comparative versus noncomparative advertising: a meta-analysis

Dhruv Grewal, Sukumar Kavanoor, Edward F. Fern,Carolyn Costley, & James Barnes

Comparative Versus NoncomparativeAdvertising: A Meta-Analysis

Previous research and reviews on comparative advertising report mixed results. The authors report the results froma meta-analysis that examines the efficacy of comparative advertising. The analysis shows that comparative adsare more effective than noncomparative ads in generating attention, message and brand awareness, levels of mes-sage processing, favorable sponsored brand attitudes, and increased purchase intentions and purchase behaviors.However, comparative ads evoke lower source believability and a less favorable attitude toward the ad. Additionalanalyses of moderator variables find that market position (sponsor, comparison, and relative), enhanced credibil-ity, message content, and type of dependent measure (relative versus nonrelative) affect some of the relationshipsbetween advertising format and cognition, brand attitudes, and purchase intentions. New brands comparing them-selves to established brands appear to benefit most from comparative advertising.

Comparative advertising has become increasinglyprevalent in the United States media. Some examplesof the many advertising campaigns using compara-

tive advertising include MCI comparing its long-distancecall prices and service to AT&T; VISA comparing the num-ber of merchants accepting its credit card to AmericanExpress; Burger King comparing its cooking method toMcDonald's; and various beverage producers challengingcompetitors in taste tests. Recent estimates indicate thatcomparative advertising formats account for one-third of alladvertisements (Neiman 1987; Stewart and Furse 1986) andclose to 80% of all television commercials (Pechmann andStewart 1990b).

Comparative advertising's increased popularity may bepartly due to the Federal Trade Commission's informalencouragement of explicit comparisons (Tannenbaum 1974;Wilkie and Farris 1975), as well as to advertisers' beliefs inits effectiveness. The FTC rationalized that explicit compar-ative advertisements deliver infonnation previously unavail-able to consumers (Wilkie and Farris 1975). The FTC'simplicit encouragement of brand comparisons, along withrelaxed restrictions and competitor and media concerns(Tannenbaum 1974), sparked the research interest of acade-micians and practitioners alike.

The effectiveness of comparative advertising, accordingto the large body of extant empirical research, is equivocal.Some investigators conclude that comparative advertising

Dhruv Grewal is Associate Professor of Marketing, University of Miami.Sukumar Kavanoor is a marketing consuitant, The Janus Group. EdwardFern Is Associate Professor of Marketing, Virginia Polytechnic Instituteand State University. Carolyn Costley Is a lecturer. University of Waikato.James Barnes is Professor of Marketing, University of Mississippi. Thisarticle has benefited from the comments and suggestions of Connie Pech-mann, Paul Miniard, Diana Grewal, and the four anonymous JM review-ers. The authors gratefully acknowledge the financial support from a Uni-versity of Miami Summer Research Grant.

provides advantages that are not associated with noncom-parative advertising (e.g., Droge and Darmon 1987; Miniardet al. 1993; Pechmann and Ratneshwar 1991; Pechmann andStewart 1990a; Rose et al. 1993). Others report that com-parative advertising produces undesirable outcomes (e.g..Belch 1981; Golden 1979; Goodwin and Etgar 1980; Levine1976; Swinyard 1981). These conflicting opinions do notseem to deter major consumer goods and service corpora-tions from using comparative advertising in their promo-tional mix.

The prevalence of comparative ads in the face of theconflicting empirical evidence about their effectiveness sug-gests that this topic should be important to researchers andpractitioners. There have been several attempts to consoli-date research findings in this area (e.g., Barry 1993; Etgarand Goodwin 1977; Pechmann and Stewart 1990b; Turgeonand Bamaby 1988). Notwithstanding the importance ofthese earlier contributions, these qualitative reviews typi-cally include only published research (cf. Turgeon and Bam-aby 1988)' To date, no one has attempted a quantitativeintegration (i.e., a meta-analysis) of this research stream. Tofill this void, we conduct a systematic review and a quanti-tative integration of all the available comparative advertis-ing research.

We systematically review the comparative advertisingliterature with two goals in mind: (1) to consolidate to theextent that is possible the comparative advertisingresearch—published and unpublished—and (2) to discover

'We have two concerns about these previous reviews. First,there is the ri.sk of bias from ignoring unpublished research whensynthesizing results across studies (Hunter and Schmidt 1990;Rosenthal 1984). Second, all of these articles use the conventionalsubjective review process. Pechmann and Stewart's (1990b) "qual-itative meta-analysis" differs somewhat from the others. Theyreport the amount of evidence available for or against a proposi-tion, based on the number of studies that support or fail to supportthe proposition.

Journal of MarketingVol. 61 (October 1997), 1-15 Comparative Versus Noncomparative /1

when comparative formats are more effective than noncom-parative formats. Constructing a general theory of compara-tive advertising is beyond the scope of our research effort.We organize the review and meta-analytical tests ofhypotheses around Lavidge and Steiner's (1961) advertisingfunctions model. Further details pertaining to the conceptualunderpinnings of the hypotheses tested in our study can befound in the original articles.

In Figure 1, we depict our Hierarchy of ComparativeAdvertising Effects Model, which is based on Lavidge andSteiner's (1961) model. This hierarchical model is used toclassify the literature's many dependent variables into mean-ingful effect categories. Using these categories we systemat-ically compare the effectiveness of comparative advertisingwith noncomparative advertising. Where possible, we testWilkie and Farris's (1975) propositions for comparisonadvertising effects.

First, we define comparative advertising and explain ourguiding framework. Second, we describe our literature-com-pilation procedures and meta-analysis methods. Third, wediscuss our findings and their implications for advertisingpractice.

A Conceptual Framework and theResearch Hypotheses

Previous research defines comparative ads (also known ascomparison ads) using two criteria. First, comparative adsexplicitly (e.g., Wilkie and Farris 1975) or implicitly (e.g..

Jackson, Brown, and Harmon 1979) compare at least twobrands in the same generic product or service class. Second,comparative ads compare the brands on specific product/ser-vice attributes (Wilkie and Farris 1975) or market positions(McDougall 1976). Thus, brands claiming to be better thantheir competitors without saying how are not using a com-parative format. Many scholars have compared the effec-tiveness of comparative ads to that of noncomparative adformats. These studies typically examine effectivenessaccording to one or more of the advertising functions inLavidge and Steiner's (1961) model.

Lavidge and Steiner (1961) separate the objectives ofadvertising into three main functions: cognitive, affective,and conative. Advertising's cognitive function providesinfonnation and facts for the purpose of making consumersaware and knowledgeable about the sponsored brand.Advertising's affective function creates liking and prefer-ence for the sponsored brand—preference presumably refersto more favorable attitudes. Advertising's affective function,therefore, is to persuade. Finally, advertising's conativefunction is to stimulate desire and cause consumers to buythe sponsored brand.

Lavidge and Steiner also specify which ad formats arebest suited for each advertising function. Based on theirguidance, we expect comparative ads to create greater spon-sored brand awareness and knowledge, stronger preferencesfor and attitudes toward the sponsored brand, higher levels ofintention to purchase the sponsored brand, and greater pur-chases of the sponsored brand than noncomparative ads do.

FIGURE 1Hierarcby of Comparative Advertising Effects

COGNITIVE

Comparative AdversusNoncomparative Ad

Attention (+)Awareness (+)

* message recall* brand recall

Processing (+)Informativeness (+)Similarity (+)Believability (-)

* message* source

y AFFECTIVE

Market Position—Sponsor, Comparison,& Relataive

Relative versus Nonrelative MeasuresSource Credibility EnhancersMessage Content

y CONATIVE i

Intentions (+)

Behavior (+)

Note: + sign if comparative is greater than noncomparative, and - sign if comparative is less than noncomparative.hypothesized. = paths not tested.

... = paths tested but not

2 / Journal of Marketing, October 1997

The Effect of Comparative Advertising on CognitionAdvertising's cognitive function moves consumers fromunawareness to knowledge about the sponsored brand byproviding information and facts about the sponsored brand(Lavidge and Steiner 1961). Because comparison formatsdeliver different infonnation to consumers than do noncom-parative formats, they are likely to produce different effectson consumer awareness and knowledge. It is not surprisingthat comparative advertising researchers have devoted muchattention to cognitive variables. Dependent measures ofcognitive outcomes include attention, message and brandawareness, number of thoughts, perceived informativeness,perceived similarity between competing brands, and believ-ability of the source and message. If comparative advenis-ing distinctly affects consumers' cognitive activities, itshould be evident in these variables.

Attention. Attention is the simple allocation of cogni-tive capacity to the advertisement (Engel, Blackwell, andMiniard 1993). One reason more of an audience shouldattend to comparative ads than to traditional ads (Wilkieand Fanis 1975) is that two brands are named in compara-tive ads. A comparative ad is personally relevant to users ofthe sponsored brand because it gives information about abrand they may use. Also, the comparative ad should bepersonally relevant to users of the comparison brandbecause it gives them information about a brand they douse. Consequently, the use of two (or more) brands in com-parative ads attracts the self-relevant selective attention ofmore people than advertisements featuring a single brand.Pechmann and Stewart (1991) predicted and found that alow-share sponsor's advertisement gets the most attentionwhen it is compared to a competitor with many users (ahigh-share brand).

