An Empirical Investigation of the Spillover Effects of Advertising and Sales Promotions in

Umbrella Branding
Author(s): Tülin Erdem and Baohong Sun
Source: Journal of Marketing Research, Vol. 39, No. 4 (Nov., 2002), pp. 408-420
Published by: American Marketing Association
Stable URL: http://www.jstor.org/stable/1558554
Accessed: 28/01/2010 14:58

Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at
http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless
you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you
may use content in the JSTOR archive only for your personal, non-commercial use.

Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at
http://www.jstor.org/action/showPublisher?publisherCode=ama.

Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed
page of such transmission.

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 Marketing Research.

http://www.jstor.org
 TULINERDEMand BAOHONGSUN*

The authorsinvestigateand findevidencefor advertisingand sales

promotion spillover effects for umbrella brands in frequently purchased
packaged product categories. The authors also capture the impact of
advertising (as well as use experience) on both utilitymean and variance
across two categories. They show that variance of the random compo-
nent of utility declines over time on the basis of advertising (and use
experience) in either category. This constitutes the first empirical evi-
dence for the uncertainty-reducing role of advertising across categories
for umbrella brands.

An Empirical Investigation of the Spillover

Effects of Advertising and Sales
Promotions in Umbrella Branding

Many companies widely practice umbrella branding.
Also, a fair amount of managerial research argues that
umbrellabrandinggeneratessavings in branddevelopment
and marketing costs over time (e.g., Lane and Jacobson
1995; Tauber 1981, 1988) and enhances marketingproduc-
tivity (e.g., Rangaswamy,Burke, and Oliver 1993).
Wernerfelt(1988) has shown analytically that a multi-
productfirm can use its brandname as a bond for quality
when it introduces a new-experience product. Umbrella
brandingis posited to both increaseexpected quality(Wern-
erfelt 1988) and reduce consumer risk (Montgomery and
Wernerfelt1992). Consumers'use experience in one prod-
uct category needs to affect their perceptionsof quality in
anotherfor umbrellabrandingto serve as a credible signal
of a new-experienceproduct'squality,becausea false signal
would be costly if the quality of the extension turnedout to
be poor. Experimentalresearch has shown some evidence
that the parent brand'sperceived quality affects the exten-
sion evaluations (Aaker and Keller 1990) and vice versa,
which indicatesthatbrandequity dilutionmay occur (Loken
and RoedderJohn 1993).
Although the cross-categorylearningeffects for umbrella
brandsare a necessary condition for umbrellabrandingto
functionas a signal (Erdem 1998), the learningeffects alone
*TulinErdem is an associate professor,Universityof California,Berke-
ley (e-mail: erdem@haas.berkeley.edu).Baohong Sun is an assistant pro-
fessor, Kenan-Flagler Business School, University of North Carolina,
Chapel Hill (e-mail: sunb@bschool.unc.edu).The authorsthankthe three
anonymous JMR reviewers for their constructive feedback. The authors
also thank Andrew Ainslie and Michael Keane for their suggestions in
regardto the updating mechanism of the stochastic component of utility
employed in this article. This researchwas supportedby NationalScience
FoundationgrantSBR-9812()67 grantedto Tiilin Erdelm.
o1!f Marketing Research
Joi)rltl
Vol. XXXIX (Novcelbhcr 200(2), 4()8-420
may be insufficientfor umbrellabrandingto generate sav-
ings in marketingcosts and enhance marketingproductivity
over time. To increasesavings in marketingcosts over time,
the marketingmix, such as advertising and sales promo-
tions, must be more effective for umbrellabrandsthan for
other brands. For example, Tauber (1981) suggests that
umbrellabrandingcan create advertisingefficiencies. Leigh
(1984) finds some experimentalevidence for better recall
and recognition performance for umbrella branding than
other branding strategies. Yet there is surprisingly little
empirical work on advertisingand/or any other marketing-
mix synergies for umbrellabrands.
The objectives of this study are twofold. First, we test for
marketing-mixstrategyspillover effects of umbrellabrands
in frequently purchased packaged product categories. We
are especially interestedin testing for own- and cross-effects
of advertising,that is, whether advertising in one product
category affects sales in another product category for
umbrella brands. However, we investigate the cross-
category effects of all marketing-mix effects, including
price, coupon availability, display, and feature. Another
relatedobjective is to shed some light on the magnitudeof
any effects that may exist. We seek to answer the following
questions: (1) What is the impact of the cross-effects on
sales? (2) How large are the cross-effects of the marketing
mix elements in relationto the own-effects? and (3) Which
marketing-mix elements seem to reveal higher cross-
effects?
Second, we explore the dynamics behind the advertising
spillover effects. Similar to use experience, advertising
spillovereffects occur for variousreasons,but they may also
transpirebecause of the uncertaintyreduction.More specif-
ically, we examine whetherthere is a reductionin consumer
uncertainty in one category due to advertising of the
40()
Advertising and Sales Promotions in Umbrella Branding
umbrellabrandin anothercategoryover time. If uncertainty
exists and advertisingreduces consumer uncertaintyabout
products,advertisingis expected to shift both utility mean
and variance.To model and test for both utility mean- and
utility variance-shiftingcross-effects of use experience and
advertising,we develop a novel and parsimoniousapproach.
Our approachenables us to test whether there is any evi-
dence for the uncertainty-reducing role of advertisingacross
categories.
We use ACNielsen scanner panel data on two cate-
gories-toothpaste and toothbrushes-to test for marketing-
mix synergies for umbrellabrands.For the first time in the
literature,on the basis of our analysis of revealedpreference
data, we find evidence for advertisingspillover effects for
umbrellabrands.Furthermore,we find that advertising(as
well as use experience) in one category serves as a mecha-
nism of reducing uncertainty in the other category for
umbrellabrands.In addition, our results indicate that there
are spillover effects of price, coupon availability,and dis-
play. Feature cross-effects exist only in the toothbrush
category.
The rest of the article is organizedas follows: In the first
section, we briefly discuss the relevantliterature.In the sec-
ond section, we outline the proposedmodel and the estima-
tion method. In the third section, we briefly describe the
data. We then outline the results and conclude the article
with implicationsand a discussion of furtherresearch.
LITERATURE
REVIEW
Previous researchin marketinghas studied the effects of
sales promotionsacross categories for complementaryand
substitute products (e.g., Mulhern and Leone 1991) and
multicategory choices (e.g., Bell and Lattin 1998; Man-
chanda,Ansari,and Gupta 1999). A separatestreamof liter-
ature focuses on cross-categorytraitsof consumerbehavior
and finds that consumer price sensitivities (Ainslie and
Rossi 1998; Erdem 1998) or sensitivities to state depend-
ence (Erdem 1998; Seetharaman,Ainslie, and Chintagunta
1999) can be correlatedacross categories.
However, there is surprisingly little published research
with respect to cross-categoryeffects in choice for umbrella
brands. Erdem and Winer (1999) have shown that brand
preferencescan be correlatedacross categoriesfor umbrella
brands.Erdem (1998) has studied how consumers'experi-
ence in one category may affect theirquality perceptionsin
another category for umbrella brands. In a two-category
context, Erdem has shown that consumer mean quality
beliefs, as well as the varianceof these quality beliefs, may
be updated through experience in either category for
umbrella brands.Thus, Erdem studied only cross-category
use experience effects.
Specifically, Erdem(1998) does not study any promotion
or advertising effects, neither own-category nor cross-
category, which is the topic of this article. Therefore,she
does not study and answerany questionsaboutthe existence
and magnitude of cross-advertisingand promotioneffects
for umbrellabrands.Furthermore,her model does not incor-
poratepurchaseincidenceand coincidence in choice, thatis,
the possibility that the stochastic component of the utility
function could be correlatedacross categories. In addition,
Erdem'sapproachonly allows for use experienceeffects due
to learning, whereas the approach we take in this article
409
allows for both learning effects and other sources through
which use experience may affect currentchoices. Further-
more, the approachErdemtakesto incorporatethe impactof
use experienceon utility varianceis difficult to apply to set-
tings in which cross-categoryvariancereductioneffects can
be due to other variablesas well (such as advertising).The
approachtaken in this articleprovidesa more parsimonious
and novel way to incorporatesuch effects. Finally,we allow
for more complex heterogeneity structuresin this article
than Erdemdoes.
Manchanda,Ansari,and Gupta(2000) also find some pre-
liminaryevidence for spillover effects in store promotional
