When !@#? (Bad Stuff) Happens...Effects of Related and Unrelated Positive Associations on the Influence of Negative Secondary Associations

Erica Mina Okada, The Wharton School, University of Pennsylvania
David J. Reibstein, The Wharton School, University of Pennsylvania
ABSTRACT - Brand associations can influence purchase likelihood even when they are not directly linked to the product attributes. This study examines the mitigating effects of positive secondary associations on purchase likelihood to counter-balance negative secondary associations. Experimental results suggest that the negative effects of negative publicity are exacerbated by corporate sponsorships in the same area, and mitigated by sponsorships in an unrelated area.
[ to cite ]:
Erica Mina Okada and David J. Reibstein (1998) ,"When !@#? (Bad Stuff) Happens...Effects of Related and Unrelated Positive Associations on the Influence of Negative Secondary Associations", in NA - Advances in Consumer Research Volume 25, eds. Joseph W. Alba & J. Wesley Hutchinson, Provo, UT : Association for Consumer Research, Pages: 349-356.

Advances in Consumer Research Volume 25, 1998      Pages 349-356

WHEN !@#? (BAD STUFF) HAPPENS...EFFECTS OF RELATED AND UNRELATED POSITIVE ASSOCIATIONS ON THE INFLUENCE OF NEGATIVE SECONDARY ASSOCIATIONS

Erica Mina Okada, The Wharton School, University of Pennsylvania

David J. Reibstein, The Wharton School, University of Pennsylvania

ABSTRACT -

Brand associations can influence purchase likelihood even when they are not directly linked to the product attributes. This study examines the mitigating effects of positive secondary associations on purchase likelihood to counter-balance negative secondary associations. Experimental results suggest that the negative effects of negative publicity are exacerbated by corporate sponsorships in the same area, and mitigated by sponsorships in an unrelated area.

Exxon’s oil spill accident in Alaska, Union Carbide’s deadly explosion in India, allegations of racially discriminatory practices at Texaco, and sexual harassment suits against Bear Stearns are all examples of negative publicity which can create a negative image in the mind of the consumer. Negative images are sometimes a function of the type of industry in which a business operates. Paper companies and chemical companies may be thought of as polluters. Negative associations are often unanticipated and/or beyond the control of the management team. Accidents such as oil spills, fatal explosions, and airplane crashes happen unexpectedly. Lrge businesses are not able to prevent bad rumors from starting and spreading, nor are they able to manage the actions of every single employee, but allegations of environmental violation, racial discrimination, or sexual harassment can tarnish the image of a company. When companies are faced with a sudden development of negative associations, they are often unprepared to respond efficiently and expediently. Corporate image impacts the bottom line. Even when associations are secondary, that is, not linked to the product itself, positive associations increase the likelihood of purchase of a brand, and negative associations decrease the likelihood of purchase. Okada and Reibstein (1996) suggested, not surprisingly, that negative secondary associations such as environmental violations and allegations of sexual harassment, have an adverse effect on a customer’s likelihood of choosing a brand as his final purchase, and positive secondary associations such as environmentally friendly business practices and corporate donations for the cure of diseases have a positive effect. In this study, we considered how a firm can build positive secondary associations to mitigate negative associations.

Corporate sponsorship, including sponsorships of philanthropic, civic, and worthy causes, is one way of cultivating goodwill, and developing favorable brand associations. Corporate policy, such as affirmative action, or publicized social criteria for doing business with certain countries, is another way to establish customer goodwill. For decades, companies have tried to enhance their image by communicating their acts of philanthropy and support of worthy causes to the consumers. Corporate sponsorship is a balance of two goals: social action and profit maximization. In general, this balance for large US, Japanese, and Western European firms is primarily profit maximization, though social goals are also important (Farmer and Hogue, 1985).