A second reason that comparative ads might receivemore attention is information content (Muehling, Stoltman,and Grossbart 1990). Users of the sponsor brand—mostlikely the lesser-known of the two—may read comparisonads because they expect new and unique kinds of informa-tion. Specific attribute infonnation presented in the compar-isons may not be accessible otherwise. Noncomparative adsare unlikely to give such explicit information. The potentialfor increasing consumers' information was one reason theFTC reputedly did not condemn comparison ads (Wilkie andFarris 1975). If consumers perceive that specific attributecomparisons provide unique and useful information, theymay attend to comparative ads more. The studies wereviewed operationalized attention as reading part of theadvertisement (e.g., Muehling, Stoltman, and Grossbart1990; Pechmann and Stewart 1990a).

H|: Consumers will attend to comparative ads more than non-comparative ads.

Awareness. Awareness is knowing the information pre-sented in the message (i.e., message awareness) and know-ing the sponsor's brand name (i.e., brand name awareness).Consumers are considered aware of a sponsor's brand ifthey can recall the brand name. Similarly, they are aware ofthe advertised message if they can report any of it. Wilkieand Fanis (1975) believe that aggregate recall levels should

be higher for comparative ads, because more people attendto them than to traditional ads.

Attention is a necessary, but not sufficient, condition formaking consumers aware of the message or the sponsoredbrand. Wilkie and Farris (1975) speculate that comparisonads attract the consumer's attention and increase awarenessof the comparison brand, but do not increase awareness ofthe sponsor brand. Nonetheless, comparative ads mayenhance consumers' brand and message awareness by pro-viding more information than do noncomparative ads. Theadditional information may cause deeper processing andmore retrieval cues. This can result in improved memory ofthe advertising message, which is favorable to the sponsoredbrand. Therefore, message recall should be greater for com-parative than for noncomparative ads (Muehling, Stoltman,and Grossbart 1990). The additional information in compar-ative ads is expected to increase the accessibility andretrieval of the sponsored brand's name from memory.

H2a: Consumers will recall more of the messages in compara-tive ads than in noncomparative ads.

H2b: Consumers will recall the sponsor's brand name more incomparative ads than in noncomparative ads.

Information processing. Comparative ads are moreinvolving for the same reasons that they attract attention(i.e., they are personally relevant). Greater involvement withcomparison ads than with noncomparative ads should causethoughts to be more elaborate (Greenwald and Leavitt 1985;Krugman 1965). As was mentioned previously, comparisonads, which make point-by-point contrasts between the com-parison brand and the sponsored brand (e.g., Swinyard1981), should induce greater mental activity and elaboration(Muehling, Stoltman, and Grossbart 1990) and more centralprocessing (Droge 1989) than noncomparative ads do. Moreelaborate message processing should generate morethoughts in response to the comparative ads than in responseto the noncomparative ads. Another reason that comparativeformats foster more elaborate processing is that they aremore stimulating (Droge 1989; Pechmann and Esteban1994). Stimulation motivates consumers to argue either insupport of or against the main appeal in the message. Thedirection depends on their previous experience and familiar-ity with the comparison brand. Both involvement and stim-ulation should work to increase the number of thoughts con-sumers generate in reaction to comparative ads. In noncom-parative ads this same degree of mental activity is not likelybecause the same amount and type of information is notavailable for processing (Muehling, Stoltman, and Grossbart1990). Information processing is typically measured by therelative number of message-related consumer thoughts thatan advertisement generates.

H3: Comparative ads elicit more consumer thoughts than non-comparative ads.

Informativeness. Perceived informativeness is not theamount of information, per se, but the degree to which con-sumers perceive or infer the advertisement to be informa-tive. Comparative ads are perceived to be more informativebecause they present and compare two brands on one ormore attributes (Goodwin and Etgar 1980; Iyer 1988; Pride,Lamb, and Pletcher 1979; Sujan and Dekleva 1987; Wilkie

Comparative Versus Noncomparative / 3

and Farris 1975). Consumers may perceive comparativeinformation to be uniquely informative because the explicitattribute comparisons help them differentiate brands. Thiskind of information is not available in noncomparative ads.Wilkie and Farris (1975) use the same argument to justifythe value of comparative ads.

H4: Consumers will perceive comparative ads to be moreinformative than noncomparative ads.

Positioning. Positioning the sponsored brand as similarto the comparison brand (i.e., association) is the explicitgoal of some comparative ads and the unintentional effect ofothers (Droge and Darmon 1987; Gorn and Weinberg 1984;Pechmann and Ratneshwar 1991; Sujan and Dekleva 1987).Because comparative ads compare two brands on one ormore attributes, consumers may link them together in thesame cognitive categories, as well as infer similarities onother attributes (Pechmann and Ratneshwar 1991). Weexpect association effects and higher brand similarity rat-ings when advertisements say that the brands are similar(i.e., associated). However, when a comparison brand is notmentioned, as in noncomparative ads, association betweenbrands is counterintuitive. Noncomparative ads do notexplicitly associate the sponsored brand with another. How-ever, noncomparative ads position the sponsored brand as amember of a product class or product type (Sujan and Dek-leva 1987), thereby inviting comparisons with other brandswithin these categories. Thus, both direct comparative andnoncomparative ads can be associative in nature (Droge andDarmon 1987). Because noncomparative ads do not referdirectly to the comparison or dominant brand in the productcategory, the association is more subtle and must be inferred(e.g., from the similarity of the package or the new product'smarket standing) (Droge and Darmon 1987). Sujan andDekleva (1987) found no differences between comparativeand noncomparative ads, for both experts and novices, in theperceived similarity of brands at the product type level.

Pechmann and Ratneshwar (1991) found that consumersperceive similarities between the comparison and sponsoredbrands even when the comparative ad differentiated thebrands. They reason that consumers categorize the spon-sored brand as similar to the comparison brand because (1)consumers assume that two brands that are compared mustbe comparable; (2) consumers learn through socializationthat leaders are copied, and they deduce that the brands areprobably similar; and (3) consumers infer from visual simi-larities that the brands are functionally similar.

H5: Consumers perceive greater similarity between the spon-sored brand and the comparison brand when sponsors usecomparative ads than when they use noncomparative ads.

Believability. Believability has two components: (1)believability of the source or source credibility and (2) adbelievability, which refers to the perceived truthfulness ofthe ad, or message. Wilkie and Farris (1975) conjecture thatcomparative ads should be more believable than noncom-parative ads. However, recent research has shown that com-parative formats may be less believable (both in source andmessage), particularly for users of the comparison brand(Swinyard 1981). Comparison brand users may attributedesperation and hostility to sponsors making comparative

claims, because these claims challenge consumers' priorbeliefs. Believing the source and the content of the messagerequires that the consumer change some of his or her exist-ing beliefs. Rather than change strongly held beliefs, skep-tical consumers may respond to comparative claims by (1)derogating the source of the message, (2) counterarguingwith the message claims, (3) making statements about thecuriosity of the claims, or (4) making other positive or neg-ative statements about the message (Swinyard 1981;Wright 1973). Belch (1981) found that consumers derogatethe source of comparative claims but not the source of non-comparative claims. Others have found that comparativeads generate more counterarguing than noncomparative adsand are perceived to be less credible (Swinyard 1981; Wil-son and Muderrisoglu 1979).

We operationalize believability of the source as the num-ber of counterarguments and the number of source deroga-tion thoughts. Believability of the message includes thebelievability, credibility, and persuasiveness of the claim.

Hf, : Consumers attribute less credibility to the source of com-parative ad claims than to the source of noncomparativeclaims.

H(,^: Consumers believe comparative claims less than non-comparative claims.

The Effect of Comparative Advertising on Affect

Affective responses to comparative advertising include con-sumers' feelings (likes or dislikes) about the advertisement(Aad), and their feelings about the sponsored brand (A^r).Attitudes are one of the consequences of the informationthat is processed during the cognition stage. The ad format(comparative versus noncomparative) should directly influ-ence affective responses. Ad format should also indirectlyaffect attitudes through the cognitive variables, though wedid not test these effects because the relevant informationwas not available.

Attitude toward the ad. Attitude toward the ad (A^j)denotes a consumer's feelings and overall attitude towardthe ad format. The extant research has shown that compara-tive ads generate fewer favorable attitudinal responsestoward the advertisement than noncomparative ads (Belch1981; Gorn and Weinberg 1984; Swinyard 1981). Oneexplanation for this is that users of the comparison brand seethe comparison as an attack on their brand, causing them toeither derogate the source of the message or counterarguewith the message content (Wilkie and Farris 1975). Empiri-cal evidence seems to support this explanation (Belch 1981;Gom and Weinberg 1984; Swinyard 1981). Droge (1989)reviewed the literature and found that comparative ads areimpersonal, less friendly and pleasant, more aggressive andintense, and less believable and honest. Attitude toward thead includes consumers' self-reports on the perceived friend-liness/offensiveness of the ad and their overall attitudestoward the ad.

