Rust Observation Parents and Children Shopping Together

Embed Size (px)

Citation preview

  • 8/3/2019 Rust Observation Parents and Children Shopping Together

    1/7

    ANGBOURNE RUST OBSERVATIONS:PARENTS AND CHILDRENSHOPPING TOGETHER:A NEW APPROACH TO THE QUALITATIVEANALYSIS OF OBSERVATIONAL DATA

    Thanks are due to all the committeemembers and their colleagues whomade the observations on which thisstudy is built. It was clear from the nar-ratives that the field observations pro-voked much good thought and not alittle amusement. Companies who lenttheir staff and resources were: ChildGrowth & Development, Children'sTelevision Workshop, Genera! Mills,General Foods, NPD, Langbourne RustResearch, and Sheridan Associates. Ob-servations were made by Ellen Sackoff,Jean O'Connor, Barbara Gussaway, D.Welcher, Allison McMorris, MarthaMontes, Julie Seyfert, Susanne Rust,Ned Rust, Susan Shannon, Jessica Lash,Karen Serrano, Mary Rae Esposito, AnnMarie Burgle, Elaine Bourke, Raisa Gil-martin, Jayne Zimmy, Scheila Jalayer,Eva Coromilas, Lee Nielsen, Dan Zee-cola, Maryanne Pagano, RoseanneSheridan, and a number of perhaps too-modest observers who provided us withobservations complete in every way ex-cept for a record of their authorship.

    F ield observations of parentsand children shopping to-gether were analyzed tofind out how their in-store be-havior varied with the age of thechild. The objective was to con-ceptualize the factors that influ-ence in-store decisions, with aneye toward developing new mar-keting strategies which wouldtarget families with children ofspecific ages.This was a qualitative researchproject, insofar as it set out togenerate new concepts, not totest a priori ones. But it differedfrom traditional qualitative re-search in using observationaldata rather than in-depth con-versations with consumers andin taking a rigorous approach togenerating, refining, and testingthe newly emerging theory.A Qualitative StudyBased onObservational Data

    Observational field studies areuniquely suited to studying chil-dren shopping with their par-ents. Interview-only designs arerestricted to sampling peoples'verbalizations. They are verywell suited to shedding light onthe thinking, overtly or covertlyvocalized, that individuals in-voke in their marketplace deci-sions. The more articulate andintrospective the respondentsare, and the more focused andinvolved they are in the deci-sion-making process, the morepowerful the quaUtative inter-view can be.

    But there are times when non-intrusive observational methodsmay provide a more completeand accurate picture of the pur-chasing experience. When thepurchasing process is more im-pulsive than planned, when theshoppers are in groups, attend-ing to each other as much as tothe shopping, when the keyplayers are children who are in-capable of introspective, verbalanalysis (either during the shop-ping process or in subsequentinterviews), it is unlikely thatverbal interviewing would get anaccurate or balanced picture ofthe factors that shape the con-sumer's behavior. Children arespontaneous creatures, living inthe here and now, unplanned,impulsive, dominated by theirtransitory mood states andsteered by the physicalstimulation of their immediateenvironments.To get a better understandingof the factors which so influencefamily purchasing, but which areso hard to find out aboutthrough verbal interviews, thisstudy turned its attention to thein-store environment, and it em-ployed nonintrusive observa-tional methods for studyingwhat goes on there.Data

    The data for this study wereanecdotal narratives about par-ents and children shopping to-gether, collected by members ofthe ARF's Children's ResearchJournal ol ADVERTISING RESEARCHJULY/AUGUST 1993

  • 8/3/2019 Rust Observation Parents and Children Shopping Together

    2/7

    O B S E R V A T I O N S

    Children are spontaneouscreatures, living in the hereand now, unplanned,impulsive, dominated hytheir transitory mood statesand steered hy the physicalstimulation of theirimmediate environments.

    Council ^Nho visited supermar-kets and toy stores. Posing asshoppers, they would wait in anaisle, watching for parents andchildren to travel down the aisletogether. On seeing a shoppingparty enter the aisle, they wouldestimate the child's age, recordsome basic information aboutthe shoppers, and then takenotes on what they said anddid. As soon as the party left theaisle, the observer would finishup the notes and wait until thenext party appeared.