activity for umbrella brands;however, they have not ana-
lyzed advertising and they have not allowed for choice
dynamics.
THEMODEL
UtilitySpecification
Considera set of consumersI = {ili = 1, 2, ..., I that,on
each purchaseoccasion, makes purchasesfrom none, one, or
both of two distinct productcategories m = 1, 2, where I is
the numberof consumers.A nonemptysubset, but not nec-
essarily all, of these consumersmakes purchasesfrom both
of the categories. Suppose that for both productcategories,
the purchasesof the consumersare observedover the period
T = tit = 1, 2, ..., T}, whereT is the time span of the period
and t denotes the week. Let J = (jlj = 1, 2, ... J} be the set
of brand options available, where j = 1 ..., J brands are
available in at least one of the two categories and J is the
numberof brands.Note that we set j = 0 to indicatethe no-
purchaseoption (choice).
Let Knj,m = 1, 2, be an indicatorvariablesuch thatK,-j =
I if brandj is available in category m, and K,n.= 0 other-
wise. Also let Dmijt,m = 1, 2, be an indicatorvariablesuch
that if consumeri purchasesbrandj at time t in categorym,
D,iji = 1, and D,,jt = 0 otherwise.Finally,denote the utility
consumeri derives from purchasingbrandj in categorym at
time (week) t by Uijt, which is postulatedto be
(1) Umijt
= (tmij +
,miPRmiit
+ ymiADmiit + XmiDlSPmijt
+ TmiFEATmijt + OmiCAVmijt+
KmiUEmijt
+ K3-m,j(miPR3- m,i,j,t + tmiAD3-m,ij,t
+ lmiDISP3 m,i,j,t + VmiFEAT3,- ,i,j,t
+ ,nmiCAV3
- m,i,j,t +(On, i UE3 m,i,j,t) + ?Emijt,
where PR,,ijtis the price paid for brandj in category m by
consumer i at t, AD,n,jtis the advertisingstock variablefor
consumer i with respect to brand j in category m at t,
DISP,nijtis the display dummy, FEAT,nitis the in-store ad
dummy, CAV,ijt is the coupon availability variable, and
UE,nijtis the use experience of brandj in category m con-
sumer i has at t. Equation 1 suggests that if brandj is avail-
able in both categories, the utility of purchasingbrandj in
one of the categories will also depend on the price, coupon
availability,advertising,display,feature,and use experience
associated with the same brandname in the othercategory.
INote that in Equation I, cross-effects are specified for umbrellabrands
only. In the empirical analysis, to ensure that the effects exist for only
umbrella,we also tested whethersuch effects exist for otherbrandsas well,
and we did not find evidence for such effects.
410 JOURNAL OF MARKETINGRESEARCH, NOVEMBER 2002
The parametersin Equation I are as follows: a,ij is the
consumer- and brand-specificconstant (brand preference),
and imi is the consumer-specific price coefficient. More-
over, Ymi,Xni, tmi, Omi, and Kmi are the consumer-specific
advertising, display, feature, coupon, and use experience
coefficients in category m (we label these own-effects
response coefficients), whereas, ini, tmi, mi, Vmi, cmi, and
()mi are the price, advertising,display,feature,coupon avail-
ability, and use experience coefficients of the price, adver-
tising, display, feature, and use experience variablesin the
other category, respectively (we label these cross-effects
response coefficients). Similar to Erdemand Keane (1996),
we specify i's utility from not purchasingany brandin cate-
gory m at t as2
(2) Umi0t = (mo + TRENDmt + emot,
where TRENDMis the time trendcoefficient, and for identi-
fication purposeswe set
(3) (mO
= 0, m = 1,2.
The use experience variable, UE,ijt, in Equation I is
defined as the exponentiallysmoothed weighted averageof
past purchases, as in Guadagni and Little's (1983) brand
loyalty variable. Its decay parameteris denoted by DUm.
Similarly,AD,nijtis defined as the exponentially smoothed
weighted average of past aggregate gross rating points
(GRPs).3 Its decay parameter is denoted by DAm. The
AD,,,ijtis updatedon a weekly basis. Note thaty,i (andr1mi),
the coefficient on the advertising stock variable, embeds
both a consumer'stelevision commercialviewing habitsand
responsivenessto the advertisements,conditionalon having
seen them. Because advertisingeffectiveness in the context
of television advertisementsdependson television and com-
mercial viewing habitsof consumers,we can justify thaty,,
(and tl,) representsadvertisingresponsivenessat the indi-
vidual level.4
We define the DISP variablesuch that DISP,nij,= I if a
display is availablefor brandj in categorym in the store that
consumer i visits at t, whereas DISP,,ijt= 0 otherwise.Sim-
ilarly,we define the FEATvariablesuch thatFEAT,ijt= 1 if
a feature is available for brandj in category m in the store
that consumer i visits at t, whereasFEAT,nijt = 0 otherwise.
Finally, CAV,,ijt is the average value of the store's and man-
2Similarto Erdem and Keane (1996), our purposeis to control for pur-
chase incidence ratherthanexplicitly model purchaseincidence decisions.
3An alternativeto GRP data is television exposure data;however,televi-
sion exposure data are not available because ACNielsen discontinuedthe
collection of such data in the early 1990s. Televisionexposuredatahave the
advantageof havingthe estimatedadvertisingcoefficient reflectadvertising
responsivenessonly because the advertisingvariableitself will capturethe
television/commercial viewing habits. However, the television exposure
data are noisy as well, because we only knew that the television was tuned
to that channel duringthe airingof a particularcommercial.One advantage
of the GRP data, in contrast,is that GRP is underthe direct control of the
firm whereas television exposure is not. Therefore, firms are more inter-
ested in models that use GRP data (Abe 1997). Finally, Tellis and Weiss
(1995) find that advertising effectiveness results are not sensitive to the
types of advertisingmeasures.
4Wealso have GRP data for different"dayparts"(e.g., primetime versus
late morning).When we used these dataandestimatedEquationI with sep-
arate coefficients for each daypart,we obtained largeradvertisingcoeffi-
cients for certain daypartsthan others, as we would expect, but the main
results are the salmewhetherthe daypall data are used or not.
ufacturer'scoupons availablefor brandj in categorym in the
store thatconsumeri visits at t. It includes the zero value for
nonavailable coupons (Keane 1997b). We introduce this
variableto avoid the endogeneityproblemcaused by includ-
ing coupons redeemedin the utility specification.Last, Emijt
in Equation 1 denotes the time-varyingstochastic compo-
nent of utility that is knownby the consumerbutnot observ-
able by the analyst.
We should note that the cross-effects specified in Equa-
tion 1 may exist for several reasons. First, price cuts,
coupons, advertising,displays, and featuresin one category
may remind the consumer about the umbrella brand and
thereforeincreasethe likelihoodof theirchoosing thatbrand
in anothercategory, conditional on productcategory deci-
sion. In the case of promotionaltools such as displays, the
promotionof the umbrellabrandin one category may help
promote the umbrella brand in consumers' consideration
sets in the second category, in which the brandis not pro-
moted, as well. Advertisingcan be expectedto increasegen-
eral brandawareness of individualproductsthat share the
same brand name. Cross-category use experience effects
may be due to habit formation.
Second, the effects of use experience and advertisingon
consumeruncertaintymay cause use experienceand adver-
tising spillover effects. If use experience affects quality
beliefs, both the mean and variance of these beliefs may
change on the basis of use experience. Therefore, if use
experienceaffects consumeruncertainty,we shouldexpect it
to affect both the utility mean and variance in a reduced-
form model. The same rationale holds for advertising.
Erdemand Keane (1996) have shown that in the context of
one category, both use experience and advertising affect
both mean and varianceof qualityperceptions.Therefore,if
advertisingspillover effects are partly due to the effect of
advertisingon consumer uncertainty,we should expect util-
ity variance(that is, the varianceof the randomcomponent
of utility) to decline over time on the basis o' advertising.
We capture these utility variance effects by adopting a
novel approach.Specifically,we allow for not only the "ran-
dom" components of utility to be correlatedacross cate-
gories but also the variancesof the randomcomponentsfor
umbrellabrandsto be updatedon the basis of past purchases
and advertising associated with umbrella brands in either
category.We cover this issue in more detail subsequently.It
may suffice to state that we attemptto capturethe effect of
use experience and advertisingas a shifter of both utility
mean and variance; the latter effect should occur mainly
because the consumer uncertaintydecreases over time on
the basis of past purchasesand advertising.
Heterogeneity Specification
It is well established that not accounting for unobserved
heterogeneitycan bias parameterestimates(Heckman 1981;
Rossi and Allenby 1993). In this article, we account for
unobservedheterogeneityas well. To do so, we assume that
brand preferences for each category A,mi= (a,n,i, aii2, ..
a1ij)T; own-effects response coefficients B,,i =(P:- i, ni,
mti , lmi, K,lni)T;and cross-effects response coefficients
mi'
Cmi = (i, t li, lli, Vm,I, i O,,1i)T, m = i, 2, where the
superscriptT denotes the transpose,are normallydistributed
with the following mean vectorand covariancematrix:
Advertising and Sales Promotions in Umbrella Branding 411

