Comparative Advertising: Issues and Problems

Michael Etgar, State University of New York at Buffalo
Stephen A. Goodwin, State University of New York at Buffalo
ABSTRACT - The recent growth in use of comparative advertising has led to increased efforts to measure its effectiveness. Empirical efforts in this area should address seven important research issues. This paper identifies and expounds upon these issues and evaluates recent empirical studies of comparative advertising, particularly with respect to the way they have addressed these issues.
[ to cite ]:
Michael Etgar and Stephen A. Goodwin (1978) ,"Comparative Advertising: Issues and Problems", in NA - Advances in Consumer Research Volume 05, eds. Kent Hunt, Ann Abor, MI : Association for Consumer Research, Pages: 63-71.

Advances in Consumer Research Volume 5, 1978      Pages 63-71

COMPARATIVE ADVERTISING: ISSUES AND PROBLEMS

Michael Etgar, State University of New York at Buffalo

Stephen A. Goodwin, State University of New York at Buffalo

ABSTRACT -

The recent growth in use of comparative advertising has led to increased efforts to measure its effectiveness. Empirical efforts in this area should address seven important research issues. This paper identifies and expounds upon these issues and evaluates recent empirical studies of comparative advertising, particularly with respect to the way they have addressed these issues.

INTRODUCTION

The general research issue involving the communications-effectiveness of comparative advertising as compared to alternative message-presentation styles cannot be resolved until a large number of research questions have been tackled and answered. It is possible that the effectiveness of comparative advertising messages will depend substantially on the specific conditions under which the ad is communicated. In particular, interaction effects need to be identified and measured, and the nature of the advertisement, the message, the promoted product, the compared-to-product(s) and the respondents must be carefully and explicitly specified.

This paper deals specifically with issues that should be addressed when studying and researching the effectiveness of comparative communications vis-a-vis other types of message appeals. In order to accomplish this, we review seven studies which empirically addressed the effectiveness of comparative advertising, our focus being placed upon elicitation of major issues and the extent to which these studies come to grip with such.

All seven studies to be reviewed used an experimental design of one form or another. In these experiments, a group of subjects were exposed to comparative ads while other groups were exposed to some form of a non-comparative communication. The effectiveness of each type of message style was generally measured along a vector of effectiveness variables (typically linked to some part, or all, of a hierarchy-of-effects model) computed from respondents' paper and pencil responses.

In order to facilitate presentation, the issues and research findings from the above-mentioned seven studies are summarized below in Table 1. In this Table, the issues are listed horizontally while the seven studies are listed vertically in approximately chronological order. Cell entries reflect positions taken on an issue and/or whether or not the specific issue was addressed.

TABLE 1

Discussion of the matrix is row-wise, reflecting our primary interest in the issues involved. In each instance, we discuss the research topic involved and then the approaches taken (if any) by specific authors. A critical review of these approaches is presented and in some cases suggestions for improvement are offered.

The issues explored pertain to five major dimensions: the characteristics of the brands which appear in the ad, the audience toward whom the ad is directed, the message design utilized, other factors relating to the ad (media utilized, multiple exposures, etc.), and the methodology used for testing and evaluating ad effectiveness.

PRODUCT CLASS/BRAND CHARACTERISTICS

It is generally accepted among marketing scholars that (regardless of the type of message-style utilized) an advertisement should develop and/or reinforce audience cognitions, attitudes, and intentions which will ultimately lead to trial and (hopefully) repeat purchase of the advertised brand. Whatever the reference point, it would seem obvious that certain characteristics of the brands presented in a communication can have crucial implications for message design. In evaluating the effectiveness of one type of message design versus another, such characteristics should be analyzed.

While in non-comparative (typically referred to as supportive) ads, designers have to be concerned only with the characteristics of the promoted brand, in comparative communications they have to consider also those of the compared-to-brands as well as the dimensions of comparison (e.g., product attributes) to be used in the ad.

