How Do Consumer Inferences Moderate the Effectiveness of Two-Sided Messages

Cornelia (Connie) Pechmann, University of California, Irvine
ABSTRACT - This paper attempts to identify the key dimensions along which two-sided nonrefutational messages differ, and then to review empirical findings regarding how each of these dimensions moderates the effectiveness of two-sided messages. The thesis of this paper is that messages which differ on certain key dimensions elicit different consumer inferences or attributions. Consequently, these messages differ in terms of how credible they are perceived to be, and in terms of how much they enhance brand attitudes, beliefs, and purchase intentions relative to a one-sided messages. Hence, by distinguishing among different types of two-sided messages, at least some of the apparent discrepancies in the literature can be resolved.
[ to cite ]:
Cornelia (Connie) Pechmann (1990) ,"How Do Consumer Inferences Moderate the Effectiveness of Two-Sided Messages", in NA - Advances in Consumer Research Volume 17, eds. Marvin E. Goldberg, Gerald Gorn, and Richard W. Pollay, Provo, UT : Association for Consumer Research, Pages: 337-341.

Advances in Consumer Research Volume 17, 1990    Pages 337-341

HOW DO CONSUMER INFERENCES MODERATE THE EFFECTIVENESS OF TWO-SIDED MESSAGES

Cornelia (Connie) Pechmann, University of California, Irvine

ABSTRACT -

This paper attempts to identify the key dimensions along which two-sided nonrefutational messages differ, and then to review empirical findings regarding how each of these dimensions moderates the effectiveness of two-sided messages. The thesis of this paper is that messages which differ on certain key dimensions elicit different consumer inferences or attributions. Consequently, these messages differ in terms of how credible they are perceived to be, and in terms of how much they enhance brand attitudes, beliefs, and purchase intentions relative to a one-sided messages. Hence, by distinguishing among different types of two-sided messages, at least some of the apparent discrepancies in the literature can be resolved.

This paper provides an organizing framework for summarizing the results of previous studies on nonrefutational two-sided messages, i.e., messages which describe potentially undesirable attributes or shortcomings of the promoted brands and yet make no attempt to refute this potentially negative information (Kamins and Assael 1987). The results of such studies seemingly are contradictory. Some studies have found that two-sided messages are more effective than one-sided messages (Goodwin and Etgar 1982; Smith and Hunt 1978), whereas other studies have found that two-sided messages are no more effective (Belch 1981; Settle and Golden 1974) or even less effective (Stayman, Hoyer and Leon 1987; Swinyard 1981) than one-sided messages.

This paper will attempt to show that two-sided messages differ on a number of key dimensions, and that messages which differ on these dimensions vary in their effectiveness because they elicit different types of consumer inferences or attributions. Therefore, by distinguishing among different types of two-sided messages, at least some of the apparent discrepancies in the literature can be resolved. Hence, this paper hopes to accomplish the following two objectives: (1) to clarify what we already know about two-sided messages, and (2) to clarify what we do not yet know, so as to stimulate and guide future research. The next section of this paper will discuss the key dimensions of two-sided messages.

THE KEY DIMENSIONS OF TWO-SIDED MESSAGES

Based on a careful review of the literature, it appears that two-sided messages differ on the following four key dimensions:

1. How important (he potentially unfavorable disclosure is perceived to be (Stayman, Hoyer and Leon 1987). A brand can be described as having one or more undesirable attributes or shortcomings that are either relatively unimportant or relatively important, either as perceived by all potential consumers or by a particular subgroup of consumers.

2. The motivation underlying the potentially unfavorable disclosure. For example, consumers might assume that a marketer is required by federal and/or state regulations to reveal its brand's potentially undesirable attributes or shortcomings. On the other hand, consumers might assume that a marketer is revealing this information voluntarily,

3. How counter-attitudinal or pro-attitudinal the message is overall. -For example, a message might attempt to persuade consumers that a very low s*are brand is superior to a high share brand, in which case it would be counter-attitudinal to most consumers. Alternatively, a message might attempt to reinforce existing favorable attitudes towards a very high share brand, in which case it would be pro-attitudinal to most consumers.

