The Risky Shift in Ad Hoc and Natural Consumer Groups: a Test of the Polarization Hypothesis and a Majority Rule Explanation

Peter H. Reingen, University of South Carolina
ABSTRACT - Group-induced change processes were examined in two experiments with the use of consumer choice dilemmas which elicited initially risky response tendencies in the subject population. In Experiment I, the prediction based on the polarization hypothesis that group discussion would lead to a shift in the risky direction was investigated in both ad hoc and natural groups, and this hypothesis was largely confirmed. Experiment II examined the impact of skewness on the risky shift in consumer groups. Skewness was manipulated in four conditions and was found to have a significant effect on shifts. However, contrary to a majority rule interpretation, significant cautious shifts failed to occur when the distributions were negatively skewed.
[ to cite ]:
Peter H. Reingen (1977) ,"The Risky Shift in Ad Hoc and Natural Consumer Groups: a Test of the Polarization Hypothesis and a Majority Rule Explanation", in NA - Advances in Consumer Research Volume 04, eds. William D. Perreault, Jr., Atlanta, GA : Association for Consumer Research, Pages: 87-92.

Advances in Consumer Research Volume 4, 1977   Pages 87-92

THE RISKY SHIFT IN AD HOC AND NATURAL CONSUMER GROUPS: A TEST OF THE POLARIZATION HYPOTHESIS AND A MAJORITY RULE EXPLANATION

Peter H. Reingen, University of South Carolina

ABSTRACT -

Group-induced change processes were examined in two experiments with the use of consumer choice dilemmas which elicited initially risky response tendencies in the subject population. In Experiment I, the prediction based on the polarization hypothesis that group discussion would lead to a shift in the risky direction was investigated in both ad hoc and natural groups, and this hypothesis was largely confirmed. Experiment II examined the impact of skewness on the risky shift in consumer groups. Skewness was manipulated in four conditions and was found to have a significant effect on shifts. However, contrary to a majority rule interpretation, significant cautious shifts failed to occur when the distributions were negatively skewed.

INTRODUCTION

The concept of perceived risk in consumer behavior was introduced by Bauer (1960), who pointed out that any action of a consumer produces social and economic consequences that cannot be anticipated with certainty. Thus, consumers tend to develop risk-handling strategies to cope with t-he hazards of buying. One such reactive strategy--an effort to reduce risk--is repeat purchasing of the same brand; i.e., brand loyalty (Cunningham, 1965). A proactive alternative is to seek additional information with regard to a decision problem (Cox and Rich, 1964). The second alternative is of particular importance in the present study.

Among the many alternatives for seeking and obtaining additional information, the notion that a consumer's peer and/or reference group serve as important sources of information and influence has received strong support (Arndt, 1967). Word-of-mouth influence appears to be especially potent when consumers face a risky choice (Perry and Lamm, 1969). Therefore, it would appear that consumers, when faced with a risky choice, may initiate some form of informal group discussion to obtain social support and legitimation for their purchase choices. Despite its relevance for research on consumer word-of-mouth behavior in general, and household decision-making (e.g., Foote, 1961) as well as group influence on innovative consumer behavior (e.g., Robertson, 1974) in particular, it is surprising, however, that very little consumer research has been directed to the possible effects of informal groups on a consumer's willingness to take greater or lesser risks under perceived risk conditions. Many findings from social psychology suggest that consumers might be willing to take greater risks after informal group discussion (e.g., Wallach and Kogan, 1965). From a consumer research point of view, Woodside (1972) concluded that this may indeed be the case. However, a more detailed examination of the relevant literature reveals that post-discussion shifts in the risky direction are not the only exception to the averaging tendency in groups observed in conformity paradigms. For example, it is clear that cautious shifts occur in some risk-taking situations (Reingen, 1973; Reingen, 1974; Stoner, 1968) and that group discussion apparently produces shifts in individual judgments and attitudes that have nothing to do with the risk-caution dimension (Home and Long, 1972; Myers and Bishop, 1970; Reingen, 1976a). Thus, the risky shift should be viewed as a special case of a more general group-induced choice shift.