Characteristics of the Advertised Brand

Product Typology. A major rationale advanced in favor of comparative advertising is that naming names of specific brands to which the advertised brand is compared, provides the consumer with more factual information and consequently aids him in making more rational brand choices (c.f., Wilkie and Farris, 1975). Consequently, comparative advertising is primarily applicable to messages which compare specific measurable product attributes such as weight, price, speed, etc., and therefore may be especially applicable for products which are primarily bought because of their functional attributes; namely, their performance along measurable dimensions, or their capacity to perform specific measurable tasks.

A large number of products, however, are bought not because of their task-related capacities, but because of the psychological or social benefits which they promise. The ability of an ad to compare and contrast performance of different brands in providing elusive benefits is more restricted. While one can establish or claim with a reasonable degree of credibility (of course, the Substantiation Doctrine of 1972 would apply in this instance; FTC Report, 1972) that one automobile brand, for example, has a better gas mileage performance than another brand, it is difficult to extend the argument and claim that one auto confers more status to its user than another brand.

The effectiveness of comparative advertising may therefore depend on whether the advertised product provides primarily functional, utilitarian benefits or social/ psychological ones. However, none of the seven studies summarized in Table 1 have explicitly considered this issue by integrating it into the experimental designs, or comparing comparative advertising effectiveness for different product classes (differing along a functional-social continuum).

Product Life Cycle Stage. While some researchers used existing brands in their experiments, others relied on new (typically fictitious) brands as the advertised stimulus. The use of existing and well-known brands in an experiment raises the issue of subjects' prior experience with and current attitudes toward such brands. It is important to recognize that such predispositional information should be controlled for. In many cases, consumers have well established perceptions and opinions about specific brands which may have required prolonged promotional effort involving the encoding of numerous messages over substantial periods of time. Not controlling for such "contamination" would tend to complicate inferences drawn from an experimental study.

Additionally, it is reasonable to expect that "one exposure'' experiments (which was the case for all seven published studies) could create insignificant marginal changes in respondents' attitudes and perceptions of well-known brands (in at least a "maturity" stage of the product life cycle), regardless of the message design utilized.

Thus, unless subject predispositional variables are taken into account (e.g., through analysis of covariance or use of a treatment-by-levels type design; Lindquist, 1956), it is preferable to develop messages for new fictitious brands.

Characteristics of The Compared-to-Brands

Naming of Compared-to-Brands. As defined by Wilkie and Farris (1975) a comparative advertisement must either specifically name the compared-to-brand(s) or refer to them in such a way as to leave no doubt regarding what they are. Additionally, the comparison must involve specific attributes of the brands.

If we utilize Wilkie and Farris' definition as a reference point, then any indirect comparison should be general so as to preclude consumer recognition of a specific comparison (e.g., Brand A lasts longer than other leading brands, as opposed to Brand A lasts longer than the leading seller).

The so-called "supportive" or non-comparative ad may also be viewed, at the most general level, as a comparative ad. Even where there is absolutely no allusion to a competitive offering, where the message communicates only information about the promoted brand, is it not fair to say that the consumer is likely to mentally make a comparison with available and known alternatives? Thus, we have a continuum of message approaches which include truly comparative communications, Brand X ads, and supportive ads. In operationalizing communication styles, care must be taken to ensure that the experimental ads truly reflect the underlying conceptual structure. For example, in the Prasad (1976) study, it is extremely likely that a number of subjects knew which brand was the "leading brand" of movie cameras. If this assumption is tenable, then Prasad has not, in fact, manipulated comparative versus Brand X ads. In effect both conditions presented comparative communications (particularly for knowledgeable respondents).

In reviewing the seven studies, we see that there is some doubt as to the existence of isomorphic relationships between concept and variable; and this is particularly the case for the so-called Brand X advertisement.

Market Position of Compared-to-Brand(s). The market position(s) of the compared-to-brand(s) are important to consider because such position(s) may affect the consumer's frame of reference as it relates to the evaluation of the brand promoted. One strategy commonly found is to compare the promoted brand to the leading brand in the pertinent generic product class though its superiority is far from established. One assumed advantage of using a leading brand for comparison is that it can potentially reduce misidentification of the promoted brand and misunderstanding of the corresponding message as well as attract the attention of current users of the leading brand and thus, at a minimum, lead to increased likelihood of greater awareness for the advertised brand.