Companies put a lot of resources into their corporate giving. Georgia Pacific runs commercials on national television to inform viewers that the company is supporting the conservation of nature in the Upper Roanoke Valley. American Airlines runs television commercials describing its program to help children in need, and encourages its customers to help out in its efforts. Working Assets, a new long distance telephone company, sends solicitation mail inviting its prospective clientele to switch to a long distance company that donates a "healthy chunk" of their profits to nonprofit organizations such as Amnesty International, Greenpeace, and Planned Parenthood. Companies such as Ben & Jerry’s Ice Cream and the Body Shop are known for allocating a percent of their revenue to charitable organizations. These and other companies are investing in activities that are outside of their line of business, and in communicating their support of these causes. Aside from helping worthy causes, do these efforts increase the sales and profits of the donating firms? Are they spending money on corporate sponsorships that are most effective in neutralizing the negative effects of unplanned negative associations, and enhancing a consumer’s purchase likelihood?

The primary question addressed in this research is: Is the best way to mitigate negative secondary associations with a positive effort in the same area (related), or with an effort in an unrelated area altogether? In the first part of the paper, we introduce theories that explain how positive and negative images affect purchase. In the following section, we describe the experiment that we conducted to test our hypotheses. We then present the results, and explain the methods of analysis that we used.

THEORETICAL DEVELOPMENT

Brand association is anything that is linked in memory to a brand, and it summarizes information about the product into one node in the consumer’s memory (Aaker, 1991). It can be anything from a logo or ymbol, to a salient feature of the brand, to a typical user. Associations can influence the recall of information as well as interpretation of facts. Secondary associations are linked to other information in memory that is not directly related to the product. Secondary associations are generally associations related to the company, such as country of origin of the firm, and spokesperson for the brand. Corporate sponsorship, corporate policy, and negative publicity, which are the topic of this research, are also secondary associations. A set of such brand associations makes a brand image. A brand image is the perception of a brand as reflected by a network of brand associations in consumer memory.

A consumer’s brand beliefs can be (1) created by the marketer, (2) formed by the consumer himself through direct experience with the product, and/or (3) formed by the consumer through inferences based on existing associations. Brand equity can be developed regardless of the source of brand beliefs (Fishbein and Ajzen, 1975) so long as the brand association is favorable, strong, and unique (Keller, 1993). Therefore, a marketer can influence a consumer’s brand belief by developing positive associations

Because secondary associations do not pertain to the product’s attributes, the effect on consumer choice may be overshadowed by other primary associations. While some consumers may place personal importance on particular negative publicity, corporate sponsorships, or policies, others may not. However, given a set of similar choices, a consumer directs more attention to otherwise less salient attributes that are discriminating (Tversky, 1977). Even to those consumers who may not necessarily make purchase decisions based on their personal beliefs and values, secondary associations may be discriminating factors in brand evaluation and choice.

Information processing strategy explains individual behavior as a consequence of how information about an object is encoded, stored and retrieved (Bettman 1979). Information about an object is stored as its associations in multiple locations in an individual’s memory. Attributes are actively associated with an object in a process called rehearsal, then stored in a systematic way. Incoming information may stimulate the activation of object-related thoughts that have been processed earlier.

When a consumer is exposed to information about a company’s negative associations, whether it be polluting, product liability, or sexual harassment, he links these associations to the brand, and stores it in memory. In evaluating the brand, the consumer retrieves associations that are linked to it in memory, including such negative associations.

