Base Rate Information, Causal Inference, and Preference

Frank R. Kardes, Massachusetts Institute of Technology
[ to cite ]:
Frank R. Kardes (1988) ,"Base Rate Information, Causal Inference, and Preference", in NA - Advances in Consumer Research Volume 15, eds. Micheal J. Houston, Provo, UT : Association for Consumer Research, Pages: 96-100.

Advances in Consumer Research Volume 15, 1988      Pages 96-100

BASE RATE INFORMATION, CAUSAL INFERENCE, AND PREFERENCE

Frank R. Kardes, Massachusetts Institute of Technology

[The author wishes to thank David M. Sanbonmatsu for his helpful comments on an earlier draft.]

ABSTRACT -

Although Kelley's (1967) attribution theory was initially designed to explain inference processes in person perception, the present study demonstrates that the theory also has direct implications for product perception In addition, the effects of distinctiveness, consistency, and consensus (base rate) information on inferences about products, and the link between inference and Preference are examined.

INTRODUCTION

Consumers are exposed to the product-related opinions and actions of other consumers every day Although some research has been conducted on the effects of the product-related opinions of others on product judgments (e.g., word-of-mouth effects, see Arndt 1967, Bauer 1967, Reingen and Kernan 1986, Richins 1983), the effects of the product-related actions of others on product judgments have been neglected. This is unfortunate because actions often speak louder than words (Amabile and Kabat 1982). The present pilot study explores some effects of the product-related actions of others on product judgments.

One theoretical framework that seems particularly well-suited for addressing this issue is Kelley's (1967, 1972a, 1972b, 1973) attribution theory According to Kelley, the inferences that people draw about the actions of others are contingent upon three types of information: consistency information (e.g., To what extent does the actor's behavior toward an object generalize across situations), distinctiveness information (e g, To what extent does the actor's behavior toward an object generalize across objects?), and consensus information (e g., To what extent does the actor's behavior toward an object generalize across people?). The patterns observed among these three types of information have been shown to have implications for judgments about the extent to which the actor's behavior is attributed to his or her personal traits and characteristics, as opposed to other factors (e.g., some situational variable may have caused the actor's behavior).

Although Kelley's model was initially designed to explain inference processes in person perception, many researchers have shown that the theory has interesting implications for consumer choice (Calder and Burnkrant 1977, Golden 1977, Hansen and Scott 1976, Mizerski et al. 1979, Settle 1972, Settle and Golden 1974, Sparkman and Locander 1980) For example, several studies have shown that the inferences consumers draw about a spokesperson's endorsement of a given product in an ad can exert a strong influence on product evaluations More favorable judgments of a product are formed when consumers infer that the endorser's behavior was prompted by dispositional factors (e.g., the endorser honestly believes that the product is a superior product) than when consumers infer that the endorser's behavior was situationally-guided (e g, the endorser is being compensated very generously). Thus, causal inference can exert an indirect influence on product perception by first influencing person perception.

However, it seems likely that causal inference can also have a direct impact on product perception Consider the case in which a consumer observes another consumer (the actor) buying a particular brand of beer Characteristics of the brand (rather than of the actor) are likely to be seen as the principle cause of the actor's behavior (a) if the actor frequently (rather than infrequently) buys this brand, (b) if the actor buys only this brand (rather than buying many different brands), or (c) if many people (rather than only a few) buy this brand The above three examples imply that consistency, distinctiveness, and consensus information, respectively, may have direct and independent effects on inferences about the extent to which the characteristics of the target brand (as opposed to other factors) cause an actor to purchase the brand However, Kelley (see Orvis, Cunningham, and Kelley 1975) argues that the three types of information are used configurally That is, when one type of information is missing, its value may be inferred on the basis of the remaining two types of information, and then all three types of information may be integrated to form an attributional judgment

