Construct Validation and Empirical Testing of Guilt Arousing Marketing Communications

ABSTRACT - An experiment was undertaken to assess the construct validity of "aroused guilt" and determine the message effects of guilt arousing marketing communications Findings suggest the validity of a guilt construct and the ability of persuasive communications to arouse guilt and affect self-attitudes. Effects on global attitudes and behavioral intentions were not evidenced.

Citation:

Morry Ghingold and Lorne Bozinoff (1982) ,"Construct Validation and Empirical Testing of Guilt Arousing Marketing Communications", in NA - Advances in Consumer Research Volume 09, eds. Andrew Mitchell, Ann Abor, MI : Association for Consumer Research, Pages: 210-214.

Advances in Consumer Research Volume 9, 1982      Pages 210-214

CONSTRUCT VALIDATION AND EMPIRICAL TESTING OF GUILT AROUSING MARKETING COMMUNICATIONS

Morry Ghingold (student), The Pennsylvania State University

Lorne Bozinoff, The Pennsylvania State University

ABSTRACT -

An experiment was undertaken to assess the construct validity of "aroused guilt" and determine the message effects of guilt arousing marketing communications Findings suggest the validity of a guilt construct and the ability of persuasive communications to arouse guilt and affect self-attitudes. Effects on global attitudes and behavioral intentions were not evidenced.

INTRODUCTION

Although advertising and promotion remain as sources of considerable interest within consumer behavior research and theory, the topic of emotion arousing persuasive communications has not received much recent attention by consumer behavior researchers. However, a large body of multidisciplinary research has been reported in the past, with most of the work focusing on fear arousing communications. Regrettably, the results to date remain equivocal. (For more complete discussions on the course and content of fear appeal research and some of the problems encountered see Higbee 1969, Ray and Wilkie 1970, and Sternthal and Craig 1974)

Recently, however, a new type of emotional appeal has been suggested as being plausible, relatively distinct, and in current use in the marketplace. Ghingold (1981) and Ghingold and Bozinoff (1981, in press) have commented on the existence and impact of guilt arousing marketing communications. They have proposed that 1) guilt arousing communications are being used by advertisers, 2) that it may be possible to theoretically account for the efficacy of guilt arousing marketing communications, and 3) that guilt arousing communications do, in fact, arouse guilt and have effects on message recipients.

It would seem that one fruitful avenue for research in this area might include efforts to validate newly proposed constructs. Although Ghingold and Bozinoff (1981, in press) have reported experimental evidence which indicated that ads intended to arouse guilt did so, three questions remain unresolved:

1) What is a guilt arousing communication and is it distinct from other forms of persuasive communication?

2) Are the affective responses (i.e., emotional arousal) to guilt arousing communications unique and predictable?

3) Do guilt arousing communications "work," i.e., what are the effects of attending to a guilt appeal?

As a means of addressing some of these questions a study was conducted with guilt arousing marketing communications. The background, description and results of this research are reported below.

BACKGROUND AND PURPOSE OF THE PRESENT RESEARCH

In the domain of marketing communications and persuasion guilt may be considered to be an a posteriori emotional response which occurs subsequent to a particular action or thought. Once aroused, guilt feelings may lead to a variety of actions, cognitive or behavioral, intended to reduce the level of guilt one experiences to a more tolerable level (Ghingold 1981). In contrast, fear results from the perception of some danger and brings about the objective of preventing or avoiding the undesired outcome, i.e.. an a priori response. Since guilt is brought about by the actual consequences of behavior and fear is caused by the anticipated consequences of behavior, theoretically there should be differences in these two forms of anxiety although grey areas of overlap may occur (Roseman 1979, Ghingold 1981).

Few investigators within consumer behavior/marketing have addressed guilt arousing persuasive communications. Wheatley and Oshikawa (1970) investigated the relationship between positive and negative advertising appeals and anxiety. Applying a general anxiety reduction model, the authors concluded that "negative communication can be effective in inducing the behavior advocated by the communicator (pg. 85)." Regrettably, a fear manipulation constituted the only negative communication investigated.

Other researchers have also demonstrated a similar emphasis on fear research in studying guilt arousing communications. Haefner (1956), Zemach (1966) and Yinon et al. (1976) reported significant effects due to guilt arousing communications, yet these researchers were applying principles from the fear paradigm dominant at the time. Notwithstanding any methodological or analytical limitations of these studies, it remains unclear whether the theories or models tested were existing fear paradigms or unique conceptualizations of guilt.

