Risk Enhancement and Risk Reduction As Strategies For Handling Perceived Risk

Citation:

Barbara J. Deering and Jacob Jacoby (1972) ,"Risk Enhancement and Risk Reduction As Strategies For Handling Perceived Risk", in SV - Proceedings of the Third Annual Conference of the Association for Consumer Research, eds. M. Venkatesan, Chicago, IL : Association for Consumer Research, Pages: 404-416.

Proceedings of the Third Annual Conference of the Association for Consumer Research, 1972      Pages 404-416

RISK ENHANCEMENT AND RISK REDUCTION AS STRATEGIES FOR HANDLING PERCEIVED RISK

Barbara J. Deering, Purdue University

Jacob Jacoby, Purdue University

Consumers must frequently make purchase decisions on the basis of inadequate information. Accordingly, important decisions often involve consequences which cannot be predicted with confidence. The possible occurrence of unpleasant post-purchase consequences and the pre-purchase uncertainty about those outcomes are incorporated in the concept of perceived risk (Bauer, 1960: Cox, 1967a). According to the risk theoreticians, perceived risk is generally handled through risk-reducing strategies such as brand loyal buying and reliance on certain sources of information (cf. Roselius, 1971). However, risk-enhancement can also constitute a viable risk-handling strategy. Under some circumstances (e.g., boredom), consumers may find the amount of perceived risk unsatisfactorily low and attempt to increase risk through their purchase decisions. A theoretical foundation for understanding risk enhancement and relevant data is now presented.

Perceived risk in a purchase situation depends upon two determinants: the clarity with which the consumer can define decision alternatives, and the consumer's involvement in the decision outcomes (Cox, 1967b). Uncertainty about decision alternatives comes from two sources. First, the consumer's knowledge of her own needs and purchase goals is frequently inadequate. Second, she must usually anticipate purchase results on the basis of those product qualities which she can assess beforehand. Such qualities may or may not have any predictive validity.

Personal involvement refers to the possible adverse consequences of poor purchase decisions. The needs and desires which the purchase is intended to satisfy may remain unaltered. Additionally, the loss of time, money, and energy depletes resources available for future purchases. Severity of perceived consequences thus depends upon the perceived importance of the purchase and the consumer's effort to carry it out.

When the buying situation is defined in terms of perceived risk, it appears to be closely related to Berlyne's (1957, 1960, 1965) concept of conflicting response tendencies. The amount of cognitive conflict evoked varies with three elements: the degree of subjective uncertainty about choosing a response, the absolute strength of the response tendencies, and the degree of incompatibility between responses. The latter two elements correspond closely to constituents of the consequences component in perceived risk. As goals represented by purchase decisions become more important, the consumer would be expected to more strongly desire and attempt to carry out the purchases. The loss of time and money invested are relevant to perceived consequences largely because they are incompatible with the initiation of subsequent responses directed toward the same ends. In most cases, a consumer cannot immediately turn to another purchase alternative if an initial choice has been costly

Subjective uncertainty is also defined similarly by Berlyne and the risk theorists. According to Berlyne (1965), subjective uncertainty varies with discrepancies in information which confront a person. These discrepancies are named "collative variables", referring to the characteristic comparison, or collation, of items of information within the present and past situations. Berlyne, like Cox, stresses the subjective evaluation of probabilities (Abelson, Aronson, McGuire, Newcomb, Rosenberg, & Tannenbaum, 1968, p. 45) as the basis for uncertainty.

Also like the risk theorists, Berlyne believes that subJective uncertainty predominates in determining the probability and direction of behavior. However, he asserts that attempts to increase response conflict are as important as attempts to reduce conflict. The function of "diversive" exploration is to introduce variation, new information, "amusement," "diversion," and "aesthetic experience" (Berlyne, 1965, p. 244). Diversive exploration is most likely to occur in monotonous environments. In such environments, a high degree of response conflict will be attractive. Diversive exploration, which does not occur in all circumstances, nonetheless contrasts sharply with the concept of pervasive risk-reducing behavior.

