Vous êtes sur la page 1sur 8

The Economic Worth of Celebrity Endorsers: An Event Study Analysis

Author(s): Jagdish Agrawal and Wagner A. Kamakura


Source: Journal of Marketing, Vol. 59, No. 3 (Jul., 1995), pp. 56-62
Published by: American Marketing Association
Stable URL: http://www.jstor.org/stable/1252119 .
Accessed: 14/09/2014 06:39
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .
http://www.jstor.org/page/info/about/policies/terms.jsp
.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact support@jstor.org.
.
American Marketing Association is collaborating with JSTOR to digitize, preserve and extend access to
Journal of Marketing.
http://www.jstor.org
This content downloaded from 111.68.103.164 on Sun, 14 Sep 2014 06:39:43 AM
All use subject to JSTOR Terms and Conditions
Jagdish Agrawal
&
Wagner
A. Kamakura
T he Economic Worth of
Celebrity
Endorsers: An Ev ent
S tudy Analysis
Celebrity
endorsement has become a
prev alent
f orm of
adv ertising
;n the United S tates.
Despite
extensiv e litera-
ture on the ef f ects of
celebrity
endorsements on consumers' brand attitudes and
purchase intentions,
little is known
about the economic v alue of these endorsements. Research on this
topic
has
typically
f ocused on theories ex-
plaining
how
celebrity
endorsements inf luence consumers' attitudes and intentions. T he authors assess the
impact
of
celebrity
endorsement contracts on the
expected prof itability
of a f irm
by using
ev ent
study methodology.
T heir
approach
assumes that the announcement of a
celebrity
endorsement
contract, usually widely publicized
in the
business
press,
is used as inf ormation
by
market
analysts
to ev aluate the
potential prof itability
of endorsement ex-
penditures, thereby af f ecting
the f irm's
expected
return. Announcements of 110
celebrity
endorsement contracts
were
analyzed.
Results indicate
that,
on
av erage,
the
impact
of these announcements on stock returns is
positiv e
and
suggest
that
celebrity
endorsement contracts are
generally
v iewed as a worthwhile inv estment in
adv ertising.
C elebrity
endorsement has become a
prev alent
f orm of
adv ertising
in the United S tates.
According
to indus-
try sources, approximately
20% of all telev ision com-
mercials f eature a f amous
person,
and
approximately
10%
of the dollars
spent
on telev ision
adv ertising
are used in
celebrity
endorsement adv ertisements
(Adv ertising Age
1987;
S herman
1985).
T he number and dollar v alue of
celebrity
endorsement contracts are
increasing,
and these
contracts constitute a
signif icant
and
growing portion
of ad-
v ertising budgets (McGill 1989).
S ev eral studies hav e examined consumers'
response
to
celebrity
endorsements in
adv ertising. Findings
show that
celebrities make adv ertisements believ able
(Kamins
et al.
1989)
and enhance
message
recall
(Friedman
and Friedman
1979). Furthermore,
celebrities aid in the
recognition
of
brand names
(Petty, Cacioppo,
and S chumann
1983),
create
a
positiv e
attitude towards the brand
(Kamins
et al.
1989),
and create a distinct
personality
f or the endorsed brand
(Mc-
Cracken
1989). Ultimately, celebrity
endorsements are be-
liev ed to
generate
a
greater
likelihood of customers' choos-
ing
the endorsed brand
(Heath, McCarthy,
and Mothers-
baugh
1994;
Kahle and Homer
1985;
Kamins et al.
1989;
Ohnian
1991). T hus,
the use of
celebrity
endorsements is an
adv ertising strategy
that should enhance the
marginal
v alue
of adv ertisement
expenditures
and create brand
equity by
means of the
"secondary
association" of a
celebrity
with a
brand
(Keller 1993).
Celebrity
endorsements are
expensiv e
f or the f irm. De-
pending
on the status of the
celebrity,
remuneration could
Jagdish Agrawal
is an assistant
prof essor,
S chool of Business and Eco-
nomics,
Calif ornia S tate
Univ ersity, Hayward. Wagner
A. Kamakura is an
associate
prof essor,
Katz Graduate S chool of Business, Univ ersity
of
Pittsburgh.
T he authors are
gratef ul
to three
anonymous
rev iewers f or their
helpf ul
comments and to Zaf ar
Iqbal
f or his able assistance on this
research.
run into millions of dollars f or sev eral
years.
A contract
may
also include a
prof it sharing plan.
Firms of ten build
special
and
costly adv ertising campaigns
around celebrities. For ex-
ample,
Coca Cola Co.
reportedly spent
$25 million in an ad-
v ertising campaign
with Bill
Cosby
as its
spokesperson
f or
Coke
(Adv ertising Age 1986). S imilarly,
IBM
spent
$40 mil-
lion in an
adv ertising campaign inv olv ing
MAS H actors
(Reuters 1987). Ov erall,
the use of celebrities as
spokesper-
sons in adv ertisements constitutes a
signif icant
inv estment
in
intangible
assets
by
the
sponsoring
f irm-an inv estment
that
management hopes
to of f set with
greater
f uture sales
rev enues and
prof its.
T he economic v alue of
strategic marketing
decisions is
receiv ing growing
interest in
marketing
literature. A number
of studies hav e examined the ef f ect of
strategic marketing
decisions such as
product
innov ation,
research and
dev elop-
ment
expenditures,
adv ertisement
expenditures, product
quality,
and consumer satisf action on f irm
prof itability
(Capon, Farley,
and
Hoenig
1990; S zymanski, Bharadwaj,
and
Varadarajan 1993).
