Do Innovations Really Pay Off?

Total Stock Market Returns to Innovation

Author(s): Ashish Sood and Gerard J. Tellis
Source: Marketing Science, Vol. 28, No. 3 (May-June 2009), pp. 442-456
Published by: INFORMS
Stable URL: http://www.jstor.org/stable/23884198
Accessed: 28-11-2016 23:43 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
http://about.jstor.org/terms

INFORMS is collaborating with JSTOR to digitize, preserve and extend access to Marketing Science

This content downloaded from 130.133.8.114 on Mon, 28 Nov 2016 23:43:39 UTC
All use subject to http://about.jstor.org/terms
Marketing Science
Vol. 28, No. 3, May-June 2009, pp. 442^56
issN 0732-23991 eissn 1526-548X1091280310442 doi 10.1287/mksc.l080.0407
©2009 INFORMS

Do Innovations Really Pay Off?

Total Stock Market Returns to Innovation

Ashish Sood
Goizueta School of Business, Emory University, Atlanta, Georgia 30322,
ashish_sood@bus.emory.edu

Gerard J. Tellis
Marshall School of Business, University of Southern California, Los Angeles, California 90089,
tellis@usc.edu

Critics often decry

long-term ansuch
projects earnings-focused
as innovation toshort-term orientation
boost a firm's ofSuch
stock price. management thatthat
critics assume eschews
stockspending
markets on risky,
react positively to announcements of immediate earnings but negatively to announcements of investments in
innovation that have an uncertain long-term pay off. Contrary to this position, we argue that the market's
true appreciation of innovation can be estimated by assessing the total market returns to the entire innovation
project. We demonstrate this approach via the Fama-French 3-factor model (including Carhart's momentum
factor) on 5,481 announcements from 69 firms in five markets and 19 technologies between 1977 and 2006.
The total market returns to an innovation project are $643 million, more than 13 times the $49 million from
an average innovation event. Returns to negative events are higher in absolute value than those to positive
events. Returns to initiation occur 4.7 years ahead of launch. Returns to development activities are the highest
and those to commercialization the lowest of all activities. Returns to new product launch are the lowest among
all eight events tracked. Returns are higher for smaller firms than larger firms. Returns to the announcing firm
are substantially greater than those to competitors across all stages. We discuss the implications of these results.
Key words: innovation; market returns; event study; Fama-French 3-factor model; high-tech marketing
History: Received: April 25, 2007; accepted: February 29, 2008; processed by Donald R. Lehmann. Published
online in Articles in Advance January 12, 2009.

Introduction some critics assert that an earnings-focused, short

Innovation is probably one of the most important term orientation on boosting stock p
forces in fueling the growth of new products, sustain- investments in innovation that typ
ing incumbents, creating new markets, transforming payoff. Such critics assume that the
industries, and promoting the global competitiveness reward efforts in innovation that typ
of nations. Even so, many researchers, analysts, and to pay off. However, accurately as
managers fear that firms do not invest enough in inno- returns to innovation may be critica
vation. According to the MIT Technology Review's how markets respond to innovati
annual survey of research and development (R&D) in firms to invest in innovation.
2004, corporate R&D spending across a broad cross- The abnormal stock market retu
section of industries is on the decline. Some go so far are one of the best means of as
as to complain that the United States may be losing its wards to innovation. Past research h
competitive edge and its famed leadership in innova- effect of innovation on firm perf
tion because of declining investment in R&D relative like sales, profits, or market sha
to other nations (Hall 1993; Council on Competitive- measures are subject to many other
ness Report 2004). Firms may underinvest in R&D ronmental factors so that the path
because of the high costs, the long delay in reaping clear. Under the assumption that t
market returns if any, the uncertainty of those returns, efficient, such returns can be ass
and the difficulty of adequately measuring them. The study (Fama 1998). The event st
increasing speed of diffusion across global markets abnormal stock market returns to
(Chandrasekaran and Tellis 2008) and the diverse pat- in an event, which is assumed to
terns of consumer adoption across products and coun- the net present value of the new
tries (Sood et al. 2009) exacerbate the challenges for early application of this method, C
firms to predict returns to new products. Moreover, report market returns of 0.25% t
442

This content downloaded from 130.133.8.114 on Mon, 28 Nov 2016 23:43:39 UTC
All use subject to http://about.jstor.org/terms
Sood and Tellis: Do Innovations Really Pay Off? Total Stock Market Returns to Innovation
Marketing Science 28(3), pp. 442-456, ©2009 INFORMS 443

Table
Table 1 Events DuringDuring
1 Events Initiation,Initiation,
Development, and Commercialization
Development, andActivities of Innovation Projects
Commercialization Activities of Inno

Phase Initiation Development Commercialization

Events unique Funding (grants, advanced Prototypes (working prototypes, New product
to this study order, funded contracts) identification of new materials, Launch (shipments, new applications)
Expansion (new development processes or equipment,
or manufacturing facilities) demonstration in exhibitions)

This research (positive and negative events are recorded separately for
announcements of all activities)
Events covered Alliances Patents New product
by prior
prior research
research (joint
(joint ventures,
ventures, acquisitions)
acquisitions) Preannouncemen
Preannounceme
one week ahead of future events) Awards (exte

Prior research Hirschey (1982) Pakes (1985) Eddy and Saunders (1980)
Jaffe (1986) Jaffe (1986) Wooldridge and Snow (1990)
Griliches (1988)
Cockburn and Griliches (1988) Erickson
Erickson and and Jacobson
Jacobson (1992)
(1992) Chaney
Chaney et
et aa
Doukas and Switzer (1992) Kelm et al. (1995) Zantout and Chaganti (1996)
Chan et al. (1992) Hendricks and Singhal (1996)
Hall (1993) Kokuetal. (1997)
Das et al. (1998) Przasnyski and Tai (1999)
Chan et et al.
al. (2001
(2001)) Nicolau
Nicolau and and Sellers
Sellers (2002)
Suárez (2002) Sorescu et al. (2003)
Suarez
Bayus et al. (2003)
Pauwels et al. (2004)
Sorescu et al. (2007)
Tellis and Johnson (2007)

new product introduction. Past research has also

mated returns to other isolated events of an inn
tion project (see Table 1). • What are the market returns to sets of activities
There are three limitations to this approach. First, of the innovation project?
returns to specific events (e.g., launch of new prod- • What structural (e.g., size) and strat
ucts) do not reveal the total returns to innovation, research productivity) variables affect
which is really the sum of all events in an innova- returns to innovation?
tion project. A focus on returns to specific events in • How do the market returns of compe
the innovation project may be one reason why mar- pare to those of the announcing firm?
kets appear to undervalue innovation. Second, a focus The rest of the paper is organized as f
on specific events cannot reveal how returns are dis- next three sections present the theory,
tributed over the entire project. Such knowledge is findings. The last section discusses the fi
useful to both understand which event of an innova- tations, and implications of the research,
tion project gets the most returns and what announce
ment strategy firms should adopt. Third, returns to
specific events may be deflated because of excessive CoriCGptual Background
announcements or inflated because of few announce- This section reviews prior findings and
ments in the innovation project. We can ascertain about markets returns to innovation. To be
this effect only by recording all announcements of the area, it begins by defining the key t
all firms throughout the innovation project and esti- sumptions of the study.
mating returns to an event after controlling for other
events and strategic and structural variables. Definitions
Hence, a researcher may arrive at erroneous esti- We define four key terms: technology,
mates of the true rewards to innovation by limiting project, event, and announcement,
the scope of study to announcements of only a new Following Sood and Tellis (2005), we de
product's introduction or any other single event. As nology as a distinct principle or platform
er as we know, there is no study on market returns ing products to serve a consumer need. F
to all events in an innovation project. This is the goal neon lamps are based on fluorescence techno
of the current study. In particular, it seeks answers to produces light by the distinct scientific p
the following questions: fluorescence. Halogen lamps are based on incandes
• How do stock markets react to each event in an cence technology that produces light by t
innovation project, after controlling for other events? scientific principle of incandescence (see A

