Marketing Impact On Firm Value

Marketing's Impact on Firm Value: Generalizations from a Meta-Analysis
Author(s): ALEXANDER EDELING and MARC FISCHER

Source: Journal of Marketing Research, Vol. 53, No. 4 (AUGUST 2016), pp. 515-534
Published by: Sage Publications, Inc.
Stable URL: https://www.jstor.org/stable/44134929
Accessed: 04-02-2020 19:45 UTC
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.
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
https://about.jstor.org/terms
Sage Publications, Inc. is collaborating with JSTOR to digitize, preserve and extend access to
Journal of Marketing Research
This content downloaded from 117.211.201.226 on Tue, 04 Feb 2020 19:45:05 UTC
All use subject to https://about.jstor.org/terms
ALEXANDER EDELING and MARC FISCHER*
The interest in the value relevance of marketing investments has given

rise to numerous studies on the marketing-finance interface. This study
integrates extant research findings and establishes empirical generalizations
on marketing's impact on firm value. Specifically, the authors conduct a
meta-analysis of prior econometric elasticity estimates of the stock market
impact of marketing actions and marketing assets. Analyses based on 488
elasticities drawn from 83 studies reveal a mean elasticity of .04 for advertising
expenditure variables and of .54 for marketing asset variables. Among
marketing assets, customer-related assets show a higher mean elasticity
of .72, compared with .33 for brand-related assets. Further analyses show
that advertising elasticities are lower in more concentrated industries
and that marketing asset elasticities are higher during recession times.
Researchers should also be aware that characteristics of the research
design (e.g., the type of firm value metric used, the omission of control
variables, or not accounting for endogeneity) may affect the estimation
results.
Keywords: meta-analysis, marketing firm value elasticity, empirical mar-

keting generalizations, hierarchical linear model
Online Supplement : http://dx.doi.org/10.1509/jmr.14.0046
Marketing's Impact on Firm Value:

Generalizations from a Meta-Analysis
i
Eve an
dou as
the M
as costs on firms' income statements have forced the mar- in
as a firm's stock market valuation.
keting profession to prove its value relevance (see, e.g.,
Marketing Science Institute's research priorities from 2014 Although there are several well-established empirical
to 2016). Numerous studies have indicated that marketinggeneralizations for sales response elasticities (e.g., Bijmolt,
action variables such as advertising expenditures (e.g., Joshi
Van Heerde, and Pieters 2005 [price]; Sethuraman, Tellis, and
Briesch 201 1 [advertising]), no such elasticity generalizations
exist with regard to the important performance metric of firm
value. Thus, the first key contribution of this study is to in-
*Alexander Edeling is Assistant Professor of Marketing, University of
Cologne (e-mail: edeling@wiso.uni-koeln.de). Marc Fischer is Professortegrate the findings of previous econometric studies and to
of Marketing and Market Research, University of Cologne, and Professoridentify and compare mean effect sizes for marketing action
of Marketing, University of Technology Sydney (e-mail: marc.fischer@ and marketing asset variables. Empirical generalizations are
wiso.uni-koeln.de). The authors appreciate comments from the JMR review
valuable for both scholars and practitioners. For researchers,
team, as well as the input from participants at the 2013 Marketing Strategy
Meets Wall Street Conference in Frankfurt, the 2013 Marketing Science
Conference in Istanbul, and the 2013 Annual Meeting Quantitative Mar- throughout this article, we adopt Srivastava, Shervani, and Fahey's
keting in Cologne. They are also grateful to Gerry Tellis and Harald van (1998, p. 2) definition of marketing assets as those "that arise from the
commingling of the firm with entities in its external environment." This
Heerde for their valuable feedback on previous versions of the article. Peter
Verhoef served as associate editor for this article. understanding includes both financial (e.g., customer equity, brand equity)
and nonfinancial (e.g., customer satisfaction, brand attitude) metrics.
© 2016, American Marketing Association Journal of Marketing Research

ISSN: 0022-2437 (print) Vol. LIII (August 2016), 515-534
1547-7193 (electronic) 515 DOI: 10. 1509/jmr. 14.0046
516 JOURNAL OF MARKETING RESEARCH, AUGUST 2016
they express what has been learned services

in the versus durable goods industries or recessionary
marketing-finance
field over the last 35 years. Marketing versus nonrecessionary
managersperiods can usemay produce
the different firm
mean effect sizes as supporting arguments value effects of both
marketing variables. and
within
outside the firm to increase marketing's accountability.
Thus, the second contribution of this In-study is to identify
vestors and analysts can incorporate determinants
meanofeffectadvertising expenditure
sizes as and marketing
reference values into their valuation asset elasticities
modelsrelated(e.g.,to Schulze,
substantive influences (e.g.,
Skiera, and Wiesel 2012). product type, region) and to identify design characteristics
Following the standard of sales (e.g., response
the type of
meta-analyses,
firm value variable used). These results
we use the marketing firm value elasticity may provide managers
as the with guidelines
effect to determine, for
size
measure. This measure expresses the example, under which circumstances
percentage change in marketing invest-
firm market valuation in terms of a 1% increase of the mar- ments are more or less effective. Marketing researchers
keting input variable. It is unit free and makes estimation can use the research design results to prevent biases in fu-
results comparable across a wide range of studies. However, ture marketing-finance studies (e.g., accounting for endo-
we observe an important difference between our study and geneity) or to interpret the results from studies that apply
prior meta-analyses on sales response elasticities. Note that different methodological approaches (e.g., use of different
firm value is a profit measure that derives from discounted firm value variables) correctly.
expected future cash flows. Whereas the relationship between We structure the remainder of this article as follows:
sales and marketing effort is monotonie, the relationship be- First, we provide a brief overview of the marketing-finance
tween firm value and marketing effort is not. Because mar- research stream. We then define the scope of the study and
keting effort has both positive and negative (cost) effects describe the literature search procedure and the database
on profit and cash flows, respectively, the profit curve has a we generated. Then, we provide the research framework for
maximum that is associated with the optimal level of mar- the meta-analysis. This is followed by the description of
keting effort. This has implications for the interpretation of the meta-analytic model and the presentation of descriptive
firm value elasticity findings, which is not as straightforward statistics and estimation results. We conclude with a dis-
as it is for sales elasticities. For example, a null finding does not cussion of the study's implications and limitations as well
necessarily imply that marketing effort has no impact on firm as suggestions for further research.
value. It may rather express that marketing is operating at
an optimal level, which implies a marketing firm value elasti- RESEARCH ON MARKETING AND FINANCE
city of zero.2 Following this efficiency argument, a negative Marketing Value Chain
elasticity signals that the firm is overinvested in the market-
ing activity or asset, whereas a positive elasticity suggests that Broadly speaking, the marketing-finance research stream
the firm is underinvested. Thus, unlike sales response meta- addresses the influence of marketing actions and marketing
analyses, our study also allows for drawing generalizable con-
assets on firm value (Srinivasan and Hanssens 2009). Our
clusions about the optimality of firm marketing behavior. underlying conceptual rationale is that marketing creates
However, differences in firm value elasticities cannot be value for the firm according to the theoretical framework
explained by disparities in optimality alone. Assume that presented in Figure 1. We distinguish between the follow-
two marketing variables are set at approximately the same ing three major categories of decision and performance
variables:
efficiency level; a gap between elasticities may still result
from a different power to drive firm value. Thus, we also
1 . Marketing actions refer to decision variables along the mar-
need to incorporate this perspective, which we call the keting mix. They are under direct control by marketing
effectiveness argument, when interpreting the empirical managers. Investors typically observe these actions and their
findings. Throughout this article, we consider both the ef- associated cost; however, the effect on firm performance is
ficiency and the effectiveness arguments to develop our less obvious to them.
hypotheses and explain the meta-analytic findings. 2. Marketing assets result from the relationship between the firm
Comparing results from individual empirical studies is and important external stakeholders such as customers and
a complex task for several reasons: First, researchers can retailers. The asset enables the firm to implement and exploit
operationalize the dependent firm value variable as market strategies for improving its efficiency and effectiveness in the
future (Srivastava, Shervani, and Fahey 1998). It represents
capitalization, the ratio of intangibles to tangibles (Tobin' s q
an important intermediate outcome variable that is driven by
or market-to-book ratio), or stock return (percentage change
marketing investments. Marketing assets are not as transparent
in stock prices). Autocorrelation issues should have a more as marketing actions to the investor community, which is
severe effect on level than on change metrics. Second, largely due to the lack of generally accepted measurement
model specification differs between studies. Whereas some standards. On the one hand, the value of brands and customers
studies incorporate only balance sheet variables (e.g., book can be measured in monetary terms. On the other hand, there
value of assets), others also control for income statement are perceptual measures available, such as brand image or
items (e.g., earnings). This may reduce a potential omitted- customer satisfaction, which describe the strength of the assets
variable bias. Third, industry and temporal characteristics in psychological terms.
are not the same across studies. Different conditions in 3. Firm performance refers to a company's accounting (top line
and bottom line) and stock market performance. Because firm
value is a future-oriented and cash-based measure, it is regarded
2Let n(X) denote firm value that depends on marketing effort, X. Firm as the ultimate performance measure (Rappaport 1998).
value elasticity is defined by en,x = (dn(X)/dX)(X/TI). At the optimum, the
Prior
first-order condition dII(X)/dX = 0 must hold. From this condition, it follows work has investigated two routes of value creation
£n,x = assuming that optimal X and n are both strictly positive. by marketing (Hanssens et al. 2014). The first route can be
Marketing's Impact on Firm Value 517
Figure 1
THEORETICAL FRAMEWORK
Marketing (Mix) Marketing Firm

Actions Assets Performance
• Communication Valuation Metrics Accounting Results

• Product
,prïce _ -Brand value «Top line (revenues)

• Distribution (+) . Customer lifetime value/ © • Bottom line
Perception Metrics Firm Value
• Brand perception . Market capitalization

• Customer satisfaction • Intangible-to-tangible ratio
• Quality perception • Stock return
described as tangible Srinivasanorand Hanssens

