Perceived Risk and Consumer Behavior: a Critical Review

Ivan Ross, University of Minnesota
[ to cite ]:
Ivan Ross (1975) ,"Perceived Risk and Consumer Behavior: a Critical Review", in NA - Advances in Consumer Research Volume 02, eds. Mary Jane Schlinger, Ann Abor, MI : Association for Consumer Research, Pages: 1-20.

Advances in Consumer Research Volume 2, 1975      Pages 1-20

PERCEIVED RISK AND CONSUMER BEHAVIOR: A CRITICAL REVIEW

Ivan Ross, University of Minnesota

The empirical research relating perceived risk to consumer behavior is summarized. The literature reveals that perceived risk has been studied in relationship to information acquisition and processing constructs such as word-of-mouth behavior and opinion-leadership, as well as to overt consumer behaviors such as new product adoption, store/brand loyalty, and modes of shopping. Recent research has been concentrated on the study of relationships between specific kinds or components of perceived risk or risk consequences and the specific relievers or reducers of these components. The reviewer offers a critique of research on perceived risk and suggests direction for future research.

INTRODUCTION

When Bauer (1960) first proposed that consumer behavior could be viewed as an instance of risk taking, he modestly hoped that the "fad" he was probably introducing would "at least survive through infancy" (p. 23). After fourteen years there is evidence that the infant is fast becoming adult. Indeed, as the list of references suggests, recent years have shown a dramatic increase in publication frequency of empirical research in this area, and current models or theories of consumer behavior broadly incorporate the perceived risk construct. Engel, Kollat and Blackwell (1973) position perceived risk specifically in the "external search and alternative evaluation" stage of decision-making (pp. 376-380) and generally attribute to it great importance: "Decision making (processes) ... occur in order to reduce perceived risk to tolerable levels (p. 59) ." Howard and Sheth (1969) conceptually deal with the construct under their term, "stimulus ambiguity", viewed as a "perceptual construct" in their theory of consumer behavior (P. 30).

The reviewer has not found the organization of the empirical literature on perceived risk for the purposes of this paper an easy task nor one which is likely to be optimally-satisfying to some (hopefully, not most) readers. In the first place, perceived risk has been studied in relation to a very large number of other consumer behavior variables--too large a number to review in detail within the space limitations imposed. And secondly, the manner in which the construct has been operationally and even conceptually defined has varied so much across the studies, that efforts at synthesis are hampered by questions of "are these two studies really talking about the same thing?" More often than not, the answer is. no.

After a discussion of the conceptualization of the construct, the reviewer has chosen to organize his summary by discussing the major consumer behavior variables to which perceived risk has been applied. Some "problematic" areas in this research tradition and suggestions for future research are at the end of the review paper.

CONCEPTUALIZATION OF PERCEIVED RISK

Bauer's initial proposition was that, "Consumer behavior involves risk in the sense that any action of a consumer will produce consequences which he cannot anticipate with anything approximating certainty, and some of which at least are likely to be unpleasant" (1960, p. 24) . Thus, the two primary structural dimensions were uncertainty and consequences which much, but not all, subsequent research in perceived risk has used in the measurement procedure. Bauer strongly emphasizes that he is concerned only with subjective (perceived) risk and not "real world" (objective) risk.

It should be noted that Bauer clearly views perceived risk as not only related to consumers ' pre-decision information acquisition and processing activity but to post-decision processes as well. Hence, he describes dissonance theory as concerned with "... ways in which people reduce perceived risk after decisions are made. People will seek out information that confirms the wisdom of their decisions" (p. 32). It would have been well if some researchers in the perceived risk area had more carefully noted the view of dissonance reduction as risk reduction processes who, as a result of failing to do so, drew equivocal conclusions from their research (e.g. Arndt, 1968a; Cox and Rich, 1964; and Schiffman, 1972). In all these cases perceived risk measures were taken after the purchase had (or had not) occurred, at which time it would be reasonable to assume that risk/dissonance reduction processes had begun, and hence would likely contaminate their response to the risk measure. Indeed, these studies might better have been addressed to postpurchase dissonance/risk reduction activity explicitly since we have no examples in the risk literature of such studies.

Cox (1967a) in his initial elaboration of Bauer's conceptualization states that it is often necessary to infer the presence of perceived risk since " ... consumers may be unable or unwilling to specify that a situation confronting them is risky (p. 36) ... (thus) ... we will assume, for operational purposes, that risk is, in some way, perceived by our subjects in those situations in which they act in such a way as to handle (e.g. reduce) risk" (p. 37). The amount of perceived risk is construed to be a function of (1? "The amount that would be lost (i.e., that which is at stake) if the consequences of the act were not favorable, and (2) the individual's subjective feeling of certainty that the consequences will be-unfavorable" (p. 37). The amount at stake "... is a function of the importance or magnitude of the goals to be attained, the seriousness of the penalties that might be imposed for nonattainment, and the amount of means committed to achieving the goals" (p. 38). Whereas certainty and consequences determine the amount of perceived risk "The nature of the risk perceived should be a function of the nature of the buying goals involved" (p. 38). Given this "two factor" view of risk structure it then follows that risk might be reduced to a "tolerable level" by either or both (1) reducing the amount at stake (e.g. reducing that which the person hoped to gain, reducing the penalties for failure, and reducing the means by which the gain was to be made) and (2) increasing the degree of certainty that loss will not occur; that is, becoming more certain that action consequences would be favorable.

While most subsequent research has employed these two dimensions specifically (e.g. Cunningham, 1967a), others have used a variant two-dimensional definition such as uncertainty and importance (e.g. Schiffman, 1972; Arndt, 1968b), and some use just one dimension (e.g. uncertainty only, Arndt, 1968a). In some cases it is difficult to distinguish whether uncertainty or consequences is being measured (e.g. "how risky is the purchase of ____"). Bettman (1973) specifically conceptualizes risk dimensionality as different from that of Cox (1967a) and Cunningham (1967a) by substituting importance for consequences/ dangers. "... the risk inherent in a brand choice situation within a product class will depend upon the degree to which a buyer believes he can construct a reasonable decision rule for making a brand choice, and the importance to him of making a satisfactory choice within that product class" (Bettman, 1973, pp. 184-185). And rather than rating the uncertainty directly, Bettman's procedure is to compute the percentage of brands falling above an acceptable level of quality to the consumer (Bettman, 1973, 1974). He reports research (1973, 1974) which supports his conceptualization as opposed to Cunningham's (1967a) but he bases his arguments largely upon his finding that when uncertainty and importance are measured as he proposes, both these components contribute variance to the overall risk ratings, whereas using Cunningham's (1967a) uncertainty and danger components, by far the most variance is explained by the danger component alone. That is, in the Cunningham procedure, when there is a great deal of danger, certainty doesn't matter and effects of certainty are felt only at low levels of danger. In the Bettman procedure, the effects of certainty are most pronounced at high levels of importance, which is what Bettman argues "should" be the case. Bettman (1972) did find that when uncertainty and danger are measured as defined by Cunningham, the two components are not independent, and that the danger component is clearly more important than the uncertainty component (also see Slovic and Lichtenstein, 1968). The issue raised here is not moot, but is very difficult to address empirically, since there may well be differences of opinion in what the conceptual definition of risk is thus leading to different views of its fundamental dimensional structure. Whether or not the relationship between the two dimensions, uncertainty and consequences (or importance),is additive or multiplicative (most have assumed it multiplicative) was tested by Bettman (1972; 1974) as a combination rule or "cognitive algebra" question, and through both graphical and statistical tests he found support for the additive rather than the multiplicative procedure.