This content downloaded from 130.133.8.114 on Mon, 28 Nov 2016 23:43:39 UTC
All use subject to http://about.jstor.org/terms
 Sood and Tellis: Do Innovations Really Pay Off? Total Stock Market Returns to Innovation
444 ' Marketing Science 28(3), pp. 442^156, ©2009 INFORMS

for details). Several new products and models (e.g., patents, and pr
hard disks, floppy drives, tapes, etc.) could be devel- ahead of futu
oped on the platform of one technology (e.g., mag- opment activ
netic storage). cause they alert competitors of progress, reduce the
We define an innovation project as the total of a firm's element of surprise, trigger imi
activities in researching, developing, and introducing sive discounting of the technical co
any new product based on a new technology, from hand, returns to development a
the initiation of the technology to about a year after itive because of reduction in o
introduction of the new product(s). For example, all of naling confidence, competence, an
Philips' research efforts in initiating, developing, and the future (Zantout and Chaganti
commercializing a compact fluorescent lamp (a new et al. 2002, Austin 1993, Pakes
product based on fluorescence technology) comprises 2007). The rival arguments for
the innovation project for that new product. market returns to development acti
We define an event as some progress in the project need for empirical research to reso
(e.g., patents or product launch). We identify seven Commercialization activities incl
such events detailed in later sections. product launch (including launches, initial shipments,
We define an announcement as the availability of in- and new applications), and awards (external recog
formation about an event either from the firm directly nition of quality). Announcements about commercial
or through other sources. ization events may lead to negative returns because
launched products fall below expectations, costs of
Market Returns to Innovation Events, Activities, promotion and launch seem high, or the c
and Projects advantages from launch seem fleeting (Crawford 1977,
We identify three distinct sets of activities in the inno- Berenson and Mohr-Jackson 1994). On the o
vation project—initiation, development, and commer- announcements of commercialization events m
cialization. Each set of activities includes key events to positive returns because they signal t
related to the overall set and may occur any time dur- itiveness of the firm, the successful co
ing the innovation project. For example, firms may innovation project, and the expansion of t
decide to enter into new alliances any time during portfolio (Sharma and Lacey 2004, Chen
the innovation project. Moreover, these events may be Akigbe 2002, Zantout and Changanti 19
either positive (patent registration) or negative (patent et al. 1991, Tellis and Johnson 2007, He
denial) (see Appendix B for details). Total market Singhal 1996, Urban and Hauser 1980, Chan
returns to the entire innovation project are the sum of Sankaranarayanan 2007, Keller and Leh
returns to all activities during the innovation project. The rival arguments for positive and ne
Currently, the literature reports rival findings about ket returns to commercialization activities su
whether returns to each of these events is negative or need for empirical research to resolve the co
positive, as summarized below.
Initiation activities include events about alliances (in- Total Returns to Innovation
eluding joint ventures and acquisitions), funding Past research has estimated returns to isol
(including grants, advanced orders, and funded con- of an innovation project (see Table 1). Th
tracts), and expansions for new innovation projects. may lead to a substantial underestimati
Announcements about initiation activities may lead total returns to innovation. We propose th
to negative returns because of high investments, long returns to innovation can only be estim
gestation periods, associated uncertainty, and high events in all sets of activities of the innova
risk of failure (Crawford 1977, Kelm et al. 1995). are included in the analysis. If the return
On the other hand, such announcements may lead entire innovation project could be estima
to positive returns as they enable market expansion, single, target event during the project, t
deter competitor entry, improve probability of sue- for other events would not be significantly d
cess and enhance firms' competitive position (Aaker from zero. That target event would be cr
1995, Suárez 2002, Anand and Khanna 2000, Das et al. important implications for firms and in
1998, Doukas and Switzer 1992). The rival arguments the other hand, if firms continue to expe
for positive and negative market returns to initiation mental returns to various events over the in
activities suggest the need for empirical research to project, ignoring certain events would resu
resolve the conflict. estimating the total returns to innovation. It would
Development activities include events about proto- also mean that firms (and investors)
types (working prototypes, demonstration in exhibi- attention to all innovation-related
tions, and new materials, equipment, and processes), mize their announcement (and inves

This content downloaded from 130.133.8.114 on Mon, 28 Nov 2016 23:43:39 UTC
All use subject to http://about.jstor.org/terms
Sood and Tellis: Do Innovations Really Pay Off? Total Stock Market Returns to Innovation
Marketing Science 28(3), pp. 442^56, ©2009 INFORMS 445

The total returns to innovation are the sum of returns Size of

to all events in an innovation project. Similarly, if a of firm is
firm has multiple innovation projects running concur- the mar
rently, the total returns to innovation to the firm are gests th
the total return to all innovation projects of the firm. the retu
In addition to completeness, the benefit of consid- of any sing
ering all events in an innovation project is that it com- (Austin
pensâtes for suboptimal or strategic announcements analysts
of the firm. For example, if the firm underpromises in event r
in early stages of an innovation project and overde- size of firm
livers in later stages, the possibly low market returns technolo
in early stages will be compensated by high returns Researc
in later stages Conversely, if a firm overpromises and duct
then underdelivery taking all events into considera- a le of
hon will compensate for possibly too-high returns in
psrlipr ^ 1
6 ' ucts and thus drive up demand for its new innova
Activities with the Highest Returns tlonus <Barney 1986' John
„ , j . , , , . , with a reputation for a regular stream of innovative
Researchers and managers may want to know which , \ , , , . , ,
. r <.• . .. ,. , . . ,.r products increases the likelihood of fruitful strategic
type of activities attracts the highest returns. We are r„. _ „ , , . , , . b
not aware of any specific study that examines this aBianceuS r et aL 1
question or any specific theory that concludes that £e P^abi ity of success w
one particular set of activities does better than oth- Hence, market returns may
ers. However, past research seems to suggest that researcuh Productivity. W
announcements of commercialization activities may tlvlty b7
experience the highest returns for several reasons. year Pnor to the date of
First, only commercialization activities signal culmi- Age of Technology. Ma
nation in terms of revenues from sales of the new projects may differ acro
product (Sharma and Lacey 2004, Chan et al. 1992). Prior research suggests
Second, based on research to date, commercialization with time (Chandy an
activities get the most attention from reporters. Christensen 1992) and th
changes from product to process innovation
Control Variables nology matures (Utterback 1974, Adner and Levinthal
Market returns during the innovation project may 2001). Hence, the improvements in product p
also be affected by the firm's announcement strategy manee might be less for older technologies. I
or structure. For this reason, we include two strategic trast, new technologies improve rapidly, open
variables (announcement frequency and research pro- opportunities and markets, and can disrupt old
ductivity) and two structural variables (size of firm nologies (Christensen 1997). Thus, market retur
and age of technology) as control variables. new technologies may be higher than those to
Announcement Frequency. Firms vary in their an- technologies. We measure age of technology a
nouncement strategy. Microsoft announces all events number of years since the first new product launch
related to the project. Other companies, for example, based on the technology,
like Apple, aggregate many events into one big an
nouncement. Some literature suggests that frequent Method
announcements reflect transparency and timeliness This section describes the method for estima
and thus would either enhance returns or at least abnormal returns to announcements during the
not lead to penalty in returns (Kelm et al. 1995, vation project in five subsections: logic of the ev
Tucker 2007, Givoly and Palmon 1982). Moreover, fre- study, model for data analysis, and sample, sources
quent and multiple announcements lead to dilution of ancj procedure for the data collection,
returns over a larger number of events and thus lower
realized returns per announcement (Chaney et al. Logic of Event Study
1991). We use two alternate measures for announce- The event study (Fama et al. 1969) is one of the
ment frequency: number of prior announcements and widely used analytical tools in financial research. T
days since last announcement. We expect returns to basic assumption underlying the method is the
be negatively correlated to the first measure and pos- cient market hypothesis, which states that a
itively correlated to the second measure. price at a particular point in time fully reflects