Wyer and Carlston (1979), however, disagree with Kelley's notions on configural information processing in attribution They argue that each type of information is useful even in the absence of knowledge about the remaining types. Moreover, they argue that the interactive effects of consistency, distinctiveness, and consensus information that have been observed in the classic tests of Kelley's model derive from the confounding of the three types of information in the stimulus materials. For example, consider the statements "John laughs at the comedian" and "John does not laugh at almost any other comedian" (McArthur 1972, 1976; Orvis et al. 1975). These statements are not explicitly tied to a given situation and may be interpreted as occurring either in a limited domain or across a wide variety of situations. To avoid this confound in the present study, all three types of information were provided explicitly and one and only one type was manipulated at a time

The Base Rate Fallacy

Although a great deal of empirical evidence indicates that consistency and distinctiveness information affect attributional judgments in a normatively prescribed manner (for reviews see Harvey and Weary 1984, Kelley and Michela 1980, Mizerski et al. 1979), several studies have shown that, under some circumstances, individuals tend to underutilize consensus information (Cooper et al. 1972; McArthur 1912, 1976; Miller et al. 1974; Nisbett and Borgida 1975; Nisbett et al. 1976)

For example, Nisbett and Borgida (1975) described two classic social psychological studies (Darley and Latane 1968, Milgram 1963) to subjects and asked them to explain why a particular participant in these studies behaved in an unexpected manner Prior to making this attributional judgment, subjects in consensus information conditions were informed that most of the participants behaved in this unexpected manner (precise distributional information was provided), whereas subjects in no consensus information conditions did not receive this information. The results indicated that consensus information had no influence on subjects' attributional judgments: extreme dispositional attributions were formed in all conditions

Nisbett and Borgida suggested that the failure to use consensus information in forming causal inferences reflects a general tendency for individuals to underutilize base rate information when assigning objects to categories (Tversky and Kahneman 1973). Although there is considerable evidence indicating that people often underutilize base rate information (for reviews see Kahneman et al. 1982, Nisbett and Ross 1980), there is also a growing literature on the conditions that moderate the use of base rate information (for reviews see Bar-Hillel 1980, Borgida and Brekke 1981, Kassin 1979, Ofir and Lynch 1984). For example, individuals are more likely to use base rate information (a) when it is presented in a concrete - rather than in an abstract-fashion (Borgida and Nisbett 1977), (b) when the social -desirability of this information is high rather than low (Zuckerman 1978), (c) when this information is not accompanied by individuating information (e g., information that highlights the idiosyncratic characteristics of the actor, Zuckerman 1978), and (d) when base rate information is presented in a within-subjects context (Fischhoff and Bar-Hillel 1984)

Finally, (e) it seems that base rate information is likely to be used in consumer settings. The product-related actions of others provide a very rich and useful source of information about products and this principle is widely applied by marketing practitioners: Advertisers often tell us that their products are the "fastest growing' or the "largest selling;" salespeople often display testimonials from satisfied customers; and fundraisers often provide long lists of names of donors to potential donors (Cialdini 1985, Kernan and Reingen 1985).

A Pilot Study

All five of the above conditions were met in the present study, and, hence, it was predicted that base rate information should influence attributions in the manner prescribed by Kelley Following Kelley, it was predicted that subjects should attribute the cause of the actor's behavior primarily to the characteristics of the target brand (as opposed to other factors) when consistency, distinctiveness, or consensus is high as opposed to low. Further, the relationship between causal inference and preference was examined.

METHOD

Subjects and Design

Fifty-three undergraduates were asked to complete a questionnaire pertaining to how consumers draw conclusions about different brands of beer on the basis of very limited information Eight of these students were non-beer drinkers and were excluded from the analysis. Subjects were randomly assigned to one of three conditions in which either situation (consistency), brand (distinctiveness), or person (consensus) generalization information was manipulated (n per cell = 15).

Stimulus Materials

In the high generalization across situations condition (high consistency), subjects read that "Dave buys Brand I (but not Brand J or Brand L) whether or not there's a sale" and "Jack does not buy Brand I" and "John does not buy Brand I."

In the high generalization across brands condition (low distinctiveness), subjects were told that "Dave buys Brand I and Brand J and Brand L when there's a sale (but not when there's not a sale)" and "Jack does not buy Brand I" and "John does not buy Brand I."