A popular empirical generalization arising from the risk-relevant as well as risk-irrelevant choice shift literature is the so-called group polarization effect: namely, that the average post-group response will tend to be more extreme in the same direction as the average of the pre-group responses (Myers and Lamm, 1976). In terms of shifts along the risk-caution dimension this generalization suggests that group discussion tends to lead to an increase in individual risk taking--a "risky" shift--on items where the predominant initial (pre-discussion) tendency is in the risky direction, and tends to enhance caution--a "cautious" shift--on items which tend to elicit initially cautious responses. Thus, the major purpose of the present study is to extend the polarization generalization to group influence on consumer risk taking with the use of consumer stimuli which pretesting suggested would elicit initially risky response tendencies. Experiment I investigates discussion-induced shifts in economic and socio-psychological risk taking in both ad hoc (i.e., assemblies of strangers) and natural (i.e., assemblies of mutual friends) groups. The latter-type groups are conspicuously absent in choice shift research, and they were included to increase the external validity of findings based on ad hoc groups. Experiment II seeks to verify a group decision rule explanation of the polarization tendency in groups.

EXPERIMENT I

Method

Subjects. Subjects were 97 female undergraduate students at the University of Cincinnati who were paid for their participation.

Materials. The risk-taking instrument utilized in the present study included four consumer buying situations. The instrument's overall structure was similar to Kogan and Wallach's (1964) choice dilemma questionnaire employed in previous risky shift investigations. Each situation presented two alternatives (X,Y), one of which (Y) is riskier than the other but also, if successful, more rewarding. Subjects indicated the lowest chance of success they would consider acceptable before recommending the risky alternative (Y) to a hypothetical friend on a scale ranging from 10 chances in 100 to 90 chances in 100 of success, plus an additional category (scored as 100) where the subject could refuse the risky alternative altogether. One situation was the following:

D., your friend, would like to buy a boot holder. While at a department store, she saw a holder (X) made out of plastic material which covers the boots and hangs them upside down to assure keeping their shape. An open bottom gives ventilation and dries the boots between uses. Out of season, the boot holder provides dust-free storage. This boot holder (X) was priced at $2.55. Right next to boot holder (X) was a holder (Y) on display which apparently had the same product characteristics as holder (X), but which was priced at $1.76. Although D. could not detect any noticeable differences between the two boot holders, she suspected that the difference in price may be due to a difference in quality. However, the salesperson at the store whom D. knew very well assured D. that boot holder (Y) was of very good quality.

The subjects were then requested to check the lowest level of chance of success (i.e., that boot holder Y will be of very good quality) that they would consider acceptable for D. to buy boot holder (Y). The other situations (Friend F, G, and M) as Well as data documenting the instrument's validity and reliability are provided in full detail in Reingen (1974). [Throughout the instrument, written explanations of the meaning of the risk-caution scale were given to the subjects. Findings by Reingen (1976b) and Reingen and Kernan(1977) suggest that standard instructions accompanying choice dilemma items are inadequate to assure the subjects comprehension of the task.] Two of the four situations (Friend D and F) represented mainly economic risk (ER), while the other two (Friend G and M) represented primarily socio-psychological risk (SPR). [In this study, economic risk was defined as potential undesirable financial consequences and socio-psychological risk as potential undesirable social consequences and psychological harm that may be associated with an incorrect purchasing decision. For data supporting the validity of the partitioning of situations on the basis of risk-type see Reingen (1974).]

Other materials included a series of trait-type scales for equivalence verification of experimental and control subjects, and a post-experimental questionnaire designed to obtain an indication of subjects' demand awareness. [Scales utilized were for situation specific self-confidence, independence, extraversion, achievement, anxiety and rigidity. Based on previous findings, these traits were selected on the basis of their correlation with risk-taking behavior.]