This content downloaded from 130.133.8.114 on Mon, 28 Nov 2016 23:43:39 UTC
All use subject to http://about.jstor.org/terms
 Sood and Tellis: Do Innovations Really Pay Off? Total Stock Market Returns to Innovation
446 Marketing Science 28(3), pp. 442-456, ©2009 INFORMS

available information up to that point (Sharpe 1964, one expe

Fama 1998). Thus, any change in the price of a stock fied period
because of arrival of new information reflects the U.S. Treasu
present value of all expected current and future prof- risk-free
its from that new information. The method has been securities h
widely used in the finance, accounting, economics, The retur
management, and marketing literatures to assess the of proje
market value of information contained in various level in the
events of interest. The market return to an event of a We es
firm is the change in the stock price of that firm due an estima
to that event, above that due to the general market announce
at the time of the event. The next subsection explains on the sto
how we compute such market returns to an event. announce
Total returns to innovation are the cumulative returns However,
to all events within an innovation project. mation perio
We next compute abnormal ret
Model event as the difference between the normal returns
We estimate abnormal returns to the event using the that would have occurre
Fama-French 3-factor model (Fama and French 1993) and foe reforns that
including Carhart's momentum factor (Carhart 1997). tdus
Prior studies in event studies have relied on the stan
dard capital asset pricing model that assumes that the ARit = Rit — E[Rit]
market portfolio is the benchmark for normal returns _ R t j_ fi ra/tr
to a stock (McKinlay 1997). However, the Fama-French — " f* ' ' ^l! mt -f- ^2t 1
3-factor model expands the completeness of the model q. + /34iUMDt] for —1 < t < 1, (2)
by adding two more factors: market capitalization and
value. More recently, Carhart proposed the addition where ARit, Rit, and E(Rif) are the abnormal, observed,
of a fourth factor, price momentum, to account for and normal returns, respectively, for announcement i
the persistence effect in returns reported by Jegadeesh and event window t. We also try windows centered on
and Titman (1993). Thus, the combined Fama-French- the date of announcement of varying widths, ±1 and
Momentum 4-factor (FFM4) model is ±2 days before and after the event.
We estimate average abnormal returns and the
t ~~ Rft — ai + fit¡(Rmt ~ Rft) + fin^MB, f-statistic 6 (Brown and Warner 1985) for the portfolio
+1B3iHML, + fi4i UMD, + eit (1) of N announcements of an event; thus,

£[«,]= 0; Var[e„] = (,)., AAR, = i £ AR„, (3)

where !=1

t: Subscri
such that —270 < t < —6; SD(7L4Rf) / 1 ip ~aap)2
i: Subscript for announcement ' i=1 1
R,: Returns to announcement i on day t; where
Rm: Returns to corresponding daily equally AARf is the average (abnormal) returns for a
weighted S&P 500; event; and T0 is the number of days in the estimatio
Rf: Theoretical rate of return attributed to an in- window, which in our case is 270 - 5 = 265, a
vestment with zero risk; Â4R = (i/r0) Dfr°, AAR,.
SMB: Returns on a portfolio of small stocks minus Note that this portfoiio Mest statistic explicitl
returns on large stocks; takes into account any potential cross-sectional
HML: Returns on a portfolio of stocks with high pendence in the abnormal returns.
book-to-market ratio minus the returns to a We compute cumulative average abnormal retur
portfolio of stocks with low book-to-market {CAR¡) in the event window as follows:
ratio;
UMD: Carhart's price-momentum factor that cap- (zi2
CAR =^DAR 15)
tures one-year momentum m returns. > '» ' y >
eit is the disturbance term and a,, /31(, p2l, fin, fin,
and a2 are the parameters of the model to be esti- where
mated. The risk-free rate represents the interest that and end

This content downloaded from 130.133.8.114 on Mon, 28 Nov 2016 23:43:39 UTC
All use subject to http://about.jstor.org/terms
Sood and Tellis: Do Innovations Really Pay Off? Total Stock Market Returns to Innovation
Marketing Science 28(3), pp. 442-456, © 2009 INFORMS 447

Table
Table 2 Sample
2 Sample
Characteristics
Characteristics

Category External lighting Display monitors Desktop memory Data transfer Printers

Number of firms 19 17 18 17 11
Total number of announcements 696 1,100 1,239 1,323 1,123
Sample period 1977-2006 1980-2006 1979-2006 1982-2006 1981-2006
Initiation activities 155 278 270 327 117
Development activities 171 305 274 183 126
Commercialization 370 517 695 813 880
Number/type of platform 5 5 5 3 4
technologies
Incandescence, arc-discharge, CRT, LCD, PDP, Magnetic, Copper/aluminum, Dot matrix, inkjet,
gas-discharge, LED, MED OLED magneto-optical, optical fiber optics, wireless laser thermal

Note. LED, light-emitting diode; MED, microwave electrodeless discharge; CRT, cathode ray tube; LCD, liquid crystal display; PDP, plasma display panel; OLED,
organic light-emitting diode.

We also estimate the cumulative average abnormal industries—external lighting, display monitors, corn
returns using alternative models, which are explained puter memory, data transfer technologies, and desk
in the Results section. We estimate the following top printer product categories (see Appendix A). We
model to ascertain the effect of hypothesized inde- identify 69 firms in the five industries and collect
pendent variables on cumulative abnormal returns 5,481 announcements from 1977 to 2006 (see Table 2).
(CARjjp); thus, There is substantial innovative activity in all the cate
gories during this period.
CARijp = a+piALijp+p2FN¡jp+/33EP¡jp+PiPRijp The present study goes further than previous st
ies in two important aspects. First, we identify all
5 >íp ^6 ijp P7 tjp P7 sJijp major firms and all technologies within each indus
+/3sAFj+f39SZj+f3wRPj+P^ATp + r¡pp, (6) Second, we collect all announcements related to
vation projects made by the firms for each activity of
where the project.
ALjjpi Announcements Sources
of alliances;
FN ¡ • : Announcements of funding;
FP¿- Announcements of expansion; Although many studies limit their focus to a
'IP'

PRAnnouncements
Iijp■ of prototypes; source of announcements, we posit that the
PtL'.
' 'IP' Announcements of patents; mation on Novation projects reaches the m
P .

PA
iyp
■ Preannouncements- through a variety of sources. So, limiting the sour
Pll: Announcements of new product launch; on'>' one Publication may not capture the da
JI in 4-i vol 4-/~v 4k\î-i w\ 1, I «-% ,-J

RQÜ': Announcements of awards; information is first released to the markets. Indeed,

AF : Announcement frequency; Glascock et al. (1987) show how the Wall Street Jour
SZ ■ Size of firm- not Pubilsb all the news and that there is
un td u ' j J.- -i_ e j.u c a lag of three days between a change in bond ratine
RP¡: Research productivity of the firm; P, , , ,J ° „ b
A r . , i by Moody s and an announcement by the Wall Street
p. ge o ec o ogy, Journal. Hence, in the int
where subscripts refer to announcement i, firm j, and prehensiven
project p, respectively. (.¡on as weg The primar
includes the Wall Street Journal), Le
Sample company websites for press releases/announcements
We use two criteria to select product categories: a rea- on technological innovations. We also include
sonable number of emerging technologies and data newswire services such as PR NeWswire, Busin
availability. We select product categories where a Newswire, and Reuters. We collect company bac
number of technologies have emerged in the last few ground information from General Business File ASA
decades and the key global players are in U.S. mar- and Yahoo! Finance
kets. The first requirement is essential to ensure that
we have a large sample of announcements and the Procedure
second is essential because we require the firm to After the selection of the industry, we identif
be listed on U.S. stock markets to assess the mar- major firms in the industry and collect informatio
ket value. On the basis of these criteria we col- each firm. We use the following key words to ident
lected data using the historical method (Golder 2000, all the announcements: name or ticker symbol of fir
Golder and Tellis 1993) on 19 technologies from five names of technology, and events of the innova

This content downloaded from 130.133.8.114 on Mon, 28 Nov 2016 23:43:39 UTC
All use subject to http://about.jstor.org/terms
 Sood and Tellis: Do Innovations Really Pay Off? Total Stock Market Returns to Innovation
448 Marketing Science 28(3), pp. 442—156, ©2009 INFORMS

project. We first sort the results based on oldest press Analysis o

report to identify the first release of information to We separat
the market. We exclude press reports appearing in into positiv
nondaily publications because of the inherent inac- negative a
curacy of determining the exact date of release of ities was ap
information. Of the remaining press reports, when announcem
multiple reports contain identical information about age return t
an event, we retain only the first press report, which the uni
we treat as the announcement. However, an event may variate me
have multiple announcements because of new infor- and contr
mation in each announcement. Finally, we include Both resu
announcements in the analysis of only firms whose returns to
data are available from the Center for Research in significantl
Security Prices (CRSP) (firms traded on NYSE, AMEX, examination
or NASDAQ) because we need price information to we used P
estimate returns. Table 4 reports these consistent estimates (refer to sec
We examine each announcement to classify it by tion on regression diagnostics for m
firm, innovation project of the firm, activity of the in- adjusted R2 for the models is at lea
novation project, and event within the set of activities. comparable to prior studies (Chaney
et al. 1997, Sorescu et al. 2007).

