i SCIENCEVol. 26, No. 6, November-December 2007, pp. 819-833ISSN 0732-23991 EISSN 1526-548X1071260610819 OOil0.1287/mksc,1070.0272

2007 INFORMS

Modeling the Determinants and Effects ofCreativity in Advertising

Robert E. Smith, Scott B. MacKenzieDepartment of Marketing, Kelley School of Business, Indiana University, Tenth and Fee Lane, Bloonxington, Indiana 47405

|smith5@indiana.edii, mackenz@indiana.edu)Xiaojing Yang

Department of Marketing, Sheldon B. Lubar School of Business, University of Wisconsin-Milwaukee, P.O. Box 413,2200 E. Kenwood Boulevard, Milwaukee, Wisconsin 53201-0413, yangxiao@uwm.edu

Laura M. BuchholzDepartment of Marketing, Kelley School of Business, Indiana University, Tenth .md Fee Lane, Bkwmington, Indiana 47405

lbuchhot@indiana.eduWilliam K. Darley

College of Business Administration, Marketing & International Business, University of Toledo,2801 Bancroft, Toledo, Ohio 43606-3390, william.darley2@utoledo.edu

Consumer perceptions of advertising creativity are investigated in a series of studies beginning with scaledevelopment and ending with comprehensive model testing. Results demonstrate that perceptions of ad cre-ativity are determined by the interaction between divergence and relevance, and that overall creativity mediatestheir effects on consumer processing and response.Key words: creativity; divergence; advertising; relevance; measurement; latent variable modelsHistory: This paper was received August 3, 2005, and was with the authors 8 months for 3 revisions;

processed by Gerard J. Tellis.

IntroductionMarketing researchers and practitioners agree that cre-ativity is one of the essential elements for advertisingsuccess in a cluttered marketplace. This philosophy isapparent in

advertising textbooks (e.g.. Belch and Belch 2004,Wells et al. 1995),

academic research (e.g., Goldenberg et al. 1999,Smith and Yang 2004, Till and Baack 2005), and

trade publications (e.g.. Advertising Age, Ad Week,Creativity).

In contrast to the importance attributed to ad cre-ativity is the limited amount of research investigatingit (Zinkhan 1993). Indeed, only a handful of empiricalstudies have been reported in the literature (see Smithand Yang 2004 for a review) and most investigate spe-cific issues regarding creativity rather than trying tomodel the overall phenomenon. Even though someinteresting questions have been addressed in the lit-erature, there is still insufficient empirical research toresolve basic issues. Accordingly, the major goals ofthis research are to provide empirical answers to thefollowing key questions:

How do consumers judge ad creativity?

What are the determinants of ad creativity? How should ad creativity be measured? Do the determinants of ad creativity interact as

some researchers have hypothesized? Does creativity mediate the effects of ad exposure

on key dependent variables like attention to the ad,ad attitude, and brand attitude?

To achieve these goals, we begin by discussingthe theoretical background for defining ad creativ-ity and its role in determining consumer process-ing and response. Next, we develop and validatescales designed to measure the key components ofad creativity (divergence and relevance). Then, wedevelop an ad processing and response model (anda mediation model) for the key constructs, and testthem in a series of pretests culminating with a finalstudy that compares Clio award-wirming ads to a ran-dom sample of network ads. Finally, we conclude bydiscussing implications from the results and neededfuture research.

Ad Creativity: Theoretical BackgroundIn the literature, definitions differ but most are sim-ilar to Leo Burnett's approach that ad creativity is

819

820Smith et al.: Modeling the Determinants and Effects of Creativity in Adi'ertising

Marketing Science 26(6), pp. 819-833, 2007 INPORMS

"the art of establishing new and meaningful relation-ships between previously unrelated things in a man-ner that is relevant, believable, and in good taste, butwhich somehow presents the product in a fresh newlight" (El-Murad and West 2004, p. 190). The main dif-ference in past definitions is whether ad creativity isdetermined by one or two factors. The first approachis to define creativity as divergence. Divergence canbe defined as the extent to which an ad containsbrand or execution elements that are different, novel,unusual, original, unique, etc. As an example of thisapproach. Till and Baack (2005, p. 49) noted: "creativeadvertisements have been consistently defined, atleast in part, as novel and/or original."

The second approach to defining ad creativity isthat it has two determ^inants: divergence and relevance(Besemer and O'Quinn 1986, Besemer and Treffinger1981, Haberland and Dacin 1992, Jackson and Messick1965, TelUs 1998, Smith and Yang 2004, Thorson andZhao 1997). Here, divergence is defined as original-ity and relevance is defined as the extent to which atleast some ad/brand elements are meaningful, use-ful, or valuable to the consumer. However, advertisingresearch has presented minimal theoretical develop-ment of divergence and relevance and has usuallyoperationalized them in a narrow manner. A majorgoal of this research is to conceptually develop thedivergence and relevance dimensions, and build validand reliable scales to measure them.

While researchers can debate the advantages anddisadvantages of the two conceptualizations, it isimportant to understand how consutners judge ad cre-ativity, because creativity (like beauty) is in the eyeof the beholder. Perceptions of an ad's divergencerequire a comparison with the consumer's experi-ences; while perceptions of relevance require a com-parison to the consumer's goals, needs, and desires.Thus it is the consumer's perceptionnot the judg-ment of researchers or advertising professionalsthat is expected to stimulate his or her interest in an ad.

Therefore it is essential to understand consumerperceptions of ad creativity. These judgments could beempirically derived by measuring consumers' percep-tions of an ad's overall creativity, and subsequentlymeasuring their perceptions of the ad's divergenceand relevance. If divergence is the only significantpredictor of overall creativity, it would suggest thatconsumers view creativity as being a function ofnovelty, originality, imaginativeness, etc. However, ifdivergence and relevance (and/or their interaction)are significant predictors of overall creativity, it wouldsuggest that consumers view creativity as a joint func-tion of these constructs.

