kardes et al_2004

Upload: lizz-focella

Post on 10-Apr-2018

223 views

Category:

Documents


0 download

TRANSCRIPT

  • 8/8/2019 Kardes et al_2004

    1/8

    The Role of Selective Information Processingin Price-Quality InferenceFRANK R. KARDESMARIA L CRONLEYJAMES J. KELLARISSTEVEN S. POSAVAC*

    This research investigates the effects of the amount of information preseinformation organization, and concern about closure on selective informationcessing and on the degree to which consumers use price as a basis for infequality. Consumers are found to be less likely to neglect belief-inconsistent mation and their quality inferences less influenced by price when concern closure is low (vs. high) and information is presented randomly (vs. ordered)small amount of information is presented. Results provide a picture of a resoconstrained consumer decision maker who processes belief-inconsistent mation only when there is motivation and opportunity.

    C onsumers frequently assume that price and quality arehighly correlated, and that as the price of a productincreases, its quality increases commensurately ("you getwhat you pay for"). This assumption exerts a powerful in-fluence on the degree to which consumers use price to inferquality (Baumgartner 1995; Bettman, John, and Scott 1986;Broniarczyk and Alba 1994; Pechmann and Ratneshwar1992; Rao and Monroe 1988, 1989). However, little isknown about other variables that influence how heavily con-sumers rely on price when inferring quality. This researchseeks to elucidate boundary conditions that govern price-based quality judgments and, in so doing, helps to reconcilesome conflicting findings concerning effects of informationformat and selective processing of belief-inconsistent in-formation reported in previous research.SELECTIVE INFORMATION PROCESSINGTHEORY

    Judgment involves generating and testing tentatively heldinterpretations, expectations, evaluations, or possibilities*Frank R. Kardes is professor of marketing. University of Cincinnati,Cincinnati, OH 45221-0145 ([email protected]). Maria L. Cronley isassistant professor of marketing, Miami University, Oxford, OH 45056([email protected]). James J. Kellaris is professor of marketing. Uni-versity of Cincinnati, Cincinnati, OH 45221-0145 ([email protected])and visiting professor, Ecole Sup6rieure de Commerce de Toulou se. France.Steven S. Posavac is associate professor of marketing. Simon School ofBusine ss, University of Rochester, Rochester, NY 14627-0100 (p osavac @simon.rochester.edu). The authors thank the editor, associate editor, and

    reviewers for their helpful input. This research was funded in part by aSummer Research Fellowship at the College of Business. University ofCincinnati.

    (Hoch and Deighton 1989). Because the evidential basejudgment are often scattered and complex, consumers osimplify the judgment formation and evaluation procesfocusing on a single hypothesis or possibility at a timeby focusing selectively on hypothesis-consistent evidwhile neglecting hypothesis-inconsistent evidence (Sanmatsu et al. 1998). When the evidence supporting a fhypothesis meets a minimum conflrmation thresholdhypothesis is accepted and information processing ceA plausible hypothesis that is considered flrst is more lito be accepted than a less focal but more viable alternhypothesis, and this leads to judgments that are moretreme and that are held with greater confidence than is ranted by the available evidence.The expectation that two variables are related encourpeople to focus selectively on cases that support thepothesis and to neglect nonsupportive cases. For examconsumers who believe that a strong positive relationexists between price and quality are likely to focus on price/high quality products and on low price/low quproducts. Unexpected information (about high price/lolow price/high quality products) tends to be neglectedpecially w hen p rocessing is difficult. Variables that incsensitivity to belief-inconsistent evidence should reducdegree to which consumers rely on price to predict qu

    Moderators of Selective Information ProcessinInformation Load. Selective information processilikely to be influenced by the amount of information a

  • 8/8/2019 Kardes et al_2004

    2/8

    3

    to proc ess beca use this information require s effortful

    Information Organization. The manner in which in-t determinant of processing difficulty (Baum gartnerBettman et al. (1988) and Broniarczyk and Alba

    list format (i.e., brands ranked by qual-effects of format on judgm ent were found in these studies,

    Concem about Closure. Another important determi-of selective information processing is the concem about

    ning judg men tal closure or downplay the costs of reaching

    for as long as possible ("freezing").As the need for closure increases, people consider less

    Concem about closure differs from involvement and theortful processing occu rs as involvement or as the need

    need for cognitive closure increases. How ever, when closuis difficult to attain because the initial evidence does nsupport any conclusion, processing effort increases as tneed for cognitive closure increases. Webster and Kruglaski (1994) obtained a correlation of .28 between the nefor cognition and concem about closure.The Prediction Task