Implicit to the questions raised by Bettman is his distinction between "inherent risk" and "handled risk." "Inherent risk is the latent risk a product class holds for a consumer, the innate degree of conflict the product class arouses in the consumer. Handled risk is the amount of conflict a product class engenders when the buyer chooses a brand from that product class in his usual buying situation. Thus, handled risk includes the effects of information -->_ risk reduction processes as they have acted on inherent risk" (Bettman, 1972, p. 394). Bettman notes that these two different types of risk have been confused in the research literature; Cunningham (1967a) using inherent risk and Cox and Rich (1966) and Spence, Engel and Blackwell (1970) using handled risk. Subsequently, only a few researchers (e.g. Lutz and Reilly, 1973) have explicitly noted the type of risk they are measuring in the sense of Bettman's distinction. That the distinction is important is demonstrated by Bettman's (1972) research which found that of nine products studied, toothpaste and margarine had the highest ratings for relative inherent risk while beer and instant coffee had the highest relative ratings for handled risk.

PERCEIVED RISK, WORD-OF-MOUTH, AND OPINION LEADERSHIP

Word-of-mouth and opinion leadership were the concepts first researched in relationship to perceived risk, perhaps because Bauer had asserted that "... one of the very important functions of opinion leaders is to reduce the perceived risk of the behavior in question" (Bauer, 1960, p. 26). Research on the manner in which physicians adopted new drugs undoubtedly influenced his view (Coleman, Katz, and Menzel, 1957; Coleman, Menzel, and Katz, 1959). This research suggested that doctors tended to rely on their colleagues, especially the more "respected" ones, early in the diffusion process, rather than on non-professional medical sources. Once the drug had become reasonably well established, personal influence seemed to play a less important role. As the severity of the disease for which the drug was to be used increased so did the propensity for doctors to rely upon professional as compared to commercial sources (Bursk, 1960; Bauer and Wortzel. 1966).

Cunningham (1964, 1966, 1967a, 1967b, and 1967c) measured the uncertainty and danger (consequences) housewives perceived in the fabric softener, dry spaghetti, and headache remedies product categories (uncertainty -- would an untried brand work as well; consequences -- how much danger would she see in trying a brand she had never used before.) In addition, brand purchase behavior, word-of-mouth activity, and various descriptive measures were obtained. Cunningham hypothesized that "... those users of a product who were high in perceived risk would reduce risk through conversation and thus a greater proportion of the high risk perceivers would be classified as 'talkers' (a respondent who discussed the product category within the last six months) than would the low risk perceivers" (1967b, p. 271). The data supported this hypothesis for headache remedies and fabric softeners but not for dry spaghetti. Regarding the direction of low of word-of-mouth as a function of perceived risk, there were product differences. Those who perceived high risk regarding headache remedies were more likely than low risk perceivers to have initiated their last conversation about the product, but the relationship was reversed for fabric softeners. Also, those higher in perceived risk for fabric softeners and dry spaghetti (but not headache remedies) were more likely to have requested information than those lower in perceived risk and were more likely to claim they had made a recommendation in their last conversation, a finding at variance with the researcher's hypothesis. Nevertheless, Cunningham views this finding as one which "... strongly supports the notion advanced ... that high perceived risk consumers are sought out by others who presumably value their expert opinion" (1967b, pp. 282-283). Thus, in his view of the data, the high risk perceiver is more apt to be an opinion leader. A problem in this research design is that subjects were retrospectively reporting their role in the word-of-mouth process. The question regarding opinion leadership was, "When you bring up the subject of products and brands, do you usually ask someone else for information or do you just suggest helpful information from your own experience?" (1967b, p. 279). Those who are high risk perceivers might report that they suggested information to others (more than low risk perceivers) as part of a risk (dissonance) reduction process. Cunningham recognizes this issue but discounts this interpretation in sustaining the conclusion reported above.

Arndt (1967b, 1967c, 1968b, 1968c) studied word-of-mouth flaw within a married student housing complex concerning the adoption process for a new brand of coffee, PERKY, and the data support his hypothesis that opinion leaders would be lower in perceived risk, contrary to Cunningham's research findings above. Arndt believes that the differences between the two studies might be explained by differences in methodology or in products chosen for study. Indeed, there were methodological differences; Arndt measured importance (How important is it to you that a new brand of coffee you have never tried before is as good as your present brand: not important, fairly important, important, or very important?") rather than consequences. Arndt also found that word-of-mouth had more effect on high- than on low-risk perceivers. That is, those who were high risk perceivers appeared to pay more attention to what they had heard, particularly to unfavorable comments. In general, Arndt views his data as supportive of the conclusion that, "word-of-mouth seemed to flow from the low to the high-risk perceivers" (1967c, p. 294), primarily because the low-risk perceivers were more likely to report having given advice about PERKY than high-risk perceivers. In most all other regards, however, the high-risk perceivers seemed to be more "active" than the low-risk perceivers: in initiating pre-purchase conversation, in overhearing comments, and in seeking information.

Various other studies have asked subjects to evaluate the importance of alternative information sources,and personal sources (i.e., word-of-mouth) are invariably rated high in importance, and there is evidence that it is in particularly "high-risk" situations where personal influence is most important, supporting Arndt's view (1967a). For example, Roselius (1971) found that those who perceived high risk for "time", "ego", or "money" loss rated word-of-mouth more helpful (as a "reliever") than did subjects in general. And Perry and Hamm (1969) found that "... the higher the risk involved in a particular purchase decision, the greater the importance of personal influence" (p. 354) in their study of social and economic risks perceived by subjects across 25 product categories. The same conclusion was reached by Lutz and Reilly (1973) in their study of the effects of social and performance risk on consumer information acquisition; word-of-mouth was the most important of the four sources of information available externally to subjects (word-of-mouth, mass media advertising, rating magazine, and sales clerks). However, they did not find that variations in levels of social risk influenced information search behavior as they had hypothesized. Finally, Sheth and Venkatesan (1968) also found that the experimentally created high-risk (regarding the hair-spray product category) group sought personal sources of information significantly more than did the low-risk group.