Other researchers disagree. Kershaw and Tannenbaum (1976) argue that where both the promoted and the compared-to-brands are relatively unknown to the audience, the viewers may confuse the brands, possibly leading to the paradoxical situation whereby a comparative ad improves consumer attitude toward (and ultimately sales of) the compared-to-brand.

Still other opinions reflect the belief that use of market leaders for comparison may be a futile exercise in increasing awareness. While a greater number of consumers might be attracted to the ad, these people may be much less prone to change their attitudes and purchase behavior patterns due to their brand loyalty; further, such brand loyal consumers may be more likely to disbelieve the message claims made by the "new" (or small, in terms of market share) brand. Perhaps, then, anchor brands should be other than the market leader.

So far, none of these arguments has been validated by actual research. Even though Golden (1976) and Mazes (1976) have manipulated the market position of the sponsored brand, to date, none of the published studies have manipulated the market position of the compared-to-brands. Clearly this area may be very important vis-a-vis practical marketing strategy implementation.

Similarity of Brands. In many generic product classes, subgroupings exist. For example, even though economy, luxury and sports cars belong to the same general product class of automobiles, consumers may use different evaluative criteria and hold different evoked sets in each group.

Comparison claims between brands from different subgroups may be tenuous. There are two points to be made in this context; first, consumer perceptual/attitudinal evaluation of ads which make "dissimilar comparisons" maybe more negative than for comparisons involving brands likely to be included in the typical evoked set; second, the legal ramifications of such a strategy are not clear. While the claims made may be literally true (e.g., VW has more trunk space than Cadillac Coupe de Ville, better gas mileage than Mercedes 220 SL, etc.) the impression made may be false. Thus, the advertiser considering a comparative message appeal between dissimilar items runs the risk of FTC involvement on the basis of advertising deception. None of the published studies have dealt with this issue. It merits empirical inquiry.

Number of Brands Compared. Some comparative ads include only one compared-to-brand; others use several. On an a priori basis, there is little behavioral evidence to indicate which approach should be more effective, if any. However, it may be reasonable to expect that a substantial increase in the number of brands compared can add to communication noise (e.g., information overload), while a modest comparison set may improve the credibility of the message. The bulk of the reported studies utilize only one brand for comparison. The exception is McDougall (1976) who used two. Experimental manipulation of this variable has not, in any case, been operationalized.

PRODUCT ATTRIBUTES ON WHICH BRANDS ARE COMPARED

Number of Attributes Presented

Comparative ads found currently in the media reflect a diversity of approaches. While some restrict discussion to one product characteristic, many make comparisons on multiple dimensions, believing that more information lends additional credibility concerning the claim (or in order to be consistent perhaps with a particular positioning strategy which management wishes to invoke). The issue of course is to find out, other things being equal, the differential effects (if any) due to presentation of differing amounts of attribute information.

Perhaps the "information overload" (Jacoby, et al., 1974) framework which suggests that consumer evaluation of ad and brand advertised conforms to some kind of inverted u-shaped function of amount of information provided in the communication is useful here.

If this argument is tenable, a comparative ad which uses one or a small number of attributes as comparison points may have more "positive" effects than similar appeals making a greater number of attribute comparisons; or, perhaps an intermediate amount of brand characteristic information would be "best". Most of the studies in Table 1 have used one or two product attributes for comparison: two (Prasad, 1976; and McDougall, 1976) have used four characteristics. None of these studies, however, have operationalized an "amount of attribute information" variable and attempted to compare the impact of change in this variable on communicative effectiveness.

Salience of Brand Attributes

The effectiveness of a comparison between two or more brands on a specific attribute dimension in terms of consumer evaluative response may depend very much on the relative importance (perception of salience) of the attribute for the targeted audience. If the discussed characteristics provide important benefits to the consumer, then the comparison may be interesting and informative, and likely to make a positive impact. Consumers who are only marginally interested in the brand attribute(s) discussed may not be influenced at all (or even negatively influenced).

