ArticlePDF Available

Two Centuries of Innovations and Stock Market Bubbles

Authors:

Abstract and Figures

The interplay between innovation and the stock market has been extensively studied by scholars across all business disciplines. However, one phenomenon remains understudied: the association between innovation and stock market bubbles. Bubbles—defined as rapid increases and subsequent declines in stock prices—have been primarily examined by economists who generally do not focus on individual characteristics of innovations or on the consequences of bubbles for their parent firms. We set out to fill this gap in our paper. Using a sample of 51 major innovations introduced between 1825 and 2000, we test for bubbles in the stock prices of parent firms subsequent to the commercialization of these innovations. We identify bubbles in 73% of the cases. The magnitude of these bubbles increases with the radicalness of innovations, with their potential to generate indirect network effects, and with their public visibility at the time of commercialization. Moreover, we find that parent firms typically raise new equity capital during bubble periods, and that the amount of equity raised is proportional to the magnitude of the bubble. Finally, we show that the buy-and-hold abnormal returns of parent firms are significantly positive between the beginning and the end of the bubble, suggesting that these innovations add value to the economy, in spite of the bubble. Our findings have important implications for managers interested in commercializing innovations and for policy makers concerned with the stability of the financial system.
Content may be subject to copyright.
This article was downloaded by: [128.194.220.55] On: 02 August 2018, At: 07:38
Publisher: Institute for Operations Research and the Management Sciences (INFORMS)
INFORMS is located in Maryland, USA
Marketing Science
Publication details, including instructions for authors and subscription information:
http://pubsonline.informs.org
Two Centuries of Innovations and Stock Market Bubbles
Alina Sorescu, Sorin M. Sorescu, Will J. Armstrong, Bart Devoldere
To cite this article:
Alina Sorescu, Sorin M. Sorescu, Will J. Armstrong, Bart Devoldere (2018) Two Centuries of Innovations and Stock Market
Bubbles. Marketing Science
Published online in Articles in Advance 16 Jul 2018
. https://doi.org/10.1287/mksc.2018.1095
Full terms and conditions of use: http://pubsonline.informs.org/page/terms-and-conditions
This article may be used only for the purposes of research, teaching, and/or private study. Commercial use
or systematic downloading (by robots or other automatic processes) is prohibited without explicit Publisher
approval, unless otherwise noted. For more information, contact permissions@informs.org.
The Publisher does not warrant or guarantee the article’s accuracy, completeness, merchantability, fitness
for a particular purpose, or non-infringement. Descriptions of, or references to, products or publications, or
inclusion of an advertisement in this article, neither constitutes nor implies a guarantee, endorsement, or
support of claims made of that product, publication, or service.
Copyright © 2018, INFORMS
Please scroll down for article—it is on subsequent pages
INFORMS is the largest professional society in the world for professionals in the fields of operations research, management
science, and analytics.
For more information on INFORMS, its publications, membership, or meetings visit http://www.informs.org
MARKETING SCIENCE
Articles in Advance,pp. 1–23
http://pubsonline.informs.org/journal/mksc/ ISSN 0732-2399 (print), ISSN 1526-548X (online)
Two Centuries of Innovations and Stock Market Bubbles
Alina Sorescu,aSorin M. Sorescu,aWill J. Armstrong,bBart Devolderec, d, e
aMays Business School, Texas A&M University, College Station, Texas 77843; bRawls College of Business, Texas Tech University,
Lubbock, Texas 79409; cVlerick Business School, 9000 Ghent, Belgium; dNetherlands Organisation for Applied Scientific Research,
2595 DA The Hague, Netherlands; eTomorrowLab, 1800 Vilvoorde, Belgium
Contact:
asorescu@tamu.edu,http://orcid.org/0000-0002-3625-0656 (AS); ssorescu@tamu.edu,
http://orcid.org/0000-0002-9338-1927 (SMS); will.armstrong@ttu.edu,http://orcid.org/0000-0002-2605-5814 (WJA);
bart.devoldere@vlerick.com,http://orcid.org/0000-0001-6722-3805 (BD)
Received: April 22, 2016
Revised: October 12, 2017; January 27, 2018
Accepted: February 4, 2018
Published Online in Articles in Advance:
July 12, 2018
https://doi.org/10.1287/mksc.2018.1095
Copyright: ©2018 INFORMS
Abstract. The interplay between innovation and the stock market has been extensively
studied by scholars across all business disciplines. However, one phenomenon remains
understudied: the association between innovation and stock market bubbles. Bubbles—
defined as rapid increases and subsequent declines in stock prices—have been primarily
examined by economists who generally do not focus on individual characteristics of inno-
vations or on the consequences of bubbles for their parent firms. We set out to fill this gap
in our paper. Using a sample of 51 major innovations introduced between 1825 and 2000,
we test for bubbles in the stock prices of parent firms subsequent to the commercialization
of these innovations. We identify bubbles in 73% of the cases. The magnitude of these
bubbles increases with the radicalness of innovations, with their potential to generate
indirect network effects, and with their public visibility at the time of commercialization.