4. How highly correlated the potentially undesirable attribute is to another potentially desirable brand attribute featured in the message. For example, the fact that a brand has more of a potentially undesirable attribute might cause consumers to infer that it has more- of a desirable attribute as well. In contrast, a brand's rating on the potentially undesirable attribute might be perceived to be unrelated to its rating on another, more desirable attribute.

The next section of this paper will discuss empirical findings regarding how these four dimensions moderate the effectiveness of two-sided messages.

A REVIEW OF FINDINGS REGARDING TWO-SIDED MESSAGES

Based on a careful scrutiny of empirical findings on two-sided messages, the following propositions have been formulated:

Proposition 1. The effectiveness of a two-sided message is moderated by the perceived importance of the potentially unfavorable disclosure. If a two-sided message discloses that the promoted brand has a potentially undesirable attribute or shortcoming that is perceived to be important by consumers, the message and message source are perceived to be more truthful or credible (Settle and Golden 1974; Smith and Hunt 1978) for the following reason. Marketers who voluntarily disclose unfavorable information about their brands do not appear to be motivated by external pressures to sell their brands. Hence, consumers apparently infer that there is a relatively high degree of correspondence between the statements made in these two-sided messages and the brands' actual merits (Smith and Hunt 1978; Stayman, Hoyer and Leon 1987; also see Correspondence Theory e.g. Jones and Davis 1965; and Attribution Theory e.g. Kelley 1973). In contrast, marketers who use one-sided messages appear to be motivated by external pressures to sell their brands. Hence, consumers apparently infer that there is less of a correspondence between the statements made in these one-sided messages and the brands' actual merits.

Nevertheless, a two-sided message which discloses that a brand has important shortcomings does not appear to be any more effective than a one-sided message at promoting favorable brand beliefs, attitudes, or purchase intentions (Belch 1981; Settle and Golden 1974; Stayman, Hoyer and Leon 1987). Rather, it appears that the benefits gained from using this type of two-sided message (namely, the enhanced credibility) are either equal to or less than the costs incurred (namely, that unfavorable information is promulgated about the promoted brand). For example, Stayman et al. (1987) determined that a two-sided message stating that an alarm clock scored poorly on styling--a moderately important attribute--was perceived to be significantly more trustworthy than a one-sided message. However, consumers exposed to this two-sided message were no more favorably disposed towards purchasing the alarm clock than consumers exposed to the one-sided message. Apparently, the gain in credibility was offset by the costs incurred. Correspondingly, Stayman et al. determined that a two-sided message stating that an alarm clock scored poorly on accuracy--a very important attribute--was perceived to be significantly more trustworthy than a one-sided message. However, consumers exposed to this two-sided message were less favorably disposed towards purchasing the alarm clock than consumers exposed to the one-sided message. Apparently, the costs incurred were greater than the benefits gained.

There is as yet no evidence that it is possible to reduce the costs of using a two-sided message (i.e., by reducing the importance of the disclaimer made) without simultaneously reducing the benefits gained (i.e., without simultaneously reducing the credibility of the overall message). For example, Stayman et al. (1987) determined that a two-sided message stating that a record store did not sell concert tickets--an unimportant attribute--was not perceived to be any more trustworthy than a one-sided message. As a result, consumers exposed to this two-sided message were no more favorably disposed towards shopping at the record store than consumers exposed to the one-sided message.

Nevertheless, it might be advantageous to make a disclosure which is perceived by the target consumers to be important to some consumers and yet not to them. Such a two-sided ad might be perceived as more credible, and yet not tarnish the brand's image in the eyes of the target consumers for whom the disclosure is relatively unimportant. Hence, it might be fruitful for researchers to explore this issue in the future.