In the high generalization across people condition (high consensus), subjects learned that "Dave buys Brand I" (but not Brand J or Brand L) when there's a sale (but not when there's not a sale)" and "Jack buys Brand I" and "John buys Brand I." In addition, all subjects received low generalization information pertaining to a second target brand (Brand 3).

Although the stimuli described above are rather artificial, they have several desirable qualities: (a) novel stimuli are useful for controlling for prior knowledge and experience effects, (b) unlike previous studies, no abstract, implicit qualifiers (e g, qualifiers such as "some" or "most") are used to manipulate generalization information, (c) buying beer is socially acceptable, (d) no individuating information is provided, and (e) the above procedure permits within-subjects comparisons.

Manipulation Check, Causal Inference, and Preference Measures

After reading the generalization information, subjects were asked to indicate how many people buy the target brand (Brand 1), how many brands the actor (Dave) buys, and how frequently the actor buys the target brand on three 11-point semantic differential scales ranging from 0 (none) to 10 (all).

Next subjects were asked to judge the extent to which the characteristics of the target brand (e.g., its taste, its quality), caused the actor to make the purchase Henceforth, this type of judgment will be referred to as a brand attribution. Brand attributions were measured on an 11-point scale ranging from 0 (the target brand's characteristics are irrelevant [other factors are much more important] to 10 (the target brand's characteristics are the sole cause of the actor's behavior)

Subjects were also asked to judge the extent to which the actor's personal characteristics (his personal tastes and preferences) and the extent to which situational factors caused his behavior Again, 11-point scales with detailed information presented at the endpoints were employed. Finally, subjects were asked to indicate which brand they personally would prefer, Brand I or Brand J.

RESULTS

Manipulation Checks

Generalization judgments are presented in Table 1 As Table 1 indicates, the situation generalization (consistency) and the person generalization (consensus or base rate) manipulations were highly effective. However, the brand generalization (distinctiveness) manipulation was very weak (but the pattern of results was in the anticipated direction).

Causal Inference

Brand attributions as a function of situation, brand, and person generalization information are presented in Table 2 As Table 2 indicates, more extreme brand attributions (e g, more extreme beliefs that characteristics of the target brand caused the actor to make the purchase) were formed in high (c g., the actor buys the target brand across situations) than in low situation generalization conditions, as predicted. Moreover, subjects tended to form more extreme brand attributions in low (e g., the actor buys only the target brand) than in high brand generalization conditions. Thus, subjects tended to use situation and brand generalization information in a manner consistent with Kelley's principles.

TABLE 1

GENERALIZATION JUDGMENTS

However, base rate (person generalization) information had only a marginally significant effect on brand attributions Subjects tended to form more extreme brand attributions in low (e g., only the actor buys the target brand) than in high (e g, the actor and others buy the target brand) person generalization conditions. This pattern is opposite to the pattern prescribed by Kelley's model.

Preference

Eighty percent of the subjects preferred the target brand in high than in low situation generalization conditions, chi-square = 5 40, p < .02. Moreover, more extreme brand attributions were formed in high than in low situation generalization conditions. Hence, in situation generalization conditions, subjects' preferences corresponded closely to their causal inferences.

Brand generalization information, on the other hand, had no influence on subjects' preferences, chi-square = .07. This finding should be interpreted cautiously, however, because the brand generalization manipulation was very weak.

Although base rate information did not have a strong influence on brand attributions, it had a strong impact on subjects' preferences, chi-square, 5.40, p < 02. Eighty percent of the subjects preferred the target brand in high than in low person generalization conditions This finding suggests that little correspondence exists between causal inference and preference in person generalization conditions.

Correlational Analyses

According to Wyer and Carlston (1979), judgments of generalization across situations (x), objects (o), and persons (p) are integrated into brand attributions (JO) in the following manner JO= x - o + p. The present data permitted a test of this model. Predicted values of JO were derived from subjects' generalization judgments The relation between predicted and obtained values of JO was statistically significant, r = .54, p < 001. The predicted situation attributions (Jx = -x + o + p) were also significantly related to observed values, r = .53, p < .001. However, base rate information was not used in the predicted manner (Jp= x + o - p), r = - 02, ns Again, the results indicate that people fail to use base rate information in a normatively appropriate manner.