The research on perceived risk, word-of-mouth, and opinion leadership would seem to support the generalization that word-of-mouth functions as an important (but not necessarily the most important, e.g. Roselius, 1969) risk reliever across most or all types of risks. The nature and direction of word-of-mouth flow, and specifically as this relates to opinion leadership, is less clear. Certainly this is a very complex issue and one not easily amenable to investigation, especially through a self-report mode.

PERCEIVED RISK, NEW PRODUCT ADOPTION AND BRAND/STORE LOYALTY

The drug adoption studies previously referred to (e.g. Coleman, et. al., 1957; and Coleman, et. al., 1959) suggest the hypothesis that those high in perceived risk for a product category would be less likely to adopt at all, or to adopt quickly, if at all, a new brand introduced within that category, and vice-versa. The research subsequently would seem to strongly support this hypothesis. Both Arndt (1967b) and Cunningham (1967b) in the studies previously referred to, found evidence to this effect (although not necessarily a clear relationship for all products studied), as did Schiffman (1972) in his study of the adoption of a salt substitute product among elderly consumers. If one accepts Sheth's (1968) assumption that the adoption of a stainless steel razor blade is a low-risk decision, then one might take his results as evidence for the same hypothesis: "... as high as 89 percent of total respondents adopted the stainless steel blades in slightly more than a year's time from the "flux of three major brands in the market in 1963" ... and ... "more than 90 percent of total respondents adopted within one year after becoming aware. Similarly, as high as 49 percent adopted the stainless steel blades immediately after they became aware" (pp. 180-181). He continues, "... it can be easily seen that for a low risk innovation also possessing strong relative advantage, the diffusion is faster both in terms of time of adoption and the mental process of adoption" (p. 181).

The hypothetical relationship between perceived risk and brand/store loyalty is closely related to its relationship with new product adoption. Loyalty should be stronger among those perceiving high-risk in the product category and for basically the same reason: "Much brand loyalty is a device for reducing the risks of consumer decisions" (Bauer, 1960, p. 25). Arndt (1967b) found that those high in perceived risk for coffee were more likely to be brand loyal and hence less likely to adopt the new coffee under study. Cunningham (1967c) similarly found supportive evidence for this relationship but it was less strong for dry spaghetti than for fabric softeners or headache remedies, thus suggesting that where risk is generally low for.the product category (e.g. dry spaghetti), brand loyalty plays a smaller role as a risk reduction process. Sheth and Venkatesan (1968) studied the development of brand loyalty as a risk-reduction process in repetitive (over time) consumer behavior and found support for their hypothesis that brand loyalty increased over time. It should be noted, however, that the development of brand loyalty (repeat selection of brands) was quicker for the low-risk than for the high-risk groups. The authors conclude that "...perceived risk" is a necessary condition only for the development of brand loyalty. The sufficient condition is the existence of well-known market brand(s) on which the consumer can rely." (p.310)

Hisrich, Dornoff, and Kernan (1972) hypothesized that, "If the product is intolerably ambiguous, perhaps the store, which might be far less ambiguous, can serve as a surrogate" (p. 435). They chose draperies, furniture, and carpeting as their "ambiguous" products, but found their data rejected the hypothesis. For all three products, and for both male and female subjects, and at every level of perceived risk, the number of store-loyal buyers was less than the number of non-loyal buyers. They conclude, "At a minimum, this suggests that these buyers did not consider repeat patronage as a viable risk-handling strategy. Indeed, depending on prior results, not shopping at a previously-patronized store might have served as a form of risk reduction" (pp. 438-439)

PERCEIVED RISK AND MODE OF SHOPPING

Noting that many women do not order any merchandise by phone, Cox and Rich (1964) hypothesized that telephone shopping creates perceived risk of sufficient magnitude to deter many women from shopping by this mode. Although in their measurement procedure, in the reviewer's opinion, there was contamination between the criterion measure (whether or not the item was purchased by phone) and the dependent variable measure (what items could be bought by telephone without worry and which would be worried about), the authors conclude that "...high perceived risk is likely to be a strong deterrent to purchasing an item by telephone" (p. 499). When respondents were asked why they had not shopped,< nearly two thirds replied "...that they had not done so because they were apprehensive of not getting what they wanted" (p. 495). Among those who did shop by phone, newspaper advertising was a favored source of information. However, that newspaper advertisements function to identify the merchandise and the store at which it is available may confound the relationship between "saw ad" and "shopped by phone" so that it might not necessarily suggest anything about the uncertainty reducing role that such advertising might play. That is, ipso facto, the person who does shop by phone as contrasted with one who does not is more apt to say she "relies" more on newspaper advertisements. In any case, the author's investigation of those items of merchandise rated higher and lower in perceived risk in shopping for them by phone suggested that, "The more decisions to be made in making a single purchase, the more important the decisions are, and the more uncertain the consumer is about making the decisions without visual inspection, the greater the risk potential of ordering the product by phone" (p. 505).

Using a paper-and-pencil questionnaire, subjects rated the overall risk they perceived in purchasing twenty different products through the ms; 1 as opposed to in a store or from a salesman in research conducted by Spence, Engel, and Blackwell (1970). Their hypothesis that people perceive more risk in buying by mail than buying in a store or from a salesman was supported, but their hypothesis that mail-order buyers of hospitalization insurance would perceive significantly less risk in mail-order buying of other products was not supported by the data, nor was there support for the hypothesis that m order buyers of hospitalization insurance would perceive significantly less risk in the mail-order purchase of such insurance than non-buyers. The authors recognize the inconsistency between the support of the first hypothesis and the lack of support for the second and third and urge that future research be directed at the question. It may well be that the overall risk measure employed obscured particular types of risk being more prominent in one mode than in the other, particularly as these risk components might have interacted with the particular products studied. One might further question the reliability of the difference scores (between perceived risk in the two buying situations) computed by the researchers and then subjected to an ANOV, as well as the logic in summing these difference scores across the twenty products for each subject to obtain an "average perceived risk difference" s core.

An issue not addressed in either of these studies is the question of the use of different shopping modes in gathering information rather than the more narrow question of actual purchase by that mode. It would be interesting to know whether or not high risk perceivers for buying by telephone are more or less inclined to gather information by phone as compared to lower-risk perceivers. It would also be enlightening were research conducted which simultaneously evaluated the way in which these alternative shopping modes would be used by consumers in coming to a decision rather than addressing the question one mode at a time. For example, one might create an environment where consumers could choose among these modes, each with fixed costs (e.g. postage, gasoline, time, etc.), in deciding on a Purchase.