    Broniarczyk and Alba (1994) presented information aboprice (in dollars), quality (in ratings on a scale from 0 100), and amount of advertising (in thousands of dollars pmonth) for 25 brands of stereo speakers. Several differeversions of this table were created. In each version, thcorrelation between price and quality was held constant zero, the correlation between amount of advertising aquality was manipulated, and the format was manipulateAfter studying the data, participants were asked to rate tquality of 10 test brands that were not included in the oriinal set of 25 brands. For each of the 10 test brands, priand amount of advertising information was presented, bno information about quality was presented. Based on paticipants' quality ratings (on a scale from 0 to 100) for ttest brands, price-quality correlation coefficients and avertising-quality correlation coefficients were computeThe results revealed that participants overestimated tstrength of the relationship between price and quality in conditions, except for the condition in which amount advertising and quality were perfectly correlated. Particpants also underestimated the strength of the relationshbetween amount of advertising and quality in all conditionexcept for the condition in which amount of advertising aquality were uncorrelated.

    Contrary to the implications of the selective informatioprocessing explanation of price-quality inference, no effefor infonnation list format (rank ordered vs. random) wobserved. Rather than rejecting the selective processing eplanation, we suggest that the explanation should be testin experiments that control for variables hypothesized moderate the likelihood of selective information processinBroniarczyk and Alba (1994) used a relatively small amouof information (i.e., information pertaining to 25 brands) each of their five experiments. Providing information aboa larger set of brands (e.g., 100 brands) may encouraselective processing to reduce processing difficulty. Hoever, even when a large amount of information is availabselective processing should be less likely when informatiis presented in a random (vs. rank-ordered) list format consumers with a low (vs. high) concem about closubecause these factors should increase sensitivity to beliinconsistent evidence, reducing the degree to which cosumers rely on price to predict quality.

    E X P E R I M E N T 1Participants and Design

    Participants were 118 undergraduates at a large mwestem university. Participants were randomly assigned

  • 8/8/2019 Kardes et al_2004

    3/8

    370 JOURNAL OF CONSUMER RESEATABLE 1

    MEAN SUBJECTIVE PRICE-QUALITY CORRELATIONS ASAND CONCERN ABOUT

    Experiment 1Experiment 2Experiment 3

    Low load/randomHigh concern Low concernabout closure about closure

    .63 .46

    A FUNCTION OF INFORMATION LOAD,CLOSURE (EXPERIMENTS 1, 2, AND 3)High load/organized

    High concern Low concernabout closure about closure.92 .84.53 .57

    INFORMATION ORGANIZATI

    High load/randomHigh concernabout closure

    .92.72.48

    Low coabout cl.71.66.22

    NOTE.The objective price-quality correlations are r = .25 (experiment 1), r= .22 (experiment 2). and r = .20 (experiment 3).

    conditions in a 2 (rank ordered vs. random list format) x2 (high vs. low conce m about closure) x 2 (Top 100 listor 100 Best Buys list) between-subjects design.

    ProcedureParticipants received a questionnaire booklet stating thatthe purpose of the study was to examine people's percep-tions of quality given specific information about a product.he instmctions described the prediction task and the im-ortance of prediction in everyday life (see Broniarczyk andlba 1994, p. 137). In high concem about closure condi-ions, participants were told, "It is not necessary to spend great deal of time pondering the questions. In fact, pleasenswer the questions as quickly as possible (while still pro-

    In low concern about closure conditions, participants wet e

    During the learning phase, participants received infor-

    Wine Enthusiast), and ther = .27, p < .05.nthusiast), and the objective p rice-quality coirelation

    as r .22, p < .05. No main effects or interactions were

    During the test phase, participants were asked to proquality inferences (on a scale from 0 to 100) for 10 hthetical wines described in terms of price. Subjective pquality correlation coefficients were computed on the of participants' quality inferences (Broniarczyk and 1994). Participants were not permitted to refer back toproduct information. Finally, participants received questpertaining to familiarity with the different varieties of wthe different brands of wine, and wine consumption.differences were found on quality inferences as a funcof these familiarity variables.Results and Discussion

    Mean subjective price-quality correlations as a funcof information load, organization, and motivation to proselectively for experim ents 1, 2, and 3 are pres ented in 1. Prior to analysis, the price-quality correlations were verted to z scores using Fis her's / to c transformation. (rank order vs. random organization) x 2 (high vs. concem about closure) analysis of variance performed scores yielded an information organization x accuracy tivation interaction, F(\, 114) = 6.89, p