There are two possible ways to develop positive associations in order to mitigate existing negative associations. One can either develop positive associations that are directly related to the existing negative association (and therefore likely to be retrieved together with the negative associations), or alternatively, one can develop positive associations that are totally unrelated (and likely to be retrieved independently of the negative associations). In a case where consumers are exposed to negative associations such as allegations that a company is engaged in environmentally unfriendly business practices, an example of a related positive association would be to sponsor green peace: an association with something environmental and positive. An example of an unrelated positive association would be the sponsorship of a scholarship fund: an association with something positive and unrelated to the environment. The first step in improving the image of a brand which has negative secondary associations would be to break or at least weaken the link between the brand and its negative secondary associations in the consumer’s memory. Prior research on priming and associative interference, as applied to secondary associations would suggest that positive secondary associations may prime negative associations in a related domain, but may interfere with negative associations in an unrelated domain. On the topic of priming, Herr (1989) found tha by unobtrusively presenting an exemplar from an attribute, that attribute becomes more salient , and more likely to be used by consumers in subsequent processing of new information. Yi (1990) suggests that contextual factors can affect consumers’ product evaluation by priming certain attributes. Bettman and Sujan (1987) showed that priming a decision criterion influenced the evaluation outcomes especially for novice consumers. On associative interference, Burke and Srull (1988) found that additional information about a brand may limit the consumer’s ability to recall old information about the brand. Increasing the number of links from the brand decreases the retrieval likelihood of each individual piece of associated information.

In a study by Tybout, Calder and Sternthal (1981), it was shown that even a rumor that is not credible, can have an adverse effect on consumers’ evaluation of a brand. In their study, the negative association was not secondary, but linked directly to a significant product attribute. Refuting the rumor did not help in raising consumer evaluation, but strategies that aimed at strengthening the brand’s other existing links were effective in damage control, and furthermore, strategies that created new positive links to the brand were most effective. This study investigates the same issue with respect to a brand’s secondary associations.

Also in the case of related positive associations, the consumer may see an inconsistency between the existing negative image and the positive association that is being developed. A related association may be discounted by the existing negative image. In some instances a related positive association may be seen as insincere or manipulative by the consumer (Campbell, 1995). For example, a tobacco company, which carries a negative image by the nature of the industry, may be seen as insincere if it sponsors cancer research.

A related positive association may serve as a reminder of the brand’s existing negative image. A related association may increase the rehearsal of the existing negative association, and strengthen the link between the brand and its negative image. In contrast, an unrelated positive association would have no effect on the rehearsal of the negative image. Developing an unrelated association will increase the number of positive links to the brand. In that sense, an unrelated positive association may have a diluting effect on the link with an existing negative image.

H1: When a brand has an existing negative secondary association, developing a related positive association is expected to strengthen the link between the brand and its negative association, thereby further exacerbating the negative effect on consumer purchase probability.

H2: When a brand has an existing negative secondary association, developing an unrelated positive association is expected to weaken the link between the brand and its negative association, thereby mitigating the negative effect on consumer purchase probability.

H1 suggests that a combination of negative publicity with related positive corporate sponsorships and policies would further decrease a consumer’s probability of purchase, as compared to the case when only negative publicity is present. H2 suggests that a combination of negative publicity with unrelated positive corporate sponsorships and policies would modify the negative impact of negative publicity.

DESCRIPTION OF THE EXPERIMENT

As part of a course requirment, 140 undergraduates at a major eastern university participated in the experiment. Airlines was chosen as the product class for two reasons. First, we wanted a product category with which the subjects were familiar. The students at the university are relatively geographically diverse, and are generally familiar with air travel. Second, the experimental setting maintains the essential elements of a real purchase situation. In purchasing an airline ticket, a consumer would typically make his travel arrangements by telephone, without actually inspecting the aircraft or the safety standards. In fact, he may reasonably call only a few airlines before making his purchase decision.

Ten attributes were identified in a pretest that was conducted on thirty doctoral students from the business and economics departments of said university. The pretest participants were asked to name the airlines with which they were familiar, and to list the qualities that they thought were important in choosing an airline for travel. From the pretest, the ten most frequently mentioned attributes were included in the survey.

Three airlines were chosen: one national carrier, one regional airline, and one low-cost carrier. The 140 subjects (I=140) each evaluated eight airlines including the three target airlines (X=3) and an additional five airlines that were decoys in the experiment, on ten attributes (K=10); that is, experimental manipulations were done on the X=3 airlines only. The subjects in the main experiment used a 6 point scale in their attribute evaluations Akxi’s, with 1 being the least favorable and 6 the most favorable.