PERCEIVED RISK AND THE RELIEVERS OF PERCEIVED RISK

Cox (1967a) very early in the perceived risk research literature noted, at least for the two consumers in his study, that reducing uncertainty was far more common than reducing unfavorable consequences as a risk reducing strategy. Sheth and Venkatesan (1968) state that, "Generally, the consumer cannot change the consequences of using a brand. He can, however, change his uncertainty about these consequences ..." (p. 307). Although there is no research directly related to this question, subsequent researchers have restricted themselves to "uncertainty reducing" strategies. However, to the extent that the certainty and the consequences dimensions of perceived risk are not independent (see Bettman, 1972), we might construe that much of the empirical evidence is functionally addressed to both dimensions simultaneously.

The reviewer has already noted one example of an instance when different levels of overall risk seem to evoke different "relievers " or risk-reducers than when risk is low; namely, the apparent important role of word-of-mouth or personal sources in general. Arndt (1967b) found that the content of perceived risk was different for high versus low risk perceivers for coffee. Those who perceived low risk denied any problem except "inconvenience" but those high in perceived risk saw "waste of money" and "husband's reaction" as risk factors. And in his study of the persuasibility of purchasing agents and chemists, Levitt (1967) found that although a high credibility company source acted to "reduce risk" in both a "high risk" (adoption of product) and a "low risk" (refer the product to someone else for serious consideration) situation, credibility was more important in the high risk situation. Thus, there is some evidence that the level of overall perceived risk might evoke different prepotencies of risk components,hence different relievers for that risk. It should be noted that research is not consistent on this point. For example, Zikmund and Scott (1973) using a multiple discriminant analysis to distinguish differences in information search activity between high versus low risk perceivers regarding lawn furniture, color TV, and stationery, did not find a statistically significant difference in these information preferences. (However, a canonical analysis performed within product categories did reveal differences.)

But most research has sought to specifically relate types of risk to types of relievers. A starting place seems to have been the difference in evaluation of product information as a function of the "performance" versus "psychosocial" goals of the consumer (e.g. Wilding and Bauer, 1968; Ross, 1972). Subsequently, most recent research in perceived risk has been focused on the relationship between risk and reliever relationships (Roselius, 1971; Perry and Hamm, 1969; Lutz and Reilly, 1973; Zikmund and Scott, 1973; Jacoby and Kaplan, 1972; Kaplan, Szybillo and Jacoby, 1974; Newton, 1967; and McMillan, 1972).

Although they did not relate their risk components to types of relievers, Jacoby and Kaplan (1972) did address the fundamental structure of these components. They identified five types of risk: (1) financial, (2) performance, (3) physical, (4) psychological, and (5) social. Considering the way in which these components grouped themselves as subjects rated these risks with regard to twelve diverse consumer products, it was clear that price seemed to be the metric ordering these products on overall risk perception. Performance risk correlated most highly with overall perceived risk more highly than any other component for eight of the twelve products, and was highly correlated for the others. For that reason, the authors suggest that performance risk could be employed as an approximation of overall perceived risk. However, this generalization may be unwarranted since most of the twelve products "seem" highly "performance" related (performance risk -- '\hat is the likelihood that there will be something wrong with an unfamiliar brand of or that it will not work properly"), and thus high correlations between this component and the overall measure ("On the whole, considering all sorts of factors combined, about how risky would you say it was to buy an unfamiliar brand of "?) may be variously interpreted. In any event, after performance risk, the next most important risks averaging across the twelve products are (in decreasing order) financial, social, psychological, and last, physical. But in regressing components on the overall perceived risk score, social entered after performance risk. The authors found that a multiple regression equation predicting overall perceived risk from component risk scores accounted for 74 percent of the variance in this criterion, and moreover, they cross-validated the regression weights derived in this study using different subjects two years later (Kaplan, Szybillo, and Jacoby, 1974) and found negligible shrinkage. The cross-validation is to be applauded since it stands out as a singular event of its kind in all the perceived risk research literature.

Roselius (1971) identified another type of loss, time loss, and studied the effect of this type of loss in comparison to three others: hazard loss, ego loss, and money loss, the latter three seemingly comprising the same "set" of losses or risks as those employed in the research by Jacoby and Kaplan (1972). There were eleven risk or loss relievers studied with respect to these losses using a five-point rating from 472 housewives on "how helpful (almost always, usually, rarely, almost never) each reliever would be for reducing the risk posed in the situation": (1) endorsements, (2) brand loyalty, (3) major brand image, (4) private testing, (5) store image, (6) free sample, (7) money-back guarantee, (8) government testing, (9) shopping, (10) expensive model, and (11) word-of-mouth. We are told only that the questionnaire "presented several generalized risky buying situations" in which "situations were not related to specific products or purchase methods," so there is no way for the reader to know the meaningfulness of the stimuli employed. Of the eleven relievers evaluated, "brand loyalty" and "major brand image" evoked the most consistently favorable response, being ranked first and second, respectively, as relievers for each of the four losses. (The authors constructed -a "net favorable percentage" quotient for each reliever which was the number of unfavorable responses subtracted from the number of favorable responses given to it by subjects, the difference divided by the total number of responses, then multiplied by 100). Some relievers were consistently rated "unfavorably" by respondents; e.g. "expensive model" was least helpful in all four kinds of losses and had a negative quotient sign, and "private test", "money back guarantee", and "endorsements" all had negative signs across the four losses. One may well question the meaningfulness of a "negative" quotient sign in that it suggests a reliever was "not helpful" (when it could on the average have been helpful) or "unhelpful" (which was not a response alternative for subjects). Some relievers had "special meaning" in the sense that high loss perceivers were particularly sensitive to them; for example, as previously noted, "word-of mouth" was a helpful reliever except for "hazard loss" ("Some products are dangerous to our health or safety when they fail"), and "major brand image" seemed to function much the same as "word-of-mouth." "Government testing" was found to be particularly helpful as a reliever for hazard loss. Six of the relievers were labeled "general-purpose" risk relievers by Roselius since there was not a significant difference in responses to them between high perceivers and "other" buyers across the types of loss; these were brand loyalty, private testing, shopping, endorsements, expensive models, and money-back guarantees. Roselius concludes that "... buyers prefer some relievers to others depending upon the kind of loss involved ... (and) ... perhaps a seller should first determine the kind of risk perceived by his customers and then create a mix of risk relievers suited for his combination of buyer type and loss type (p. 61).