Because a
growing
number of f irms
are
inv esting
in
celebrity
endorsement to enhance the v alue
of
adv ertising
dollars and build brand
equity,
a natural
ques-
tion is: What are the economic returns f rom the inv estment
in this f orm of
adv ertising?
Our
purpose
is to assess the
prof itability
of the
celebrity
endorsement
adv ertising strategy.
Prev ious studies on the
impact
of
celebrity
endorsements hav e
prov ided
v aluable in-
sights
into consumers'
responses
to adv ertisements that
por-
tray
celebrities. Our
premise
is that it is
important
f or a man-
ager
to know whether a
celebrity
endorsement
strategy gen-
erates a f av orable consumer
response,
as well as understand
whether the economic returns
justif y
the costs associated
with such a
strategy.
Measuring
the ov erall ef f ect of
adv ertising expenditures
on sales is
problematic,
and a direct assessment of the ef -
f ectiv eness of a
celebrity
endorsement on a f irm's
prof itabil-
ity may
be
impossible.
In
addition,
because
adv ertising
ef -
Journal of
Marketing
Vol. 59
(July 1995),
56-62
56 / Journal of
Marketing, July
1995
This content downloaded from 111.68.103.164 on Sun, 14 Sep 2014 06:39:43 AM
All use subject to JSTOR Terms and Conditions
f ects accrue ov er
time,
current
prof its may
not
accurately
re-
f lect the true
prof itability
of a
giv en campaign.
T he dif f icul-
ty
associated with
isolating
and
measuring
the
prof it
associ-
ated with a
giv en celebrity
endorsement
campaign
led us to
use an alternativ e measure. T his is the
expected prof it
asso-
ciated with a
celebrity
endorsement
campaign
as ref lected in
the abnormal returns of a f irm. Abnormal returns are mea-
sured
using
ev ent
study methodology. Although
a f irm's
prof it
is inf luenced
by
sev eral
f actors,
and
isolating
the con-
tribution of
any
one v ariable is
dif f icult,
the ev ent
study
methodology prov ides
a means and
unique opportunity
to
assess the
impact
of a
particular strategy
on a f irm's
expect-
ed f uture
prof its.
Ev ent
S tudy Methodology
Ev ent
study methodology
measures the
magnitude
of the ef -
f ect that an
unanticipated
ev ent has on the
expected prof -
itability
and risk of a
portf olio
of f irms associated with that
ev ent. T he
theory underlying
ev ent
study methodology
is the
ef f icient market
hypothesis (Fama
et al.
1969). According
to
this
theory,
the
price
of a
security
is the
present
v alue of f u-
ture cash f lows
expected
f rom a f irm's assets
and,
at
any
giv en
time,
ref lects all the av ailable inf ormation about the
f irm's current and f uture
prof it potential.
If
any
new inf or-
mation
resulting
f rom an
unexpected
ev ent is believ ed to af -
f ect a f irm's current and f uture
earnings,
the
security price
changes
as soon as the market learns of the ev ent.
T heref ore,
stock
prices
are v iewed as reliable indicators of a f irm's
v alue. T he amount of
change
in the
price
of a
security
af ter
an
ev ent,
relativ e to its
pre-ev ent price,
would ref lect the
market's unbiased estimate of the economic v alue of that
ev ent
(Brown
and Warner
1985).
T o examine whether an
ev ent had
any impact
on the f irm's
v alue,
abnormal return,
which is the
change
in stock
price
af ter it has been
adjusted
f or
changes resulting
f rom
general
market
mov ements,
is
measured.
T his
methodology
is well
accepted
and has been
widely
used in a
v ariety
of
disciplines,
such as
f inance, accounting,
law, organizational
behav ior,
and business
strategy.
More re-
cently,
the
methodology
has also been
applied
to assess the
impact
of some
marketing-related
ev ents on a f irm's
prof -
itability;
such ev ents
include,
f or
example,
a
change
in a
company's
name
(Horsky
and
S wyngedouw 1987),
new
product
introduction
(Chaney, DeVinney,
and Winer
1991),
brand
lev eraging (Lane
and Jacobson
1995), product
recalls
(Jarrell
and Peltzman
1985),
and
regulations
and
rulings
on
f alse
adv ertising (Peltzman 1981).
Details of ev ent
study
methodology
are
widely
av ailable in the literature
(Horsky
and
S wyngedouw
1987;
S chwert
1981).
We also
present
a
summary
of this
methodology
in the
Appendix.
Our rationale f or
applying
ev ent
study methodology
to
celebrity
endorsements is:
Although
it
may
be
impossible
to
measure the direct
impact
of a
celebrity
endorsement on a
f irm's f uture
prof its,
we can
inv estigate
whether the decision
to incur this
adv ertising expenditure
is v iewed as wise
by
in-
v estors. Because of the substantial costs
usually
associated
with
celebrity
endorsements,
we
argue
that a f irm's contrac-
tual
agreement
with a
celebrity
is a
major
ev ent with
poten-
tial f inancial
implications.
S uch contracts are
f ormally
an-
nounced
by
f irms and
usually
receiv e wide
cov erage
in the
media.
T heref ore,
when an announcement is made about a
celebrity
endorsement
contract,
inv estors make
independent
judgments
on the f uture
prof it impact
of the contract. T hese
judgments
are then
immediately
ref lected in the f irm's stock
returns.