Modeling Ad Divergence. Both theoretical per-spectives share the belief that the leading character-istic of creative ads is their divergencethey contain

elements that are novel, different, or unusual in someway. Indeed, as noted above, some researchers equatedivergence with creativity. Given the prominent roleit plays in determining creativity, it is surprising thatno research has investigated consumer perceptions ofad divergence. In fact, most advertising studies havelimited measures of divergence, which fall far shortof the conceptual development of this construct in thepsychology literature. For example. Smith and Yang(2004) review the pioneering creativity research byGuilford (1950, 1956, 1967) and Torrance (1987), toidentify seven primary indicators of divergence (flu-ency, flexibility, originality, elaboration, resistance topremature closure, unusual perspective and synthe-sis), and seven secondary indicators (empathic per-spective, provocative questions, future orientation,humor, richness and colorfulness of imagery, fantasy,and expression of feeling and emotion).

A major goal of this research is to empiricallyreduce this list of possible divergence factors toinclude only those that are directly related to diver-gence in an advertising context. It is important to notethat the divergence factors are conceived of as con-ceptual determinants of divergence rather than asreflections of it. Thus, these characteristics should bemodeled as forrtmtive (rather than reflective) indicatorsof divergence (see Diamantopoulos and Winklhofer2001, Jarvis et al. 2003).

Modeling Ad Relevance. The second characteristicof creative ads prominent in the literature is relevance tothe consumer. Using the second theoretical perspective,in addition to being divergent, creative ads must con-tain elements that are meaningful, appropriate, useful,or valuable to the audience in some way.

In marketing, there has been a long interest in therelevance component of ad creativity. Often referredto as involvement, there is a rich background on whatmakes an ad "personally relevant" to consumers, andhow this relevance can be expected to influence adprocessing and response (see, for example, Macinnisand Jaworski 1989). Thus the relevance component ofcreativity reflects the extent to which ad elements aremeaningful, useful, or valuable to the consumer, and itcan be achieved in the following ways:

Ad-to-Consumer Relevance. "Ad-to-consumer rele-vance" refers to situations where the ad contains exe-cution elements that are meaningful to consumers.This type of relevance is achieved when stimulusproperties of the ad create a meaningful link to poten-tial buyers. For example, using Beatles music in anad could create a meaningful link to Baby Boomers,thereby making the ad relevant to them.

Brand-to-Consumer Relevance. "Brand-to-consumerrelevance" refers to situations where the advertisedbrand (or product category) is relevant to potentialbuyers. This type of relevance occurs when an adestablishes a mearungful link between the brand and

Smith et al.: Modeling the Determinants and Effects of Creativity in AdvertisingMarketing Science 26(6), pp. 819-833, 2007 INFORMS 821

the consumer. For example, the advertisement couldshow the brand being used in circumstances familiarto the consumer (Mishra et al. 1993, Thorson and Zhao1997).

Ad-to-Brand Relevance. A third possibility in termsof relevance is how well the ad relates to the brand(e.g., Ang and Low 2000). However, this connec-tion is not directly indicative of the ad's relevanceto the consumer as applied in the creativity litera-ture. Accordingly, it is not expected that ad-to-brandrelevance will predict consumer perceptions of adcreativity.

Thus, another goal of this research is to develop avalid model of relevance in an ad creativity context.As before, the relevance factors are conceived of as theconceptual determinants of ad relevance rather thanas reflections of it, and should be modeled asfortnativeindicators (see Diamantopoulos and Winklhofer 2001,Jarvis et al. 2003).

Consumer Processing and Response. Another goalof this research is to examine the effects of ad creativityon consumer processing and response. Traditionally,creative ads have been expected to attract more atten-tion from consumers because their divergence con-trasts with noncreative ads {Smith and Yang 2004).While it seems likely that creative ads will stand outin ad clutter (Wells et al. 1995, p. 451), careful exam-ination of the full range of consumer processing andresponse variables has not been achieved. To providenew evidence about where the effects of ad creativitywill be manifested, we collected measures of consumerprocessing and response. The selection of these vari-ables was guided by the widely referenced ad modelof Maclnnis and Jaworski (1989) and included threeprocessing variables: amount of attention allocated tothe ad, motivation to process the ad, and depth of adprocessing; and three response variables: ad attitude,brand attitude, and purchase intentions.

Divergence by Relevance Interaction Effect.Another important issue in conceptualizing ad cre-ativity is the possibility of an interaction effectbetween ad divergence and ad relevance. Specifically,past models in advertising (Smith and Yang 2004),marketing (Im and Workman 2004), consumer creativ-ity (Burroughs and Mick 2004), and social psychology(Mumford and Gustafson 1988) have conceptualizedcreativity as requiring both high divergence and highrelevance. According to these models, ads that are lowin both divergence and relevance lack both prerequi-sites, and therefore are noncreative. Ads that are highin divergence and low in relevance may attract theconsumer's attention (because of divergence), but willhave limited effectiveness (because of low relevance).Ads that are low in divergence and high in relevancecan be ineffective in a cluttered marketplace becausethey fail to engage the consumer.

Creative ads possess high levels of both divergenceand relevance. These ads can attract the attention ofconsumers and channel it to relevant issues, therebyenhancing the impact on ad processing and response(i.e., making them significantly more effective). More-over, based on the above discussion and given theimportance attributed to ad creativity, it seems reason-able to predict that the effects of high divergence andhigh relevance should be more than additive. Indeed,Smith and Yang (2004) predict that there will be a fan-shaped interaction effect between ad divergence andad relevance. While this interaction effect is stated orassumed in many previous ad creativity studies, it hasnever been empirically examined.

Scale Development and PretestingValid and reliable scales for the ad divergence andad relevance constructs were developed over a seriesof six pretests involving 1,250 respondents. The finalmeasurement scales are presented in Appendix A andthe pretest process is summarized in Appendix B.