Moreover, we find that parent firms typically raise new equity capital during bubble peri-
ods and that the amount of equity raised is proportional to the magnitude of the bubble.
Finally, we show that the buy-and-hold abnormal returns of parent firms are significantly
positive between the beginning and the end of the bubble, suggesting that these innova-
tions add value to their firm and to the economy, in spite of the bubble. Our findings have
important implications for managers interested in commercializing innovations and for
policy makers concerned with the stability of the financial system.
History:
K. Sudhir served as the editor-in-chief and Harald van Heerde served as associate editor for
this article.
Funding:
The authors gratefully acknowledge financial support from the Marketing Science Institute
and from Mays Business School at Texas A&M University. B. Devoldere acknowledges financial
support from the Belgian Intercollegiate Centre for Management Science.
Supplemental Material:
Data and the online appendix are available at https://doi.org/10.1287/
mksc.2018.1095.
Keywords:
radical innovation
stock market bubbles
capital raised
diffusion
visibility
network effects
1. Introduction
A sizable stream of research has examined the inter-
play between innovation and the stock market. In
marketing, the focus has been on short- and long-
term abnormal returns associated with new prod-
uct introductions. The marketing literature generally
finds a positive impact of innovation on stock prices
(e.g., Pauwels et al. 2004, Sood and Tellis 2009, Warren
and Sorescu 2017) and concludes that many new prod-
ucts and technological innovations have spurred the
growth of their parent firms and even of entire indus-
tries and countries (Tellis et al. 2009).
In turn, the literature in financial economics tends
to take a more macro view of innovation. A significant
part of this literature has studied the causes and con-
sequences of technological revolutions—periods when
technological advances bring abrupt and long-lasting
changes to our society (Lamoreaux et al. 2007, Perez
2002). A few authors working in this domain suggest
that the stock market plays a critical role in financing
innovations launched during technological revolu-
tions, particularly beginning with the 1920s (Hsu et al.
2014, Neal and Davis 2007, O’Sullivan 2007, Perez
2002). Others observe that such revolutions tend to be
associated with stock market bubbles—a rapid increase
and subsequent decline in stock prices—citing, as
examples, radio stock prices during 1929–1930 and In-
ternet stock prices during 1999–2000 (DeMarzo et al.
2007, Pástor and Veronesi 2009, Shiller 2015).
Are certain types of innovations more likely to be
associated with stock market bubbles than others? If
an innovation produces a bubble, are firms that com-
mercialize it affected by the bubble, and if so, in what
way? Can firms raise new capital during bubble peri-
ods and use it to accelerate the diffusion of the innova-
tion? Extant literature does not provide clear answers
to these questions. The lack of insight is due, in part, to
the fact that the bubble literature in financial economics
tends to view innovation as an undifferentiated out-
put generated by an aggregate production function,
1
Sorescu et al.: Two Centuries of Innovations and Stock Market Bubbles
2Marketing Science, Articles in Advance, pp. 1–23, ©2018 INFORMS
rather than a collection of products with distinct char-
acteristics. Moreover, studies in this area rarely incor-
porate a formal statistical measurement of bubbles,
using instead ex post observations of sudden rises and
declines in stock prices, and little attempt is made
to link such price movements to specific innovations.
A notable exception is by Frehen et al. (2013), who show
that the South Sea Bubble of 1720—commonly thought
of as a single occurrence by historians—actually con-
sisted of two separate bubbles, each associated with
contemporaneous innovations in two distinct indus-
tries: insurance and the Atlantic trade. Finally, most
papers that study bubbles tend to focus on only one or
two salient innovations, such as rail (Campbell 2012) or
the Internet (Pástor and Veronesi 2009).