Results Initiation Activities. Across all categories, the

Market response to announcements using the event returns to all events
study method suggests that the cumulative average 0.6% (f = 3.7). The fin
abnormal returns to all announcements in the sam- by announcing their
pie are positive (see Table 3). Across all categories, average, returns to init
the cumulative average abnormal returns to all launch. At the event lev
announcements are 0.4% on the event day. This result for positive announcem
holds even when examined at the individual cate- funding (0.9%, t = 2.3)
gory level. Moreover, the returns are the highest on f = 2.2) (see Table 4).
the day of the announcement and not significantly were not significantly dif
different from zero for event windows longer than ative announcements
five days (±2 days around the day of announcement) tion of alliances (-0.3
(see Table 3 and Figure 2). Hence, in the rest of the funding to projects
paper, we use the abnormal returns for an event win- plans (—0.6%, t = —
dow of only one day and use the term returns to mean results is that while f
abnormal returns. The returns that we report are for forthcoming joint v
the FFM model (Equation (1)), although a subsequent have other indicators
subsection explores returns by other methods. events, such as the dissol
We present the results in four subsections: analysis tures before the actual f
of returns, analysis of total returns, additional analy- the actual negative even
ses, and test of robustness. not that bad.

Table 3 Descriptive Statistics: Abnormal Returns to an Average Event by Category tor Various Wi

AAR (event day) CAAR (±1 day) CAAR (±2 days)

Percentage of
Category N Est. (%)
Est, /-value p-valuea positive p-valueb Est. (%) /-value Est. (%) /-value

All 5,481 0.4 7.4 <0.0001 52 <0.0001 0.5 14.7 0.5 3.3
Lighting 696 0.9 6.3 <0.0001 56 <0.0001 1.1 13.7 1.4 3.6
Monitors 1,100 0.8 3.5 <0.0001 51 0.015 0.7 5.7 0.4 0.7
Memory 1,239 0.3 2.7 0.0135 51 0.004 0.5 9.3 0.4 1.4
Data transfer 1,323 0.2 2.8 0.0047 51 0.004 0.2 4.6 0.3 1.5
Printers 1,123 0.1 1.8 0.1301 51 0.026 0.1 1.6 0.3 1.5

"The p-value is estimated using the Brown-Warner (1985) approach.

"The p-value is estimated by sorting the 265 average abnormal returns from minimum to maximum and calculating how far away from the tail in rank the
event average abnormal return is for these 265 values. We thank the anonymous reviewer for suggesting this.

This content downloaded from 130.133.8.114 on Mon, 28 Nov 2016 23:43:39 UTC
All use subject to http://about.jstor.org/terms
Sood and Tellis: Do Innovations Really Pay Off? Total Stock Market Returns to Innovation
Marketing Science 28(3), pp. 442-456, ©2009 INFORMS 449

Table
Table4 Average
4 Abnormal
Average Returns Abnormal
to Various Events During
Returns
Innovation Projects
to Various Events During Innovation

Univariate (Equation (3)) Multivariate (Equation (6))

Positive
Positive only only Negative only Positive
Negative only only Negative
Positive only only Allb
Negative only All"

Announcements
Announcements N Est. (%) f-valuea N Est. (%) f-valuea
< El C CD
Est. (%) f-value Est. (%) f-vaiue
f-value Est. (%) f-vaiue
f-value

Intercept -0.02 -0.1 0.6 4.6 0.2 1.0

Alliances 878 0.6 5.1 34 -0.02 -0.1 0.5 3.3 0.2 0.4 0.4 2.6
Funding 154 0.9 2.3 18 -1.3 -0.6 0.7 2.1 -1.1 -1.4 0.4 2.4
Expansion 181 0.6 2.2 29 -0.6 -0.9 0.4 1.1 -0.3 -0.2 0.2 0.7
Prototypes
Prototypes 776 1.0 9.0 21 -4.2 -5.9 0.6 3.5 -2.3 -2.4 0.5 2.6
Patents 218 1.6 4.0 85 -1.6 -2.5 1.4 4.9 -1.8 -4.4 0.4 1.6
Preannouncements
Preannouncements 762 1.2 8.8 39 -4.7 -9.6 0.9 5.3 -3.2 -4.3 0.6 3.6
Launch
Launch 2,106 0.2 2.5 16 -4.7 -7.2 0.2 1.6 -2.2 -2.2 0.01 0.1
Awards 488
488 1.2 5.2 0.8 3.9 0.0 1.8 0.7 3.0
Announcement
Announcement 1,8E—05
1.8E—05 1.0 -7.9E-08 -3.9 2.4E-05 1.4
frequency"
frequency0
Size of firm"
firmd —8.6E—08 -4.2 —8.0E-05
-8.0E-05 -1.07 —8.2E-08 -4.0
Research
Research —5.8E—05
—5.8E—05 -0.8 1.3E-05 0.4 —5.7E-05
-5.7E-05 -0.8
productivity
productivity
Age
Ageofoftechnology
technology 3.4E-05 1.1 1.3E-05 0.4 2.6E-05 0.8
Adj. fl2
Adj. R2 2.48
2.48 2.24 1.48

Estimated using the Brown-Warner (1985) method (Equation (4)).

"Announcement frequency measured as the number of prior announcements.
"Positive announcements coded as "1" and negative announcements coded as 1."
"Size of firm was also measured as the number of different technologies in which a firm invests.

Development Activities. Across all categories, the Activities with the Highest Returns. We find that
returns to all development activities are 0.9% (t = 5.5). the highest returns are for development activities
At the event level, we find that market returns are (see Figure 1). Across all categories, the returns for
strongly positive for announcements of successful the development (D) activities are significantly greater
demonstration of prototypes (1.0%, t = 9.0), patents than those for the initiation (I) activities (f = 2.7) and
(1.6%, t = 4.0), and preannouncements (1.2%, t = 8.8) the commercialization activities (f = 4.0). At the indi
(see Table 4). A majority of the positive announce- vidual category level, the returns to development are
ments on patents are from firms announcing award more than commercialization (C) activities or initia
of patents. Surprisingly, negative returns for nega- tion activities in all five categories,
tive announcements are even higher in absolute value Resultg for Strategic and structural Variables. The
than positive returns. For example, returns are -4.2% results of the anaiysis of strategic and structural vari
(f = -5.9) for delays in product development dead- ables estimated via the model in Equation (6) (see
lines or failure to meet expected performance lev- Table 4) are as follows
els, -1.6% (f = -2.5) for denial of patents or patent . A higher (or lower) number of prior announce
infringement suits, and —4.7 (f = —9.6) for post- ments or longer time since the last announcement
ponement, delay, deferral, shelving, or suspension of
launches. Figure 1 Average Abnormal Returns Figure
(AAR) in Each
1 Average AbnormalSet of(AAR)
Returns Activities
in Each Set ofof
Activities of
Commercialization. Across all categories, the re- Innovation Project
Innovation Project

turns to all commercialization activities are 0.3%

1-41—
(t = 2.5). At the event level, market returns are pos
itive for announcements of launch of new products l.o
(0.2%, t = 2.5) and receipts of awards (1.2%, t = 5.2) g 0.8

ïJÊÊàiL.s
(see Table 4). In contrast, market returns to delays os o.6
in product launches, cancellation of plans to launch < 04
products, and product recalls because of malfunctions
have a negative return of —4.7% (f = —7.2).
-0.2 1^""'"'■"" ! , '"Ti
All I.iphtinp Mrvnitrvrc lv if .... ' rx..,„ 1 377 ! '
In summary, we find that market returns to neg- "u z '"""au 'Lighting
Lighting1
MonitorsMonitors
Memory Data' Printers
Memo[y 1 Data ' Printers
ative announcements are negative across all events. transfer
transfer

However, the absolute value of the market returns is

All Lighting Monitors Memory Data transfer Printers
higher for negative announcements than for (%) positive
(%) (%) (%) (%) (%)
announcements. This result is consistent■ iI with
0.4 theory
0.8 0.4 0.4 0.2 0.20
0.20

and findings that losses loom larger

■
B D
than
0.8
0.8 1.0
1.0
gains
1.0 0.5 0.6 1.30
1.30
■C 0.3 0.9 0.8 0.1 0.2 0.04
(Kahneman and Tversky 1979).