    Two hundred records werecollected. Data were a conve-nience sample, gathered when-ever the observers were in thestore and parents came downthe aisles with their children.Although one could not, withconfidence, extrapolate thequantitative patterns in thesedata to the population at large,the purposes of this study werequalitative. The intent was toexamine the patterns of parent-child interactions and to developways of conceptualizing whatgoes on. For that, the sampleserved well. Exhibit 1 showssome representative records.Analysis

    The anaiysis procedure fol-lowed the principles ofGrounded Theory Development,a method for generating theoryfrom case studies first articulatedby Glaser and Strauss (1967),

    which has become an importanttool in academic sociology andanthropology. (See Strauss andCorbin (1990) for a good back-ground and introduction to thismethod of qualitative analysisand theory building.) Rust (1971,1985) has extended the methodto the study of a number ofproblems in consumer behavior.

    Grounded Theory Develop-ment brings structure to the in-ductive side of social science.Hypotheses are generated,tested, and refined through asystematic process of coding andrecoding of the original observa-tions. By keeping the researchergrounded in the raw observa-tions, rather than getting lockedinto a preset matrix of scores orratings, it avoids some of thebuilt-in hmitations of more tradi-tional, deductive approaches. Hypothesis generation, or induc-tion, was done by contrastingthe cases involving the veryyoungest children againstthose with the very oldestones. Testing applied the hypotheti-cal attribute definitions one ata time to the whole data set:evaluating how well they dif-ferentiated old-child fromyoung-child cases. Refinement was another induc-tive phase, rooted in study ofthe case records. It relied ondeviant case analysis: contrast-ing the deviant cases againstthe conforming ones, modify-

    ing the hypotheses to excludeas many of the deviants aspossible, and Integration was accomplishedby testing all of the hypothe-ses as a set, using a logisticregression model to identifythe ones which made uniquecontributions and droppingthose which added nothing toour ability to predict the age ofthe children involved in eachcase.

    "Young" child cases werethose involving children aged and younger. "Old" child caseinvolved children from 6 to 14There were 97 young-child casand 103 old-child cases in thedata set.For keeping track of therecords, searching throughthem, rating them vis a vis varous attributes, and evaluatinghow closely they were associatwith age of child, SOLIDGROUND, a specialized text adata base program written bythe author, was used. Otherprograms, some commerciallyavailable, are reviewed in Field

    ing and Lee (1991).Hypothesis Generation

    The central question of thestudy was how the cases witholder children differed fromthose with younger children.The analysis took the form of asearch for attributes whichwould most consistentiy differentiate the old from young chicases.Rather than start the search forming hypotheses before thefield work and collecting data confirm (or refute) them.Grounded Theory Developmenbegins hypothesis generationwith the cases themselves. Usa strategy called "theoreticalsampling," the analyst identifia limited set of cases that highlight the distinction for which theory is being soughtin this

    case, age-of-childand usesthem as the springboard forinduction.Twenty-four cases were cho-sen for analysis in the hypothesis-generation phase: twelvewith very little kids (involving to 3-year-oIds) and twelve withrelatively old ones (10- to 13-year-olds). These were read ovand over, and a list was made hypotheses about how theymight differ. Those which con-

  • 8/3/2019 Rust Observation Parents and Children Shopping Together

    3/7

    O B S E R V A T I O N S

    sistently differentiated the old-from the young-child cases werecarried on to the next phase.The number of cases to use ineach high-contrast set is a com-promise. Too few cases increasethe risks of false hypotheses(which fail when tested againstthe full data set) and of missedhypotheses (which would haveshowed up if a larger sample ofcases had been contrasted). Buton the other hand, too manycases strain the channel capacityof the analyst, meaning that thesets cannot be apprehended intheir entirety, the induction pro-cess is hobbled, and the hypoth-eses tend to become piecemealand erratic. Contrast-sets ofabout 12 cases each appear towork well: they are smallenough to be comprehensibleyet large enough to get at themany of the major patterns un-derlying the data.Hypothesis Testing

    For each hypothesized attrib-ute, all 200 cases in the data setwere content analyzed and clas-sified. A blind rating system wasused: the rater had no cluesabout the age of the child in acase before the attribute-ratingwas made. This helped shieldthe rater from "wishful think-ing" and ensured that the classi-fications remained objective.When the ratings were com-pleted, cross tabs were run toevaluate how well the attributesdifferentiated old- from young-child records.Some of the hypothesized at-tributes clearly identified young-child cases. For example, 28 per-cent of the cases with kids under6 involved some physical in-volvement with the productother than picking up or carry-ing it (such as poking or playingor consuming it) versus only 11percent of the cases with olderchildren.