-A7 ii -Al- to be more advertisingsensitive than the averageconsumer

Bli B, in the toothpaste category. The corresponding cross-
categorycorrelationscan be calculatedfrom the covariance
C,i Ci matrixn. We denotethese cross-categorycorrelationsby A.6
(4) - Nc
A2i A2 Choice Probabilities
B2i B2 Let us rewriteEquation1 for each categorym, wherem =
1, 2, as
-C2i -C2.
-' (5) mmijt Vmijt
+
?mijt'
-E;A nAA2
"A1B, "AlC, A,C2
YBB
]EBI nBC nBiA, nBB2
I"AB2 nB,C2
whereVmijtis the "deterministic"partof the utility.
B11C,
?C| nciA2 "CIB nCC2
Let 0 denote the vector of parameters(Al, A2, BI, B2, Cl,
!AIB2
BBIA2 !CiA2
?A2 i"A2B2
r[
A2C2 C2, d(ZAI), d(EA2), d(ZBI), d(IB2), d(ECI), d(IC2),
HAB nClB EB2 w(n))T. The vectors d(EAI), d(AA2), d(ECI), and d(IC2)
A,B2 A2C2
IIAIC,
!BiB,
nBC containthe diagonalelements of the correspondingmatrices
nAC2 -CC2 nA2C2 A 2C2 ZC2
(i.e., the variances of the heterogeneity distributions of
The vectors Al, B , Ci are the means in Category I and brandpreferencesand cross-effects response coefficients).
the vectors A2, B2, C2 are the means in Category 2. The The vectors d(EBi) and d(EB2) contain the diagonal ele-
block diagonal matrices EAI, E?B,ICI, IA2, ?B2, and ?C2 ments(variancesof own-effects responsecoefficients)of the
are the covariancematricesof A l, Bij, Cli, A2i,B2i,and C2i, correspondingmatrices, as well as the off-diagonal ele-
ments, or covariances, that we allowed to be nonzero.
respectively.The diagonal elements ofA AI, B 1,ECi, IA2,
Finally,the vector w(n) contains the covariancesamong the
ZB2, and ZC2themselves containthe variancesof the corre-
sponding heterogeneity distributions. Finally, F! are the brand-specificconstantsand advertisingand price sensitivi-
ties, as well as cross-categorycovariancesto be estimated.
appropriatecovariancematrices. Note that though we used the covariancematricesin defin-
Within-CategoryCorrelations ing the parametervectorfor notationalconvenience,we esti-
Given the large numberof covariancesto be estimatedas mate the correspondingstandarddeviationsand correlations
shown in Equation4, we allow only a subsetof these covari- introduced previously, instead of the variances and
ances to be nonzero for parsimony.5We allow heterogeneity covariances.
in all brand-specific constants and allow own and cross Recall thatall coefficients in EquationI are allowed to be
marketing-mix response and use experience coefficients; heterogeneousacross consumers,and their distributionsare
however,we estimate and reportin the "Results"section of given in Equation4. Let (pidenote the multivariatestandard
the article only the following covariances (correlations) normalvectorthatgeneratesthese coefficients.To be able to
write down the choice probabilities,we need to specify a
among the heterogeneity distributions: (I) correlations
distribution for Emijt
among the brand specific constants and price (as well as
We assume that ?,mijtj e J is given by
advertising)sensitivities, (2) pairwise correlationsbetween
own price (and advertising) sensitivities and all other (6) = Ymiit
?miit -Et[Ymiit],
marketing-mixsensitivities (e.g., display), and (3) pairwise
correlationsbetween own price (and advertising)sensitivi- where y,ijt is the generator of ?,ijt, whose prior distribution
ties and use experience. is
Cross-CategoryCorrelations
(7) ,
Ainslie and Rossi (1998) suggest that consumersensitiv- Y,io
Y2ij0_ N([o]P p I1
ities to marketing variables (own-effects of the marketing
mix in our model) may be correlated across categories. Note that Et[.] indicates expectation conditional on the
Therefore, a person who is more price sensitive than the informationat time t. Equations 6 and 7 indicate that the
average consumer in one category can be also more price means of the randomcomponentsare always zero, whereas
sensitive than the average consumer in another category. the assumption is that the initial variancesare one and the
Therefore,we also estimate nBI B2 (the covariancematrixof initial covariancebetween the randomcomponentsis p.
Bl and B2). This means that we permit consumer price, A parsimonious and new way of capturing the utility
advertising, display, feature, coupon, and use experience variance-shiftingrole of use experienceandadvertisingis to
sensitivities to covary across the two categories. For exam- allow the initial utility variance,which we set to unity,to be
ple, a consumer who is more advertisingsensitive than the updatedconditionalon past purchases.Neitheradvertising's
averageconsumer in the toothbrushcategory may also tend
6Finally,note that we adopt the recent classical inferencetechniquesto
sHowever, in the empirical analysis, we tested the sensitivity of our model and estimate unobserved heterogeneity. An alternative approach
results to the assumptions made about covariance structureby allowing would have been Bayesiantechniques(e.g., Rossi, McCulloch,andAllenby
more covariances(correlations)to be nonzero. For example, we estimated 1996). However, the main advantageof such techniques is the ability to
the covariancesbetween brand-specificconstants,a few covariancesamong providehousehold-levelparameterestimates, which is particularlyuseful if
the cross-responsecoefficients, and so forth,but these did not improvethe the research objective includes micromarketingimplications, which is
model fit significantly and did not change the parameterestimiates. beyond the scope of this article.
412 JOURNAL OF MARKETINGRESEARCH, NOVEMBER 2002

impact on the consumer utility mean nor its impact on the

m,i,j,t - 1 12,i,j,t - I im,i,j,t - I
consumer utility variance across categories for umbrella "bimijt Dm,i,j,t-
brandshas been empiricallyverified in previouswork. D3-m,i,j,t - (3-m,i,j,t -1 b2mijt 3-m,i,j,t - 112,i,j,t- I
To capturethe effects of both use experience and adver- 012,i,j,t - I b3mijt am,,
m,i,j,t-
2 2
12 b
tising on the varianceof the randomcomponent,thatis, util- 03-m,i,j,t- 2+ i,3-m _ 4mijt _ 12,i,j,t-I
ity variance,let the informationobtainedby consumeri by and with Equation 12, the updatingformulas for the vari-
purchasingbrandj before time t fromcategorym be denoted ances are obtainedfrom
by xmijt.This informationcan be about attributes,including
quality,or match informationbetween consumertastes and (14) ,mi,2 = E{[Ymijt
- Et(
t)]2}
m = 1,2
productofferings and the like, which are not observedby the
analyst. Let x,ijt be given by 12i.jt=
E {[Yijt- Et (ylijt)][Y2ijt - Et(Y2ijt)]}

(8) = +
Xmijt Ymijt 6mijt' Note that the precedingequationssuggest that the higher
the precisionof informationfromuse experience(l/(' n)and
where 6i.jt is an i.i.d. random error term that reflects the
level of noise in the informationobtained.We suppose that advertising (I/'2 n) are (or the lower the use experience
variability, 'n,, and advertising variability,oG2, are), the
- more consumerswill update.
(9) ,mi.t N(O, o62).
Given Equations 12 to 14, the distributionof ?1ijt, m = I,
Note that additionalinformationcomes from weekly adver- 2 at time t is given by
tisements. Denoting this informationwith z,ijt , we let
-. 2 0
j= 1,2,...,J.
(10) = + 1 -c2ij't-
L[2ijt O _101
(T 12ijt 11t2
02i 12ijt
mijt Ymijt mijt' ijt