H7: Consumers' attitudes toward comparative ads are less pos-itive than their attitudes toward noncomparative ads.

Attitudes toward the sponsor's brand. Whether con-sumers' attitudes toward the ad (A^j) can transfer to the


sponsor's brand is arguable. MacKenzie, Lutz, and Belch(1986) support the notion that feelings and attitudestoward the sponsored brand (A(,r) are positively influ-enced by attitude toward the ad (Agji)- Muehling (1987)provides evidence that shows that comparative ads have apositive influence^on attitude toward the sponsor (Aj,r)across five different types of comparative ad formats.However, Droge (1989) found that Agj predicts A r fornoncomparative ads, but not for comparative ads. Itappears that the question of whether attitudes toward thecomparative ad (A^^) transfer to the sponsored brand isnot settled.

Another factor influencing attitudes toward the spon-sored (A(,r) brand is the association with the comparisonbrand. To the extent that consumers perceive the sponsor'sbrand to be similar to the comparison brand we might expecttheir attitudes toward the sponsored brand (Ab,.) to mirrortheir attitudes toward the comparison brand (Droge and Dar-mon 1987). If attitudes toward the comparison brand arepositive, perceived similarity may lead to more positive atti-tudes toward the sponsored brand (A(,r). It must be notedthat in several studies, the comparison brands are marketleaders, and consumers are likely to have positive attitudestoward them. In these cases, we expect attitudes toward thesponsor's brand (A(,r) to be more favorable when the adver-tisement is comparative than when it is noncomparative.Finally, because comparative ads allow consumers to haveclear and focused perceptions (Wilkie and Farris 1975) andto differentiate the brands (Droge 1989), they should have amore favorable effect on consumers' attitudes toward thesponsored brand (A^r) (Goodwin and Etgar 1980; Gom andWeinberg 1984; Levine 1976; Mazis 1976; Shimp and Dyer1978) than noncomparative ads do.

Hg: Consumers' attitudes toward the sponsored brand (A(,r) aremore positive when the advertisements are comparativethan when they are noncomparative.

The Effect of Comparative Advertising on Conation

The conative component may be the most important becauseit shows whether comparative advertising affects con-sumers' purchase behavior (Pechmann and Stewart 1990a).Previous research frequently operationalized conation aspurchase likelihood and infrequently as actual purchasebehavior (e.g., coupon redemption by Demirdjian [1983]and Swinyard [1981]). Nevertheless, the effect of ad formaton the consumer's intention to purchase and actual purchaseis equivocal. Lavidge and Steiner (1961) reason that if acomparative ad positively influences cognitive and/or affec-tive responses, it should favorably influence behavioralintentions and behavior. Droge (1989) provides partial sup-port: She found a stronger relationship between attitudetoward the sponsored brand (A(,r) and purchase intention forcomparative ads than for noncomparative ads. Additionally,Droge found that cognition has no direct effect on intention;it is mediated by attitude. Swinyard (1981) found no differ-ence between comparative and noncomparative ads onintention to purchase. However, he found a significant effecton purchase behavior due to noncomparative ads' one-sidedclaims (i.e., coupon redemptions). Presumably, if the adver-tising claim is accepted and causes a favorable attitude

toward the sponsored brand (A(,r), the consumer is morelikely to purchase it.

Hy : Consumers' intentions to purchase the sponsored brandare greater when the ad is comparative than when it isnoncomparative.

Hgi,: Consumers are more likely to purchase the sponsoredbrand when the ad is comparative than when it is non-comparative.

Six Potentiai Moderating Variabies

We used three criteria to select from the many potential vari-ables moderating the relationship between ad format andconsumer response. Our analysis of moderators is limited tothose that (1) are theoretically relevant, (2) provide a suffi-cient number of effect sizes to test the relationship, and (3)are important to advertisers. Studies examining cognitionprovided little guidance as to what factors should moderatethe ad format/cognition relationship or why they do. There-fore, our hypothesis development is limited to attitudetoward the brand (A|,r) and intention to purchase; nohypotheses were developed for the cognitive dependentvariables. The moderators are (1) the sponsored brand'scompetitive position (Iyer 1988; Mazis 1976; Pechmann andStewart 1990a; Shimp and Dyer 1978), (2) the comparisonbrand's market position (Iyer 1988; Mazis 1976; Pechmannand Stewart 1990a; Shimp and Dyer 1978), (3) the spon-sored brand's relative market position (Iyer 1988; Mazis1976; Pechmann and Stewart 1990a; Shimp and Dyer 1978),(4) the comparative ad's credibility (e.g., Kavanoor, Grewal,and Blodgett 1997), (5) the factual content of the message(Iyer 1988), and (6) the nature of the dependent measure,that is, relative versus absolute (Miniard et al. 1993; Rose etal. 1993; Snyder 1992).^

Relative market position. Market position is thought tomoderate the relationship between ad format (comparativeversus noncomparative), ad persuasiveness (Pechmann andStewart 1991), and intent to purchase the sponsor's brand(Pechmann and Stewart 1990a). Consumers are more likelyto be positively influenced in favor of a low-share sponsoredbrand when it is compared to a high-share brand than whenit is compared to a moderate- or low-share brand. There arethree reasons for this. First, comparing low-share brandswith high-share brands draws interest to the advertisementbecause of the high-share brand's personal relevance tomany more consumers (Pechmann and Stewart 1990a). Sec-ond, comparing a low-share brand to one with a high shareelevates the image of the brand with the lower share (Pech-mann and Stewart 1990a). Third, comparing a new brandwith an established brand or the market leader may be con-sidered novel, receive more attention, and be perceived asmore informative (Iyer 1988). This can result in a greaterimpact on consumers' attitudes and purchase intention.When the ad is noncomparative, any consideration of mar-ket position by consumers is based on implicit ad cues andshould not have as great a systematic effect on their behav-ior as when the market share information is explicitly stated.Therefore, we expect that infonnation about market position

^Examining the main effects of these six variables is outside thescope of this study.


has a substantial positive effect on consumers' responses tothe ad when it is comparative, but little or no effect when itis noncomparative.

Thus, when the sponsored brand is new or fictitious, therelative effectiveness of comparative advertising should begreatest, and when the sponsored brand is an establishedbrand or market leader we should notice decreases in the rel-ative effectiveness of comparative ads (new brands > estab-lished followers > market leaders).

H|o: Consumers should have more favorable sponsored brandattitudes and greater purchase intentions after viewingcomparative ads, versus noncomparative ads, when thesponsor's brand is new to the category than when it is anestablished hrand or market leader.

Tbe comparison brand's market position also mightaffect the influence of ad format. There is relatively littleadvantage in using a new brand as the comparison brand,because it is relevant to fewer consumers and is less infor-mative than established brands. Altematively, establishedbrands and market leaders are relevant to more consumers,are more informative, and are suitable targets for brandassociation strategies (new brands < established followers <market leaders).

H||: Consumers' sponsored brand attitudes should be mostpositive and their purchase intentions greatest after view-ing comparative ads, versus noncomparative ads, whenthe comparison brand is the market leader. The differencebetween comparative and noncomparative ads shoulddecrease when the comparison brand is an established fol-lower, a new brand, or a fictitious brand.

Thus, sponsors with new or unfamiliar brand names arelikely to benefit from comparative advertising more than theestablished brands or market leaders. Conversely, when thesponsored brand is a market leader, it should benefit lessfrom comparative advertising than will established brandsand new brands (Pechmann and Stewart 1990a, 1991).

H|2: Consumers' sponsored brand attitudes should be morepositive and their intentions to purchase greater afterviewing comparative ads, versus noncomparative ads,when the sponsor's brand market position is less than thecomparison brand. This effect should be attenuated whenthe sponsor's brand market position is greater than orequal to the comparison brand.

Source credibility enhancers. Previously, we hypothe-sized that consumers perceive comparative ads to be lessbelievable than noncomparative ads (H^^ and Hgj,) and thatthey induce source derogation and counterarguing. Bolster-ing the credibility of comparative ads should decrease con-sumers' tendencies to derogate the source and counterarguewith the major claims. Enhanced source credibility shouldbenefit comparative ads more than noncomparative ads,because the latter are inherently more believable. Whencredibility is high, comparative ads will be more persuasivethan noncomparative ads. When credibility is low, compar-ative ads will be equal to or less persuasive than noncom-parative ads. Swinyard (1981) found that one-sided non-comparative claims were more acceptable to consumersthan one-sided comparative claims. The opposite was truewhen credibility was enhanced by both sides of the claim

being presented. Steps to reduce or moderate the negativeeffects of source derogation and counterarguing andincrease the effectiveness of comparative advertising rela-tive to noncomparative ads include (1) using two-sided mes-sages (Etgar and Goodwin 1982; Swinyard 1981), (2) estab-lishing credible sources (Gotlieb and Sarel 1991), (3) sub-stantiating the claim (Golden 1979), and (4) attending to thecontent of the message (Iyer 1988).