After rating the airlines on each attribute, the subjects were asked how likely it would be for them to choose each of the airlines in traveling from City A to City B, given that all eight airlines, including the X=3 airlines and the five decoys, had service between the two cities. This estimated the purchase factor PFxi. Again, this task was done using a 6 point scale. The subjects rated the PFxi’s for the five decoys also, which served as another measurement of the effect of secondary associations on a consumer’s purchase likelihood (Table 1).

TABLE 1

Finally, the subjects were asked how many times they fly in a year, whether or not they have flown each of the airlines in the past, and in which of the three airlines’ frequent flyer programs they were enrolled.

The subjects were randomly assigned to one of four experimental conditions. Three-fourths of the subjects read a page of fabricated news about activities in the airline industry, to expose them to different combinations of negative and positive secondary associations. These subjects were exposed to these manipulations before they did the attribute ratings and before they rated the probability of airline choice. The four groups were:

1) The negative group (NEG) who read a manipulation page that included information relating each of the three companies to a negative activity: environmental violation or sexual harassment.

2) Like NEG, the related group’s (REL) manipulation page included negative information about the three airlines, but it also provided information about each of their involvement in a related positive activity as well. The airlines that were linked to sexual harassment were also supporting a battered women’s center, and those that were linked to environmental violations were also donating to the National Park Service or actively participating in a recycling program.

3) The unrelated group (UNR) read about the companies’ involvement in the same negative activities as well, but they also read about their involvement in an unrelated positive activity: sponsorship of ancer research, or participation in local community work. In each experimental condition, all three companies were manipulated simultaneously and in the same way.

4) The fourth group was the control group (CON) who executed only the task of evaluating the 30 combinations of Akxi’s, the 3 PFxi’s, and the five PFdi’s for the decoys. [Subscript x denoted the target companies, and subscript d denotes the decoys.]

EXPERIMENTAL RESULTS

The subjects who participated in the experiment flew, on average 5.5 times a year, ranging from zero to 24 times a year. With 140 subjects evaluating three airlines each, 140 x 3=420 data points were collected in this survey. There were 35 subjects in each of the four experimental conditions: control, negative, related, and unrelated.

We used three methods to measure the effect of experimental condition on the purchase likelihood (1) by comparing the signs and significance levels of the experimental condition parameter estimates in an OLS regression based on a "purchase factor" model, (2) by comparison of the mean PFxi’s by experimental condition, and (3) by comparison of the relative ratios of the PFX’s of the three target airlines to the PFD’s of the other five decoy airlines by experimental condition.

1. OLS Regression

The impact on purchase likelihood of different combinations of negative and positive associations is analyzed in the context of a "purchase factor" model. The purchase factor model that we develop in this analysis is derived from the basic concept of a simple additive conjoint model (Green and Rao, 1971, and Green and Srinivasan, 1978). If there are K determinant attributes for a product class, and brand x (x product class) has attribute level Akxi for each of the K attributes, the part-worth utilities for the K attributes to consumer i are Vki(Akxi). The purchase factor model retains the basic concept of the additive model, that the utility of the brand is the sum of the partworths, but also incorporates key ideas from the hybrid conjoint model (Green, 1984 and Green, Goldberg, and Montemayor, 1981). Some simplifications were made to arrive at the basic purchase factor model used in this analysis. First, as in the hybrid model, the attribute levels, Akxi’s are the consumer’s own subjective ratings, which can be expressed numerically on a rating scale. In the purchase factor model constructed here, the Akxi’s are further assumed to be linear. A unit change in Akxi is constant. The second simplification is that, the Vki’s are weights that are allocated to each of the K attributes, and these weights, Wk’s can be estimated for a consumer population, thereby assuming homogeneity. By these assumptions, utility can be expressed as:

(1) Uxi=Ek Wk Akxi .