Divergence Measures. A major goal of this re-search is to identify the ways that ads can achievedivergence. To accomplish this goal, we started withthe list of 14 divergence factors developed by Gullford(1950, 1956, 1967) and Torrance (1987) and appliedto advertising by Smith and Yang (2004). We thenexamined the face validity of each item and elimi-nated those that overlapped with established market-ing constructs (i.e., empathic perspective, humor, andexpression of feeling and emotion) based on sugges-tions of experts (Pretests 1 and 2). Next, we eliminatedfactors that are more closely related to the diver-gent thinking process (i.e., how people come up withunusual ideas) than to perceptions of ad divergence(i.e., resistance to premature closure and provocativequestions). Finally, to reduce conceptual redundancyand achieve parsimony, we combined fantasy, futureorientation, and unusual perspective into a single fac-tor called imagination, leaving the following sevenmain indicators of ad divergence:

FlexibilityAds that contain different ideas orswitch from one perspective to another.

FluencyAds that contain a large number ofideasmore than expected.

OriginalityAds that contain elements that arerare, surprising, or move away from the obvious andcommonplace.

ElaborationAds that contain unexpected details,or finish and extend basic ideas so they become moreintricate, complicated, or sophisticated.

SynthesisAds that combine, connect, or blendnormally unrelated objects or ideas.

Artistic ValueAds that contain artistic verbalimpressions or attractive colors or shapes.

822

Tabie 1

nam

Smith et al

Rotated Fador Matrix for Exploratory Factor Analysis on

.: Modeling the Determinants andMarketing Science

Divergence Measures

Factors

1 2 3 4

Effects ofCreativiti/ in Advertising26(6), pp, 819-833, 2007 INFORMS

5 6 7

The ad was out of the ordinary.The ad broke away from hahit-bound and stereotypical thinking.The ad was unique.The ad contained a large number of ideas.The ad contained more concepts than most ads.The ad had many ideas, more than expected.The ad contained ideas that moved from one subject to another,The ad contained different ideas.The ad shifted from one idea to another.The ad connected objects that are usually unrelated.The ad contained unusual connections.The ad brought unusual items together.The ad contained numerous details.The ad finished basic ideas so that they become more intricate.The ad contained more details than expected.The ad allowed me to form images I have not directly experienced.The ad made something unreal come to life.The ad allowed me to form vivid mental images.The ad was visually/verbally distinctive.The ad made ideas come to life graphically/verbally.The ad was artistically produced.Initial eigenvalues.

0.720,690,69

0.36

0.30 0.310.750,780,87

0.310.27

0.27

0,830,770,79

0.28

0.35

0.28

0.300,99 2,98

0,850,860,82

0.290.27

1,72 9,30

0,800,630,70

0,63

0.690,780,66

0,700.29 0.790,30 0.541,12 0.93

Notes. (1) Extraction method: Maximum likelihood method, (2) Rotation method: Varimax method, (3) Loadings lowerthan 0.27 werenot included in the table. Loadings between 0,27 and 0,4 were italicized, (4) Factors: 1 ^ Originality; 2 = Fluency; 3 = Flexibility;4 ^ Synthesis; 5 = Elaboration; 6 ^ Imagination; 7 ^ Artistic Value,

ImaginationAds that help consumers formvivid mental images, or make something unreal cometo life.

Measurement Model for Ad Divergence. Scaleitems were developed for each of the divergence fac-tors and pretested in a series studies (see Appendix A,Part 1 for final measures). Following preliminary testsof their face validity, an exploratory factor analysisof the measurement items (Appendix A) was con-ducted using the combined data from Pretests 5 and 6and a maximum likelihood extraction method. Ini-tially, seven factors were retained because the screeplot revealed a significant drop in variance extractedafter this point. Together, these factors accounted fora total of 84.17

Smith et al.: Modeling the Determinants and Effects of Creativity in AdvertisingMarketing Science 26(6), pp. 819-833, 2007 INFORMS 823

Figure 1 Measurement Model for Divergence

Source. Diamantopoulos and Winklhofer 2001, Jarvis et al. 2003.Divergence is a second-order composite latent tactor, which is jointly determined by tive equally important first-order factors. Each individual first-order

dimension o! divergence is further measured reflectively by three items. In addition, three global reflective measures were added for the second-order compositelatent construct of divergence.

series of pretests (see Appendix A, Part 2 for finalmeasures). Tests of the psychometric properties ofthese measures using the Pretests 5 and 6 data indi-cated that they were all valid and reliable. The AVEsranged from 0.69 to 0.84 with an average of 0.78;and the as ranged from 0.87 to 0.95 with an averageof 0.92.

Measurement Model for Ad Relevance. For theo-retical reasons, we expected only ad-to-consumer andbrand-to-consumer relevance to be related to ad cre-ativity. To test this premise, we exam i^ned the cor-relations between scale scores for the three types ofrelevance and overall creativity. The results, displayedin Table 3, show a positive association between adcreativity and ad-to-consumer relevance (r = 0.46,p < 0.01) and a positive association between ad cre-ativity and brand-to-consumer relevance (r = 0.31,p < 0.01) as hypothesized. However, the associationbetween overall ad creativity and ad-to-brand rele-vance was negative (r ~ -0.31, p < 0.01). Accordingly,the measurement model specification for overall adrelevance included only ad-to-consumer and brand-to-consumer subdimensioris as shown in Figure 2.

Processing and Response Measures. The selec-tion of the processing and response variables wasbased on the ad processing model of Maclnnis and

Table 3 Correlation Matrix for Relevance

Measures 1 2 3

-0.31-

Jaworski (1989). Scales were developed to measure:amount of attention allocated to the ad, motivationto process the ad, depth of ad processing, ad atti-tude, brand attitude, and purchase intentions. Thesescales were pretested over a series of six studies andresults showed the items presented in Appendix A,Part 3 were valid (AVEs averaged 0.77) and reliable(as averaged 0.91).

Production Quality Measures. To better establishdiscriminant validity, it is important to distinguishan ad's creativity from its production quality. There-fore, four items were developed to measure the pro-duction quality of the ads. One was an estimateof the ad's overall quality, while the other three

Figure 2 Measurement Model for Relevance*

Brand-to-consumerrelevance

1,2.3.4.