    Exhibit 1Sample Young-Child and Old-Child Data RecordsCase #171 (young child)Observer; Julie SeyfertStore: Supermarket: Waldbaum's, Mt. Kisco. NYAisle: Breakfast cerealsShopping party: Mother, 1 girl age 7, 1 boy age 1. Both children in cart.No shopping list visible, no coupons visible.Field Observations:Party enters section, girl in bottom of cart, boy in top. Girl immediately points to CountChocula box and exclaims something to the effect of "Look at those eyes!" Boy then joins in"Oh yes, lets get that one, can we get that one ?" Mom says something in Spanish as she picksup the box of Count Chocula. Group talks in Spanish, a few things are saidneutral intonemom puts Count Chocuta in cart and Vi/heels away.Field Thoughts:Mom seems willing to please, didn't look at side panels or anything. Ch ildren were obv iouslydrawn to the box and seemed to make decision solely on eyes on box. Girl in bottom of cartless vocal than boy. (Count Chocula picture on the box had plastic piece over the face thatmade the eyes appear to move as the viewer walked by.)Case #49 (old child)Observer: S. AdragnaStore: Supermarket: Shop Rite, NJAisle: Breakfast cerealsShopping party: Mother, 1 girl age 12. No shopping list visible, coupons visible.Daughter holds coupon for STRAWBERRY WHEAT SOUARESobviously a planned pur-chase. Coupon in hand, she searches the shelf, finds the brand, reaches for it on tip-toe offthe top shelf, and continues holding it as they exit the aisle.Case #25 (young child)L- RustSupermarket: Grand U nion, Pleasantville, NYBreakfast cerealsFather, 1 boy age 2Observer:Store:Aisle:Shopping party:No cart visible.Dad holding list, wanders up and down aisle. Boy, too, independently. Boy keeps turningboxes, looking at their backs. Paints at one. Comments puzzle," He passes FREAKIESstops, says, "Nea t!" Tilts box back and forth looking at the hologram. Dad exits aisle vkfithouttaking anythingboy lingers, then skips along after. Hear Dad in next aisle"Adam! Whichdo you want, one of these or these?"Case #144 (old child)Observer: S. JahayerStore: Toy store: Toys "R" Us, Carle Place, NYVideogamesMother. 1 girl age 12Aisle:Shopping party:No cartA mother and a daughter walked into the video software section and the girl was looking forTetris video. She asked her mother, "Are these in alphabetical orde r?" Then the mother said,"Yes, they should be ." Then the girl said, "Here, I got it," and the mother said, "Oh , is it $35?"Then the girl took the video and they walked avt/ay.

    Some attributes pointed to theolder children. For example, 27percent of the observations ofchildren over 5 demonstratedsome sort of mutual discussionor compromise between motherand child versus 10 percentamong young-child cases.Hypotheses which uncoveredno trend at all were droppedfrom further consideration.Refining HypothesesIf a hypothesis survived the

    full-sample evaluation, it wasnot automatically accepted in itsoriginal form. Instead, an at-tempt was made to see if itcould be improvedmodified toaccount better for the dataviadeviant case analysis.An attribute's deviant caseswere the ones it incorrectly ac-counted for: an old-child casewith a young-child attribute, forexample. The operational as-sumption of this analysis wasthat the problem probably lay in

    Journal of ADVERTISING RESEARCH^JULY/AUGUST 1993 67

  • 8/3/2019 Rust Observation Parents and Children Shopping Together

    4/7

    O B S E R V A T I O N S

    the definition: that the attributewas inaccurately or incompletelydefined. By going back to thecase records and methodicallycontrasting the deviant caseswith the conforming ones, sys-tematic differences could oftenbe found, and the definitionsmodified accordingly.Integration