T heref ore, measuring
the abnormal returns of f irms
announcing
endorsement contracts enables us to examine
the market's v aluation of the net economic worth of celebri-
ty
endorsements.
Results
Data
We def ine the ev ent
day
as the date when the announcement
of a contract or a
f orthcoming
contract between a f irm and a
celebrity
f irst
appeared
in the
print
media. We consulted
major newspapers
f or announcements of contracts between
January
1980 and December 1992. We also conducted a
thorough
search of other
newspaper
and newswire serv ices
in Lexis/Nexis data f iles to search f or
any leakage
of f orth-
coming
contracts,
and other
f irm-specif ic
ev ents f rom f iv e
days
bef ore to one
day
af ter the announcements of the con-
tracts. Our search led to the identif ication of 207 cases of en-
dorsements f or which announcement dates were av ailable.
Ninety-sev en
cases were excluded f rom the
analysis
because
they
inv olv ed
priv ately
held
companies
or
f oreign
f irms f or
which stock data were unav ailable.
Hence,
our f inal
sample
size was 110 observ ations. For most of the cases
(83%),
the
ev ent
day
is the date when the news about an actual contract
appeared
in the
print
media. We f ound no other
prior
news
of a
f orthcoming
contract f or these cases. For the
remaining
cases,
there was
prior
inf ormation f rom the
print
media of
f orthcoming
contracts;
and f or
these,
the ev ent
day
is the
f irst
day
when the news of the
f orthcoming
contract became
public through print
media. T he
sample
of 110 cases
repre-
sents announcements of
celebrity
endorsement
by
35 f irms
inv olv ing
87 celebrities. Each announcement contained
pri-
marily
inf ormation
regarding
the names of celebrities and
the
sponsoring
f irm.
Only
a f raction of the announcements
(31%)
also contained inf ormation on the contract
v alue,
and
eight
cases were associated with the endorsement of new
brands. Data on stock returns were collected f rom the f iles
of the Center f or Research in
S ecurity
Prices at the Univ er-
sity
of
Chicago.
Abnormal Returns Due to Announcement of
Celebrity
Contracts
Following
Brown and Warer's (1985) study,
we estimated
the
parameters
of the market model
(Equation
3 in the
Ap-
pendixl)
f or each f irm
by regressing
its actual returns on the
returns of a
weighted portf olio
of securities
using
an esti-
mation
period
of 239
days (t
-
244 to t
- 6
days
relativ e to
the ev ent
day,
t
=
0).
T he estimated market model
parame-
ters
(a
and
B3)
were then used in
Equation
4 to estimate ab-
normal returns to the
portf olio
of 35 f irms f or 110 an-
nouncements on and around the ev ent
day. Figures
1 and 2
1All
equations
mentioned
subsequently
can be f ound in the
Ap-
pendix.
Celebrity
Endorsers/ 57
This content downloaded from 111.68.103.164 on Sun, 14 Sep 2014 06:39:43 AM
All use subject to JSTOR Terms and Conditions
FIGURE1
Frequency
Distribution
Abnormal Returns
92
20
------------------------------------ -- -- --
u.
l)
4 15 -------------- ------------------
0
m
z
5-- ---------------------
o ,
I .A? . .. ll. .. l.. .1
-3.75 -3.25 -2.75 -2.25 -1.75 -1.25 -.75 -.25 .25 .75 1.25 1.75 2.25
ABNORMAL RET URN
2.75 3.25 3.75 4.25 4.75 5.25 5.75 6.25
FIGURE2
Frequency
Distribution
S tandard Abnormal Returns
-
mol-
ii i i i i i i i i i i i ii i i i i i i i i i i i
. .*
:::: ::::::::::::::::::1:1
-3.60 -3.20 -2.80 -2.40 -2.00 -1.60 -1.20 -.80 -.40 0.00 .40 .80 1.20 1.60 2.00 2.40 2.80
S T ANDARDIZED ABNORMAL RET URN
present
the distribution of abnormal returns
(Equation 4)
and standardized abnormal returns
(Equation 5), respectiv e-
ly
on the ev ent
day
f or all 110 ev ents.
T able 1
presents
the
av erage
abnormal returns f or the
110 announcements on the ev ent
day,
as well as f or a win-
dow of ? 10
days
around the ev ent
day.
Results show
that,
on
av erage,
announcements of
celebrity
endorsement con-
tracts are associated with
positiv e
excess returns. T he av er-
age
abnormal return f or the ev ent
day
is
.44%,
which is sta-
tistically signif icant (t
=
2.39, p
<
.05).2
T his
percentage rep-
resents the
largest gain
in excess returns ov er a
period
of
2We used the Center f or Research in
S ecurity
Prices market
v alue
weighted portf olio
of all NYS Eand AMEX stocks toserv e
as the
proxy
f or the market
portf olio
when
estimating
each f irm's
market model. S imilar results tothose we
report
here were ob-
tained when the
equally weighted
market
portf olio
was used. T he
results at the ev ent
day
were:
Av erage
abnormal return
=
.53%;
T -
testl
=
2.74;
and T -test2
=
2.92.
58 / Journal of
Marketing, July
1995
30
25 -
co
m
20-
4
0
U.
O 15-
1J
z
10
-
5
0
.
- - - - - - - -
.
::-- 1::1::: --~--------------
, . . .