  • 8/8/2019 Kardes et al_2004

    4/8

    37

    Wine Enthusiast, and 10 brands were se-

    Specif-r = .22,.05, for both lists, the quality ratings ranged from 95 to

    1 except that measured

    During the prediction phase, participants were asked to

    4.00, NS, to number of cases produced, x' = 8 0 1 ,and to quality ratings, x ' = 4 .00, NS (median splitsThe number of cases produced per year was unre-d to price (r == .29, NS) and to q uah ty (r .07,When predictors are uncorrelated, "the correlation of

    the 42-item Need for Cognitive Closure Scale= .82) was administered to measure concem about clo-and participants were blocked into high versus lows based on a median split performed on their scoreser and Kruglanski 1994). A multiple regression anal-

    t(52) =p

  • 8/8/2019 Kardes et al_2004

    5/8

    72 JOURNAL OF CONSUMER RESEA

    prediction task was important to me personally." Two itemsassessed the need for cognition (a = .78): "I don't like tohave to do a lot of thinking" (R); "I try to avoid situationsthat require thinking in depth about something" (R). Aspredicted, the concern about closure manipulation had astronger effect on the concern about closure, F(\, 159) =5.68, p < .02, than on involvem ent, F( l, 159) = 1.44, NS ,or need for cognition, F(\, 159) < 1, NS.Finally, participants were shown a list of digital camerabrand names and were told that some of the brands werepresented earlier and some were not. Recognition confidencewas measured on scales ranging from one (definitely wasnot presented) to 10 (definitely was presented). A total of24 brand names were presented, with six "belief consistent"brands (i.e., high quality/high price or low quality/lowprice), six "belief inconsistent" brands (i.e., high quality/low price or low quality/high price), and 12 filler brands.Results and Discussion

    A 2 (rank order vs. random organ ization) x 2 (high vs.low concem about closure) analysis of variance performedon z scores yielded an organization x accuracy motivationinteraction, F(l, 157) = 4.78, /;> < .03. Follow -up testsshowed that the effect of information organization on price-quality inference was more pronounced for high accuracymotivation (M = .22 vs. .48, f(157) = 2.31, p < .02} thanin low accuracy motivation conditions (M = .53 vs. .57, NS ).A 2 (rank order vs. random organization) x 2 (high vs.low accuracy motivation) x 2 (belief consistent brands vs.belief inconsistent brands [within-subject factor]) analysisof variance performed on brand name recognition scoresyielded a main effect for belief consistency (M = 6.61 vs.5.86), f d , 157) = 26.69, p < .001. More important, resultsshow ed a significant organ ization x belief cons istency in-teraction, F(\, 157) = 6.17, p < .02. Recognition confi-dence was higher for belief-inconsistent cases in random(vs. rank ordered) list format conditions (M = 6.14 vs.5.60), r(157) = 2.56, p < . 01. Recognition confidence forbelief-consistent cases did not differ as a function of format(M = 6.54 vs. 6.57, NS).There was also a significant co ncem about closure xbelief consistency interaction, F(l, 157) = 4.17, p < .02.Recognition confidence tended to be higher for belief-in-consistent cases in low (vs. high) concern about closureconditions (M = 6.04 vs. 5.67) r(157) = 1.71, p < .08.Recognition confidence for belief-consistent cases did notdiffer as a function of the concern about closure (M 6.52 vs. 6.69, NS).The results support the selective processing hypothesisaccount using a different manipulation of concem aboutclosure, a different product category, and a recognition con-fidence measure of the degree to which participants pro-cessed information selectively. Recognition confidence wasgreater for belief-consistent information when infonnationwas presented in a rank-ordered list format because such aformat increases the ease with which consumers can focuson belief-consistent cases. When information was presented

    in a random format, more balanced processing occurscause consumers are more likely to encounter belief-insistent cases unintentionally.EXPERIMENT 4

    The goal of experiment 4 was to examine the extewhich the effects of concem about closure on price-quinference generalize across products (interior house and boxed chocolates) and across data sets having (, - .80) versus low (r - .20) objective price-qualityrelations. High information load (i.e., 50 brands) and adom format were held constant, and concem about clowas measured as a trait (Webster and Kruglanski 1994addition to measuring price-quality inferences, price-qucovariation judgments were assessed.Participants and Design