This content downloaded from 130.133.8.114 on Mon, 28 Nov 2016 23:43:39 UTC
All use subject to http://about.jstor.org/terms
 Sood and Tellis: Do Innovations Really Pay Off? Total Stock Market Returns to Innovation
450 Marketing Science 28(3), pp. 442 456, ©2009 INFORMS

within a project does not lead to higher returns. The average ret
results remain similar even if we code the prior num- rabie to e
ber of positive or negative announcements separately. prior stud
• Returns are higher for smaller firms than for than the mean
larger firms. Hence, ignoring the totality of events of innovation
• The age of technology does not have an effect on when estimating returns severely
the market returns to innovation. total returns to innovation.
• Firms with higher research productivity (across To estimate the dollar
projects) do not have higher returns per announce- we first compute doll
ment than firms with lower research productivity. thus,
We used alternative measure of research produc- CARDijp = CARijp*SOj*SPj, (9)
tivity—the number of technologies a firm invests in. where
,
We find that returns for firms that invest in a few
technologies is higher than for firms that invest across CAR£V Returns ^ dollars for announcement ;;
a broad set of technologies (f = -3.2). S0/: Number of shares outstanding for firm ; on
day of announcement i;
Analysis of Total Returns SPp Price of shares for firm j at the end of that
The sum of returns to all events within an innova- trading day.
tion project of a firm provides the total returns to that We then follow the same procedure as describ
innovation project. We exclude firms where data on above to compute the dollar value of returns to an
shares outstanding are not available from CRSP. We event or an innovation project and that for the who
then calculate the returns to each project as the sum project. Across the five markets, the average return to
of returns to all announcements for that project; thus, an event is $49 million, while the average total retur
to any project is $643 million. Again, taken across or
TRjp = RALjV + RFNjp + REPjp + RPRjp within categories, returns to projects are substantially
+ RPT¡p + RPAjp + RPLjp + RRQjp, (7) more than the returns to individual events'

where TR is total returns to firm j for project p, and Additional Analyses

RALjp, RFNjp, REP: , RPRip, RPT]p, RPAjpr RPLip/ and no,w Presei* two addihonal analyses: returns o
RRqI are returns to all announcements of alliances, first relative to later announcements and returns r
funding, expansion, prototypes, patents, preannounc- tlve to comPehtors
ements, new product launch, and awards for project p, First Announcement. Readers may suspect that t
respectively. first announcement of an innovation project would
We estimate the average return to a project across yield higher returns than an
the sample as The reason may be that the first announcement tells
H TRjp of a whole new project or product by the firm. Sub
p~ } ' sequent announcements may not have as big an in
where / is the total number of projects in the sam- Rational or signaling impact (Kleine
pie. Table 5 shows that the total returns (averaged l , Naë^-As^yag and Manceau 200
„ , • w m o o/ ti, . , i , , this hypothesis. We define the first announcement
across all categories) is 10.3%. The total returns by , , , , . ,
, . l , i o i o/ c • , • ... i o co/ as the first ever release of information on an ínno
category is about 13.1% for projects in lighting, 19.8%
for projects in monitors, 7.02% for projects in memory vatlon Prolect and later ann
products, 7.4% for projects in data transfer, and 3.8% announcements during the pr
c ... • , . „ , , ,, . , We find that the difference between the returns to
for projects m printers. More important, the simple ,
the first announcement of any project and the returns
to any later announcement (second, third, or all sub
Table 5 Total Abnormal Returns to Innovation by Category
Table 5 Total Abnormal Returns to Innovation by Category sequent) are not significantly different from zero. We
Total
Total abnormal
abnormal Total
Total abnormal
abnormal also compare the returns to the first announcement in
returns (%) returns ($M) each set of activities with later announcements within
Stage (Equation (7)) (Equation (8)) the same set of activities and the results are simi

All 10.3 972 lar. These results belie the expectation that the first
Lighting 13.1 712 announcement is more important. A possible reason
Monitors 19.8 1,275 might be that later announcements may have equally
7.02 446
Memory large (or larger) returns because what they lack in
Data transfer 7.4 2,635
Printers 3.8 432
"news" value they make up for by indicating increas
ing confidence that the project will succeed.

This content downloaded from 130.133.8.114 on Mon, 28 Nov 2016 23:43:39 UTC
All use subject to http://about.jstor.org/terms
Sood and Tellis: Do Innovations Really Pay Off? Total Stock Market Returns to Innovation
Marketing Science 28(3), pp. 442-456, ©2009 INFORMS 451

Table
Table6 Effect
6 of Effect
Innovation on of
Abnormal Returns to Competitors
Innovation on

Differen ¡ñ 2.5% levels. Second, we rep

Difference in
abnormal
abnormal returns returns removing observations w
toto competitors
competitors vs. vs. with potentially undue in
Competitors
Competitors announcing firmannouncing firm age on the results
Phase f-value f-value
higher than 4/n (Cook 1979). The r
Category Est. (%) Diff. (%)
original results in both cases for a
All 1 0.1 0.7 -0.3 2.5 for new funds, where the coeffici
D 0.1 2.5 -0.7 5.1
but no longer significant.
C 0.1 2.3 -0.2 2.2
We also test for presence of autocor
Lighting 1 -0.1 -0.6 -0.9 2.3
D 0.0 -0.4 -1.1 2.9
using the Durbin-Watson statistic af
C 0.1 1.6 -0.7 2.4 outliers. The tests fail to reject bo
Monitors 1 0.1 1.1 -0.4 1.3 of no autocorrelation in the errors
D 0.1 0.7 -4.7 3.1 tive hypotheses of positive and ne
C 0.1 0.8 -0.8 3.1 tion, respectively, for z'th order au
Memory 1 0.1 1.6 -0.3 0.9 0 < i < 4.
D 0.1 1.1 -0.4 1.7 The White test is significant (Pr > x2 —< 0.0001) and
C 0.0 -0.1 -0.1 0.8
suggests potential heteroskedasticity of residuals. We
Data transfer 1 0.0 0.3 -0.1 0.6
plot the residuals versus fitted values to investigate
D 0.2 1.2 -0.4 1.8
C 0.1 1.5 -0.1 0.5
any patterns of increasing residuals. No such patterns
are visible. We also reestimate the model after remov
Printers 1 -0.2 -1.5 -0.5 1.7
D 0.5 3.1 -0.9 2.0 ing observations to maintain a constant bound on the
C 0.1 1.5 0.2 -1.5 variance of residuals; the results are similar.
In both the level of set of activities and individual
Note. I, initiation; D, development; C, commercialization.
events of analysis, multicollinearity is not a problem
among the control variables, as indicated by the coef
Returns Relative to Competitors. How do the re- ficient variance-decomposition analysis and the
turns of the announcing firm affect returns to com- dition indices.
petitors in each of the three set of activities? Most past Alternative Methods to Estimate Market Re
studies suggest that competitors experience negative We use three other models to estimate the "no
returns in such a situation (Zantout and Tsetsekos returns to verify the robustness of our results—
1994, Chen et al. 2005, Ferrier and Lee 2002, Akhigbe market adjusted, and market model (McKinlay
2002). We extend the analysis to examine the returns First, we used the mean return model (Equatio
to a firm relative to its competitors at various sets of in which the firm is expected to generate the
activities of the innovation project. We create a port- return that it averaged during a previous estimati
folio of all firms that made no announcement on the period. Second, we used the market-adjusted r
day the focal firm makes an announcement. model (Equation (11)) in which the firm is exp
Consistent with the findings of prior literature, we to generate the same return as the rest of the ma
find that in all three sets of activities, the returns to Third, we used the market model where the firm
competitors are negative (see Table 6). These results expected to generate the same return as a port
hold even if we expand the definition of competitors °1 stocks used to represent the overall market
te include all firms across categories in our sample tion (12)). Thus,
not making the announcement or use wider windows r _ r¡ _|_ e (10)
around the day of the announcement (e.g., ±1 or
±2 days). Rit — Rmt + Sit> (11)