Translating utility into probability of purchase becomes difficult, because (a) the translation from utility to preference is not deterministic (Eliashberg and Hauser, 1985), and (b) the transformation from preference to purchase probability is not a linear function. To circumvent these difficulties, we introduce the concept of "purchase factor". Purchase factor PFxi is an index of how likely it is for consumer i to purchase brand x. PFxi is a positive monotonic transformation of purchase probability, but it is not bound between 0 and 1, and it is not necessarily linear.

The effects of secondary associations on purchase likelihood are also incorporated into the model. As we saw in our previous research, negative secondary associations, such as involvement in environmentally unfriendly business practices, and allegations of sexual harassment should decrease a consumer’s likelihood of purchase. In the present study, we analyze how effective different types of positive secondary associations are in mitigating an existing negative effect. The PFxi model measures the comparative effects of being exposed to (a) negative secondary associations only, (b) negative secondary associations combined with related positive associations, and (c) negative associations combined with unrelated positive associations.

(2) PFxi=Ek Wk Akxi + n NEGxi + r RELxi + u UNRxi + C .

In (2), NEG, REL, and UNR are indicator variables for the negative, related positive and unrelated positive experimental conditions respectively, and n, r, and u estimate the effects of these experimental conditions on purchase likelihood. In the way that we designed the experiments, the (un)related positive group was exposed to negative and (un)related positive associations, therefore r (u) estimates the interaction of negative and (un)related positive associations. H1 predicts r to be more negative than n, and H2 predicts u to be greater than n.

Using OLS regression, the Wk’s, n, u, and r, in Equation (2) were estimated for the subject population, along with the three experience covariates.

An analysis of the correlation of the variables indicates that as expected, the correlations between the Akxi’s and PFxi were high (about 0.5) except for FARE which had low negative correlations. There was in general, relatively high correlation among the ten Akxi’s as well. Of the 45 possible two way correlations, 23 were greater than 0.5. FARE was the only variable that had low and negative correlations with the rest of the attributes. The correlation among the experimental group variables and the demographic covariates was generally low, which suggests that the subjects were randomly assigned to the four experimental groups, and that experience was relatively evenly distributed. The correlation among the experimental group variables and attribute ratings were also low.

The regression estimated the Wkx’s, the effects of experimental conditions, and demographic covariates on the purchase factor. The results are summarized in Table 2. The estimates that are significant at the 5% level are in bold. Eight of the ten attributes had positive estimates, which is directionally as predicted. FARE had a negative estimate, which is also directionally as predicted. Only three attribute estimates, SAFE, FLET, and COMF, had significant t statistics, but due to the high correlation among the attributes, the effects of the non-significant attributes on PFxi may have been picked up by the three significant attribute coefficients. Omitting an attribute could increase the t-statistic of an attribute that is highly correlated to the omitted attribute.

The coefficients of NEGxi, RELxi, and UNRxi, that is n, r, and u respectively, are estimates of the effect of the experimental conditions relative to the control condition. All three of these estimates were directionally as hypothesized, but not significant. H1 predicted r to be negative and/or less than n. r=-0.18518n= -0.01492<0, which is consistent with H1, but neither r nor n was significant. H2 predicted u to be positive and/or greater than r. r= -0.18518<0<u=0.2348, which is consistent with H2, but again u was also directionally as predicted but not significant. The results of OLS were suggestive but certainly not conclusive regarding the hypotheses.

Our model (2) measures only the direct effects of the experimental conditions on PFxi. If the different combinations of secondary associations impacted the PFxi’s through the Akxi’s, this halo effect (Wilkie, McCann, and Reibstein, 1973) would not be captured by n, r, and u. In order to test the halo effect, we compared the attribute ratings by experimental condition. For each of the ten Akxi’s, we did two way comparisons for Akxi:CON-Akxi:NEG as a manipulation check, Akxi:NEG-Akxi:REL and Akxi:NEG-Akxi:REL as a check for H1, Akxi:NEG-Akxi:UNR as a check for H2, and Akxi:REG-Akxi:UNR as a check for H1 and H2. No single pairwise difference was significant, but all ten Akxi:REL’s were less than the respective Akxi:NEG’s, which suggests that secondary associations, especially the combination of negative associations and a related positive, may have both a direct and indirect effect through the attributes on Pfxi.