Overall, the academic literature in marketing and fi-
nancial economics provides little systematic evidence
of the link between external financing, innovation, and
bubbles. However, understanding the extent to which
the stock market plays a role in helping new indus-
tries develop and flourish is a topic of high relevance
to managers seeking to finance innovation and to pol-
icy makers who seek to foster economic growth while
maintaining the stability of the financial system. Man-
agers might also be interested in learning whether cer-
tain types of innovations are more likely to be associ-
ated with disruptive patterns in their firm’s stock price,
and whether such patterns impact the eventual diffu-
sion of the innovation.1
Our paper seeks to fill this gap through a study
positioned at the interface of marketing, finance, and
economics. Using the list of radical innovations iden-
tified by Chandy and Tellis (2000) as a starting point,
we assemble a sample of 51 major innovations intro-
duced between the years 1825 and 2000. For each of
these innovations, we collect data on firms that com-
mercialized it—which we call parent firms—and test
whether these firms experienced bubbles in their stock
prices following commercialization.
To detect bubbles, we look for significant differences
between stock market prices and intrinsic values, using
the econometric method proposed by West (1987). We
then estimate a number of cross-sectional models to
understand how these bubbles relate to various char-
acteristics of each innovation. Next, we measure the
amount of new equity capital raised by parent firms
during bubble periods. Finally, we estimate the relation
between this new capital raised and the subsequent
visibility of the innovation in the public domain.
We present six important findings. First, we detect
bubbles in approximately 73% of the innovations stud-
ied (37 out of 51). By contrast, the largest collection
of innovations previously associated with bubbles can
be found in the book by Shiller (2015), who describes
nine possible bubbles, and argues that they tend
to be associated with major shifts in technological
paradigms.2Second, we show that the magnitude of
these bubbles is positively related to the contempora-
neous level of visibility of each innovation. To measure
this visibility, we obtain from the Google Books Ngram
Viewer the annual frequency with which the innova-
tion is mentioned in public books, through time. Third,
we find that bubbles are more likely to occur for inno-
vations that have a higher degree of radicalness and
for innovations that are more likely to generate indi-
rect network effects. Fourth, we find that parent firms
raise significantly more equity capital during bubble
periods compared to the average firm in the market,
and the amount of this new capital is proportional to
the magnitude of the bubble. Fifth, we show that the
new capital raised during bubble periods is positively
associated with a faster and stronger increase in the
visibility of the innovation after the bubble. Finally, we
show that the buy-and-hold abnormal returns of par-
ent firms are significantly positive between the begin-
ning and the end of the bubble, suggesting that these
innovations add value to their firms and to the econ-
omy, in spite of the bubble.
These findings have important implications for man-
agers and policy makers. Managers of firms commer-
cializing innovations that are radical (Chandy and
Tellis 1998) or innovations that can generate indirect
network effects (Stremersch et al. 2007) should be
aware of a potentially disruptive pattern in their stock
prices in the form of bubbles. At the same time, these
bubbles may create a short window of opportunity
during which equity capital could be raised on favor-
able terms. If this capital is judiciously invested, it can
help accelerate the diffusion of the innovation. In terms
of policy implications, our results suggest that bub-
bles might allow for faster building of the infrastruc-
ture needed for innovation to achieve its full economic
potential. Therefore, at least from this narrow perspec-
tive, bubbles appear to act as an indirect, voluntary tax
that some people pay to invest in what is essentially a
public good. This implication is consistent with Olivier
(2000), whose mathematical model predicts that stock
market bubbles are beneficial to value creation, invest-
ment, and growth.
Our results also support the views expressed by for-
mer Federal Reserve Chairman Alan Greenspan, who
opined that while bubbles are precipitated by exagger-
ated perceptions of economic growth, they also fuel the
diffusion of the underlying innovation by providing
the capital required for its commercialization (Lansing
2008). Without a bubble, he argued, a major innovation
might not take off or achieve its full potential. Observ-
ing that “there were no bubbles in the Soviet Union,
Greenspan concluded that policy makers should not
attempt to quash bubbles since doing so could impede
innovation and lead to suppressed economic activity
over the long term (Guha 2008, p. 2).