H|3: Consumers' attitudes toward the sponsor's brand will bemore positive and their purchase intentions greater afterviewing comparative ads, versus noncomparative ads,when the credibility of the ad is enhanced. The relativeeffectiveness of comparative ads will decrease when thecredibility of the ad is not enhanced.

Message content. Another moderating factor is the mes-sage content of the advertisement. The credibility of a com-parative claim may be enhanced by including factual infor-mation in the content of the message (Iyer 1988). Factualinformation (i.e., information that is objective in nature)compared with evaluative information (i.e., information thatis subjective in nature), elicits fewer counterarguments(Edell and Staelin 1983), encourages more support argu-ments (Edell and Staelin 1983), and enhances the visual-imaging abilities of consumers (Iyer 1988; Paivio 1969;Paivio, Yuille, and Madigan 1968). Thus, factual informa-tion may be perceived as more informative. As a result, con-sumers are less likely to derogate the source of the messageand argue against the comparative claim.

H14: Consumers' attitudes toward the sponsor's brand will bemore positive and their purchase intentions greater afterviewing comparative ads, versus noncomparative ads,when the message content is factual rather than evaluativein nature. The relative effectiveness of comparative adsover noncomparative ads will decrease when the messagecontent is evaluative.

Relative versus nonrelative measures. Finally, Miniardand colleagues (1993) bave alerted us to a measurement fac-tor that may attenuate the magnitudes of the comparative/noncomparative advertising effects in previous research.Rose and colleagues (1993) and Miniard and colleagues(1993) demonstrate that a relative scale, such as "How likelyis it that brand X is more durable than Brand Y?" produceslarger effects than absolute scales, such as "How likely is itthat Brand X is durable?" They hypothesize that a dependentmeasure's sensitivity to an advertising effect depends on thecorrespondence between the frame of the advertising mes-sage and the response frame. If the message is framed as acomparison between a new brand and established brand, themessage is likely to be encoded and stored in memory in rel-ative terms (i.e., the sponsored brand relative to the compar-ison brand). Because noncomparative messages are notcoded in relative terms, they are not encoded or stored assuch by the targeted consumers. Likewise, the dependentmeasure can be framed in relative or absolute terms.Miniard and colleagues show that if the frame of the depen-dent measure does not correspond to the frame used in themessage, the advertising effect may not be detected. Theircompelling arguments are supported by empirical evidence.Therefore, we examine whether the relationship between the


comparative ad frame and its effect is moderated by the waythe dependent variable is framed (i.e., relative versus non-relative measures).

H15: The effects of comparative advertising versus noncompar-ative advertising on consumers' attitude toward the spon-sor's brand and purchase intention should be greater whenrelative measures are used than when nonrelative mea-sures are used. Nonrelative dependent measures shouldattenuate the magnitude of the comparative advertisingeffect.

Review ProceduresSampling Frame and Search ProcessWe synthesize 22 years (1975-1996) of empirical research(published and not published) on the effectiveness of com-parative advertising. The search for relevant studies beganwith thorough computerized searches using ABl/INFORM,DIALOGUE, and INEOTRAC. Additionally, the searchexamined the Business Periodicals Index, ERIC documents,PSYCHLIT, Social Sciences Citation Index, and Disserta-tion Abstracts. Next, we examined all journals containingresearch on advertising. This effort included the Journal ofMarketing, Journal of Marketing Research, Journal of Con-sumer Research, Journal of the Academy of Marketing Sci-ence, Journal of Advertising, Journal of AdvertisingResearch, all journals that were referenced in the reviewedarticles, and the proceedings from national and regional pro-fessional associations, including the American MarketingAssociation, Association for Consumer Research, AmericanAdvertising Association, and Academy of Marketing Sci-ence. We also examined all of the reference lists from thecomparative advertising articles that we had collected.Einally, we contacted several comparative ad researchers sothat we could obtain their unpublished working papers. Thesearch produced more than 100 published and unpublishedarticles (including dissertations) on comparative advertis-ing. However, only 77 studies (54 published and 23 unpub-lished) compared comparative advertising to noncompara-tive advertising and provided the necessary information tocalculate effect size estimates.3

following types of articles were not included in our meta-analysis articles because they did not provide the necessary infor-mation: conceptual and/or review pieces, content and frequencyanalyses (e.g., Jackson, Brown, and Harmon 1979), those thatexamined general perceptions about comparative advertising (e.g.,Rogers and Williams 1989), articles that did not contrast compara-tive and noncomparative ads (e.g., Etgar and Goodwin 1982;Muehling 1987), studies that made no brand comparisons (e.g.,political advertising—see Hill [1989]—and price comparison stud-ie.s—.see review by Grewal and Compeau [1992]), and studies inwhich the empirical or the conceptual domain did not provide sta-tistics that could be converted to d-statistic effect size. We alsoexcluded studies that used analyses fundamentally different fromother studies (e.g., Pechmann and Stewart [1991] used ANOVA inwhich the total degrees of freedom was based on total number ofads used and not on the number of subjects).

Effect Size ComputationEffect size estimate is defined as an estimate of the degreeto which the ad format/effectiveness relationship is presentin the population of comparative advertising research. Theterm effect size refers to the sample estimate of the popula-tion effect regardless of the particular effect size indicatorused (Fern and Monroe 1996). We converted the various teststatistics (F, t, Z, chi-square) to Cohen's d following the for-mula suggested by Fern and Monroe (1996) and Rosenthal(1984). Cohen's d is the ratio of the difference between themeans to the pooled within-group standard deviation. Theeffect size (weighted d) was weighted using the relevantsample size and reliability information (see the Appendix).We corrected the effect sizes for attenuation due to mea-surement errors if a reliability estimate was available for atleast one study for a particular dependent variable (see theAppendix). On the basis of the available reliability estimateand the number of items used in individual studies, we com-puted reliabilities using the Spearman-Brown formula(Hunter and Schmidt 1990). Thus, the rating scale measures,such as informativeness, ad believability, ad attitude, brandattitude, and purchase intentions, were corrected.

Coding for Meta-Analysis VariabiesSome studies provide more than one effect size estimate. Weaccumulated results and calculated the average effect withina study when an article gave separate reports for differentproducts but used the same subjects, because separate effectsize estimates for each of the multiple products in a singleexperiment would weigh each study's contribution to theaverage effect size, proportional to the number of productsit investigated. This procedure eliminated the disproportion-ate influence from studies that examined multiple products.When subjects or dependent variables differed, we calcu-lated separate effect size estimates across products. Theindependent variable used to partition studies for calculatingthe effect size estimates was ad format (comparative ad ver-sus noncomparative ad). Comparative ads were those thatexplicitly or implicitly compared the sponsor's brand withanother brand in the product category—noncomparative adsdid not. In the few studies using three ad levels (e.g., directcomparative. Brand X, noncomparative), we used directcomparative versus noncomparative ads to calculate theeffect size estimate. The d-statistic was coded positive if thecomparative ad produced more favorable results than thenoncomparative ad and negative otherwise.

Moderating variables were included in our analysiswhen (1) they were theoretically justified by either previousempirical or conceptual articles (e.g., Etgar and Goodwin1977; Grossbart, Muehling, and Kangun 1986), (2) theywere used in a least two previous studies and allowed statis-tical analysis, and (3) they were managerially relevant. Theuse of these criteria resulted in the six moderating variablesdiscussed previously. We limit the analysis to the dependentvariables most commonly used across studies (e.g., cogni-tion: message awareness, informativeness, similarity, andmessage believability; affective responses: attitude towardthe sponsor's brand; and conation: intention to purchase thesponsor's brand).


Analysis and ResultsMethod of AnalysisOur objective was to discover whether comparative adver-tising is more effective tban noncomparative advertising. Tothis end, we calculated tbe simple average effect size esti-mate (d), the weighted average effect size, the standard errorof the estimate, tbe 95% confidence interval (Hunter andSchmidt 1990), and tbe file drawer N, which indicates howrobust the results are (Rosenthal 1984). If the confidenceinterval does not include zero, the effect is significant.

The file drawer N tells us how vulnerable our analysis isto future null results. It also represents the number of unpub-lished null-effect studies (i.e., the studies that have not beenpublished because of null results) that would be needed tobring the significant effect (i.e., tbe simple average d in Table1) down to the level of Just significant. The file drawer N wasestimated for tbree levels of "just significance"—.05, .10,and .15. For example, to bring the significant effect of ad for-mat on attention down to tbe level of just significant at a =.05, it would require 32 null-results to be uncovered andadded to our analysis. In Table 1, we summarize all of thisinformation. Where applicable, tbe reported results are cal-culated from effect sizes corrected for attenuation due to theunreliable dependent measure. Positive mean effect sizesindicate that tbe effect of comparative advertising is greaterthan that of noncomparative advertising on the dependentvariable. Negative mean effect sizes indicate that tbe effectof noncomparative advertising is greater.