where rijtm is an i.i.d. randomerror term that reflects the We then can write the choice probabilitiesconditional on 0
level of noise in the informationfrom the advertisement.It and (pias
is given by
(16) = 1,Dt =
1;p)= ) tjk(0,(Pi),
Probjt(D!ijt
(11) rmiit N(O, m 2)
where q)tjk(-)is the appropriatecumulativedistributionfunc-
Given these, we then assume that the randomcomponent tion whose probabilitydistributionfunction is determined
of utility is updated according to Bayesian updatingrules from Equation 15. Note that this is a 2J + 2 dimensional
and invoke the Kalmanfilter for updatingthe distributionof integral.Therefore,the probabilityof consumer(household)
i makingthe sequenceof purchasesgiven by D,,,jt, m = 1, 2,
Y,nijtand therefore of ?,j1.t There is no updatingfor j = 0,
which correspondsto no purchase.To derivethe Kalmanfil- j = 0, , ..., J, t = 1, ..., Ti, conditionalon 0 and (pi,is
ter, we regress T J J
Ymiit- Et- I[Ymiit] (17) Prob (0, i) = n DlijtD2Ijtt ( pi).
on t=l j =O k=0
m = 1, 2
Xmijt- Et_ I[xmijt]and Zmijt- Et_ i[Zmijt],
Integratingover(Qi,we obtain
by suppressing the intercept,and we obtain the following
updatingformulasfor the expectations: (18) Probi(0) = I Prob(0,(p )f/(pi 0)d(pi,
= - n
(12) Et[Ymiit] Et _[Ymijt] + blmijt{xmijt Et,- [mijt]}
+ E - I[Y3 - where Q2is the domain of the integration,whereasf((p10) is
b2mijt {x3- m,i,j,t m,i,j,t 1)
the multinomialprobabilitydistributionfunctionfor pi,con-
+b3mijt{Zm,ij,t
- Et
[nmi,j,t ]) ditional on 0.
Thus, we estimate a multivariate multinomial probit
+ b4mijt{z3 -m,i,j,t m,i,j, t]'
-
Et[Z3 model, which has been adopted in some recent articles in
marketing as well (e.g., Manchanda,Ansari, and Gupta
where bk,,lit, k = 1, 2 and m = 1, 2 are solved from
1999). This model allows randomcomponents to be corre-
2 2
lated across the m categories, which Manchanda,Ansari,
-
Dm,i,jt-j , t- I + 6m and Gupta label as "co-incidence."Thus, factors that are
(13)
D3- m,i,j,t - 1o12,i,j,t -I unobservedby the analystbut knownby the consumercould
DDm i jt - I(Y 2 be correlatedacrosscategories.However,for the first time in
m,i,j,t - I
-
Dm,i,j,t 11(12,i,j,t - I the economics and marketingliterature,we also incorporate
into this model a process by which the varianceof the ran-
2
D CY- 11,i,,-- II..
IniJDi.,t- I 12I,i,j,t m,i,j,t - I ni, j,t - I dom component (as well as the covariance of this compo-
D - Im,,j,
2 ++ 03
G 2 D 0 2, i,j.t - nent across the m categories) is updatedover time, on the
- -
i,,t ,3 - n D3-m,i,t - I
2 basis of past purchasesof the consumersand past advertise-
D3 1- m i.j.t - lI312,i,,jt -D0II 12Y j,i j,- 2 I2 ++ I m ment. The extent of the updatingis determinedby l/o'1n in
D3 -m.i.,l t - I<3-12,i, j.t -I the case of use experienceand by 1/o', in the case of adver-
Advertising and Sales Promotions in Umbrella Branding 413

tising; that is, the more precise the informationsource is, the
more updatingthere will be.
Given Equation18, the log-likelihoodfunctionto be max-
imized is
I ParameterEstimatesand Goodnessof Fit
(19) LogL(O)= Xln[Probi(0)],
i=l
Note that the calculationof Probi(O)in Equation19 requires
the evaluation of multiple integrals,which in turn requires
Monte Carlo simulation techniques. We use a probability
simulatoras the Monte Carlo method (Geweke 1991; Haji-
vassiliou, McFadden,and Ruud 1994; Keane 1990, 1994).
As our estimation method, we adopt the method of simu-
lated moments (MSM) first developed by McFadden(1989)
and extended by Keane (1990).
EMPIRICALANALYSIS
Data
The models are estimatedon scannerpanel data provided
by ACNielsen for toothpasteand toothbrushes.Given new
productfeatureintroductionsand the existence of long-term
experience attributes(e.g., cavity-fightingpower of a tooth-
paste), quality uncertaintymay exist even in the relatively
matureproductcategoriestypically studiedin scannerpanel
research (Erdem and Keane 1996). Also note that because
most frequentlypurchasedproductcategories are not high-
involvementgoods, consumerstypicallyrely more on brand
names and advertisingthan on active search to resolve any
quality uncertainty.
The panels cover 157 weeks from the end of 1991 to the
end of 1994. The data sets includehouseholdsfrom two test
markets in Chicago and Atlanta.We use the Chicago panel
for model calibration.We use the Atlanta panel to assess
out-of-sample fit (i.e., the predictivevalidityof the models).
The analysis includes four brandsin each category.In both
the calibration(Chicago) and the prediction(Atlanta)panel,
we randomlydraw the panel members from the sample of
households that do not buy multiple brandsin a given cate-
gory on the same purchaseoccasion.
The first two brands(Brand 1 and Brand2) are available
in both categories. Table I providessummarystatistics.The
weekly GRPs are plotted in Figure 1. The price used in the
study is the price paid without any coupons. Using a price
value net of coupons introducesa seriousendogeneityprob-
lem; therefore, we model coupon effects by using the
coupon availabilitymeasurewe have defined previously.
We estimatedthe model describedpreviouslyand various
nested models. Table 2 reports the results for five nested
models and the full model. The sixth nested model has the
multivariateprobit specification with correlatederrorsand
allows for updatingerrorson the basis of past purchasesbut
does not allow the errorsto be updatedon the basis of past
advertising.The fourth nested model (NM4) has the multi-
variate multinomial probit specification with correlated
errorsacross the categories but does not allow for updating
errorson the basis of purchasesand advertising.The third
nested model (NM3) allows for neitherupdatingnor corre-
lated errors across the two categories. The second nested
model (NM2) is similar to NM3 but does not allow for use
experience, price, coupon availability,advertising,display,
and featurein one category to affect the utility of the other
category.Finally,the first nested model (NM ) is similarto
NM2 except that heterogeneitydistributionsare assumedto
be independentacross the two categories. Both in-sample
and out-of-sample, the full model fits the data statistically
betterthan all other nested models. Therefore,we focus on
the parameterestimates obtained by estimating the full
model.
In a given category,the results indicatethatcoupon avail-
ability,advertising,displays, features,and use experienceall
have significant, positive effects on consumer utility,
whereasprice has the expected negativeeffect. Thereis sig-
nificant heterogeneityin consumer tastes and responses to
use experienceand marketing-mixvariables.The time trend
in the no-purchasespecification is statisticallyinsignificant.
Given the objectives of this article, the importantparam-
eter estimates are associated with the cross-effects of
marketing-mixvariables(and use experience),as well as the
cross-categoryparametersthat govern the process of updat-
ing the utility error term variance over time through use
experience and advertising.These parameterestimates are
markedin bold in Table 3.
We find thatuse experience in one categoryhas a positive
impact on brandchoice probabilitiesin the other category

Table 1
SUMMARYSTATISTICS

Average
Brand Name MarketShare AveragePrice" GRP Frequencyof Display Frequencyof Feature AverageCouponti
Toothpaste
Brand 1 31.3% $1.81 13.54 2.0% 2.9% $.15
Brand2 20.0% $1.87 27.07 1.6% 2.8% $18
Brand3 10.6% $1.81 0 1.4% 3.2% $.23
Brand4 9.5% (71.4%) $2.67 0 1.2% 1.8% $.23
Toothbrush
Brand 1 10.2% $2.36 12.62 1.2% 2.6% $23
Brand2 21.8% $1.99 19.75 1.1% 3.1% $19
Brand5 19.4% $2.36 22.84 .6% 3.2% $.27
Brand6 17.3%(68.7%) $1.96 0 .7% 3.3% $19
t'Weightedaverage price per 50)ounces for toothpaste.Weightedaverageprice per unit for toothhrush.
'Average nonzero coupon value of toothpaslenormalizedat 50()ounces. Averagenolnero coupon value of toothbrushper unil.
 :

414 JOURNAL OF MARKETINGRESEARCH, NOVEMBER 2002

Figure 1
WEEKLYGRPs

Plot of Weekly GRP for Toothpaste

100
Xl 'I
,

Week

Brand 1 ------ Brand 2 [

Plot of Weekly GRP for Toothbrush

100

a-
c: 50
o . * .,
^'.a,.r, A 3,
tTIIT
,... / ' -
o 1 ,iI' . . . ,,, , i,,f T i (, ? 1nr 1 r ir 1 11 11I i T 1 1l
1 9 17 25 33 41 49 57 65 73 81 89 97 105 113 121 129 137 145 153
Week