More recently, Kelman (1979) has discussed the relationship between various emotional states and subsequent changes in attitude. Citing dissonance theory, Kelman suggested that guilt, as an emotional response, was among the most likely moral dilemmas to lead to attitude change, although he did not frame his conceptualization in the context of persuasive communications. Extending this view, Ghingold (1981) argued in favor of the relevance of guilt appeals. Defining guilt as a psychological construct in a persuasion context, Ghingold proposed an initial conceptualization of guilt arousing persuasion. (For a more detailed review of past research on guilt appeals and greater explication of both the guilt construct and a proposed guilt persuasion motel, see Ghingold 1981).

Ghingold and Bozinoff (1981, in press) offered the first empirical study on guilt appeals research in the consumer behavior literature. They presented findings of an investigation on the persuasive effect of guilt arousing marketing communications, reporting that guilt could in fact be aroused in subjects and that the aroused guilt had cognitive consequences on certain post-exposure attitudes.

Given this rather hesitant beginning, guilt appeal research in consumer behavior remains very limited. Notwithstanding Ring's (1981) call for more research in the area, to date little is truly known about guilt arousing communications and persuasion. Therefore, to extent the limited existing findings concerning guilt arousing communications the present study was undertaken.

The goals of the present research were of an exploratory nature. Validation of a full model of guilt arousing persuasion was not undertaken. Rather, present interest was more narrowly focused on two issues of theoretical and practical significance. First, does a communication, designed to be guilt arousing, in fact arouse feelings that reflect guilt and not other emotions or affects? Second, given that guilt has been aroused, what, if any, are the effects on consumer cognitions or behavior (e.g., attitudinal and behavioral effects)?

These research questions have been formulated into the following hypotheses which are to be tested in this study.

H1: The construct "aroused guilt" is measurable, unique and distinguishable from other affective construct

H2: Guilt can be predictably aroused in experimental subjects with marketing communications.

H3: Message relevant attitudes can be favorably influenced through the use of persuasive guilt arousing communications.

H4: Message relevant behavioral intentions can be favorably influenced through the use of guilt arousing communications.

Thus, the study is an effort to establish construct validity for the abstraction "aroused guilt" and determine the relationship between attention to a guilt arousing persuasive communication and message effects.

METHOD

Subjects

A total of ninety-one subjects were recruited from various undergraduate marketing courses. During regular class sessions, students were invited to participate in a study of advertising. Participation was voluntary, although all of the students invited agreed to participate. The sample was approximately 60% male.

Since the principal aim of the present research is the detection of significant relationships between variables of theoretical significance, the use of a convenience sample of students is defensible (Sternthal, Dholakia and Leavitt 1978). In cases where individual differences are not of theoretic interest, homogeneous samples (such as students in particular classes at a particular institution) are appropriate for theory-oriented research (Kruglanski 1975). Rather,interest is centered on the causal relationship between the operationalizations of the independent variables (i.e., internal validity). Hence, the use of a student sample in the present research permits the testing of inferences regarding the levels of guilt aroused by particular marketing communications and the persuasive effects of such communications.

Procedure

Subjects in all classes were randomly assigned booklets containing one of five print ads and twenty-six questions. Four of the ads, which solicited donations to a charitable organization to benefit overseas underprivileged children were designed to partially span the range of guilt arousing communications. Two ads were selected from a group of advertisements appearing in national news magazines which were rated on "guilt arousal" by five independent judges. One ad was judged to induce a low level of guilt and the other ad was judged to arouse a high level of guilt. The experimenters then modified the two ads in an effort to increase the level of guilt aroused by the high guilt and reduce the level of guilt aroused by the low guilt ad. This made a total of four ads, two low guilt arousing ads and two high guilt arousing ads. A fifth ad, which was for ar airline, served as a control ad.

Although it may be argued that university students are not the target population for the advertisements used in the study, for the purposes of this experiment the stimuli arc appropriate. As noted above, present interest is focused on the causal relationship between the independent and dependent variables, i.e., internal validity. Since the cognitive structures of students should be as prone to guilt arousal as those of other segments of the population, the stimuli selected should allow for the testing of inferences regarding aroused guilt and its consequences.

The first page of the booklet contained several Likert type questions dealing with present attitudes and behavior. These questions served as premeasures. The advertisement followed the premeasures. After fully reading the ad, subjects completed the remaining questions in the booklets.