Results of Main Effects Hypothesis TestsCognitive component. Consistent with Wilkie and Far-

ris's (1975) predictions and Hi, the comparative format gen-erates more attention than noncomparative ads. Also, com-parative ads are more effective than noncomparative inincreasing advertising message awareness (H2a), brand nameawareness (H2b), and increased message processing (H3).

The null effects of comparative advertising on the infor-mativeness of the ads (H4) and positioning, (i.e., perceivedsimilarity of the sponsored brand to the comparison brand)(H5) are unexpected. Before concluding that comparativeads are not more informative, consider that the averageeffect size is positive and the lower bound of the confidenceinterval (-.01) just barely includes zero. Arguably, compar-ative ads may be as informative or more informative thannoncomparative ads depending on whether the notion ofmarginal significance is accepted. Comparative ads do notappear to be more effective in creating the perception thatthe sponsored brand is similar to tbe comparison brand.

Source believability is less for comparative tban non-comparative ads (Hga). Tbis finding is consistent witb thenotion tbat consumers wbo are exposed to comparative adsmay use source derogation as an information processingstrategy (Belch 1981). When consumers derogate the sourceof a persuasive communication, they also may infer that theadvertiser and the message are not credible (H^^). However,tbe negligible and nonsignificant difference in tbe averagemessage believability is inconsistent with tbis explanation.If the source is less credible, we would expect the messageto be less credible, unless consumers differentiate betweenthese two origins of credibility. Additionally, several studiesuse fictitious sponsor brands, which should be intuitivelyless credible. Finally, the lack of difference in claim believ-ability might be explained by otber study factors, such as thetypes of claim substantiation that were used in the research(e.g., Droge 1989; Grossbart, Muehling, and Kangun 1986;Park, Lessig, and Lee 1990). Claim substantiation mightenhance message believability for botb comparative andnoncomparative ads; however, the effect should be greaterfor comparative ads. We examine this possibility in ouranalysis of moderators.

Affective component. The results indicate that compara-tive ads generate more negative attitudes toward the ad(H7), but more favorable attitudes toward the sponsored

TABLE 1Main Effects Hypothesis Tests

DependentVariables

AttentionlUlessage AwarenessBrand AwarenessProcessingsinformativeness''SimilaritySource BelievabiiityAd Believabiiity''Ad Attitude»Brand Attitude^)Intention'^Overt Behavior

Ho

HiH2a

H3H4H5HeaH6bH7HsHgaH9b

N

31824

535238

362042476

SimpleAverage d

.59.43.38.11.21.07

-.30-.07-.30.29.27.46

WeightedAverage d

.47

.27

.33

.36

.11-.17-.26

.01-.28

.23

.20

.46

StandardError

.17

.13

.12

.06

.05

.13

.10

.06

.09

.05

.04

.15

95%Confidence

Interval

LCi

.13

.01

.09

.03-.01-.42-.45-.13-.45

.13

.12

.16

UCI

.81

.52

.57

.29

.23

.09-.07

.12-.20

.32.28.75 .

File

.05

32136160

5NCNC40

NC10020220748

Drawer

.10

155968

0NCNC16NC40808021

N

.15

933370

NCNC

8NC20393812

aEffect size is homogeneous.'•Effect size was corrected for measurement error. The uncorrected simple average d's are informativeness (.19), ad believability (-.06), adattitude (-.25), brand attitude (.26), and intention (.25).

Note: NC means the file drawer N was not calculated, because the 95% confidence interval contained zero.LCI = lower confidence interval. UCI = upper confidence interval.


brand (Hjj), than noncomparative ads. These findings areinconsistent with the affect transfer explanation.^ Perceivedsimilarity to the comparison brand does not account formore favorable brand attitudes in comparative ads, becausethe effect of ad format on similarity is not significant. Fur-ther research may provide explanations for this apparentparadox.

Conative component. Compared to noncomparative ads,comparative ads appear to increase both purchase intention(Hga) and purchase behavior (H91,). For many advertisers,the most important findings are that comparative ads influ-ence behavioral intentions and actual buying behavior morethan noncomparative ads.

The hierarchy of comparative advertising effects. It isnot our goal to test the hierarchical model outlined in Figure1. However, we examined the correlations among the effectsizes associated with the three major constructs in the modelto determine if it has face validity. All of the correlations arepositive (ranging from .53 to .93) and, with one exception,significant. The relationship between message awarenessand brand attitude has a large correlation coefficient (r =.93), but with only three observations, it was not significant.

Resuits of Moderator Hypothesis Tests

To determine whether the relationships uncovered in ouranalysis of main effects would be affected by other causalfactors, we analyzed several moderator variables. Thedependent variables for this analysis are cognition (messageawareness, infonnativeness, similarity, and ad believabil-ity), brand attitude, and purchase intention. Note that whenanalyzing the impact of a moderator variable on a givendependent variable, we use all available data points (i.e.,studies that explicitly manipulate a specific moderator as anindependent variable provide multiple data points). Thenumber of observations used in each hypothesis test variesbecause of missing data. On the basis of our hypotheses, sixvariables were coded as moderators. The results are pre-sented in Table 2.

Moderator effects on cognition. As was mentioned pre-viously, there is little theory about the effects of moderatingvariables on the relationship between ad format and cogni-tive variables. Nevertheless, we tested these effects to com-plete our analysis; we have no hypotheses about these mod-erators. Because relative measures and the factual/evalua-tive nature of message content are not included in studies oncognition, we only examined the moderating effects of thesponsored brands' market position, the comparison brands'market position, the relative market position of the brands,and ad credibility enhancers (i.e., attempts to substantiatethe message claim).

Message awareness is greater for comparative ads whenthe comparison brand is the established brand than when itis the market leader. Note, however, that effect size esti-mates were not available when the market leader is thesponsored brand or when a new brand is used for the com-parison brand. These omissions from this research stream

"•As a reviewer noted, these studie.s used single ad exposures,and repetition may be necessary for affect transfer to occur.

make intuitive sense, because the market leader has little togain by comparing itself to an established or new brand.Remember that we found that comparative ad messages arerecalled more than noncomparative messages. This disparityincreases when relative market position is considered. Whenthe sponsored brand's relative market position is equal to orgreater than the comparison brand's market position, thecomparative ad's message awareness increases substantiallyover that of noncomparative ads. Finally, comparative adscreate more message awareness when credibility enhancersare not used. Credibility enhancers may bewilder or confuseconsumers, causing them to focus on the source or reasonsfor the credibility claim rather than on the content of themessage (Belch 1981).

Informativeness differs across ad formats depending onthe levels of the sponsored brand's market position. Appar-ently, consumers perceive the informativeness benefit fromcomparative ads to be greater for established sponsoredbrands than for new brands or market leaders. When thesponsored brand is the market leader noncomparative adsare perceived to be more informative than comparative ads.The infonnativeness effect size estimate is larger for estab-lished comparison brands than the market leader, but thecontrast is nonsignificant. Again, there appears to be no rea-son to put a new brand in the role of comparison brand, andno effect sizes were obtained for this level. The moderatingeffect sizes of relative market position and ad credibilityenhancers are of approximately the same magnitude and arenot statistically significant.

Perceived similarity between the sponsored and compar-ison brands is significantly greater for comparative adswhen the sponsored brand is an established brand than whenit is a new brand. This is not surprising, because comparisonbrands are almost always established brands, and sponsoredbrands that are established in the market are objectivelymore similar to the comparison brand. Likewise, the spon-sored brand in comparative ads is perceived to be more sim-ilar to the established comparison brand than the marketleader, but the effect is not significant. Generally, compara-tive ads appear to be more effective when the sponsoredbrand's relative market position is equal to or greater thanthe comparison brand's, but this effect is not statistically sig-nificant. Ad credibility enhancers do not moderate the adformat/similarity relationship.

Ad believability does not differ significantly across dif-ferent levels of the sponsored brand's market position, thecomparison brand's market position, or ad credibilityenhancers. Relative market position significantly moderatesthe relationship between ad format and ad believability. Forsponsored brands with market positions equal to or greaterthan the comparison brand, noncomparative ads are per-ceived to be relatively more believable than when the spon-sored brand's position is less than the comparison brand's.

Moderator effects on brand attitude. Comparative adsgenerate more positive consumer attitudes toward the spon-sored brand than noncomparative ads when the sponsoredbrand is new to the category (H|o). This effect decreaseswhen the sponsored brand is well established and reverseswhen the sponsored brand is the market leader (i.e., non-comparative ads are more effective than comparative ads).


(0

ra

ICM OUJ re

om

mila

rity

(A

w<uH-

umbe

r

z

u

Sti

•5

1UJ

3

5>1

Ci

tud

ies

(A

Siz

e

CO

t~~CMCM

CO • *

CO 0q CO

(0(01)

IE

ror

c

m

ber

E3

(0

(0

"o

4..*

0a>

Eff

—>t i

ies

pni

(A

a>N(A

(010

a>

II

.a 0)

2CA N .