    One hundred and sixty two undergraduates were domly assigned to conditions in a 2 (high or low conabout closure) x 2 (high or low objective price-qucone lation s) x 2 (product catego ry: interior house paiboxed chocolates) mixed design with two between-sufactors and one within-subject factor (product categorProcedure

    Participants received information about 50 brands oterior house paint. The brand name, the type of paint low-luster, or semi-gloss), the price per gallon, and quratings were provided for each brand. The objective pquality correlation was high (/ .80) for half of theticipants and low (/- = .20) for the remaining participAfter the leaming phase, participants were asked to the quality of 10 hypothetical brands based on type of and price. During the leaming phase, the type of paintunrelated to both quality and price in high, x" = 1.40a n d x ' = 4.5 1, NS, respectively, and in low objective pquality conelation conditions, x" = -72, NS, and x4.51, NS, respectively. During the prediction phase, type was unrelated to price, and x" = .28, NS . Subjeprice-quality correlations were computed using participquality inferences.Next, covariation judgments were assessed. The instions stated, "A positive relationship exists when oneiable increases as another increases. A negative relatioexists when one variable decreases as another variablcreases." Participants were asked to indicate the direand the strength of the relationship between price and ity, type of paint and quality, and price and type of pon scales ranging from 10 (perfect negative relationto -f 10 (perfect positive relationship). The midpoint ofscale was labeled 0 (no relationship).

    After completing the judgment tasks for paint, particiwere asked to perform the same judgment tasks for aferent product category, receiving information abou

  • 8/8/2019 Kardes et al_2004

    6/8

    3

    .80) for half of the participan ts and low (r = .20) for

    x ' = 9.64, NS, and x ' == 8.54, NS , respectively, and= 8.98, NS , and X" = 8.54, NS , respectively. D uring

    and x" = 3.33, NS . After the learning phase, partic-

    Finally, the 42-item Need for Cognitive Closure Scale= .86) was administered to measure concern about clo-and participants were blocked into high versus lowon their scoreser and Kmg lanski 1994). A multiple regression treat-

    p < .05.

    A 2 (high or low concem about closure) x 2 (high orw objective correlation) x 2 (product category) analysisn about closu re (M = .78 vs. .67), F(l, 158) =p < . 0 0 1 , objective price-quality conelation (M =vs. .66), f"(l, 158) = 23.63, p < .001, and for product(M = .75 vs. .69), F ( l, 158) = 5.63,/? < .02. Sub-

    These effects were qualified by a significant concem about158) = 7.00, p < . 01 . Follow -up tests show that theon subjective price-qua lity correlations w ere more pro-

    = .56 vs. .77, r(158) = 5.34, p < .001) than under highabout closure conditions (M = .74 vs. .81 , NS ).

    A 2 (concern about closure) x 2 (objective price-qualitylation) x 2 (product category ) analysis of variancefects for concem about closure (M = 5.52 vs. 4.68),158) = 5.20, p< .02, and for objective price-qualityp

  • 8/8/2019 Kardes et al_2004

    7/8

    374 JOURNAL OF CONSUMER RESEA

    some studies show better memory performance for belief-consistent information (e.g., Rothbart, Evans, and Fulero1979), and others show better memory for belief-inconsis-tent information (e.g., Srull 1981). The present research re-solves these mixed findings: belief-consistent informationhas a memory advantage when information load is high,and belief-inconsistent information has a memory advantagewhen information load is low. Attention to belief-consistentinformation is likely to occur at an earlier and more basicstage of information processing that does not require abun-dant cognitive resources. By contrast, inconsistency reso-lution requires effortful elabor ative proce ssing that is likelyto occur at a later stage of information processing (Senguptaand Johar 2002; Wyer and Smll 1989). Effortful processesoccurring at a relatively late stage of the judgment processare disrupted more heavily by cognitive load manipulations,relative to less effortful proce sses occu rring at earlier stages(Johar and Simmons 2000; Meyers-Levy and Stemthal 1993).

