Special Session Summary Relative Salience in Consumer Decisions

Citation:

France Leclerc (2002) ,"Special Session Summary Relative Salience in Consumer Decisions", in NA - Advances in Consumer Research Volume 29, eds. Susan M. Broniarczyk and Kent Nakamoto, Valdosta, GA : Association for Consumer Research, Pages: 59-61.

Advances in Consumer Research Volume 29, 2002     Pages 59-61

SPECIAL SESSION SUMMARY

RELATIVE SALIENCE IN CONSUMER DECISIONS

France Leclerc, University of Chicago

SESSION OVERVIEW

The relative salience of the information used in a decision has been shown to impact the outcome of a choice in a number of ways (e.g., Tversky, Sattah and Slovic 1988). In this session, two new factors expected to alter the relative salience of information and as such the choice outcome, are discussed.

In the first paper, Leclerc et al. propose that the status of an item within a category is more salient than the status of the category. For example, everything else being equal, people should prefer a high-position member in a low-position category (e.g., a good paper in a bad journal, a good model of a low-status brand, etc.) than a low-position member in a high-position category (e.g., a bad paper in a top journal, a bad model of a top brand, etc). The authors provide evidence supporting this conjecture in the "judgment and choice" domain as well as in the perceptual domain.

The second factor proposed by Hsee and Rottenstreich is whether the object under consideration is affect-poor or affect-rich. The authors argue that some goods are associated with vivid imagery and keen sensation and evoke "hot" affect whereas others are relatively pallid or "cold". The authors show that compared to affect-poor objects, affect-rich objects activate evaluation by feeling and are therefore less sensitive to quantity.

In their work, Kahneman et al. report that when presented in isolation, a good evokes a context of similar objects and as such will be evaluated relative to the category it evokes. However, when objects from different categories are presented (e.g., physical injury and financial injury), then the category membership becomes salient and is the basis of the comparison. The research in this third paper is in a different setting, namely punitive damages, and willingness to pay for public goods.

In looking at these three papers together, the paper by Kahneman et al. can be thought of as providing additional evidence for the salience of the position of an item within a category proposed by Leclerc et al. Kahneman and his colleagues show that this is the case as long as the object is evaluated in isolation or (as shown by Leclerc et al.) if there other categories involved, the categories are relatively similar (e.g., models within a brand). However, when the categories compared are dissimilar such as physical injuries and financial injuries, or to use the Hsee and Rottenstreich’s distinction when one is affect-rich and the other is affect-poor, then the category membership become the basis of the comparison.

 

"BEST OF THE WORST OR WORST OF THE BEST!"

France Leclerc, University of Chicago

Christopher K. Hsee, University of Chicago

Joseph C. Nunnes, University of Southern California

Suppose that Mr. L is in the market for a digital camera. He consults a specialized magazine and finds out that there are currently four brands of digital cameras on the market and that each brand has four different models. The magazine is also providing information on the overall quality of all sixteen existing products. Among the 16 products, two of them have the same quality level. One of those models is the lowest quality model of Brand A whereas the other is the top quality model of Brand B. The fact that the two models have equivalent quality ratings of course implies that on average Brand A offers higher quality models that Brand B. After reviewing the information in the magazine, is Mr. L likely to find one of the two models more attractive, even though the two models have the same overall quality ratings. If yes, which one would it be, the higher quality model of Brand B, the inferior brand or the lowest quality model of Brand A, the superior brand? More generally, would people prefer a low-status member in a high-status category or a high-status member in a low-status category, holding the objective quality of the items constant?

This is the general question addressed in this research. In this example, as well as in a number of shopping decisions, there are three types of information that can be used to evaluate an item. The first one is the absolute quality ratings: in our example, if Mr. L focuses mainly on the quality ratings then the two models should be judged equally attractive. That is, one could evaluate an item as a function of its position in a "universe" of items (which is what the quality rating indicates). In doing so, these two items would be evaluated equally. Although normative theory would predict that this should be the basis for the evaluation, we know from the multiple demonstrations of context effects that information from the context may influence the evaluation of an item. In a decision task like the one previously described, the context may highlight two additional types of information. One of them is the fact that an item belongs to a category (in this case a brand) and category membership (Brand A or Brand B) can be used to evaluate the item. Said differently, an item can be evaluated based on the position of the category it belongs to in a "universe" of items. If this is the case, Mr. L would find Brand A model more attractive than Brand B, Brand A being the higher average quality brand. Finally, a third type of information that can be used in evaluating an item is the position of the item within the category (best quality in Brand B or lower quality in Brand A). When using this perspective, an item is evaluated as a function of its position in the category it belongs to. If this is the main type of information used in the decision, then Mr. L is likely to consider the highest quality model (in Brand B the inferior brand) as being more attractive.