TABLE 2

BRAND ATTRIBUTIONS

DISCUSSION

The results indicate that brand and situation generalization information can have direct and independent effects on consumers' brand attributions. Moreover, causal inferences based on situation generalization information seem to translate directly into preferences. Further, it should be emphasized that care was taken to ensure- that subjects would be sensitive to base rate information: (a) base rate information was presented in a very concrete fashion, (b) the actor's behavior was not socially undesirable, (c) no individuating information about the actor was provided, (d) base rate information was manipulated in a within-subjects context, and (e) base rate information was presented in a consumer setting Nevertheless, subjects clearly underutilized base rate information in their attributional judgments

Why was base rate information underutilized? One possibility is that subjects may have failed to attend to this information. According to Trope's (1986) two-stage model of attribution, people must first attend to information (the identification stage) before they can use it (the inferential stage). Trope maintains that people do use base rate information when they attend to it The problem is they often do not attend to it. However, this was clearly not the case in the present study The manipulation check indicated that subjects did attend to base rate information Nonetheless, this information was underutilized.

A second possibility is that subjects may have employed the representativeness heuristic (Kahneman and Tversky 1973) in order to simplify the judgment task. However, this possibility seems unlikely for two reasons: (a) very little informal ion was provided to subjects and, hence, information reduction may have been unnecessary, and (b) no individuating information was presented.

Nisbett et al. (1976) offered an alternative to Kahneman and Tversky's representativeness explanation According to Nisbett et al, individuating information is more concrete and more vivid tan base rate information and, consequently, individuating information is weighted much more heavily than base rate information. However, Nisbett et al.'s vividness explanation cannot account for the results of the present study, because base rate information was presented in a very concrete and vivid manner and because no individuating information was provided.

The "relevance" explanation (Bar-Hillel 1980, Ofir and Lynch 1984) has recently been offered as an alternative to the representativeness and vividness explanations People may ignore base rate information simply because they fail to see the implications of this information for the judgment task at hand Rather than attempting to apply information that they do not know how to apply, they ignore it.

How can we determine what information is relevant for a given judgment on an a priori basis? What are the defining features of information relevance? According to Kelley, there are three types of attribution-relevant information: consistency, distinctiveness, and consensus information However, consensus information is often treated as if it were irrelevant. Bar-Hillel suggests that information is ordered in terms of perceived relevance and dominated items are ignored If consistency and distinctiveness information are perceived as more relevant than consensus information, then subjects may ignore consensus information even if they believe that it is somewhat relevant.

One factor that may influence perceived relevance is information specificity (Bar-Hillel 1980, Ofir and Lynch 1984) If information about a product class and information about a specific subset of that product class is available, the latter type of information should be more relevant for judgments about a specific brand within that subset Clearly, it should be easier to see the implications of a given piece of information for a given judgment when the information and the judgment are at the same level of specificity. However, in the present study, base rate information was underutilized even though very specific information was provided

Yet another possibility is that instead of failing to see the implications of the base rate information for the judgment task at hand, subjects may have seen multiple, opposing implications As a consequence, several opposing psychological forces may have been operating simultaneously For example, popularity appeals may foster favorable inferences about the attributes of the target brand. After all, products that are sought after by many people must possess some desirable characteristics Moreover, consumers may learn that it is often "safe" to follow the lead of others On the other hand, products that can be owned by anyone are often less attractive than products that are only for a select few People like to think of themselves as being discriminating and unique, and these characteristics are reflected in their desire to purchase products that are, in some sense, unique or uncommon

Finally, the present data suggest that current notions about the boundary conditions of the base rate fallacy are underconceptualized Although a rather lengthy list of variables have been identified as moderators of the base rate fallacy, the processes that underlie this phenomenon are not well understood. What factors lead consumers to view a given piece of information as causally relevant? Once an item of information has been identified as causally relevant, how is this item integrated with other pieces of relevant information? Although Kelley maintains that causally relevant information is processed configurally, the present data suggest that different types of generalization information can be processed independently. What conditions determine whether configural versus elementistic information processing will occur in causal inference? These and other fundamental questions about consumer inference processes are left for future research.