2. Comparison of the mean PFxi’s

In the second approach to measuring the effect of the different combinations of negative publicity and corporate sponsorships, we compared the average PFxi’s by experimental condition, which would capture the combined direct and indirect effects. H1 predicts PFREL<PFNEG, and H2 predicts PFUNR>PFNEG. The overall mean PFxi’s, as well as the PFxi’s of each of the three airlines individually are shown by experimental condition in Table 3.

A simple comparison of the overall average PFxi of the four experimental conditions indicates differences in the predicted direction. First of all, subjects responded negatively to negative associations. PFNEG=3.276<PFCON=3.371 which serves as a manipulation check. Secondly, combining a negative association with a related positive association significantly magnifies the negative effect on consumer choice. PFREL=2.829<PFNEG=3.276, which would support H1. Also, combining a negative association with an unrelated positive association significantly mitigates the negative effect on consumer choice. PFNEG=3.276<PFUNR=3.419, which would support H2. In fact, PFCON=3.371<PFUNR=3.419, which suggests that the mitigating effect of an unrelated positive association is very strong and it more than offsets the negative effects of negative associations. [The significance of the six pair-wise comparisons of the PF's were determined using a one-sided Z test at the 5% level (greater than Z.05 = 1.645).]

Significant differences are indicated in bold in the table. Referring to the differences in the overall average PF’s, PFCON-PFNEG was not significant, which suggests that the negative effect of negative associations when presented alone is not significant. However, when the negative associations are combined with related positive associations, PFCON-PFREL becomes significant, which indicates that in the presence of a consumer’s knowledge of a negative association, a related positive association increases the magnitude of the negative effect. This is supportive of H1. The significant (a<.05) Z statistic of the difference PFNEG-PFREL also supports H1. A consumer is significantly (a<.05) less likely to choose an airline when a negative association is combined with a related positive association, than when the negative association alone is presented. A related positive association exacerbates the negative effect of negative associations. The significance (a<.05) of the difference PFREL-PFUNR supports a combinationof H1 and H2. A consumer is significantly less likely to choose a product when a combination of a negative and a related positive association is presented, than when a combination of a negative and an unrelated positive association is presented. H1 and H2 jointly hypothesize that an unrelated positive association is more effective than a related positive association in mitigating the negative effects of negative associations.

TABLE 2

OLS ESTIMATE OF PURCHASE FACTORS

Each of the three airlines is also considered individually (Table 4). American had one significant (a<.05) difference: PFAA:NEG-PFAA:UNR. This significant negative difference indicates that a consumer is more likely to choose AA when a negative association is combined with an unrelated positive association, than when the negative association is presented alone. This supports H2. An unrelated positive association mitigates the negative effect of negative associations. Southwest had three significant (a<.05) statistics, which were the same three pairs as in the overall PFxi. Valujet had one significant (a<.05) difference: PFVJ:REL-PFVJ:UNR.

3. Comparison of the Decoys’ PFD’s

In the third approach to measuring the effect of negative and positive associations on consumer choice, we looked at the average purchase factor of the five decoy airlines "PFD" by experimental conditions. Because the subjects were given the same set of choices that included the three target airlines and the five decoy airlines in all four experimental conditions, an increase (decrease) in a subject’s likelihood of choosing airline x could result in an increase (decrease) in PFxi, and/or a decrease (increase) in the PF’s of all other airlines. For example, a subject a in the negative group and another subject b in the control group may concur on their rating of PFx, i.e. PFxa=PFxb. However, subject a may indeed be less likely to choose airline x than subject b, if (1) a’s PFDa is greater than b’s PFDb, or (2) if the ratio PFxa / PFDa is less than PFxb / PFDb. We therefore make a comparison of the aggregate PFD, and of the ratio PFX/ PFD across experimental groups.