direct. (2009) provide aArigorous
vastcon- st
has investigated the effect
ceptual structure of marketin
of the marketing-finance research and the
line results (i.e., sales and
first narrative summary revenues) with
of empirical findings. The hand-
intermediate assetbook metrics
edited by Ganesan (2012)(for an overv
extends Srinivasan and
Parsons, and Schultz Hanssens' s work2001).
by providing detailedThe effec
reviews of the lit-
sumed to be positive erature in- that
specific areas such asis,
product investme
innovation, adver-
lead to a positive tising,
sales and so on. response. At
These conceptual reviews improve our th
keting actions induce
understanding ofcosts, leading
the marketing-finance interface. However, to
effect on bottom-line results.
they do not produce These
empirical generalizations that can be opp
motivated marketing-finance
obtained from a meta-analytic approach. research
the relationship between marketing
Conchar, Crask, and Zinkhan (2005) provide the first ac
firm value to demonstrate whether the revenue or cost ef- meta-analysis of empirical marketing-finance work. This
fect dominates. Regarding the most-often studied variable, analysis focuses on the effect of advertising expenditures
advertising expenditures, most studies have reported a pos-on firm value and offers insightful generalizations about the
itive effect on firm value (e.g., Joshi and Hanssens 2010),role of advertising for generating firm value. Our study is
but some studies have found a negative effect (e.g., Lu also quantitative in nature and complements and extends
and Beamish 2004). prior work in important ways. First, in addition to adver-
The second route, the intangible or indirect route, ac- tising, we collect and analyze firm value effects across
counts for the notion that marketing actions usually first many more variables, including brand and customer assets,
lead to a change in customers' mindset, which in turn re- marketing capabilities, and marketing actions covering
sults in purchase decisions and thus in higher revenues, price, product, distribution, and online communication.
profits, and eventually firm value. Recent "mindset metric" Second, we extend Conchar, Crask, and Zinkhan' s sample
research (e.g., Hanssens et al. 2014; Stahl et al. 2012) has considerably by including 55 (compared with 15) advertising-
overcome the limitations of previous studies that have related studies. Considering additional marketing variables
focused on pairwise relationships between marketing ac- further increases our sample to 100 studies in total. Finally, we
tions and marketing assets (e.g., Yoo, Donthu, and Lee use a different effect size measure, firm value elasticity, and
2000) and marketing assets and firm performance (e.g., consider new moderators such as the control for endogeneity
Anderson, Fornell, and Mazvancheryl 2004; Barth et al. in empirical models.
1998). Their results support the coexistence of both a
DESCRIPTION OF DATABASE
tangible, direct route and an intangible, indirect route of
value creation through marketing initiatives. Scope of Study
We defined four criteria for including a study in our meta-
Prior Review Studies
analysis: First, the study must use an econometric model. This
Table 1 positions our meta-analytic study relative to means that we exclude portfolio analyses (16 studies) in
existing review studies on the marketing-finance interface. which portfolios are constructed using descriptive statistics
ff 1
si
Ö 05 05Ö1/5
1/5 3
s
■« 'S I =3
•sí i 'S 'a 1-1 'a •» 3
1 -5 "8 S , 5 S -S •» s
^ Ö s £ , § £ g -e £
g<->
<-> 2o
g 2oggUU -g
-g «1
«1& M
C --Ö
C 33 §yyZ3
Z3 o
J3'S
@ ^ 00 ÏÏ ' e 00 U 'c c3 'C y
@
fe^ .S
? 00
§ a 'g ïÏÏ äS a,
e .H
u 00
a ä
M 1 U 'c S | u c3 'C S
1» fe mSÕ a ï I« a, o s a i S
fe -S g B Jt B S o s §" <» o o S §
I fi! ì ¡i Il <» 1 1 1 i
1 I'll v1.? *'S'Il! 1 i
š -5 $ <s > ^ g *9'% S | S ęś s ģ N
a. a.
Ã s -ź > c
s u « 'C
^ «-ź
íá «c'G
w S« aOJ
X! xxi
.2 N2 L: "O
o "O
^ Ã |UT3 u I c ^ ^ íá ā c w g OJ §á t xxi 2 ü 'S o
filili
.SOS .SOS
o ^ III 1.iSH ShoS¿
i 5 î
co
o ShoS¿
LU
D
• ••• • ••••• ^
D
I-
CO
^ o
I"
§ .Q G o
UJ Co U
-a g r
>
LU ^
00 g cd
§
.Já
'§ «o °P<D
r
CC SC O c« o S
LU
O
£ļ cL 52 Mou*c
»2 g 'C-Ö
-Ög«2 S xi
3 *2 'S "2 ^ 3
^ if ^
Z cc 12çjS"52 «2 »2 -O g«i *c ö o ü *cö G & ö
Öa•£>
.2 •£> _ c
_ &Û
<
z
< &o 8 "* y y
«*«i-cu uü ëfi
GGgG
y y &Û
« .S .8 go -« H
-2«yy¿VV to
£
*3 G fa Ö0Ö0 ÜUhS.2 00.2 sÛMgSw) S
u. *3 S *53 3 CG C 5 .2 S 00.2 G P ¿ O (U G V G G 5
'-S ^ *§ "g * g I 5 "8 ? « 'fi •§ -B (U '8 g s 'S jž
8 '-S S ^ S ^ -8 s § g S 'fi ^^-S-sIs-giS-g ? « 'fi g jž &
z
I-
&-Š& S ^ â il -8 s 2 3 c.ï 3<£ûfiO |cQU I & >»
•3 < 8 SS « u £ Q s
LU K a
* >-h • • • •••• • • *4)
*o
cc O O O O
<
2
'3
2
2 2 2 T3
'5 '5 '5 2
LU
1/1 C« C« «3 .2
« , Já Já Já Já E
>
i
E'
« 'o
^
Ü m
, 'a§mSm§ 'a
£ § to
60
S a
§s
oo
Ł
.s
^ Ö «í G caGoíOeaGjôeaG O
£ 'c .a B .as rs .as «R .ss &
0) p
-Q 2
J UcflË
£ s 3 ¿ a¿'cia¿"cla¿ »
UcJjËË^ &iËË^ 'LiË Ö
^^sjássjáss^sjáss^sjáa *>
H^3
HÔ ^ ^ ^^sjássjáss^sjáss^sj
^ ^ ]¿ %> M 1¿ %> M V %> M M ^
û
Z
<
|lg M |lg S£wwU
S£wwU M S5CO
2£comU g1 |lg
g 2£comU M M g^lf
• •••••••••••••s* u
<
D % I
£
LU ■M C ° « On & S © = O 2 O I «
C ° « On
«cr,Co-H<S
© O
CŅ
O
(Ņ
«
£
I «cr,Co-H<S 1*8 =Ł ¿ ci CŅ ri (Ņ I £
O
Z
O
o |<§|
o,o,I s -H
-H g S- - --S; < 1
0
z 5 1
fe
- ea
sp*
p* >>
-a >>
-S
ra
i -S
3 3
Q jü -G G vî
•rei5
•rei5 jü
•h ii GÖ ? ß
h ö
ö ■«
LU
LL
O
§^ļ >P
^■S ^ a Ü ^ ^ ^
5 Sil ° a -3 ^ " I -
^ ¿ř ^"8 a " 1 - -8
LU
>
1^1 .G e Ss O 1 "O !
GC
LU
.G
►5 S S
O
1 11 ■« 1 2
< s s g 2
xi 2
£ £ Ë
^ <l> >
4) >0)&> ¿á
Cd
«P
(V eg *Ö *Ö «S "O ^
^ *ö S 2 "ö <u 5
^ » I ! I i 1 9
ê a a 8 .S «Ä «3
73 2Ž Já
S « a
ÎQ S ^ Cd G
■s"§"2Ci
3 0 ā 9 0§ -- 2"" ã"e"
S
u. s i 9 I - § S -s
ï
2O
"S
II
O -SG ss
g N-S È È
^^ <U
Q) G
I Q)
I SO3 3
. s G
. G
so 'C 5 íG 5
U O H es
(e.g., mean brand value ratings) and

database includes studies usingthen compared
data sets from a time period w
regard to their average return.
of 40 years (1971-201
Second,
1), which stem the
from North
dependent
and South
iable must be firm valueAmerica, Europe, and Asia. The 100 studies
or shareholder value, contribute 621
respect
Although most studies focus marketing firmon shareholder
value elasticities. Of these, only value
164 (26.4%) th
flects the market value of equity,
are directly reported as others consider
elasticities or interpretable as such bot
market value of equity and because ofdebt (e.g.,
a multiplicative model through
specification. the u
We could not include
Tobin' s q). Because shareholder valueall 621 elasticities
is the in the sub-
key firm
variable in the vast majority of studies,
sequent multivariate meta-analysis for we use firm
the following rea-
and shareholder value sons: interchangeably
First, we identified only few elasticities throughou relating to
study. We exclude studies the with cash
marketing action flow
variables product(8(n studies)
= 43 [6.9% of all o
observations]),
variables (17 studies) as the dependent online communicationvariable. (n = 35 [5.6%]),
Altho
these metrics are closely distribution
related (n = to
8 [1.3%]),
firm and price (n = 7 [1.1%])
value, theyand to ar
ceptually different. Third, marketing the model
capability variables (nmust include
= 32 [5.2%]). Using these a
elasticities
interval-scaled firm-specific as dependent variablesvariables
marketing in a multivariate meta-
(mar
mix, marketing assets, and analytic marketing
regression model would be unfeasible because of T
capabilities).
we exclude event studies thethat investigate
small number of degrees of freedom. the effect
The remaining
categorical independent observations
variable related(event
to the marketing occurs or ad-
action variable not
cause we cannot derive an elasticity
vertising expenditures (n =estimate
298 [48.0%]) and tofrom marketing t
studies. Fourth, the study assets must
(n = 198 [31.8%])
either allow for multivariate elastic
report analyses.
directly or include information
Because advertising that enables
expenditures us to
are antecedents to calc
mar-
elasticities (e.g., coefficient estimates
keting assets, we conduct two andseparate mean
meta-analysesvaluon
focal variables are required). Because
these variables. Second, for eachthe reporting
of the two types of in- o
ticities is much less commondependent forvariables, we excluded outliers outside the in-
marketing-finance m
than for sales response models, weelasticity
terval of the mean had plus toorcalculate
minus three standardelas
in most cases (for details deviations
on these (Bijmolt, Van Heerde, and Pieters 2005).
calculations, see Thisthe
Appendix). results in final sample sizes of 296 advertising expenditure
elasticities from 55 studies and 192 marketing asset elas-
Database Compilation
ticities from 42 studies. The total sample size of 488 is
As a starting point, we
similar reviewed all
to the sample sizes in relevant
previous stu
meta-analyses (e.g.,
included in Conchar, Crask,
Albers, Mantrala, and Zinkhan
and Sridhar 2010; n = 506).(2005)
In the Web
Srinivasan and Hanssens (2009).
Appendix, we list the Next, we
studies that are applied
included in the meta-
word search (e.g., "marketing
analysis. firm value," "mark
stock return") in several online databases (e.g., EB
RESEARCH FRAMEWORK
Google Scholar). We also did a manual interdiscipl
search in leading journals
In from
Figure 2, we marketing, managem
organize the various potential drivers of
finance, and accounting
firmresearch.3
value elasticity in two To diminish
major categories: substantivepu
tion bias (Rust, Lehmann, and
drivers and Farley
research 1990),Substantive
design characteristics. we ch
drivers
the Social Science Research explain the variance
Network of elasticities bypaper
working differencesda
and requested unpublishedthat arise
work from different
from marketing
authors activities/assets
who andha
sented marketing-finance from different product
research atand market conditions.
major Previous
conference
Marketing Science Conference, Marketing
meta-analyses (e.g., Strategy
Sethuraman, Tellis, and Briesch 201 1) M
Wall Street Conference, have
Marketing Dynamics
shown that the effectiveness of marketing Confer
variables
Finally, we conducted a varies
cross-reference search
across product types. In addition, to ide
market conditions
additional relevant studies. such as geographic region, recessionary periods, or the
structure of competition could affect firm value elasticities.
Database Scope Research design characteristics entail factors that are re-
The literature search resulted in the identification of lated to the type of data, modeling and estimation deci-
sions, and the control for other firm influences. A crucial
100 studies published or written (working papers) between
1977 and 2013 that satisfied our criteria. Notably, the
data-related decision includes the choice of the dependent
proportion of unpublished studies (20%) is substantially
firm value variable - that is, whether to use market capi-
talization, intangibles-to-tangibles ratio (e.g., Tobin's q),
larger than the mean proportion in marketing meta-analyses
(6%) identified in a study by Eisend and Tarrahi (2014),
or stock return as the dependent firm value variable. Im-
reducing the risk of a meta-analysis selection bias. portant
The questions researchers face concerning model and
estimation decisions pertain to whether to account for
endogeneity or heterogeneity, which could potentially bias
specifically, we reviewed the following leading journals: Journal of
estimated
Marketing , Journal of Marketing Research , Marketing Science , Journal of effect sizes. Similarly, we investigate whether
the Academy of Marketing Science , International Journal of Researchthein inclusion or omission of important variables such as
earnings, market
Marketing, Marketing Letters , Management Science , Administrative Science share, or firm growth affects firm value
Quarterly , Academy of Management Journal , Academy of Management
elasticities. We also study the influence of the manuscript
Review , Strategic Management Journal , Journal of Finance , Review of
status (published or not) on effect estimates. Finally, we
Financial Studies, Journal of Financial Economics, Journal of Financial and
Quantitative Analysis, Journal of Accounting and Economics, Review consider
of the interaction effect between time and the type
Accounting Studies, and The Accounting Review. of marketing asset.
Figure 2
RESEARCH FRAMEWORK
Substantive Drivers
r ~ ~ ~
ļ Marketing Influence Interactions Product and Market Conditions

i
ļ • Advertising expenditures vs. • Type of marketing asset x Time «Product type

i marketing asset • Region
! • Type of marketing asset (brand- • Market concentration
i related vs. customer-related) • Time
ļ • Economic condition
Publication-Related
/ Firm Value Y
V Elasticity f
Research Design ' ManuscriPt status
Characteristics ^ -
1 Data Characteristics Model and Estimation Control for Other Firm

i Characteristics Influences
1 • Type of firm value variable
, (dependent variable) • Endogeneity • Earnings
1 • Temporal aggregation • Heterogeneity • Market share
I • Structure of data (e.g., panel • Estimation method »Growth
1 data) • Functional form • R&D
ļ • Duration of the effect • Leverage
' • Size
I • Competition
i • Risk
i
information (e.g., 10% increase in costs that are I directly