T 4 ( 1Î V. 4 R-it = ai + R-mt + Eit' (12)

Tests of Robustness

We carry out a number of analyses to test the robust- where Ru and Rmt are the period f retur
ness of the results including regression diagnostics, r*ty ' and market portfolio, respective
alternative method to estimate returns, alternate mar- the zero mean disturbance term. The estim
ket index, nonparametric tests, and accounting for the dow for a11 three models is the sam
lack of clean estimation period. tion W" For each firm z and event date l>
a v* _ v _ v (i m
Regression Diagnostics. We examine the impact •' " " v '
of residuals (outliers) on the outcome and accuracy AR*t = Rit — Rmt, (14)
of the regression results. First, we repeat the regres
° AO AD* D ( ^ i O D \ C\ C\
sion after trimming

This content downloaded from 130.133.8.114 on Mon, 28 Nov 2016 23:43:39 UTC
All use subject to http://about.jstor.org/terms
 Sood and Tellis: Do Innovations Really Pay Off? Total Stock Market Returns to Innovation
452 Marketing Science 28(3), pp. 442-456, ©2009 INFORMS

Figure 2 Cumulative Average Abnormal Returns (CAAR) Using OLS Mark

model
FFM4 model

Market-adjusted model

Mean-adjusted model
Mean-adjusted model

OLSmarket
OLS market model
model

where AR*t, R„ a¡, and /§,■ are the abnormal estimate market parameters prior to the event
return, mean firm return, and parameter estimates clean; i.e., there is no other announcement ma
of market-adjusted model, respectively. The plots of by the firm in that period. Because we examin
CAAR in Figure 2 using all models—mean, market, multiple announcements made by the same firm
and market-adjusted models—demonstrate that the over the entire innovation project, this assump
CAAR was not much different with the use of these tion is violated. We remove the dates of all pr
models. Similarly, there were no significant differ- announcements made by the firm from the esti
ences in the reported results for the hypotheses with tion period (Brown and Warner 1985) and reestimate
the use of these alternate models as well. the returns. The results do not change much with this
Alternate Market Index. We use the equally correction,
weighted market index to estimate the abnormal re
turns in Equation (1) as recommended by Brown and Discussion
Warner (1980, 1985). We also reestimate the returns This section summarizes our findings and discusses
using the value-weighted market index to ensure im lications and iimitations.
robustness. The results are not materially different
from those presented. Summary of Findings
Nonparametric Tests. We use the Wilcoxon sign- The current research leads to these major findings
rank test to test the null hypothesis that the observed • Total market returns to an innovation projec
returns are symmetrically distributed around 0 and $643 million, substantially greater than $49 mil
the proportion of observed sample securities having the returns to an average event in the innova
positive returns is equal to 0.5. This situation would project.
be true if markets do not respond favorably to posi- . Of three sets of activities of innovation (in
tive news of technological innovations. The Wilcoxon tion, development, and commercialization), returns t
sign-rank test uses both the sign and the rank infor- the development activities are consistently the high
mation and is therefore more powerful than the sim- across and within categories and the returns to
plet binomial sign test. The results reject the null mercialization the lowest. Moreover, returns to in
(p = 0.001) and support our findings that market tion occur, on average, 4.7 years ahead of launch
returns to innovation are positive. » Returns to the new product launch are the lowe
Accounting for the Lack of Clean Estimation among all eight events tracked.
Period. An assumption intrinsic to the market- • Returns to negative events are higher in abs
adjusted model is that the estimation period used to value than those to positive events.

This content downloaded from 130.133.8.114 on Mon, 28 Nov 2016 23:43:39 UTC
All use subject to http://about.jstor.org/terms
Sood and Tellis: Do Innovations Really Pay Off? Total Stock Market Returns to Innovation
Marketing Science 28(3), pp. 442—456, © 2009 INFORMS 453

• Returns are consistently higher for small firms progress in their

than for large firms, and for those that focus on a few vaporware. How
rather than many technologies. all positive announcements
• Returns to the announcing firm are substantially from zero for a
greater than those to competitors across all stages. ments, firms shou
• The number of prior announcements or time events. Otherwise,
since the last announcement has no effect on the mar- increasing marke
ket returns to innovation. nouncements. These findings are also consistent with
• Returns to the first announcement of an inno- recent findings in marketing literature that sug
vation project are not different from returns to later markets react positively to new product introducti
announcements. Similarly, results for older technolo- but discount short-term promotions.
gies and projects are not different from those for Fourth, returns are highest for developmental a
newer ones. ities. Returns are higher for development activities
over startup activities probably because star
Implications and Contributions to Practice ities involve heavy up fron
This study has several implications for managers. ditures and resources with
First, markets respond promptly and substantially several years away. Ret
to announcements about innovation at all stages of opment activities over c
the innovation project. When considering the value of because development ac
innovation, it is inappropriate to limit the analysis to reduction of uncertaint
only one or another event in the innovation project. of the expected returns
The frequently cited undervaluation of innovation izations. Thus, it is import
(Hall 2009, 1993; Hall et al. 1993) may be due not to progress in development b
markets not appreciating the full value of innovations developments.
immediately, but to researchers computing returns to Fifth, when announcing in
isolated events in an innovation project. Following the not seem to suffer any disa
approach described in this study, managers can com- firms. Rather, small firms s
pare the costs of an innovation project to the average than large firms, ceteris pa
returns they can get to better assess the value of any this is that large firms are
innovation project they plan to undertake. covered by the investment
For example, AXT Inc. develops and markets news from small firms is
three product lines of high-performance compound positive surprise than that fr
semiconductor substrates—gallium arsenide (GaAs)
substrates, indium phosphide (InP) substrates, and Limitations and Futur
single-element substrates. Between 2000 and 2003, the several Citations that can be
firm made various announcements on the develop- In a11 categories, the hi
ment of new products, allocation of resources to the consistently for announc
three innovation projects, and expansion of manu- activities. However, we co
facturing facilities. With our approach, we estimate or why this ocÇurs' Secon
the total return to the three innovation projects to flve industries because of
be $29.3 million. These returns are substantial when a comprehensive set of an
compared to total R&D expenditures of $11.9 million ab°ut, novation project
durine this period include firms not listed on the stock markets. Future
Second, the findings on various announcement research might explore whether the same results hold
strategies indicate that a mere increase or decrease in for such Fourth' the results may be affected by
either the frequency or total number of announce- a potential selection bias as firms can be more seg
ments does not lead to an increase or decrease in tive about the type of announcements made during
returns. The median number of prior announcements initiation and development than during the market
in our sample is 2 and the 90th percentile is 9. More- stage,
over, the first announcement of a project is no more
important than later announcements. These results .r.c n0^V f S, t t .. .
. r, , ... . . ,. The studv benefited from grants by Don Murray to the Urn
ímply that the markets are efficient, and firms can- vers,ty ¿ Southern
not game the system by overannouncing or through Global Innovation> the
multiple announcements of a single event. ter for Research in Techn
Third, the absolute value of a negative announce- School of Management, a
ment is greater than that for a positive announce- and from the research
ment. Thus, firms should be careful not to exaggerate Harris, Angie Zerillo

This content downloaded from 130.133.8.114 on Mon, 28 Nov 2016 23:43:39 UTC
All use subject to http://about.jstor.org/terms
 Sood and Tellis: Do Innovations Really Pay Off? Total Stock Market Returns to Innovation
454 Marketing Science 28(3), pp. 442-456, ©2009 INFORMS