~~~~. . I I
This content downloaded from 111.68.103.164 on Sun, 14 Sep 2014 06:39:43 AM
All use subject to JSTOR Terms and Conditions
T ABLE1
Excess Returns of 110 Announcements of
Celebrity
Endorsements
Av erage Cumulativ e %of +
Ev ent Abnormal
Av erage
Abnormal Abnormal
Day
Return
(%)
T -S tatisticl T -S tatistic2 Return
(%)
Returns
-10 .26 1.28 .93 .26 49.1
-9 .30 1.61 1.98* .56 58.2
-8 -.28 -1.11 -1.01 .28 44.5
-7 -.11 -.36 -1.51 .17 38.2
-6 .02 .55 .11 .19 44.5
-5 -.21 -.84 -.32 -.02 47.3
-4 .15 1.32 .81 .13 54.5
-3 -.00 -.74 -1.07 .13 38.2
-2 -.05 -.23 -.31 .08 44.5
-1 .09 .54 1.42 .17 56.4
0 .44 2.39* 2.63* * .61 57.3*
+1 -.21 -.96 -.55 .40 45.5
+2 -.26 -1.51 -1.69 .14 41.8
+3 -.02 .30 .47 .12 50.0
+4 .04 .33 -.01 .16 45.5
+5 .20 .89 .04 .36 52.7
+6 .19 .97 .18 .55 53.6
+7 -.32 -1.47 -1.70 .23 34.5
+8 -.17 -.87 -.37 .06 55.5
+9 -.02 .00 -.09 .04 46.4
+10 .18 .74 -.12 .22 43.6
T -S tatisticl = Based on the
parametric
test in
Equation
6.
T -S tatistic2 = Based on Corrado's
(1989) nonparametric
test.
* * p <= .01.
* p < =.05.
? 10
days
around the ev ent
day.
T he results indicate
that,
on
av erage,
f irms
announcing
contracts with celebrities
experi-
enced a
gain
of .44% in excess returns.3 T his estimate has
the same order of
magnitude
as abnormal returns
reported
in
other
marketing-related
ev ent studies. For
example, Horsky
and
S wyngedouw (1985) report
.64%
positiv e
excess re-
turns on
av erage
in
response
tothe
changes
in a f irm's
name;
Chaney, DeVinney,
and Winer
(1991) report
.25% abnormal
return with the announcement of new
products;
and more re-
cently,
Lane and Jacobson
(1995) report
a .32% abnormal
return due tothe announcement of brand
lev eraging.
We
perf ormed
an additional test to f urther examine the
signif icance
of our results. T he distribution of excess returns
on the ev ent
day, presented
in
Figures
1 and
2, depict
a dis-
tribution that is skewed
slightly
to the
right.
T his skewed
distribution
suggests
that the
assumption
of a normal distri-
bution
implicit
in the t-test in
Equation
7
might
be v iolated.
T ocircumv ent this
problem,
Corrado
(1989)
has
suggested
an alternativ e
nonparametric
rank test. T his test uses
only
ordinal inf ormation about the returns on the ev ent
day,
rather
than inf ormation on the
magnitude
of returns
(Corrado
1989).
As a
result,
the test statistic is not inf luenced
by
the
3For nine
cases,
there was
leakage
of news
regarding "negotia-
tions". Because such
negotiations
hav e f ailed in the
past
on some
occasions
(f or example,
Madonna f or LA Gear and
Meryl S treep
f or American
Express),
we
ignored
such rumors.
Howev er,
consid-
ering
the date of
publication
of rumors as the ev ent
day produced
v ery
similar results. Results f or the ev ent
day
were:
Av erage
ab-
normal return =
.38, T -testl =
2.33,
and T -test2 =
2.42;
and CAR
(Cumulativ e Av erage Return)
f or -1 to0
days
was .41, T -testl
=
1.86.
v ariance in the distribution of returns on the ev ent
day.
As
indicated in Column 4 of T able
1,
the t-v alue of the ev ent
day
estimates deriv ed f rom this test is
2.63,
which is statis-
tically signif icant.
T his result f urther conf irms the
signif i-
cance of excess returns on the ev ent
day4.
T he last column
presents
the
percentage
of the 110 abnormal
returns,
which
are
strictly positiv e
f or each
day, showing
that 57.3% of all
cases had a
positiv e
abnormal return on the ev ent
day.
T his
proportion
was
signif icantly (at p
<
.05) larger
than the av -
erage percentage
of
positiv e
returns observ ed ov er the esti-
mation
period (47.0%).
Cumulativ e Abnormal Returns
It is standard
practice
in an ev ent
study
to examine the cu-
mulativ e excess returns f or v arious windows
surrounding
the ev ent
day
f or tworeasons.
First,
analyzing
abnormal re-
turns
surrounding
the ev ent
day
allows f or
uncertainty
re-
garding
the actual date of the ev ent.
S econd,
it allows the re-
searcher to
capture
the cumulativ e ef f ect of an
ev ent,
be-
cause the ef f ect
may
be
spread
ov er sev eral
days
surround-
ing
the ev ent
day
because of the
gradual av ailability
of in-
f ormation and
interpretation
of the ev ent's
impact
on f uture
f irm
prof itability.
T able 2
presents
the cumulativ e
av erage
returns
(CAR)
f or v arious windows
surrounding
the ev ent
day.
Of all win-
dows
considered, only
the one f rom -1 to0
days
showed a
signif icant
CAR,
with a v alue of .54%. T he
signif icance
of
4Figures
1 and 2 show some of the abnormal returns as
potential
outliers. We
replicated
our
analyses
af ter
excluding
outliers and ar-
riv ed at
essentially
the same conclusions
reported
in the text.