incurred) and the upside (potential revenuee increase).
As a result of these opposite effects, the advertising
i
elasticity for firm value should be lower and f closer
to zero. m
The situation is different for marketing r assets. Accordin
to Srivastava, Shervani, and Fahey (1998), e marketing asse
can lead to accelerated and enhanced future cash flows
t that
come at a lower risk. For example, firms with strong r brands
benefit from faster trials of new products and h price pre-
miums through a higher perceived value of the offering. m
Strong customer relationships provide opportunities
a for
the influence of all other moderator variables on firm value loyalty and cross-selling. Recent mindset metrics studies
elasticity estimates in Table 2. Here, we also provide a de- have shown that the long-term sales elasticity of a per-
tailed description of our variable operationalizations and ceptual asset variable such as "brand liking" is, on average,
their use in previous meta-analyses. Note that our expec- indeed more than 16 times larger than for advertising
tations about a variable's influence on firm value elasticity (Srinivasan, Vanhuele, and Pauwels 2010). In terms of
can differ across the advertising expenditures model (AEM) cash outflows, investors will evaluate investments in mar-
and the marketing assets model (MAM). keting assets as more beneficial than advertising expendi-
tures for two reasons. First, the costs of achieving a lift
Advertising Expenditures Versus Marketing Assets in a marketing asset (e.g., 10% increase in brand equity)
Advertising expenditures and marketing assets have dif- were incurred in the past and have no bearing for the fu-
ferent roles in a firm's value-creation process (see Figure 1). ture profitability of the firm. Second, because marketing as-
Expenditures are a flow variable (i.e., flow of money dur- sets are much more "sticky" than advertising expenditures
ing a specific period of time), whereas marketing assets (Hanssens et al. 2014), there is less need for future in-
are a stock variable (i.e., value of the asset at a specific
vestments to keep them at a certain level than for adver-
moment of time) (Hanssens and Dekimpe 2008). As such, tising. Given that firms benefit more from marketing assets
they differ considerably with respect to their contribu- than from advertising expenditures in terms of future re-
tion to the discounted future cash inflows and outflows venues and suffer less with respect to current and future
of the firm. Advertising expenditures represent a current costs, we expect this difference in effectiveness to translate
investment into a future uncertain sales uplift. Thus, in-to higher firm value elasticities for marketing assets than
vestors need to trade off between advertising expenditure for advertising expenditures.
ss Su Su g i
-§
1 Ë
U pq
Sí c/5
fS Ë
çj
i s" il SL
a. 03 3 - pag
SS co fj 2 SS Ü2
SS Il co fj IS SS 11 Ü2
s L I
1 °! S
S â 'S J g
I¡ini
3 o I§ 3
-
„ S" s
g t f £ S 3
u O- J g -f -S
s * u < S ž J i g 'fi 's
1 -s
«S'z'z'
8 S
S «
fe! u
u -ā
" " s
S¿. m q a © «
I SIS «« M ! '3 h ,H
Ü ,H G G es (n
S CS M M
w ■s Ü 1 1 G 1 -8 I = (n
-I
•& ce S 3 -o Ja 3 =
LU
Q 'S II ce 8 Ł S E g, §
O I 3 S I S a 4 I I
2
0
I-
1
řv
řv aí
g. xx
<D <D
xx 8. IMIII
O O >>
S o 5 5 g -S
< o ^ a a o 0 ļg G
B' o o «, Jš 0 S .SP
1 ¿3 ¡z; ¡z; c . g " "
^ W) O 2 60 «
Č Ê-H
LU <ß $ .2 $G g'ftûl
o ^ Õ
X £ S s .s *3 g
UJ S Sá
X -S £ 8^s g>
.s g *3
°
g g
2
&
CM f ^ j -S ~ 5 .S ^ g u
S ^ G Ö U c/ì £ 3
Z
S ¡z; - G -g ^ ^ c/ì cd c«
»«
inniïilîls
-g ^ cd c«
-Q ~
cö Q
I- ^ m
I- ^ CO
D
CO
LU mil
ü H o 8 • -
(U ^ TD &û 2
OQ
< £ ^ Ö 8 u .£ G
« Ē c S 3 G
ÛC
< psšal
f " ««Ē "> c O
S 3M
G
> #3.ē.ē.ss
OC
1
cc
LU
ill. ,1'ì 1 II î! Sfip ¡ s . ifsilïl I ,
Û
O UH
73 'S!^ §„i
73 111!
S^ „ifiS°ļļ2
sii!
fi SfS3S3so
f"£
* Pl'ili
S*£-JnSsi^SSSÛiīiSsoģ^
P "£ ¡t? Srfí^lšíSl
S g-JnSsi^SSSÛiīiSsoģ^
| o^sä^liol'ovüiiis^ ^^ SS ï !¡ 1 ,
2
Utcli ® § S S3S3so o û"i| g s-Sí'fi » S = I5 S S í i I
•It3£ Utcli
o, G®^ -S
* *i-i
J I•.sO1 G
** obh
g .í^£. ■Í3r^l-Sl«-i¡|!vÎ-§U-2¡
>^X e"0 P3 R* S s-Sí'fi >^X»O
2 5 "O » SA
= 2<L>
S S íî <+i
-S I
•§ o, ~ Ž SI el S-2 g Sor^M 6 g s o
« jš ~ ^ Ž « SI Sfliā<š?fi&tf-BSļāS£>§-aJ£l
I i 3 I I «s •asS|<S|d|^É|,|sll|l^SS)i^Ìc(Ì'lc'IÌ e
IÃ"
S«ļ|l ^«ä"
«s I Sñ"'5^g>u§S¿S6PoS2§e-q--§5S^0»«>S«|
Sñ"'5^g>u§S¿S6PoS2§e-q--§5S^0»«>S«|
! I 8 ! t! g 1 1 2 ®J Š § 1 ! ! S g
a vS
lesili
a Ige
^«ä" Pni ii Sñ"'5^g>u§S¿S6PoS2§e-q--§5S^0»«>S«| »I» M»SÏ H|| Sili 'Ji" « Iff Ił! -I g
l¿s|| S ģ s J Jí S ^ S ? ã I 1 'n .2 1 Š IT* J> sí ^ 1 1 ■§,£ ^ E | g y
i-îîll l¿s|| Ill|ililļ|ii|li3 S ģ s J S ^ S ? ã 'n £ .2 1 Š 8ļiiļis;iiii|li J> sí ■§,£ Sm E g y
H of S" S 8 M ¿3 > G <43 ößU S. S § cd hh 8 £ 3 ^ O'S cz> .5 £ 43 8 8 § Sm c3 ¿3 ,2 'S S
w o
ll-ôês II "S
B 8-1 -51 i Sjlsla lif!
■a
■a-Ul Qcluii II«^•SSS.sb
s Qcluii «^•SSS.sb gSls
gSls§ 1 I i
-g li-sl I I 1 «itttfll S'il §
I i|gi?V |ť?° I
I 3 |H 1* ° s
■a !§ ł « g" Í
I ■a eC ^ s* « ^ a," m E j
le Ö S eC 1 s* S 1 a," ŠjH m si j
* M U E 2 & & «g S 00 S
4 * iisaijiei
< co eî le
< <i co<i
< <i S¿i
< §SPO
è
I ■! il ! tl-a3
I 9 .s I -S ¿ § 'i - I Ē
1 s .s 1 1 8 c ! a 1 - 8 1 Ē
« Ö - h ¿sê! c a E g S Ö
11 J UM«
•-■gao - I gh-g8*
.S 2g- "5 &¡ .S i
•s
c ^cÏ3t||
tt^«ÎS"2J
-2 Is •OO'S'"
•OO'S'" S M!
H3»| l&S »"Soi
§ s§ ; s» S
5 g »
t s :§ 3
5 fi fiÉ 312§ -c
"sul-ñ
13 .y
5 ^ « -S ; a «gß M
t a•?
•? Ç^ 5 g13
^ 13 £ p12
£ S c/j ¡2 >-»i S -c p >-» 2
•Ssaî »*« c/j S «
§|Sa -Í2o :3 i 8 &
lii! §|Sa 1§! :3 fils i &
- - Ills Ijp »
I - S - S |"S=Jáfe|| -S I U 8 » § § § §
I « » läi lî seji 8 i i i i
ì «S
I«S
5Xl« -s
5 1»551
•- £>
i - Si 2
5 i -- *
_C
ť 1 9 « 1§9 IS
* io 2(ÍťCS (u^CccD
i 8Łs|s&
8 
8 (u^CccD
8 8
<L> g- <L> g
"S i Xl i £> « 5 1 _C 1 - .a 2
§ * * i,:- 1 .a i 1 tē§l g 'Ł u 'Ł o u
§■ 8 8 eži -ta -S a i g -a 'S u 1 O o o o
¿J en C/5 ,2^-SrCeSS a g « -a E 8 u 3 £ £ O ¡Z¡ o z o z o
,2^-SrCeSS Silnice « g .ž4 8 ß 3 I £ 8
S -s 5 S ^
o
!¿Ü!Í<g S ~ S bSí
bSí «O.a <iSO
g S ~s s■§
s -2
=3 i-S2 5-SI ss 'S
'S ss
CM W
^ ÍuuèÔT3M(fl -S S ^ S SsO? p
^S f^ ss fi
-S 1S 1-sC -ail
S '3d-SÊ e^ *e s*s s8». ?8^
® 2
■S
Ie0 z 22 ■* r~^ C 'p <S o 22 ws £ j_i
d >gü<UOU¿§ 22 ■* r~^ 'p <S o g> 22 «í £ ws j_i £
H O rm • rt
rt-5-5
O "O P "O
ü T3 r? H
S« ppj=^ü ü
"1 "1
.2 C W)
W) r r
6 fi
.s 8 s a s fc. % « S *fi s ■«
•s -8.1 s Sil % III « S *fi s ā S'
&¡^ ^^ 8I <
g ą ^^3
s »„S «.a«ï .a.S e^ i g
^2«>%25
U ~ M > £ ^ ~
Q< O O L- 3 ^ e
MÛScs-Sûûi
MÛScs-Sûûi fi1i-8a||gPJ
U M*rt
'S S L-O§ 3-û
i S ê
Q 2 e .2 o ~ V, ¿ - . 2 u 2 ~ $ sfjg g
o
C "S oÜ5.2
Ü5.2 5 gI
3 J= u m h g g
S
o¡3.>0 1>. ź S
-g
u o -o -g
r -zi°
c5 O ce°
odu cem2¿|
-73 g
o o ~3 12^ ^
S «ť ļļ u
3 3
jf o
C o O 0) o cd 3 h § Oh o¡3.>0 3 0.) CJ ^ >. <4-h C CÄ Wh #, ^Tx -73 D O o ^ ""Sj ^2 = S
S 1 o 8.(2 O 0) § cd ¿Bi § Oh -a^ÛS 3 0.) CJ ^ go? <4-h C S | .g CÄ «2 Wh .è1 <» #, ^Tx 8 JS D g «s ^ « ""Sj .a
'Coa^pQ -â ^ ^ ! o § a ° il ¿2 'a •§ S S -i
s 11 'Coa^pQ si/à īsi -â ^ *41 ^ ! o § s-î a il ii§ ¿2 'a •§ S Vntml 11 S -i -5
•B .22 I o o 'l těl ¡ ^ -S -i ! ¡ $ fc. 8 J 1 üfSSlfill ¡1 t
INI!» 5 ï i s 1 1 s i«â§> fc. I«! Is I
1 ilSŠN
I ilSŠN 1|i iBž>St
Ēi1iJi s ¡?2'i 5lig
5lig lili
-S «g i«â§> í-lí
51§īlā|&S GZ» llillflM
5 -3 51§īl
° 1 s :s a 8 S ilf^l^õ II. s !=>■§! i ¡1^1 lêi II
isô'ôeü«
ri a s 8 acScSc-p
ô'ôeü« !!!>
5 o 5 ouuô.S3rt^°ocg
leliā'lgS'š
ņ ÔÛ -r3 ô.S3rt^°ocg
^ ^ ^ ^-is ma
«i'j« ma «fl-M^sái
'3^c*o „ ^ ^c*o
«îgqjja „^ ¡1^1
S'j 8 3pjC<u2c«n^:
3pjC<u2c«n^: ^ «i « -i cc'3 -^ «ñ «1 SOa-SaöSS^Sa^
«îgqjja 60 M n SOa-SaöSS^Sa^
& £ Xì'ìà -c G ™ .cnu0?-c
~ ù, C0Ö0
& w llilllUiíllll 3pjC<u2c«n^: - ñ «1 60 M n lllllllliilïl SOa-SaöSS^Sa^ il C0Ö0
ill sã«« 1 S 111 .1

sã«« .c S 2ž'-°^t2 tSCT't« o
f§»S
OOÂC'-S'VPÎÎË-, OOÂC'-S'VPÎÎË-,
- ^ C/D ti «+H*J I-I3Cd¿_ 1 -a I
-HZJ^ § - ^ •& ?
-! OOÂC'-S'VPÎÎË-, 1 ! 1 ?i ! ! I ! 2 ^ 1 _ S C/D I ti &I ¡ fli «+H*J sil» I-I Cd _ Ili -HZJ^ ili
|i^^ s
s S11 fi
fiÌ51| ««Sá0 111
««Sá0 |D3|D3 iil H
Illfllll
04i H
řS04 ā
řS o<
ā o< z
z HI §<£ lllll
§<£ eooôeooô lllllll
SJSosSJ SJSosSJ
1 8 i • g • s • • i 'S
Jh H 00 o W S W UH O
• • s • • ^ • • • §•••&••
2 < Q S
1 ^
e* U S i* U M Ž fi I
I§
e Ö
uuu§
uu§Ë Ë IuË ISi* ^Ö a"
S MI
fi Í
Í
4 8 1 ï
^55 55 o oS
-OÜCÖÜ« 0> jrT
•a 1 -OÜCÖÜ« a•a
».1*18* S 1*88
-1 8. Il >,*
0> » jrT
»
•3 ? £••§ g- s S 5 3 g
•3 > g 1 i g- s 5 -s g s « « =3 1 i
-o5S
SfiiS«? fl.So .2 Ö s
1-g-g»8!! fl.So s ¡5 s -a S s
w,
S"!
>° 8
0
Gi?"S
U
2 iaSi« ë*!
ü
S5
s ¡5 8 "3° °«1 *
iaSi
cq
S«.
*P c -S
c
.22 H
2
Ôû
55
II ¿d
w, ^ >,'5b a O .g cc 8 £ 2 3 Ą
5 ÍS8 8¿| 111 £ 5,1 ï
1
^ ^5I 5
« i.s
8j¡-8&¡g
.sii.
ï
«
|ll.
^ *^ p
* p
f| õ
1
§
a g -g g a >
«* &
g
| >p
g g,i
£
p £
o
§« 8
o 8 "S
"S I Nuu
Cx, &Û •« Öß c« 2l3 ^
1 Cx, 11 &Û •« siili Öß <u ?? lili SS ^ T3