Consequently, pretesting the salience of brand attributes should be a basic initial step in any research design for testing effectiveness of comparative communications. Yet none of the published studies reported in Table 1 have selected product attributes according to such a pretest; therefore, it may be that, for the subject population studied, marginal or non-salient attributes were used for comparison. Golden (1976) did measure product attribute importance in her study, but apparently only after selection for the experiment. She found that selected attributes were rated as important; but she had no way of knowing if others, not included, were relatively more important.

RESPONDENTS

Type of Respondents

Researchers studying effectiveness of comparative advertising have employed two distinct groups of subjects: students and female heads of households (housewives). Use of adult housewives has generally been lauded due to the ostensible increase in external validity of the study results. Furthermore, it may be likely that housewives are more concerned with product classes which have been utilized in the various studies, and consequently are more involved in the task presented them regarding assessment of ad and brand advertised. Still, use of students as subjects may be entirely acceptable and consistent with improved external validity if product classes advertised are salient to them (i.e., they purchase and consume brands from such product classes). Two of the studies in Table 1 (Ogilvy and Mether, 1975; and McDougall, 1976) have used female heads of households as subjects; the others have used students. Only Wilson (1976) and Prasad (1976) explicitly measured product/brand relevance explicitly.

Knowledge of the Product Class/Brands

In order to derive evaluative conclusions from a message presented by a comparative communication (or any type of appeal, for that matter), the consumer will draw upon knowledge gained about the compared-to-brands. Certainly, lack of such knowledge will have some impact on the evaluation process; at the extreme, evaluations of the comparison claims may be meaningless to a particular viewer heretofore unknowledgeable about the product class and/or brands promoted and compared. Of the studies reviewed, only Wilson's checked for degree of prior consumer knowledge.

Usage Rates. A useful surrogate for product knowledge, perhaps, is usage rate which details the extent to which the consumer has been buying or using various brands from the pertinent product class(es). Usage rates were not incorporated in any of the studies in Table 1.

Brand Loyalty. Related to usage is the notion of brand loyalty. Consumers who exhibit substantial brand loyalty to compared-to-brand(s) in comparative communications may be likely to resent the comparative claim(s), perhaps in response to cognitive dissonance. Whether or not, in principle, this cognitive response is general, the point remains that respondents should be measured on this basis to permit an examination of such effects. As with the above suggested measures, brand loyalty measurements permit increased control of error variation, and can be utilized in experimental designs via covariance analysis or directly in the form of a leveling variable.

Brand loyalty was incorporated in two studies: Golden (1976) and McDougall (1976). Golden measured brand loyalty by subjects' reported relative frequency of purchase of the advertised and compared brands, and used it as a covariate in an analysis of variance design. McDougall measured brand loyalty by assessing subjects' degree of agreement with a semantic differential statement with endpoints "loyal to the (specific) brand" and "different brands are purchased", or "don't buy". He then compared the means of responses regarding subjects' opinions as to the reliability and helpfulness of the ads between loyal and non-loyal subjects, by using a simple t-test.

It should be mentioned, too, that Prasad (1976) measured his subjects regarding the number who already (prior to the test) preferred the compared-to-brand (in this instance, Kodak XL movie camera); he found, via nonparametric tests, that there was no evidence of "selective recall" between Kodak-preferred and non-Kodak-preferred groups.