    Prior research on the effects of information format onprice-quality inference has obtained mixed results: somestudies have found that easy-to-process formats (i.e., graphs)improve inference (Baumgartner 1985) and some havefound null effects (Bettman et al. 1988; Broniarczyk andAlba 1994). Our research helps to reconcile these apparentlydisparate findings. Easy-to-process formats have no effecton price-quality inference when information load is lowbecause selective information processing is unlikely underthese conditions. A different pattern of results emerges underconditions of high information load because high load en-courages consu mers to simplify a judgm ent task by focusingselectively on belief-consistent information. Selective in-formation processing is particularly likely when informationload is high, information is presented in a rank-ordered listformat (rather than a random format or perhaps a graphicalformat), and concem about closure is high.Prior research on sensitivity to objective degrees of cor-relation is mixed, with some studies showing insensitivity(Crocker 1981) and some showing sensitivity (Bettman etal . 1986). The present research suggests a resolution to thesemixed findings: insensitivity to objective degrees of corre-lation is more likely when selective processing is likely (vs.unlikely). The likelihood of selective processing depends onthe amount of information that is presented, the manner inwhich this information is presented, and on the concernabout closure.

    [Dawn Iacobucci and David Glen Mick ser\'ed as editorsand Joel Huber ser\'ed as associate editor for this article. ]REFERENCES

    Baumgarttier, Hans (1995 ), "Oti the Utility of Consu mers' Theoriesin Judgments of Covariation," Joumal of Consiuner Research,21 (March), 634-43.Bettman, James R., Elizabeth H. Creyer, Deborah Roedder John,

    and Carol A. Scott (1988), "Covariation Assessment inOrder Data," Journal of Behavioral Decision Making,239-54.Bettman, James, R., Deborah Roedder John, and Carol A.(1986), "Covariation Assessment by Consumers," JourConsumer Research, 13 (December), 316-26.Broniarczyck, Susan M. and Joseph W. Alba (1994), "Theorsus Data in Prediction and Correlation Ta sks," OrganizaBehavior and Human Decision Processes, 57 (Jan117-39.Cacioppo, John T, Richard E. Petty, Jeffrey A. Feinstein, aBlair G. Jarvis (1996), "Dispositional Differences in Ctive Motivation: The Life and Times of Individuals Vain Need for Cognition," Psychological Bulletin, 11197-253.Crocker, Jennifer (1981), "Judgments of Covariation by SPerceivers," Psychological Bulletin, 90 (September), 27Hoch, Stephen J. and John Deighton (1989), "Managing Consumers Learn from Experience," Joumal of Mark53 (April), 1-20.

    Johar, Gita Venkataramani and Carolyn J. Simmons (2000),Use of Concurrent Disclosures to Correct Invalid InfereJoumal of Consumer Research, 26 (March) . 307-22.Kmglanski, Arie W. and Donna M. Webster (1996), "MotiClosing ofthe M ind: 'Seizing' and 'Freezing'," PsycholReview, 103 (April). 263-83.Meyers-Levy, Joan, and Brian Stemthal (1993), "A Two-FExplanation of Assimilation and Contrast Effects." Jof Marketing Research, 3 0 ( 3 ) . 3 5 9 - 6 8 .Pechmann, Cornelia and S. Ratneshwar (1992), "Consumevariation Judgments: Theory or Data Driven?" JoumConsumer Research, 19 (December), 373-86.Rao, Akshay R. and Kent B. Monroe (1988). "The ModeEffect of Prior Knowledge on Cue Utilization in Productuations," Joumal of Consumer Research, 15 (Septem253-64.(1989), "The Effect of Price. Brand Name, and Store on Buyers' Perceptions of Product Quality: An IntegReview," Joumal of Marketing Research, 26 (Au351-57.Rothbart, Myron, Mark Evans, and Solomon Eulero (1979)call for Confirming Events: Memory Processes and the tenance of Social Stereotypes,'" Joumal of Experimentacial Psychology, 15 (4), 343-55.Sanbonmatsu. David M.. Steven S. Posavac. Frank R. KardeSusan P. Mantel (1998), "Selective Hypothesis Testing,chonomic Bulletin and Review, 5 (2), 197-220.

    Sengupta, Jaideep and Gita Venkataramani Johar (2002), "Eof Inconsistent Attribute Information on the Predictive of Product Attitudes: Toward a Resolution of Opposingspectives." Joumal of Consutner Research, 29 (June), 3Smll, Thomas K. (1981), "Person Memory: Some Tests osociative Storage and Retrieval Models." Joumal of Emental Psychology: Human Leaming and Memoiy, 440-63 .Webster, Donna M. and Arie W. Kruglanski (1994), "IndiDifferences in Need for Cognitive Closure," Joumal osonalir\- and Social Psychology, 67 (December), 261 -7Wyer. Robert S. and Thomas K. Smll (1989), Memoiy andnition in Its Social Context, Hillsdale, NJ: Erlbaum.

  • 8/8/2019 Kardes et al_2004

    8/8