Sorescu et al.: Two Centuries of Innovations and Stock Market Bubbles
Marketing Science, Articles in Advance, pp. 1–23, ©2018 INFORMS 3
We do not claim to establish causal relations in our
paper. Instead, our goal is to provide a number of sta-
tistical tests to study stock market bubbles across the
largest set of innovations used in this literature. As
is the case with most papers that focus on first-order
research questions, we are only able to answer a sub-
set of questions that readers might deem important.
The data collection for this project was laborious and
extensive. Data for innovations from the 19th century
and from the earlier part of the 20th century were chal-
lenging to obtain, and in many cases extensive manual
searches had to be performed in newspapers and other
contemporaneous sources. We nevertheless hope that
our insights into the economic consequences of inno-
vation will serve as impetus for additional research in
this area.
The rest of the paper proceeds as follows. In Sec-
tion 2, we present an overview of the theory that
supports the existence of stock market bubbles, and
we review the literature that links innovations with
bubbles. We also present theoretical arguments that
support the relation between individual innovation
characteristics, stock market bubbles, and external fi-
nancing. In Section 3, we discuss the economic con-
sequences of bubbles. In Section 4, we describe the
data collection process and empirical methodology.
Section 5presents the results, and Section 6concludes
with a discussion of the implications of our findings
for researchers, managers, and policy makers.
2. Innovation and Stock Market Bubbles
We begin this section with a discussion of the economic
theory of bubbles. We continue with a brief overview
of the literature that links innovation and bubbles, and
conclude with a set of empirical predictions that link
bubbles with specific innovation characteristics.
2.1. Asset Price Bubbles
Bubbles occur when the price of an asset exceeds its
intrinsic value by a significant amount and for a suffi-
ciently long period of time. The literature in financial
economics distinguishes between “behavioral” bub-
bles and “rational” bubbles (Scherbina and Schlusche
2014). Behavioral bubbles occur when investors deviate
from the rational expectations model, perhaps because
they overreact to news or fail to update their beliefs in
a Bayesian manner. By contrast, rational bubbles occur
despite all agents being perfectly rational and having
access to the same information set.
We adopt the rational bubble approach because it
is based on the rational expectations paradigm pre-
vailing in financial economics. Fama (1998) argues that
researchers should not reject rational expectations until
(or unless) the alternative (behavioral) paradigm can
do a better job at explaining existing empirical obser-
vations within a unified body of knowledge. While the
literature in behavioral finance has evolved over the
past 20 years, to our knowledge, there is no unified
behavioral theory of bubbles. Indeed, Scherbina and
Schlusche (2014) discuss no fewer than 18 papers
grouped into four different categories of behavioral-
based models of bubbles. In turn, each category is
based on a unique set of underlying cognitive biases
(such as self-attribution or conservatism). Collectively,
these 18 papers provide little guidance for our research
because there is no systematic way to determine, ex
ante, which cognitive biases apply to various innova-
tions or time periods. By contrast, the rational bubble
approach allows us to develop a parsimonious set of
testable predictions that apply uniformly to all innova-
tions in our sample.
A common theoretical explanation for rational bub-
bles is the mathematical, discrete-time model of
Blanchard and Fischer (1989). Versions of this model
have been used in other papers, including those by
Blanchard and Watson (1982), Froot and Obstfeld
(1991), Rappoport and White (1993), White (1990), and
Scherbina and Schlusche (2014). The model explains
how bubbles may arise even if all economic agents act
rationally. Blanchard and Fischer (1989, pp. 213–223)
posit that the price of an asset in one period is equal
to the discounted value of its expected price and cash
flows during the next period. They then seek to find
the theoretical value of the price during the current
period, expressed only in terms of future cash flows
and discount rates.
Blanchard and Fischer (1989) show that the solu-
tion to their model is not unique. There are, in fact,
an infinite number of solutions. One of them (the fun-
damental solution) is the well-known present value of
future cash flows, which defines the intrinsic value of
an asset. All other solutions include a “bubble” compo-
nent: a positive quantity that is artificially added to the
intrinsic value. We know very little about this bubble
component. Blanchard and Fischer’s (1989) theory tells
us only that it is expected to grow at a rate r, equal to
the discount rate of the asset. This property is not very
useful because it allows for an infinite number of out-
comes. For example, any random number that grows
through time at rate ris a potential bubble. Ideally, we
would like to obtain more specific properties for this
bubble so that we may learn something about its mag-
nitude and determinants.