How would Mr. L evaluate the digital cameras? Which type of information is likely to have a stronger impact on judgment? Empirical evidence for narrow framing and bracketing suggests that people focus their attention on a subset of the information. In the types of decisions previously described, the narrower subset information is the category, thus the position of an item within a category is likely to receive greater attention than the two other types of information. For example, in one experiment, subjects were told that they were in the market for a car, they were given a list of all car models within a brand described on the number of horsepowers in the engine. Half of the subjects were given a list of Audi cars a better category) and told that they were interested in the model at the bottom of the list (worst model on the list), the Audi 4. The other half was given a list of VW cars (a worse category) and were told that they were considering the model at the top of the list (best model on the list), the PassatV6. Both cars were described as having 190 horsepowers and all subjects were asked the amount of money they were willing to pay to get the considered car. Subjects in the VW condition were willing to pay significantly more than subjects in the Audi condition supporting the dominating role of position within the category in the evaluation. A series of six experiments was conducted and shows the following: 1) the position of an item within a category has a significant impact on its evaluation, 2) this overweighting of the position in the category can be observed in the perceptual domain as well as in judgment and choice and 3) situational factors (priming, instructions) can affect the salience of the other types of information as well.

 

"PANDA, MUGGER AND MUSIC: ON THE AFFECTIVE PSYCHOLOGY OF VALUE"

Christopher K. Hsee, University of Chicago

Yuval Rottenstreich, University of Chicago

In this research, we propose a process-based interpretation of evaluation. Building on earlier dual-process models (e.g., Chaiken & Trope, 1999; Epstein, 1994; Kahneman & Frederick, 2001; Sloman, 1996), we draw a distinction between two types of evaluations: evaluation by calculation and evaluation by feeling. We suggest that evaluations by calculation rely on rules and algorithms that are sensitive to the quantity or magnitude of the object being valued (e.g., the number of dollars to be earned, the number of lives to be saved, etc.). Evaluations by feeling, on the other hand, reflect one’s gut feelings toward a prototype of the target object (e.g., Kahneman & Frederick, 2002) and are relatively insensitive to quantity.

A value function relates the quantity dimension of an object to one’s valuation of the object. We propose that the shape of the value function will vary, depending on whether the underlying evaluations are by calculation or by feeling. Value functions that reflect evaluations by calculation will be relatively linear and steepBthat is, sensitive to variations in quantity. Value functions that reflect evaluations by feeling will be relatively flatterBthat is, insensitive to variations in quantity.

A critical question concerns the factors that determine whether a person will engage in evaluation by calculation or evaluation by feeling. There are likely to be many such factors. In this research, we explore one such factor: whether the object under consideration is affect-poor or affect-rich. We propose that, even controlling for their monetary values, some outcomes are associated with vivid imagery and keen sensation and evoke "hot" affect (either positive or negative), whereas other outcomes are relatively pallid or "cold." We describe the former type of outcomes as "affect-rich" and the latter type as "affect-poor."

Our main hypothesis is that compared to affect-poor objects, affect-rich objects are more likely to activate evaluation by feeling and are therefore less sensitive to quantity. If this hypothesis holds, then we should observe a Quantity (less vs. more) x Affect Richness (poor vs. rich) interaction effect.

We report a set of studies that tested this prediction. All the studies had four between-subject conditions, constituting a 2 (quantity) x 2 (affect-richness) design. These studies involved different types of quantity manipulations, different tyes of affect-richness manipulations and different dependent variables.

In one study, for example, we asked college students to indicate their willingness to donate money to save panda(s) that fellow students had found in a remote area. The quantity manipulation in this study was the number of pandas found: either one or four. The affect richness manipulation was the manner in which the number of panda(s) was conveyed: either by a representation in which one dot was displayed for each panda found (affect-poor) or by a representation in which one picture(s) of a panda was displayed for each panda found (affect-rich). We found the predicted quantity x affect-richness interaction: people were more sensitive to the quantity variation (one vs. four pandas) in the dot (affect-poor) condition than in the affect-rich (picture) condition.

In another study, we asked college students to assume the perspective of a juror and recommend a jail term for a person convicted of mugging a fellow student on campus. In this study, the quantity manipulation was the number of times the person had mugged before: either once or five times. Before recommending a jail term, some respondents were asked to think about the feelings of the mugger’s victim(s) (the affect-rich condition), and some respondents were not asked to do so (the affect-poor condition). Again, we secured the predicted interaction effect: The length of the recommended sentences was less sensitive to the number of previous offenses when respondents had imagined the feelings of the victim(s) than when they had not.

In a related paper (Rottenstreich & Hsee, 2000), we found the same type of interaction when the quantity dimension was probability. For example, people who indicated willingness to pay for a lottery ticket were less sensitive to the probability of winning if the prize was affect rich (e.g., a coupon for a trip to Europe) than if it was affect poor (e.g., a coupon for a tuition deduction).

We consider these findings as evidence for a dual-process model of evaluation: affect-rich outcomes activate evaluation by feeling, which is qualitative in nature, whereas affect-poor outcomes trigger evaluation by calculation, which is more quantitative in nature. This research has important implications for common analyses of value and probability weighting functions as well as for the scope and duration neglect effects recently discussed in the literature.