Drawing on Berlyne's formulation, Howard and Sheth (1969) posited two primary purposes for consumers' information search: such search helps fulfill specific purchase goals, either immediately or in the future, and also provides stimulation when the unpredictability of the buying situation is unpleasantly low. This latter, diversive exploration tends to alternate with the former, problem-specific, search as the consumer's attitude toward the product shifts between curiosity and boredom.

Certain evidence suggests that consumers do not act exclusively to minimize the uncertainty and importance of purchases. Buying a brand repeatedly rather than switching involves relatively less uncertainty about product performance. However, repeat buying accounts for only a small proportion of all buying decisions (Frank, 1967). Using available information sources, particularly when the product is new, reduces perceived risk. But consumers confronted with a new brand frequently try it without consulting anyone beforehand (Arndt, 1967). Though mail-order buying generally appears riskier than in-store or salesman-mediated buying, the consumers who shop by mail do not perceive less risk in such buying than do consumers who do not buy by mail (Spence, Engel, & Blackwell, 1970).

Studies dealing specifically with perceived risk offer few data relevant to risk-enhancement for several reasons. First, risk enhancement is more likely under conditions of low perceived risk. Products representing extremely low perceived risk are infrequently used in such studies. They may be eliminated for their apparently trivial and non-involving characteristics--just those characteristics which would be expected to elicit diversive exploration. For instance, after a lengthy series of trials consumers may react to boredom with risk-enhancing behavior. Trial sequences considerably longer than those customarily used (e.g. Sheth & Venkatesan, 1968. Swan, 1969) may be of interest. Second, if product preference and search behavior are expected to both increase and decrease as a function of risk, at least three risk levels are necessary for interpreting risk effects. Use of two risk-levels (Cox & Rich, 1964; Spence et al, 1970) or emphasis on a single component of risk (Perry & Hamm, 1969) is relatively uninformative. Third, risk enhancement involves selecting products and items of information which increase the complexity of the consumer's decision. Such diversive stimuli must be available, perhaps in the form of product alternatives, but have not yet been made available in previous studies.

HYPOTHESES

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. The most acceptable alternative should represent a compromise between the risk-enhancing and risk-reducing tendencies.

The acceptable range of perceived risk may vary among individuals and groups of individuals, dependent on their previous purchase experience and involvement in unpredictable non-consumer contexts (Howard & Sheth, 1969). Individual differences in risk-taking propensity have been related to willingness to try new products (Popielarz, 1967) and susceptibility to advertising messages (Barach, 1969). Copley and Callom (1971) grouped industrial buyers according to risk perceived across 12 buying situations and found that the relationship of search behavior to perceived risk varied across groups.

Despite individual variability, however, certain decisions probably represent acceptable risk for most consumers. Low cost, easy availability, repetitive use, and low social visibility all decrease the importance and unpredictability of a purchase. Optimal perceived risk would be unlikely to be associated with products extremely low or high in all these attributes. That is, most consumers will prefer purchase alternatives at neither extreme of the perceived risk continuum.

This should be true whether risk is designated in a situation- or object-oriented manner. In the object-oriented approach (Sandell, 1968), individuals evaluate particular products in terms of the associated unpredictability and importance. The situation-oriented approach recognizes that certain situations can be expected to- alter uncertainty and/or importance across products. For instance, introduction of a new brand represents risk increase because the consumer's knowledge becomes incomplete to some degree. The addition of a guarantee usually increases the consumer's information regarding the product, thus representing a form of risk reduction.

If, as the risk-optimization process suggests, perceived risk and preference are positively related in purchase decisions involving extremely low risk and negatively related in decisions involving extremely high risk, then interaction between product and situation risk is implicit. Either the product or situation can be conceptualized as the determinant of the basic level of risk involved in a decision, with the other risk element providing increases and decreases relative to that level. For instance, if product risk is considered basic, then a positive relationship between situation risk and preference would he expected at low levels of product risk and a negative relationship at higher levels. Conversely, negative and positive relationships between preference and product risk would be expected in situations high and low in risk, respectively.

