Martin Source: Journal of Public Policy & Marketing, Vol. 16, No. 2 (Fall, 1997), pp. 205-216 Published by: American Marketing Association Stable URL: http://www.jstor.org/stable/30000445 . Accessed: 23/01/2014 22:00
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp
.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.
American Marketing Association is collaborating with JSTOR to digitize, preserve and extend access to Journal of Public Policy &Marketing.
http://www.jstor.org
This content downloaded from 207.62.63.163 on Thu, 23 Jan 2014 22:00:33 PM All use subject to JSTOR Terms and Conditions
Children's
A
Intent
of
Advertising:
Mary C. Martin
The author conductsa meta-analysisinvestigatingthe extent to which children understand the intent of advertising. The results show several studycharacteristicsthat have contributedto variance in results of prior research efforts. The author uses these results to for public policymaking. develop suggestionsfor further research and implications
The
30,000 television commercials a year for various types of products (Condry, Bence, and Scheibe 1988). In addition, the percentageof commercials for toys and games during children's programminghas more than tripled since the late 1970s (Condry, Bence, and Scheibe 1988). Given the amountof advertisingtargetedat children, the concern about the effects of advertisingon children is alive and well. For example, the Children'sTelevision Act (14 U.S.C. 303), passed by Congress in 1990, reinstateda ban on program-length commercialsand issued a policy that a related advertisementmust be separatedfrom the startor close of a toy-based programby some portion of an unrelated program.In August, 1996, a mandatefrom Washington stipulated that broadcastersadhere to the Children's Television Act of 1990 by providingthree hours of educational programming each week (Eggerton 1996). Most recently, the Food and Drug Administration's (FDA) restriction of cigarette advertising directed at youths has fueled a heateddebate. Given this continuing concern, over the past three decades in the marketing,advertising,communications,and developmental literature, researchers have assessed the effects of advertising on children. This effort continues today. For example, researchershave produced empirical evidence of cigaretteadvertising'seffect on childrenand/or its relation to their smoking behavior (e.g., Mizerski 1995; Pechmannand Ratneshwar1994; Pollay et al. 1996). Other areas recently investigated include the effects of retrieval cues on children's memory and brandevaluations(Macklin 1994) and adolescents' skepticism toward advertising (Boush, Friestad,and Rose 1994). This concern about the effects of advertisingon children has two components:social and political.The social concern involves consumers', particularly parents', beliefs that For example, Hite advertisingaffects childrendetrimentally. and Eck (1987) find that consumers view children's advertising as manipulative,promotingmaterialism,stifling creativity, and disrupting parent-child relationships. This
MARY C. MARTIN is Assistant Professorof Marketing,Department of Marketing,The Belk College of Business Administration,The University of NorthCarolinaat Charlotte.The authorthanksSteve Brown, Jim Gentry, Ravi Sohi, and Steve Wise for their support and assistance on this project.
social concern also has involved the formation of public interest groups with the intent of protectingchildren from advertising. For example, in 1978, Action for Children's Television (ACT) petitioned the Federal Communications Commission(FCC) to ban all television advertisingdirected to children. The politicalconcernstems from the criticismsand actions of consumersand public interestgroups and has resultedin both self-regulation by televisionnetworksand governmental of children's One particular regulation advertising. questionis whetherand to what extent childrenunderstand the intentof (i.e., Do childrenknow whatcommercialsare and advertising what they are tryingto do?). The public policy issue relevant to children'sunderstanding of ad intentis fairness:"Areyoung children;who possess only limitedcapabilitiesfor evaluating commercialpersuasion,fair targetsfor advertising" (Kunkel andRoberts1991,p. 58)? Froma legal perspective, if children of the persuasiveintentof advertising, are unaware all advertisementsaimedat them are,by definition,unfairand/ormisleading.In turn,it has been proposedthatchildrenwho do not understand the intentof advertising are influencedmoreeasily by advertising, expressinggreaterbeliefs in commercialcontent and makingmore purchase requeststhanchildrenwith a of intent(Adleret al. 1977). greaterunderstanding advertising this research stream is Thus, importantgiven its value in as affecting past and future social values as well expressing public policy efforts by regulators,such as the FCC and the FederalTradeCommission (FTC), to protect childrenfrom when factors influencingthe relaadvertising.In particular, between and tionship age understanding of advertising intent are understood more clearly, policymakers will be better able to establish guidelines for the regulationof children's advertising(or, more basically, whether any guidelines are needed at all).
The ResearchQuestion
The purpose of this meta-analysisis to examine the literaturethathas investigatedthe extent to which childrenunderstand the intent of advertising. Specifically, those studies that address the questions "What is a commercial?" and "Why is a commercialon television/What is a commercial trying to do?" are the focus of this meta-analysis.This area of study was selected purposefully to serve as a starting point for meta-analysesof researchon children and advertising. Given the diversity and numberof studies that have 205
This content downloaded from 207.62.63.163 on Thu, 23 Jan 2014 22:00:33 PM All use subject to JSTOR Terms and Conditions
206
addressedthe relationshipof advertisingto other variables, this arearepresentsa small proportion of studies in the literature.For example, 21 articles (with 23 effect sizes) met the criteriafor inclusion in this meta-analysis. Most studies have found a positive relationshipbetween of ad intent. That is, as children get age and understanding betterthe intentof advertising.Howolder, they understand ever, thoughthis appearsto be an unambiguous finding, it is not clear what factors influence the strengthof this relationship. Specifically, this meta-analysis addressesthe following questions: 1. Do the studiesassessingthe understanding of ad intentin children sharea common effect size (i.e., aretheyhomogeIf not,whatstudycharacteristics moderate oraccount neous)? forthe variance in effectsizes acrossstudies? 2. What do the results of the meta-analysissuggest for researchers andpublicpolicymakers? After a brief review of the literature,I describe the method. Then, I present the results, including a series of analyses of the involving study characteristicsthat act as moderators relationship between age and understandingof ad intent. Finally, I providea discussion of the findings and directions for furtherresearchand public policymaking.