Measurement Instruments

Immediately after reading the advertisement, the subjects were asked to indicate how well a series of 13 adjectives described their present feelings. Four of the adjectives were guilt related, taken from Haefner's (1956) list of guilt indices. Three other affective domains (annoyance, joy, fatigue), were included in the list based on items t taken from clusters reported by Nowlis (1965). These domains were chosen because they spanned a wide range of emotional states. These measures served two purposes. First, they offered the opportunity of a test of convergent and discriminant validity. Second, the guilt related adjectives allowed for a manipulation check, and served in part as a measure of aroused guilt.

The next series of questions were self-attitude measures relevant to the communication (e.g., I have a clear conscience regarding the plight of the underprivileged). These self-dispositions also served as measures of aroused guilt.

The final series of questions contained measures of more general or deep rooted attitudes towards charities and donating behavior, as well as behavioral intention measures relevant to the content and recommendations of the communication (e.g., I plan to donate to charitable organizations more than I have in the past).

RESULTS

Construct Validation: Tests of Convergent and Discriminant Validity

The validity of a guilt construct was assessed by Campbell and Fiske's (1959) multitrait matrix. (Multiple methods were not available.) As noted above, the 4 adjectives designed to measure guilt were embedded in a list of 13 adjectives. The remaining 9 adjectives were designed to measure elation, fatigue and aggression (Nowlis, 1965). There were 3 adjectives designed to measure each of these three affective domains. Table 1 presents the complete correlation matrix for these thirteen adjectives. In order to show convergent validity, the correlations between the four guilt adjectives should be significantly greater than zero (Campbell and Fiske, 1959). As reported in Table 1, the correlations between the four guilt adjectives (shown within the triangle in Table 1) are all significantly greater than zero. The second component of construct validity, discriminant validity, can be demonstrated by showing that the correlations between the guilt adjectives are significantly greater than the correlations between the guilt adjectives and the other adjectives. The correlations between the guilt adjectives (shown within the triangle in Table 1) are almost without exception greater than the correlations between the guilt adjectives and the other adjectives (shown in the rectangLe in Table 1).

As an additional test of construct validity, the multitrait matrix was factor analyzed. If there is a distinct guilt construct, the four guilt adjectives should all load highly on the same factor (Huba and Hamilton, 1976). Table 2 reports that the four guilt adjectives do load highly on the same four factor solution (VARIMAX rotation). These findings provide preliminary support for the first hypothesis. Guilt does appear to be a separate and distinct construct in this experimental setting.

TABLE 1

MULTITRAIT MATRIX

TABLE 2

VARIMAX ROTATED FACTOR MATRIX

Message Effects of Guilt Arousing Marketing Communications

A series of univariate and multivariate analyses of variance (ANOVA, MANOVA) were performed on the data. During the first analysis tests were made for significant differences between treatment groups with respect to the premeasures. No significant differences were found between the treatment groups for any of the premeasures in both the univariate and multivariate analyses. This suggested that the randomization procedure was successful.

Measures of Guilt Arousal

Two sets of measures were used as indices of guilt arousal. One set was comprised of the guilt related adjectives in the adjective list and the other set consisted of three measures of individual self-attitudes relevant to the content of the communication (e.g., the ad makes me feel partly to blame for the plight of the poor). In the univariate analysis three of the four guilt related adjectives showed significant differences between treatment groups (p < .02), with the high guilt treatments yielding higher measures of aroused guilt. Overall multivariate tests also indicated highly significant differences between the treatment groups based on a linear composite of the four measures (p < .0022). (Table 3 summarizes these findings.) Analysis of the self-attitude measures indicated significant group differences in both the univariate and multivariate cases. Two of the three univariate analyses were at or near significance (p < .07) and the overall multivariate tests combining the measures were highly significant (p ' .003). Again, the high guilt treatments yielded mean score differences in the expected direction. (Table 4 summarizes these findings.) It should be noted, however, that the efforts to modify the original stimuli were not entirely successful. In fact, for the guilt arousal adjectives the results of the modifications were often opposite from what was expected. Nonetheless, the overall pattern of the results, from control to low treatments (averaged) to high treatments (averaged) do provide some support for the second hypothesis although more testing is clearly in order.