    Taken one at a t ime, all ofthese attributes differentiated thecases by age with some degreeof success. But there were anumber of overlaps, with two ormore attributes pointing to manyof the same cases.A Logistic Regression was runon the data to identify the small-es t set of attributes that woulddifferentiate the cases. Five ofthe original attributes proved toadd nothing to our overall abilityto differentiate the ages of thechildren. The ones which re-mained are described in the fol-lowing section.Final List of Attributes

    Pointing. Young children weremuch more likely to point atproducts or other things in thestore: 27 percent of the 97 youngchildren's cases and 9 percent ofthe 103 old children's cases in-volved this attribute. Pointingw as not dependen t on theirriding in the shopping cart:young children were more likelyto point whether they were inthe cart or walking. Pointinggives younger children a way toindicate desire, even when theylack the symbolic skills orknowledge to communicateverbally.Child in Cart. Not many kidsolder than age 6 rode around inor on the shopping cart. Nearlya third of the children under 4did so (31 percent of under 6s,11 percent of 6 and overs). Thismay not be a profound insight.

    Younger children were muchmore likely to exhibit somesort of physical involvementwith products or displays orpackaging . . ,

    but it has implications (consid-ered later) that may be of impor-tance to marketers.Physical Involvement. Youn-ger children were much morelikely to exhibit some sort ofphysical involvement with prod-ucts or displays or packagingover and above the functionalcontact involved in picking up apackage and carrying it or put-ting it in the cart (27 percentyoung, 11 percent old). Youngkids would explore things tactu-ally, play with them, open them,consume them, or manipulatethem in one way or another.This was often done while sit-ting in the cart but by no meansalways. One little girl, for exam-ple, walked down the cerealaisle, systematically turning ev-ery box around backwards. Itappeared to be pure physical/sensory play.

    Parent Firmly Denies/Ignoresthe Child's Request. Althoughparents often turned down pur-chase requests, whatever the ageof the asker, they were morelikely to be firm and unyieldingwith their younger children.Kids between 6 and 9 sometimesnegotiated successfully, follow-ing an initial turn-down. Withchildren aged 10 and older, par-ent acceptance appeared to bemore automatic. This may havebeen a function either of alteredpower relationships or of more-educated children who knewahead of t ime what mom wouldaccept.Labeling. A certain amount ofthe dialogue between parents

    and younger children in storesinvolved in communicat ing thenames of things (12 percentyoung, 0 percent old). Namelearning is a key developmentatask of preschool children. Theare hungry to learn the nameseverything around them, andparents appear eager to nur turthem this way. Research basedon videotapes of children whilethey watch TV has shown thatpreschoolers consistently payclose attention whenever the Tmaterial involves show-and-tellor labeling (Rust, 1972). Thestore environm ent is full of ob-jects to learn about, and as parents browse along with their lit le children, they spend a certaamount of time identifying thethings around them.

    Teamwork. Parent interactiowith older children often re-flected a degree of t eamwork: adivision of labor with coordination and communicat ion be-tween the members of the shopping party and a set of sharedobjectives (20 percent of old-child cases, 5 percent of youngchild cases).Preplanning. Shopping witholder children more oftenshowed signs of prior planning(12 percent old, 1 percentyoung). This sometimes showeup in the dialogue, when theywould refer directly to prior coversations and intent ions, andsometimes in the fact that thechild would refer to a list, orbring out a coupon that hadbeen saved for use on this trip.Reading. Children aged 5 orunder were never seen readingthings, either on displays orpackaging. Although a numberof 4- and 5-year-olds in the poulation have some ability toread, the ones we saw were nospontaneously motivated to doso in the store environment. Anumber of older children didorient to the text stimuli arounthem (7 percent old, 0 percent

    68 Journa l of ADVERTISING RESEARCHJULY/AUGUST 1993

  • 8/3/2019 Rust Observation Parents and Children Shopping Together

    5/7

    O B S F R V A T 1 O N S

    young).These eight attributes, takenas a set, were correctly associ-ated with the age of the children11 percent of the time.Looking forMarketing Implications