The Relationship Between Age and Understanding of Ad Intent: Potential Moderator Effects Measurement of Understandingof Ad Intent
Measurementof children's understandingof the intent of advertisinghas been primarilythrough verbal assessments, includingpersonalinterviews(e.g., Macklin 1983) and written questionnaires(Boush, Friestad,and Rose 1994). In the 1970s, almost all of the studies that examined children's of ad intent were personal interview assessunderstanding ments (for a review of these studies, see Macklin 1987). The idea that children are able to articulatetheir understanding of ad intent, however, has been questioned. For example, Macklin (1987) proposes that this task may be too difficult for children and that the conclusion that children do not understandad intent is problematic. Thus, several studies of ad intent through nonhave assessed the understanding verbal tasks (Donohue, Henke, and Donohue 1980; Kunkel 1988; Macklin 1985, 1987). These tasks include having childrenpoint at picturesto demonstratetheir understanding of ad intent (Donohue, Henke, and Donohue 1980; Kunkel 1988; Macklin 1987) or enact the behavior that the product spokespersondesired (Macklin 1987).
LiteratureReview
Children and Advertising
The researchliteratureon childrenand advertisingis extensive: Over the past two decades, academic researchershave attemptedto determinewhat effects advertisinghas on children. The researchon children and advertisingcan be classified into two primarycategories, each with several subcategories (Young 1990): 1.Cognitive activated as a result of watching andlisprocesses teningto advertising a. attention to advertisements b. abilityto distinguish between andprograms advertising c. understanding of ad intent d. interpretation of advertising content e. memory foradvertisements awareandrecall, (recognition nessof advertisements) f. otherprocesses involved mediators) (cognitive 2. What thechilddoes withthe information processed a. effecton knowledge, andvalues attitudes, b. effecton otherpeople(e.g., parents) c. effect on choice/consumption behavior or othertypesof behavior behavior) (e.g., antisocial Given the proliferation of researchin this area,reviews of the literature have been made throughout the past two decades (e.g., Adler et al. 1980; McNeal 1987; Young 1990). These reviews are summariesof the literature,perhaps using vote-countingmethodsto drawconclusions as to what has been done and what should be done next. However, as Young (1990, p. 39) writes, The research literature on childrenand televisionadvertising shouldnot be regarded as a series of detached experimental each one painstakingly investigations, gettingone step nearer has beenfitfulandcentered on parthe goal of truth. Progress
This content downloaded from 207.62.63.163 on Thu, 23 Jan 2014 22:00:33 PM All use subject to JSTOR Terms and Conditions
Journal of Public Policy & Marketing Althoughmore specific distinctionsbetweenthe five types of intent discussed by Blosser and Roberts (1985) may be has not taken these differencesinto important,the literature consideration. For example, unlike Blosser and Roberts (1985), most studiesseem to equate"selling"intentwith "persuasive"intent.In addition,with the exceptionof two studies (Macklin 1985; Robertsonand Rossiter 1974), the literature has not assessedempiricallythe differencesbetweeninformational/assistive intent and persuasive/selling intent and/or focused on one of these types. Therefore,of primary concern in the meta-analysis is the commonlyuseddistinction between intentand informational/assistive intent. persuasive/selling
207
Type of Ad Exposure
In the literature,measurementof children's understanding of the intent of advertisinghas occurred in the context of various researchdesigns. Surveys or interview formats, for example, have been used (e.g., Ward, Wackman, and Wartella 1975, 1977). In this type of researchdesign, subjects are interviewedand asked generalquestionsabouttheir understandingof ad intent (e.g., "What are commercials for?""Why do they have commercialson television?")and are not exposed to commercialsand/orprogramming. When the assessment of the understandingof ad intent has occurred in conjunctionwith other researchobjectives (e.g., assessment of the degree to which childrencan distinguish between commercials and programming),measurement of understandingof ad intent has generally occurred after the subjects have been exposed to a series of commercials only or commercialsand other types of programming. For example, Rubin (1974) assesses understandingof ad intent in addition to other factors, such as recall of specific commercial elements. In this study, subjects were exposed to one of two types of commercialspriorto questioning. In othercases, commercialswere embeddedin othertypes of programming.For example, Stutts,Vance, and Hudleson (1981) conduct an experiment in which children were exposed to one of threevideo segmentscontaininga cartoon and commercial, with segments differingas to type of separatordevice between the cartoonand commercial.After the experiment,the subjects were asked general questions with respect to their understandingof ad intent. Blosser and Roberts (1985) expose subjects to different versions of videotapes containingnews, an educationalspot, a commercial, and a public service announcementand then ask them of ad intent. general questions with respectto understanding It is possible that the type of ad exposure (i.e., "no ad," "ad only," or "ad mixed"-advertisements and other types of programming) may moderatethe strengthof the relationbetween of ad intent. For examship age and understanding ple, subjects may be better able to articulatethe intent of advertising after exposure to an advertisementversus no exposure to an advertisement.Merely viewing an advertisement, especially in an unrealisticsetting where no (or few) distractions occur, may "cue" subjects to the advertisement's informativeand/orpersuasive/sellingintent.