1 |OI|Ē11IE|1| SouĒ«ou I I I I i i i Ö
3 -«
-«SouĒ«ou
Ü
~g|S «U
S I I I Ic
O I || -a.SáD
w
«
SÖ
fl
3 ^ ^ ^ u u -a.SáD
§ ~
I&«U
« soisl
u i 2 u iS c2fl-d
t 2'ö■§e 'ö
Ë Ioeo
Ë ô ô^ o
^ S á* »£ £>S§« Z Z o Z o Z o Z o o-s =3 ' §
Ł =3
Ł » 12
i ë
•S S U -S r 9- 8 S
tS'S^-OW'Sr^SH'S g"U T3 S c
9- S &a a5 5£' 7?
£
§l'!S<§g7M8s -SPS c 7?
22üoü2tií3 -S § "i Ł-I J3'P3„
üoü2tií3 J3'P3„13 'i 13 Ig2®
2® =
s Ilidis üoü2tií3 J3'P3„ 111- Jì 2® >
o
cm w
a) z
S Igü'ä ¡s &"ll J | ss •§
Îgü'ä I ^ I-8 til ¡s !
-2 ^
,co z
i- O
TÍ gg í?|j
M 'S&O«í3 8 §
O «í3 £ ^^^ §T3
CQII
TÍ !|-og|| M 'S O O -sili ^ l|lx. T3 II
o
lelilí SI 8% 1 1 i « B
«•Sfi
5 2
5
ēl 6 aa"!
«•Sfi
8 S 8
8s*<•
igS gã ĒĒ •&
" •§
¿ ¿ SM § i=3
•§ i8 8§1£
I 2 S g~
~
s=öSÍ3
<Š§l§L.Ssž Śle I (2-a¿ ||
II 1|Sî S U r _ £.s § | 3 Š 11
-4) *SS Ē "O U ^ g ^ S - û H ü g 2 « ^ ^ V3 cd J
¿•g s d 13 .a 3 8 'S 11 » g 'S S 3§ 'Z ? "2 š >;í g g £> ï g ^ § - 8 í H ' ü I g -S 2 13 « ^ s ^
Já ¿•g I I 'i. .a * -a c -a » I 'Z I ? S I š -f >;í 'g -i g -S g I .s ^ a s g S « ä S s & ^ S
c 'S £• >| S -a -M ë á -c 1-c •" -c ^ = u -s S. f ■- I "? I ° -c S
! c 'SI £• il S l'I -M i l'i. á -c l'I 1-c •" .glrf.s.8 -c = u S S. fl*'§«¿ f ■- "? S I S -c I 2 S Î,
.§ 'SI ai II il iii i l'i. N l'I Si I .glrf.s.8 s 8 1 1 .a S 8 fl*'§«¿ Ï I 2 1 i S 1 S e? I 5 Î, 1 II
10-i ai
PDPD &"ÍálS
"ÍálS $« fá«-
^ ^^ 4)^ Qí ^JrH
i^g «n^^HS 1 öS'ÖH^SG^CCjS
^ «n^^HS ° g aS'SSS^â e? s ° <$
öS'ÖH^SG^CCjS Ilu
•« 0 Sc PD I« 8e^ 4) e^G JrH >Ö on
ç Sc«
^ ç Sc« ^ rt>
^ rt> «2 !r
2 0w««3
2 02 so>
n 2~ Xw?,n US
3 ?,2 2HH8 -2
Ü G .S
& .S
^ a^ řa £
Tuř ><îg à« 2 P
§ Ö & Ö.2 ö ö.g Ö« Ö 2 « Ü «3 c -S § ö I ē J £ ä1 S c ÖÖ a;-
•fí S2 -q 'S 2 & 'S Š ^ ö "öS ö.g 5 S ^ Ü ^ -o o já r 5 ö - Ł J o ä1 o S « « c fi ri a;- S°°
I~ē f| f & s«l& ft- !&2!s^ä!§5«!3| 1 1 S.S
ja i? 3 S S§" lÎllsIâ^^^ÂÍ^Jí s^õ §73
(d 3 ļ rtS>0«iS « cd S cd Q 0«iS
73 ^ c 73 J OH^ P u &1« cd
73 oOhJ "OcdOH CTíK>A(L)CC>C>>
i±3«L)Ohh<L>CTíK>A(L)CC>C>> i- 3 c3 Jgdj CXc3fiË^
s
« "Sai
e S £ S
s
" u b B « 11 g.
Eu -I s I s« ï « „
Eu Ö| -I lš «I s > il s«
1 lil! JIHU ¡Ü liiHiJli > 1¡ âi||i<
^ ¿i-§l •lilHl
1 r|50
^ ě I • |Sol£0
lil iiliiliil ¡s .rullili
.g • qSOÏSO'SO&SO
• s • • m<Z »••m<Z ^ ! •53Ž5
53Ž5
J^S
• 5 -s s
ggw So p2 -§ S í «Z
Su «2 X
brand metrics such

In addition to the effectiveness argument, we as image,
argue likeability,
that, or attitude (e.g.,
Bird, Channon, and
from an efficiency perspective, advertising firmEhrenberg
value 1970). The value of op-
elas-
timizing
ticities are also lower compared with customer
asset firmmetrics has been
value intröduced only
elas-
recentlytends
ticities. Note that firm value elasticity (e.g., Blattberg
toward andzero
Deightonthe1996; Reinartz,
Thomas,
closer the firm operates at the optimal and Kumar
level for the2005). respective
Thus, we argue that because
marketers are more
marketing variable, whereas underspending experienced with managing brand-
(overspending)
related metrics Advertising
leads to a positive (negative) elasticity. compared with customer-related
ex- metrics,
they manage
penditures are a decision variable, brand-related metrics can
and management more efficiently. In
directly and permanently change combination, the leveltheoftwo arguments
this suggest the following
variable.
There are also no concerns about the measurement of hypothesis:
advertising expenditures. Finally, firms have engaged in
H2: Marketing asset elasticities are higher for customer-related
optimal advertising budgeting for more than 50 years asset variables than for brand-related asset variables.
(Hanssens, Parsons, and Schultz 2001), so they are very
experienced in finding optimal advertising levels. InIndustry
con- Concentration4
trast, the value of marketing assets has only been In
ap-more strongly concentrated industries, fewer and
preciated for two decades, if at all. Asset metrics such as firms compete with one another. In such oligopolistic
larger
brand equity are intermediate outcome variables, markets,
which competitive reactivity is particularly pronounced
are not easy to measure. They are sticky - that is, management
(Gatignon 1984). An increase in advertising by one firm
cannot quickly change the level from one period directly to the affects its competitors, which often retaliate im-
next but has to wait for several periods until its activities
mediately to maintain their share-of- voice levels. This questions
materialize. In addition to advertising, many other factors
the effectiveness of advertising actions in highly concen-
(e.g., product features, retail coverage) drive the asset trated
value. markets, leading to lower returns to advertising. Em-
Consequently, finding the optimal level for a marketing pirically, Danaher, Bonfrer, and Dhar (2008) find lower sales
asset is much more complex than for advertising expen- response to advertising in the presence of competitive inter-
ditures. The limitations in measurement, together with the With the expectation that such top-level effects will
ference.
notion that marketing assets are less deeply understood translate to bottom-level earnings and, eventually, firm value,
by non-marketing-oriented chief executive officers and
we hypothesize:
chief financial officers, suggest that firms are probably
still underinvested in brands and customer relationships. H3: Advertising expenditure elasticities are lower for more
strongly concentrated industries.
Thus, from an efficiency point of view, marketing asset
elasticities should be further away from zero than ad-
Recession
vertising elasticities (because of a larger suboptimality)
During recessions, customer demand declines. Adver-
with a positive sign (because of a likely underspending).
tising expenditures and marketing assets have been shown
Combining this reasoning with the effectiveness argument
to balance revenue peaks and slumps and to generate less
leads to the following hypothesis:
volatile cash flows (Srinivasan and Hanssens 2009). In
addition, especially during periods of financial turmoil,
Hļ: Firm value elasticities are smaller for advertising expenditures
than for marketing assets. investors' risk tolerance decreases (Hoffmann, Post, and
Pennings 2013) and they look for "safe harbors," with adver-
Brand-Related Versus Customer-Related Marketing Assets
tising spending and marketing assets serving as investment-
Brand-related metrics focus on value created through
decision surrogates for financial information such as profits.
Firms
(product) brands while customer-related assets focus onalso tend to cut marketing effort during recessions,
leading to a potential underinvestment (Deleersnyder et al.
value created through customer relations. Researchers
2009). Thus, we expect elasticities to be higher during
have conceptualized marketing value chains in which
recessionary times.
brand equity antecedes customer equity (e.g., Rust, Lemon,
and Zeithaml 2004; Stahl et al. 2012). Drawing on the
H4: (a) Advertising expenditure elasticities and (b) marketing
hierarchy-of-effects models of consumer behavior, they asset elasticities are higher during recessions.
argue that companies need to win the hearts and minds of
Type of Firm Value Variable
consumers (i.e., building brand equity) before acquiring
and retaining satisfied customers (i.e., building customer
The choice of the dependent variable should affect elasticity
equity). Indeed, Stahl et al. (2012) show strong empirical
estimates. Whereas market capitalization and intangibles-to-
evidence that customer-based brand equity partially tangibles
me- ratios are level measures, stock return by defini-
diates the impact of marketing investments on customertion is a first-difference metric. "Levels models" with highly
asset metrics such as acquisition and retention rates.autocorrelated
Be- dependent and independent variables suffer
cause the link between customer-related assets and firm
from a spurious regression problem, which leads to downward-
value is closer than for brand-related assets, customer- biased standard errors and thus to an overreporting of sig-
related firm value elasticities should be larger than brand- nificant effects (Mizik and Jacobson 2009). This problem
related elasticities.
does not arise in "differences models." Thus, we expect
We arrive at the same conclusion when taking the effi-
ciency perspective. Although achieving an optimal level is
by no means easy for either brand or customer metrics, 4 Very low variance of the concentration variable in the marketing as-
firms have longer experience with managing and monitoring
sets data set prevents us from studying this variable in the MAM.
elasticities from marketthe capitalization and