The following study examined preference as a function of perceived risk associated with both products and situations. The specific hypotheses entertained were: (1) purchase preference is curvilinearly related (inverted-U) to the amount of risk associated with a set of ordered product alternatives; (2) this relationship occurs whether risk is determined by the products or the information provided in the purchase situation; (3) product and situational determinants interact to affect preference.

METHOD

Overview

A written questionnaire which described three shopping situations was administered to two independent samples of female homemakers. Twenty products remained constant across these three situations, and ten questions measured the theoretical components of perceived risk associated with each of these products. Respondents role-played to indicate how they would react in each situation. The 20 products were selected from 56 pre-tested products and represented a wide total range of perceived risk with relatively small dispersion among judgments of risk associated with any single product.

Subjects

Each sample was randomly selected from a list of residences in each of two cities. Sample 1 was drawn from Richardson, Texas, a suburb of Dallas with a population of approximately 50,000. Sample 2 was drawn from Lafayette and West Lafayette, Indiana, two non-suburban cities with a combined population of approximately 100,000. In Sample 1, 118 respondents, or 88% of those contacted, returned questionnaires. In Sample 2, 111 returns represented 96% of the number distributed. Respondents' failure to follow instructions necessitated eliminating two questionnaires in each group.

Administration

Questionnaires were personally delivered to respondents' homes and retrieved three days later. The deliverer described the study as an academically-sponsored project, involving objective questions about consumer behavior. Each person who completed a questionnaire later received $2.00 in payment ant a letter explaining the purpose of the study.

Measures of Product Risk

Ten questions (see Table 1) were constructed to assess components of perceived risk for 20 products. These items were responded to on a 9-point scale in which high values on each scale indicated a high degree of danger or uncertainty. The order of products was randomized across questions; question order was randomized across subjects. Responses to selected questions were combined to form three composite measures. As defined by Cox (1967b, 1967c), perceived risk requires the combination of two general risk components; purchase-relevant uncertainty and consequences. Each risk component also has several dimensions. For comparison, each of the three composite measures emphasized somewhat different dimensions of the uncertainty and consequences components.

The first composite measure (CM-1) combined responses to two questions used in previous studies (Cunningham, 1967a, 1967b) and labeled A and B in Table 1. CM-1 was obtained by multiplying the individual's scores on these items, yielding a score range from 1-81.

In the second composite measure (CM-2), the uncertainty components emphasized individual-specific differences in ability to predict product attributes. Questions C, D, and E were combined as follows to give a measure with a range of 1-81: CM-2 = (C), (D + D)/2.

In the third composite measure (CM-3) of perceived risk, consequences were again represented by ratings of importance (D) and investment (B), as in CM-2.

The unpredictability component included the perceived unpredictability of product dependability with repeated use (F), product construction and materials (G), results of product failure (H3, and the degree (I) and kind (J) of goal fulfillment involved. The formula for obtaining CM-3 was: CM-3 = (D + E)/2. (F + G + H + I + J)/5.

TABLE 1

QUESTIONS MEASURING PERCEIVED RISK

A. How certain are you that a brand name of this product you haven't tried will work as well as your present brand?

B. We all know that not all products work as well 85 others; compared to other products, how much danger would you say there is in trying a brand of this product that you have never used before?

C. How confident would you say you are about judging the quality of the product?

D. Buying a product that gives you good results may be more important for some products listed than for others. How important would you say it is for this Product to satisfy you?

E. The investment you make when you buy a product includes your time and energy as well as money. In terms of the time, money, and overall effort required to buy this product, how much would you say you invest?

F. Can most shoppers guess ahead of time how dependable this product will be if it is used over and over again?

G. Before buying this product, can almost anyone tell how good its materials are and how well it's put together?

H. Can almost any shopper predict what the bad results will be if this product fails?

I. In general, does this product tend to fulfill your expectations?

J. Is it obvious why someone like yourself would want this product?

Measures of Situational Risk

Each respondent was given the opportunity to win one of the 20 products, except the four most expensive, as a prize in a drawing to be made from every 20 questionnaires. Respondents indicated on a 9-point scale how much they would like to receive each gift as a prize. Subsequent questions, ordered alternately across subjects, assessed product preferences in two situations: (1) products were offered with a full money-back guarantee against "any failure to perform satisfactorily during normal use"; (2) products were described as new brands, soon to be marketed for the first time. The guaranteed and new products represented, respectively, risk-reducing and risk-enhancing circumstances.