majorchanges (for a comprehensivereview, see Wilson and Weiss 1992). Specifically, in 1974, a policy was issued by the FCC that limited the amountof advertisingduringchildren's television programmingto 9.5 minutes per hour on weekends and 12 minutesper houron weekdays.These limits, according to the FCC, "struck a balance between the needs of children,who were judged uniquely susceptibleto commercial influence, and the needs of broadcasters,who were dependent upon advertising revenue to maintain the children's programofferings"(Kunkeland Roberts 1991, p. to sepa61). In addition,that policy called for broadcasters rate programcontent from commercial messages targeted towardchildrenby prohibitingprogram-length commercials that within the body of the promote products (programs story) and host selling (the use of the same charactersin commercials as are featured in adjacent programming). "Bumpers"also were required during all children's proand commercials gramming,which separatedprogramming with five-second segments, such as "And now a word from our sponsor." In 1984, many of the FCC's previous policies were reversed,driven by a shift in the U.S. government'sphilosophy toward deregulation. The FCC argued that marketplace forces should determine the acceptable amount of commercial content during children's programming. No longer was the amountof advertisingduringchildren'sprocommercials banned. gramminglimited or program-length However, the practiceof host selling was still prohibited. Then, in 1990, the Children'sTelevision Act was passed commercialswas by Congress. The ban on program-length reinstatedby the FCC, and a policy that a relatedadvertisement must be separatedfrom the startor close of a toy-based program by some portion of an unrelated program was issued. However, the definition of a program-length commercial was narrowed,which some have claimed is redundantwith host selling, a tacticalreadyprohibited by the FCC. Given these changes in regulation,it is possible that the natureof children's advertising,because it variedover several time periods,may affect the strengthof the relationship between age and children's understanding of ad intent. For commercials,which were banned example, program-length in 1990, may blur the distinctionbetween commercialsand of programmingand thus reduce children's understanding ad intent. Four time periods that correspondto changes in regulation of children's advertising were examined in the meta-analysis: pre-1974, 1975-1984, 1985-1990, and 1991-present. Publication year thus served as a proxy for U.S. legislation enacted at the time of the study.
Publication Type
Meta-analysts generally are concerned with potential sources of bias that may exist in a body of literature.One type of bias that has received considerable attention from meta-analystsis publicationbias (for a discussion of publication bias, see Begg 1994). Most commonly, publication bias is conceptualizedas selective publication.That is, the decision to publish research is influenced by the results of the study, and those studies producingstatistically significant results are more likely to be published.To test for the presence of publication bias in the literatureon children's understandingof ad intent, the type of publication was
Publication Year
Beginning in 1974, policies were establishedto protectchildren from advertising. Since then, the regulation of children's advertising in the United States has seen several
This content downloaded from 207.62.63.163 on Thu, 23 Jan 2014 22:00:33 PM All use subject to JSTOR Terms and Conditions
208
includedas a potentialmoderator by examiningtwo types of publications: published journal articles and books/edited books. The books/edited books category represents an researchliterature formatand thus may be underrepresented to bias. subject publication
Method
Identification of Articles
A computerizedsearchon Psychlit was conducted,covering the publishedliteraturefrom 1972 to the present.The Business PeriodicalsIndex (BPI), which covers publishedliterature from 1984 to the present,also was searched.The keywords used in these searches were "advertising" and "children." Issue-by-issue searches of conference proceedings, including Advances in ConsumerResearch, Developments in MarketingScience, and Enhancing KnowledgeDevelopment in Marketing,also were conducted.In addition,extensive reference lists by McNeal (1987) and Young (1990) were consulted, as was McNeal's (1991) publishedbibliography of research conducted on advertising and children. Reference lists of articles included in the meta-analysisand related articles also were searched. Finally, an active researcherin the area was contactedbut had no unpublished data relevantto the researchquestion.
Coding of Studies
Each study that met the selection criteria was coded for the potential moderatoreffects discussed previously. In addition, each study was coded for gender, race, nationality,and socioeconomic status of the subjects. In most studies, the subject pool was mixed with respect to these features, and data were aggregated;thus, using these featuresas potential moderatoreffects was not possible. As was stated previously, publication year served as a proxy for U.S. legislation enacted at the time of publication. This proceduredid not account for publication lag time; thus, those studies whose publication date was at or near "cutoff' years (e.g., Blosser and Roberts 1985) were inspected for actual data collection times. No indicationof the time of datacollection was given in any study except for Ward, Wackman, and Wartella's (1977), which specified the year of data collection as 1973. Hence, this study was included in the pre-1974 group of studies. Young's (1987) study was eliminatedfrom the analyses for publicationyear because it involved Britishchildren.A list of included studies and their relevantcodes are presentedin Table 1.