Proposition 2. The effectiveness of a two-sided message is moderated by the perceived motivation underlying the potentially unfavorable disclosure. A marketer's perceived motivation for using a two-sided message has never been explicitly identified as being an important moderating variable, but it would also would appear to moderate the effectiveness of two-sided messages. If consumers perceive that a marketer is required to make a potentially unfavorable disclosure, or has some other ulterior motive for doing so, they are not likely to infer that the two-sided message is any more credible than a one-sided message. For example, since most consumers realize that the health warnings on cigarette ads are required, they do not infer that cigarette advertisers are more trustworthy or truthful than other advertisers. Hence, required disclosures of any type--to the extent that they are important--almost undoubtedly are detrimental to brand attitudes and sales, which is why such disclosures are vehemently opposed by most marketers.

Proposition 3. The effectiveness of a two-sided message is moderated by how counter-attitudinal or pro-attitudinal the message is overall. This particular proposition comes from the comparative advertising literature (e.g., Swinyard 1981). Comparative ads typically claim that very low share brands are superior to very high share brands. Since such comparative ads are highly counter-attitudinal (particularly to users of the very high share brands), they are perceived to be less credible and they elicit more counter-arguing than noncomparative ads (e.g., Belch 1981). Hence, it has been suggested that comparative ads be worded as two-sided messages in order to enhance their perceived credibility and to reduce counter-arguing.

Based on this same logic, it would appear that the more counter-attitudinal a message, the more likely it is to benefit from a two-sided format. However, as stated previously, there is as yet no evidence that a two-sided message can enhance the credibility of a message without simultaneously reducing its overall persuasiveness because of the negative disclosures that are made. Hence, it is unclear whether counter-attitudinal messages (including comparative ads) can benefit from being two-sided. Once again, further research is sorely needed.

Proposition 4. The effectiveness of a two-sided message is moderated by how highly correlated the potentially undesirable attribute is to another potentially desirable attribute featured in the message. The correlation between the attributes featured in two-sided messages never before has been identified as being important in moderating the effectiveness of two-sided messages. However, related empirical findings on consumer inferences suggest that the relative effectiveness of a two-sided message will be contingent on the inferences consumers make when a potentially undesirable attribute is highly correlated with another desirable attribute featured in the message (e.g., Ford and Smith 1987). Consumers can make at least two types of relevant inferences, each of which will be discussed below.

First, consider a situation in which the promoted brand's rating on a desirable attribute featured in a two-sided message is ambiguous. When the brand is described in a two-sided message as having an unfavorable rating on an attribute that is highly correlated with this desirable attribute, consumers may infer that the brand has a more favorable rating on this desirable attribute. Hence, a two-sided message may promote more favorable brand beliefs, attitudes, and purchase intentions than a one-sided message. However, the two-sided message probably will not be perceived as any more trustworthy or credible than the one-sided message, particularly if consumers realize that the marketer had an ulterior motive for disclosing the potentially unfavorable information. Consumers are quite accustomed to marketers discounting their brands on relatively unimportant attributes so as to more effectively position their brands as being superior on more important, highly correlated attributes.

For example, most consumers have been exposed to two-sided ads acknowledging that Loreal hair color is more expensive than other competing brands. Furthermore, consumers probably realize that the marketer of Loreal has an ulterior motive for acknowledging that Loreal is more expensive--to more effectively position it as being a superior quality hair color. Hence, consumers probably do not infer that the two-sided Loreal ad is any more trustworthy or credible than the one-sided ads for competing brands of hair color.

Prior studies of two-sided messages often used messages in which the discounted attributes were highly correlated with the other featured attributes, although probably only inadvertently. For example, Etgar and Goodwin (1982) copy-tested a two-sided ad stating that the advertised brand of beer was a premium beer with a full bodied taste, but was more expensive and higher in calories. The one-sided ad portrayed the advertised brand favorably on all four dimensions. It was found that the two-sided ad was not perceived to be any more believable than the one-sided ad. However, the two-sided ad was more effective at persuading respondents that the advertised beer was of higher quality, and at enhancing purchase intentions of this beer. It appears that by describing the advertised beer as being inferior on two dimensions (that is, as being higher in calories and more expensive), the two-sided ad more effectively positioned it as being superior on other, highly correlated dimensions (that is, as being a premium beer with a full bodied taste). Since respondents in the study preferred a premium beer with a full bodied taste over a less expensive or lower calorie beer (according to pretest data), the two-sided ad thereby promoted more favorable brand beliefs and purchase intentions.