Ad-to-consumer relevanceBrand-to-consumer relevanceAd-to-brand relevanceOverall creativity

0 .71"0.040.46*-

0.090.31

"Correlations significant at 0.01 level.

Source. Jarvis et al. 2003.Relevance is a second-order composite latent factor, which is jointly

determined by two equally important first-order factors: ad-to-consumer rel-evance and brand-to-consumer relevance. These first-order dimensions ofrelevance are measured reflectively b"^ four and five items. In addition, threeglobal reflective measures were added for the second-order composite latentconstruct of relevance.

824Smith et al.: Modeling the Determinants and Effects of Creativity in Advertising

MarkeHng Science 26(6), pp, 819-833, 2007 INFORMS

focused on the quality of the audio elements, visualelements, and production elements. The items shownin Appendix A, Part 4 were found to exhibit adequatevalidity (AVE = 0.64) and reliability {a = 0.87) basedon pretests. The production value scale was then usedas a covariate in the analyses that follow.

Overall Ad Creativity. To examine the determi-nants of consumer perceptions of ad creativity, anunbiased overall measure of perceived ad creativitywas needed that did not prompt subjects to includeeither divergence or relevance. Accordingly, afterviewing the ad, respondents were asked first to reporttheir judgments on how creative the ad was. Becauseno cues were provided in the definition of ad cre-ativity, respondents had to use their own judgments,which allows us to examine whether consumers judgead creativity as divergence or divergence plus rele-vance. The items used to measure overall creativityare shown in Appendix A, Part 5.

Overall Divergence and Relevance. To fullyaccount for the hypothesized relationships and forthe purpose of identification of the proposed second-order composite latent factor measurement model,global measures of overall divergence and overall rel-evance were also needed (see MacKenzie et al. 2005,Jarvis et al. 2003). Accordingly, respondents Vi^ ereasked to report their judgments on how divergentand how relevant the ad was on the scales shown inAppendix A, Part 6.

Final Study 'To provide an independent and conservative test ofthe models, a final study was conducted that com-pared creative ads to a random sample of networkads for national brands. The creative ads {n = 39) wereselected from Clio award winning ads from 2003.' TheClio awards are given to reward creative excellence inadvertising and design and are selected by a panel ofadvertising experts. The average ad, were a randomlygenerated set of network television ads {n = 39) fornational brands.- In generating the random sample ofnetwork ads, procedures similar to those of Till andBaack (2005) were used.^

' Brands for the creative ads included: Unif Green Tea, Budlight,Daily Telegraph, DeMorgan Newspaper, T-fal, CBS, AirlineArgentina, Deita Airline, Hallmark, Mastercard, Nike, AmericanExpress, Thai Insurance, Olympus, Toyota, Beatle, iPod, Sony PS2,and Honda.^ Brands for the random sample of ads included: AAA, Accuchek,Accura, All State, Ultramax, Colgate, Covergirl, Dodge, Excedrin,Ford, Icy Hot, KFC, Loreal, Maybelline, Vain, Golden Corral,FifthThird Bank, Value City, McDonald, Heavenly Ham, NicodermCQ, Flexall, Neutragena, Goodwill, Bounty, and Verizon Wireless.^ Four networks (ABC, CBS, NBC, and FOX) were each assigned anumber (1-4); each day of the week was assigned a number (1-7);

Two hundred tv^^enty-three student consumers froma major public university participated in the exper-iment for credit. The data were collected in a com-puter lab with up to 28 respondents participating atone time and wearing headphones for privacy. Par-ticipants were asked to view the target ad and thenfilled out the questionnaire at their own pace. Toreduce the possibility oi auto-checking, we reversedthe anchor points on every other item throughout thequestionnaire.

Psychometric Properties of the Divergence andRelevance Scales. Because the divergence and rele-vance scales are new, it is important to demonstratetheir reliability and validity. To test the factor struc-ture for divergence, a second-order composite latentfactor model was estimated using LISREL 8.7. Asshown in Table 4,

The hypothesized model of divergence fit thedata well (CFI = 0.95; NFI = 0.92; RMSEA = 0.086;SRMR = 0.10; x'^ = 371.83; df ^ 125; p < 0.001).

All 15 of the items loaded significantly on theirhypothesized first-order factors (p < 0.01).

The item reliabilities all exceeded 0.50. All five of the paths from the divergence sub-

dimensions to the second-order divergence factorwere significant {p < 0.01).

To test the factor structure for relevance, a second-order composite latent factor model was estimatedusing LISREL 8.7. As shown in Table 5,

The hypothesized model of relevance fit the datawell (CFI = 0.98; NFI = 0.97; RMSEA = 0.092; SRMR =0.043; x^ = 166.65; df =^ 53; p < 0.001).

AH nine of the items loaded on their hypothe-sized first-order factors (p < 0.01).

The item reliabilities all exceeded 0.50. In addition, both of the paths from ad-to-

consumer relevance and brand-to-consumer relevanceto the second-order divergence factor were significant{p < 0.01).

To gauge the validity of the divergence and rel-evance dimensions (i.e., construct reliability, conver-gent validity, and discriminant validity) in context

each hour of programming from 6:00 A.M. to 2:00 A.M. was assigneda number (1-20). Random numbers were drawn for each of thethree variables to designate the network, day, and time to recordcommercials, and the specified hour of programming was tapedfrom air in the spring of 2003. Each possible network commercialwithin the hour pod was assigned a number from 1-25. Researchassistants drew a random number and then counted the ads fornational brands until the random number was reached. That ad wastransferred to a CD and represented one of the random networkads. This procedure was repeated until 56 ads were designated.After eliminating duplicate ads, poorly recorded ads, and missingads (because of war coverage), there were a total of 50 differentnetwork ads, 39 of which were randomly selected for inclusion inthe final study.

Smith el al.: Modeling fhe Determinants and Effects of Creativity in AdvertisingMarki-ling Science 2f)((i), pp. 819-833, 2007 INFORMS 825

Table 4 Completely Standardized Measurement Parameter Estimatestortile Divergence Construct

Measures

Originality:The ad was out of the ordinary.The ad broke away from habit-bound and

stereotypical thinking.The ad was unique.

fiexibiiity;The ad contained ideas that moved from

one subject to another.The ad contained different ideas.The ad shifted from one idea to another.