Celebrity
Endorsers/ 59
This content downloaded from 111.68.103.164 on Sun, 14 Sep 2014 06:39:43 AM
All use subject to JSTOR Terms and Conditions
T ABLE2
Cumulativ e Abnormal Returns f or Windows
S urrounding
the Ev ent
Day
Cumulativ e
Av erage
T ime Abnormal
Interv al Return
(%)
T -statistic
-10;
+10 .23 .61
-10;
-2 .08 .49
-5;
+5 .18 .44
-5;
-2-.1 -. -24
-2; 0 .48 1.54
-1;
0 .54 2.04*
0;
+1 .24 .99
-1;
+1 .33 1.12
+1;
+5 -.24 -.42
+1;
+10 -.39 -.50
-1;
+10 .15 .38
* S tatistically signif icant
at the .05 lev el. We
computed
the T -statistics
using
Brown and Warner's
(1985) Equation A.6.
CAR f rom -1 to 0
days
is
plausible because,
f or some
ev ents,
there is a
lag
of at least one
day
between the an-
nouncement of a contract and
av ailability
of inf ormation
through print
media. As a
result,
the
positiv e
abnormal re-
turn on
Day
-1
may
ref lect the
impact
of those announce-
ments that were made on
Day -1,
but were not
printed by
the
media until
Day
0. In
summary,
ev idence of
signif icant pos-
itiv e returns on the ev ent
day
and the cumulativ e
av erage
re-
turn ov er two
days (including
the ev ent
day) suggest that,
on
av erage,
the market
quickly
reacts
positiv ely
to announce-
ments of
celebrity
endorsement contracts in
adv ertising.
T he
absence of
signif icant
cumulativ e ef f ects
beyond
the
2-day
period
around the ev ent is consistent with the ef f icient mar-
ket
assumption implicit
in ev ent
study methodology (Fama
et al.
1969).
T o test whether our results were caused
by hav ing
a sub-
stantial
subsample
associated with sof t drinks
(37.3%)
and
athletic shoes
(22.7%),
and
using
a substantial
proportion
of
sports
celebrities
(55.5%),
we
perf ormed
a
regression analy-
sis. We
regressed
the standardized abnormal returns on the
ev ent
day
on a set of
dummy v ariables, identif ying
the an-
nouncements associated with sof t drinks
(S of t-dummy)
or
athletic shoes
(S hoe-dummy),
and
sports celebrity
endorsers
(S port-dummy).
None of the v ariables had
any signif icant
impact
on abnormal returns
(F3,105
=
1.57; p
<
.20).
Our re-
sults
suggest
that ev ent
study
results are not inf luenced
by
the inclusion of a
large portion
of cases f rom the sof t drink
or athletic shoe industries. Nor was ev idence f ound that the
market
assigns sports
celebrities more or less v alue than it
does
non-sports
celebrities.
Conclusions and Extensions
Our
analyses
indicate that the market reacted both
positiv e-
ly
and
negativ ely
to the announcement of dif f erent
celebrity
endorsement contracts,
which is ref lected in both
positiv e
and
negativ e
abnormal returns on the ev ent
day.
Howev er,
on
av erage,
inv estors seem to v alue
positiv ely
the use of
celebrities in adv ertisements. A
signif icant percentage
of
positiv e
abnormal returns to the
sponsoring
f irms was ob-
serv ed on the ev ent
day;
on
av erage,
the
sponsoring
f irms
recorded a
gain
of .44% excess returns in their market v alue
as a result of
announcing
contracts with
celebrity
endorsers.
Looking
at sev eral "windows" around the ev ent
day,
we
f ound that the interv al between one
day
bef ore the ev ent and
the ev ent
day
was the
only period
with
signif icant
cumula-
tiv e abnormal returns.
Ov erall,
these results
clearly
indicate
a
positiv e impact
of
celebrity
endorsements on
expected
f u-
ture
prof its,
which lends
objectiv e,
market lev el
support
to
the use of
celebrity
endorsers in
adv ertising.
Recent
reports
in the business
press suggest
that there
are
decreasing
returns associated with
using
celebrities in
adv ertising.
For
example,
the costs associated with
celebrity
endorsements are
rising;
some celebrities endorse sev eral
products,
sometimes ev en
switching
their endorsements to
riv al
brands;
the
negativ e publicity generated by
some
celebrities has added the
potential
risk of
negativ e impact;
and
surv eys
of consumers' reactions to
product
endorse-
ments rev eal that
only
a f raction of consumers react
posi-
tiv ely
to endorsements.
Howev er,
the
widespread
and
per-
sistent use of celebrities in
adv ertising suggests
that market-
ing managers
continue to believ e that
celebrity
endorse-
ments are a worthwhile
component
of the
adv ertising
strate-
gy, despite
the costs inv olv ed. Our results
suggest
that this
belief is
generally
shared
by
inv estors as well. T he
positiv e
av erage
abnormal returns f ound in this
study
ref lect the mar-
ket's
general
belief that the
expected
incremental
gain
f rom
celebrity
endorsements exceeds the incremental costs of ad-
v ertising
due to such contracts.
Because stockholders are one of the
major
stakeholders
in the
f irm,
shareholder v alue
analysis
has been adv ocated
and
practiced (Arzac 1986; Day
and
Fahey 1990; Rappaport
1986)
as a means of
assessing
the f inancial
consequences
of
strategic
decisions. Recent studies show a
positiv e
associa-
tion between abnormal returns and inv estments in
intangible
assets,
such as innov ativ eness
(Chaney, Dev inney,
and
Winer
1991)
and brand
quality (Aaker
and Robertson
1994).