Web Appendix also presents the marketing intangible
firm value
tangibles models to be higher
elasticity database, than
which includeselasticities
the 621 elasticities and from
return models. corresponding coded information concerning substantive and
research design characteristics.
H5: (a) Advertising expenditure elasticities and (b) marketing
asset elasticities from models using market capitalization or
Meta-Analytic Model and Estimation
intangibles-to-tangibles ratio as the dependent variable are
higher than those from models using stock return as the Following Bijmolt and Pieters (2001), we model the elas-
dependent variable. ticity as a function of the selected independent variables
using hierarchical linear modeling (HLM). An important
Interaction Effects assumption of ordinary least squares regression is that the
We considered interaction effects in our meta-analysis. errors are not correlated (Greene 2012). However, the fact
To reduce the number of potential interactions, we focus that there can be multiple measurements of the elastic-
only on interactions between substantive drivers. Inter- ity in one study leads to a violation of this assumption.
actions among study design characteristics (e.g., using Determinants are observed either at the study level (e.g.,
stock return as dependent variable and accounting for endo- publication status) or at the measurement level (e.g., inclu-
sion of an earnings variable). Therefore, measurements of
geneity) are difficult to interpret from a conceptual point of
view. Considering all substantive drivers, we end up with tenthe elasticity are not independent within one study, and
two-way interactions for both advertising and marketing additional study-specific factors might exist that the in-
cluded independent variables do not control for. To account
asset elasticities. In addition, we consider the operationali-
zation of the marketing asset (monetary vs. other) as a for potential within-study error correlations, we use an
potential interaction variable for marketing assets. Because HLM with the measurement of an elasticity as the lower
variables are measured in categories, the effective number level and with the study from which an elasticity is derived
of interaction variables is even larger. This aggravates the as the higher level.
multicollinearity problem that is introduced by interaction In addition, marketing firm value elasticities are not true
variables and plagues many meta-analyses (e.g., Albers, parameters but are estimated with error (Sethuraman, Tellis,
Mantrala, and Sridhar 2010; Sethuraman, Tellis, and Briesch and Briesch 201 1). To account for this measurement error in
2011). We therefore use a strict procedure to select in- the dependent variable, we weight each observation with
teraction variables. First, the joined categories of two a normed variance (i.e., the absolute value of the ratio of
variables must show at least 5% of total observations. the estimated elasticity and its standard error; for a similar
Second, the variance inflation factor must not exceed approach,
10, see Bezawada and Pauwels [2013]). Because
which signals potentially severe collinearity issues. Third, information on statistical significance (e.g., standard errors,
the interaction effect must significantly add to the t-values)
ex- is not available for all observations, this slightly
planatory power of the model according to a likelihood- reduces the sample sizes to 269 for the advertising ex-
ratio test that passes the 10% level. Following this pro- penditures data set and to 178 for the marketing assets
cedure, we are left with only one interaction variable data set.
for
marketing asset elasticities out of 18 tested effects for
advertising and 11 for marketing assets. This is the in- RESULTS
teraction between time (metric variable) and the type of
Descriptive Analysis
customer-related asset (dummy variable). Because the type
of marketing asset has two categories, the main effectOverview. of Table 3 shows descriptive statistics for all
time de facto measures the effect of time for brand-related marketing firm value elasticities with at least 30 observa-
assets. The interaction effect measures the extent to which tions.5 All marketing elasticities, with the exception of
product, are, on average, significantly different from zero
the effect of time is different for customer-related compared
with brand-related assets. ( p < .10). We discuss advertising and marketing asset elas-
ticities in more detail subsequently. Online metrics include
valence and volume of online reviews as well as web
RESEARCH METHOD
traffic variables. Marketing capabilities refer to "a firm's
Data Coding ability to understand and forecast customer needs better
Two judges fulfilled the coding. Following Geyskens than its competitors and to effectively link its offerings
et al. (2009), the first author and a second judge who is not to customers" (Krasnikov and Jayachandran 2008, p. 1)
an author of this article coded a random sample of 30 and include variables such as marketing efficiency. The
studies. Coding agreement was greater than 90%. After relatively large mean elasticities for these metrics (.22
resolving any remaining inconsistencies, the first author for online metrics and .55 for capabilities) suggest that
coded all other studies. To achieve a high degree of trans- they indeed drive firm value. The null finding for product,
parency regarding these coding decisions and to enable re- however, is surprising in light of the prominent role of
searchers and managers to conduct further analyses, we product innovation. This might be because it is not the
provide two databases: First, because the vast majority of volume but the quality of innovation that is eventually
elasticities had to be calculated on the basis of parameter relevant for investors (Sorescu and Spanjol 2008), or
estimates and descriptive statistics (457 of all 621 elas-
ticities, 354 of 488 elasticities relating to advertising ex- 5The variables distribution (M = .33, Mdn. = -.06) and price (M = -.08,
penditures and marketing assets), we present the calculation Mdn. = -.00) have too few observations (eight and seven, respectively) to
procedures for each article in the Web Appendix. Second, obtain any inferences from their analysis.
Table 3
DESCRIPTIVE STATISTICS
Marketing Assets
Brand-Related Customer-Related
Advertising Expenditures Overall Assets Assets Product Online Metrics Marketing Capabilities
na 296 192 89 103 42 35 31
M .04 .54 .33 .72 .03 .22 .55
t-value (H0: M = 0) 5.29*** 8.23*** 3.67*** 7 94*** ļ 50 3.95** 1.78*
SD .12 .92 .86 .93 .12 .33 1.73
Mdn .02 .27 .09 .59 .00 .07 .11
Min -.37 -2.74 -.43 -2.74 -.44 .01 -.06
Max .77 4.72 4.72 4.59 .32 1.08 7.14
***/? < .01 (two-sided tests).

aWe excluded observations outside the in
because the market the means (medians) considers

of the brand-related and customer- inv
close to the related elasticities are .33 (.09) and
optimum (on .72 (.59) andaverage
both differ
to fully understand significantly from zero the(p < .01). Thisrelation
suggests a sub-
ness and firm value. stantially stronger firm value impact of customer metrics
Advertising expenditure elasticities. Figure 3, Panel A, compared with brand metrics, a result that we further in-
shows the frequency distribution of the advertising ex- vestigate in the subsequent multivariate analysis. Market-
penditure elasticities with n = 296. The mean elasticity is ing asset elasticities are also heterogeneously distributed
.04, with magnitudes ranging from -.37 to .77. The median (Q = 2,407.08, d.f. = 177, p < .01, 12 = 92.65), so an inves-
is even lower, at .02. Notably, 23% of all observations are tigation of moderators is warranted.
negative, suggesting that investors occasionally weight the Model-free evidence. Before we turn our focus on the
cost dimension of advertising expenditures more stronglymultivariate analysis of moderator variables, we aim to
than the revenue dimension or that firms in these studies are detect differences in firm value elasticities by comparing
overspending. Nevertheless, the mean elasticity is signif- means across variable categories. To be comparable with
icantly positive at .04 (p < .01). Note that the recent meta- the subsequent HLM results, we use the same samples
analysis by Sethuraman, Tellis, and Briesch (2011) finds a (i.e., n = 269 for advertising elasticities and n = 178 for
mean short-term elasticity of sales with respect to adver- marketing asset elasticities). Table 4 shows the results of
tising of .12 (.24 long-term). Apparently, the firm value this univariate analysis. Overall, 1 1 mean-difference tests
elasticity is substantially lower than the sales elasticity, among advertising elasticities and 15 mean-difference tests
which can be explained by the conceptual difference in the among marketing asset elasticities turn out to be at least
dependent variable. Unlike sales, firm value incorporates marginally significant (p < . 10). These findings already sug-
revenues and costs, which are both affected by advertising gest that there is systematic variation of elasticities that
expenditures; thus, it is not only the positive effect of ad- can be explained by moderating influences. Indeed, we find
vertising that is included in firm value effects. In addition, support for our expectation that advertising elasticities are
the close-to-zero mean elasticity suggests that the average smaller (1) in stock return models and (2) if endogeneity
firm operates at near-optimal spend levels. is corrected for. In addition, the tests provide extensive
The Q test (Q = 3,668.56, d.f. = 268 ,/? < .01), which tests support for the econometric regularity that effects are bi-
the null hypothesis that results differ only because of sam- ased if important variables (i.e., earnings, size, and com-
pling error, and the I2 statistic (92.69%), which is the per- petition) are omitted.
centage of total variability in a set of effect sizes due to Table 4 also reveals that this omitted- variable problem
true heterogeneity, indicate that advertising expenditure also seems to drive elasticity differences for marketing as-
elasticities are heterogeneously distributed. This finding sets. The omitted variables here are growth, research and
warrants the study of moderator variables (Huedo-Medina development (R&D) effort, leverage, and size. The differ-
et al. 2006). ence test also supports H2, which predicts that elasticities
Marketing asset elasticities. In Figure 3, Panel B, we for customer-related assets exceed those for brand-related
present the frequency distribution of the marketing asset assets.
elasticities with n = 192. Mean and median elasticity are .54

Overview of Meta-Analytic Results
( p < .01) and .27, respectively, and thus are significantly
larger (p < .01) than the observed mean (median) adver- Table 5 presents the maximum likelihood estimation
tising expenditure elasticity. Thus, Hļ receives support. results of the HLM for advertising expenditure elasticities
However, the range of elasticities is also larger, with a (columns 3-5) and marketing asset elasticities (columns 6-8).
minimum of -2.74 and a maximum of 4.72.6 Furthermore, We find seven statistically significant parameters (p < .10,
one-sided test if sign prediction is possible, two-sided test
6Note that the magnitude of a firm value elasticity may easily exceed 1
otherwise) in the AEM and nine significant effects in the MAM.
even if the underlying sales elasticity is relatively small. In the Web Ap- The overall fit of the models is satisfactory: the pseudo R2
pendix, we illustrate this with a numerical example. (squared correlation between estimated and actual dependent
Figure 3
FREQUENCY DISTRIBUTION OF ELASTICITY ESTIMATES
A: Advertising Expenditure Elasticities
60% -
50.7%
50% -
H
■
3 30% - H
■
"■ 20% * ł«* ■ 13.9%
■ ■ ■
0% -I - - i - ■■ - i - ^B - , - ^B - , - ^B - , - BB - , - ■■
<-.1 -Ito -.05 -.05 toO Oto .05 .05 to .1 .Ito .15 .15 to .2 >.2
B : Marketing Asset Elasticities

70% -
58-9%
60% -
I
I 30% -
20% ' H 13.5%
10% - 7 3% WĚĚ 5.2% 5.2%
0% -I -
■ H H ■ H M
<-.5 -.5 to 0 Oto .5 .5 to 1.0 1.0 to 1.5 1.5 to 2.0 >2.0
Notes: The percentages in Panel A sum to 100.1% as a result of rounding error.
variable) amounts to .876 in the AEM and of