Fortunately, a special case of the rational bubble the-
ory allows us to narrow down the set of bubble solu-
tions and obtain useful testable implications. This spe-
cial case was proposed by Froot and Obstfeld (1991),
who introduced the concept of intrinsic bubbles, a subset
of rational bubbles. With intrinsic bubbles, the magni-
tude of the bubble component is proportional to the
contemporaneous value of the asset’s cash flows (or
to the value of another proxy that is correlated with
cash flows).
Sorescu et al.: Two Centuries of Innovations and Stock Market Bubbles
4Marketing Science, Articles in Advance, pp. 1–23, ©2018 INFORMS
We mentioned earlier that the solution to the
Blanchard and Fischer (1989) model may or may not
include a bubble. While both scenarios are theoretically
possible, only one of them will be observed in reality—
either the asset price includes a bubble or it does not.
Mathematics alone cannot help us determine which of
these scenarios will prevail. For this, we need to know
more about the underlying economic conditions.
In Web Appendix A, we provide a simplified ver-
sion of the Blanchard and Fischer (1989) model, along
with an economic interpretation of their solution. We
also discuss the intuition behind Froot and Obstfeld’s
(1991) theory of intrinsic bubbles and explain how its
predictions relate to our paper. We argue that bubbles
are more likely to occur when the following four con-
ditions are met: (1) the asset is expected to live for an
indefinite period of time (as is the case with stocks),
(2) no bubble-free close substitute exists for the asset,
(3) investors have exploding expectations about the
future cash flows generated by the asset, and (4) short
selling is costly or subject to constraints. When we over-
lay the theory of intrinsic bubbles on the basic model of
Blanchard and Fischer (1989), we obtain the additional
prediction that the magnitude of a bubble is propor-
tional to a proxy for the cash flows generated by the
asset at the time when the bubble occurs. These eco-
nomic implications form the basis for our predictions
about the determinants and consequences of bubbles
associated with innovations. We present these predic-
tions in Sections 2.3 and 3, respectively.3
2.2. Relation Between Innovation, External
Financing, and Bubbles
A review of the bubble literature reveals several aspects
that warrant further inquiry. First, a few authors note
that stock market bubbles occur during industrial rev-
olutions; however, this observation is made primarily
in case studies, books, and in the popular press (Gross
2009, Kindleberger and Aliber 2005, Perez 2002, Shiller
2015, Wood 2006).4Perhaps because of the tendency
to look for bubbles during these periods of intense
technological progress, only a limited set of bubbles
has been examined in academic studies. For instance,
Campbell (2012) studies the British railway bubble
of the 1840s, Ofek and Richardson (2003) study the
Internet bubble of 1999–2000, and Pástor and Veronesi
(2009) study the Internet bubble and the American rail-
roads before the Civil War, while Frehen et al. (2013)
establish a link between the stock market bubbles of
1720 and innovations that at the time were taking place
in the insurance and Atlantic trade industries.
Second, with the exceptions mentioned above, this
literature tends to focus on marketwide bubbles, rather
than on industry-specific bubbles spurred by individ-
ual innovations. Moreover, academic papers in this
stream tend to view innovation as an undifferenti-
ated aggregate output, which precludes the study of
product-specific characteristics. For example, Nicholas
(2008) evaluates the extent to which the stock mar-
ket during the 1920s responded to contemporaneous
changes in the intellectual capital of U.S. firms, mea-
sured using citation-weighted patents assigned to each
firm. He finds that excess stock returns during 1928–
1929 are positively related to firms’ intellectual capi-
tal, but does not evaluate the relation between stock
returns and specific products.
Third, the literature does not provide a unified view
on why bubbles might be associated with innovations,
and why some bubbles might be larger than others.
This could be a consequence of different philosoph-
ical perspectives regarding the underlying causes of
bubbles. For example, several authors take the view
that bubbles are caused by irrational investors (e.g.,
Ferguson 2008, Perez 2002, Shiller 2015). Others adopt
a more rational view of bubbles but tend to focus on
only one or two bubbles and, therefore, lack context to
systematically examine common patterns across inno-
vations. As a result, the rationale presented in support
of the bubble varies from investor myopia (Campbell
2012) to changes in the systematic risk of parent firms
(Pástor and Veronesi 2009).