Lutz and Reilly (1973) specifically undertook an investigation of the effects of social and performance perceived risk on information acquisition. Subjects rated which of five types of information they would use in making a purchase decision for each of nine products selected by pretest to represent different levels of social and performance risk combinations. They found that when performance risk was low or moderate, subjects opted for "buy" (take their chances and pick a brand without search for product information), but when products were high in performance risk, "direct observation and experience" was the most preferred route and "buy" the least. Over all levels of performance risk, word-of-mouth was the most important of the four information sources external to the consumer (following "buy" and "direct observation and experience"), but contrary to the hypothesis, variations in social risk level did not have any influence on consumers' information search behavior. In a similar sort of research design, Perry and Hamm (1969) tested the effect of social risk and economic risk on subjects' evaluations of seven alternative sources of information and found that "... the higher the risk involved in a particular purchase decision, the greater the importance of personal influence" (p. 354).

Zikmund and Scott (1973) conducted personal interviews with housewives to evoke specific risk consequences associated with purchasing in eight different product classes. These risk variables were then factor analyzed to identify principal risk dimensions. They also measured the traditional uncertainty and consequences of perceived risk and comPuted an overall risk measure by multiplying these two components. Considering color TV, lawn furniture and stationery, the factor analysis showed that all three products had "... dimensions relating to quality or reliability and the reaction of significant others who might judge a purchase" (p. 410). A "new" risk factor, not previously identified, was "future opportunity lost", which is associated with the expectation that an improved or lower cost product may be available at a future time which would be precluded by a current purchase -- both color TV sets and lawn furniture have this risk associated with them. They also found a "shopping frustration" factor associated with lawn furniture and stationery but not with color TV. The authors concluded that the research "... illustrates an important reason for investigating perceived risk in a multidimensional fashion. Consumers evaluate products on the basis of a few principal attributes and each represents a potential source of risk. Further, these attributes vary across product classes. Disaggregating perceived risk into product-specific components in this fashion provides much more information about why a consumer perceives risk than overall measures such as social or performance risk" (p. 411).

PERCEIVED RISK AND PERSONALITY

Cunningham (1967a) had suggested that some people have a generalized tendency to perceive either high or low risk across a range of products, but this hypothesis has not been specifically tested subsequently. Cox (1967a) noted a "clarifier"-"simplifier" cognitive style difference in the two subjects he studied intensively; the one tending to seek information to clarify or reduce ambiguity ("clarifier") and the other reducing ambiguity by keeping out disturbing cognitive elements ("simplifier").

The perceptual/cognitive style construct, "category width", or "broad" versus "narrow" categorizers, has been specifically studied in relation to perceived risk. Pettigrew (1956) observed that "broad categorizers seem to have a tolerance for type I errors: they risk negative instance in an effort to include minimum positive instances. By contrast, narrow categorizers are willing to make type II errors. They include many positive instance by restricting their category ranges in order to minimize the number of negative instances (p. 532). Popielarz (1967) reasoned that broad categorizers would express greater willingness to buy new products than narrow categorizers and that people with broad category ranges would be more likely to perceive smaller qualitative differences between products of a given product class than narrow categorizers. For each of six products, subjects indicated their willingness to buy each of four qualitatively different brands; the brands differed in newness of products themselves and in the buyers' familiarity with the brand name of the manufacturer (two levels of newness and two of familiarity). They then rated the extent to which they saw brands as qualitatively different. Category width scores correlated with willingness to buy in the hypothesized direction, but only for male subjects was the prediction supported that broad categorizers would perceive smaller qualitative differences among products; for female subjects, the relationship was reversed.

Schiffman (1972) also found that the broad categorizer was more apt to have adopted a new product (salt substitute) than the narrow categorizer, but the measure he employed to measure category width (he refers to the construct as "error tolerance") seems to the reviewer to be criterion contaminated (see "The criterion and construct definition problem" section of this review). And although not specifically related to the perceived risk concept, Barach (1969) found that broad categorizers were more persuaded (i.e. "switched" more to the advertised brand in a Schwerin test) than were narrow categorizers.

In sum, there is evidence that the perceptual or cognitive construct, category width, is related to willingness to adopt new products; broad categorizers being more willing than narrow categorizers to adopt. However, the construct has not been specifically related to perceived risk. We do not know that broad categorizers are less prone to perceiving risks as a generalization.

Self-confidence as a personality construct has also been studied in relationship to perceived risk. Although Hisrich, et. al. (1972) found a significant inverse relationship between perceived risk and generalized self-confidence, Zikmund and Scott (1973) and Cunningham (1967a) found no relationship between the two. However, Cunningham (1967c) does report that those subjects medium in self-confidence were more likely to be brand loyal (high brand commitment) than those low or high in self-confidence. The effect of self-confidence on perceived risk remains unclear.

THE CRITERION AND CONSTRUCT DEFINITION PROBLEM

Several investigators have recognized that the criterion problem has not been adequately addressed (e.g. Spence, et. al. , 1970, p. 369). Unless we can "know" what kinds of behaviors are manifestations of risk then it is always equivocal that we are really (validly) measuring risk. The criterion problem is at the same time a construct definition problem, and vice-versa. Researchers have assumed that there is risk in decision-making simply because, using the instruments they have developed to measure it, they have "measured" it. Naturally, one could make the same observation about research difficulties associated with other similar hypothetical constructs or intervening variables such as "personality", "self-confidence", "dissonance", "attitudes", and so on. But the reviewer would suggest that the criterion/construct definition problem in perceived risk research appears to be more troublesome and less adequately dealt with than in programmatic research in these other areas. One can find numerous examples among the studies reviewed herein where the "criterion contamination" problem is severe,rendering "findings" of many of these studies equivocal, at best.

For example, Arndt (1968a) measured (and thus defines) perceived risk by asking housewives, "How sure would you be of picking the best brand of (product class): very sure, quite sure, sure, not too sure, or unsure?" (p. 3). A criterion variable is "innovativeness" which he measures by asking when the consumer made her first purchase of the product (ranging from not at all to three or more years ago). He not surprisingly finds a strong negative correlation between perceived risk and innovativeness. Now if a consumer has never bought any brand in the product category (and in his study, 40 percent had never bought soft margarine and 74 percent had never bought electric toothbrushes or electric dishwashers) one should not be surprised to find that she is rather "unsure" that she would be able to "pick the best brand" in that product category.