Experimental design and procedure. The standard risky shift design and procedures were employed. As far as the ad hoc groups were concerned (40 subjects), the subjects in the experimental condition (33 subjects) first completed the risk-taking instrument individually. For each situation, they were allowed three minutes to make their decisions. The sequence of the situations in the instrument, which was presented as an "Opinion Questionnaire," was varied systematically across subjects. After the subjects completed the instrument, they were scheduled for a subsequent experiment--de-scribed as a decision-making experiment--four to seven days after the pretest. Upon arrival for the experiment, the subjects were randomly assigned to three-person groups where they sat in chairs arranged in a circle, discussed the situations for a maximum of seven minutes each, and arrived at a consensus decision for each situation. After this, the subjects once more responded privately to the instrument, and they individually completed the various personality scales and the post-experimental questionnaire. Seven of the 40 subjects were randomly selected to serve in the control condition. These subjects responded to the risk-taking instrument twice, four to seven days apart. After the retest, the control subjects completed the personality scales and the post-experimental questionnaire. The various instructions that were utilized were those by Wallach and Kogan (1965).

The subjects comprising the natural groups (a total of 57 subjects) came either from a large dormitory or two sororities. The dormitory subjects responded to advertisements that had appeared on bulletin boards. Groups of friends were recruited by contacting one member of this subject pool and requesting her to bring two mutual friends of the same sex to the experimental sessions. The sorority subjects responded to a questionnaire which asked them to list friends who were members of their sorority. To obtain groups of three mutual friends of the same sex, the data supplied by the sorority subjects were used to establish a sociometric choice matrix. From this matrix, groups of mutual friends were formed, and they were subsequently scheduled for the experimental sessions. Of the total number of 19 groups of three mutual friends each, three were randomly selected to serve as the control condition. While otherwise the same procedure as described above was utilized, the subjects in the natural group case were of course not randomly assigned to groups.

Computation of shift scores. For analyzing the results of the discussion conditions, a group pre-to-post shift score was computed for the ER-situations and SPR-situations by calculating for each group the arithmetic mean per item for the initial Opinion Questionnaire, and subtracting the pre scores from the post scores. A group pre-to-consensus shift was calculated for each group by taking the arithmetic mean of the consensus scores and subtracting the pre scores from the consensus scores. For both shift scores, a negative value indicates a shift in the risky direction.

Results

Validity checks. Insignificant differences between pretest scores for the experimental and control subjects were obtained for the ER-situations (ad hoc: t = 1.37, p >.15; natural: t = .22, p > .80) and SPR-situations (ad hoc: t = .28, p > .75; natural: t = .52, p > .60). Thus, the assignments of subjects to experimental and control conditions resulted in equivalent initial risk-taking scores between conditions. Furthermore, no significant differences in initial risk taking were found between ad hoc and natural group subjects (ER-situations: F = 1.18, p > .05; SFR-situations: F < 1). This rules out the possibility that differences in mean shifts between ad hoc groups and natural groups can be attributed to differences in initial risk taking. [Similarly, no significant differences (p > .05) in personality scores between experimental and control subjects and between ad hoc and natural group subjects were found. However, one natural group had to be eliminated from the shift score analysis due to likely demand awareness. The post-experimental questionnaire data revealed also that the ad hoc groups were indeed groups of strangers. None of the subjects indicated that a friend was in their group. The subjects in the natural groups had been friends for 19 months on the average. With one exception, they indicated that they planned to stay friends.]

Shift analysis. Table 1 shows the means of (individuals') pretest, group consensus, and (individuals') posttest scores for each type of risk and type of discussion group. Regardless of type of risk or group, the average pretest scores were clearly in the risky direction of the risk-caution scale. Thus, according to the polarization hypothesis, group discussion should enhance this initial tendency and a risky shift should materialize.

In the ad hoc group case, the average group consensus scores were riskier than the corresponding pretest decisions for the ER (mean shift = -11.67, t = 3.63, p < .003) and SPR-situations (mean shift = -9.69, t = 4.60, p < .001). Similarly, the individuals' posttest scores were riskier on the average for the ER (mean shift = -10.91, t = 3.32, p < .005) but only marginally riskier for the SPR-situations (mean shift = -3.94, t = 1.37, p < .10). Insignificant shifts in test-retest scores were found for the control subjects (p > .10).