I.IE o •o

0)

u o>

10 CMCNJ

CO CMCM 0 0

0 0

COi

COCM

in 00

CM IDCM O

COCM

00 10CO CM

CO

CJ)

CO O

C35CO

CD

UJ

O COCM

CO OiO CM

inCO

CO

1 - CO

o o

CO

( O CMCM

CO CMCM CO

CO

CO CO

CJ)

CMCM

O 00 CO CO

CO CM I - 1 -

cu

tabl

ist

CD CO

2 UJ

ader

cu_J

T3C

2COc0CO

COQ .

E00V

onso

r

Q .

in

•0cCO

mc0cnCOd .

E00Al

onso

r

CO

C/J . . .

cu 0> Z

0

0

1T30

s

cIBJ(

<D

;uod

CO

0

siti

0CL

-iCO

T3CCO

c0CO

COdLUO

0

c0

siti

0

1

E>

elat

i

DC

^ ^CM

c

;itio

UJ

§.

I

• = <U

— c

b c• D <"

lo rt.3 —TS COCO >

LL LU

c

oocuO)COcncn ^-^cu •*

^£

c00

T3CCO

new

CD

0

c0CO

COCL

0

aiue

N

etI-

• 0

cCO

£ o

05 T3

2§

N 0,

1

tive

ela

DC

^ •

c

"Scu

of d

e

Q .

cu>150

0

—cu

cai

cn

'iabl

e

>

c0 iti

CO0

COr~

nd's

nB

rai

c

aris

oC

omp

cCO

tion

CO

1ECO

T 3CCO

Q

c0Q .

CO

0

_C0

1p

sil

CO

CO

si;

The

el

te:

0

T3C0

COr iJBJ

. 0

CD

•§

CD

fth

ais

on 0

0 .

JOO

CD

CO

03

>1

SJ

Q

CDCO

the

tio

c

CO

c

1.c• . =

•2fl)CD

£

ditio

ntio

n.co

ndi

COT3

c.a0)

0

c0

iti

• 0

c8CO

C

1"0

son

g.

com

i

c0

~COCL

fc00CD

CO

§

ati

"-

this

ned

CO

CD

:=3

1CDC0

II

1

COCD

X>

tai

d)0

wer

eth

e c

inpV

CO

COa)

iled

eg6BCD

CO

I0 '

CD

ort

he

sf

-vai

u

NCD

c

BCO0

CD

ifica

nc

cO)cn


04 0)

on

snt

enti

S!(0

"i3Q.

a>

Att

it'a

nd

CD

ivab

ilil

»!iag pi

00

jmb

er

z

(0

|2!

Num

ber

Test

Num

ber

tistic

alue

)S

tal

(Z-v

"5

Eff

eciti

c

VJ

Sta

l

o

Eff

ect

o

tatis

t

in

°

Eff

e

(0a>

ludi

J/)

Siz

e

(an

n

(Z-V

lies

0)

Siz

-vai

u

Ma>

Stu

di

«

Siz

evei

o2Hi

Mod

i

CDCM

inCO

.29

-aCMCM

•

CM

COi n

CM

CD

CDO

l'

(U

isor

ed b

rand

's

oa.CO

CJ)

i n

CO

-o

o

iri

CD

CDo

r

cu

stab

lis

LU

o

irket

pos

ition

(H

i

E

CM

i n

!;::

-DOCO

in

CDo

1

TJCO

nO

-*"

CO

CO

o

u

CO

CO

. ^

cuz

paris

on b

rand

'sC

omi

n

CM

in

in

COCM

in

CO

1.48

CM

q

TJcux:

Sta

blis

UJ

irket

pos

ition

(H

i

E

COCO

COCO

COCO

-o

CO

CO

CO

COCM

CO

oq

TJCO(U

CO

n

inCO

COCO

CD

2.59

a

CM

0

r

rand

mc0S

UB

Com

p

V

pons

o

CO

tive

mar

ket

Rel

al

CM

COCM

inCM

1

CO

CO

r

TJ

2mc0S

UB

Com

p

Al

osuod

CO

g

0Q .

qCM

0CM

0

-DinCOCO

CM

CM.9

8

••-

7

cn

redi

bilit

yA

d C

I

CM

CO

CM

CO

CJ)

CMCM

r

0

z

5)0

CO

cu

CM

0CO

. 0

03

CM

1

1

lo

tjCO

u.

iess

age

cont

ent

tTJ

i n

0

in

i n

1

1

<u>

valu

ati

LU

ininCO

in

CO

Sq

COCM

CO

elat

ive

DC

of d

epen

dent

Typ

e

T -

•r0J

COCM

in0

onre

la

cuCO0cn

cuX)CO

iBA

3 co

n-an

d'i

new

br

the

•5c0CO

com

par

3 th

e

CO

.2CD

ilue r

§N

, th

etio

nos

ike

t p

raECO

- 0cro

on B

rer

s to

fl)

fl)

Z-v

alu

hird

eti

£

crocg

cond

md'

sd

ert

lea

0 th

etio

n t

c00CO

T3cra

-DszCO

2

3 es

ta

sz

Com

pais

on o

fio

n and

com

pari

posi

tD

the

arke

t

CO

T3

m73CD

0COc0CL

C/3

d's.

For

_cn

J ave

rai

sim

ple

The

effe

ct s

izes

ai

Not

e:

fl)

CD

dZ

-va

c0fl)

CD

ditio

n, t

0

rand

's c

hed

bl

CO

ditio

n to

the

est

ab

cf0

cond

itibr

and'

s

fl)

the

l(

0

c

:ioco

ndit

d's

cro

5

f the

ne

0

ipar

isoi

t0ufl)

the

com

paris

on o

fhi

p.el

atio

ns

—CO

ined

thex

am

COCD

one

or

few

er s

tudi

II

J

CO

test

-0fl)

le-t

ail

[0fl)

fl)gCO

othe

r

c-

< .0

5, tl

3 w

ith p

to©

0ra

fl)

es w

er

X)ra

ra

cogn

iti

CD

jes

for

Z-v

ali

ifica

nce

test

s on t

l-05

0.01

0.00

1.

O) V V Vc/3 C i CL caCO .Q 0 "O

Comparative Versus Noncomparative /11

Contrary to H||, the difference in attitude (between compar-ative and noncomparative ads) toward the spon.sored brandis most positive when the comparison brand is established inthe product category rather than when it is the market leader.When the sponsored brand's market share is less than thecomparison brand, comparative ads result in significantlymore positive attitudes toward the sponsored brand (H12);otherwise noncomparative ads result in more positive atti-tudes than comparative ads. These results indicate that com-parative advertising is most effective in enhancing the spon-sored brand's attitude when the sponsor is new to the cate-gory and the comparison brand is established in the marketbut is not the market leader.

As we expected, enhancing the credibility of compara-tive ads leads to a significant increase in the favorablenessof attitudes toward the sponsored brand (H13). Contrary toH|4, when comparative ads contain evaluative messages,their impact on attitudes is greater than when they are fac-tual. Factual information may be more informative, but itdoes not appear to influence consumers' attitudes morethan evaluative information. This seems to suggest thatcomparative ads do not foster attribute-based attitude for-mation; rather, they seem to inspire an affective response.Stronger conclusions must await further research on theserelationships.

Relative attitude scales account for greater effects due tocomparative advertising than do absolute scales. For the lat-ter scales, noncomparative ads generate more positive atti-tudes than comparative ads. According to Rose and col-leagues (1993) and Miniard and colleagues (1993), thismakes sense because relative measures correspond to therelative nature of comparative ads, and absolute measurescorrespond to the nonrelative nature of noncomparative ads

Moderator effects on intention to purchase. Contrary toHio, the sponsored brand's market position does not affectthe relationship between ad format and intention to pur-chase. However, the comparison brand's market positionsignificantly moderates the relationship between ad formatand intention to purchase the sponsored brand (H||). Con-sumers are more likely to purchase the sponsored brandwhen the comparison brand is the market leader. When thesponsored brand's market share is less than the comparisonbrand's, comparative ads result in significantly greaterintention to purchase the sponsored brand than noncompar-ative ads do (H12).

Compared to noncomparative ads, comparative ads withenhanced credibility significantly increase intention to pur-chase (H13). It was noted previously that comparative adsare less believable than noncomparative ads (see Table 1).However, purchase intention is higher when comparativeads are used. This inconsistency is resolved when the stud-ies are differentiated on the basis of credibility enhancers.First, the use of credibility enhancers as a partitioning vari-able indicates that comparative ads with enhanced credibil-ity are not significantly less believable than noncomparativeads. Second, more credible comparative ads increase themagnitude of the effect size, which indicates higher levels ofintention to purchase the sponsored brand.

The moderating effect of message content on the rela-tionship between ad format and purchase intention is sup-ported. Comparative ads account for higher intention to pur-chase than noncomparative ads when they contain factualinformation instead of evaluative infonnation (H14). Appar-ently, factual information is more effective in enhancingintention to purchase, but evaluative information is moreeffective in increasing the favorableness of attitudes towardthe sponsored brand. This seems to make sense, particularlyif the attitudes are affect based.