Brand 1 ------ Brand2 Brand5

Table 2
MODELSELECTION

NMI NM2 NM3 NM4 NM5 FM

In-Sample(Chicago)
-LL 21323.4 20731.6 20212.5 20101.2 19510.4 18925.3
AIC 21417.4 20831.6 20336.5 20226.2 19637.4 19054.3
Out-of-Sample (Atlanta)h^
BIC 21839.0 21280.1 20892.3 20786.9 20208.0 19633.9
-LL 13260.5 12936.2 12712.2
122179.6 12367.5 2592.7
"Numberof observations= 59,032 (includingno purchases),numberof households(hhs) = 376; 167 hhs made621 purchasesof toothbrush;345 hhs made
2880 purchasesof toothpaste; 136 hhs purchasedboth toothbrushand toothpaste.
hNumberof observations= 30,301 (includingno purchases),numberof hhs = 193; 92 hhs made 198 purchasesof toothbrush;186 hhs made912 purchases
of toothpaste;85 hhs purchasedboth toothbrushand toothpaste.
Notes: LL = log likelihood;AIC = Akaike informationcriterion;BIC = Bayesian infornation criterion.

for umbrella brands. We also find statistically significant
cross-categoryeffects of advertisingin both categories.Fur-
thermore,we find evidence thatcoupon availabilityand dis-
play (significantat the p <. 10 level) in the toothbrushcate-
gory affects consumer utility in toothpaste positively for
umbrellabrands.Coupon availability,display,and featurein
the toothpastecategory affect consumerutility in the tooth-
brushcategory positively as well. Finally,prices have nega-
tive cross-category effects in both categories. Therefore, a
price cut in one category for an umbrellabrand increases
consumer utilities and thus the choice probabilitiesfor that
brandin the other category.
Turningour attention to the random component of the
consumerutility and the process of utility varianceupdating
based on use experience and advertising,we find the fol-
lowing results: Random components are correlatedacross
the two categories.Recall that we estimatedan initialcorre-
lation betweenthe randomcomponents,which we find to be
positive and significant. However, our results indicate that
both osa and oln are statisticallysignificant in both cate-
gories. Therefore,dependingon the degree of the precision
of information (i.e., 1/o,Smor I/oG,n), the variance and
covarianceo' randomcomponentsacross the two categories
are updatedwith use experienceand advertising.This result
Advertisingand Sales Promotionsin UmbrellaBranding 415

Table3
ESTIMATION
MODEL

Parameters Toothpaste(m = I) Toothbrush(m = 2)

Brandspecific constantsan,j Brand I -2.07(.19) -3.77(.32)

Brand2 -2.54(.28) -.26(.10)
Brand3 (5)a -3.29(.22) -.86(.21)
Brand4 (6) -3.31(.17) -1.03(.26)
No purchaseh 0 0
Standarddeviationof brandspecific constant nctmj Brand I 1.51(.34) 1.02(.18)
Brand2 1.35(.20) .47(.19)
Brand3 (5) 1.00(.22) .23(. 11)
Brand4 (6) .58(. Il) .11(.07)
No purchase 0 0
Trend,, .015(2.53) .002(1.10)
.010(2.57) .001(1.73)
Mean price coefficient P1, -1.21(.18) -.95(.13)
Standarddeviationof the price randomeffect 7p,. .56(.16) .45(.12)
Mean coupon coefficient on .82(.12) .95(.17)
Standarddeviationof the coupon randomeffect (o,m .55(.16) .46(.20)
Mean ad coefficient y,, .12(.05) .15(.05)
Standarddeviationof the ad randomeffect (9ml .05(.02) .07(.04)
Mean display coefficient kn 1.49(.26) 1.36(.33)
Standarddeviationof the display randomeffect o(,, 1.20(.25) .75(.23)
Mean featurecoefficient x,, 1.09(.40) 1.29(.20)
Standarddeviationof the featurerandomeffect on .82(.43) .43(.59)
Mean use experience coefficient Km 2.74(.33) 2.10(.23)
Standarddeviationof the use experience randomeffect (oc 1.86(.30) 1.10(.23)
Mean cross-price coefficient 3 - m -.11(.04) -.18(.08)
Standard deviation of the cross-price random effect ag_ m .08(.13) .08(.08)
Mean cross-coupon coefficient 3 - m .45(.18) .47(.17)
Standard deviation of the cross-coupon random effect 3 _ m .23(.09) .20(.08)
Mean cross-ad coefficient 13 m .011(.005) .013(.005)
Standard deviation of the cross-ad random effect a3 - m .008(.005) .011(.004)
Mean cross-display coefficient 13 _ m .097(.055) .147(.077)
Standard deviation of the cross-display random effect 3-_m .091(2.58) .100(1.15)
Mean cross-feature coefficient V3_ m .132(.083) .170(.037)
Standard deviation of the cross-feature random effect av3 - n .085(.044) .095(.144)
Mean cross-use experience coefficient (03_ m .82(.25) .69(.14)
Standard deviation of the cross-use experience random effect o _-m .45(.45) .20(.09)
Use experience smoothing parameterDUM .78(.12) .88(.25)
Media smoothing parameterDA, .28(.10) .56(.17)
Correlationbetween constanttelrmsand price pa,i,p,, Brand I .15(.09) .13(.08)
Brand2 .19(.19) .17(.06)
Brand3 (5) .07(.11) -.08(.09)
Brand4 (6) -.29(.05) -.25(.1 )
Correlationbetween constantterms and advertisingPamrn, Brand I .12(.04) .14(.05)
Brand2 .09(.05) .15(.07)
Brand3 (5) .06(.03) .08(.06)
Brand4 (6) .07(.05) .07(.05)
Correlationbetween price and ad coefficients pi,,,ny -.17(.08) .16(.06)
Correlationbetween price and display coefficients pp,,lmX -.25(.10) -.23(. 11)
Correlationbetween price and featurecoefficients Ppi,Tm -.17(.07) -.22(.10)
Correlationbetween price and coupon coefficient Ppnmom -.18(.07) -.13(.07)
Correlationbetween price and use experience coefficients P,c;m -.21(.09) -.20(.1 1)
Correlationbetween ad and display coefficients Pyv.m .11(.06) -.18(.()8)
Correlationbetween ad and featurecoefficients Ppini .08(.04) .13(.()05)
Correlationbetween ad and coupon coefficients P,nl,9 .16(.07) .10(.09)
Correlationbetween ad and use experience coefficients Py,,n, .27(.12) .23(.13)
Cross-categorycorrelationbetween price coefficients AIP2 .63(.13)
Cross-categorycorrelationbetween ad coefficients AyI.2 .44(.10)
Cross-categorycorrelationbetween display coefficients AX.i2 .16(.10)
Cross-categorycorrelationbetween featurecoefficients At,t2 .13(.08)
Cross-categorycorrelationbetween coupon coefficients A0102 .14(.10)
Cross-categorycorrelationbetween use experience coefficients A,i,e2 .42(.18)
Cross-category correlation between error terms p (in the initial period) .19(.08)
Advertising variability o%for toothbrush .22(.1 1)
Advertising variability a6 for toothpaste .74(.31)
Experience variability CYn for toothbrush .16(.06)
Experience variability Oc for toothpaste .27(. 1 )
aToothpastebrandsare I, 2, 3, and 4, and toothbrushbrandsare 1, 2, 5, and 6.
hNo-purchaseoption constantis set to zero for identificationpurposes.
Notes: Figuresin boldface indicateparameterestimates associatedwith the cross-effectsof marketing-mixvariablesand the cross-categoryparamleters
that
govern the process of updatingthe utility errorterm varianceover time.
416 JOURNAL OF MARKETINGRESEARCH, NOVEMBER 2002
confirms the finding in the literaturethat use experience
may reduce uncertainty across categories for umbrella
brands (Erdem 1998) and establishes the new result that
advertisingdecreases consumer uncertaintyacross product
categories for umbrellabrandsas well.
Note also thatthe precisionof use experienceinformation
is found to be higher than that of advertisingin both cate-
gories, which is an intuitive result because use experience
should provide less noisy informationthan advertisingdoes
in most cases. Also, the ratioof advertisingvariabilityto use
experience variability is higher for toothpaste than tooth-
brushes; that is, although advertisingprovides more noisy
informationthandoes use experiencein both categories,the
relative noisiness of advertising informationis greater for
toothpastethan toothbrushes.
Finally, we estimated several within-categoryas well as
across-category correlations, as discussed in the previous
section. Not all the correlationsbetweenbrandconstantsand
price (advertising) responses are statistically significant.
The statistically significant estimates suggest that in the
toothpaste(toothbrush)category,consumerswho like Brand
I (Brand2) more than the averageconsumertend to be less
price sensitive than the average consumer.The reverse is
true for Brand 4. Therefore,the correlationbetween brand
specific constants and price sensitivity is positive for some
brandsand negative for others. The estimates also suggest
thatconsumerswho like Brands 1, 2, and 3 in toothpasteand
Brands 1 and 2 in toothbrushesmore than the averagecon-
sumertend to be more advertisingsensitivethanthe average
consumer. Finally, in both categories, there are statistically
significantcorrelationsof price sensitivity with advertising,
display, feature, and use experience sensitivities. Advertis-
ing and featuresensitivities and advertisingand display sen-
sitivities are significantly correlated in both categories as
well. Advertising and coupon availability sensitivities, as
well as advertisingand use experiencesensitivities, are sig-
nificantlycorrelatedonly in the toothpastecategory.
Given the signs of the parameterestimates, these com-
bined results indicate that in the two categories we study,
there seems to be an individual-specific trait we call
"marketing-mixsensitivity" (i.e., some consumers are in
general more sensitive to marketing-mixstrategies).Indeed,
similar results have been found in otherrecent work as well
(Arora2000). The only result that seems somewhatsurpris-
ing is the negative sign of the correlationbetween use expe-
rience and price sensitivities. Although these correlation
terms do not necessarily imply any causal links, we might
still have expected the reverseresultto hold. This is because
high use experience involves positive purchase feedback,
and consumers who are sensitive to this may be expected to
be less price sensitive (Seetharaman,Ainslie, and Chinta-
gunta 1999). However, note that the estimate we obtain is
not necessarily counterintuitive.This is because more use
experience-sensitive people may have utility fartherout in
the tail of the distributionand may need a largerprice coef-
ficient to have the same price elasticity.This would be con-
sistent with high sensitivity to use experience being associ-
ated with high price sensitivity.Anotherexplanationfor the
result is that consumers with higher sensitivity to use expe-
rience may have smaller considerationsets (Rajendranand
Tellis 1994); this may increasethe abilityof use experience-
sensitive consumers to make price comparisons,which may
result in higher price sensitivities in these two categories,
ceteris paribus.Finally, we also find that consumer price,
advertising,and use experience sensitivities are correlated
across the two categories we study.
Policy Simulations
To evaluatethe impactof temporarypolicy changes (i.e.,
a change in one of the marketing-mixelements in the second
week) in each category, we ran the following simulations
separatelyfor Brand I and Brand2: a 20% increase in (1)
coupon availability,(2) advertising,(3) displays, and(4) fea-
tures and (5) a 20% decrease in price. We reportonly the
impact of the policy change on the cumulative purchases.
The impactof the temporarychangeon cumulativesales can
be conceptualizedas the long-termeffect of these temporary
policy changes. In regardto the short-termeffects of a tem-
porarypolicy change (i.e., the impactof a policy change in
the beginningof the second week on sales in that week), it
may suffice to indicatethat the short-termeffects are larger.
It is also worthwhile to note that we found the long-term
effects to be largerthanthe short-termeffects in advertising.
Table 4 reports the results for the respective temporary
policy changes adoptedby Brand I in toothpasteand tooth-
brushesseparately.Table5 reportsthe correspondingresults
for Brand2. In bothTables4 and 5, the first half of the table
compares the baseline figures with the cumulative "sales"
figures when a temporarypolicy change was adopted in
Week 2 (orjust beforeWeek 2 in the case of advertisingand
coupon availability).In particular,the first row indicatesthe
baseline figures.The next five rows indicate the impacts of
a change in pricing, couponing, advertising, display, and
featurepolicy, respectively,in toothbrushon toothpasteand
toothbrushsales. The following five rows revealthe impacts
of a change in pricing, couponing, advertising,display, and
featurepolicy, respectively,in toothpasteon toothpasteand
toothbrushsales. The lower half of the table reportsthe per-
centage change in sales due to the respectivepolicy changes,
again separatelyfor toothbrushesand toothpaste.The tables
do not include the featureeffects in toothpaste,because the
cross-effects of features were found to be statistically
insignificantfor toothpaste.
To assess the impact of a 20% increase in Brand I's
advertising frequency in toothbrushes on the toothpaste
sales of Brand I, for example, comparethe baseline figure
of 1326, which reflects the baseline cumulative sales of
Brand I, with the correspondingpostintervention(a tempo-
rary increasein advertisingfrequencyby Brand I in tooth-
brushes at PurchaseOccasion 2) figure of 1431. This com-
parison reveals an 8% increase in toothpastesales due to a
20% increase of advertising frequency of Brand I in the
toothbrushcategory.
The resultson spillovereffects (cross-effects)can be sum-
marized as follows: Overall, coupons and then advertising
have the largest cross-effects on sales in both categories
except for Brand2 in the toothpastecategory,for which fea-
tures (more so than any other marketing-mixelement in the
toothpastecategory)affect sales the most in the toothbrush
category, followed by coupons and advertising. In most
cases, the smallest cross-effects are the price cross-effects.
The cross-effectsof coupon availabilityon sales rangefrom
5% to 13%acrossproductcategoriesand across Brand I and
Brand2. The ~
?
~ ~
advertisingcross-effects
~ ~ rangefi-om40%to 8%.
I
Advertising and Sales Promotions in Umbrella Branding 417