As far as the natural groups were concerned, significant mean pre-to-group consensus (mean shift = -9.89, t = 3.51, p < .002) and pre-to-post risky shifts (mean shift = -8.22, t = 3.21, p < .005) were found for the ER-situations. In terms of the SPR-situations, the average pre-to-group consensus shift (mean shift = -4.34) was marginally significant (t = 1.48, p < .08), but the pre-to-post shift in the risky direction was insignificant (mean shift = -3.11, t = .96, p > .15). Again, changes in test-retest scores for the control subjects were found to be insignificant (p > .10).

When mean shifts between ad hoc and natural groups were compared, no significant differences (p > .10) in mean pre-to-group consensus and pre-to-post shifts were found. [Since no specific differences in mean shifts between types of groups were hypothesized, all tests were two-tailed. Otherwise, in the experimental conditions, the tests were one-tailed.]

TABLE 1

PRETEST, CONSENSUS, AND POSTTEST MEAN SCORES FOR AD HOC AND NATURAL GROUPS AND RISK TYPE

DISCUSSION

Generally speaking, the above results are in agreement with the predictions derived from the polarization hypothesis. The group consensus decisions were riskier than the individuals' pre-discussion decisions regardless of type of risk or type of group. However, only for economic risk situations were the individual post-discussion decisions significantly riskier than the pre-discussion ones in both types of groups. This seems to suggest that discussion-induced enhancement of initially risky positions is inhibited when the situations represent primarily socio-psychological risk and that, therefore, risky shifts may not be equally pervasive across all types of risk. Nevertheless, while the results not always coincided with commonly accepted levels of statistical significance, they were unequivocally in the predicted direction. Thus, the present findings are viewed to substantiate the major empirical generalization arising from the choice shift literature that group interaction tends to enhance choice tendencies which are initially preferred in the subject population. What is particularly important about the present study is that, generally speaking, the effects obtained, regardless of type of group. The results of this study suggest that generalization of findings from groups of strangers to groups of friends may not he as potentially invalid as was previously believed (e.g., Cartwright, 1971).

From the viewpoint of consumer behavior research, the findings imply that informal word-of-mouth processes enhance consumer risk taking when the predominant initial tendency is in the risky direction. True, a natural skepticism surrounds shifts discovered in a hypothetical situation where group members are not confronted with the consequences of their decisions. Yet Runyan's (1974) work suggests that generalization of such phenomena to "real" problems is far from invalid. Also, although this study's findings apply to group-induced shifts in risk positions--and hence imply nothing about individual vis-a-vis group risk propensities--the results are quite applicable to the typical consumer psychology scenario where an individual forms an opinion which is then subjected to group influence. This scenario, which has been alluded to by several investigators (e.g., Woodside and DeLozier, 1975), should be monitored by those who seem to accept the "two-step flow" as an article of faith if they would have the dynamics of the process work to their benefit. Several strategies can be used by persuaders, but they must first recognize what works under which conditions. Robertson (1974), for example, offers correlational evidence on the relationships among risk perception, communication proclivity and innovation proneness--evidence to which this study lends experimental substantiation.

While Experiment I extended the empirical generality of the polarization phenomenon to risk taking in consumer settings, it did not provide an adequate explanation of this phenomenon. Experiment II investigates the possibility that the choice shift is the result of skewed distributions of subjects' initial choices.

EXPERIMENT II

A variety of explanations of the polarization phenomenon have been tested. However, while the evidence produced by the literature on the effects of group discussion on individual choice generally supports a group polarization interpretation, consensus among investigators concerning the specific mechanisms by which these effects are obtained has yet to materialize. The theoretical controversy surrounding choice shifts has focused on three theories: Interpersonal comparison, informational influence, and group decision rule explanations (Myers and Lamm, 1976).