Means

4.064.07

4.36

3.14

3.233.16

Standarddeviation

1,801,76

1.74

1.62

1.601.52

Factorloadings

0.920,76

0.84

0.86

0.830.84

Synthesis:The ad connected objects that are usually

unreiated.The ad contained unusual connections.The ad brought unusual items together

l^laboration:The ad contained numerous details.The ad finished basic ideas so that they

become more intricate.The ad contained more details than

expected.

Artistic value:The ad was visually/verbally distinctive.The ad made ideas come to life

graphically/verbally.The ad was artistically produced.

Overall divergence:The ad was different.The ad was uncommon.The ad was unusual,

Originality factorFlexibility factorSynthesis factor ' 'Elaboration factorArtistic value factor

3.95 1.93 0.87

3,703.77

3.594.00

11

11

.91

.88

,51.50

0.840,90

0.730,65

3.48

4.474.47

4.33

1.53

1.841.79

1,90

0.770.85

0.79

4.434.213.86

1711,651.62

0.880.830.77

0,230.180.230.150.20

Notes. (1) All estimates significant {p < 0.01). (2) CFI = 0.95; NFI = 0.92;RMSEA - 0,086; SRMR = 0.10; x^ = 371.83; df = 125; p < 0.001.

of each other, a 24-item confirmatory factor modelwith 7 intercorrelated first-order factors was tested{i.e., 3 items for each of 5 divergence subdimensions,plus 4 items for ad-to-consumer relevance and 5 itemsfor brand-to-consumer relevance). The results sug-gest that

The hypothesized model was consistent with thedata (CFI = 0.98; NFI ^ 0.95; SRMR = 0.044; RMSEA =0.055; x^ = 401.48; df = 231; p < 0.001).

All 24 items had significant loadings {p < 0.01)on their intended factors.

In addition, as shown in the lower diagonal entriesof Table 6,

The five divergence subdimensions and the tworelevance dimensions were all reliably measured (a'sranged from 0.79 to 0.94), and

Exhibited reasonable levels of convergent valid-ity (AVE ranged from 0.59 to 0.79).

3.853,883.663.61

3,773.893.833.673.74

3.833.623.73

t-791.821.781.77

1.871.871.891.842.11

1.811.671,91

0.860.830.910,86

0.920.920.920.930.73

0,870,790,77

0.400.45

Table 5 Completely Standardized Measurement Parameter Estimatesfor the Relevance Construct

Standard FactorMeasures Means deviation loadings

Ad-to-consumer relevance:The ad was meaningful to me.The ad was appropriate for me.The ad was useful to me.The ad was valuable lo me.

Brand-to-consumer relevance:The product or brand was meaningful to me.The product or brand was appropriate to me.The product or brand was useful to me.The product or brand was valuable to me.I do NOT care about this product/service. (R)

Overall relevance:The viewing experience was relevant to me.The viewing experience was useful to me.Overall, the ad and the brand were NOT really

applicable to me. (B)Ad-to-consumef relevance factorBrand-to-consumer relevance factor

Notes. (1) All estimates significant (p < 0,01), (2) CFI = 0.98; NFI = 0.97;RMSEA = 0.092; SRMR - 0.043; x^ = 166,65; df - 53; p < 0,001,

Finally, the results showed that a reasonable level ofdiscriminant validity was achieved as is evident fromthe fact that for every pair of constructs,

The intercorrelations were significantly less than1.0 {p < 0.05) (Anderson and Gerbing 1998), and

The squared intercorrela tions were always lessthan the AVE for the constructs (Fornell and Larcker1981).

Psychometric Properties of the Overall Diver-gence, Overall Relevance, and Overall CreativityScales. It is also important to assess the validity ofthe overall creativity, overall divergence, and over-all relevance measures. Thus, a 10-item confirmatoryfactor analysis (CFA) (4 global measures of overallcreativity, 3 global measures of overall divergence,and 3 global measures of overall relevance) was con-ducted. The CFA was performed separately from theprevious analysis because these constructs are at dif-ferent levels of measurement. Results show the fol-lowing:

The hypothesized model was consistent with thedata (CFI = 0.98; NEI = 0.98; SRMR = 0.059; RMSEA =0.083; r = 82.51; df = 32;p< 0.001),

That all 10 items had significant loadings(p < 0.01) on their intended factors,

Further analysis showed that overall creativity,overall divergence, and overall relevance were all reli-ably measured (a's were 0.78, 0.88, and 0.83, respec-tively),

Exhibited reasonable levels of convergent valid-ity (AVE were 0.65, 0.71, and 0.63), and

Met the requirements of discriminate validity asthe squared intercorrelations were always less than

826 Smith et al.: Modeling the Determinants and Effects of Creativity in AdvertisingMarketing Science 26(6), pp. 819-833, 2007 INFORMS

Table 6 Construct Intercorrelations and Reliability Estimates for the Divergence and Relevance Factors

Measures

1. Originality2. Flexibility '^, 3. Synthesis4. Elaboration5. Artistic value6. Ad-to-consumer relevance7. Brand-to-consumer relevance

a

0.890.880.910.790.850.920.94

0710.720.760.590.650.750.79

1

1.000.270.530.310.540.350.24

2

1.000.400.530.270.050.00

3

1.000.430.300.130.06

4

1.000.410.190.10

5

1.000.330.16

6

1.000.77

7

1.00

the AVEs for the constructs {the correlations were: cre-ativity and divergence (0.76), creativity and relevance(0.30), and divergence and relevance (0.42); all corre-lations are significantly less than 1.0).

The Role of Divergence and Relevance in Predict-ing Ad Creativity. The next goal was to examine theentire measurement model to see if ad divergence andad relevance predict ad creativity as hypothesized. Ascriterion measures, we used both consumers' percep-tion of ad creativity and whether the ad received aClio award from ad creativity experts.