This content downloaded from 207.62.63.163 on Thu, 23 Jan 2014 22:00:33 PM All use subject to JSTOR Terms and Conditions
Journal of Public Policy & Marketing Table 1. of IncludedStudies CodedCharacteristics Measurement Type of Ad Typeof Intent of Intent Assessed Exposure Ad mixed Verbal Persuasive/selling No ad Verbal Persuasive/selling No ad Verbal Persuasive/selling Ad mixed Verbal Persuasive/selling Ad only Nonverbal Persuasive/selling No ad Verbal Persuasive/selling Ad mixed Verbal Persuasive/selling Ad mixed Nonverbal Persuasive/selling No ad Verbal Persuasive/selling Ad only Nonverbal Persuasive/selling Ad only Nonverbal Persuasive/selling Nonverbal Informational/assistive Ad only Nonverbal Informational/assistive Ad only No ad Verbal Persuasive/selling Verbal Informational/assistive No ad Ad only Verbal Persuasive/selling Ad mixed Verbal Persuasive/selling Ad mixed Verbal Persuasive/selling No ad Verbal Persuasive/selling No ad Verbal Persuasive/selling Ad mixed Verbal Persuasive/selling No ad Not included Persuasive/selling Ad mixed Verbal Persuasive/selling Publication Year 1985-1990 1991-present 1975-1984 1975-1984 1975-1984 1975-1984 1975-1984 1985-1990 1975-1984 1985-1990 1985-1990 1985-1990 1985-1990 1975-1984 Pre-1974 Pre-1974 1975-1984 1975-1984 Pre-1974 Pre-1974 1991-present 1985-1990 1975-1984
209
Study BlosserandRoberts (1985) andRose(1994) Boush,Friestad, et al. (1981) Butter Christenson (1980)a andDonohue Henke, Donohue, (1980) andHawkins Faber, Perloff, (1982) GainesandEsserman (1981) Kunkel (1988) Macklin (1983) Macklin (1985)-Study I Macklin (1985)-Study 2 Macklin (1987)-Study 1 Macklin (1987)-Study 2 andHenke(1978) Meyer,Donohue, Robertson andRossiter (1974) Rubin (1974) andStutts (1982) Stephens Stutts, Vance,andHudleson (1981) Ward (1972) andWartella Ward, Wackman, (1977) WilsonandWeiss(1992) Young(1987) andGianinno Zuckerman (1981)
Publication Type article Journal article Journal article Journal article Journal article Journal article Journal book Book/edited article Journal article Journal article Journal article Journal article Journal article Journal article Journal article Journal article Journal article Journal article Journal article Journal book Book/edited article Journal book Book/edited book Book/edited
derive one effect size. In one study (Kunkel 1988), means deviationsfor youngerand older subjectswere and standard provided, so I calculated a Hedge's g and then convertedit to r by a formulaprovidedby Rosenthal(1991, p. 24). For the other studies included in the meta-analysis,r was calculated by correlatingage and grade level with understanding of ad intent. A spreadsheetwas used to enter freof intent quencies of age and grade level and understanding reportedin each study and to calculate a correlationcoefficient between the two variables. In some studies, understandingof ad intentwas reportedin low, medium,and high levels (e.g., Ward,Wackman,and Wartella1977). In others, understandingof ad intent was recordedmerely as success (understandingof intent exhibited) or not success (understandingof intentnot exhibited).For example, Stephensand Stutts(1982) providedpercentagesof three-,four-, and fiveyear-olds who correctly and incorrectly responded when asked, "Why are commercialson TV?" of ad intent was In some studies in which understanding measuredthroughverbalresponses,a series of questionswas used. For example, Blosser and Roberts (1985) asked subjects threequestionsto ascertainintent:"Whatis a commercial?" "Why are they on TV?" "Whatdo they do?" In that study,responsesto the questionswere pooled andcoded, and for those who correctlyarticpercentageswere then reported ulated intent.In other studies that measuredintentthrougha series of questions, percentages were reported for correct/incorrectresponses to each question (e.g., Stutts, Vance, and Hudleson 1981). When percentages were reportedfor each question, I calculatedcorrelationsfor each and then calculatedan averageof the correlationsfor use in the meta-analysis. Ward, Wackman, and Wartella (1977)
provide a correlationof .66 between gradeand intent.Howcalever, this value cannotbe confirmedthroughspreadsheet culations. Thus, I calculated an average correlationof the three questions designed to measuread intentand used it in the meta-analysis. In the experimentalstudies (Kunkel 1988; Stutts, Vance, and Hudleson 1981; Wilson and Weiss 1992), no effect on nor did age of intentwas found for treatments understanding interact with treatments.Thus, deriving an effect size was not problematic. In all studies but two, sample sizes were reportedclearly. However, in Meyer, Donohue, and Henke's (1978) study, sample sizes variedacross threequestionsmeasuringunderstandingof ad intent and were not reportedfor each. Thus, a sample size of 177 was used in the meta-analysis,a conservative estimate derived from informationprovidedin the article. Similarly, in Robertsonand Rossiter's (1974) study, a conservativeestimateof samplesize was derivedand used in the meta-analysis(n = 274), because sample sizes varied across several variablesof interestin that study. In Table 2, I list the studies includedin the meta-analysis of of understanding along with the method of measurement ad intent, correlationsbetween age and intent, average correlationsbetween age and intent,and sample sizes for each. Before averages were calculated,each r was converted to a Fisher's z (Rosenthal 1991, p. 21). Fisher's z values are provided in parenthesesafter each correlation. To estimate a mean effect size (a common correlation coefficient) among the studies in the meta-analysis,each r was transformedto z and a weighted average of the studies (z,) was calculated. Individualstudy effects were weighted by an estimate of the inverse of the variance(N - 3) to give
This content downloaded from 207.62.63.163 on Thu, 23 Jan 2014 22:00:33 PM All use subject to JSTOR Terms and Conditions