Similarly, Swinyard (1981) copy-tested a two-sided ad stating that the advertised grocery store charged lower prices but had fewer amenities, e.g., consumers had to bag their own groceries. The one-sided ad simply stated that the advertised grocery store charged lower prices. It was found that the two-sided ad was more effective at persuading respondents that the advertised store actually charged lower prices. It appears that by describing the advertised store as being inferior on one dimension (that is, as having fewer amenities), the two-sided ad more effectively positioned it as being superior on another, highly correlated dimension (that is, as having low prices). However, it appears that in the aggregate, the respondents in the study preferred a store with amenities and somewhat higher prices over a store without amenities albeit lower prices. Hence, the two-sided ad actually was less effective than the one-sided ad at persuading respondents in the aggregate to shop at the advertised store and redeem the store coupon they had been given. However, it is possible that the two-sided ad was more effective than the one-sided ad at persuading respondents in the store's target market (i.e. respondents who considered low prices to be more important than amenities) to shop at the advertised store. Unfortunately, the data only were analyzed at the aggregate level.

Many studies also have been conducted to determine what inferences consumers make when information about a brand is missing, and in several of these studies the missing attributes were highly correlated with the known attributes. Furthermore, these "missing information" studies are related to studies of two-sided messages, although they generally have not been perceived as such. The reason why these "missing information" studies are related is that one-sided messages essentially are messages in which the (unfavorable) information provided in two-sided messages is "missing".

In one such "missing information" study, Meyer (1981) exposed respondents to a two-sided message stating that a pizza parlor charged low prices but had a plain decor. The one-sided message simply stated that the pizza parlor charged lower prices. Meyer essentially replicated Swinyard's key result: it was found that the two-sided message was more effective than the one-sided message at persuading respondents that the pizza parlor actually charged low prices. Furthermore, Meyer replicated these results using a non-price attribute. He also exposed respondents to a two-sided message stating that a pizza parlor had excellent (tasting) pizza but a plain decor. The one-sided message simply stated that the pizza parlor had excellent pizza. It was found that the two-sided message was more effective than the one-sided message at persuading respondents that the pizza parlor actually had excellent pizza.

Since the two-sided messages promoted more favorable brand attitudes than the one-sided messages, Meyer concluded that consumers replace missing information with discounted average values. However, an alternative or rival explanation for why the two-sided messages promoted more favorable brand attitudes is that respondents inferred that a pizza parlor with a plain decor was more likely to charge lower prices and to have excellent pizza, i.e., respondents perceived these attributes to be highly correlated. Hence, research should be conducted to examine the validity of this rival explanation.

There is yet another reason why the correlation between the attributes featured in two-sided messages might moderate the effectiveness of two-sided messages. Consider a second type of situation in which the promoted brand's rating on a desirable attribute featured in a two-sided message is known with some degree of certainty. When the brand is described as having a favorable rating on this desirable attribute, if this desirable attribute is highly correlated with another attribute, consumers may infer that the brand has a less favorable rating on the other attribute if exposed to a one-sided message than if exposed to a two-sided message. Hence, once again, a two-sided message may promote more favorable brand beliefs, attitudes, and purchase intentions than a one-sided message. However, the two-sided message probably will not be perceived as any more trustworthy or credible than the one-sided message, particularly if consumers realize that the marketer had an ulterior motive for disclosing the potentially unfavorable information. Consumers are quite accustomed to marketers reassuring them that although their brands may have shortcomings, these shortcomings are lesser in magnitude than might be expected.