 

"JUDGMENT REVERSAL IN CROSS-CATEGORY COMPARISONS"

Daniel Kahneman, Princeton University

David Schkade, University of Texas

Ilana Ritov, Hebrew University

Cass Sunstein, University of Chicago

In this study we demonstrate a new kind of judgment reversal in two important real-world tasks: the setting of punitive damages, and willingness to pay for public goods using contingent valuation (CV) surveys. We find in both of these tasks that rating and dollar responses are quite different when a case or a cause is evaluated on its own, than when it is directly compared to an item from another category.

The standard design for studies of preference reversals involves judgments of the same two objects under two conditions, which Hsee et al. (1996; 1999) have labeled Separate Evaluation (SE) and Joint Evaluation (JE). Participants in the SE condition evaluate each object in isolation. Participants in the JE condition report a choice or a comparative evaluation of the two objects. The original studies of preference reversals involved a possible confound, because the conditions differed both in the procedure of evaluation (pricing or choice) and the context of evaluaion (SE or JE). We adopt the procedure suggested by Hsee et al., which eliminates the confound by requiring participants in the JE condition to rank the two objects on the scale of evaluation used in SE, and then to make a numerical evaluation.

Some attributes of judgment objects are more (or less) salient in JE than in SE. Two specific variants of this generic explanation have been proposed. The prominence account asserts that the most important attributes are weighted more heavily in choice than in SE (Tversky, Sattath and Slovic, 1988). The evaluability account asserts that some attributes are much easier to evaluate in a direct comparison than in isolation (Hsee, 1996). Prominence and evaluability are not mutually exclusive, and may explain reversals in different situations. The present article adds cross-category comparison (CCC) as third mechanism that can alter the relative salience of attributes.

Based on norm theory (Kahneman and Miller, 1986), our first hypothesis is that, in SE, cases of each kind are spontaneously compared to other cases of the same kind, yielding evaluations that are relative to that category. Our second hypothesis is that physical injury is more prominent than financial injury, and that cases of physical injury will therefore evoke more indignation and higher punitive damages in CCC than in SE. We make similar predictions for public causes, where we expect threats to human health to be more prominent than environmental problems.

Jury-eligible citizens from Travis County, Texas (n=1,035) were recruited by a survey firm. The stimuli were three pairs of legal cases and three pairs of public causes. Prominence was varied in each pair. The cases were short summaries of a realistic incident of personal injury or financial harm. The defendant was always a large company described as having "annual profits of around $150 million", and $500,000 in compensatory damages had already been awarded to the plaintiff. Two thirds of respondents were asked whether punitive damages were appropriate, and if so, in what dollar amount. Other respondents rated the appropriate severity of punishment on a scale ranging from 0 ("no punishment") to 8 ("extremely severe punishment"). Public causes (threats to public health or environmental issues) were described in a short paragraph, along with a proposed mitigating intervention. Two thirds of respondents were asked to indicate the maximum amount that they would be willing to pay (WTP) to contribute to the intervention. Other respondents rated the satisfaction that they would derive from contributing to the cause, ranging from 0 ("no satisfaction at all") to 6 ("extreme satisfaction").

Participants were randomly assigned to one pair of legal cases and to one pair of public causes. Every participant received two numbered envelopes, each containing both a legal case and a public cause. To implement separate evaluation (SE), the participant was required to respond to both the case and the cause in Envelope 1 before opening Envelope 2. To implement cross-category comparison (CCC), the questionnaire in Envelope 2 first required an explicit comparison of the new item to the corresponding item in Envelope 1 (e.g., "Which defendant deserves more punishment?"), as well as a numerical response to the new item on the same scale. Thus, each participant provided data for both the SE (Envelope 1) and CCC (Envelope 2) conditions.

Our major hypothesis was that cases or causes drawn from a more prominent category would elicit relatively higher responses in joint than in separate evaluation, producing an interaction between Mode of Evaluation (CCC vs SE) and Order of presentation. The predicted interactions are present in all four cells of the design, and highly significant in each case (each p<.001): the response to a prominent item is enhanced by comparing it to a previously judged low-prominence item, but a low-prominence item is not trivialized by a comparison to a more prominent one. The ANOVA tests of this asymmetry are significant in each case (p<.05). The same interaction shows up both in ratings and in dollar measures of attitdes, for both legal cases and public causes.

Our findings illustrate a new type of judgment reversal, which has significant implications for several real-world domains. We have observed, with dollar responses as well as with ratings, an interaction that is predicted jointly from norm theory (Kahneman and Miller, 1986) and from our analysis of preferences as attitudes (Kahneman, Ritov and Schkade, 1999). These are the steps in the argument. (i) If a judgment object encountered in isolation (SE) spontaneously evokes a context of similar objects, the norms to which its attributes will be compared are context-dependent and category-relative. (ii) If the affective response to an object is inherently comparative, then the differences between the frames of reference evoked in SE and in CCC will yield attitudes that actually differ in intensity. (iii) If dollar responses to attitude objects are attitude expressions, dollar judgments in SE and in CCC will be predictably incompatible. (iv) The observation of similar category-relativity effects in dollar responses and in ratings eliminates an interpretation of the results as a linguistic artifact, which would be possible if ratings were our only source of evidence.

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