Appendix A. Operating Principles of Sampled Technologies

Technology Principle

External lighting
Incandescence Generate light by heating up thin metallic wires with an electric current
Arc-discharge Emit light by arc formed between two electrodes oppositely charged by an electric current
in a high-pressure gas chamber
Gas-discharge
Gas-discharge Electrons
Electronsexcited
excitedbybypassing
passinganan
electric current
electric in in
current a low-pressure gas gas
a low-pressure chamber emitemit
chamber lightlight
Light emitting
emitting diode
diode Emission
Emission of
of the
the light
lightin
inn-p
n-ptransition
transitionzone
zoneunder
underinfluence
influenceofofananelectric
electric
potential
potential
Microwave electrodeless
Microwave electrodeless discharge
discharge Emission
Emissionof
oflight
lightby
bymicrowaves
microwavesfrom
frominduction
inductioncoil
coilinside
insidethe
the bulb
bulb toto excite
excite the
the gasgas
Display monitors
monitors
Cathode ray
ray tube
tube Form
Form an
an image
image when
when electrons,
electrons,fired
firedfrom
fromthe
theelectron
electrongun,
gun,converge
convergetotostrike
strike
a screen
a screen
coated with
with phosphors
phosphors of
of different
different colors
colors
Liquid crystal
crystal display
display Create
Create an
an image
image by
bypassing
passinglight
lightthrough
throughmolecular
molecularstructures
structuresofofliquid
liquidcrystals
crystals
Plasma display
display panel
panel Generate
Generate images
Images by
bypassing
passingaahigh
highvoltage
voltagethrough
througha alow-pressure
low-pressureelectrically
electricallyneutral
neutral
highly
highly
ionized atmosphere
atmosphere using
using the
the polarizing
polarizingproperties
propertiesof
oflight
light
Organic light
light emitting
emitting diode
diode Generates
Generateslight
lightby
bycombining
combiningpositive
positiveand
andnegative
negativeexcitons
excitons(holes
(holes
emitted
emitted
byby
anodes
anodes
and electrons
electrons emitted
emitted by
by cathodes)
cathodes) in
in aapolymer
polymerdye
dyethrough
throughthe
theprinciple
principle
of electroluminescence
electroluminescence

Desktop memory
memory
Magnetic Record data by passing a frequency modulated current through the disk drive's magnetic
head, thereby
thereby generating
generating aa magnetic
magneticfield
fieldthat
thatmagnetizes
magnetizesthe
theparticles
particlesofofthe
thedisk's
disk's
recording
recording surface
surface
Optical Store data using the laser modulation system, and changes in reflectivity are used to
store and retrieve data
Magneto-optical Record data using the magnetic-field modulation system but read the data
with a laser beam

Computer printers
Dot-matrix Create an image by striking pins against an ink ribbon to print closely spaced dots that form
the desired image
Inkjet Form images by spraying ionized ink at a sheet of paper through micro-nozzles
Laser Form an image on a photosensitive surface using electrostatic charges, then transfer the
image on to a paper using toners, and then heat the paper to make the image permanent
Thermal Form images
Images on paper by heating ink through sublimation or phase change processes
Digital data transfer
Cu/AI Transmit data in the form of electrical energy as analog or digital signals
Fiber optics Transmit data in the form of light pulses through a thin strand of glass using the principles
of total internal reflection
Wireless Encode data in
In the form of a sine wave and transmits it with radio waves using
a transmitter-receiver combination

Source. Adapted from Sood and Tellis (2005).

Appendix B. Examples of Positive and New Patents

Negative Announcements Positive: Universal Display Corporation announces
issuance of the 14th patent in the organic light em
Joint Ventures project.
Positive: Cree Research and Philips sign joint a
new laser diodes will increase optical storage ca
ARPA provides $4 million funding.
Preannouncements
Negative: Hitachi, GE dissolve lighting joint ventu
Positive: Sony Corporation of Japan said on Tuesday
New Funds ^ W'" launch a home-use optical-type videodisc player,
"Laser Max," on April 21.
Positive: Intel to invest $100 million in Hitachi, Ltd.'s
„ . , „ Negative: Sony to delay mass production of digital audio
int venture Elpida Memory Inc.-DJ.
joint venture Elpida Memory Inc.-DJ. tape (DAT) heads
Negative: Storage Technology loses loan.
Product Launch

New Prototypes Positive: Sony expands 5.25-inch magneto optical library

Positive: IBM says it set record for bits of data on disk. line to include permanent WORM config
Negative: Gentex delays new LED technology. Negative: Sony to delay mass production of DAT heads.

This content downloaded from 130.133.8.114 on Mon, 28 Nov 2016 23:43:39 UTC
All use subject to http://about.jstor.org/terms
Sood and Tellis: Do Innovations Really Pay Off? Total Stock Market Returns to Innovation
Marketing Science 28(3), pp. 442-^156, ©2009 INFORMS 455