Interpersonal comparison theories argue that the choice shift is due to normative processes. Subjects have a tendency to present themselves favorably in comparison to relevant others. Exposure to others' positions during group interaction leads an initially moderate subject to abandon his initial choice for one which will make him at least as extreme as most other members so as to maintain his image of social desirability. Detailed versions of these approaches appear in Cartwright (1971) and Pruitt (1971). Informational influence explanations attribute shifts to persuasive arguments that are generated during group discussion and that favor the initially preferred alternative. The shifts in choice following discussion are suggested to result from persuasive arguments that were only partially available to the average subject prior to discussion. More thorough presentations of informational influence explanations are presented by Vinokur and Burnstein (1974) and Myers and Lamm (1976). Group decision rule explanations suggest that social decision schemes for combining individual choices predict shifts. While a variety of possible social decision schemes have been outlined by Davis (1973), most prominent has been the suggestion of majority rule (e.g., Cartwright, 1971; Lamm, Trommsdorff and Rost-Schaude, 1973; Reingen, 1974). Due to skewness in the distribution of initial choices (Fraser, Gouge and Billig, 1971; Home, 1970), some group members find themselves in the minority. A shift in the group mean is created when the minority members move toward agreement with the majority opinion. A more complete review of this approach can be found in Myers and Lamm (1976).

Despite its substantial intuitive appeal and positive empirical evidence (Lamm, et al., 1973; Moscovici and NFve, 1973; Reingen, 1974), the explanation of majority rule has been largely abandoned in favor of more-substantiated theories, especially the ones based on informational influence. Central to this negative evaluation of majority rule has been the critical finding that skewness cannot account for group polarization. With choice dilemmas (Kogan and Wallach, 1964) one would expect positively skewed distributions (i.e., groups with an initially risky majority) to yield a risky shift and negatively skewed distributions to produce a cautious shift. Contrary to these predictions, Wallach and Mabli (1970) obtained a risky shift when they created three-person groups so that each group would have a cautious majority. As Cartwright (1971) observes, however, this study may not provide an adequate test of the skewness issue, since the groups were formed on the basis of total scores for ten items of the Kogan and Wallach (1964) choice dilemma questionnaire. An analysis for each item would appear to have been necessary. A more serious challenge comes from a study by Vinokur (1969) who manipulated the skewness of the within-group distributions of initial choices such that groups were either positively or negatively skewed. While skewness had an effect on shifts, it could not, by itself, account for the risky shift phenomenon. The groups in the negative skewness condition (i.e., those with a cautious majority) shifted toward risk which is evidence against the majority rule viewpoint. However, a more detailed investigation of the effects of skewness should be informative (Cartwright, 1971). Experiment II investigates the skewness issue using different stimulus materials and a similar but more complete design than Vinokur's (1969). Thus, the present investigation tests for the generalizability of choice dilemma findings, which is important for choice-shift researchers; and it provides a more detailed study of the majority rule interpretation.

More specifically, Experiment II allowed for a more effective manipulation of skewness. In the Vinokur (1969) study, the mean (absolute) skewness was .78 which does not rule out a constellation of initial scores of his five-member groups of, for example, 1,5, 5,5, and 5 chances in 10 (absolute skewness = .80). Although in this hypothetical example the scores show a relatively more cautious majority, the median value of 5 chances in 10 clearly does not indicate the group's majority leaning in the cautious direction of the scale. In fact, it has been suggested that for choice dilemmas the psychological midpoint of the risk-caution scale is probably not 5 chances in 10, or even 6 chances in 10 (Myers and Aronson, 1972). Therefore, based on the polarization generalization, one would expect such a group to shift in the risky direction despite the negative skewness in initial choices. Since the mean of initial choices of the subjects in Vinokur's (1969) negative skewness condition was 5.18, it is unlikely that many of his groups had similar constellations of pretest scores. Nevertheless, the present study completely avoided this potential problem by constructing groups whose majorities clearly favored either the risky or cautious pole of the scale. In addition, the manipulation of skewness was stronger in two of four conditions such that the effects of skewness had a maximum chance to occur.

In order to test the hypothesis that skewness can account for group polarization, groups with different distributions were formed according to the subjects' initial risk-taking positions. In the first two experimental conditions, groups were either strongly or slightly positively skewed with a majority taking greater risks. In the third condition, the initial choices of the individuals comprising a group were symmetric with zero variance, while in the fourth condition, the groups were strongly negatively skewed. If the hypothesis is valid, risky shifts should occur in the first two conditions and the risky shift should be greatest in the condition of strongly positively skewed groups. No shift should be found in the third condition but a cautious shift should be expected in the strongly negatively skewed groups condition. Thus, direction as well as magnitude of shifts are hypothesized.