A number of questions remain unanswered. Perhaps
the most salient relates to the scale of the relation
between innovation and bubbles: Does this relation
extend to all innovations or is it confined to technolog-
ical revolutions? Because innovations vary on a contin-
uum in terms of technological advances and benefits
they bring to consumers, it is important to understand
where on this continuum the boundary that could trig-
ger a bubble might lie.
What happens after the bubble is also of interest.
A number of authors have argued that the collapse
of bubbles is disruptive to economic activity (Perez
2002, Wood 2006), but are there other, more desirable
outcomes? Rapp (2014, p. 21) proposed that “when
stock prices increase, firms can raise cash from exist-
ing and new investors at lower costs.” Thus, bubbles
could create a favorable environment for firms to raise
new equity capital during the commercialization phase
of new products. However, there is no systematic evi-
dence that supports this assertion. Finally, we know
very little about the longer-term prospects of products
whose parent firms are subject to bubbles. Do bubbles
help or hurt the visibility and eventual diffusion of
innovations?
We address some of these open questions using a
novel data set that spans multiple industries across the
19th and 20th centuries. In Section 2.3, we leverage the
theories proposed by Blanchard and Fischer (1989) and
by Froot and Obstfeld (1991) to make predictions on the
determinants and consequences of bubbles associated
with innovation.
Sorescu et al.: Two Centuries of Innovations and Stock Market Bubbles
Marketing Science, Articles in Advance, pp. 1–23, ©2018 INFORMS 5
2.3. Which Innovations Are More Likely to Be
Associated with Bubbles?
We mentioned earlier that the literature in financial
economics has not examined how differences in inno-
vation characteristics may impact stock prices. By con-
trast, research in marketing has examined the differen-
tial effect of radical versus incremental innovation on
stock prices. When compared with incremental innova-
tions, radical innovations—defined as “new products
that (1) incorporate substantially different technology
from existing products and (2) can fulfill key customer
needs better than existing products” (Chandy and
Tellis 1998, p. 475)—are associated with stronger long-
term stock performance and also with higher firm risk
(Pauwels et al. 2004, Sood and Tellis 2009, Sorescu
and Spanjol 2008). The overarching conclusion from
this stream of research is that the superior benefits
provided by radical innovation eventually result in a
broader adoption of these products, which translates
into higher revenues and profits.
To understand the relation between bubbles and spe-
cific innovation characteristics, we return to the the-
oretical framework of intrinsic rational bubbles pre-
sented in Section 2.1. Recall that from Blanchard and
Fischer (1989) and Froot and Obstfeld (1991) we had
derived a set of conditions that are more conducive to
the emergence of bubbles. We now turn to interpret-
ing these conditions within the context of our current
research.
Two of the four conditions that underlie the for-
mation of bubbles are met by all innovations in our
sample. Specifically, the first condition—indefinitely
lived assets—is met by all stocks. Unless a firm is
in bankruptcy, the expectation is that it will con-
tinue to operate for the foreseeable future. The fourth
condition—costly short selling—is also likely to be met
by all stocks in our sample. Danielsen and Sorescu
(2001) provide several reasons why investors find it dif-
ficult to establish short positions in the stock market. For
example, investors do not immediately get access to the
proceeds from short selling, investors face high search
costs for finding a willing security lender, and there is
always a risk of a “short squeeze,” that is, of a premature
request to liquidate the short position before investors
can earn a profit (also see Akbas et al. 2017).
The second and third conditions are met by some
but not all firms. The second condition—the absence of
a close substitute in the stock market—is more likely to
be met in the case of innovations whose characteristics
are very different from those of existing products. By
definition, radical innovations meet these criteria (at
the product category level), since they provide signifi-
cantly higher consumer benefits and incorporate a new
technology relative to existing products. Because of
these characteristics, parent firms that commercialize
radical innovations are exposed to risks previously
unknown to investors, which cannot be easily hedged
using existing assets. The more radical the innova-
tion, the fewer close-substitute stocks are available in
the stock market.5Because close-substitute stocks are
necessary to arbitrage away mispricing and prevent
the bubble from growing, a lower number of close-
substitute stocks may lead to bigger bubbles. Thus, our
first prediction is that the higher the level of radicalness
of innovations, the larger the magnitude of the innovation
bubble.