    This sort of qualitative analysisholds out special promise formarketers. The conceptual struc-ture that is embodied in the at-tribute definitions is not arbi-trary. It is firmly grounded inthe consumers' behaviorsopti-mized and tested as a conceptualframework to account for whatthey actually do. As theory, itprovides a structure for predict-ing how consumers will behavein new circumstances in the fu-ture. As grounded theory, itprovides some measure of assur-ance that those predictions willbe correct.A subsequent paper (Rust, inpress) will consider some of thepractical, marketing implicationsof the substantive findings. Anexample is presented here.Little Children Ride in Carts,s o . . . . To reach them, one mightdirect the information on the cart tothe child inside. Advertising infor-mation on the carts need not allbe directed to the mom. Thechild taking a ride is a captive,and sometimes restless, audi-ence, hungry for focus and stim-ulation. Many moms would bepleased to have their childrenconcentrating on in-cart mediarather than having them boredand meddlesome and frustratedat their captivity. There is anopportunity to communicatemarketing information to theshopping party in such a waythat it provides genuine benefitsto both parent and child.

    Display products at cart height.Although bottom shelves may begood for the walking or toddlingchild, children in carts may have

    a harder time seeing, or notic-ing, products there.Incorporate the cart-riding childin advertising copy imagery. If thecopy scenarios echo the shop-ping scenarios, more shoppers(parents and children) are likelyto make the link between ad andstore.Give children packaging or pro-motional material that can keep themoccupied inside the cart.Make packaging and display ma-terials stand out and be noticeablefrom a middle-of-the-aisle dis-tance. Children riding in cartsare confronted with a staggeringarray of stimuli in the averagetoy or grocery store. We know,from studies of child develop-ment, that little children havevery limited abilities to scanbroad arrays of novel stimuliand make much sense out ofthem. Proportions are important,and recognizable elements arepivotal. Faces, particularly thoseof familiar characters, are re-markably effective at getting no-ticed in cluttered environments.

    Research ConsiderationsObservationai ResearchNeeds Discipline. Before goinginto the field, it is important torecognize that a complete de-scription of what goes on is im-possible to achieve. If you wan-der into a behavioral space andwrite down all that hits you, youare not likely to produce scientif-ically useful data. Your percep-tual filters will be inconsistentand shifting. If you are lookingat parents and children, you areapt to notice things that relate tothe custodial/control aspects ofwhat goes on (stemming fromyour role as a nurturant/protec-tive adult within the humanfamily). And if you are comingin as a marketer, you are likelyto put extreme focus on thosebehaviors that you believe toforecast the ultimate brand selec-

    Grounded TheoryDevelopment bypasses theforcing of observations intopreestablished codes and usesobservations instead as astimulus for creating brandnew codes . . .

    tionone consequence of whichwill be to only let you see thingsthat confirm your implicit beliefsabout what determines brandselections. This kind of undisci-plined observation is not likelyto challenge your preconcep-tions, and so it functions lessthan optimally as a scientifictool.Sometimes, Checklists AreHeipful. One way to avoid theproblems of undisciplined obser-vation (though not the one wechoose here) is to set up a com-pletely closed-ended codingscheme beforehand, and reducedata collection to a methodicalchecklisting of the behavior thatunfolds. There are t imes whenthis is the right way to go. Suchtimes are comparable to the con-ditions in which you would do asurvey study based on a closed-ended quest ionnaire. The valid-ity of such research depends onknowing exactly what questionsto ask and how to ask them.But Sometim es You NeedConcepts, NotNumbers. If thetopic is something less narrowlydefined, if the important vari-ables are not clearly conceptual-ized, then a different approachis needed. Grounded TheoryDevelopment bypasses the forc-in g of observations into prees-tablished codes and uses obser-vations instead as a stimulus forcreating bran d new codescodes which, when refinedthrough a sequence of repeatedtesting and modification, are"grounded" in the concrete

