pearson, Christine M. and Ian I. Mitroff (1993), "Froni CrisisProne to Crisis Prepared: A Framework for Crisis Management,"Academy of Management Executive, 1 (1), 48-59. ,

Petty, Richard E. and John T. Cacioppo (1977), "ForeWaming,Cognitive Responding, and Resistance to Persuasion," Joumalof Personality and Social Psychology, 35 (9), 645-55.

and Jon A. Krosnick (1995), Attitude Strength:Antecedents and Consequences. Mahwah, NJ: LawrenceEribaum Associates.

Pomerantz, Eva M., Shelly Chaiken, and Rosalind S. Tordesillas(1995), "Attitude Strength and Resistance Process," Journal ofPersonality and Social Psychology, 69 (September), 408-j-19.

Rogosa, David (1988), "Myths About Longitudinal Research," inMethodological Issues in Aging Research, K.W. Schaie, et al.,eds. New York: Springer.

Sawyer, Alan (1973), 'The Effects of Repetition of Refutationaland Supportive Advertising Appeals," Journal of MarketingResearch, 10 (February), 23-33.

Schriver, Steve (1997), "Customer Loyalty: Going, Going ...,"American Demographics, 19 (September), 20-23.


Shiv, Baba, Julie A. Edell, and John W. Payne (1997), "FactorsAffecting the Impact of Negatively and Positively Franfied AdMessages," Joumal of Consumer Research, 24 (December),285-94.

Skowronski, John J. and Donal E. Carlston (1987), "SocialJudgment and Social Memory: The Role of Cue Diagnosticity inNegativity, Positivity, and Extremity Biases," Joumal ofPersonality and Social Psychology, 52 (4), 689-99.

and (1989), "Negativity and Extremity inImpression Formation: A Review of Explanations,"Psychological Bulletin, 105 (January), 131-42.

Stemthal, Brian, Ruby Dholakia, and Clark Leavitt (1978), 'ThePersuasive Effect of Source Credibility: Tests of CognitiveResponse," Joumal of Consumer Research, 4 (March), 252-60.

Weinberger, Marc G., Jean B. Romeo, and Azhar Piracha (1991),"Negative Product Safety News: Coverage, Responses, andEffects," Business Horizons, (May/June), 23-31.

Wood, Wendy, Nancy Rhodes, and Micheal Biek (1995), "WorkingKnowledge and Attitude Strength: An Information-ProcessingAnalysis," in Attitude Strength: Antecedents and Consequences,Richard E. Petty and Jon A. Krosnick, eds. Mahwah, NJ:Lawrence Eribaum Associates, 283-13.

Wright, Peter (1974), 'The Harassed Decision Maker: TimePressures, Distractions, and the Use of Evidence," Joumal ofApplied Psychology, 59 (May), 555-61.

Zaichkowsky, Judith L. (1985), "Measuring the InvolvementConstruct," Joumal of Consumer Research, 12 (December),341-52.

One of the world's leading business journals zf bfis now available in English: ^

sbrSchmaleiibach Business Review

' addresses the international scientific community with excellent andinnovative artides. It covers topics especially in accounting, finance, marketing,decision theory, and neo-institutionalism. Thus, s iar attempts to bringtogether the most important management areas. The approach of notspecializing in one specificsubject, such as accounting or finance,stands in the tradition of G^^^business research journals.

The first issue sfarts

Documents

Consumer Response to Negative Publicity