Our results
prov ide
additional
empirical
ev idence that in-
v estors not
only pay
close attention to
marketing strategies
but also react to them. Our results also
prov ide
f urther
sup-
port
f or the need f or shareholder v alue
analysis
in
conjunc-
tion with
analysis
of
marketing strategies.
Our results also
may
of f er some
insights
to f irms con-
sidering
the use of
celebrity
endorsers. T o
date, managers
considering
the
hiring
of
celebrity
endorsers f or
adv ertising
could
only
turn to results of consumer
experiments
and sur-
v eys
f or
guidance.
Our
results,
based on a
large sample
of
celebrity
endorsement contracts,
prov ide
new ev idence that
celebrity
endorsements are
generally
v iewed as a
prof itable
adv ertising strategy.
Howev er,
measurement of abnormal re-
turs is
only
the f irst
step
in
assessing
the economic v alue of
celebrity
endorsement. A
manager,
when
considering
the
use of
celebrity
endorsements,
must
identif y
the
appropriate
celebrities who will
potentially
enhance the v alue of inv est-
ing
in
adv ertising.
Prev ious studies on the ef f ectiv eness of
celebrity
endorsements
prov ide
v aluable
insights regarding
the
celebrity
characteristics that consumers v iew
positiv ely.
Further research could
surv ey
inv estments
analysts regard-
60 / Journal of
Marketing, July
1995
This content downloaded from 111.68.103.164 on Sun, 14 Sep 2014 06:39:43 AM
All use subject to JSTOR Terms and Conditions
ing
the characteristics of celebrities and circumstances when
celebrity
endorsements are v iewed either more
positiv ely
or
negativ ely.
Results of such
surv eys may prov ide
additional
inf ormation f or the selection of
appropriate
celebrities.
Celebrity
endorsement contracts are not
only
on the rise
but the
av erage compensation paid
to celebrities is also in-
creasing. Reports
f rom the annual
surv eys
of Forbes
(Lane
1994) magazine
rev eal that
many
celebrities earn f ar more
money
f rom their endorsement contracts than f rom their
usual f ields of endeav or. In
addition,
the abnormal returns
reported
here
actually suggest that,
on
av erage,
celebrities
are worth more than the costs of
hiring
them. T hese results
should be of interest to
agencies
that
represent
celebrities in
contract
negotiation.
T he results of our
study,
as well as other recent market-
ing-related
ev ent
studies,
should
prov ide impetus
f or
apply-
ing
this
methodology
to other areas.
Managers
of ten
prean-
nounce their
marketing
decisions
through
interv iews or
press
releases. T heir decisions
may
be associated
with,
f or
example,
the introduction of new
products,
a
change
in ad-
v ertising agencies,
or the
dev elopment
of new
segments
or
markets.
Apart
f rom
competitors,
consumers,
and distribu-
tors,
these announcements also
prov ide
inf ormation to in-
v estors
regarding
the f irm's commitment and intentions
(El-
ishberg
and Robertson
1988; Horsky
and
S wyngedouw
1987).
Inv estors' reactions to such
announcements,
ref lected
in
positiv e
or
negativ e
abnormal returns,
serv e as f eedback
to
managers regarding
the inv estors' assessments of their
decisions
(Rappaport 1987).
On the basis of interv iews with
top
executiv es,
Webster
(1981)
identif ies the lack of under-
standing
about the f inancial
implications
of
marketing
deci-
sions as one of the
major
issues
f acing marketing manage-
ment. Ours and other ev ent studies show that ev ent
study
methodology prov ides unique opportunities
f or
marketing
managers
to assess the ramif ications of their decision on the
f irm's
prof itability.
Finally,
a cav eat must be
placed
on this and other stud-
ies that hav e
employed
ev ent
study methodology.
We hav e
attributed the excess return obtained at the date of the con-
tract announcement to inv estors'
expectations regarding
the
f uture
prof itability
of the
celebrity's
endorsement. T he im-
plicit assumption
here is that inv estors make inf erences
about the net
present
v alue of the f irm's inv estment in the
celebrity
contract and that the outcome of this inf erence
by
inv estors will be ref lected in the f irm's stock
price.
In other
words,
we assume that the announcement of an endorsement
contract is
interpreted directly
as inf ormation about the
f irm's f uture use of the
celebrity
in
adv ertising.
T his as-
sumption
is based on the
f indings
of
prev ious
studies that
demonstrate the
potential
ef f ect of
celebrity
endorsement on
consumer attitudes and intentions.
T he announcement of such contracts could also send
other
signals
to f inancial markets that
might
not be
directly
related to the usef ulness of the
celebrity
in
adv ertising
(Klein
and Lef f ler 1981;
Nelson
1974).
For
example,
the an-
nouncement of a contract could act as a
signal
to inv estors
of the conf idence a f irm has in the
superiority
of its
products
or of its commitment to
particular
brands. In
contrast,
we as-
sume that inv estors use these announcements as inf orma-
tion:
T hey
consider the costs and
potential
benef its of the
celebrity
endorsement and react to the announcement ac-
cordingly. Obv iously,
as in
any
ev ent
study, v erif ying
whether the observ ed abnormal returns attributed to each an-
nouncement are due to inv estors' v aluation of the
celebrity
endorsements or their
response
to some unrelated
signal
is
v irtually impossible,
because it would
require
identif ication
of all the hidden
signals implied by
each ev ent.