First, we examined whether consumers judge adcreativity as divergence or divergence plus relevanceusing LISREL 8.7 by regressing consumer perceptionsof ad creativity on divergence, relevance, and theirinteraction, while controlling for ad production qual-ity"* The results reveal that divergence had a signifi-cant positive relationship with ad creativity (j8 = 0.35,z 4.88, p < 0.001), as did the interaction betweendivergence and relevance (/3 = 0.36, z = 8.11, p 0.05).

network ads. Comprehensive model testing was thenused to provide empirical answers as summarizedbelow.

How Do Consumers Judge Ad Creativity? Resultsshow that divergence is the leading indicator of cre-ativity but its interaction with relevance also plays asignificant role. This finding suggests that the defini-tion of ad creativity as divergence plus relevance ismost consistent with the data. However, the lack ofa main effect for relevance may explain why someresearchers do not include it in the definition of adcreativity.

What Are the Determinants of Ad Creativity? Inthis research, divergence was shown to be the lead-ing indicator of ad creativity. The list of 14 possi-ble divergence factors was reduced conceptually andempirically until five unique determinants remained:originality, flexibility, synthesis, elaboration, and artis-tic value. This finding suggests that research onad creativity should include measures of these fivefactors. Empirical results showed ad-to-consumer rel-evance and brand-to-consumer relevance were sig-nificant predictors of ad creativity while ad-to-brandrelevance had a negative correlation.

How Should Ad Divergence and Ad Relevance BeMeasured? Over a series of studies, the measures ofad divergence and ad relevance developed here wereshown to be reliable and valid. This result is impor-tant as it allows advertising researchers to use ver-ified scales, which could increase the comparabilityof results across studies and manipulations. Althoughthese scales were developed for TV ads, they canbe readily adapted to any media or marketing mixelement.

In addition, the scales could be used by market-ing managers or expert panels to judge the creativityof a proposed ad or a campaign over time. To facili-tate this usage, the mean ratings, standard deviations,and ranges for the divergence and relevance scalesacross all 189 ads used in this research are provided inAppendix D for use as preliminary norms in judgingad creativity.

Do the Determinants of Ad Creativity Interact?For the first time, we examined how ad creativityinfluences the processing of, and response to, adver-tising using a wide range of dependent variables.Results showed some main effects for both divergenceand relevance, but these were qualified by signifi-cant interaction effects on all six dependent variables(replicated in Pretests 5 and 6). These findings rein-force the need to include relevance measures in cre-ativity studies and expand the known effects of adcreativity well beyond the consumer's attention to thead. This implication is important because the multi-plicative nature of ad divergence and ad relevance

830 Smith et al.: Modeling the Determinants and Effects of Creativity in AdvertisingMarketing Science 26(6), pp. 819-833, 2U07 INFORMS

has been widely suggested but never empirically val-idated before this study.

Does Creativity Mediate the Effects of Ad Expo-sure on Key Dependent Variables? Finally, weshowed that creativity, in general, did mediate theeffects of divergence and relevance on the dependentvariables. This result furthers our understanding ofthe relationships among the key variables and rein-forces conceptual models that give ad creativity a cen-tral role.

Limitations and Future Research. Because internalvalidity is critical in theory testing and scale develop-ment, our studies were conducted in a lab setting tocontrol for extraneous variables. This procedure limitsthe findings to similar situations and shows a need forfuture research that replicates these results in realisticmarketplace settings. Also, the use of student partic-ipants in the final samples may restrict the general-izability of these results so research is needed usingmore diverse consumer segments.

Future research is also needed to explore the rela-tionship between divergence and relevance. Thesevariables can be tested in a variety of situations toexamine boundary conditions that n:\ay facilitate orameliorate the consistent interaction effect found inthe final study and the pretests. In addition, the diver-gence and relevance factors developed here could beextended to other marketing mix elements such asnew product development (Im and Workman 2004),or the idea generation process itself (Toubia 2006).

In addition, it seems time for marketing researchto move beyond the attentional effects of ad creativ-ity and investigate its role in the persuasion process.For example, research could examine whether creativeads make consumers more curious and open-minded(Kardes et al. 2004, Kruglanski and Webster 1996).Both of these effects can be predicted theoretically andboth could have a major influence on making con-sumers less defensive when they process persuasivemessages. It could also be hypothesized that exposureto creative ads might produce a priming or framingeffect making novel product features more salient indecision making.

Finally, there is currently no research that comparesad creativity processing and effects across impor-tant demographic and/or psychographic factors suchas age, gender, education, culture, and consumeridentity. Studies of these issues would advance ourknowledge of how ad creativity influences consumerbehavior. It definitely seems time to address this areaof advertising that is frequently discussed and intu-itively important.

AcknowledgmentsThis research was supported by grants from the KelleySchool of Business. The authors thank Adam Duhachek,

four anonymous reviewers, the area editor, and editor fortheir helpful comments and advice.

Appendix A. Measurement Scales

Part 1. Measures of DivergenceOriginalityThe ad's ideas are rare, surprising, or move away from the obviousand commonplace.

The ad was "out of the ordinary." ,The ad broke away from habit-bound and stereotypical

thinking.The ad was unique.

FlexibilityThe ad luid different ideas and shifted from one type of subjectmatter to another.

The ad contained ideas that moved from one subject toanother.

The ad contained different ideas.The ad shifted from one idea to another.

SynthesisThe ad combined or connected normally unrelated objects or ideas.

The ad connected objects that are usually unrelated.The ad contained unusual connections.The ad brought unusual items together.

ElaborationThe ad provided numerous details. The ad finished, extended, anddetailed basic ideas so they become more intricate or sophisticated.

The ad contained numerous details.The ad finished basic ideas so that they become more

intricate.The ad contained more details than expected.

Artistic ValueThe ad had striking visual and/or verbal elements.

The ad was visually/verbally distinctive.The ad made ideas come to life graphically/verbally.The ad was artistically produced.

Fluency'The ad contained a large number of ideasmore than expected.