For example, most consumers have been exposed to two-sided ads for certain brands of luxury cars acknowledging that such cars are quite expensive, albeit less expensive than might be expected. Furthermore, consumers probably realize that the marketers of these cars have an ulterior motive for acknowledging how expensive their cars are--to insure that consumers will not infer that these luxury cars are even more expensive than they actually are. Hence, consumers probably do not infer that these two-sided car ads are any more trustworthy or credible than one-sided car ads.

As stated earlier, many studies have been conducted to determine what inferences consumers make when information about a brand is missing, and in many of these studies the missing attributes were highly correlated with the known attributes. Furthermore, these "missing information" studies are related to studies of two-sided messages, since one-sided messages essentially are messages in which the (unfavorable) information provided in two-sided messages is "missing".

In one such "missing information" study, Huber and McCann (1982) exposed respondents to a two-sided message stating that a beer was low in price but only of average quality. The one-sided message simply stated that the beer was low in price. It was found that the two-sided message was more effective at enhancing purchase intentions for the beer than the one-sided message. A post-hoc interpretation of Huber and McCann's results is that the two-sided message was more effective than the one-sided message for following reason. Since the one-sided message described the beer as having a favorable rating on price (e.g., as being low in price), and since price generally is perceived to be highly correlated with quality, respondents expose to the one-sided message may have inferred that the beer had an unfavorable rating on the missing attribute (e.g., that it was low in quality). Hence, respondents exposed to the two-sided message actually may have been reassured by it. Even though the two-sided message acknowledged that the beer was only of average quality, at least it reassure respondents that the beer was not of even lesser quality. If so, this may explain why respondents exposed to the two-sided message evaluated the beer more favorably than respondents exposed to the one-sided message. Hence, research also should be conducted to examine the validity of this post hoc explanation of Huber and McCann's results.

It should be noted that Huber and McCann's results appear to be quite robust. For example, a study by Johnson and Levin (1985) essentially replicated Huber and McCann's (1982) results using television sets rather than beer. In addition, a follow-up study by Lim, Olshavksy and Kim (1988) was able to replicate Huber and McCann's results using non-price attributes of cars, although only when respondents' inferences were prompted. Apparently, it was necessary for Lim et al. to prompt respondents to make inferences since the correlation between the attributes Lim et al. used (i.e., comfort and gas mileage) was lower and/or 1< salient than the correlations between the attributes used in the earlier two studies (i.e., price and quality).

IMPLICATIONS OF THIS REVIEW

Empirical and theoretical work on two-sided messages has focused almost exclusively on one key dimension along which two-sided messages differ: the relative importance of the shortcomings or unfavorable attributes that are disclosed. Based on the review presented in this paper, this variable does appear to moderate the effectiveness of two-sided messages. The more important the shortcoming that is disclosed, the more trustworthy or credible the message and message's source are perceived to be-assuming that the disclosure is perceived to be made voluntarily rather than solely to comply with regulatory statutes or for some other ulterior motive. However, the more important the shortcoming that is disclosed, the less effective the message is at enhancing favorable brand attitudes, beliefs, and purchase intentions precisely because increasingly unfavorable information is promulgated about the brand.

What this suggests is that two-sided messages are of limited value to marketers, even for enhancing the credibility of highly counter-attitudinal messages (such a comparative ads claiming that very low share brands are superior to leading brands). One possible way to use a two-sided message effectively is to disclose a shortcoming that consumers will perceive to be important to some consumers and yet not important to them personally, and to insure that consumers will perceive that this disclosure is being made voluntarily and with no ulterior motive. However, the relative effectiveness of this type of two-sided message must still be empirically studied.

Furthermore, much more attention probably should be focused on yet another key dimension on which two-sided messages differ: the extent to which the unfavorable and favorable attributes are perceived to be correlated. Based on the review presented in this paper, this variable also appears to moderate the effectiveness of two-sided messages. In fact, based on a post-hoc analysis of prior studies pertaining to two-sided messages, it appears that the only types of two-sided messages that appear to be effective at enhancing brand beliefs, attitudes, and purchase intentions are those in which the potentially undesirable attributes are highly correlated with other desirable brand attributes featured in these messages.