210 Table2.
Children's Understanding of the Intent of Advertising IncludedStudiesand TheirEffectSizes Correlation BetweenAge and Intenta Average Correlation BetweenAge and Intenta
Study
Blosser and Roberts(1985)
.59 .54
(.68) (.60)
.56 (.64)
90
427
80
30d
Donohue, Henke, and Donohue (1980) Faber,Perloff, and Hawkins (1982) Gaines and Esserman(1981) Kunkel(1988) Macklin (1983)
.18
(.18)
97
.25e (.26)
67
104 152 35
Macklin (1985)-Study I Macklin (1985)-Study 2 Macklin (1987)-Study 1 Macklin (1987)-Study 2 Meyer, Donohue, and Henke (1978) Robertsonand Rossiter (1974)
.15 .08
(.15) (.08)
.28e (.29) .49e (.53) .45e (.49) .37e (.38) .49e (.54)
.47f (.51) .02 (.02) .52 .32 .33 .14 .57 .57 .48 .46 (.58) (.33) (.34) (.14) (.65) (.65) (.52) (.50) .43 (.45)
-
274
Rubin (1974) Stephens and Stutts (1982) Stutts, Vance, and Hudleson (1981) Ward(1972)
72
104 108
.44 (.48)
.47 (.51)
64
This content downloaded from 207.62.63.163 on Thu, 23 Jan 2014 22:00:33 PM All use subject to JSTOR Terms and Conditions
211
547
94
Young (1987)
123
64
of averages wereperformed areprovided in parentheses after eachcorrelation. aCalculations on Fisher's z transformations of thecorrelation which coefficients, of correct werenotgivenperquestion butin aggregate. bPercentages responses usedto calculate thecorrelation coefficient andsellingintent combined. correct to persuasive cPercentages represented responses wasbasedon a group dEffect size estimate of subjects in the"commercial only"experimental group. eCorrelation wasprovided by theauthors. forpersuasive intent was usedexceptin the moderator fortypeof intent, in whichthecorrelation for assistive intent fInthisstudy,the correlation analysis was used. of .66 between levelandawareness of thepurpose of television wasreported to be consistent, a commercials However, grade by theauthors. gAcorrelation wascalculated correlation foreachquestion, andthenanaverage Inaddition, because sizesvaried eachcorcorrelation wascalculated. sample perquestion, relation wasweighted theaverage size in calculating correlation. by thesample hThe correlation between of ad intent atTimeI (versus Time2) wasused. age andunderstanding
more precise estimates greater weight (Hedges and Olkin 1985, p. 231).
Q = I ( n, - 3)(z,- z)2,
where z is the mean of the weighted z-transformed correlations. The Q statistic has an approximate chi-squaredistribution with k - I degrees of freedom,where k is the number of studies included in the analysis. A nonsignificantQ statistic indicateshomogeneityof effect size. If homogeneityis not achieved, combiningestimatesof effect size across studies is misleading, because the studies do not share a common underlyingeffect size (Hedges 1986, p. 383). In addition, nonhomogeneity of effect size suggests the potential presence of moderatorvariables.
Results
For23 studyeffects, a cumulativeN of 2934 and an estimate of r of .37 (with a 95% confidence interval of .34 to .40) resulted from the meta-analysis. However, homogeneity across all 23 studies was not achieved(Q = 112.36,p < .001). For the analyses for publication year, one study (Young 1987) was eliminatedfrom the total pool of studies,because the subjects used were Britishchildren;therefore,the study did not have relevance (because publicationyear servedas a proxy for U.S. legislation enacted at the time of the study). Homogeneity across the remaining 22 studies was not achieved (Q = 111.68, p < .001, cumulativen = 2811). Because homogeneity of effect size was not achieved in the entire set of studies, moderatoranalyses were undertaken for each potential study effect described previously
of ad intent, type of intent (measurementof understanding assessed, type of ad exposure, publicationyear, and publication type). The studies were categorizedaccordingto each potential study effect and tested for homogeneity within each subgroup (Qw). For example, the set of studies that used verbal assessments of intent was tested to determine whetherthose studies sharea common effect size; a separate test of the homogeneity of the set of studies that used nonverbalassessmentswas conducted.Consistentwith previous meta-analyses(e.g., Brown and Peterson 1993; Brown and Stayman 1992) and recommendationsof methodologists (e.g., Hedges 1986; Hedges and Olkin 1985), studies were deleted in the meta-analysis if homogeneity still was not achieved after subgroupingthe studies. The study with the largest weighted deviation from the mean effect size was eliminated,the mean effect size recalculated,and the homogeneity test repeated(Hedges and Olkin 1985, p. 256). This process was conducted until homogeneity was achieved in each subgroup. Then, tests were conducted to assess whethersignificant differencesamong mean effect sizes for the subgroups exist (QB). The results of the moderator analyses are presentedin Table 3. In all subgroups,except informational/assistive intent as a type of intent, homogeneity was achieved by eliminating one or several outliers. In the meta-analysis,the amount of data eliminatedrangedfrom 10.9%to 42.9%. There are differing opinions as to the amount of data that should be deleted to achieve homogeneity (see Hedges and Olkin 1985, p. 250). However, when outliers were eliminated,the results of the between-group homogeneity tests did not change, so deletion of studies did not change the conclusions drawnfrom testing for between-groupdifferences.For example, the mean effect size for studies using verbal assessments was significantly greater than the mean effect size for studies using nonverbalassessmentbefore as well as
This content downloaded from 207.62.63.163 on Thu, 23 Jan 2014 22:00:33 PM All use subject to JSTOR Terms and Conditions
This content downloaded from 207.62.63.163 on Thu, 23 Jan 2014 22:00:33 PM All use subject to JSTOR Terms and Conditions
213
Publication Type
Finally,a significantdifferencewas foundbetweenthe effect sizes for publicationtype. The effect size for books/edited books (r = .55) is muchhigherthantheeffect size for published reasonfor the variance journalarticles(r = .37). No apparent in effect sizes exists. Forassessmentof the existenceof publication bias, the statisticalsignificanceof each effect size was examined (i.e., Is the effect size significantlydifferentfrom different fromzero(p zero?).Botheffect sizes aresignificantly < .001; using a test describedby Hedges and Olkin 1985, p. biasis limitedin thisbody 231). Thus,it seems thatpublication of literature; the reasonfor the more frequentpublicationof work (i.e., books and edited journalarticlesthanunpublished is not the statistical books) significance of their findings, such findings. becausebothtypes of workhave produced
Discussion