TABLE 3A

GUILT AROUSAL ADJECTIVES  -  CELL MEANS (5 POINT SCALES)

TABLE 3B

UNIVARIATE ANALYSES OF VARIANCE FOR GUILT AROUSAL ADJECTIVES

TABLE 3C

MULTIVARIATE ANALYSIS OF VARIANCE FOR GUILT AROUSAL ADJECTIVES

TABLE 4A

MEASURES OF GUILT AROUSAL  -  CELL MEANS (5 POINT SCALES)

TABLE 4B

UNIVARIATE ANALYSES OF VARIANCE FOR SELF-ATTITUDE MEASURES OF GUILT AROUSAL

TABLE 4C

MULTIVARIATE ANALYSIS OF VARIANCE FOR SELF-ATTITUDE MEASURES OF GUILT AROUSAL

Measures of General Attitudes Toward Charities and Donating Behavior

Three measures were included to assess post-exposure attitudes toward charities and donating behavior. No between group differences were found in either univariate or multivariate analyses. It was evident that there were no meaningful changes in global attitudes following processing of the guilt arousing communication. Although differences in self-blame and other guilt related self-attitudes were observable between the treatment groups, the more global (and possibly value-bound) attitudes were not affected by the experimental manipulations. As a result, the third hypothesis could not be confirmed.

Measures of Behavioral Intention

Three measures were used to assess behavioral intentions relevant to the ad content. The measures included a general reactive measure (i.e., I would like to do something about the way I feel about...), a measure of anticipated future donating behavior, and a measure of interest in obtaining further information on the underprivileged. Of the three measures, only the general reactive measure showed significant differences between treatment groups (p < .03). Overall multivariate tests approaches statistical significance (p < .07) but the patterns of group means were irregular and less predictable than in prior analyses. As a result, the findings for this section of the analysis were inconclusive and it is not possible to confirm the fourth hypothesis of meaningful treatment group differences for the measures of behavioral intentions.

DISCUSSION

Four major findings emerge from this study. First, there is preliminary evidence of an "aroused guilt" construct which is separate and distinct from several other constructs. A set of adjectives were developed which operationalized this construct and which can now be used in future research in this area. Second, it was shown that guilt can be aroused by persuasive communications. More importantly, it was shown that guilt can be aroused in a predictable manner. Those ads designed to arouse the least guilt did so, while those ads designed to arouse the most guilt also did so. It was further shown that the level to which guilt can be aroused is sufficient to change message-relevant self-perceptions or self-attitudes. This means that guilt arousing communications can be designed which create guilt feelings that surpass the individual's guilt tolerance threshold and therefore result in certain changes in cognitive structure.

Third, it was also shown that guilt could not be aroused for the extent that general or deep rooted attitudes and behavioral intentions will change. This finding should not be totally unexpected. It may be unrealistic to expect a single presentation of a guilt arousing communication or for that matter, any persuasive communication, to impact upon broad based attitudes or behavioral intentions. A more realistic test would require multiple presentations, ideally over an extended period of time.

Another reason for the lack of any changes in general attitudes or behavioral intentions may be due to the sample used. It may be unrealistic to expect university students to change behavioral intentions regarding the plight of underprivileged children until they have started to raise children of their own. It may also be unrealistic to expect students who tend to have limited financial resources to favorably change their attitudes toward charities and donating behavior or increase their intentions to donate to charities, due to guilt arousal. Rather, students with aroused guilt emotions can be expected to feel some guilt but simply cannot change their behavior because of financial considerations.

A further explanation for the failure to find any general attitude or behavioral intention changes in light of the significant guilt arousal may be that the subjects engaged in some mediating cognitive responses such as counterarguing. A wide array of counterarguments are possible when one is confronted by an advertisement soliciting charitable donations. Examples may include the arguments that there are poor people in the United States who should be helped first, that students cannot afford to help others, that most of the money donated never reaches the intended recipients, or that the underprivileged are responsible for their own plight.

It is appropriate at this point to consider two possible limitations of the experimental design which may impact on the above conclusions. Firstly, by having subjects examine advertisements and then rate their own guilt feelings there is the danger of a strong demand artifact. Although this may have taken place, the rival hypothesis of a demand artifact would not account for the higher levels of guilt indicated by the subjects in the high treatments (on average) as compared to the low treatments (on average) or the control treatment. Furthermore, although the subjects in the control treatment were given the same measurement instrument, in the vast majority of the cases they indicated the least aroused guilt.