Personality Traits

While we do not advocate use of general personality inventories (Kassarjian, 1971), we do feel that certain care- fully selected measures may shed additional light on consumer response to comparative vis-a-vis non-comparative communications. In particular, those traits which pertain to the ability and motivation of the consumer's processing of information varying in terms of uncertainty, ambiguity, etc., are of some interest. While none of the six authors incorporated personality trait measures in the external designs a couple can be mentioned here. The Mehrabian-Russell Arousal-Seeking Tendency Scale (Mehrabian and Russell, 1974) which measures the general tendency or proneness of curiosity-seeking/variety- seeking behavior is one example. Consistent with a point mentioned by Wilkie and Farris (1975) that comparative ads are likely to be perceived as more novel than "ordinary'' supportive message appeals, we should find that consumers high on arousal-seeking tendency would evaluate comparative ads more positively than those consumers "low" on arousal-seeking tendency. A second example is Kelman and Cohler's (1967) test for "cognitive style". This test identifies two major styles regarding consumer's characteristic mode of resolving cognitive uncertainty: clarifiers (people who react "positively" to uncertainty by seeking clarification, additional information, etc.) and simplifiers (people who simplify their environments by keeping out obtrusive cognitions, and tend to avoid uncertainty at "all costs"). Assuming that the information encoded in comparative communications may arouse the cognition of uncertainty or ambiguity (at least as compared to a more traditional supportive communication), it may be reasonable to suppose that clarifiers will explore comparative communications more thoroughly, and generally evaluate them more favorably (perhaps be affected more by them), than would be the case for simplifiers.

OTHER DESIGN CONSIDERATIONS

Medium

The studies presented in Table 1 cover a good range of media-vehicles. Several experiments invoked mock print advertisements, others used radio and/or television (video-taped) ads. McDougall (1976) suggests that for comparative advertising, print media may have substantial advantages over audio-visual media because comparative advertising includes relatively more information (data not only about the promoted brand but also about additional brands) and correspondingly more complex messages, and thus may be communicated more effectively via the print medium where the consumer can spend more time with the message, and do so at his/her own pace. Audiovisual approaches automatically restrict message-consumer interaction time.

Two of the summarized studies (Mazis, 1976; and Ogilvy and Mather, 1975) have operationalized their communications via audio-visual or only audio vehicles. The remaining studies used print medium only. Mazis also compared print versus radio media. In this study he found insignificant differences on major response variables between message modalities, shedding some light on the previously mentioned opinion that comparative advertising is likely to be more "effective" in print rather than audiovisual media. The issue is still open, however, for future study.

Number of Ads Presented

Some of the researchers exposed their subjects to only one advertisement and compared the results between conditions using essentially factorial experimental designs. Others exposed subjects to several ads, all of which were studied (Wilson, 1976) or only the one which was tested (e.g., Prasad, 1976; here the rationale of using a portfolio approach was to create a more "natural" format and simulate a "real life" exposure situation).

It is possible that the portfolio approach inadvertently creates unnecessary demand characteristics; subjects' attention is spread out, a memory set is likely to be aroused, etc. It is important to come to grips with the potential trade-off between "real-life" simulation and confounding demand characteristics. (We recognize that if the primary response domain pertains to memory, a portfolio approach is virtually mandatory: more on this in the last major section of the paper.)

Frequency/Intensity of Exposure

Cognitive response to advertising may be affected by the frequency of message exposure. Repetition effects (Sawyer, 1973; Ray and Sawyer, 1971) deserve careful study, but to date the experiments published have invariably been single-exposure. It is difficult to assess those studies that have focused upon recall (aided or unaided) measures in evaluating comparative versus non-comparative ads when only one single exposure has taken place. Replication designs ought to be considered by future researchers; furthermore, recall/memory measures should be taken at realistically spaced time intervals rather than immediately (or shortly) after the advertisement exposure. All of the studies measuring recall suffer from this potential problem, although Prasad (1976) readministered his measures to subjects one week after his single-exposure.

OTHER MESSAGE CHARACTERISTICS

Message characteristics which may affect cognitive, affective, or behavioral responses to communications include the issues of substantiation (factual or subjective claims), sidedness of the message, and the relative position of the comparative claims within the advertisement.

Substantiation

Several critics of comparative advertising (Kershaw, 1976, Chevins, 1975) have suggested that when the comparisons include generalized, non-substantiated statements, the consumer may become misinformed as well as confused. Claims of superiority, for example, which are not supported by independent (or company) tests may reduce the credibility of the message. McDougall (1976) tested for the relative effectiveness of use or non-use of substantiated claims across various message appeals. Using separate regressions for comparative substantiated, noncomparative substantiated, comparative unsubstantiated, and non-comparative unsubstantiated messages, with measures of ad reliability and helpfulness as dependent variables, McDougall found that, overall, substantiated claims receive greater acceptance in terms of reliability and helpfulness; non-comparative substantiated claims were found to be superior to all of the others. Golden (1976) also manipulated claim substantiation. She found no differences in message-appeal effect.