Cox and Rich (1964) have a similar contamination issue wherein they measure perceived risk in telephone shopping by asking respondents to sort 20 cards, each bearing the name of a particular item of merchandise, into two piles: (1) "Items you feel could be bought by telephone without too much worry over getting just what you want"; and (2) "Items which you would worry about if ordered by phone" (p. 499). All respondents in this portion of the study had ordered by phone at least once during the year prior to the survey. The 10 items about which the respondents had the most "worry" were designated "high perceived risk" items and the other 10, the "low perceived risk" items. The authors then determined the relationship between the perceived risk ratings merchandise items received and the frequency with which an item was mentioned as being purchased by the telephone in their last phone orders. They find that "... knowing that an item is high in perceived risk allows us to predict that in 90 percent of the cases the item will be a medium or low frequency of telephone purchase item. Knowing the item is low in perceived risk allows us to predict that in 85 percent of the cases the item will be medium or high in frequency of phone purchase." (p. 499) Although the authors caution that "... it is not possible to demonstrate the direction of causality ... that, "... it would seem reasonable to conclude that high perceived risk is likely to be a strong deterrent to purchasing an item by telephone" (p. 499). One might argue that this observed relationship demonstrates that people say they are apt to be less worried about doing something they have in fact done before than they are to worry about doing something they have not done before. The construct of perceived risk has become obscured by defining (measuring) it in a situation-specific context (ordering by telephone) which is the same context used to "validate" and interpret the effect of the construct. The logic here (or the lack of it) is akin to the assertion that "not having a telephone is likely to be a strong deterrent to purchasing an item by telephone." Although probably "correct", we have not learned much about why some people shop by telephone more than do others.

Contamination problems also seem present in Schiffman's (1972) study of new product trial (a salt substitute) among 100 elderly (average age, 74) consumers. All households had received a coupon worth 30 cents on the purchase of the salt substitute regularly priced at 59 cents. After two weeks, 17 percent had redeemed the (household coded) coupon, soon after which time interviews were held with the female member of all households. "Taste risk" and "health risk" were measured as well as the importance of each of these two types of risk as follows (p. 107): "Would you say you are quite certain, somewhat certain, or not certain that a new brand of salt substitute would taste as good as regular salt (Taste Risk)? And, "Would you say it is not important, of some importance, or quite important for you (or your husband) to get a salty flavor into your food (Taste Risk Importance)? For Health Risk the question was, "Would you say you are quite certain, somewhat certain, or not certain that a new brand of salt substitute would be better for one's health than regular salt?", and the Importance of Health Risk was evoked by, "Would you say there is no danger, some danger, or much danger in using a new brand of salt substitute in place of regular salt?" Since these questions were asked after adoption had occurred, and assuming that taste and health considerations were indeed primary (as author's pilot study had indicated), then we have a backward contamination built into the design in that the retrospective recollection of "risks" among those who had already adopted the salt substitute would be expectedly "lower" than the risk perception of those who had rejected the product, presumably for the very risk "reasons" contained in the risk measure. Further contamination occurs between Schiffman's measure of respondents history of new product adoption (how many of ten new food products introduced in the 18 months prior to the study had the household tried) and his measure of "perceived error tolerance" (another way of saying broad versus narrow categorizers) which was measured by the questions: "Who is a wiser consumer: (1) a person who tries a new food product which turns out to have a poor taste or (2) a person who does not try a new food product and later learns it has a good taste?", and, "Do you prefer to: (1) try a new food product when it first comes out, or (2) wait and learn how good it is before trying it?" (p. 107). The strong positive correlation between the "inclusion" (broad) style and the criterion measure, the number of new food products adopted, is a clearly contaminated relationship which only demonstrates that if a person knows they've tried new food products lately they will say they try new foods when they first come out. What this demonstrates about even the existence of a "new risk-handling variable, perceived error tolerance", quite aside from the question of what such a construct could or does have to do with new product trial, is unclear.

THE "LEAP" TO CONCLUSIONS UNWARRANTED BY THE DATA

This age-old problem is not escaped in the perceived risk empirical research tradition; not withstanding the "aiding and abetting" of this propensity by journal editors who demand "marketing management implications" to that which they publish, one has reason to be critical of such conclusions, interesting as they may be, especially since many of us who skim the journals almost always read the "findings" and/or "conclusions" sections if nothing else. We would be led astray by many of the conclusions we would read in this area.

For example, Sheth (1968), despite his caveats that the empirical research on the diffusion of innovations has been lacking in theoretical foundation and rather has proceeded "on the grounds of convenience and ease in implementation" (p. 175), and that in the study of diffusion it is important "to consider the adopter's perception of the magnitude of risk involved in an innovation", concludes after his research on the adoption of stainless steel razor blades among 601 college males that "... it can be easily seen that for a law risk innovation also possessing strong relative advantage, the diffusion is faster both in terms of time of adoption and the mental process of adoption" (p. 181). In fact, Sheth does not measure whether or not or the extent to which risk was perceived by adopters but simply asserts that this would be an example of a "low risk" innovation. Further, we know nothing about subjects' perception of "relative advantage" of this product nor anything about the "mental process" of adoption (or non-adoption). Although Sheth may certainly be correct in his theoretical argument that the adoption process for a "miracle drug" may be quite different than that for a stainless steel razor blade in part because of the differences in "perceived risks" among adopters for these products, we find no data in this study which are relevant to this very important issue. If speed of adoption/diffusion is all we're interested in, then why not simply contrast sales curves at retail for razor blades, new drugs, or whatever? Why bother asking adopters when they adopted if this is not specifically to be related to some measure of risk perception?

Perry and Hamm (1969) specifically measure only two risk components, "social" ("how the purchase decision will affect the opinion other people hold of the individual") and "economic" ("how the purchase will affect the individuals ability to make other purchases") (p. 351), and yet draw conclusions with respect to "high-risk purchase situations" despite the fact that they did not include other components of risk in their conceptualization. They also conclude that "These findings suggest that promotional strategies in a high-risk purchase situation ... should ... emphasize the social benefits of the purchase more than the economic ones" (p. 354) on the grounds that the social risk component contributed more variance to total risk scores than did the economic risk component. But the variance finding alone does not suffice as a basis for arguing that "social benefits" would be more "risk reducing" than "economic" ones, although the finding does suggest an interesting hypothesis which ought to be empirically tested.

FUTURE RESEARCH NEEDS

1. Unobtrusive measurement or at least disguised measures of overt risk- reduction processes:

Since the usual research designs have measured risk perception simultaneously with risk relieving preferences or activities in a paper-and-pencil or other self-report mode, the subjects' "set" to be rational and to give "proper" answers is a likely bias which may, for example, account for the relatively low importance attributed to media advertising as a risk-reduction information source. This research mode also is likely to sensitize subjects to their "perception" of perceived risk and may therefore motivate them to behave as if there was risk when, without the intrusive measure, they might not have done so. Admittedly, there is a danger in separating in time the measurement of risk (less in the "inherent" than in the ''handled" sense, probably) from the measurement of risk handling processes, but to the extent that it can be done, the two measurements should not be so "nakedly" exposed to one another. Embedding either or both within other measures or other disguises are recommended. Unobtrusive measures have obvious advantages where the subject of the investigation would seem to be so especially sensitive to these potentially reactive measures. Certainly the subject's preferences for and utilizations of some information (e.g. advertising, salespersons, etc.) could be unobtrusively measured by the ingenious researcher.