RESULTS

Composite Risk Measures

Products were ranked according to each of the composite measures of risk. Rankings of the two samples were closely related: the Spearman rank-order correlation between the two samples of CM-1 equalled .9323, and was high also for CM-2 (rs = .9097) and CM-3 (rs = .9301).

The mean of the two sample ranks for each product was calculated for each composite risk measure. The mean of the two sample ranks for each product was similar across the three risk measures: rs for CM-1 and CM-2 was .8925; for CM-1 and CM-3, rs = .8684; for CM-2 and CM-3, which used the same measure for the consequences component, rs = .8820. The mean of the three risk measures was then computed for each product and used to classify products into one of five perceived risk levels. Each level was relatively homogeneous in terms of numerical indices of perceived risk for the products included in that level. The final mean product ranks (the cross-measure mean of cross-group ranks) and corresponding risk measures appear in Table 2.

TABLE 2

RANK, MEAN PERCEIVED RISK AND FINAL RISK LEVEL OF 20 PRODUCTS IN COMBINED SAMPLES

Analysis of Product Preferences

Product preferences provided cell entry data for a 2 (Samples) X 3 (Situation Risk) X 5 (Risk Level) analysis of variance, with repeated measures on the second and third factors. The two samples were treated as the first, between-subject factor. An unweighted means solution was applied for unequal sample sizes (Winer, 1962, pp. 374-378).

The model for a repeated-measures analysis is appropriate if the variance-covariance matrices are equal and the pooled variance-covariance matrix is symmetrical (Winer, 1962, pp. 369--374). Each portion of the within-cell variation was tested for homogeneity of variance, using Hartley's F test. Insignificant results were obtained for variation due to interaction between subjects within groups and situation risk (Fmax .95 [2,223]), product risk max (Fmax .95[2.546]), or both product and situation risk (Fmax .95[2,892]). Complete assumptions about the variance-covariance matrices were not tested. Consequently, the analyses used conservative degrees of freedom (Greenhouse & Geisser, 1959). The conservative procedure is negatively biased, i.e., the critical value is over-large, leading to errors in the direction of not rejecting false null hypotheses.

Results of the 2 X 3 X 5 analysis of variance are summarized in Table 3.

TABLE 3

ANALYSIS OF VARIANCE OF PRODUCT PREFERENCE FOR PRODUCTS, SITUATIONS, AND SAMPLES

Sample main effect was insignificant; variation within risk treatments accounted for most (w2 =.78) of the total response variation. Situation and product risk factors each produced significant main effects (respectively, F = 184.655, df = 1,225, p < .0005; F = 47.257, df = 1,225, p < .0005). According to Newman-Keuls analysis of means representing main effects, preference decreased significantly (p < .01) between each of the first three product levels and changed significantly at higher levels. Preference in the guarantee situation was significantly (p < .01) greater than in the other two situations, which did not differ significantly.

As expected, both aspects of risk contributed to product preference. Interaction between product and situation risk factors was significant (F = 23.186, df = 1,225, p < .0005). The interaction (see Figure 1) was evaluated in terms of the effect of one risk factor at each single level of the other risk factor. Analysis of variance of simple product risk effects yielded significant results for the guarantee (F = 104.145, df = 1,225, p < .0005), free gift (F - 100.497, df = 1,225, p < .0005), and new brand situations (F = 268.947, df = 1,225, p < .0005). Within each of the latter two situations, mean preference decreased significantly (p < .01) between product risk levels 1 and 2, and levels 2 and 3, according to Newman-Keuls tests. At levels 3 and 4, means of preference for a free gift did not differ, and preference increased (p < .01) at level 5. Preference for a new brand increased (p < .01) between levels 3 and 4 and decreased (p < .01) at level 5. Within the guarantee situation, mean preference was highest for products at the lowest product risk level. Preference decreased at product level 2 and remained stable across levels 3, 4, and 5.