Wilson's study (1976) also bears on this issue in that all of his ad manipulations involved subjective (unsubstantiated) claims. He found that unsubstantiated comparative ads were significantly perceived as more confusing and less believable than their unsubstantiated non-comparative counterparts.

One-Sided versus Two-Sided Appeals

In his working paper, Mazis (1976) provides an interesting perspective regarding sidedness of message appeals. Drawing upon work done by Hovland, et al., (1949), and integrating McGuire's Inoculation Theory (1961), Mazis presents a conceptual framework for assessing message sidedness with respect to comparative or non-comparative advertising. Specifically, Mazis points out that comparative communications can be constructed along conventional lines by mentioning only positive comparisons in the advertised products favor (one-sided comparative ad), or alternatively by mentioning both positive and negative facts about the promoted brand (two-sided comparative ad).

From the managerial perspective, it may be the case that such communications are not only more believable, but that the overall image of the brand (and perhaps the company sponsoring it) is enhanced due to the "refreshing" glimpse at honesty. It is plausible to assume that most consumers would applaud an attempt at "telling it like it really is" rather than only mentioning positive benefits accruing to brand consumption. From the public policy point of view, such two-sided comparative ads may provide even more objective information upon which the consumer may make a brand choice. This is entirely consistent with the stance taken by the FTC.

Operationally, of course, there is always the chance that such a communication problem may win an ad agency an Effie, but lose the company precious sales and market share. It deserves careful study. Mazis initiates such a study in his second experiment reported in the above-mentioned working paper. The results are clouded, unfortunately, by an unrealistic operationalization of the negative appeal in the two-sided communication; nonetheless, he has made an important first step in studying the interesting, and potentially useful (managerially and public policy-wise) message-appeal strategy.

Relative Positioning of the Claim(s)

In rounding out this section, we mention that effectiveness of comparative advertising communications may depend, in part, on the positioning of the comparative claims (e.g., at the beginning, the middle or end of the ad). None of the seven studies explored this issue, although admittedly it is a minor point in contrast to some of those previously discussed. Also related to this "recency-primacy" issue is the question of whether the compared-to-brand(s) might be mentioned first or last in the message.

GENERAL ISSUES PERTAINING TO TESTING AND METHODOLOGY

Independent Variables

Manipulations of message appeals require close attention to proper operationalizations. Accordingly, scrutinization of the isomorphism between concept and variable is imperative. Where correspondence is considerably less than 1 to 1, the researcher is not in a position to make conclusive inferences with respect to the response effects. Thus, it is important that the researcher fully understands what is meant by "comparative advertisement", "Brand X" ad, and "supportive" communication so that reasonable working definitions can be operationalized. The same is true whatever independent variable is selected for study.

We have seen, for example, that there is some doubt as to whether or not Prasad's (1976) "Brand X" ad was, in fact, truly perceived as a Brand X ad or rather a standard comparative ad (at least for some of the subjects in his study). As a general rule, it is advisable to thoroughly pretest these manipulations, utilizing different subjects from the same population. Such pretesting is the exception rather than the rule in published studies.

Dependent Variables

Without getting into the debate as to what constitutes reasonable measures of advertising effectiveness, we can say that perhaps the majority of advertising academics and certainly practitioners are comfortable with some sort of hierarchy of effects model (e.g., Lavidge and Steiner, 1961).

In such models the consumer is viewed as responding to advertising stimuli according to a hierarchy such as "awareness", "comprehension", "interest", "liking", "intention'', and "behavior". As pointed out by Lavidge and Steiner (1961), such a portrayal of response to stimuli covers three stages: cognitive - affective - and conative (behavioral). The advertising researcher, over the years, has developed a variety of measures which more or less tap dimensions of each of these three stages. Depending upon the purpose of the ad, whether copy pretesting or market post-testing is the primary focus, combinations of these measures are gathered.