2. Experimental manipulation of risk:

With the exception of research by Sheth and Venkatesan (1968), there is no research in this area which directly manipulates risk by experimental design (although several studies expose subjects to different risk "sets" in evoking preferences for alternative kinds of information, e.g. Hart, 1974). There can and should be more research experimentally manipulating products/ services and purchase/use situations (e.g. use by purchaser only versus use by others, purchase by telephone versus in-store, etc.). Use of such designs would give the researcher more power, especially in addressing theoretical relationships within the risk model (e.g. kind of risk perceived and preference for risk relievers).

3. Mathematical formalization of risk structure:

As noted by Nicosia (1969) in his review of the Harvard studies (Cox, 1967c), the perceived risk literature "...reflects the shortcomings of a too direct dialogue between verbal hypothesis and empirical data (and their statistical manipulations) without the benefits of an intervening formalization" (p. 165). There is no true model of the perceived risk construct as it relates specifically to information acquisition, transmission, or handling. Taylor's (1974) efforts may be helpful in this regard as well as some of the work on risk enhancement strategies as stimulated by Berlyne (1965) reflected especially in Copley and Callom (1971).

4. Risk enhancement strategies:

Berlyne (1965) has asserted that increasing response conflict can be as important as attempts to reduce conflict, especially in monotonous environments wherein persons may engage in "diversive" exploration (p. 244). Deering and Jacoby (1972) and Copley and Callom (1971) have made provocative initial efforts in studying the conditions under which such risk enhancing activity may occur, and the question seems worthy of further study. Deering and Jacoby specifically hypothesized that "Given purchase alternatives which encompass a wide range of risk, maximal preference should be manifested for alternatives which are neither extremely high nor low in perceived risk" (p. 406) and found some support for this hypothesis b ut the relationship was more complex than expected. Copley and Callom studied industrial search behavior and did identify a group which behaved as the "Berlyne Curve" would suggest; however, only 8 percent of the subjects in their study behaved in this way. And Venkatesan's (1973) work in "novelty-seeking" may be a conceptually related construct which ought generate empirical research in consumer behavior.

5.   The literature reviewed suggests that perceived risk is a function of intrapersonal variables (e.g. personality), product differences, and situation differences (e.g. for "self" or for "other" -- see Hart, 1974; and Reingen, 1974). There may well be demographic correlates but the research obscures such relationships because of variations in the demographic composition of subjects across studies (see Brown, 1969; Spence, et. al., 1970; Cunningham, 1967a; and Hasty, 1969). Thus, we need "richer" research designs to simultaneously address these variables.

REFERENCES

Andreasen, A. Attitudes and customer behavior: a decision model. In L. E. Preston (Ed.), New research in marketing. Berkeley: University of California Institute of Business and Economic Research, 1965. Pp. 1-16.

Arndt, J. Word-of-mouth advertising and informal communication. In D. F. Cox (Ed.), Risk-taking and information-handling in consumer behavior. Boston: Harvard University Press, 1967. Pp. 188-239. (a)

Arndt, J. Perceived risk, sociometric integration, and word-of-mouth in the adoption of a new food product. In D. F. Cox (Ed.), Risk-taking and information-handling in consumer behavior. Boston: Harvard University Press, 1967. Pp. 289-316. (b)

Arndt, J. Role of product-related conversations in diffusion of a new product. Journal of Marketing Research, 1967, 4, 291-295. (c)

Arndt, J. New product diffusion: the interplay of innovativeness, opinion leadership, learning, perceived risk and product characteristics. Markeds Kommunikasjon, 1968, 2, 1-9. (a)

Arndt, J. A test of the two-step flow in diffusion of a new product. Journalism Quarterly, 1968, 45, 457-465. (b)

Arndt, J. Profiling consumer innovators. In J. Arndt (Ed.), Insights into consumer behavior. Boston: Allyn and Bacon, 1968. Pp. 71-83. (c)

Arndt, J. Intrafamilial homogeneity for perceived risk and opinion leadership. Journal of Advertising, 1972, 1, 40-47.

Barach, J. A. Advertising effectiveness and risk in the consumer decision process. Journal of Marketing Research, 1969, 6, 314-320.

Bauer, R. A. Consumer behavior as risk-taking. In R. S. Hancock (Ed.), Dynamic marketing for a changing world. Chicago: American Marketing Association, 1960. Pp. 389-398. Cited from D. F. Cox (Ed.), Risk-taking and information-handling in consumer behavior. Boston: Harvard University Press, 1967. Pp. 23-33.

Bauer, R. A. Risk handling in drug adoption: Role of company preferences. Public Opinion Quarterly, 1961, _, 546-549.

Bauer, R. A. Source effect and persuasibility: A new look. In D. F. Cox (Ed.), Risk-taking and information-handling in consumer behavior. Boston: Harvard University Press, 1967. Pp. 559-578.

Bauer, R. A., and Wortzel, L. H. Doctor's choice: The physician and his sources of information about drugs. Journal of Marketing Research, 1966, 3, 40-47.

Berlyne, D. E. Structure and direction in thinking. New York: Wiley, 1965.

Bettman, J. R. Perceived risk: A measurement methodology and preliminary findings. In M. Venkatesan (Ed.), Proceedings, Third Annual Convention of the Association for Consumer Research, 1972, 394-403.

Bettman, J. R. Perceived risk and its components: A model and empirical test. Journal of Marketing Research, 1973, 10, 184-189.

Bettman, J. R. Information integration in perceived risk. University of California, Los Angeles, Center for Marketing Studies. Working paper number 15, May, 19 74.

Brown, H. E. The role of personality and perceived risk in the purchase of branded headache and pain and cold remedies. Unpublished doctoral dissertation. Ohio State University. 1969.

Bursk, E. Consumer behavior research seminar manuscript. Boston: Harvard Business School, 1960.

Coleman, J., Katz, E., and Menzel, H. The diffusion of an innovation among physicians. Sociometry, 1957, 20, 253-269.

Coleman, J., Menzel, H., and Katz, E. Social pressures in physicians' adoption of a new drug. Journal of Chronic Diseases, 1959, 9, 1-19.

Copley, T. P. and Callum, F. L. Industrial search behavior ant perceived risk. In D. M. Gardner (Ed.), Proceedings, Second Annual Convention of the Association for Consumer Research, 1971, 208-217.