Not surprisingly, these are the only types of two-sided messages that advertisers appear to be using at the present time. One two-sided message of this type that has recently been used in Southern California is an ad for a fast food restaurant chain called Carl's Junior. The ad acknowledges that Carl's Junior is not as fast as other fast food restaurants in order to persuade consumers that its food is fresher (because it is prepared when ordered rather than in advance). Yet another recent example is the ad for Frusen Gladje ice cream which acknowledges that Frusen Gladje is fattening in order to persuade consumers that it is very rich and delicious. Other recent examples of two-sided messages are the ads for luxury cars which acknowledge the cars' high prices in order to preclude consumers from inferring that the cars are even more expensive than they really are. Given that this particular type of two-sided message appears to be the most prevalent, it is somewhat surprising that this type of two-sided message never has been explicitly studied.

Unfortunately, the conclusions made in this paper regarding the relative effectiveness of two-sided messages with correlated attributes are tentative at best. These conclusions were based primarily on post hoc analyses of empirical studies of two-sided messages as well as somewhat related "missing information" studies. Hence, these tentative conclusions must be verified by conducting additional research. It is hoped that this review will stimulate additional research on the relative effectiveness of two-sided messages, particularly those in which the desirable and undesirable attributes featured in the messages are highly correlated.

REFERENCES

Belch, George (1981), "An Examination of Comparative and NonComparative Television Commercials: The Effect of Claim Variation and Repetition on Cognitive Response and Message Acceptance", Journal of Marketing Research, 18 (August), 333-349.

Etgar, Michael and Steven M. Goodwin (1982), "One-Sided Versus Two-Sided Comparative Message Appeals for New Brand Introductions", Journal of Consumer Research, 8 (March), 460465.

Ford, Gary T. and Ruth Ann Smith (1987), "Inferential Beliefs in Consumer Evaluations: An Assessment of Alternative Processing Strategies", Journal of Consumer Research, 14 (December), 363-371.

Huber, Joel and John McCann (1982), "The Impact of Inferential Beliefs on Product Evaluations," Journal of Marketing Research, 19 (August), 324333.

Johnson, Richard D. and Irwin P. Levin (1985), "More than Meets the Eye: The Effect of Missing Information on Purchase Evaluations," Journal of Consumer Research, 12 (September), 169-177.

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

Kamins, Michael A. and Henry Assael (1987), "Two-Sided Versus One-Sided Appeals: A Cognitive Perspective on Argumentation, Source Derogation, and the Effect of Disconfirming Trial on Belief Change," Journal of Marketing Research, 24 (February), 29-39.

Kelley, Harold E. (1973), "The Process of Causal Attribution", American Psychologist, 107-128.

Lim, Jeen-Su, Richard W. Olshavsky, and John Kim (1988), "The Impact of Inferences on Product Evaluations: Replication and Extension," Journal of Marketing Research, 15 (August), 308316.

Meyer, Robert J. (1981), "A Model of Multiattribute Judgments Under Attribute Uncertainty and Informational Constraint", Journal of Marketing Research, 18 (November), 428441.

Settle, Robert B. and Linda L. Golden (1974), "Attribution Theory and Advertiser Credibility", Journal of Marketing Research, 11, 181-185.

Smith, Robert E. and Shelby D. Hunt (1978), "Attributional Processes and Effects in Promotional Situations", Journal of Consumer Research, 5 (December), 149-158.

Stayman, Douglas, Wayne Hoyer and Robert Leon (1987), "Attribute Importance in Discounting Product Features in Advertising", paper presented at the American Marketing Association Summer Educator's Conference in Toronto, Canada.

Swinyard, William R. (1981), "The Interaction Between Comparative Advertising and Copy Claim Variation", Journal of Marketing Research, 18 (May), 175-186.

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