The seven studies summarized have utilized all or part of such hierarchical responses. Some (Prasad, 1976; Mazis, 1976; Ogilvy and Mather, 1975) stressed memory or recall ability occasioned by the various manipulated advertisements. Such measures reflect the researcher's concern for assessing the cognitive dimension. Others utilized a battery of questions designed to shed light on the entire spectrum, with the exception of memory measures (Wilson, 1976).

However, both evaluations of the ad itself and the brand advertised should be measured. Some of the studies looked at one or the other (e.g., Golden, 1976; McDougall, 1976), but not both.

Pretests and Manipulation Checks

It is extremely important that the researcher verify that his/her manipulations are in fact being perceived in a manner which is consistent with his/her objectives. This is true of independent variable operationalizations as well as dependent measures (scales) administered and gathered.

Thus, pre-tests/manipulation checks are crucial to ensure internal validity. While such checks can be made after the experiment is completed by using the same subjects, it may be more advisable to perform them before the experiment by using different subjects drawn randomly from the same population from which the experimental respondents will be selected. Such a pre-test strategy mitigates against likely demand characteristics operating (Sawyer, 1975). Thus, use of scale which is not discriminating will yield statistically insignificant re-suits when, in fact, such results may exist. Golden's study suffers in this respect. All her subjects regardless of experimental condition appeared to generally use the same small range on her intention scale. Perhaps the scale was not explained adequately; perhaps an ll-point subjective probability scale should have been utilized instead of the 7-point scale which was used; perhaps the independent variables were operationalized poorly.

Another related point concerns the nature of the measurement level of the scales used to tap dependent responses. Unless such scales are thoroughly examined and pre-tested so as to permit an assumption that they are interval- scaled, it is difficult to argue in favor of parametric analytical techniques. And, since parametric techniques (e.g., analysis of variance, regression) permit assessment of interaction effects (whereas assessment of such by use of non-parametric techniques is not clear cut), it behooves the researcher to strive for development of dependent measures which can be assumed to be at least interval scaled. This requires the discovery that subjects perceive distances between each number in a given scale as reflective of approximate equal psychological distances.

Covariates

When a researcher does not collect predispositional measures which presumably relate to informational processing of advertising stimuli, error variance may be unnecessarily large, resulting in non-significant differences.

To reduce error variation (or adjust for extraneous sources of error), the researcher might design experiments which include important predispositional variables (e.g., treatment-by-levels designs ((also referred to as randomized block designs)); see Lindquist, 1956; or Winer, 1971) as one (or more) of the factors. Alternatively, he might consider the use of such predispositional variables as covariates, where any contaminating influences may be statistically adjusted for. The Golden study is illustrative of this approach: she gathered a measure of brand loyalty which she used to statistically adjust responses to treatment combinations. Clearly, assessment of main and interaction effects (as well as the various multiple comparisons which may be of interest) can be clouded by failure to reduce reasonable sources of confounding error variation (e.g., product class usage experience, brand loyalties, evoked sets, related personality traits).

SUMMARY AND CONCLUSIONS

The purpose of this paper has been to explore the various issues involved in assessing the communications effectiveness of comparative advertising. Using seven recent studies as reference points, we have endeavored to point out the complexities of comparative advertising research. No matter what the scope of an individual study, there are several crucial considerations necessary. These involve the advertised product class (brand) characteristics (e.g., product class, salience, life cycle stage), the characteristics of the compared-to-brand(s) (e.g., market position(s), number of brands compared), product class (brand) attributes on which comparisons are made (e.g., number of attributes and their salience), the subjects (e.g., type, usage rates, brand loyalty, personality traits), the advertisement design (e.g., media-vehicle utilized, frequency/intensity of ad exposure), other message characteristics (e.g., use of substantiation, sidedness of appeals), and methodology used (e.g., experimental designs, manipulation checks, covariates).