The effect of comparative advertising on intention topurchase is greater when the dependent variable is measuredwith relative measures than when it is measured absolutely(H15). These results are consistent with Miniard and col-leagues' research and our previous findings regarding theeffects of this moderator on brand attitude.

Managerial ImplicationsThe inferences that can be drawn from our meta-analysisdepend on the interpretation of the reported average effectsizes. For ease of interpretation, we converted the effectsizes in Table 1 to correlation coefficients (r = d / [d^ +4] 1/2). Most of the correlations are relatively small, rangingin absolute value from .10 to .23. However, low values of rdo not mean that these findings are unimportant. Rosenthaland Rubin (1982) provide an intuitive and helpful way tointerpret small effects which they call the binomial effectsize display (BBSD). We use the BESD to look at theincrease in the effectiveness of comparative ads over non-comparative ads. If there are no differences due to theeffects from the ad format used, responses to the dependentvariables would be the same for those who view the com-parative ad and those who view the noncomparative ad (i.e.,r = 0) and there would be a 50:50 chance that the compara-tive ad would be more effective than the noncomparative ad.The average correlation coefficient for message awarenessis .23, which means that comparative ads are read by 23%more people than noncomparative ads (i.e., the difference ineffectiveness is 23%). The eight managerial implicationsthat follow are based on similar BESD analyses. Keep inmind, however, that the studies producing these effects arecontrolled experiments with single exposures to the stimu-lus, and the consumers' responses were measured immedi-ately after the stimulus materials were presented. Also, thenumber of studies used to calculate the effect sizes variesacross independent and dependent variables. Thus, cautionshould be exercised in generalizing these findings to spe-cific industry applications.

1. Message awareness (number of message points recalled)and brand awareness (brand name recall) are greater forcomparative ads than for noncomparative ads, 12% and16% respectively.

2. Source believability and ad attitude are negatively corre-lated with ad format (r = -. 13 for both) which means thatcomparative advertising is 13% less believable and gener-ates 13% less favoi"able attitudes toward the ad than non-comparative advertising.

3. Comparative advertising generates 22% more purcha.sesthan noncomparative ads.

12 / Journai of Marketing, October 1997

4. When the comparison brand i.s established, 36% more mes-sage points are recalled in comparative ads than noncom-parative ads.

5. Comparative advertising is responsible for 28% more favor-able attitudes toward new sponsored brands than noncom-parative ads are. This advantage is 5% for established spon-sored brands.

6. When comparing the sponsored brand with an establishedbrand, comparative ads seem to realize 22% more favorablespon.sored brand attitudes than noncomparative ads do.

7. Enhancing the credibility of comparative ads may increasethe favorable attitude toward the sponsored brand 21%beyond that of noncomparative ads.

8. By using the market leader rather than another establishedbrand as the comparison brand, intentions to purchase thesponsored brand may be 17% above what would beexpected from noncomparative ads.

In general, comparative advertising appears to be moreeffective than noncomparative advertising, particularlywhen the sponsored brand is new to the product categoryand when the comparison brand is the market leader. Appar-ently the advantages of comparative advertising help the"underdog." The major exception appears to be that the dif-ference in message recall between comparative and non-comparative ads is greater when the sponsored brand isestablished than when the sponsor is new to the category.

LimitationsMeta-analyses are informative, but they have limitations, aswell. First, our findings are based on secondary data; there-fore, we cannot make strong causal statements about the rela-tionships we have uncovered. Second, though meta-analysisproponents say that it is appropriate to test effects from justa few studies, we remain cautious. Recognizing the small nfor some of the cognitive dependent variables (attention, pro-cessing, brand awareness, and source believability), we urgereaders to be careful about drawing unwarranted conclusionsfrom these results. Third, some of the moderating variablesexhibit heterogeneous effect sizes; hence, other variablesmay account for some of the differences among these stud-ies. Typical of meta-analyses, there were insufficient studiesto conduct further tests for homogeneity. Therefore, morecomplex relationships remain unexplored. Fourth, for meta-analysis purposes, it is assumed that the effect sizes beingaccumulated are from a homogeneous population of studiesand that any causes of heterogeneity can be isolated andidentified. This often is not the case (Fern and Monroe 1996).There are many differences across a population of studies, andtypically there are not enough observations to test for all ofthe.se differences. Therefore, the effect sizes we report shouldbe treated as averages and interpreted with care.

ConclusionsOur meta-analysis reveals that comparative ads generallyelicit (1) more attention to the ad, (2) greater message andbrand awareness, (3) increased information processing, (4)more favorable brand attitudes, and (5) increased purchaseintentions and increased purchase behaviors. Moreover,comparative ads evoke lower source believability and less

positive attitudes toward the ad. No differences were foundfor informativeness, believability of the advertised message,and positioning as similar to the comparison brand.

We identify several factors that moderate the relation-ship between ad format and attitude toward the sponsoredbrand and intention to purchase the sponsored brand. Thosefactors that enhance the effects of comparative advertisinginclude relative market position, whether an attempt wasmade to enhance credibility, whether the message content isfactual, and the type of scale used to measure the effect.

With the exceptions noted previously, on average, com-parative advertising is more effective than noncomparativeadvertising. Moreover, this increase in effectiveness maytranslate into substantial advantages for those firms that useit. The analysis of moderating variables indicates that newbrands that compare themselves to established brands bene-fit most from comparative advertising. Whether enhancedcredibility is desirable depends on the advertising goal.Enhanced credibility has a positive effect on sponsoredbrand attitudes and intention to purchase the sponsoredbrand; however, it has a negative impact on message aware-ness. The differences in advertising goals should be seri-ously considered when deciding between comparative andnoncomparative advertising.

AppendixThe F, t, z, and Chi-square values (with 1 dO were convertedto Cohen's d. Within cumulation was carried out when nec-essary. Hunter and Schmidt (1990) recommend withincumulation if there is replication within a study, because theindividual effect sizes are not likely to be independent ofeach other.

The weighted average of effect sizes D is given byZ«(W|*di)/Z"W|, where d| is the individual effect size, w; isthe weight associated with individual effect size. The weightis equal to the sample size (Nj) for uncorrected effect sizes.If the effect size is corrected for attenuation due to unrelia-bility associated with the measure of dependent variable,then the weight is equal to N|*alpha. Effect sizes were cor-rected for attenuation due to unreliability by dividing theeffect sizes with the square root of alpha.

The confidence interval is computed around theweighted average effect size. It is actually a second-orderanalysis. The weighted average effect size in a given meta-analysis is only an estimate (even if all the currently avail-able studies have been included in the analysis) of thedomain-wide population effect size. The confidence intervalgives the range within which the domain-wide effect sizemay be found. The confidence interval is given by weightedaverage effect size +/- 1.96*standard error(Epsilon). Vari-ance(Epsilon) is the sampling variance divided by K (thenumber of studies cumulated) if the effect sizes are homoge-nous and observed variance divided by K, otherwise.

Homogeneity is checked with the formula (K*observedvariance) divided by sampling variance.

The formula for file draw is K*(D - dj.)/d(., where K isthe number of studies cumulated, d is the simple average ofindividual effect sizes, and d(. refers to the critical d value.


Because the denominator will be zero for d . = 0, that com-parison cannot be done.

To do a two-group moderator analysis, the effect sizes

were converted to r and then Z , where Z,. = .5*Logg([1 + r]/[l - r]). Moderator procedures suggested by Rosen-thal (1984) were followed.

REFERENCESBarry, Thomas E. (1993), "Twenty Years of Comparative Advertis-

ing in the United States," International Journal of Advertising,12 (November), 325-50.

Belch, George E. (1981), "An Examination of Comparative andNoncomparative Television Commercials: The Effects of ClaimVariation and Repetition on Cognitive Response and MessageAcceptance," Journal of Marketing Research, 18 (August),333^9.

Demirdjian, Zohrab S. (1983), "Sales Effectiveness of Compara-tive Advertising: An Experimental Investigation," Journal ofConsumer Research, 10 (December), 362-64.

Droge, Cornelia (1989), "Shaping The Route to Attitude Change:Central Versus Peripheral Processing Through ComparativeVersus Noncomparative Advertising," Journal of MarketingRe.searclh 26 (May), 193-204.

and Rene Y. Darmon (1987), "Associative PositioningStrategies Through Comparative Advertising: Attribute VersusOverall Similarity Approaches," Journal of MarketingResearch, 24 (November), 377-88.

Edell, Julie A. and Richard Staelin (1983), "The Information Pro-cessing of Pictures in Print Advertisements," Journal of Con-sumer Research, 10 (June), 45-61.

Engel, James F., Roger D. Blackwell, and Paul W. Miniard (1993),Consumer Behavior, 7th ed. New York: The Dryden Press.

Etgar, Michael and Stephen M. Goodwin (1977), "ComparisonAdvertising: Issues and Problems," in Advances in ConsumerResearch, Vol. 5, H. Keith Hunt, ed. Provo, UT: Association forConsumer Research, 63-71.

and (1982), "One-Sided Versus Two-Sided Com-parative Message Appeals for New Brand Introduction," Jour-nal of Consumer Research, 8 (March), 460-65.