Cox, D. F. Risk handling in consumer behavior -- an intensive study of two cases. In D. F. Cox (Ed.), Risk-taking and information-handling in consumer behavior. Boston: Harvard University Press, 1967. PP. 34-81. (a)

Cox, D. F. The influence of cognitive needs and styles on information handling in making product evaluations. In D. F. Cox (Ed.), Risk-taking and information-handling in consumer behavior. Boston: Harvard University Press, 1967, Pp. 370-393. (b)

Cox, D.F. Risk-taking and information-handling in consumer behavior. Boston: Harvard University Press, 1967. (c)

Cox, D.F. and Rich, S.V. Perceived risk and consumer decision-making -- the case of telephone shoPPing. Journal of Marketing Research, 1964, 1, 32-39.

Cunningham, S.M. Perceived risk as a factor in product-oriented word-of-mouth behavior: A first step. In L.G. Smith (Ed.), Reflections on Progress in Marketing. Chicago: American Marketing Association, 1964. Pp. 229-238.

Cunningham, S.M. Perceived risk as a factor in the diffusion of new product information. In R.M. Hass (Ed.), Science, Technology. and Marketing. Chicago: American Marketing Association, 1966. Pp. 698-721.

Cunningham, S.M. The major dimensions of perceived risk. In D.F. Cox (Ed.), Risk-taking and information-handling in consumer behavior. Boston: Harvard University Press, 1967. Pp. 82-108. (a)

Cunningham, S.M. Perceived risk as a factor in informal consumer communications. In D.F. Cox (Ed.), Risk-taking and information-handling in consumer behavior. Boston: Harvard University Press, 1967. Pp. 265-288. (b)

Cunningham, S.M. Perceived risk and brand loyalty. In D.F. Cox (Ed.), Risk- taking and information-handling in consumer behavior. Boston: Harvard University Press, 1967. Pp. 507-523. (c)

Deering, Barbara J. and Jacoby, J. Risk enhancement and risk reduction as strategies for handling perceived risk. In M. Venkatesan (Ed.), Proceedings, Third Annual Convention of the Association for Consumer Research, 1972, 404-416.

Engel, J.F., Kollat, D.T., and Blackwell, R.D. Consumer behavior. (2nd ed.) New York: Holt. Rinehart. Winston. 1973.

Green, P.E. Consumer use of information. In J.W. Newman (Ed.), On knowing the consumer. New York: John Wiley and Sons, 1966. Pp. 67-80.

Hart, E.W., Jr. Consumer risk-taking for self and for spouse. Unpublished doctoral dissertation, Purdue University, 1974.

Hasty, R.W. Some demographic correlates o, risk perception in the purchase of electric housewares. Unpublished doctoral dissertation, University of Colorado. 1969.

Hisrich, R.D., Dornoff, R.J., and Kernan, J.B. Perceived risk in store selection. Journal of Marketing Research, 1972, 9, 435-439.

Howard, J.A. and Sheth, J.N. The theory of buyer behavior. New York: John Wiles and Sons, 1969.

Jacoby, J. and Kaplan, L. The components of perceived risk. In M. Venkatesan (Ed.), Proceedings, Third Annual Convention of the Association for Consumer Research, 1972, 382-393.

Kaplan, L., Szybillo, G.J., and Jacoby, J. Components of perceived risk in product purchase: A cross-validation. Journal of Applied Psychology, 1974, 59, 287-291.

Levitt, T. Persuasibility of purchasing agents and chemists: Effects of source, presentation, risk, audience competence, and time. In D. F. Cox (Ed.), Risk-taking and information-handling in consumer behavior. Boston: Harvard University Press, 1967. Pp. 541-558.

Lutz, R. J. and Reilly, P. J. An exploration of the effects of perceived social and performance risk on consumer information acquisition. In S. Ward and P. Wright (Eds.), Advances in Consumer Research. Volume 1. Proceedings of the Fourth Annual Convention of the Association for Consumer Research . 1973 393-405.

McMillan, J. R. The role of perceived risk in vendor selection decisions. Unpublished doctoral dissertation, Ohio State University, 1972.

Newton, D. A. A marketing communications model for sales management. In D. F. Cox (Ed.), Risk-taking and information-handling in consumer behavior. Boston: Harvard University Press, 1967. Pp. 579-602.

Nicosia, F. M. Perceived risk, information processing, and consumer behavior: A review article. Journal of Business, 1969, 42, 162-166.

Perry, M. and Hamm, B. C. Canonical analysis of relations between socioeconomic risk and personal influence in purchase decisions. Journal of Marketing Research, 1969, 6, 351-354.

Pettigrew, T. F. The measurement and correlates of category width as a cognitive variable. Journal of PersonalitY, 1956, 26, 532-544.

Popielarz, D. T. An exploration of perceived risk and willingness to try new products. Journal of Marketing Research, 1967, 9, 368-372.

Reingen, P. H. Choice shifts along a dimension of risk tested with consumer-related stimuli. Unpublished doctoral dissertation, University of Cincinnati. 1974.

Roselius, T. Consumer rankings of risk reduction methods. Journal of Marketing, 1971, 35, 56-61.

Ross, I. Structure of information seeking and personality in consumer decision-making. Journal of Business Administration, 1972, 3, 55-67.

Schiffman, L. G. Perceived risk in new product trial by elderly consumers. Journal of Marketing Research, 1972, 9, 106-108.

Sheth, J. N. Perceived risk and diffusion of innovations. In J. Arndt (Ed.), Insights into consumer behavior. Boston: Allyn and Bacon; 1968. Pp. 173-188.

Sheth, J. N. and Venkatesan, M. Risk-reduction processes in repetitive consumer behavior. Journal of Marketing Research, 1968, 5, 307-310.

Slovic, P. and Lichtenstein, Sarah. Relative importance of probabilities and payoff in risk taking. Journal of ExPerimental Psychology MonograPhs, 1968, 78, 1-18.

Spence, H. E., Engel, J. F., and Blackwell, R. D. Perceived risk in mail-order and retail store buying. Journal of Marketing Research, 1970, 7, 364-369.

Taylor, J. W. The role of risk in consumer behavior. Journal of Marketing, 1974, 38, 54-60.

Venkatesan, M. Cognitive consistency and novelty-seeking. In S. Ward and T. S. Robertson (Eds.), Consumer behavior: Theoretical sources. New York: Prentice Hall. 1973.

Wilding, J. and Bauer, R. A. Consumer goals and reactions to a communication source. Journal of Marketing Research, 1968, 5, 73-77.

Zikmund, W. G. and Scott, J. E. A multivariate analysis of perceived risk, self-confidence and information sources. In S. Ward and P. Wright (Eds.), Advances in Consumer Research. Volume 1. Proceedings of the Fourth Annual Convention of the Association for Consumer Research, 1973, 406-416.

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