Quality Awards Das, S., P. K. Sen, S. Sengupta. 1998. Impact of strategic alliances
on firm"energy
Positive: EPA names Lexmark International valuation. Acad.
starManagement J. 41(1) 27-42.
printer partner of the year." Dollinger, M. ]., P. A. Golden, T. Saxton. 1997. The effect of repu
tation on the decision to joint venture. Strategic Management J.
18(2) 127-140.
References Doukas, }., L. Switzer. 1992. The stock market's valuation of R&D
spending and market concentration. J. Econom. Bits. 44 95-114.
Aaker, D. A. 1995. Strategic Market Management, 4th ed. John Wiley
& Sons, New York. Eddy, A. R., G. B. Saunders. 1980. New product announcements
Adner, R., D. Levinthal. 2001. Demand heterogeneity and technol and stock prices. Decision Sci. 11(1) 90-97.
ogy evolution: Implications for product and process innova Erickson, G., R. Jacobson. 1992. Gaining comparative advantage
tion. Management Sci. 47(5) 611-628. through discretionary expenditures: The returns to R&D and
Akhigbe, A. 2002. New product innovations, information signal advertising. Management Sci. 38(9) 1264-1279.
ing and industry competition. Appt. Financial Econom. 12(5) Fama, E. 1998. Market efficiency, long-term returns, and behavioral
371-378. finance. ]. Financial Econom. 49(3) 283-306.
Anand, B. N., T. Khanna. 2000. Do firms learn to create value? Fama, E., K. French. 1993. Common risk factors in the returns on
The case of alliances. Strategic Management ]. 21(Special Issue) stocks and bonds. J. Financial Econom. 33(1) 3-56.
295-315.
Fama, E., L. Fisher, M. C. Jensen, R. Roll. 1969. The adjustment of
Austin, D. 1993. An event study approach to measuring innova stock prices to new information. Internat. Econom. Rev. 10(1)
tive output: The case of biotechnology. Amer. Econom. Rev. 83(2) 1-21.
253-258.
Ferrier, W., H. Lee. 2002. Strategic aggressiveness, variation, and
Barney, J. B. 1986. Strategic factor markets: Expectations, luck, and surprise: How the sequential pattern of competitive rivalry
business strategy. Management Sci. 32(10) 1231-1241. influences stock market returns. J. Managerial Issues 14(2)
Bayus, B. L., G. Erickson, R. Jacobson. 2003. The financial rewards of 162-180.
new product introductions in the personal computer industry.
Foster, R. 1986. Innovation: The Attacker's Advantage. Summit Books,
Management Sci. 49(2) 197-210. New York.
Berenson, C., I. Mohr-Jackson. 1994. Product rejuvenation: A less
Givoly, D., D. Palmon. 1982. Timeliness of annual earnings
risky alternative to product innovation. Bus. Horizons 37(6)
51-58. announcements: Some empirical evidence. Accounting Rev.
57(3) 486-508.
Brown, S., J. Warner. 1980. Measuring security price performance.
J. Financial Econom. 8(3) 205-258. Glascock, J. L., W. N. Davidson III, G. V. Henderson, Jr. 1987.
Announcement effects of Moody's bond rating changes on
Brown, S., J. Warner. 1985. Using daily stock returns: The case of
equity returns. Quart. J. Bus. Econom. 26(Summer) 67-78.
event studies. J. Financial Econom. 14(1) 3-31.
Carhart, M. 1997. On the persistence of mutual fund performance.Colder, P. N. 2000. Historical method in marketing research with
/. Finance 52(1) 57-82. new evidence on long-term market share stability. J. Marketing
Res. 37(2) 156-72.
Chan, L., J. Lakonishok, T. Sougiannis. 2001. The stock market val
uation of research and development expenditures. J. Finance Golder, P., G. J. Tellis. 1993. Pioneering advantage: Marketing fact
56(6) 2431-2457. or marketing legend. J. Marketing Res. 30(2) 158-170.
Chan, S., J. Kissinger, J. Martin. 1992. The market rewards promisHall, B. H. 1993. The stock market's valuation of R&D investment
ing R&D—And punishes the rest. J. Appl. Corporate Finance 5(2) during the 1980's. Amer. Econom. Rev. 83(2) 259-264.
59-66. Hall, B. H. 2009. The financing of innovation. S. Shane, ed. The
Chandrasekaran, D., G. J. Tellis. 2008. The global takeoff of new Blackwell Handbook of Technology and Innovation Management.
products: Culture, wealth, or vanishing differences. Marketing Blackwell Publishers, Oxford, UK. Forthcoming.
Sci. 27(5) 844-860. Hall, B. H., E. Mansfield, A. B. Jaffe. 1993. Industrial research dur
Chandy, R., G. J. Tellis. 2000. The incumbent's curse? Incumbency, ing the 1980s: Did the rate of return fall? M. N. Baily, P. C.
size and radical product innovation. J. Marketing 64(3) 1-17. Reiss, C. Winston, eds. Brookings Papers on Economic Activity:
Chaney, P. K., T. M. Devinney, R. S. Winer. 1991. The impact of Microeconomics, Vol. 2. Brookings Institution, Washington, DC.
new product introductions on the market value of firms. J. Bus.Hendricks, K. B., V. R. Singhal. 1996. Quality awards and the mar
64(4) 573-610. ket value of the firm: An empirical investigation. Management
Chen, S. S., K. W. Ho, K. H. Ik, C. Lee. 2005. The wealth effect Sci. 42(3) 415^436.
of new product introductions on industry rivals. /. Bus. 78(3) Hirschey, M. 1982. Intangible capital aspects of advertising and
969-996.
R&D expenditures. J. Indust. Econom. 30(4) 375-390.
Christensen, C. M. 1992. Exploring the limits of the technology
Jaffe, A. B. 1986. Technological opportunity and spillovers of R&D:
S-curve, part I: Component technologies. Production Oper. Man
Evidence from firms' patents, profits, and market value. Amer.
agement 1(4) 334-357. Econom. Rev. 76(5) 984-1001.
Christensen, C. M. 1997. The Innovator's Dilemma: When New Tech
Jegadeesh, N., S. Titman. 1993. Returns to buying winners and sell
nologies Cause Great Firms to Fail. Harvard Business School
Press, Boston. ing losers: Implications for stock market efficiency. J. Finance
48(1) 65-91.
Cockburn, I., Z. Griliches. 1988. Industry effects and appropriability
measures in the stock market's valuation of R&D and patents.John, G., A. M. Weiss, S. Dutta. 1999. Marketing in technology
Amer. Econom. Rev. 78(2) 419-423. intensive markets: Toward a conceptual framework. J. Market
ing 63(Special Issue) 78-91.
Cook, R. D. 1979. Influential observations in linear regression.
J. Amer. Statist. Assoc. 74 169-174. Kahneman, D., A. Tversky. 1979. Prospect theory: An analysis of
decisions under risk. Econometrica 47(2) 263-291.
Council on Competitiveness. 2004. Innovation: The new reality for
Keller, K. L., D. R. Lehmann. 2006. Brands and branding: Research
national prosperity. 21st Century Innovation Working Group
recommendations, version 2.1, December 15, Council on Com findings and future priorities. Marketing Sci. 25(6) 740-759.
petitiveness, Washington, DC. Kelm, K. M., V. K. Narayanan, G. E. Pinches. 1995. Shareholder
Crawford, M. 1977. Marketing research and the new product failurevalue creation during R&D and commercialization stages.
rate. J. Marketing 41(2) 51-61. Acad. Management J. 38(3) 770-786.

This content downloaded from 130.133.8.114 on Mon, 28 Nov 2016 23:43:39 UTC
All use subject to http://about.jstor.org/terms
 Sood and Tellis: Do Innovations Really Pay Off? Total Stock Market Returns to Innovation
456 Marketing Science 28(3), pp. 442-456, ©2009 INFORMS

Sharpe,
Klein, B., K. B. Leffler. 1981, The role W. F.
of market 1964.in
forces Capital asset prices: A
assuring
librium under
contractual performance. J. Political Econom. conditions of risk. ]. Fina
89(4) 615-641.
Sood,
Koku, P., H. Jagpal, P. V. Viswanath. 1997. A.,
The G. J.
effect ofTellis. 2005. Technologica
new product
announcements and preannouncements innovation.
on stock J. Marketing
price. J. Mar 69(3) 152-168.
ket Focused Management 2(2) 183-199. Sood, A., G. M. James, G. J. Tellis. 2009. F
Le Nagard-Assayag, E., D. Manceau. 2001. newModeling
model for thepredicting
impact ofmarket penetr
product preannouncements in the context Marketing Sci. 28(1)
of indirect 36-51.
network
externalities. Internat. J. Res. Marketing 18(3) A.,
Sorescu, 203-219.
V. Shankar, T. Kushwaha. 20
McKinlay, C. A. 1997. Event studies nouncements
in economicsand andshareholder
finance. value: Don
J. Econom. Literature 35(March) 13-39. can't keep. /. Marketing Res. 44(3) 468-48
Suárez,
Nicolau, J. L., R. Sellers. 2002. The stock M. M.
market's V. 2002.
reaction toStock market respo
quality
certification: Empirical evidence from global alliances:
Spain. Evidence
Eur. ]. Oper. Res.from Spanish c
142(3) 632-641. Rev. 14(6) 401-108.
Pakes, A. 1985. On patents, R&D, and Tellis,
the G. J., J.
stock Johnson.
market rate2007.
of The value of
26(6) 758-773.
return. j. Political Econom. 93(2) 390-409.
Paulson Gjerde, K. A., S. A. Slotnick, Tucker,
M. J. J. W. 2007. Is openness
Sobel. 2002. penalized?
New Stock returns around
product
innovation with multiple features and earnings warnings. Accountingconstraints.
technology Rev. 82(4) 1055-1087.
Management Sci. 48(10) 1268-1284. Urban, G. L., J. R. Hauser. 1980. Design and Marketing of New Prod
Pauwels, K., J. Silva-Risso, S. Srinivasan,ucts. Prentice
D. M. Hall, Englewood
Hanssens.Cliffs, NJ. 2004.
New products, sales promotions, Utterback,
and firm value:
J. M. 1974. The
Mastering the case
dynamics of
of innovation. Sci
ence,68(4)
the automobile industry. /. Marketing New Series 183(4125) 620-626.
142-156.
Woolridge,
Przasnyski, Z. H., L. S. Tai. 2002. Stock J. R., C. Snow. 1990. Stock
performance of market
Malcolmreaction to
baldrige national quality award winning companies.
strategic investment Total ]. 11(5)
decisions. Strategic Management
Quality Management 13(4) 475^188. 353-363.

Sankaranarayanan, R. 2007. Innovation and

Zantout, Z., the 1996.
R. Chaganti. durable goods
New product introductions, share
monopolist: The optimality of frequent holders'new-version
wealth, and first-mover advantages. ]. Financial Strate
releases.
Marketing Sci. 26(6) 774-791. gic Decisions 9(3) 49-61.
Zantout,product
Sharma, A., N. Lacey. 2004. Linking new Z. Z., G. P. Tsetsekos.
outcomes1994. The to
wealth
mareffects of announce
ket valuation of the firm: The case of ments
the of R&D
U.S. expenditure increases. /. Financial Res. 17(2)
pharmaceutical
205-216.
industry. ]. Product Innovation Management 21(5) 297-308.

This content downloaded from 130.133.8.114 on Mon, 28 Nov 2016 23:43:39 UTC
All use subject to http://about.jstor.org/terms