Analysis of simple effects of situation risk indicated significant (df = 1,225, p < .0005) effects at each level of product risk. All inter-mean comparisons of preference across situations at each level of product risk were significant (p < .05) in Newman-Keuls tests, with one exception. Preference at product risk level 2 differed insignificantly for the guaranteed and new products.

FIGURE 1

INTERACTION BETWEEN PRODUCT AND SITUATIONAL RISK

The hypothesized attractiveness of an optimal amount of risk implies higher preference at one risk level and lower preference at adjacent levels. Given a comprehensive range of risk levels, curvilinear relationship of preference to risk levels would be expected. Accordingly, trend analyses were made for the two risk main effects. Orthogonal coefficients were derived for unequal product risk intervals and the unequal sample sizes (Kirk, 1968, pp. 513-517). Situational risk intervals were assumed to be equal. For product risk, the linear, quadratic, and cubic trend components were significant (Flin = 480.6788, df = 1,56, p < .0005; Fquad= 142.5231, df = 1,56, p < .0005; Fcubic = 11.4949, df = 1,56, p < .0005). Contrary to expectation, the linear trend clearly dominated, contributing 85% of the variance. The trend of situational main effects was predominantly quadratic (X2 = .87), although both linear and quadratic trend components were significant (Flin = 822.9858, df = 1,112, p < .005 Fquad = 136.5222. df = 1,112, p < .0005).

Trend analyses were conducted to comPare the pattern of effects of one risk factor at each level of the other factor. Linear, quadratic, cubic, and quartic components of simple product risk effects were significant (df = 1,56, p < .005) at each level of situation risk. The quadratic component of product risk accounted for more of the treatment variance (w2 = .55) in the guarantee situation. In the gift and new brand conditions, the linear trend component dominated (w2 = .66; w2 = .89, respectively). The relative contribution of the quadratic component decreased markedly between the gift (w2 = .28) and the new brand conditions (w2 = .01). The pattern of product risk effects thus tended to differ at different levels of situational risk.

Trend analysis of simple situation risk effects produced significant (df = 1,112, p < .005) linear and quadratic components at all levels of product risk. The linear trend component dominated at the lowest product risk level (w2 = .88). Quadratic trend accounted for more treatment variance at levels 2 and 3 (w2 = .76; w2 = .85, respectively). At product risk levels 4 and 5, linear and quadratic components contributed equally (w2lin = .49; w2quad = .51, at level 4; w2lin = .55 and w2quad = .45 at level 5).

In the 2 X 3 X 5 analysis, all interactions involving the samples were significant (see Table 2). To clarify the effect of risk factors within each sample, simple main effects for each risk factor, and simple interactive effects for the product-situation combination were calculated. Simple product risk effects were significant for both Sample 1 (F = 106.044, df = 1,116, p < .0005) and Sample 2 (F = 158.561, df = 1,109, p < .0005), as were simple situation risk effects (for Sample 1, F = 63.334, df = 1,116, p < .0005, for Sample 2, F = 60.409, df = 1,109, p < .0005). Simple product-situation interaction occurred for Samples 1 and 2 (respectively, F = 10.065, df = 1,116, p < .001; F = 41.480, df = 1,109, p < .0005). Situation and product risk and their interaction influenced product preference within as well as across the two samples. Additionally, trend analysis of the simple main risk effects yielded no notable deviations from the trends reported for the combined samples. The trends for product and situational risk effects were attributable primarily to linear and quadratic components, respectively, within each sample.

Although the effect of each risk factor was similar across samples, interaction between the risk factors varied somewhat with samples (Figure 2).

Analysis of variance of simple interaction between samples and product risk indicated significant interaction in the gift (F = 12.460, df = 1,225, p < .001). and new brand situations (F = 12.460, df = 1,225, p < .001). For the latter situation, Newman-Keuls analysis indicated significantly (p < .05) greater preference for Sample 1 for product risk levels 2 and 3, and significantly (p < .001) less preference at product risk level 5. A single significant inter-mean difference occurred in the free gift situation. Preference for Sample 2 was significantly (p < .05) greater at product risk level 3. The preceding significant inter-mean differences were sufficient to give significant simple interaction between samples and situational risk at product risk levels 2, 3, 4, and 5.