TABLE 3

AVERAGE PURCHASE FACTOR BY EXPERIMENTAL CONDITION

TABLE 4

Z STATISTICS FOR THE DIFFERENCES IN PURCHASE FACTORS AMONG EXPERIMENTAL CONDITIONS

TABLE 5

PURCHASE FACTORS OF DECOY

In comparing the PFD’s, shown in Table 5, the predicted changes across experimental groups are in the opposite direction of the predicted changes in PFxi. H1 predicts PFD:REL>PFDD:NEG, and H2 predicts PFD:UNR<PFD:NEG.

The significant differences were between CON and UNR (Z=2.363), and between REL and UNR (Z=2.371). The significant difference of PFD:REL-PFD:UNR supports H1 and H2. Subjects were more likely to choose the decoy airlines when the three target airlines combined negative publicity with related positive corporate sponsorships, than when they combined negative publicity with unrelated positive corporate sponsorships. This is consistent with the direct results shown earlier.

We also compared the ratio PFX / PFD across experimental conditions in Table 6. The predicted changes in this ratio are in the same direction as the predicted changes in PFxi. H1 predicts RATIOREL<RATIONEG, and H2 predicts RATIOUNR> RATIONEG.

RATIOREL=0.796<RATIONEG=0.866, which directionally supports H1. RATIOUNR=1.007>RATIONEG=0.866, which directionally supports H2.

LIMITATIONS OF THE STUDY, AND DIRECTION FOR FUTURE STUDIES

There are some limitations to this study. This study addresses the mitigating effects of a new positive secondary association on an existing negative ssociation. In the study design, both the positive and the negative associations were presented together. Therefore, the negative associations may not have been linked strongly enough with the airlines for retrieval to occur.

TABLE 6

RATIO OF PURCHASE FACTORS PFX/PFD

TABLE 7

A FULL FACTORIAL DESIGN AND THE CELLS CORRESPONDING TO THE CONDITIONS IN THE PRESENT EXPERIMENT

A limitation regarding the model is in the simplifying assumption of a linear relationship between PFxi and the experimental conditions. Equation (2) implies that the effect of the different experimental conditions is independent of the consumers’ purchase likelihood. In reality the n, r, and u parameters may depend on PF. For example, a consumer j who has a strong preference for one brand y over the other choices may look only for confirmatory evidence of his intended choice. Such a consumer would score high on PFyj, and his rating will be influenced more positively by a positive association, r and/or u, than it would be influenced negatively by negative associations, n. In this case, the magnitude of n will be a decreasing function of PFyj, and the magnitude of r or u will be an increasing function of PFyj. In another example, a consumer k who has a strong preference for brand q may have high expectations, and use his expected PFqk as a reference. Due to loss aversion, such a consumer may be adversely affected by negative publicity (which may decrease PFqk) more than he would be positively affected by positive associations (which may increase PFqk). In this case, the magnitude of n would be an increasing function of PFq, and r and/or u would be a decreasing function of PFqk. The interaction of the level of PF and n, r, and u was not considered in this study, and is one possible direction for future research on this topic.

Regarding the experimental design, a complete design would be a 2 x 3 factorial. The first factor would be the presence or absence of negative publicity about a company concerning a specific issue, such as environmental issues. The second would be three levels of positive associations that links the company to one of (1) a positive action in the same area as the negative publicity, (2) a positive action in an unrelated area, or (3) no positive action.

Our study used a between subject design. In order to increase the number of subjects in each experimental condition, we tested only four of the six conditions of a full factorial design. Of the possible six cells of a full design, the ones that were tested in the present study are indicated in Table 7.

Finally, our study uncovered some interesting results suggesting the possibility of both direct and halo effects of secondary associations on purchase likelihood. This is also a topic for future research.