.623 in the relationship,
the expected MAM. elasticities from models that
In the following subsections, we discuss ignore
theseendogeneity
regressionare significantly higher than those from
results in detail. models that incorporate it (coefficient = .056, p < .05).
Endogeneity is most often conceptualized in the form of
Estimation Results of the AEM
simultaneity between stock market performance and the
With respect to the substantive drivers, only the industry setting of advertising budgets.
concentration variable is significant. Consistent with H3, We find three other significant effects for research design-
advertising firm value elasticities in medium-high con- related variables for which we did not have prior ex-
pectations. Elasticities derived from cross-sectional data
centrated industries are significantly lower than those in
industries with low concentration (coefficient = -.063, p < are significantly larger than those from purely time-series or
.05). Note that in more strongly concentrated industries, a panel data (coefficient = .044,/? < .05). Thus, the effect found
few large firms compete with each other. If one competi- in the sales response meta-analysis by Assmus, Farley, and
tor increases the advertising effort, this usually affects its ri- Lehmann (1984) is replicated when firm value is used as the
vals in a noticeable manner. The rivals are likely to directly performance variable. Furthermore, elasticities from models
counter the attack by the aggressor (see, e.g., the "cola war" estimated with generalized least squares (coefficient = .034,
between Pepsi and Coke). As a result, volume gains are limited p < .10) and with other estimation methods such as nonlinear
and profit decreases. least squares (coefficient = .038, /? < .10) are marginally
With respect to research design characteristics, we find significantly larger than elasticities from models estimated
six significant effects. In accordance with H5a, we find that with ordinary least squares. Following Capon, Farley, and
elasticities estimated from models that use the level vari- Hoenig (1990), we do not draw any normative conclusions
ables market capitalization (coefficient = .102, p < . 10) and from this result, because the choice of estimation method
intangibles-to-tangibles ratio (coefficient = .032,/? < .05)always as depends on the specific research context.
the dependent firm value variable are higher than elastic-It would be worthwhile to compare our results with those
ities from models using stock return. Obviously, a spurious- of Conchar, Crask, and Zinkhan's (2005) meta-analysis
regression problem arises in levels models, leading to on advertising effects on firm value. However, we refrain
elasticities that are biased upward. In addition, in support from doing so because they use unstandardized regression
528 JOURNAL OF MARKETING RESEARCH, AUGUST 201 6
Table 4
COMPARISONS OF MEANS
Advertising Expenditure
Elasticity (n = 269) Marketing Asset Elasticity (n = 178)
Expected Difference Expected Difference
Variable Level Difference0 n Mean (p-Value)b Difference0 n Mean ( p-Value)b
Substantive Drivers
Marketing Influence
Type of marketing-asset variable Brand-related asset N.A. N.A. N.A. N.A. 88 .396
Customer-related asset N.A. N.A. N.A. N.A. + 90 .801 .003
Product and Market Conditions
Product type Across product types Base 204 .036 Base 163 .679
Only durables -/+ 23 .024 .799 N.A. N.A. N.A.
Only nondurables -/+ 21 .089 .658 N.A. N.A. N.A.
Only services -/+ 21 .064 .965 -/+ 15 -.251 .000
Geographic region United States Base 251 .047 Base 157 .634
Other -/+ 18 -.038 .000 -/+ 21 .355 .232
Concentration Low Base 37 .031 N.A. N.A. N.A.
Medium to high - 232 .043 .300 N.A. N.A. N.A.
Timec Earlier than mean Base 92 .040 Base 72 .771
Later than mean -/+ 177 .042 .945 + 106 .485 .061
Recession No recession in data period Base 137 .050 Base 39 .838
Recession in data period + 132 .032 .255 + 139 .534 .095
Research Design Characteristics
Data Characteristics
Type of firm value variable Stock return Base 53 .007 Base 96 .567
Market capitalization + 71 .042 .034 + 44 .586 .993
Intangibles to tangibles + 145 .053 .001 + 38 .703 .407
Temporal interval Up to one month Base 20 .087 Base 22 .175
Longer than one month -/+ 249 .038 .301 -/+ 156 .661 .006
Structure of data Time series (pure and panel) Base 157 .034 Base 146 .587
Purely cross-sectional -/+ 112 .051 .246 -/+ 32 .663 .699
Model and Estimation Characteristics
Endogeneity Accounted for Base 50 .015 Base 35 .222
Not accounted for + 219 .047 .008 + 143 .693 .006
Heterogeneity Accounted for Base 117 .052 Base 111 .649
Not accounted for -/+ 152 .033 .190 -/+ 67 .531 .411
Estimation method Ordinary least squares Base 199 .034 Base 116 .622
Generalized least squares -/+ 56 .030 .952 -/+ 33 .980 .379
Other -/+ 14 .188 .061 -/+ 29 .083 .017
Functional form Additive Base 171 .043 Base 117 .478
Multiplicative -/+ 64 .053 .829 -/+ 49 1.028 .002
Other -/+ 34 .011 .056 -/+ 12 .056 .000
Duration of the effect Short-term Base 244 .043 Base 167 .663
Long-term -/+ 25 .021 .132 -/+ 11 .112 .095
Control for Other Firm Influences
Earnings variable Included Base 108 .023 Base 104 .531
Omitted -/+ 161 .053 .049 + 74 .699 .134
Market-share
variable Included Base 38 .165 Base 17 .800
Omitted + 231 .021 .000 + 161 .580 .529
Growth variable Included Base 56 .035 Base 28 .236
Omitted + 213 .042 .278 + 150 .669 .018
R&D variable Included Base 181 .043 Base 20 .176
Omitted -/+ 88 .036 .656 -/+ 158 .655 .013
Leverage variable Included Base 94 .053 Base 20 -.005
Omitted -/+ 175 .035 .253 -/+ 158 .677 .004
Size variable Included Base 210 .022 Base 89 .332
Omitted -/+ 59 .109 .001 -/+ 89 .870 .000
Competition variable Included Base 47 .116 Base 36 .726
Omitted -/+ 222 .025 .002 -/+ 142 .569 .402
Risk Accounted for Base 53 .021 Base 82 .466
Not accounted for -/+ 216 .046 .036 -/+ 96 .716 .082
Publication-Related Factors
Manuscript status Published Base 249 .044 Base 118 .586
Unpublished - 20 .011 .126 - 60 .630 .372
aIf more than two means are compared, we apply either the Tukey t
groups), respectively.
bOne-sided test if there was an expected difference (+ or -), two-side
cWe dichotomized this variable through a mean split to allow for gro
scale.
coefficients as effect size and test differences in a bivariate rather fast stock market reaction to information con-
rather than a multivariate manner. This renders compari- tained in marketing asset metrics, which tends to be diluted
sons meaningless. with longer time intervals. Finally, contrary to the AEM,
elasticities are significantly lower if cross-sectional (vs.
Estimation Results of the MAM
longitudinal/panel) data are used (coefficient = -.995, p <
In the MAM, we find three significant main effects and a .05). In Table 2, we mention that there is no agreement
significant interaction effect among the substantive drivers. on the direction of influence. These inconsistent findings
First and foremost, H2 is supported by the multivariate add to the mixed empirical evidence on the effect of data
aggregation on marketing elasticities (Albers, Mantrala,
analysis, because the elasticity is significantly higher for
customer-related asset variables compared with brand- and Sridhar 2010 [longitudinal > cross-sectional]; Assmus,
related asset variables, which serve as the base category Farley, and Lehmann 1984 [cross-sectional > longitudinal];
(coefficient = 1. 180;/? < .01). Thus, the stock market seems Kremer et al. 2008 [insignificant effect]).
to acknowledge that customers are relational assets (Morgan
Additional Analyses and Robustness Checks
2012) that represent the realized value of marketing ac-
tions in terms of cash flow and are closer to firm value. In To obtain more insights from the data and to ensure the
contrast, brands are regarded as reputational assets (Morganrobustness of the results, we performed further analyses that
2012) representing intangible value potentially created by involve the analysis of negative over positive advertising
marketing initiatives. They are an antecedent to customer elasticities, outliers, method bias-corrected elasticities, mul-
behavior and are further away from firm value in the mar- ticollinearity, exclusion of cases due to missing information
keting value chain. The positive interaction effect be- on uncertainty, and short- versus long-term effects.
tween customer-related assets and time (coefficient = . 143, Analysis of positive and negative elasticities. A potentially
p < .05) suggests that this difference in value relevance insightful analysis would be to study the conditions un-
increases over time. In our view, the finding goes handder which negative versus positive advertising elastici-
in hand with the trend from product-centric thinkingties occur.7 Figure 3 reveals that almost one-quarter of
toward customer-centric thinking over the past two decades elasticities are negative, suggesting that firms are over-
(Shah et al. 2006). invested in advertising, whereas positive elasticities imply
Consistent with H4b, we find that marketing asset elastic-that firms are underinvested in advertising. We adopted a
ities are significantly greater during recessionary comparedlogistic regression approach to analyze the drivers of neg-
with nonrecessionaiy periods (coefficient = 5.178, p < .01). ative versus positive elasticities. The model includes all sub-
This supports the argument that firms with strong brands and stantive drivers and the manuscript status as predictors.
customer relationships are, to some extent, protected againstWe do not have a theory of how research design charac-
the general downturn financial markets face during an eco-teristics could explain negative elasticities or advertising
nomic downturn (Johansson, Dimofte, and Mazvancheryl overspending, respectively. We also estimated a multino-
2012). Note that the large size of the coefficient does not implymial logit model with a third category of elasticity estimates
that the firm value elasticity is in the one-digit range. To obtain that are not significantly different from zero. This class (n =
the unconditional effect, we have to account for the values 128) represents firms with optimal advertising levels. The fit
of all other moderator variables (for the unconditional effectof these models, in terms of McFadden's R2 and classifi-
sizes, see Table 4). In addition, we find significantly lower cation rates, was not satisfactory at all. This prevents us
marketing asset elasticity for service firms than for other firms from drawing meaningful conclusions. Separating elastic-
(coefficient = -2.569, p < .01). Recall that a negative elasticityities into negative and positive values probably reduces the
implies that firms are overinvested in the marketing asset. Cus-variance and information content so that meaningful insights
tomer acquisition and retention are of utmost importance toin our moderators cannot be obtained.
service firms and determine their success in the marketplace. Outlier-robust analyses. We reanalyzed both meta-
As a result, intense competition might have driven investmentsanalytic models using a least absolute deviation (LAD)
in marketing assets over their profit maximum. The findingestimator. This approach is based on a median regression
by Anderson, Fornell, and Rust (1997) that the return-on- and is thus less affected by outlying observations (Greene
investment elasticity for customer satisfaction is lower for 2012; for an application of LAD in a meta-analysis, see
service firms than for product firms supports our argument. Smith and Huang 1995). The Web Appendix presents the
We obtain five significant effects with respect to researchweighted LAD results for the two models including all
design characteristics. Consistent with the AEM and H6b, observations as well as weighted least squares (WLS) re-
elasticities are significantly higher in level than in returnsults for comparison purposes. Note that neither LAD nor
models (coefficient for market capitalization: 1.453, p <WLS account for the hierarchical error structure. If outliers
.01; coefficient for intangibles-to-tangibles ratio: .389, p < were a severe problem, WLS results should substantially
.10). With respect to the inclusion of relevant firm-specificdiffer from LAD results. However, we find the signs of
variables in the model specification, we find the expected the coefficients to be generally consistent across both es-
positive omitted-variable bias for earnings (coefficient = timation techniques. Most importantly, we do not find re-
.277, p < .10). This result reinforces the call by Jacobson versals in signs for significant effects. Therefore, outliers
and Mizik (2009) to always include a measure of accoun-do not seem to bias the estimation results. The similarity of
ting profitability in models when relating firm value to LAD and HLM, which accounts for the correlation of error
marketing assets. In addition, elasticities are significantly
lower if time intervals of more than a month are used in
the model (coefficient = -.642, p < .10). This suggests a7We thank a reviewer for this suggestion.
530 JOURNAL OF MARKETING RESEARCH, AUGUST 201 6
Table 5
HLM RESULTS
A EM (n = 269) MAM (n = 178)

Expected Estimate Expectation Expected Estimate Expectation
Variable Level Sign (SE) Supported Sign (SE) Supported
Intercept - .084 (.080) -.267 (.651)
Substantive Drivers
Marketing Influence
Type of marketing asset variable Brand-related asset N.A. N.A. Base
Customer-related asset N.A. N.A. + 1.180 (.444)*** V
Product and market conditions
Product type Across product types Base Base
Only durables -/+ -.035 (.037) N.A. N.A.
Only nondurables -/+ -.047 (.029) N.A. N.A.
Only services -/+ -.009 (.030) -/+ -2.569 (.707)***
Geographic region United States Base Base
Other -/+ -.130 (.092) -/+ -.601 (.546)
Concentration Low Base N.A. N.A.
Medium to high - -.063 (.029)** N.A. N.A.
Time Mean year of data period -/+ .002 (.002) + -.086 (.058) n.s.
Recession Months of recession + .009 (.059) n.s. + 5.178 (1.377)***
Interaction Effect
Customer-related asset x Time N.A. N.A. + .143 (.083)** v*
Research Design Characteristics
Data Characteristics
Type of firm value variable Stock return Base Base
Market capitalization + .102 (.062)* ť + 1.453 (.297)*** V
Intangibles to tangibles + .032 (.018)** f + .389 (.298)* *
Temporal interval Up to one month Base Base
Longer than one month -/+ -.013 (.038) -/+ -.642 (.383)*
Structure of data Time series (pure and Base Base
panel)
Purely cross-sectional -/+ .044 (.023)* -/+ -.995 (.464)**
Model and Estimation
Characteristics
Endogeneity Accounted for Base Base
Not accounted for + .056 (.026)** + .029 (.136) n.s.
Heterogeneity Accounted for Base Base
Not accounted for -/+ .004 (.017) -/+ .013 (.206)
Estimation method Ordinary least squares Base Base
Generalized least squares -/+ .034 (.018)* -/+ .013 (.143)
Other -/+ .038 (.022)* -/+ .076 (.294)
Functional form Additive Base Base
Multiplicative -/+ -.026 (.042) -/+ .116 (.248)
Other -/+ -.009 (.041) -/+ -.059 (.268)
Duration of the effect Short-term Base Base
Long-term -/+ .028 (.017) -/+ .644 (.430)
Control for Other Firm Influences
Earnings variable Included Base Base
Omitted -/+ .016 (.033) + .277 (.194)*
Market-share variable Included Base Base
Omitted + -.038 (.027) n.s. + .219 (.324) n.s.
Growth variable Included Base Base
Omitted + -.002 (.025) n.s. + .100 (.142) n.s.
R&D variable Included Base Base
Omitted -/+ -.029 (.047) -/+ -.069 (.265)
Leverage variable Included Base Base
Omitted -/+ -.004 (.024) -/+ .162 (.545)
Size variable Included Base Base
Omitted -/+ .000 (.016) -/+ -.079 (.267)
Competition variable Included Base Base
Omitted -/+ -.015 (.020) -/+ .047 (.265)
Risk Accounted for Base Base
Not accounted for -/+ .009 (.047) -/+ .227 (.239)
Publication-Related Factors
Manuscript status Published Base Base
Unpublished - .046 (.074) n.s. - -.122 (.484) n.s.
*p < .10.
**p < .05.
***p < .01.
Notes: n.s. = not significant (p > .10); N.A. = not applicable. Observations were weighted by the absolute value of the ratio of elasticity and standard error. We
used a one-sided t-test if sign prediction was possible (+ or -) and a two-sided t-test otherwise (-/+). Expected sign: + = positive relationship (compared with base
level); - = negative relationship; -/+ = ambiguous relationship.
terms within studies, strong mediating