The third condition—exploding expectations of fu-
ture cash flows—is more likely to be satisfied by in-
novations believed to offer quasi-unlimited revenue
potential. We propose that network goods meet this
condition. Network goods and services are those
whose value to adopters increases as a function of the
number of other people who use them (e.g., Gupta
et al. 1999, Hall 2005). Direct and indirect network
externalities have been distinguished in the literature
(Katz and Shapiro 1986). The steam engine exemplifies
direct externalities, because the utility derived from
rail transportation increased along with its diffusion
(the size and geographic reach of the rail network).
A good example of indirect externalities is given by
the recent growth in the app industry, spurred by
smartphones and tablets. In turn, apps provide smart
devices with increased functionality, arguably broad-
ening those devices’ appeal and potentially strength-
ening their diffusion (e.g., Stremersch et al. 2007).
Assessing the future cash flows of an innovation is
particularly difficult if it requires an estimation of the
magnitude of network effects. Network effects could
exponentially increase the adoption of a technology
within and across industries, indicating that the inno-
vation has the potential to generate exploding future
cash flows. Investors—particularly optimistic ones—
will factor this possibility into their valuation. There-
fore, we suggest that the magnitude of a bubble increases
with an innovation’s potential for network externalities.
Finally, we derive a third testable implication from
the intrinsic bubble model of Froot and Obstfeld (1991).
Their research suggests that the magnitude of an intrin-
sic bubble is proportional to the contemporaneous cash
flows generated by the underlying innovation, or to
any exogenous variable that is correlated to these cash
flows. In support of this prediction, Froot and Obstfeld
(1991) provide macro-level evidence of intrinsic bub-
bles obtained from aggregate stock returns and divi-
dends in the United States during the period from 1900
to 1988.
Recall that the intrinsic bubble theory requires that
we measure innovation-specific cash flows that are con-
temporaneous with the bubble. In the absence of specific
information about cash flows—which are not directly
observable—investors are likely to anchor their valua-
tions on metrics that are visible and that, in their opin-
ion, might be correlated to cash flows. We propose that
Sorescu et al.: Two Centuries of Innovations and Stock Market Bubbles
6Marketing Science, Articles in Advance, pp. 1–23, ©2018 INFORMS
the public visibility of the innovation around the time
of the bubble is a good exogenous proxy for the extent
to which it has been adopted by the public, and there-
fore for the cash flows that it generates. To develop
this proxy, we use the Google Books Ngram Viewer, an
Internet search engine that provides annual frequen-
cies of any search terms that appear in print up to 2008.
We obtain from Google Ngram the annual frequencies
for each innovation at the time of the bubble.
While admittedly imperfect, the Google Ngram
proxy has the distinct advantage of not suffering from
retrospective bias because it contains information that
was publicly known at the outset of the bubble. In ad-
dition, Google Ngram data are available for all inno-
vations in our sample, which provides us with a con-
sistent proxy for cash flows that is comparable across
several innovations and time periods. In turn, this con-
sistency facilitates cross-sectional analysis in a sample
that is otherwise quite heterogeneous. Thus, our third
prediction is that the magnitude of a bubble increases with
the public visibility of the innovation as captured by its
contemporaneous Google Ngram frequency of mentions in
printed books.
In sum, we predict that the magnitude of innovation
bubbles is increasing in the degree of radicalness, in
the potential for network externalities, and in the con-
temporaneous visibility of the innovation. In Section 3,
we examine the consequences of bubbles for the parent
firms and for the public visibility and eventual diffu-
sion of the innovation.
3. What Are the Consequences of
Bubbles?
3.1. The Economic Value Added of Bubbles
The intrinsic bubble theory of Froot and Obstfeld
(1991) is based on the premise that when bubbles oc-
cur, they do so because the underlying asset—in this
case, the stocks of parent firms—adds significant value
to the economy. This prediction is corroborated by
Olivier (2000, p. 133), who argues that “the real impact
of a [rational] bubble depends on the type of asset
that is being speculated on. Speculative bubbles on
equity raise the market value of firms, thus encourag-
ing entrepreneurship, firm creation, investment, and
growth.” Therefore, we expect innovations that are sub-
ject to bubbles to have, on average, positive net present value
(NPV) despite the presence of bubbles. We test this predic-
tion in our paper, with NPV measured as the abnormal
returns earned by the stocks of parent firms, from the
beginning to the end of the bubble.