The ad contained a iarge number of ideas.The ad contained more concepts than most ads.The ad had many ideas.

Imagination"The ad caused you to form vivid mental images, or made some-thing unreal come to life.

The ad allowed me to form images I have not directlyexperienced.

The ad made something unreal come to life.The ad allowed me to form vivid mental images.

Part 2. Measures of RelevanceRelevance of the Ad to You:

The ad was very meaningful to me.The ad was appropriate to me.The ad was useful to me.The ad was valuable to me.

This dimension was removed from the model in the final study.

Smith et al.: Modeling the Determinants and Effects of Creativity in AdzvrtisingMarketing Science 26{6), pp. 819-833, 2007 INFORMS 831

Relevance of the Brand to You:The product or brand was meaningful to me.Tlie product or brand was appropriate to me.The product or brand was useful to me.The product or brand was valuable to me.I do NOT care about this product/service. (R)

Relevance of the Ad to the product or brand*:The product or brand was the primary focus of the ad,The product or brand was NOT a central character in the

ad, it was more a background component. (R)The heart of this ad was what it said about the product

or brand.The product or brand did NOT seem to be related to

what went on in the ad. (R)The ad presented useful information about the product

or brand.

Part 3. Processing and Response MeasuresAmount of Attention:

The ad demanded my attention.1 examined the main elements of the ad very carefully.I tried to carefully evaluate the brand information pro-

vided in the ad.I spent considerable time analyzing the ad's message.

Motivation lo Process the Ad:I had a strong desire to examine the ad.I was highly motivated to read the ad.I really wanted to understand the ad.I was very interested in the ad.

Depth of Processing:I related parts of the ad to my own life.I used my imagination to go beyond the information pre-

sented in the ad.1 was able to imagine using the product in the ad.

Ad Attitude: What is your overall evaluation of the advertise-ment!

Bad/GoodUnpleasant/PleasantUnfavorable /FavorableNot Likeable/Likeable

Brand Attitude: Wftflf is your overaii evaluation of the advertisedbrandl

Bad/GoodUnpleasant/PleasantUnfavorable/FavorableNot Likeable/Likeable

Purchase Intentions: What (s the probability that you mil pur-chase the advertised brand in the future?

Unlikely/LikelyImprobable/ProbableImpossible / Possible

Part 4. Measures of Production Quality of the AdThe audio elements of the ad (e.g., music, voice-overs,sound effects, etc.) were of high quality.

The visual elements of the ad (e.g., images, colors, light-ing, etc.) were of high quality.

The production elements of the ad (e.g., expensive stag-ing, celebrities, action scenes, special effects, etc.) were ofhigh quality.

Overall, it must have cost a lot of money to producethe ad.

Part 5. Global Measures of Overall CreativityOverall Creativity: All tilings considered, tww creative was thead compared to the average TV adl ^

In general, the ad was very creative.The ad should win an award for creativity.The ad was not very inventive and displayed little cre-

ativity in its design. (R)Please rate the ad's overall creativity on the following

scale (1-20 points).Part 6. Global Measures of Overall Divergence andRelevanceOverall Divergence: All things considered, how unusual was thead compared to the average TV ad7

The ad was different.The ad was uncommon.The ad was unusual.

Overall Relevance, Meaningfulness, and Usefidness:The viewing experience was relevant to me.The viewing experience was useful to me.Overall, the ad and the brand were NOT really applicable

to me. (R)

Appendix B. Summary of PretestsStudy Purpose Sample Ads Independent variables

Pretest 1 Preliminary item generation

Pretest 2 Check face validity of scales with experts

Pretest 3 Examine scale psychometrics

Pretest 4 Measure ad creativityPretest 5' Test creative versus noncreative ads

13 ad agency creative directors12 advertising creative directors, and13 marketing managers340 student consumers

372 student consumers246 student consumers

I'a'test 6" Test creative ads versus average network ads 254 student consumers

Not usedNot used

2 real magazine ads

3 TV ads100 TV ads

100 TV ads

Creative/noncreative ads;prtxressing involvement

Creative/average ads

Mn Pretest 5, the procedure was the same as the final study except for the manipulation of ad creativity. In Pretest 5, the 50 creative adscame from award-winning ads from Ad Week. The 50 noncreative ads were regional/local ads recorded from TV.

''In Pretest 6, the same 50 ads from Ad Week were used as creative ads. However, 50 ads randomly recorded from major TV network(please see the description ot' the final main study) were used as the average ads.

Smith et al.: Modeling the Determinants and Effects of Creativity in AdvertisingMarketing Science 26{6), pp. 819-833, 2007 INFORMS

Appendix C. Summary of Pretest Results for Divergence Measurement Model

Measures

Originality (AVE = 0,81/0.61, a = 0.93/0,81):The ad was out of the ordinary.The ad broke away from habit-bound and stereotypical thinking.The ad was unique.

FlexibUity (AVE = 0,78/0,75, a = 0.92/0.90):The ad contained ideas that moved from one subject to another.The ad contained different ideas.The ad shifted from one idea to another.

Synthesis (AVE = 0,89/0,80, a ^ 0,96/0,92):The ad connected objects that are usually unrelated.The ad contained unusual connections. The ad brought unusual items together.

Elaboration (AVE = 0.67/0.67, a = 0.87/0,86):The ad contained numerous details.The ad finished basic ideas so that they become more intricate.The ad contained more details than expected.

Artistic value (AVE = 0.75/0,63, = 0,89/0,81):The ad was visually/verbally distinctive.The ad made ideas come to life graphically/verbally.The ad was artistically produced.

Overall creativity (AVE = 0,90/0,79, a = 0,95/0,90):The ad was creative.The ad was innovative.