REFERENCES

Amabile, T M and L G Kabat (1982), "When Self Descriptions Contradict Behavior: Actions Do Speak Louder than Words," Social Cognition, 1, 311-335.

Arndt, Johan (1967), "Role of Product-Related Conversations in the Diffusion of a New Product," Journal of Marketing Research, 4 (August), 292.

Bar-Hillel, Maya (1980), "The Base-Rate Fallacy in Probability Judgments," Acta Psychologica, 44 (3), 211-233

Bauer, Raymond A. (1967), "Source Effects and Persuasibility: A New Look," in Risk Taking and Information Handling in Consumer Behavior, ed Donald F. Cox, Boston: Harvard Business School, 559-578.

Borgida, Eugene and Nancy Brekke (1981), "The Base Rate Fallacy in Attribution and Prediction," in New - Directions in Attribution Research, Vol. 3, eds John H. Harvey, William J Ickes, and Robert F Kidd, Hillsdale, NJ Lawrence Erlbaum, 63-95.

Borgida, Eugene and Richard E Nisbett (1977), "The Differential Impact of Abstract vs Concrete Information on Decisions," Journal of Applied Social Psychology, 7 (July), 258-271.

Calder, Bobby J. and Robert E Burnkrant (1977), "Interpersonal Influence on Consumer Behavior An Attribution Theory Approach," Journal of Consumer Research, 4 (June), 29-38

Cialdini, Robert B (1985), Influence: Science and Practice, Glenview, IL: Scott, Foresman and Company.

Cooper, Joel, Edward E Jones, and S. Mark Tuller (1972), "Attribution, Dissonance and the Illusion of Uniqueness," Journal of Experimental Social Psychology, 8 (January), 45-47.

Darley, John M and Bibb Latane (1968), "Bystander Intervention in Emergencies: Diffusion of Responsibility," Journal of Personality and Social Psychology, 8 (April), 377-383.

Golden, Linda L (1977), "Attribution Theory Implications for Advertisement Claim Credibility," Journal of Marketing Research, 14 (February), 115-117

Hansen, Roben A. and Carol A Scott (1976), "Comments on Attribution Theory and Advertiser Credibility," Journal of Marketing Research, 13 (May), 193-197

Harvey, John H. and Gifford Weary (1984), "Current Issues in Attribution Theory and Research," in Annual Review of Psychology, Vol 35, eds Mark R. Rosenzweig and Lyman W Porter, Palo Alto, CA: Annual Reviews, 427-459.

Kahneman, Daniel and Amos Tversky (1973), "On the Psychology of Prediction," Psychological Review, 80 (October), 237-251.

Kahneman, Daniel, Paul Slovic, and Amos Tversky, eds (1982), Judgment Under Uncertainty: Heuristics and Biases, Cambridge: Cambridge University Press.

Kassin, Saul M. (1979), "Consensus Information, Prediction, and Causal Attribution: A Review of the Literature and Issues," Journal of Personality and Social Psychology, 37 (November), 1966-1981

Kelley, Harold H. (1967), "Attribution Theory in Social Psychology," in Nebraska Symposium on Motivation, Vol 15, ed David Levine, Lincoln, NB: University of Nebraska Press, 192-238.

Kelley, Harold H. (1972a), "Attribution in Social Interaction," in Attribution: Perceiving the Causes of Behavior, eds Edward E Jones, David E. Kanouse, Harold H Kelley, Richard E. Nisbett, Stuart Valins, and Bernard Weiner, Morristown, NJ General Learning Press, 1-26.

Kelley, Harold H. (1972b), "Causal Schemata and the Attribution Process," in Attribution: Perceiving the Causes of Behavior, eds. Edward E Jones, David E. Kanouse, Harold H. Kelley, Richard E Nisbett, Stuart Valins, and Bernard Weiner, Morristown, NJ: General Learning Press, 151-174

Kelley, Harold H. (1973), "The Processes of Causal Attribution," American Psychologist, 28 (February), 107-128.