Table 4
RESULTSFOR POLICYCHANGESBY BRAND1
SIMULATION

Sales of Toothpaste Sales of Toothbrush

Brand I Brand2 Brand 3 Brand4 Brand I Brand 2 Brand5 Brand6
Baseline 1326 777 424 353 99 256 160 106
Toothbrush
Price cut by 20% 1405 734 402 339 121 245 154 101
Coupon increaseby 20% 1464 713 384 319 118 245 155 103
Advertisingincrease by 20% 1431 734 388 327 125 141 153 102
Display increase by 20% 1419 730 397 334 124 243 152 102
Featureincrease by 20% 116 246 155 104
Toothpaste
Pricecut by 20% 1748 519 334 279 105 252 159 105
Coupon increaseby 20% 1763 571 299 255 112 248 158 103
Advertisingincrease by 20% 1665 594 338 283 107 252 157 105
Display increase by 20% 1659 600 340 281 105 253 159 104
Featureincrease by 20% 1641 604 343 292 104 255 158 104

Toothpaste Toothbrush
BrtandI Brand2 Brand3 B-rand4 Brand I Brand2 B-rand5 B-rand6
Toothbrush
Price cut by 20% .0596 -.055 -.052 -.04 .2222 -.043 -.038 -.047
Coupon increase by 20% .1041 -.082 -.094 -.096 .19/9 -.043 -.031 -.028
Advertisingincrease by 20% .0792 -.055 -.085 -.074 .2626 -.449 -.044 -.038
Display increase by 20% .0701 -.06 -.064 -.054 .2525 -.051 -.05 -.038
Featureincrease by 20% .17/7 -.039 -.031 -.019
Toothpaste
Price cut by 20% .3183 -.332 -.212 -.21 .0606 -.016 -.006 -.009
Coupon increaseby 20% .3296 -.265 -.295 -.278 .1313 -.031 -.013 -.028
Advertisingincrease by 20% .2557 -.236 -.203 -.198 .0808 -.016 -.019 -.009
Display increase by 20% .2511 -.228 -.198 -.204 .0606 -.012 -.006 -.019
Featureincrease by 20% .2376 -.223 -.191 -.173 .0505 -.004 -.013 -.019
Notes: The cross-effects are indicatedin bold and the own-effects are indicatedin italics. The figures in bold and italics refer to the impactsof the policy
changes on the sales of the umbrellabrandimplementingthe policy change.