In this paper, we looked at how secondary associations such as negative publicity, corporate sponsorships, and corporate policies affect consumer choice. As the name suggests, secondary associations are brand images that are not directly related to the attributes of the product, and they may not be important attributes in a consumer’s brand selection process. The effects on purchase choice were expected to be relatively small.

The results of this study suggest that a company should consider any negative associations that may exist, in choosing what philanthropic/civic causes to support and sponsor. Developing positive associations that are too closely related to an existing negative association may only serve to promote a social cause, and may actually be detrimental to consumers’ evaluation of the product. Developing unrelated positive secondary associations on the other hand, may be effective in mitigating existing negative associations.

REFERENCES

Aaker, David A. (1991), Managing Brand Equity. New York: Macmillan.

Bettman, James R. (1979), An Information Processing Theory of Consumer Choice, Reading, MA: Addison-Welsey.

Bettman, James R., and Mita Sujan (1987), "Effects of Framing on Evaluation of Comparable and Noncomparable Alternatives by Expert and Novice Consumers," Journal of Consumer Research, 14 (September) 141-154.

Burke, Raymond R., and Thomas K. Srull (1988), "Competitive Interference and Consumer Memory for Advertising," Journal of Consumer Research, 15 (June) 55-68.

Campbell, Margaret C. (1995), "When Attention-Getting Advertising Tactics Elicit Consumer Inferences of Manipulative Intent: The Importance of Balancing Benefits and Investments," Journal of Consumer Psychology, 4(3) 225-254.

Eliashberg, Joshua, and John Hauser (1985), "A Measurement Error Approach for Modeling Consumer Risk Preference," Management Science, 31, 1-35.

Farmer, Richard N., and W. Dickerson Hogue (1985), Corporate Social Responsibility, 2nd ed. Lexington, MA: Lexington Books.

Fishbein, Martin, and Icek Ajzen (1975), Belief, Attitude, Intention, and Behavior: An Introduction to Theory and Research. Readying, MA: Addison-Wesley Publishing Company.

Green, Paul E. (1984), "Hybrid Conjoint Analysis: An Expository Review," Journal of Marketing Research, 21 (May) 155-159.

Green, Paul E., Stephen M. Goldberg, and Mila Montemayor (1981), "A Hybrid Utility Estimation Model for Conjoint Analysis," Journal of Marketing, 45, 33-41.

Green, Paul E., and Vithala R. Rao (1971), "Conjoint Measurement for Quantifying Judgmental Data," Journal of Marketing Research, 8, 355-362.

Green, Paul E., and V. Srinivasan (1978), "Conjoint Analysis in Consumer Research: Issues and Outlook," Journal of Consumer Research, 5, 103-123.

Herr, Paul M. (1989), "Priming Price: Prior Knowledge and Context Effect," Journal of Consumer Research, 16, (June) 67-75.

Keller, Kevin L. (1993), "Conceptualizing, Measuring, and Managing Customer-Based Brand Equity," Journal of Marketing, 57 (January), 1-22.

Okada, Erica M., and David J. Reibstein (1996), "Quantifying the Effects of Positive/Negative Associations on Consumer Choice," unpublished working paper.

Russo, Jay E., and Barbara E. Dosher (1983), "Strategies for Multiattribute Binary Choice," Journal of Experimental Psychology: Learning, Memory, and Cognition, 9 (October), 676-696.

Tversky, Amos (1977), "Features of Similarity," Psychological Review, 84 (July), 327-352.

Tybout, Alice M., Bobby J. Calder, and Brian Sternthal (1981), "Using Information Processing Theory to Design Marketing Strategies," Journal of Marketing Research, XVIII (February), 73-79.

Wilkie, William L., John M. McCann, and David J. Reibstein (1973), "Halo Effects in Brand Belief Measurement: Implications of Attitude Model Development," Proceedings, Fourth Annual Conference, Association for Consumer Research, (November).

Yi, Youjae (1990), "The Effects of Contextual Priming in Print Advertisements," Journal of Consumer Research, 17 (September) 215-222.

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