Effects of divergence components on creativity:Originality factorFlexibility factorSynthesis factorElaboration factorArtistic value factor

Pretest 5Standardized

estimates

0,910,880,91

0.880.880.89

0.95

0.92

0.770.830,84

m-0.920.80

0.950.95 ., ,

0,250.190.260,170.24

Pretest 6Standardized

estimates

0.750,750,84

0.830.880.R9

0,900.870,90

080

0.86

0.850.880.641 ir

0.890.89

0.190 ^0.260.19(1.20

Notes. (1) All estimates significant (p

Smith et al.: Modeling the Determinants and Effects of Creativity in AdvertisingMarketing Science 26(6), pp. 819-833, 2007 INFORMS 833

ReferencesAnderson, J. C, D. W. Gerbing. 1998. Structural equation mtxleling

in practice: A review and recommended two-step approach.Psych. Bull. 103(May) 411-423.

Ang, S. H., S. Y. M. Low. 2000. Exploring the dimensions of adcreativity. Psych. Marketing 17(October) 835-854.

Belch, G. E., M. A. Belch. 2004. Introduction to Advertisitig and Promo-tion: An Integrated Marketing Communications Perspective, 6th ed.McGraw-Hill/Irwin, Homewood, IL.

Besemer, S. P., K. O'Quinn. 1986. Analyzing creative products:Refinement and test of a judging instrument. /. Creative Behav.20(2)115-126.

Besemer, S. P., D. J. Treffinger. 1981. Analysis of creative products:Review and synthesis. /. Creative Behav. 15(3) 158-178.

Burroughs, J. E., D. G. Mick. 2004. Exploring antecedents and con-sequences of consumer creativity in a problem-solving context./. Consumer Res. 31(September) 402-411.

Busemeyer, ). R., L. E. Jones. 1983. Analysis of multiplicative com-bination rules when the causal variables are measured witherror. Psych. Bull. 93(3) 549-562.

Diamantopouios, A., H. M. Winkihofer. 2001. Index constructionwith formative indicators: An alternative to scale development./. Marketing Res. 38(2) 269-277.

El-Murad, J., D. C. West. 2(X)4. The definition and measurementof creativity: What do we know? }. Advertising Res. 44(2)188-201.

Fornell, C, D. E Larcker. 1981. Evaluating structural equation mod-els with observable variables and measurement error. /. Mar-keting Res. 18(February) 39-50.

C)Oldenberg, ]., D. Mazursky, S. Solomon. 1999. The fundamentaltemplates of quality ads. Marketing Sci. 18(3) 333-351.

Guilford, J. P. 1950. Creativity. The Amer. Psychologist 14 444-454.Guilford, J. P. 1956- The structure of intellect. Psych. Bull. 53

267-293.Guilford, I. P. 1967. The Nature of Human Intelligence. McGraw-Hill,

New York.Haberland, G. S., P. A. Dacin. 1992. The development of a measure

to assess viewer's judgments of the creativity of an advertise-ment: A preliminary study. J. P. Sherry, jr., B. Sternthal, eds.Advances Consumer liesearch. Vol. 19. Association for ConsumerResearch, Provo, UT, 817-825.

Im, S., J. P. Workman, Jr. 2004. Market orientation, creativity, andnew product performance in high-technology firms. /. Market-ing 68(April) 114-132.

Jackson, P. W., S. Messick. 1965. The person, the product, and theresponse: Conceptual problems in the assessment of creativity.;. Personality 33(March-December) 309-329.

Jarvis, C. B., S. B. MacKenzie, P M. Podsakoff. 2003. A criticalreview of construct indicators and measurement model mis-specification in marketing and consumer research. /. ConsumerRes. 30(2) 199-218.

Joreskog, K. G., D. Sorbom. 1982. Recent developments in structuralequation modeling. /. Marketing Res. 19(4) 404-416.

Kardes, F. R., M. L. Cronley, J. J. Kellaris, S. S. Posavac. 2004. Therole oF selective information processing in price-quality infer-ence. /. Consumer Res. 31(2) 368-374.

Knjglanksi, A. W., D. M. Webster. 1996. Motivated closing of themind: "Seizing" and "Freezing." Psych. Rev. 103(2) 263-283.

Maclnnis, D. J., B. J. Jaworski. 1989. Information processing fromadvertisements: Toward an integrative framework. J. Marketing53(October) 1-23.

MacKenzie, S. B., P. M. Podsakoff, C. B. Jarvis. 2005. The prob-lem of measurement model misspecification in behavioral andorganizational research and some recommended solutions./. AppL Psych. 90au]y) 710-730.

Mishra, S., U. N. Umesh, D. E. Stern, Jr. 1993. Antecedents of theattraction effect: An information processing approach. /. Mar-keting Res. 30(August) 331-349.

Mumford, M. D., S. B. Gustafson. 1988. Creativity syndrome: Inte-gration, application, and innovation. Psi/ch. Bull. 103(January)27-43.

Smith, R. E., X. Yang. 2004. Toward a general theory of creativ-ity in advertising: Examining the rtile of divergence. MarketingTheory HI/2) 29-55.

Tellis, G. J. 1998. Advertising and Sales Promotion Strategy. Addison-Wesley, Reading, MA.

Thurson, E., X. Zhao. 1997. Television viewing behavior as anindicator of commercial effectiveness. W. D. Wells, ed. Mea-suring Advertising Effectiveness. Lawrence Erlbaum Associates,Mahwah, N], 221-237.

Till, B. D., D. W. Baack. 2005. Recall and persuasion: Does creativitymatter? /. Advertising 34(3) 47-57.

Torrance, E. P. 1987. Using the Torrance Test of Creative Viinkingto Guide the Teaching of Creative Behainor. Scholastic TestingService, Benesnville, IL.

Toubia, O. 2006. Idea generation, creativity, and incentives. Market-ing Sci. 25(5) 411^25.

Wells, W., J. Burnett, S. Moriarty. 1995. Advertising Principles andPractice. 3rd ed. Prentice-Hall, Inc., Englewood Cliffs, NJ.

Zellner, A. 1962. An efficient method of estimating seeminglyunrelated regression equations and tests for aggregation bias./. Amer. Statist. Assoc. 57 348-368.

Zinkhan, G. M. 1993. Creativity in advertising: Creativity in thejournal of advertising. /. Advertising 22(June) 1-4.