Fern, Edward F. and Kent B. Monroe (1996), "Effect Size Esti-mates: Issues and Problems in Interpretation," Journal of Con-sumer Research, 23 (September), 89-115.

Golden, Linda L. (1979), "Consumer Reactions to Explicit BrandComparisons in Advertisements," Journal of MarketingResearch, 16 (November), 517-32.

Goodwin, Steven and Michael Etgar (1980), "An ExperimentalInvestigation of Comparative Advertising: Impact of MessageAppeal, Information Load, and Utility of Product Class," Jour-nal of Marketing Research, 17 (May), 187-202.

Gom, Gerald J. and Charles B. Weinberg (1984), "The Impact ofComparative Advertising on Perception and Attitude: SomePositive Findings," Journal of Consumer Research, 11 (Sep-tember), 719-27.

Gotlieb, Jerry B. and Dan Sarel (1991), "Comparative AdvertisingEffectiveness: The Role of Involvement and Source Credibil-ity," Journal of Advertising, 20 (March), 38^5 .

Greenwald, Anthony G. and Clark Leavitt (1985), "AudienceInvolvement in Advertising: Four Levels," Journal of Con-sumer Research, II (June), 581-92.

Grewal, Dhruv and Larry D. Compeau (1992), "Comparative PriceAdvertising: Informative or Deceptive?" Journal of Public Pol-icy and Marketing, 11 (Spring), 52-62.

Grossbart, Sanford, Darrel D. Muehling, and Norman Kangun(1986), "Verbal and Visual References to Competitors in Com-parative Advertising," Journal of Advertising, 15 (March),10-23.

Hill, R. (1989), "An Exploration of Voter Responses to PoliticalAdvertisements," Journal of Advertising, 18 (December),14-22.

Hunter, John E. and Frank L. Schmidt (1990), Methods of Meta-Analysis. Beverly Hills, CA: Sage Publications.

Iyer, Easwar S. (1988), "The Influence of Verbal Content andRelative Newness on the Effectiveness of ComparativeAdvertisements," Journal of Advertising, 17 (September),15-21.

Jackson, Donald W., Jr., Stephen W. Brown, and Robert R. Harmon(1979), "Comparative Magazine Advertisements," Journal ofAdvertising Research, 19 (December), 21-26.

Kavanoor, Sukumar, Dhruv Grewal, and Jeff Blodgett (1997),"Ads Promoting OTC Medications: A Multi-Study Examina-tion of the Effect if Ad Credibility and Ad Format on Beliefs,Attitudes and Purchase Intentions," Journal of BusinessResearch, forthcoming.

Krugman, Herbert E. (1965), "The Impact of Television Advertis-ing: Learning Without Involvement," Public Opinion Quar-terly, 29 (Fall), 349-56.

Lavidge, Robert J. and Gary A. Steiner (1961), "A Model for Pre-dictive Measurements of Advertising Effectiveness," Journal ofMarketing, 25 (October), 59-62.

Levine, Philip (1976), "Commercials That Name CompetingBrands," Journal of Advertising Research, 16 (December),7-14.

MacKenzie, Scott B., Richard J. Lutz, and George E. Belch (1986),"The Role of Attitude Toward the Ad as a Mediator of Adver-tising Effectiveness: A Test of Competing Explanations," Jour-nal of Marketing Research, 23 (May), 130^3.

Mazis, Michael B. (1976), "A Theoretical and Empirical Exami-nation of Comparative Advertising," working paper. Collegeof Business Administration, University of Florida,Gainesville.

McDougall, G. H. G. (1976), "Comparative Advertising: AnEmpirical Investigation of the Role of Consumer Information,"Working Paper No. 167, Department of Marketing, Universityof Western Ontario.

Miniard, Paul W., Randall L. Rose, Michael J. Barone, and Ken-neth C. Manning (1993), "On the Need for Relative MeasuresWhen Assessing Comparative Advertising Effects," Journal ofAdvertising, 22 (September), 41-57.

Muehling, Darrel D. (1987), "Comparative Advertising: The Influ-ence of Attitude-Toward-the-Ad on Brand Evaluation," Journalof Advertising, 16 (December), 4 3 ^9 .

, Jeffrey J. Stoltman, and Sanford Grossbart (1990), "TheImpact of Comparative Advertising on Levels of MessageInvolvement," Journal of Advertising, 19 (December), 41-50.

Neiman, Janet (1987), "The Trouble With Comparative Ads," AdWeek, (January 12), BR4-5.

Paivio, A. (1969), "Mental Imagery and Associative Leaming andMemory," Psychological Review, 76 (May), 241-63.

, C. Yuiile, and S. A. Madigan (1968), "Concreteness,Imagery, and Meaningfulness Values of 925 Nouns," Journal ofExperimental Psychology, 76 (January), 1-25.

Park, C. Whan, V. Parker Lessig, and D. H. Lee (1990), "The Leveland Nature of Product Knowledge and Ad Format Strategies,"in Current Issues and Research in Advertising, James Leigh andClaude R. Martin, Jr., eds. Ann Arbor: University of Michigan,125-54.

Pechmann, Cornelia and Gabriel Esteban (1994), "PersuasionProcesses Associated With Direct Comparative and Noncom-parative Advertising and Implications for Advertising Effec-tiveness," Journal of Consumer Psychology, 2 (4), 403-32.


and S. Ratneshwar (1991), "The Use of ComparativeAdvertising for Brand Positioning: Association Versus Differ-entiation," Journal of Consumer Research, 18 (September),145-60.

• and David W. Stewart (1990a), "The Effects of Compara-tive Advertisement on Attention, Memory and Purchase Inten-tions," Journal of Consumer Research, 17 (September),180-91.

and (1990b), "The Development of a ContingencyModel of Comparative Advertising," Working Paper No.90-108, Marketing Science Institute, Boston.

and (1991), "How Direct Comparative Ads andMarket Share Affect Brand Choice," Journal of AdvertisingResearch, 31 (December), 47-55.

Pride, William M., Charles W. Lamb, and Barbara A. Pletcher(1979), "The Informativeness of Comparative Advertisements:An Empirical Investigation," Journal of Advertising, 8 (June),29-35.

Rogers, John C. and Terrell G. Williams (1989), "ComparativeAdvertising Effectiveness: Practitioners' Perceptions VersusAcademic Research Findings," Journal of AdvertisingResearch, 29 (October/November), 22-37.

Rose, Randall L., Paul W. Miniard, Michael J. Barone, Kenneth C.Manning, and Brian D. Till (1993), "When Persuasion GoesUndetected: The Case of Comparative Advertising," Journal ofMarketing Research, 30 (August), 315-30.

Rosenthal, Robert (1984), Meta-Analytic Procedures for SocialResearch. Beverly Hills, CA: Sage Publications.

and Donald B. Rubin (1982), "A Simple, General PurposeDisplay of Magnitude of Experimental Effect," Journal of Edu-cational Psychology, 74 (April), 166-69.

Shimp, Terence A. and David C. Dyer (1978), "The Effects ofComparative Advertising Mediated by Market Position Spon-

soring Brand," Journal of Advertising, 3 (Summer), 13-19.Snyder, Rita (1992), "Comparative Advertising and Brand Evalua-

tion: Toward Developing a Categorization Approach," Journalof Consumer Psychology, 1 (I) , 15-30.

Stewart, David W. and David H. Furse (1986), Effective TVAdvertising: A Study of 1000 Commercials. Lexington, MA:Lexington Books.

Sujan, Mita and Christine Dekleva (1987), "Product Categoriza-tion and Inference Making: Some Implications for ComparativeAdvertising," Journal of Consumer Research, 14 (September),372-78.

Swinyard, William R. (1981), "The Interaction Between Compara-tive Advertising and Copy Claim Variation," Journal of Mar-keting Research, 18 (May), 175-86.

Tannenbaum, S. (1974), "Policy Statement and Guidelines forComparative Advertising," American Advertising Agencies,April.

Turgeon, Normand and David J. Barnaby (1988), "ComparativeAdvertising: Two Decades of Practice and Research," in CurrentIssues and Research in Advertising, James Leigh and Claude R.Martin, Jr., eds. Ann Arbor: University of Michigan, 41-65.

Wilkie, William L. and Paul Farris (1975), "Comparison Advertis-ing: Problems and Potential," Journal of Marketing, 39 (Octo-ber), 7-15.

Wilson, R. Dale and Aydin Muderrisoglu (1979), "An Analysis ofCognitive Respon.ses to Comparative Advertising," in Advancesin Consumer Research, Vol. 7, Jerry C. Olson, ed. San Fran-cisco, CA: Association for Consumer Research, 566-71.

Wright, Peter L. (1973), "The Cognitive Processes MediatingAcceptance of Advertising," Journal of Marketing Research, 10(February), 53-62.


comparative versus noncomparative advertising: a meta-analysis

Documents

advertising campaigns

comparative advertisings

comparative adsare

comparative advertising

noncomparative advertising

university of miami

brand awareness

marketing consuitant