    Journa l of ADVERTISING RESEARCH^JULY/AUGUST 1993' 69

  • 8/3/2019 Rust Observation Parents and Children Shopping Together

    6/7

    O B S E R V A T I O N S

    data.Poss ibilities for Future Re-search. Do More with TheseData. It v^'ould be a straightfor-v^fard task to do more groundedanalyses from this (or a similar)data base. One could look atother age breaks: How to targetthe child over 9? What aboutchildren under 4?Compare Dads with M oms.Fathers clearly differ from moth-ers in the ways they interactwith their kids. Are dads reallypushovers? How are they differ-ent? Are there ways marketerscould target dad/child shoppingparties?Study Gatekeeping. A look atgatekeeping dynamics would beinteresting. You could separateall the interactions where par-ents yielded from those wherethey hung tough and do ananalysis of what was differentabout those cases.Hybrid Designs. Observa-tional studies need not rely ex-clusively on passive measures.They could certainly be set upwith an intercept interview . . .either in the store {with the co-operation of the retailer) or inthe street or parking lot.Analyze Open-ends. The an-ecdotal records analyzed in thisproject were short segments oftext, typed in verbatim. They aredirectly comparable to the verba-tim responses to open-endedquestions on questionnaires. Nomodification of current proce-dure would be required to do agrounded analysis on openends. The research could be setup , for example, as a search forthe attributes which differentiatethe open-ends of trier-acceptorsfrom trier-rejectors.Most open-end coding isbased on ad hoc, a priori codes.The codes are selected either totap some preexisting theory ofthe marketer {e.g., mention ofkey benefits is presumed to begood, inconsistency in brand

    labeling is presumed to be bad),or to fit some formal require-ments, like making sure that allphrases can get at least onecode. In my experience, scoresbased on such coding schemesseldom correlate closely to pur-chase-related variables, perhapsbecause they are insufficientlyrelated to the psychological dy-namics of the consumer.Grounded Theory Developmentis a good way to get insightsinto what those dynamicsmight be.Observat ionai Measures andGrounded Analysis Can MakeUnique Contributions to Qualita-tive Research. Field observa-tions have a richness to themthat make them ideal for quahta-tive studies. How people behavespontaneously in the act ofshopping or consuming productsis often different from the de-scriptions they give in inter-views. They are much more re-sponsive to social and environ-mental stimuli than they areconsciously aware. And one wayresearchers can get a handle onall these influences (which are ofparamount importance to mar-keters) is by observing the be-haviors as they occur in thefield.

    Perhaps the most useful dis-tinction between qualitative andquantitative research is not inthe style of data gathering, or inthe degree of structure involvedin questioning consumers, butrather in the purpose of theanalysis. All scientific statementshave both a quantitative and aqualitative component: some-thing happens to some degree.The quantitative component isan expression of the degree, thequalitative component is the def-inition of what it is that hap-pens. Quantitative research putsits emphasis on how oftenthings happen; qualitative re-search puts its emphasis onwhat those things are. Quahta-

    tive researchers, in marketing,are the ones who keep the minof the industry on trackproving the conceptual structure founderstanding the consumer.Grounded theory developmecan use both qualitative andquantitative inputsas we didherebut its focus is on creatithe most powerful conceptual-izations possibleconcepts thaare grounded in the data. In thsense, it is essentially qualitativresearch. It has the potential toenhance the already great contbution that qualitative researchmakes to modern market ing.

    ReferencesFielding, Nigel G., and Ray-mond M. Lee, eds. Using Computers in Qualitative Research.Lond on: Sage Pu blications, 199Glaser, Barney G., and AnselmL. S t rauss . The Discovery ofGrounded Theory: Strategies forQualitative Research. Chicago: Adine Publishing Company, 196Rust, Langbourne W. Attributesof Sesame Street that Influence Prschoolers' Attention to the TelevisScreen. New York: Children'sTelevision Workshop, 1972.Rust , Langbourne W. "UsingTest Scores to Guide the ConteAnalysis of TV Materials." Jounal of Advertising Research 25, 5{1985): 17-23.Strauss, Anslem, and JulietCorb in . Basics of Qualitative Re-search. Newbury Park, CA: SaPublications, 1990.

    70 Journal of ADVERTISING RESEARCHJULY/AUGUST 1993

  • 8/3/2019 Rust Observation Parents and Children Shopping Together

    7/7