Howev er,
at
the
aggregate lev el,
whether f irms use announcements of
celebrity
endorsements to
signal
other intentions
(e.g.,
en-
tering
a new market
segment, increasing
adv ertisement ex-
penditures)
or whether the market
interprets
those an-
nouncements as hidden
signals
can be assessed
by surv eys
of
managers
and inv estment
analysts.
T hus,
Eliashberg
and
Robertson's
(1988) study
on inf ormational and
signalling
v alues of new
product preannouncements
could
prov ide
a
basis f or f urther research in this area.
Appendix
A
S ummary
of the Ev ent
S tudy Methodology
T he
price
of a
capital
asset i at time t
(Pit)
is the
present
v alue of the f uture
expected
cash f lows f rom the asset
(1) Pit=
Zk
=
1,dit + k/(l
+
rit+ k)k
where
dit
+
k
is the
expected
cash f low f rom asset i in
period
t +
k,
and
ri
+
k
is the discount rate of the cash f lows ac-
counting
f or its estimated riskiness.
T he
impact
of an ev ent on the f irm's
potential prof itabil-
ity
can be measured
by comparing
the stock return at the
ev ent
day:
(2) Rit
=
(Pit
+
dit
-
Pit- )/Pit- 1,
to the "normal return" that would be
expected
if the ev ent
did not take
place.
T his "abnormal" return measures the
change
in the stock v alue due to an ev ent.
Among
v arious
models,
the one used most of ten f or es-
timating expected
or "normal" returns is the market model
(Fama 1970). According
to the market model,
the normal re-
turn to asset i at time t can be
expressed
as a linear f unction
of the returns f rom a
portf olio
of all marketed assets
Rmt,
(3) Rit
=
aoi
+
iRMt
+
eit.
T he abnormal return
eit
f or a
security
i at the date t of an
ev ent
can, thus,
be obtained as
(4) eit
=
Rit
-
,oi
-
iRmt.
T he
commonly accepted hypothesis
of ef f icient markets
implies
that the dev iation
eit
is a random v ariable with mean
equal
to
0,
because the dev iations between the actual returns
to asset i and their
expected
v alues,
conditional on all av ail-
able inf ormation at time t -
1,
should not be
systematically
dif f erent f rom 0. T o test if abnormal returns resulted because
of an
ev ent,
test the
hypothesis
that the cross-sectional mean
of
eit
at the ev ent
day
is dif f erent f rom 0. T he
hypothesis
is
tested as f ollows.
Estimate the coef f icients
oi
and
3i
f rom
Equation
3
through
a linear
regression
ov er an estimation
period
of T
days (f or example,
-244 to -6
days
relativ e to ev ent
day
=
0)
Celebrity
Endorsers / 61
This content downloaded from 111.68.103.164 on Sun, 14 Sep 2014 06:39:43 AM
All use subject to JSTOR Terms and Conditions
f or each f irm i
announcing
a
celebrity
endorsement contract.
T he standardized abnormal return f or the ev ent
day
t
=
0 is
(5) 6i0
=
eio/S i,
where
S i
is the standard dev iation of the
regression
residuals
that were obtained
prior
tothe contract announcement.
Because the ev ent
study
is conducted ov er
multiple
ev ents observ ed f or sev eral f irms f or v arious calendar
dates,
the ef f ect of a
particular type
of ev ent
(e.g., celebrity
en-
dorsement
contracts)
can be tested
by
f irst
computing
the
av erage
of standardized abnormal returns ov er all observ ed
ev ents:
(6) Et
=
i =
,n Eit/N
where N is the number of ev ents
being
studied
among
mul-
tiple
f irms and t
represents
the ev ent
day (which
f alls on dif -
f erent calendar
days
f or dif f erent
ev ents).
T otest whether the
av erage
abnormal return is dif f erent f rom 0 on the ev ent
day
t =
0,
use the test
statistic,
(7) Zt
=
i= 1,n Eit/N,
which is distributed unit normal f or
large
N.
REFERENCES
Aaker,
Dav id A. and Robert Jacobson
(1994),
"T he Financial In-
f ormation Content of Perceiv ed
Quality,"
Journal
of Marketing
Research,
31
(May),
191-201.
Adv ertising Age (1986),
"E. F. Hutton's
S pokesman
Idea a 'Cos'
Celebre"
(April 21),
1
(1987),
"Adwhirl"
(July 13),
88.
Arzac, Enrique
R.
(1986),
"DoYour Business Units Create S hare-
holder Value?" Harv ard Business Rev iew,
64
(January/Febru-
ary),
121-126.
Brown, S tephen
J. and Jerold B. Warner
(1985), "Using Daily
S tock Returns: T he Case of Ev ent S tudies,"
Journal
of
Finan-
cial Economics,
14
(March),
3-31.
Capon,
Noel,
John U.
Farley,
and S cott
Hoening (1990),
"Deter-
minants of Financial Perf ormance: A
Meta-Analysis," Manage-
ment S cience,
36
(October),
1143-59.
Chaney,
Paul
K., T imothy
M.
Dev inney,
and Russell Winer
(1991),
"T he
Impact
of New Product Introductions on the Market Value
of Firms," Journal
of Business,
64
(4),
573-610.
Corrado,
Charles J.
(1989)
"A
Nonparametric
T est f or Abnormal
S ecurity-Price
Perf ormance in Ev ent S tudies," Journal
of
Fi-
nancial Economics, 23
(August),
385-95
Day, George
S . and Liam
Fahey (1990), "Putting S trategy
into
S hareholder Value
Analysis,"
Harv ard Business Rev iew,
68
(March/April),
156-62.