further position in the
supports value robustnes
the chain, management
main-model results (see Table
is well advised to5).
focus on optimizing these assets when
Method-corrected effect settingsizes
advertising. Following
budgets, for example.Albers,
Notably, this isM
trala, and Sridhar (2010), we "correct"
not limited each
to advertising but embraces the elasticit
entire marketin
mix. Marketing
surement for the statistically assets are also built method
significant from investments biaint
accounting for endogeneitythe distribution network, product quality,
in advertising modelsand so on. an
ting an earnings variable Ourin analysis
marketing reveals a higher firm value
asset elasticity of
models).
tain a mean method bias-corrected
customer-related assets compared elasticity of as
with brand-related -
advertising expenditures, sets. Moreover,is
which thenot
gap seems to increase over time. The-
significantly dif
from zero (p > . 10), and oretically,
of .42 this (p < supports
finding .01) the for marketing
existence of a hierarchica
Further robustness and effect structure among assets,
collinearity checks.which has Webeen suggested
perf
several additional robustness
in previouschecks.
research (e.g., First, we Here,
Stahl et al. 2012). check the
multicollinearity. Maximumbrand is considered a means to win new
variance customers and
inflation facto
8.513 in the AEM andtransform
6.716 theminintothe
satisfiedMAM indicat
and loyal customers. The
moderate levels of multicollinearity. However,
brand signals future growth opportunities for the firm, b
several combinations of variables exist with a bivariate but these also come with higher risk. In contrast, a loyal
correlation greater than 1.501 in both models, we deleted customer base promises less volatile cash flows for the
each of the affected variables one at a time to assess the future. Investors seem to value the better predictability
robustness of the results and found no substantial differ- of financial performance from customer metrics (Himme
ences (details of these results are available on request). and Fischer 2014).
Second, we lose a considerable number of observations Managing marketing assets. Marketing departments are
because they are missing information about statistical sig- under ever-increasing pressure to show the value relevance
nificance, so we imputed the mean- and median-normed of their marketing investments to maintain their influence
variance from the reduced sample as weights for all caseswithin the firm (Verhoef and Leeflang 2009). The large
for which this information is missing. The results, which average elasticity of .54 for marketing assets implies that
are available from the authors, correspond very closely to brand and customer assets are generally not yet at their
the main-model results. optimal levels. This is good news for marketing managers
Third, we pool short- and long-term elasticities in our for at least three reasons. First, it emphasizes the value
meta-analysis, which, according to Albers, Mantrala, and relevance of marketing. Second, it suggests that there is still
Sridhar (2010, p. 841), is "not meaningful when carryoverroom for further marketing investments to drive firm value.
effects ... are heterogeneous across study settings." Therefore,Finally, it offers a direct link to daily marketing practice
we performed an additional analysis using only short-term because managing marketing assets is well known to mar-
elasticities (AEM: n = 244; MAM: n = 167). The results, keters and more actionable than managing shareholder re-
which can be obtained from the authors on request, areturns. The emerging mindset metrics literature supports these
similar to the results from the models that include long- findings (e.g., Hanssens et al. 2014).
term elasticities and account for the long-term character Managing advertising expenditures. At first glance, the
with the use of a dummy variable. average advertising expenditure elasticity of .04 might
suggest that advertising does not contribute much to firm
DISCUSSION
value. The wide dispersion of elasticities below and above
Substantive Implications for Managers and Researchers zero, however, implies the opposite. Advertising is a valu-
able activity. Some firms seem to overinvest, others to un-
Marketing value chain. Marketing scholars have sug-
gested various chain-of-effects models of how marketing derinvest, but there are also many firms that manage their
advertising expenditures very well with respect to financial
actions contribute to firms' financial performance (e.g.,
Lehmann 2004; Rust et al. 2004; Srivastava, Shervani, objectives.
and Marketers may learn from these firms and their
management tools (see, e.g., Bayer's budget allocation
Fahey 1998). Our meta-analysis enables us to generalize
approach in Fischer et al. 2011).
the productivity chain on empirical grounds. We find
Competition and economic recession. The structure of
support for the hierarchy of effects as proposed in the
competition and the state of the economy are important
literature and our theoretical framework (see Figure 1).
conditions that alter the effectiveness of marketing decision
Marketing-mix decisions such as advertising spending do In more concentrated markets, the effectiveness
making.
translate into financial results for firms that are appreciated
of advertising to drive firm value seems to be lower. Com-
by the stock market. However, not every single adver-
petitive reactivity usually increases when there are fewer com-
tising dollar improves financial performance. Advertising
petitors. The sales response literature offers opposing results
expenditures need to be successfully convertedon intothe in-
impact of competitive advertising reactivity on sales
termediate performance metrics before they can influence
elasticities. Whereas Gatignon (1984) finds increasing elas-
other financial outcome variables such as sales, profits, andDanaher, Bonfrer, and Dhar (2008) show that elas-
ticities,
firm value. There is also an optimal investment level that
ticities actually decrease. Irrespective of the true sales effect,
maximizes firm value, and firms need to manage this the net effect on the bottom line seems to be lower in more
level.
An understanding of the role of marketing assets in the
concentrated markets. Managers should consider this in their
value chain helps in managing marketing activities. Brand
decision making to avoid a potential overinvestment.
and customer assets are important intermediate outcomeOur analysis also adds to the understanding of the ef-
variables that directly move firm value. Because of fecttheir
of recessionary periods on the marketing-performance
Marketingabout
relationship. There is an ongoing debate asset elasticities
whether are significantly
firms lower if tem-
should increase, decrease, or maintain their
poral intervals level
longer thanof adver-
a month are used, probably
because the stock
tising during tough economic conditions. market reacts Lilien,
Srinivasan, within a shorter period of
and Sridhar (2011) show that some time. We recommend
firms benefit usingfrom
more disaggregated
in- data, if
available,
creasing advertising during recessions such as daily
in terms oforprofits
weekly data, andthat more closely
stock returns, whereas others do capture investors' reaction
not. Similarly, time, Heerde
Van as some previous studies
et al. (2013) and Steenkamp and investigating
Fang (2011) marketing identify
assets have done op- (Luo, Raithel, and
Wiles 2013;on
posing effects of economic contractions Tirunillai
the and Tellis 2012).
advertising
Type of dependent
sales effectiveness in the United Kingdom andvariable. The finding that both ad-
the United
vertising expenditure
States, respectively. The nonsignificant effectand inmarketing
the AEM asset elasticities tend
to be higher if firms
reflects these mixed findings. Nevertheless, firm valueareis measured
advised in levels (e.g., market
capitalization, Tobin' s prosperous
to invest in marketing assets in economically q) instead of changes such as stock
returns provides
periods. Shareholder returns to marketing supportare
assets for Mizik
greaterand Jacobson (2009).
during an economic downturn when They consumer
argue that working in levels, and in
confidence isparticular with
Tobin' sconditions,
low. Ou et al. (2014) show that in such q, is not advisable. Models with market capitali-
consumers
regard a value-for-money advantage zation as the dependent
(value equity) variable
andmay the
suffer from severe
autocorrelation
high credibility of strong brands (brand problems.
equity) asThe measurement of the denomi-
important
drivers of their loyalty intentions. nator
Thus, in Tobin's q (asset replacement
stronger assets help value) is prone to mea-
surement errors.
firms retain customers and thus attenuate We follow their
the negative suggestion to use stock
financial
return as the dependent
consequences of recessions. An alternative firm value variable.
view suggests that Note that although
firms tend to be underinvested inwe could not include assets
marketing event studies in our analysis,
during a the depen-
dent variable
recession. The managerial implication, here also reflects
however, is the changes in shareholder value.
same.
Increasing the asset in better times, Inclusion
when of control
more variables. We find only one signif-
financial
resources are available, is advisableicant omitted-variable
given bias with respect to the earnings
that marketing
assets are quite sticky. variable in the MAM. However, we suggest including an
earnings-related
Implications for structural modelers. Ourvariable
analysisin all marketing
also firm value models
because the detection
carries an important message for structural of value relevance
modelers. A key implies a signif-
assumption of many structural models icant effectis that
that is not reflectedbehave
firms in contemporaneous ac-
optimally (i.e., they maximize cashcounting
profit performance
or the (Jacobson
net present and Mizik 2009).
Accounting
value of cash flows, respectively; e.g., for endogeneity. Ifet
Chintagunta simultaneity
al. between firm
2006). With regard to advertising, value
this andassumption
advertising expenditures
is well is not
in accounted for in
model specification,
line for the average firm under average market elasticities are biased upward. Therefore,
conditions
across industries over the past 40 years.should
researchers The mean
control method
for such potential reverse-causality
effects
bias-corrected advertising elasticity in advertising
does not differ firm value models (e.g., by applying
signifi-
cantly from zero. The distribution instrumental
of advertisingvariables estimation
elasticities techniques,
in by specifying
Figure 3, however, demonstrates structural
that many models that may need to account
elasticities de-for nonoptimal
firm
viate substantially from zero. Thus, behavior;do
firms see not
previous discussion).
set optimal
advertising budgets in these cases,Finally, we note that
assuming our study
they want is consistent
to with several
maximize shareholder value. The situation is even more other meta-analyses in marketing in finding that the majority of
severe with respect to brand and customer assets. Here, even
potential determinants are insignificant. We agree with Farley,
Lehmann, and Sawyer (1995) that this is a reassuring pattern
the average elasticity, which is significantly different from
zero, implies that firms do not behave optimally. of robustness rather than something to worry about. We also
concur with Sethuraman, Tellis, and Briesch's (201 1) opinion
Given that a structural model usually focuses on a specific
market, these findings cast serious doubts on one of the key
that nonsignificant effects in a meta-analysis do not imply that
subsequent studies should ignore these factors. For example,
assumptions of these models. We suggest that structural mod-
elers be more open to considering alternative assumptions unobserved heterogeneity remains an important issue in
about firm decision making that better reflect actual firm marketing-finance
be- studies, and a growing number of studies
account for it (e.g., Anderson, Fornell, and Mazvancheryl
havior, even though this behavior may not be consistent with
profit maximization. 2004). The same logic applies to the inclusion of risk factors
in marketing firm value models. We advise researchers to
Methodological Implications test as many model specifications and estimators as possible
The insights from our meta-analysis hold important im- to show the robustness of econometric results.
plications for further research in the field of marketing-
LIMITATIONS AND DIRECTIONS FOR
finance. The subsequent discussion can be interpreted as a
FURTHER RESEARCH
roadmap for researchers. We conclude that the following
decisions on research design are critical: (1) the temporal This study has some limitations offering fruitful avenues
for further research. First, we were not able to include
aggregation level of data, (2) the type of dependent variable
elasticities from all available marketing-finance interface
used, (3) the inclusion of control variables, (4) and whether
to account for endogeneity. studies because some of them did not provide the necessary
Temporal aggregation. According to the efficient mar- information to calculate elasticities. The exclusion of these
kets hypothesis (Fama 1970), the stock market reacts com- elasticities may attenuate the generalizability of our results
to some extent. We follow Albers (2012) in recommending
pletely and instantly to all publicly available information.
Bird, M., C. Channon,