Method

A total of 42 male graduate business students at Iona College volunteered for participation in the experiment. [The results in Experiment I are based on female subjects. However, comparable shifts for both male and female subjects have been found in choice dilemma research (Pruitt, 1971).] The subjects were run in three groups of 15, 15, and 12.

Subjects were given a questionnaire which contained the four consumer-related choice dilemmas for which they were asked to indicate the minimum level of odds. They were given three minutes for each situation. After the subjects completed their initial choices, the questionnaires were collected and the subjects were given a duchy task for the next few minutes. During that time, the initial data for the first situation were processed and groups of three persons each were formed based on the subjects' initial risk-taking positions. Groups with strong positive skewness make up Experimental Condition I (majority mean = 30 chances in 100), those with slight positive skewness Experimental Condition II (majority mean = 30 chances in 100), those with identical initial risk-taking choices Experimental Condition III, and the ones with strong negative skewness Experimental Condition IV (majority mean = 65 chances in 100). Skewness was computed by subtracting the median of the initial individual choices from their mean. Positive skewness reflects a relatively more risky majority, while negative skewness implies a relatively more cautious majority. After the determination of groups for the first situation, they were placed far apart in a large room to prevent inter-group interaction. The groups were instructed to discuss the first situation (of which they had received a new form), to reach a consensus decision during a seven-minute time span, and to indicate the group decision on the new form. If the subjects in a group were unable to reach a consensus decision within the prescribed time limit, they were instructed to mark their individual decisions. Upon termination of the time limit, all subjects were reformed into new groups (which were determined in the meantime) for the next situation. The same procedure was repeated until the last situation was completed. [Since the individual post-discussion decisions tend to correspond closely to the group's decision, no individual post-test was included in Experiment II.] After the experiment the subjects were debriefed and dismissed.

In total number of three-person groups that were formed in the course of the experiment was 56. Of this total number of groups, two had to be eliminated because of a misunderstanding of the task, two groups were unable to reach a consensus decision, and seven groups were formed which could not be assigned to either one of the four conditions. [Data for these seven groups were not included in the analysis.] A successful attempt was made to establish as many groups as possible on the same situation across the experimental conditions (X2 = 3.82, p > .05). Further, all the groups that were included in Experimental Conditions I and IV were matched as closely as possible on the basis of their absolute magnitude of skewness.

Group shifts were computed by subtracting the mean of the initial risk-taking choices from the group consensus decision, or from the mean of the subjects' decisions that followed the group discussion in the two cases where the groups were unable to reach a consensus decision.

Results

The experimental results are summarized in Table 1. Analysis of variance rejected the null hypothesis of equal mean shifts across the experimental conditions (F = 17.59, p < .001). Specifically, significant shifts toward risk occurred in Experimental Condition I (strongly positive skewed groups) (X = -13.59, t = 5.17, p < .001) and Experimental Condition II (slightly positively skewed groups)(X = -5.50, p < .005), while no shift was found in Experimental Condition III (symmetric groups). The risky shift in Experimental Condition I (strongly positively skewed groups) was significantly greater than the one in Experimental Condition II (slightly positively skewed groups) (t = 2.09, p < .05). These results were as expected. However, in Experimental Condition IV (strongly negatively skewed groups), the hypothesize shift in the cautious direction was insignificant (X = 2.05, t = .40, p > .10).

TABLE 2

INITIAL MEAN RISK-TAKING SCORES, MEAN SKEWNESS AND MEAN SHIFT

Does skewness account for any proportion of the variance of the risky shift? A significant product-moment correlation (r = -.56, p < .001) between skewness and shift for all the groups in the four experimental conditions suggests that skewness does influence shifts. However, since there is a difference between the means of the initial individual risk-taking choices of the four conditions (F = 12.29, p < .001), it could be argued that the difference in shifts between the conditions is due to this difference rather than to the manipulation of skewness. However, partialling out mean initial individual risk taking yields a correlation of -.69 (p < .001), which largely rules out this interpretation. Hence, the effect of skewness of shifts accounts for up to 48% of the variance of group shifts, and it can be concluded that skewness does influence shifts. [In the Vinokur (1969) study, skewness could account for only 9% of the variance of group shifts. The higher variance accounted for in the present study may likely be attributable to the stronger manipulation of skewness.]