Eliashberg,
Jeshoshua and T homas S . Robertson
(1988),
"New
Product Preannouncement Behav ior: A Market S ignalling
S tudy,"
Journal
of Marketing
Research,
25
(August),
282-92.
Fama, Eugene
F.
(1970),
"Ef f icient
Capital
Markets: A Rev iew of
T heory
and
Empirical
Work," Journal
of
Finance,
25
(2),
383-417.
,
Lawrence Fisher,
Michael C. Jensen,
and Richard Roll
(1969),
"T he
Adjustment
of S tock Prices toNew Inf ormation,"
International Economic Rev iew,
10
(1),
1-21.
Friedman, Hershey
H. and Linda Friedman (1979),
"Endorser Ef -
f ectiv eness
by
Product
T ype,"
Journal
of Adv ertising
Research,
19
(October),
63-71.
Heath, T imothy B.,
Michael S .
McCarthy,
and Dav id L. Mothers-
baugh (1994), "S pokesperson
Fame and Viv idness Ef f ects in
the Context of Issue-Relev ant
T hinking:
T he
Moderating
Role
of
Competitiv e S etting,"
Journal
of
Consumer Research,
20
(March),
520-34.
Horsky,
Dan and Patrick
S wyngedouw (1987),
"Does It
Pay
to
Change
Your
Company's
Name? A S tock Market
Perspectiv e,"
Marketing S cience,
6
(Fall),
320-34.
Jarrell, Gregg
and S am Peltzman
(1985),
"T he
Impact
of Product
Recalls on the Wealth of
S ellers,"
Journal
of
Political
Economy,
93
(3),
512-36.
Kahle, Lynn
B. and Pamela M. Homer
(1985), "Physical
Attrac-
tiv eness of
Celebrity
Endorsers: A S ocial
Adaptation Perspec-
tiv e," Journal
of
Consumer Research,
11
(March),
954-61.
Kamins,
Michael
A.,
Meribeth J.
Brand,
S tuwart A.
Hoeke,
and
John C. Moe
(1989),
"T wo-S ided Versus One-S ided
Celebrity
Endorsements: T he
Impact
on
Adv ertising
Ef f ectiv eness and
Credibility,"
Journal
of Adv ertising,
18
(2),
4-10.
Keller,
Kev in Lane
(1993), "Conceptualizing, Measuring,
and
Managing
Customer-Based Brand
Equity,"
Journal
of
Market-
ing,
57
(January),
1-22.
Klein,
B. and K. B. Lef f ler
(1981),
"T he Role of Market Forces in
Assuring
Contractual Perf ormance,"
Journal
of
Political Econ-
omy,
89
(4),
615-41.
Lane,
Randall
(1994),
"T he Forbes
All-S tars," Forbes, (December
19),
266-78.
Lane,
Vicki and Robert Jacobson
(1995)
"S tock Market Reactions
toBrand Extension Announcements: T he Ef f ects of Brand Atti-
tude and
Familiarity,"
Journal
of Marketing,
59
(January)
63-77.
McCracken,
Grant
(1989),
"Whois the
Celebrity
Endorser? Cul-
tural Foundations of the Endorsement Process,"
Journal
of
Consumer Research,
16
(December),
310-21.
McGill, Douglas
C.
(1989),
"S tar Wars in Cola
Adv ertising,"
New
York T imes, (March 22),
D1.
Nelson, Philip (1974), "Adv ertising
as Inf ormation,"
Journal
of
Po-
litical
Economy,
82
(July/August),
729-54.
Ohnian,
Roobina
(1991),
"T he
Impact
of
Celebrity S pokespersons'
Perceiv ed
Image
on Consumers' Intention toPurchase," Jour-
nal
of Adv ertising
Research,
31
(February/March),
46-54.
Peltzman,
S am
(1981),
"T he Ef f ects of FT C
Adv ertising Regula-
tion,"
Journal
of
Law and Economics,
24
(December),
40348.
Petty,
Richard E.,
John T .
Cacioppo,
and Dav id S chumann
(1983),
"Central and
Peripheral
Routes to
Adv ertising
Ef f ectiv eness:
T he
Moderating
Role of Inv olv ement,"
Journal
of
Consumer
Research,
10
(S eptember),
135-46.
Rappaport,
Alf red
(1986), Creating
S hareholder Value: T he New
S tandardf or
Business
Perf ormance.
New York: T he Free Press.
(1987)
"S tock Market
S ignals
to
Managers,"
Harv ard
Business Rev iew, (Nov ember/December),
57-62.
Reuters, (1987), (April 3).
S chwert,
G. William
(1981), "Using
Financial Data toMeasure Ef -
f ects of
Regulation,"
Journal
of
Law and Economics,
26
(April),
121-58.
S herman,
S tratf ord P.
(1985),
"When You Wish
Upon
a S tar,"
For-
tune, (August 19),
68.
S zymanski,
Dav id M.,
S under G.
Bharadwaj,
and P.
Rajan
Varadarajan (1993),
"An
Analysis
of the Market S hare-Prof -
itability Relationship,"
Journal
of Marketing,
57
(July),
1-18.
Webster,
Frederick E.
(1981), "T op Management's
Concerns about
Marketing:
Issues f or the 1980's," Journal
of Marketing,
45
(S ummer),
9-16.
62 / Journal of
Marketing, July
1995
This content downloaded from 111.68.103.164 on Sun, 14 Sep 2014 06:39:43 AM
All use subject to JSTOR Terms and Conditions

Vous aimerez peut-être aussi