that authors report dimensionless and A.S.C.
elasticities Ehren
in additi
unstandardized regressionand Brand Usage," Journal of Marketin
coefficients.
Second, like any other 307-14.
meta-analysis, our study r
on partially subjectiveBlattberg,
data Robert C. and John Deighton
coding. (1996), Manage
The Market-
provisio
ing by the makes
the meta-analytic database Customer Equity Test,"
our Harvard Business Review deci
coding ,
transparent. 74 (July/August), 136-44.
Third, it might be worthwhile to study additional mod- Capon, Noel, John U. Farley, and Scott Hoenig (1990), "De-
erators. However, we are limited by the actual occurrence terminants of Financial Performance: A Meta-Analysis," Man-
or measurability of study characteristics (Farley, Lehmann, agement Science , 36 (10), 1143-59.
and Sawyer 1995). For example, the question of whether Chintagunta, Pradeep, Tiilin Erdem, Peter E. Rossi, and Michel
advertising expenditures and marketing assets are more Wedel (2006), "Structural Modeling in Marketing: Review and
value relevant for young versus mature companies cannot Assessment," Marketing Science , 25 (6), 604-16.
be answered, because an average firm/brand age is given in Conchar, Margy P., Melvin R. Crask, and George M. Zinkhan
only a very small number of studies. Future studies should (2005), "Market Valuation Models of the Effect of Advertising
put more focus on the heterogeneity in firm value effects and Promotional Spending: A Review and Meta-Analysis,"
between different industries and firms. Journal of the Academy of Marketing Science , 33 (4), 445-60.
Finally, we included only elasticities pertaining to the Danaher, Peter J., André Bonfrer, and Sanjay Dhar (2008), "The
Effect of Competitive Advertising Interference on Sales for
marketing-mix variable advertising in our meta-analytic
model. The meta-analyses on price (mean elasticity of -2.62 Packaged Goods," Journal of Marketing Research, 45 (April),
211-25.
in Bijmolt, Van Heerde, and Pieters [2005]) and personal
Deleersnyder, Barbara, Marnik G. Dekimpe, Jan-Benedict E.M.
selling (mean elasticity of .34 in Albers, Mantrala, and
Steenkamp, and Peter S.H. Leeflang (2009), "The Role of Na-
Sridhar [2010]) imply a much higher sales response effect of
tional Culture in Advertising's Sensitivity to Business Cycles,"
these instruments. Given that recent studies (e.g., Srinivasan,
Vanhuele, and Pau weis 2010) have shown that the effects Journal of Marketing, 46 (October), 623-36.
Eisend, Martin and Farid Tarrahi (2014), "Meta-Analysis Selection
of price and distribution on brand equity metrics are also
Bias in Marketing Research," International Journal of Research
significantly higher than the effect of advertising, it is es-
in Marketing , 31 (3), 317-26.
pecially surprising that these variables have received so little
Fama, Eugene F. (1970), "Efficient Capital Markets: A Review
attention in marketing-finance studies. For instance, future
of Theory and Empirical Work," Journal of Finance , 25 (2),
studies could assess the shareholder value effect of skimming 383^17.
versus penetration strategies or whether the breadth of a dis-
Farley, John U., Donald R. Lehmann, and Alan Sawyer (199
tribution channel (exclusive vs. intensive distribution) has a
"Empirical Marketing Generalization Using Meta-Analysi
stock market impact.
Marketing Science , 14 (3), G36-46.
Fischer, Marc, Sönke Albers, Nils Wagner, and Monika F
REFERENCES (201 1), "Dynamic Marketing Budget Allocation Across Countri
Products, and Marketing Activities," Marketing Science , 30 (
Albers, Sönke (2012), "Optimizable and Implementable Aggregate
568-85.
Response Modeling for Marketing Decision Support," Inter-
Ganesan, Shankar, ed. (2012), Handbook of Marketing and Finance .
national Journal of Research in Marketing , 29 (2), 111-22.
Cheltenham, UK: Edward Elgar.
Selling Elasticities: A Meta- Analysis," Gatignon, Hubert (1984), "Competition
Journal as a Moderator of the Ef-
of Marketing
Research , 47 (October), 840-53. fect of Advertising on Sales," Journal of Marketing Research ,
Anderson, Eugene W., Claes Fornell, 21 (November),
and Sanai 387-98. K. Mazvancheryl
(2004), "Customer Satisfaction and Geyskens, Inge, Rekha Krishnan,
Shareholder Value," Jan-Benedict
JournalE.M. Steenkamp,
of Marketing, 68 (October), 172-85. and Paulo V. Cunha (2009), "A Review and Evaluation of Meta-
Analysis Practices in Management Research," Journal of Man-
agement, 35 (2),
faction, Productivity, and 393-419. Profitability
Greene, William
Goods and Services," H. (2012), Econometric Analysis ,Science
Marketing 7th ed. New
York: Prentice Hall.
Assmus, Gert, John U. Farley, and Don
"How Advertising Hanssens,
Affects Dominique M. andSales:
Marnik G. Dekimpe (2008), "Models
Meta- An
for theMarketing
Results," Journal of Financial-Performance Effects of Marketing,"
Research in Handbook
of Marketing Decision Models,
Barth, Mary E., Michael B. Berend Wieringa, ed. New York:
Clement, G
Kasznik (1998), Springer, 501-23. Values
"Brand and Capit
Review of Accounting Studies , 3 (1/2), 4
Bezawada, Ram andResponse
Koen Models: Econometric
Pauwels and Time
(2013),
Marketing Organic Boston: Kluwer Academic Publishers.
Products? How Organ
and Promotions Drive Retailer Performa
and Gokhan
keting , 77 (January), 31-51. Yildirim (2014), "Con
Bijmolt, Tammo H.A. Guiding and Marketing
Rik PietersMix Decisions
(20
Marketing 534-50. when
Studies Contain Mul
Marketing Letters , 12 Hoffmann, Arvid(2),
O.I., Thomas Post, and Joost M.E. Pennings
157-69.
(2013), "Individual Investor Perceptions and Behavior During
Empirical Generalizations on thethe Financial Crisis," Journal
Determinants of of Price
Banking and Finance, 37 (1),
Elas-
60-74.
ticity," Journal of Marketing Research , 42 (May), 141-56.
Himme, Alexander and Marc Fischer (2014), "Drivers of the Cost of

"Return
Capital: The Joint Role of Non-Financial onInternational
Metrics," Marketing: Using Custom
Journal of Research in Marketing , 31 (2), 224-38.
keting Strategy," Journal of Marketin
Schulze, ,Christian,
Hirschey, Mark (2009), Managerial Economics Bernd Skiera, an
12th ed. Mason,
OH: Cengage Learning. "Linking Customer and Financial Metr
Huedo-Medina, Tania B., Julio Sánchez-Meca, Fulgencio
The Leverage Marín-
Effect in Customer Bas
Marketing,
Martínez, and Juan Botella (2006), "Assessing 76 (March),
Heterogeneity in 17-32.
Meta-Analysis: Q Statistic or I2 Index?" Psychological
Sethuraman, Methods
Raj, Gerard, J. Tellis, and
11 (2), 193-206. "How Well Does Advertising Work? Ge
Jacobson, Robert and Natalie Mizik (2009), "Customer
Analysis Satisfaction-
of Brand Advertising Elastici
Based Mispricing: Issues and Misconceptions,"
ing Research,
Marketing
48 (June),
Sci- 457-71.
ence , 28 (5), 836-45. Shah, Denish, Roland T. Rust, A. Parasur
Johansson, Johny K., Claudiu V. Dimofte, and Sanai K. George S. Day (2006), "The Path to
Mazvancheryl (2012), "The Performance of Global Brands in Journal of Service Research, 9 (2), 1
the 2008 Financial Crisis: A Test of Two Brand Value Measures," Smith, V. Kerry and Ju-Chin Huang (19
International Journal of Research in Marketing , 29 (3), Quality? A Meta-Analysis of Hedonic
235-45. Journal of Political Economy, 103 (1)
Joshi, Amit and Dominique M. Hanssens (2010), "The Direct and
Sorescu, Alina B. and Jelena Spanjol (2
Indirect Effects of Advertising Spending on Firm Value," Journal on Firm Value and Risk: Insights fr
of Marketing, 74 (January), 20-33. Goods," Journal of Marketing, 72 (M
Krasnikov, Alexander and Satish Jayachandran (2008), "The Rel-Srinivasan, Raji, Gary L. Lilien, and
ative Impact of Marketing, Research-and-Development, and Opera- "Should Firms Spend More on Resear
tions Capabilities on Firm Performance," Journal of Marketing, Advertising During Recessions?" Journ
72 (July), 1-11. 49-65.
Kremer, Sara T.M., Tammo H.A. Bijmolt, Peter S.H. Leeflang, and Srinivasan, Shuba and Dominique M. Hanssens (2009), "Marketing
Jaap E. Wieringa (2008), "Generalizations on the Effectiveness ofand Firm Value: Metrics, Methods, Findings, and Future Di-
Pharmaceutical Promotional Expenditures," International Jour- rections," Journal of Marketing Research, 46 (June), 293-312.
nal of Research in Marketing , 25 (4), 234-46.
Lehmann, Donald R. (2004), "Metrics for Marketing Matter," Metrics in Market Response Mod
Journal of Marketing, 68 (October), 73-75. proach," Journal of Marketing Rese
Lu, Jane W. and Paul W. Beamish (2004), "International Di- Srivastava, Rajenda K., Tasadduq A.
versification and Firm Performance: The S-Curve Hypothesis," (1998), "Market-Based Assets and Sh
The Academy of Management Journal , 47 (4), 598-609. work for Analysis," Journal of M
Luo, Xueming, Sascha Raithel, and Michael A. Wiles (2013), "The2-18.
Impact of Brand Rating Dispersion on Firm Value," Journal of Stahl, Florian, Mark Heitmann, Donald R. Lehmann, and Scott A.
Marketing Research , 50 (June), 399-415. Neslin (2012), 'The Impact of Brand Equity on Customer Ac-
Mizik, Natalie and Robert Jacobson (2009), "Financial Markets quisition, Retention, and Profit Margin," Journal of Marketing,
Research in Marketing," in "Commentaries and Rejoinder to 76 (July), 44-63.
'Marketing and Firm Value: Metrics, Methods, Findings and Steenkamp, Jan-Benedict E.M. and Eric (Er) Fang (2011), "The
Future Directions,'" Journal of Marketing Research , 46 (June), Impact of Economic Contractions on the Effectiveness of R&D
320-24. and Advertising: Evidence from U.S. Companies Spanning Three
Morgan, Neil (2012), "Marketing and Business Performance,"Decades," Marketing Science, 30 (4), 628-45.
Journal of the Academy of Marketing Science , 40 (1), 102-19.Tirunillai, Seshadri and Gerard J. Tellis (2012), "Does Chatter
Ou, Yi-Chun, Lisette de Vries, Thorsten Wiesel, and Peter C. Really Matter? Dynamics of User-Generated Content and Stock
Verhoef (2014), "The Role of Consumer Confidence in Creating Performance," Marketing Science, 31 (2), 198-215.
Customer Loyalty," Journal of Service Research , 17 (3), 339-54. U.S. Department of Justice (2015), "Herfindahl-Hirschman Index,"
Rappaport, Alfred (1998), Creating Shareholder Value: A Guide for (accessed February 14, 2015), [available at http://www.justice.
Managers and Investors , 2nd ed. New York: The Free Press. gov/atr/public/guidelines/hhi.html] .
Reinartz, Werner, Jacquelyn S. Thomas, and V. Kumar (2005), Van Heerde, Harald J., Maarten J. Gijsenberg, Marnik G. Dekimpe,
"Balancing Acquisition and Retention Resources to Maximize and Jan-Benedict E.M. Steenkamp (2013), "Price and Advertising
Customer Profitability," Journal of Marketing , 69 (January), Effectiveness over the Business Cycle," Journal of Marketing
63-79. Research, 50 (April), 177-93.
Verhoef, Peter C. and Peter S.H. Leeflang (2009), "Understanding
Rust, Roland T., Tim Ambler, Gregory S. Carpenter, V. Kumar, and
Rajendra Srivastava (2004), "Measuring Marketing Productivity: the Marketing Department's Influence Within the Firm," Journal
Current Knowledge and Future Directions," Journal of Mar- of Marketing, 73 (March), 14-37.
keting , 68 (October), 76-89. Yoo, Boonghee, Naveen Donthu, and Sungho Lee (2000), "An
Examination of Selected Marketing Mix Elements and Brand
Equity," Journal
mating Publication Bias in Meta-Analysis," of the Academy
Journal of Marketing Science, 28 (2),
of Marketing
Research , 27 (May), 220-26. 195-211.

Marketing Impact On Firm Value

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Marketing Impact On Firm Value

Uploaded by

Copyright:

Available Formats

Marketing's Impact on Firm Value: Generalizations from a Meta-Analysis

Author(s): ALEXANDER EDELING and MARC FISCHER

The interest in the value relevance of marketing investments has given

Keywords: meta-analysis, marketing firm value elasticity, empirical mar-

Online Supplement : http://dx.doi.org/10.1509/jmr.14.0046

Marketing's Impact on Firm Value:

© 2016, American Marketing Association Journal of Marketing Research

they express what has been learned services

Marketing (Mix) Marketing Firm

• Communication Valuation Metrics Accounting Results

,prïce _ -Brand value «Top line (revenues)

Perception Metrics Firm Value

• Brand perception . Market capitalization

described as tangible Srinivasanorand Hanssens

(e.g., mean brand value ratings) and

ļ Marketing Influence Interactions Product and Market Conditions

ļ • Advertising expenditures vs. • Type of marketing asset x Time «Product type

1 Data Characteristics Model and Estimation Control for Other Firm

information (e.g., 10% increase in costs that are I directly

ililiJ, ? Cfl u C ^ O lili rt 2 > llilllUiíllll 3pjC<u2c«n^: - ñ «1 60 M n lllllllliilïl SOa-SaöSS^Sa^ il C0Ö0

ill sã«« 1 S 111 .1

1 Cx, 11 &Û •« siili Öß <u ?? lili SS ^ T3

brand metrics such

elasticities from marketthe capitalization and

***/? < .01 (two-sided tests).

because the market the means (medians) considers

elasticities with n = 192. Mean and median elasticity are .54

A: Advertising Expenditure Elasticities

B : Marketing Asset Elasticities

Notes: The percentages in Panel A sum to 100.1% as a result of rounding error.

variable) amounts to .876 in the AEM and of

A EM (n = 269) MAM (n = 178)

terms within studies, strong mediating

Bird, M., C. Channon,

Himme, Alexander and Marc Fischer (2014), "Drivers of the Cost of

You might also like