Discussion

The following results were consistent with the hypothesis that skewness can account for the risky shift. Significant shifts in the risky direction were found for groups which had an initially more risky majority (positively skewed groups), with the strongest risky shift occurring in that condition where groups had the greatest positive skewness. No shift was observed in the case where the groups were symmetric with zero variance regarding the subjects' initial risk-taking choices. However, incongruous with the hypothesis was the insignificant shift in the predicted cautious direction for groups which had an initially cautious majority (strongly negatively skewed groups). Since the strongly positively and strongly negatively skewed groups were virtually identical in terms of absolute skewness, the shifts should have been in opposite directions and of similar magnitude. How can one explain this inconsistency?

First, regression toward the mean cannot be completely ruled out; but, this is an unlikely possibility since regression cannot account for the risky shift found in the condition of strongly positively skewed groups. Regression should have affected both conditions equally.

Secondly, the lack of significance of the cautious shift found in the strongly negatively skewed groups condition may have been due to a balancing effect. A post hoc analysis of shift scores on a group basis indicates a minority domination in six of thirteen groups in this condition (i.e., a risky shift occurred in these groups). Findings suggest that people who initially take greater risks are more confident in their positions (Burnstein, 1969) and are perceived by their fellow group members as possessing greater ability (Jellison and Riskind, 1971). This may have translated itself in these groups into greater minority assertiveness and influence during group discussion.

Nevertheless, while skewness as in Vinokur's (1969) study was found to have an overall significant effect on shifts, the results in the final analysis are unfavorable to a majority rule interpretation. Significant cautious shifts failed to occur despite the ideal condition of strong negative skewness in the data. Thus, the present results are in high agreement with Vinokur's (1969).

CONCLUSIONS

Experiment I provided substantial directional support to the proposition that group discussion tends to enhance choice tendencies which are initially preferred in the subject population. For both ad hoc and natural consumer groups, discussion-induced enhancement of initially risky positions was observed. These consumer risky shifts were especially strong and significant when the stimuli represented primarily economic rather than socio-psychological risk.

Experiment II examined a skewness explanation of this polarization tendency in consumer groups. The overall hypothesis that consumer risky shifts can be explained by the shifting of a group's more cautious minority toward the initially riskier positions of the majority--thereby creating a shift in the group mean--received equivocal support. While skewness had an effect on shifts, significant cautious shifts failed to occur when the distributions were negatively skewed.

REFERENCES

J. Arndt, Word of Mouth Advertising (New York: Advertising Research Foundation, 1967).

R. A. Bauer, "Consumer Behavior in Risk Taking," Proceedings, American Marketing Association, 1960.

E. Burnstein, "An Analysis of Group Decisions Involving Risk," Human Relations, 22 (1962), 381-395.

D. Cartwright, "Risk Taking by Individuals and Groups: An Assessment of Research Employing Choice Dilemmas," Journal of Personality and Social Psychology, 20 (1971), 361-378.

D. F. Cox and S. U. Rich, "Perceived Risk and Consumer Decision Making--the Case of Telephone Shopping," Journal of Marketing Research, 1 (1964), 32-39.

S. M. Cunningham, "Perceived Risk as a Factor in Product-Oriented Word-of-Mouth Behavior: A First Step," Proceedings, American Marketing Association, 1965.

N. N. Foote, Household Decision Making (New York: University Press, 1961).

J. H. Davis, "Group Decisions and Social Interaction: Theory of Social Decision Schemes," Psychological Review, 80 (1973), 97-125.

C. Fraser, C. Gouge and M. Billig, "Risky Shifts, Cautious Shifts, and Group Polarization," European Journal of Social Psychology, 1 (1971), 7-30.

W. C. Home, "Group Influence on Ethical Risk Taking: The Inadequacy of Two Hypotheses," Journal of Social Psychology, 30 (1970), 237-238.