The first purpose of this meta-analysis is to determine of ad intent in whether studies assessing the understanding children share a common effect size. Given the results, the second purposeof this meta-analysisis to provideguidance for futureefforts by researchersand public policymakers. The results suggest that these studies are variant with respect to effect size. Although the relationshipis generally positive, the strengthof relationshipvaries across studies, and thus drawingconclusions from the entire set of studies is problematic.Several study characteristics were examined and found to moderate the relationshipbetween age and understandingof ad intent. Differences in methods across the studies led to much of the variance in results of studies of assessing the relationshipbetween age and understanding ad intent. First, the results suggest that nonverbalassessments may be more appropriate for young childrenthan verbal assessments. A lower correlationfor nonverbalassessments sugof ad intentis morestableacrossage gests thatunderstanding levels. Nonverbalmeasures seem to disclose understanding of ad intent that was previously undetectedbecause of children's inabilitiesto verbalizethatunderstanding. Second, the results suggest that the distinctionbetweentypes of intent is an importantone. Understanding of the informational/assistive intentof advertisingentails comprehension of the notion that "commercialstell you about things." However, understanding of the persuasive/selling intent of advertising is more sophisticatedand entails comprehensionof the notion that "commercialstry to make you buy things."Third, the resultsof the meta-analysissuggest thatthe type of exposure before measurement of understanding of ad intent("adonly," "ad mixed," and "no ad") affects the level of understanding displayed by children. For example, confusion may occur when a child sees an advertisement in isolation,thusdecreashis or her of ad intent. Fourth,the results ing understanding with respect to publicationyear appearto be beneficial for the advocates of regulationof children's advertising.Fifth, researchersin this area should feel comfortablethatpublication bias does not seem to exist in this body of literature. Guidance for Researchers Several issues should be considered by researchers who engage in future efforts in this area. Specifically, some of
Type of Ad Exposure
A significant between-class effect was found with respectto the type of advertising exposure before measurementof understandingof ad intent. The effect sizes for "no ad" exposure (r = .39) and "ad mixed" exposure (r = .45) are higher than the effect size for "ad only" exposure (r = .29). This suggests thatconfusion may occur when a child sees an advertisementin isolation, because it does not enable him or her to distinguish between programmingand commercials and thus to articulate or demonstrateunderstandingof ad intent.Five of the six studies in the "adonly"category,however, were also studies that used nonverbalassessments of understandingof ad intent. This may explain partiallythe relatively low correlationfor this groupof studies. Publication Year A significantbetween-classeffect was found with respectto publication year. The effect size for pre-1974 studies (r = .51) is higher than the effect sizes for 1975-1984 studies (r = .30), 1985-1990 studies (r = .24), and 1991-present studies (r = .10). This suggests that policies establishedto protect childrenfrom advertising,beginningin 1974, have been successful in enhancing children's understandingof ad intent. That is, perhapsthe understanding of ad intent has been enhanced in younger children,and thus understanding across ages is more stable (hence a lower correlation).
This content downloaded from 207.62.63.163 on Thu, 23 Jan 2014 22:00:33 PM All use subject to JSTOR Terms and Conditions
214
the moderatingvariablesused in this meta-analysis should be manipulated directly in future experimentation. For example, further research should manipulate the type of exposure ("adonly," "ad mixed,"and "no ad") to clarify its relationship with age and understanding of ad intent. Researchersalso should note that an importantissue here is external validity. The "ad mixed" condition, in which children are exposed to cartoons, commercials, and separator devices between the cartoon and commercial, seems most consistent with the currentlegislation, the Children'sTelevision Act of 1990. In addition, future experimentationshould measure both informational/assistiveand persuasive/selling intent. For example, the conceptual and methodological difference between these two types of intentis not clear. Conceptually, these types of intent seem to fall on a continuum from low levels of understanding(the informational/assistivefunction) to higher levels of understanding (the persuasive/selling function). However, the literature(and, consequently, this meta-analysis)has not treatedthemconsistentlyas such. For example, Roberts(1982) argues thatthese functions are distinct and thatthe perspectivea child is able to take determines his or her level of understanding of ad intent. Specifically, if a child is able to take the buyer's perspective,then he or she will understandthat advertising's function is to show products that are available to buy. Conversely, if a child is able to take the seller's perspective, then he or she will understand thatadvertising'sfunctionis to persuadethe audience in order for the advertisersto profit (for a discussion of these perspectives,see Macklin 1987). Thus, further research should determine whether the types of intent are independent of each other (i.e., whether understandingof each type of intent develops simultaneously and/or at the same rate) or whether an understandingof informational/ assistive intentis subsumedby a higherlevel of understanding of persuasive/sellingintent. Although the results of the meta-analysis suggest that nonverbal measures of intent may detect understandingin young children that verbal measures cannot, the development of reliable nonverbal measures should continue. Althoughsome studies claim to measurethe persuasive/selling intent through nonverbal measures (Donohue, Henke, and Donohue 1980; Macklin 1985), this is questioned by Macklin (1987), who argues that a nonverbaltask of asking children to point to "what Toucan Sam wants you to do" of the selling intentof advermay not assess understanding tising. Rather,Macklin (1987) suggests that this task may assess a child's acceptance of advice from the product In addition,Macklin(1985) finds dramatic spokescharacter. differences in understanding of ad intent when the response alternatives are altered in a "point to the picture" task. Specifically, when a more difficult set of nonverbal measures is developed (i.e., a set with additionalphoto alternatives that illustrate the advertised product), children's demonstrationof understanding of ad intent is lower. Thus, furtherresearchshould continue to develop nonverbalmeasures with the goals of determiningwhetherthese tasks are ascertaining informational/assistiveintent, persuasive/selling intent,or some other type of information,and the extent to which the results are affected by the nature of the task. Developments in verbal assessments should continue as