The display cross-effects range from 3% to 7%. Similarly,
the feature cross-effects in toothbrush (the impact of a
change in the display activity in toothpasteon toothbrush
sales) are 5% for both Brands I and 2. Finally,price cross-
effects range from 3% to 6% as well in both categories.
The simulation results for a temporarypolicy change by
Brand I (Brand2) in the toothbrushcategorysuggest thatas
a percentageof the own-effects, cross-effects on Brand I's
(Brand 2's) toothpaste sales are 27%, 54%, 30%, and 28%
(23%, 61%, 43%, and 64%) for price, coupon availability,
advertising, and display, respectively. Thus, in the tooth-
pastector
category,couponavailabilityandthen advertisinghave
the largest cross-effects as a percentageof own-effects for
Brand I, and displays and coupons have the largest cross-
effects as a percentageof own-effects for Brand2. The sim-
ulation results for a temporarypolicy change by Brand I
(Brand 2) in the toothpastecategory suggest that as a per-
centage of the own-effects, cross-effects on Brand I's
(Brand2's) toothbrushsales are 19%,40%, 32%, 24%, and
21% (16%, 61%, 43%, 64%, and 29%) for price, coupon
availability, advertising, display, and feature, respectively.
Thus, in the toothbrushcategory, as in the toothpastecate-
gory, cross-effects as a percentageof own-effects are largest
for coupons and advertisingfor Brand 1. For Brand2, in the
toothbrushcategory, cross-effects as a percentageof own-
effects are largestfor features,because of a policy change in
the toothpastecategory, followed by coupons and advertis-
ing. In contrast,in the toothpastecategory these effects are
largestfor display and coupons, followed by advertising.As
these numbersshow, as a percentageof own-effects, cross-
effects are substantialfor all marketing-mixelements.
MANAGERIAL IMPLICATIONS, CONCLUSIONS,AND
FURTHERRESEARCH
We tested for the spillovereffects of advertisingand other
marketing-mix/salespromotion strategies (price, coupon,
display,and feature)in umbrellabrandingon scannerpanel
data.We found that advertisingspillover effects, along with
the use experiencespillovereffects, affect both utility mean
and variance. The latter effect provides evidence for the
uncertainty-reducingrole of advertising (along with use
experience) across categories for umbrellabrands.To cap-
ture the varianceeffect in a parsimoniousway, for the first
time in the literature,we estimated a multivariatemulti-
nomialprobitmodel, in which the variancesandcovariances
of random components of utilities across categories were
updatedon the basis of use experience and advertising.
Although there were some differences across categories
and brands,marketing-mixspillover effects were largestfor
coupon availability,generally followed by advertising.The
relativelylargecoupon effects, as well as advertisingeffects,
can be explained by various behavioralphenomenarelated
to brandequity; for example,couponingor advertisingby an
umbrellabrandmay increaseconsumer awarenessof prod-
418 JOURNAL OF MARKETINGRESEARCH, NOVEMBER 2002

Table 5
RESULTSFORPOLICYCHANGESBY BRAND2
SIMULATION

Sales of Toothpaste Sales of Toothbrush

Brand I Brand2 Brand3 Brand4 Brand I Brand2 Brand5 Brand6
Baseline 1326 177 424 353 99 256 160 106
Toothbrush
Price cut by 20% 1294 813 419 354 85 307 140 89
Coupon increaseby 20% 1270 867 403 340 84 305 142 90
Advertisingincrease by 20% 1306 829 404 341 85 296 146 94
Display increaseby 20% 1311 814 410 345 95 275 152 99
Featureincreaseby 20% 96 271 153 101
Toothpaste
Price cut by 20% 1206 966 382 326 95 266 157 103
Coupon increaseby 20% 1207 963 383 327 94 269 156 102
Advertisingincrease by 20% 1226 950 385 319 94 267 158 102
Display increase by 20% 1237 924 392 327 95 264 158 104
Featureincreaseby 20% 1246 915 390 329 93 269 157 102

Toothpaste Toothbrush
Br-andI Brand2 Brand3 Brand4 Brand I Brand2 Brand5 Brand6
Toothbrush
Price cut by 20% -.024 .0463 -.012 .0028 -.141 . 992 -.125 -.16
Coupon increaseby 20% -.042 .1158 -.05 -.037 -.152 .1914 -.113 -.151
Advertisingincrease by 20% -.015 .0669 -.047 -.034 -.141 .1563 -.088 -.113
Display increase by 20% -.011 .0476 -.033 -.023 -.04 .0742 -.05 -.066
Featureincrease by 20% -.03 .0586 -.044 -.047
Toothpaste
Price cut by 20% -.09 .2432 -.099 -.076 -.04 .0391 -.019 -.028
Coupon increaseby 20% -.09 .2394 -.097 -.074 -.051 .0508 -.025 -.038
Advertisingincrease by 20% -.075 .2227 -.092 -.096 -.051 .043 -.013 -.038
Display increase by 20% -.067 .1892 -.075 -.074 -.04 .0313 -.013 -.019
Featureincrease by 20% -.06 .1776 -.08 -.068 -.061 .0508 -.019 -.038
Notes: The cross-effects are indicatedin bold and the own-effects are indicatedin italics. The figures in bold and italics refer to the impactsof the policy
changes on the sales of the umbrellabrandimplementingthe policy change.