W. C. Home and C. Long, "Effect of Group Discussion on Universalistic-Particularistic Orientation," Journal of Experimental Social Psychology, 8 (1972), 236-246.

J. M. Jellison and J. Riskind, "Attribution of Risk to Others as a Function of Their Ability," Journal of Personality and Social Psychology, 20 (1971), 413-415.

N. Kogan and M.A. Wallach, Risk Taking: A Study in Cognition and Personality (New York: Holt, Rinehart and Winston, 1964).

H. Lamm, G. Trommsdorff and E. Rost-Schaude, "Group Induced Extremization: Review of Evidence and a Minority-Change Explanation," Psychological Reports, 33 (1973), 471-484.

D. G. Myers and S.J. Aronson, "Enhancement of Dominant Risk Tendencies in Group Discussion," Psychological Reports, 30 (1972), 615-623.

D.G. Myers and G.D. Bishop, "Discussion Effects on Racial Attitudes," Science, 169 (1970), 778-779.

D.G. Myers and H. Lamm, "The Group Polarization Phenomenon,'' Psychological Bulletin, 83 (1976), 602-627,

S. Moscovici and P. NFve, "Studies on Polarization of Judgements, III. Majorities, Minorities and Social Judgements," European Journal of Social Psychology, 3 (1973), 479-484.

M. Perry and C.B. Hamm, "Canonical Analysis of Relation Between Socioeconomic Risk and Personal Influence in Purchase Decisions," Journal of Marketing Research, 6 (1969), 351-354.

D.G. Pruitt, "Choice Shifts in Group Decisions: An Introductory Review," Journal of Personality and Social Psychology, 20 (1971), 339-360.

P. H. Reingen, "Risk Taking by Individuals and Groups with the Use of Industrial Product Purchasing Situations As Stimuli," Journal of Psychology, 85 (1973), 339-345.

P. H. Reingen, "Choice Shifts Along a Dimension of Risk Tested with Consumer Related Stimuli," Ph.D. dissertation, University of Cincinnati, Cincinnati, Ohio, 1974.

P. H. Reingen, "Effect of Group Discussion on Honesty-Dishonesty in Business Related Situations: A Social Comparison Test," Working paper, Center for Marketing Studies, College of Business Administration, University of South Carolina, Columbia, South Carolina, 1976(a).

P. H. Reingen, "Do Subjects Understand the Choice Dilemma Questionnaire," Journal of Social Psychology, 1976, forthcoming (b).

P. H. Reingen and J.B. Kernan, "Risky Shift or Confused Subjects," Journal of Social Psychology, 1977, forthcoming.

T. S. Robertson, "Group Characteristics and Aggregate Innovative Behavior: Preliminary Report," in J.N. Sheth, ed., Models of Buyer Behavior (New York: Harper and Row, 1974), 310-327.

D. L. Runyan, "The Group Risky Shift Effect As a Function of Emotional Bonds, Actual Consequences, and Extent of Responsibility," Journal of Personality and Social Psychology, 29 (1974), 670-676.

A. Vinokur, "Distribution of Initial Risk Levels and Group Decisions Involving Risk," Journal of Personality and Social Psychology, 13 (1969), 207-214.

A. Vinokur and E. Burnstein, "The Effects of Partially Shared Persuasive Arguments on Group Induced Shifts: A Group Problem Solving Approach," Journal of Personality and Social Psychology, 29 (1974), 305-315.

M.A. Wallach and N. Kogan, "The Role of Information, Discussion, and Consensus in Group Risk Taking," Journal of Experimental Social Psychology, 1 (1965), 1-19.

M. A. Wallach and T. Mabli, "Information Versus Conformity in the Effects of Group Discussion on Risk Taking," Journal of Personality and Social Psychology, 14 (1970), 149-156.

A. G. Woodside, "Informal Group Influence on Risk Taking,'' Journal of Marketing Research, 9 (1972), 223-225.

A. G. Woodside and M.W. DeLozier, "Effects of Word of Mouth Advertising on Consumer Risk Taking," Journal of Advertising, (1976), forthcoming.

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