This content downloaded from 207.62.63.163 on Thu, 23 Jan 2014 22:00:33 PM All use subject to JSTOR Terms and Conditions
Journal of Public Policy & Marketing itation of this body of literature, because research has demonstrateddifferent levels of understanding of ad intent between white and African-Americanchildren. For example, Meyer, Donohue, and Henke (1978) comparetheirfindings to Ward's (1972) and find that 56% of the AfricanAmerican children in their study never showed any awareness of the intentof advertising,whereas85% to 90% of the white childrenin Ward's(1972) studyhad at least a minimal level of understanding. Donohue, Meyer, and Henke (1978) also find that39%of theirsampleof African-American children had no understanding of ad intent, whereas 18.9% of their sample of white children had no understanding of ad intent. They thereforeconclude (p. 57) that this is "disturbing in the sense that without the ability to understandthe manipulativeand biased approachtakenby advertisers,millions of younger black children may well be vulnerableto the influence of commercials....[T]hese same black children watch so much TV and so many commercials,far more than their white counterparts." Second, public policymakersshould ask whetheralternative explanationsfor these findings exist. The interpretation of the results with respect to publicationyear is not certain because it cannot be regardedas causal. That is, there is no assurance that the results of the meta-analysisare attributable to changes in legislation. These lower effect sizes for later time periods may indicateincreasedunderstanding due to other factors, such as increasingefforts by public school systems to educatechildrenaboutadvertising.Alternatively, this result may be a function of the fact that more recent studies have used more nonverbalassessments. Third,publicpolicymakersshouldrecognizethatthe small numberof recent studies assessing children's understanding of ad intent representsa limitation.The most recent studies were publishedin 1992 (Wilson and Weiss 1992) and 1994 (Boush, Friestad, and Rose 1994). Because the Children's Television Act representsa majorchange in how advertising is presentedto childrenand was passed only two years prior to the publicationof Wilson and Weiss (1992), our knowledge of the effectiveness of this act is severely lacking. It is hoped that this meta-analysisawakens public policymakers and researchersto the need for more work in this area,parassessed in the ticularly work that integratesthe moderators and the made meta-analysis suggestions previously. This work then can be appliedto answeringa more basic question essential to public policymakers: How and in what ways should childrenbe educatedas to the intentof advertisers?
215
as children's ability to distinguish between programming and commercials, would establish with more confidence its status as a moderator.Then policymakerswould be better able to establish and/orrevise guidelines for the regulation of children's advertising.As Kunkel and Roberts (1991, p. 69) write, "researchis indeed a necessarycomponentin the policy-making process, at least in those instances in which its bearing is clear and direct."I believe that these research findings in combinationwith those of futureresearchefforts (both primarystudies and meta-analyses)will have a clear and direct impact.
References
Z. Friedlander, Richard Gerald S. Lesser, Laurene P.,Bernard Adler, JohnR. Rossiter, ThomasS. Robertson, and Scott Meringoff, ontheEffects Ward on (1977),Research of Television Advertising DC:U.S.Government Children. Office. Washington, Printing -, Gerald S. Lesser,Laurene Thomas S. RobertMeringoff, andScottWard son,JohnR.Rossiter, (1980),TheEffects of Teleon Children.Lexington,MA: Lexington vision Advertising Books. Bias,"in TheHandbook of Begg, Colin B. (1994), "Publication Research Harris V. Hedges,eds. Synthesis, Cooperand Larry New York: RussellSageFoundation, 399-409. B. Bengen, andT. G. Johnson Bever,T. G., M. L. Smith, (1975), "YoungViewers'Troubling Responseto TV Ads,"Harvard Business 53 (November/December), 109-20. Review, and ScottWard(1972),"A Cognitive Blatt,Joan,Lyle Spencer, of Children's Reactions to Television AdverDevelopment Study andJ. P. Murray, DC: eds. Washington, stein,G. A. Comstock, U.S. Government Office,452-67. Printing F. Roberts in Blosser, (1985),"AgeDifferences BetsyJ.andDonald Children'sPerceptions of Message Intent,"Communication TV Advertising "Adolescent Toward andKnowledge Skepticism of Advertiser 21 (June), Journal Tactics," Research, of Consumer 165-75. StevenP. andRobert A. Peterson and Brown, (1993),"Antecedents of Salesperson Job Satisfaction: Consequences Meta-Analysis and Assessmentof Causal Effects,"Journal of Marketing 30 (February), 63-77. Research, M. Stayman andDouglas andConse(1992),"Antecedents of Attitude Toward theAd:A Meta-Analysis," Journal of quences
ConsumerResearch, 19 (June),34-5 1. Butter, Eliot J., Paula M. Popovich, Robert H. Stackhouse, and Research, 12 (October),455-84. Boush, David M., MarianFriestad,and GregoryM. Rose (1994), tising," in Televisionand Social Behavior,Vol. 4, E. A. Rubin-
A Final Note
Finally, it is hoped thatresearcherswill make efforts to conduct and accommodate future meta-analyses. As was the case here, the results of meta-analyses in other areas relevant to childrenand advertisingwill be important in guiding future research efforts and shaping public policy efforts. Accommodation for these efforts will be provided by researchers who report clearly information necessary for calculations of effect sizes and all pertinentaspects of their methodology when conductingprimaryresearch.The unanswered issues noted previously reinforcethe need for future meta-analyses. In particular,using publication year as a moderatorin meta-analysesthat address relatedareas, such
Peter Gilbert Inforof Consumer Christenson, (1980),"TheEffects mation Announcements on Children's of Processing Perceptions Commercials and Products," doctoral dissertation, unpublished of Communication, Stanford Department University. "Children's of TV Commercials and (1982), Perceptions
Products:The Effects of PSAs," Communication Research, 9
491-524. (October),
Condry,John, PatriciaBence, and Cynthia Scheibe (1988), "NonprogramContent of Children'sTelevision,"Journal of Broadcasting and ElectronicMedia, 32 (3), 255-70.