The seven studies addressed only some of the research questions identified in this paper. Notably, there is an important gap in that the bulk of the studies failed to experimentally manipulate (or control error variation via covariance analysis) major product-, subject-, and situational-related variables. While it is true that exploration of all research questions raised in this paper are beyond the scope of any individual study, we suggest that future research should address the specific issues presented.

Careful attention to the identified issue areas should contribute to the development of body of knowledge which could then provide answers as to when, if, and how comparative advertising can be used.

REFERENCES

A. G. Chevins, "A Case for Comparative Advertising," Journal of Advertising, 4:2(1975), 31-36.

Federal Trade Commission Report. "Staff Report to The Federal Trade Commission on The Ad Substantiation Program Together with Supplementary Analysis of The Submissions and Advertisers' Comments," Committee on Commerce, U.S. Senate, Washington, D.C., July, 1972.

L. L. Golden, "Consumer Reactions to Comparative Advertising,'' in B. B. Anderson, Advances in Consumer Research, Volume III (Chicago: Association for Consumer Research, 1976), 63-67.

C. I. Hovland, A. A. Lumsdaine and F. D. Sheffield, Experiments on Mass Communication (Princeton: Princeton University Press, 1949).

J. Jacoby, D. E. Speller and C. A. Kohn, "Brand Choice Behavior As A Function of Information Load," Journal of Marketing Research, 11(1974), 63-69.

H. H. Kassarjian, "Personality and Consumer Behavior: A Review," Journal of Marketing Research, 8(November, 1971), 409-418.

H. Kelman and R. Cohler, "Test for Cognitive Style," (Reported and adapted) in D. Cox, Risk-Taking and Information Handling in Consumer Behavior (Cambridge: Harvard University Press, 1967).

A. G. Kershaw and S. I. Tannenbaum, "For and Against Comparative Advertising," Advertising Age, (July 5, 1976), 25-28.

R. J. Lavidge and G. A. Steiner, "A Model for Predictive Measurements of Advertising Effectiveness," Journal of Marketing, (October, 1961), 59-62.

E. F. Lindquist, Design and Analysis of Experiments (Boston: Houghton Mifflin, 1956).

M. B. Mazis, "A Theoretical and Empirical Examination of Comparative Advertising," Unpublished paper, University of Florida, 1976.

G. H. G. McDougall, "Comparative Advertising: An Empirical Investigation of its Role in Consumer Information," Working Paper #167, School of Business Administration, The University of Western Ontario, London, Canada, December, 1976.

W. McGuire and D. Papageorgis, "The Relative Efficacy of Various Types of Prior Belief-Defenses in Producing Immunity Against Persuasion," Journal of Abnormal and Social Psychology, 62(May, 1961), 327-337.

A. Mehrabian and J. A. Russell, An Approach to Environmental Psychology (Cambridge: MIT Press, 1974).

Ogilvy and Mather. "The Effects of Comparative Television Advertising That Names Competing Brands," Research Report, 2 East 48th Street, New York, 1975.

V. K. Prasad, "Communications-Effectiveness of Comparative Advertising: A Laboratory Analysis," Journal of Marketing Research, 13(May, 1976), 128-137.

M. L. Ray and A. G. Sawyer, "A Laboratory Technique for Estimating the Repetition Function for Advertising Media Models," Journal of Marketing Research, 8(February, 1971), 20-29.

A. G. Sawyer, "The Effects of Repetition of Refutational and Supportive Advertising Appeals," Journal of Marketing Research, 10(February, 1973), 23-33.

A. G. Sawyer, "Demand Artifacts in Laboratory Experiments in Consumer Research," Journal of Consumer Research, 1 (March, 1975), 20-30.

W. L. Wilkie and P. W. Farris, "Comparison Advertising: Problems and Potential," Journal of Marketing, 39 (October, 1975), 7-15.

R. D. Wilson, "An Empirical Evaluation of Comparative Advertising Messages: Subjects' Responses on Perceptual Dimensions," in B. B. Anderson, Advances in Consumer Research, Volume III (Chicago: Association for Consumer Research, 1976).

B. J. Winer, Statistical Principle in Experimental Design, 2nd edition (New York: McGraw-Hill, 1971).

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