Another point of interest is the low coefficient of power evident in this experiment. Although this is a limitation from a design point of view, it should be noted that as power decreases it may take larger differences in means of the dependent measures to show significance. Hence, one may choose to trade off power (or sample size) for the ability to detect truly meaningful differences in the operationalizations of the dependent variables. These limitations, however, have been recognized by the experimenters.

IMPLICATIONS

At the theoretical level, several implications flow from these results. First, given that another emotional appeal besides the fear appeal has been found, the question of how many other emotions such as envy, anger and so forth can be used in marketing communications is raised. Second, the generalizability of the findings must be explored. It was suggested that other populations may have a greater propensity to change their behavior due to guilt arousal. Third, the issue of individual differences remains unanswered. It is entirely possible that there may be individual differences in susceptibility to guilt arousing communications. This has obvious segmentation ramifications. In addition, another unexplored area is the extent to which these findings can be generalized to other products and services. At the managerial level, this study suggests that advertising copy writers may have a new general strategy with which to work. Given ads can be purposefully designed to create guilt feelings which can on a single presentation change self-attitudes, guilt arousal should now be thought of as a potential copy strategy. What remains unanswered are the ethics of such a strategy. Whether advertisers should have the right to arouse emotions (negative or positive) in the public mind remains a deeply philosophical and troublesome issue.

REFERENCES

Campbell, D. T. and Fiske, D. W. (1959), "Convergent and Discriminant Validation by the Multitrait-Multimethod Matrix," Psychological Bulletin, 56 (2), 82-105.

Ghingold, Morry (1981), "Guilt Arousing Communications: An Unexplored Variable," in Advances in Consumer Research, Vol. VIII, Kent B. Monroe, ed., Ann Arbor, MI: Association for Consumer-Research, 442-448.

Ghingold, Morry and Bozinoff, L. (1981, in press), "The Persuasive Effect of Guilt Arousing Marketing Communications: A Preliminary Investigation," Proceedings of the 1981 Administrative Sciences Association of Canada Conference.

Haefner, D. P. (1956), Some Effects of Guilt Arousing and Fear Arousing Persuasive Communications on Opinion Change, unpublished technical report.

Higbee, K. (1969), "Fifteen Years of Fear Arousal: Research on Threat Appeals, 1953-1968," Psychological Bulletin, 72 (December), 426-444.

Huba, G. J. and Hamilton, D. L. (1976), "On the Generality of Trait Relationships: Some Analyses Based on Fiske's Paper," Psychological Bulletin, 83 (5), 868-876.

Kelman, H. C. (1979), "The Role of Action in Attitude Change," paper prepared for the 1978-79 Nebraska Symposium on Motivation, Department of Psychology, University of Nebraska, Lincoln.

King, R. L. (1981), "Consumer Response to Advertising: A Discussion of Three Studies," in Advances in Consumer Research, Vol. VIII, Kent B. Monroe, ed ., Ann Arbor, MI: Association for Consumer Research, 449-451.

Kruglanski, A. (1975), "The Two Meanings of External Validity," Human Relations, 28, 653-659.

Nowlis, Vincent (1965), "Research With the Mood Adjective Check List," in S. S. Tomkins and C. E. Izard, eds., Affect. Cognition and Personality, New York: Springer Publishing Company.

Ray, M. L. and Wilkie, W. L. (1970), "Fear: The Potential of an Appeal Neglected by Marketing," Journal of Marketing, 34 (January), 54-62.

Roseman, I. (1979), "Cognitive Aspects of Emotion and Emotional Behavior," paper presented at the 87th Annual Convention of the American Psychological Association, New York City.

Sternthal, B. and Craig, C. S. (1974), "Fear Appeals: Revisited and Revised," Journal of Consumer Research, 1 (December), 22-34.

Sternthal, B., Dholakia, R. and Leavitt, C. (1978), "The Persuasive Effect of Source Credibility: Tests of Cognitive Response," Journal of Consumer Research, 4 (March) 252-260.

Wheatley, J. J. and Oshikawa, S. (1970), "The Relationship Between Anxiety and Positive and Negative Advertising Appeals," Journal of Marketing Research, 7 (February), 85-89.

Yinon, Y., Bizman, A., Cohen, S. and Seger, A. (1976), "Effects of Guilt-Arousal Communications on Volunteering to the Civil Guard: A Field Experiment," Bulletin of the Psychonomic Society, 7 (6), 493-494.

Zemach, M. (1966), The Effects of Guilt-Arousing Communications on Acceptance of Recommendations, unpublished doctoral dissertation, Yale University.

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