This content downloaded from 207.62.63.163 on Thu, 23 Jan 2014 22:00:33 PM All use subject to JSTOR Terms and Conditions
216
Donohue, Thomas R., Lucy L. Henke, and William A. Donohue Journal (1980), "Do Kids Know WhatTV CommercialsIntend?" Research,20 (October),51-57. of Advertising TimothyP. Meyer,and Lucy L. Henke (1978), "Blackand --, White Children:Perceptionsof TV Commercials,"Journal of 42 (October),34-40. Marketing, Witha Mandatefrom Eggerton,John(1996), "TheNext Generation: to ProvideChildren'sShows, TV Programmers Look Washington to Combine FCC Friendlinesswith FinancialSuccess," Broadcasting and Cable, 126 (August 12), 18-19. M. Perloff,andRobertP. Hawkins(1982), Faber,RonaldJ., Richard "Antecedents of Children'sComprehension of Television AdverJournal 26 575-84. (2), of Broadcasting, tising," Gaines,Leslie andJune Esserman(1981), "A Quantitative Study of Young Children's Comprehensionof Television Programsand Commercials,"in TelevisionAdvertisingand Children: Issues, Researchand Findings,June F. Esserman,ed. New York:Child ResearchService, 96-105. Hedges, Larry V. (1986), "Issues in Meta-Analysis,"Review of Researchin Education,13, 353-98. and IngramOlkin (1985), Statistical Methodsfor MetaAnalysis.New York:AcademicPress. Hite, RobertE. and Randy Eck (1987), "Advertisingto Children: Attitudes of Business Vs. Consumers,"Journal of Advertising 40-53. Research,27 (October/November), and Host-SellingTelevision ComKunkel,Dale (1988), "Children 71-92. mercials,"Communication Research, 15 (February), and Donald Roberts (1991), "Young Minds and MarketJourplace Values:Issues in Children'sTelevision Advertising," nal of Social Issues, 47 (1), 57-72. Macklin, M. Carole (1983), "Do Children UnderstandTV Ads?" Journalof AdvertisingResearch,23 (February/March), 63-70. S the Selling Intent (1985), "Do Young ChildrenUnderstand of Commercials?" Journalof Consumer Affairs, 19 (2), 293-304. of the Informational (1987), "Preschoolers'Understanding Function of Television Advertising," Journal of Consumer Research, 14 (September),229-39. (1994), "The Effects of an AdvertisingRetrievalCue on Young Children'sMemoryand BrandEvaluations," Psychology and Marketing,11 (May/June), 291-311. McNeal, James J. (1987), Children as Consumers:Insights and Implications.Lexington,MA: LexingtonBooks. (1991), A Bibliographyof Researchand Writingson Marto Children.New York:LexingtonBooks. ketingand Advertising Thomas R. Donohue, and Lucy L. Henke Meyer, Timothy P., Journal of' (1978), "How Black ChildrenSee TV Commercials," Research, 18 (October),51-58. Advertising Mizerski, Richard (1995), "The Relationship Between Cartoon TradeCharacter Recognitionand AttitudeTowardProductCategory in Young Children,"Journal of Marketing,59 (October), 58-70. Paget, K. Frome,D. Kritt,and L. Bergemann(1984), "Understanding StrategicInteractions in Television Commercials: A DevelopmentalStudy,"Journalof AppliedDevelopmentalPsychology,5 (2), 145-61. Pechmann,Cornelia and S. Ratneshwar(1994), "The Effects of Antismokingand CigaretteAdvertisingon Young Adolescents' Perceptions of Peers Who Smoke," Journal of Consumer Research,21 (September),236-51. Pollay, Richard W., S. Siddarth,Michael Siegel, Anne Haddix, Robert K. Merritt,Gary A. Giovino, and Michael P. Eriksen (1996), "The Last Straw? Cigarette Advertising and Realized
This content downloaded from 207.62.63.163 on Thu, 23 Jan 2014 22:00:33 PM All use subject to JSTOR Terms and Conditions