Kelley, Harold H. and John L Michela (1980), "Attribution Theory and Research," in Annual Review of Psychology, Vol. 31, eds Mark R. Rosenzweig and Lyman W Porter, Palo Alto, CA: Annual Reviews, 457-501

Kernan, Jerome B. and Peter H. Reingen (1985), "Behavioral Influence: A New Look in Persuasion Research," in Research in Consumer Behavior, ed. Jagdish N Sheth, Greenwich, CT: JAI Press, 159-199

McArthur, Leslie (1972), 'The How and What of Why: Some Determinants and Consequences of Causal Attribution," Journal of Personality and Social Psychology, 22 (May), 171-193.

McArthur, Leslie (19-76), "The Lesser Influence of Consensus than Distinctiveness Information on Causal Attributions: A Test of the Person-Thing Hypothesis," Journal of Personality and Social Psychology, 33 (June), 733-742.

Milgram, Stanley (1963), "Behavioral Study of Obedience," Journal of Abnormal and Social Psychology, 67 (4), 371-378

Miller, Arthur C., Barry Gillen, Charles Schenlter, ant Shirley Radlove (1974), "The Prediction ant Perception of Obedience to Authority," Journal of Personality, 42 (March), 371-378.

Mizerski, Richard W., Linda L. Golden, and Jerome B. Kernan (1979), "The Attribution Process in Consumer Decision Making," Journal of Consumer Research, 6 (September), 123-140.

Nisbett, Richard E. and Eugene Borgida (1975), "Attribution and the Psychology of Prediction," Journal of Personality and Social Psychology, 32 (November), 932-943.

Nisbett, Richard E., Eugene Borgida, Rick Crandell, and Harvey Reed (1976), "Popular Induction: Information is not Necessarily Informative," in Cognition and Social Behavior, eds John S. Carroll and John W. Payne, Hillsdale, NJ: Lawrence Erlbaum, 113-133.

Nisbett, Richard E. and Lee Ross (1980), Human Inference: Strategies and Shortcomings of Social Judgment, Englewood Cliffs, NJ: Prentice-Hall

Ofir, Chezy and John G. Lynch (1984), "Context Effects on Judgment Under Uncertainty," Journal of Consumer Research, 11 (September), 668-679

Orvis, Bruce R., John D. Cunningham, and Harold H. Kelley (1975), "A Closer Examination of Causal Inference: The Roles of Consensus, Distinctiveness, and Consistency Information," Journal of Personality and Social Psychology, 32 (October), 605-616.

Reingen, Peter H and Jerome B. Kernan (1986), "Analysis of Referral Networks in Marketing: Methods and Illustration," Journal of Marketing Research, 23 (November), 370-378.

Richins, Marsha L (1983), "Negative Word-of-Mouth by Dissatisfied Consumers: A Pilot Study," Journal of Marketing, 47 (Winter), 68-78.

Settle, Robert B. (1972), "Attribution Theory and Acceptance of Information," Journal of Marketing Research, 9 (February), 85-88.

Settle, Robert B. and Linda L. Golden (1974), " Attribution Theory and Advertiser Credibility," Journal of Marketing Research, 11 (May), 181-185

Sparkman, Richard M and William B. Locander (1980), "Attribution Theory and Advertising Effectiveness," Journal of Consumer Research, 7 (December), 219-224.

Trope, Yaacov (1986), "Identification and Inferential Processes in Dispositional Attribution," Psychological Review, 93 (July), 239-257

Tversky, Amos and Daniel Kahneman (1973), "Availability: A Heuristic for Judging Frequency and Probability," Cognitive Psychology, 5 (September), 207-232.

Wyer, Robert S and Donal E. Carlston (1979), Social Cognition, Inference, and Attribution, Hillsdale, NJ: Lawrence Erlbaum.

Zuckerman, Miron (1978), "Use of Consensus Information in Prediction of Behavior," Journal of Experimental Social Psychology, 14 (March), 163-171.

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