DISCUSSION

Consumers' product selections provided strong support for the hypothesized interaction between situational and product determinants (Hypothesis 3). Additionally, each determinant alone also affected product preference, as posited by Hypothesis 2. The latter hypothesis also stipulated a curvilinear relationship between preference and the continuum of risk associated with a set of ordered alternatives. However, the data indicated that preference tended to decrease with increasing product risk and then to stabilize at the higher risk levels. Similarly, the guarantee situation elicited the greatest preference, but preference did not differ between the well-known and new brand situations.

FIGURE 2

INTERACTION OF SITUATION RISK, PRODUCT RISK, AND SAMPLE

Although the predicted curvilinear relationship did not occur for situational and product determinants separately, this does not preclude the possibility of the predicted relationship occurring for combined situational and product risk. However, as joint determinants, product and situation risk interacted to elicit greatest preference when both risk components were small. Preference decreased slightly as both risk components increased, with the rate of decrease being greater for the higher risk situations. In contrast, a risk-optimization process should produce increases and subsequent decreases in preference (i.e., an inverted-U function) with increases in perceived risk.

Despite the apparent simplicity of the overall interaction between risk components, the underlying relationships were complex. Within the "safest" (i.e., guaranteed products) situation, preference stabilized at a low level of product risk. In the intermediate risk (i.e., well-known brand) situation, product risk and preference showed a consistent negative relationship. And within the riskiest (i.e., new brand) situation, important sample differences emerged. One sample behaved much as it had within intermediate risk. In the other sample, the relationship between preference and perceived risk tended toward a U-shaped pattern.

Sample differences were one of the more interesting findings of the study. One sample of consumers responded to greater situational risk with decreases in preference, except when the products involved represented very high risk. With these products, preference was lowest in the intermediate risk situation and high when either a guaranteed or new brand of product was offered. The second sample provided this same U-shaped function for situation risk when products were intermediate (levels 2 and 3) in risk and a negative risk-preference relationship for other products. It is notable that risk enhancement in each case involved selection of a new product brand rather than particular products representing higher risk.

To clarify sample differences, several measures of search for information and of emotional response to purchase situations were examined for each sample. The sample with risk-enhancing preferences (e.g., selection of new products) at the highest product level experienced significantly less worry, relative to the other sample, when shopping for products at that level. In contrast, risk-enhancing preferences at product level 3 were associated with less desire for increased product variety at that level. It appears that those who enhanced risk at a particular level were relatively more satisfied, in terms of the variety offered or associated worry, with risk at that level. Additionally, those who enhanced risk at a lower product level seemed to attribute greater risk to the risk-enhancing situation. They were more likely to read and deliberate about purchasing a new brand. Despite the specification of product and situational risk factors, individual and group differences in _interpreting product characteristics persisted.

Identifying consumers who would enhance risk at a relatively low or high risk level would be useful in anticipating market response to product changes. In the present study, age information was available for homemakers in the samples. Those who enhanced risk at a relatively low level were equally represented across all intervals from 20 to 80 years. Risk enhancement at higher levels was more likely when the homemakers were between 20 and 40 years of age. Since these years tend to require the greatest homemaking activity, this may accustom the consumer to a higher level of risk. Attempts to enhance risk may center about that higher level.

That risk enhancement occurred at certain risk levels and for certain groups suggests two areas for future research. First, it is possible that risk enhancement centers on a range of perceived risk, with risk reduction behavior occurring both below and above that range. This implies that not all products contribute equally to the risk-optimization process. With products representing very low risk, the effort involved in selecting new products may not be justified by the transient risk enhancement.

The second area concerns identification of groups who respond similarly to perceived risk. Self-confidence, both product-specific and general, is one psychological variable that has been used with some success (Barach, 1969). Although other variables such as cognitive complexity, category width, and tolerance for ambiguity should be examined, certain experiential measures may be equally helpful. Making many purchases perceived as high in risk, or a relatively high level of shopping and other homemaking activities, should, over time, accustom a consumer to a relatively high level of perceived risk. The amount of risk perceived in a purchase situation would be established relative to this higher norm.