ucts that share the same umbrellabrand(Aaker 1991) and
may strengthenthe brandimage (Keller 1993). Given that
clipping the coupons and using them requiresmore cogni-
tive elaborationon the partof consumers,such cross-effects
may even be largerfor coupons than for advertising,as our
results indicate.
We also found thatadvertising(as well as use experience)
shifts both utility mean and variance across categories.
Thus, we found that the varianceof the randomcomponent
of utility in each categorydeclined as a resultof advertising
(and use experience) in eithercategoryover time. The cross-
category effects of advertisingon the utility mean may be
due to the factors discussed previously,such as awareness,
as well as advertising'seffect of increasingquality beliefs
underuncertainty.However,the effect of advertisingon util-
ity varianceis evidence for learningdue to advertisingunder
uncertainty (the variance of quality beliefs is decreasing
over time because of advertising).Thus, we find evidence
for the uncertainty-reducingrole of advertisingacross cate-
gories, which gives rise to advertising spillover effects,
though the effects on mean utility may be a combinationof
different impacts (relatedor not relatedto uncertainty).
Our empirical results obtained from the analysis of
revealed preference data have several other managerial
implications. First, the results confirm the expectation that
umbrella brandinggenerates savings in branddevelopment
and marketingcosts over time and enhances marketingpro-
ductivity.These results call for integratedmarketingcom-
munications across products that share the same brand
name. Work on brandequity suggests this for advertising,
butour resultssuggest thatthe need for integrationexists for
all the marketing-mixelements, includingsales promotions
and especially couponing strategies, given the relatively
large cross-couponingeffects we found.
Second, our results imply that there are some asymme-
tries across productcategoriesand across brands.Forexam-
ple, for Brand 2, featuringtoothpastewill have the largest
cross-effects on toothbrushsales, whereas displaying the
toothbrushproduct(or giving a coupon for it) will have the
largestcross-effectson toothpastesales, ceteris paribus.Fur-
thermore,as a percentageof own-effects, cross-effects of a
policy change in toothbrushseem to affect toothpastesales
more than the other way around in the case of Brand 2,
whereas the magnitudesof cross-effects as a percentageof
own-effects are fairly similar across the two categories for
Brand 1. Therefore,it would be importantfor brandman-
agers to know how specific cross-effects work across cate-
gories for their brands.
Although it was not the focus of this article,we also esti-
mated and found evidence for several within- and cross-
category correlations in response coefficients. First, we
Iound that marketing-mixsensitivities are generally posi-
tively correlated within a category. We also found that
depending on the brandin question, some consumers who
Advertising and Sales Promotions in Umbrella Branding
like certain brands may be more or less price sensitive than
the average consumer. These results indicate that in these
two categories, there are segments that, overall, are
marketing-mix sensitive, as well as segments that are
marketing-mix insensitive. Second, the finding that con-
sumers' marketing-mix sensitivities tend to be correlated
across categories suggests that there is a strong individual-
specific component to marketing-mix sensitivity.
There are a few avenues for further research. We showed
evidence for both utility mean- and utility variance-shifting
effects of advertising and use experience, and the the vari-
ance effects indicate evidence for the uncertainty-reducing
role that advertising and use experience play across cate-
gories for umbrella brands. The processes by which these
spillover effects occur are worth exploring in more detail.
However, as discussed by Keane (1997a), because of identi-
fication issues that arise with respect to models estimated on
panel data, it is often not feasible to differentiate among all
possible behavioral processes using scanner panel data. Sur-
vey or experimental data would be needed to provide a
detailed account of all the underlying behavioral processes.
REFERENCES
Aaker, David A. (1991), Managing Brand Equity.New York:The
Free Press.
and Kevin L. Keller (1990), "ConsumerEvaluationsof
BrandExtensions,"Journalof Marketing,54 (January),27-41.
Abe, Makoto (1997), "A Household-LevelTelevision Advertising
Exposure Model,"Journal of MarketingResearch,34 (August),
394-405.
Ainslie, Andrewand PeterE. Rossi (1998), "Similaritiesin Choice
Behavior Across Product Categories,"MarketingScience, 17
(2), 91-106.
Allenby,Greg M. and James M. Ginter(1995), "UsingExtremesto
Product Design Products and Segment Markets,"Journal of
MarketingResearch, 32 (4), 392-403.
Arora,Neeraj(2000), "UnderstandingLoss Aversionand the Role
of Multiple Reference Points:A HierarchicalBayes Approach,"
workingpaper,MarketingDepartment,Universityof Wisconsin.
Bell, David and James Lattin (1998), "Shopping Behavior and
ConsumerPreferencefor Store Price Format:Why 'Large Bas-
ket' ShoppersPrefer EDLP,"MarketingScience, 17 (1), 66-88.
Erdem,Tulin (1998), "An EmpiricalAnalysis of UmbrellaBrand-
ing,"Journal of MarketingResearch,34 (August), 339-51.
and Michael P. Keane (1996), "Decision-MakingUnder
Uncertainty: Capturing Choice Dynamics in TurbulentCon-
sumerGoods Markets,"MarketingScience, 15 (1), 1-20.
and Russell Winer (1999), "Econometric Modeling of
Competition: A Multi-category Choice-Based Mapping
Approach,"Journal of Econometrics,89, 159-75.
Geweke, John (1991), "Efticient Simulation from Multivariate
Normal and Student-t distributions Subject to Linear Con-
straints,"in ComputerScience and Statistics Proceedingsof the
Twenty-ThirdSymposiumon the Interface,E.M. Keramidas,ed.
Alexandria,VA: AmericanStatisticalAssociation, 571-78.
Guadagni,Peter M. and John D.C. Little (1983), "A Logit Model
of BrandChoice Calibratedon Scanner Data,"MarketingSci-
ence, 2 (Summer), 203-38.
Hajivassiliou, Vassilis, Dan McFadden, and Paul Ruud (1994),
"Simulation of Multivariate Normal Orthant Probabilities:
Methods and Programs,"WorkingPaperSeries DP 31021. New
Haven,CT: Yale University,Cowles Foundation.
419
Heckman, James B. (1981), "Heterogeneityand State Depen-
dence," in Studies in Labor Markets, S. Rosen, ed. Chicago:
Universityof Chicago Press, 91-139.
Ho, Teck, ChristopherS. Tang, and David Bell (1998), "Rational
Shopping Behavior and the Option Value of VariablePricing,"
ManagementScience, 44 (2), 145-60.
Keane, Michael P. (1990), "FourEssays in EmpiricalMacro and
Labor Economics," doctoral dissertation,Departmentof Eco-
nomics, Brown University.
(1994), "Simulation Estimation for Panel Data Models
with Limited DependentVariables,"in Handbookof Statistics,
G.S. Maddala,C.R. Rao, and H.D. Vinod, eds. New York:Else-
vier Science Publishers,545-71.
(1997a) "CurrentIssues in Discrete Choice Modeling,"
MarketingLetters,8 (3), 307-22.
(1997b), "Modeling Heterogeneityand State Dependence
in ConsumerChoice Behavior,"Journal of Business and Eco-
nomnicStatistics, 15 (3), 310-27.
Keller, Kevin L. (1993), "Conceptualizing,Measuring,and Man-
aging Consumer-BasedBrandEquity,"Journalof Marketing,57
(1), 1-22.
Lane, Vicki R. and RobertJacobson(1995), "Stock MarketReac-
tions to BrandExtensionAnnouncements:The Effects of Brand
Attitudeand Familiarity,"Journalof Marketing,50 (1), 63-77.
Leigh, James H. (1984), "Recalland RecognitionPerformancefor
UmbrellaPrintAdvertisements,"Journalof Advertising,13 (4),
5-18.
Loken, Barbara and Deborah Roedder John (1993), "Diluting
Brand Equity: When Do Brand Extensions Have a Negative
Impact?"Journalof Marketing,57 (July), 71-84.
Manchanda,Puneet, Asim Ansari, and Sunil Gupta (1999), "The
'Shopping Basket': A Model for MulticategoryPurchaseInci-
dence Decisions,"MarketingScience, 18 (2), 95-114.
,and (2000), "Multi-categoryBranding
Effects: A Shopping Trip Based Analysis," paper presentedat
the 2000 MarketingScience Conference, Universityof Califor-
nia, Los Angeles (June).
McFadden,Daniel (1974), "ConditionalLogit Analysis of Qualita-
tive Choice Behavior,"in Frontiersof Economletrics,P. Zarem-
bka, ed. New York:Academic Press, 105-42.
(1989), "A Methodof SimulatedMomentsfor Estimations
of Discrete Response Models Without Numerical Integration,"
Econometrica,57 (5), 995-1026.
Montgomery, Cynthia A. and Birger Wernerfelt(1992), "Risk
Reductionand UmbrellaBranding,"Journalof Business, 65 (1),
31-50.
MulhernJ. Francis and Robert P. Leone (1991), "ImplicitPrice
Bundlingof Retail Products:A MultiproductApproachto Max-
imizing Store Profitability,"Journalof Marketing,55 (October),
63-76.
Rajendran,K.N. andGerardJ. Tellis (1994), "ContextualandTem-
poralComponentsof ReferencePrice,"Journalof Marketing,58
(January),22-34.
Rangaswamy,Arvind, Raymond Burke, and Terence A. Oliver
(1993), "BrandEquity and the Extendibilityof BrandNames,"
International Journal of Research in Marketing, 10 (March),
61-75.
Roberts,John and PrakeshNedungadi(1995), "StudyingConsid-
eration in Consumer Decision Process: Progress and Chal-
lenges," International Jour-lnalof Research in Marketing, 12 (1),
3-7.
and Glenn Urban(1988), "ModelingMultiattributeUtility,
Risk and Belief Dynamics for New Consumer Durable Brand
Choice," Managemenllt Science, 34 (2), 167-85.
420 JOURNAL OF MARKETINGRESEARCH, NOVEMBER2002

Rossi, Peter E. and Greg M. Allenby (1993), "A Bayesian
Approach to Estimating Household Parameters,"Journal of
MarketingResearch,30 (2), 171-82.
, Robert E. McCulloch,and Greg M. Allenby (1996), "On
the Valueof HouseholdInformationin TargetMarketing,"Mar-
keting Science, 15 (4), 321-40.
Russell, Gary,S. Ratneshwar,andAllan D. Shocker(1999), "Mul-
tiple Category Decision-Making:Review and Synthesis,"Mar-
keting Letters, 10 (3), 301-18.
Seetharaman,P.B., Andrew Ainslie, and Pradeep K. Chintagunta
(1999), "Investigating Household State Dependence Effects
Across Categories," Journal of Marketing Research, 36
(November),488-500.
Tauber,E.M. (1981), "BrandFranchiseExtension: New Product
Benefits from Existing Brand Names," Business Horizons, 24
(March/April),36-41.
(1988), "BrandLeverage:Strategyfor Growthin a Cost-
Controlled World," Journal of Advertising Research, 28
(August/September),26-30.
Tellis, GerardJ. and Doyle L. Weiss (1995), "Does TV Advertising
Really Affect Sales?"Journalof Advertising,24 (3), 1-12.
Wernerfelt,Birger(1988), "UmbrellaBrandingas a Signal of New
ProductQuality:An Exampleof Signaling by Posting a Bond,"
RANDJournalof Economics, 19 (3), 458-466.

CALL FOR PAPERS

APPLIED PROBABILITYMODELS AND MEASUREMENT
IN MARKETING
Due to an increasedavailabilityof dataandalso to an increasedinterestamongstthe practitioners,
one can notice more and more applicationsof statisticaltechniquesand stochasticmodels in
marketing.It is not uncommon that marketingdecisions are made based on the outcomes of
sophisticatedmodels for empiricaldata.As developmentsin this areaproceedwith rapidpace, it
is importantto once in a while establishthe currentstate of the art concerningthese empirical
models. It is for this reason that the journalAppliedStochasticModelsin Businessand Industry
is a
(Wiley) publishing special issue on
Applied Probability Models and Measurement in Marketing.
All empiricalworkrelatedto the probabilisticmodellingof choice,preference,and/orbehaviouris
welcome.Analysisof largedatasets is also encouraged.Paperscan be submittedelectronicallyto
the Guest Editors:
AlbertBemmaor PhilipHansFranses
ESSEC or ErasmusUniversity
BP 105 P.O.Box1738
95021 Cergy Pontoise Cedex NL-3000DR Rotterdam
France The Netherlands
bemmaor@essec.fr franses@few.eur.nl
The deadlinefor submissionis June 30, 2003. All paperswill be refereed,and there is no a priori
guarantee of ultimate acceptance.