For individuals in both samples, the norm of risk experienced in actual shopping activities may have obscured the relationship between perceived risk and product preference. The highest obtained average value for perceived risk (X - 36.76) was less than half the range on the scales provided. Purchases incorporating greater perceived risk may occur. For example, the monetary commitment required for home furnishings or insurance probably exceeds that of any of the products studied, and social implications may be equally as great. If such purchases were included for consideration, it is possible that they would elicit the decrease in preference predicted for extremely high risk. The generally negative relationship found between risk and preference may thus reflect an insufficient range of perceived risk.

An accurate a priori estimate of the range of risk associated with products may be difficult. Highest perceived risk was associated with purchasing a car, car tires, hair spray, and a headache remedy. Risk apparently derives from diverse consequences of product use. Jacoby and Kaplan (1972) detected five varieties of perceived risk: functional, physical, social, financial, and psychological. Relative contribution of each variety to overall perceived risk varied across products. Such diversity in components of risk may affect the relationship between risk and product preference or search.

Product preference is one of many ways of handling perceived risk (Roselius, 1971). Both the relevance of information to particular product characteristics and the sources of information may vary with perceived risk. Activities which precede and follow product choice must also be examined to evaluate the possibility of risk optimization.

REFERENCES

Abelson, R. P., Aronson, E., McGuire, W. M., Newcomb, T. M., Rosenberg, M. J., & Tannenbaum, P. H. (Eds.), Theories of cognitive consistency: A source book. Chicago: Rand-McNally, 1968.

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

Barach, J. A. Advertising effectiveness and risk in consumer decision process. Journal of Marketing Research, 1969, 6, 314-333.

Bauer, R. A. Consumer behavior as risk taking. In R. S. Hancock (Ed.), Dynamic marketing for a changing world. Chicago: American Marketing Association, 1960.

Berlyne, D. E. Uncertainty and conflict: A post of contact between information theory and behavior theory concepts. Psychological Review, 1957, 64, 329-339.

Berlyne, D. E. Conflict, arousal, and curiosity. New York: McGraw-Hill, 1960.

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

Copley, T. P., & Callom, F. L. Industrial search behavior and perceived risk. Proceedings, Second Annual Convention of the Association for Consumer Research, 1971, 208-31.

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

Cox, D. F. Introduction. In D. F. Cox (Ed.), Risk-taking and information- handling in consumer behavior. Boston: Howard University Press, 1967. (b)

Cox, D. F. Risk handling in consumer behavior--Intensive study of two cases. In D. F. Cox (Ed.), Risk-taking and information-handling in consumer behavior. Boston: Harvard University Press. 1967. (c)

Cox, D. F., & Rich, S. U. Perceived risk and consumer decision making--The case of telephone shopping. Journal of Marketing Research, 1964, 1, 32-39.

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. (a)

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

Frank, R. E. Is brand loyalty a useful basis for market segmentation? Journal of Advertising Research, 1967, 7, 27-33.

Greenhouse, S. W., & Geisser, S. On methods in the analysis of profile data. Psychometrika, 1959, 24, 95-112.

Howard, J. A., & Sheth, J. N. The theory of buyer behavior. New York: John Wiley, 1969.

Jacoby, J., & Kaplan, L. B. The varieties of perceived risk. Purdue University, Purdue Papers in Consumer Psychology. Paper No. 118, 1972.

Kirk, R. E. Experimental design: Procedures for the behavioral sciences. Belmont, California: Brooks/Cole Publishing, 1968.

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

Popielarz, D. T. An exploration of perceived risk and willingness to try new products. Journal of Marketing Research, 1967, 4, 368-373.

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

Sandell, R. G. Effects of attitudinal and situational factors on reported choice behavior. Journal of Marketing Research, 1968, 5, 405-408.

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

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

Swan, J. E. Experimental analysis of predecision information seeking. Journal of Marketing Research, 1969, 6, 192-297.

Winer, B. J. Statistical principles in experimental design. New York: McGraw-Hill, 1962.

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