3.2. Capital Raised in the Stock Market
A critical question of interest to managers is how to
raise the capital needed to finance innovations. This
is particularly true for innovations that are fueled by
new technologies, which require significant capital
investment. While it might be natural today to assume
that this new capital could be easily raised by issuing
new equity, historical evidence suggests that this was
not always the case. Indeed, prior to the mid-1920s,
innovations were financed mostly with private capital
raised from the owners’ personal network (Lamoreaux
et al. 2007, O’Sullivan 2007). The stock market began
to play a more significant role in financing innovation
beginning with the aircraft and radio industries in the
1920s (O’Sullivan 2007). Since that time, firms have
increasingly relied on equity capital raised from public
sources.
We expect that the presence of bubbles could make
it easier for parent firms to raise equity capital because
one consequence of overvalued stocks is a temporary
reduction in the firm’s cost of equity during the bub-
ble period. In turn, this could make it attractive for
managers to raise new equity capital in the form of
initial public offerings (IPOs) or seasoned equity offer-
ings (SEOs). This new capital would be raised by issu-
ing overvalued shares, which is favorable to the firm’s
existing shareholders, but not to its new shareholders.
Although managers would normally want to issue
new equity when their stock is overvalued, most of the
time they are not able to do so without paying a penalty
because of adverse selection. Indeed, investors usually
view SEOs as a signal of overvaluation, and a mere
announcement of an SEO typically results in an imme-
diate decline in the market value of the firm (see, e.g.,
Myers and Majluf 1984). In a nonbubble environment,
this decline would offset any benefits obtained from
issuing overvalued stock. The bubble environment,
however, might be different: it might provide a unique
opportunity for managers to raise cheap equity capi-
tal without worrying about sending negative signals
to the market. If this were true, we would expect par-
ent firms to raise an abnormally high amount of equity
capital during bubble periods. We provide empirical
evidence on this issue in our paper.
3.3. Postbubble Trajectory of Innovation
We have so far focused on the consequences of bubbles
for parent firms. We now turn to the potential conse-
quences of bubbles for the innovations themselves. Can
a bubble affect the public visibility, and ultimately the
diffusion, of an innovation, and if so, in what way?
The diffusion literature has examined several mile-
stones in the life cycle of a new product. The first mile-
stone is the takeoff time, defined by Golder and Tellis
(1997, p. 257) as “the point of transition from the intro-
ductory stage to the growth stage of the product life
cycle,” a point when a dramatic increase in sales occurs.
Determinants of the takeoff time include a reduction in
the price of the innovation and the level of market pen-
etration (Golder and Tellis 1997). Another determinant
of takeoff that is particularly relevant to our context is
Sorescu et al.: Two Centuries of Innovations and Stock Market Bubbles
Marketing Science, Articles in Advance, pp. 1–23, ©2018 INFORMS 7
the performance of the parent firm’s stock in the early
commercialization stages: a strong positive abnormal
stock performance has been found to predict takeoff
in the subsequent year (Markovitch and Golder 2008),
suggesting that takeoff could follow shortly after a bub-
ble has emerged. Finally, the parent firm’s ability to
raise cheap equity capital during a bubble essentially
reduces the cost of the innovation, which can poten-
tially reduce its price and shorten its time to takeoff
(Golder and Tellis 1997). Therefore, we expect that the
magnitude of the bubble and the capital raised during the
bubble are positively related to how quickly takeoff occurs.
We do not measure diffusion directly. Rather, we use
as a proxy the time series of visibility obtained from
Google Ngram, beginning with the commercialization
date of each innovation. Just as visibility could serve as
an ex ante proxy for the innovation’s cash flows, it can
also serve as an ex post proxy for its diffusion. Using
this proxy, we estimate the takeoff of each innovation as
the inflection point in the time series of Google Ngram
visibility. In Web Appendix D we show, for a subsam-
ple of 25 innovations, that the takeoff estimated from
visibility correlates highly with the takeoff estimated
from product adoption data.
The literature on new product diffusion also makes
predictions about the shape of the diffusion curve after
takeoff. Initial models predicted a monotonic increase
in sales up to the peak of growth, followed by a
slowdown (e.g., Peres et al. 2010). Other authors have
argued that the diffusion curve follows a chasm and
saddle pattern, where sudden decreases in sales fol-
low the initial rise (Golder and Tellis 2004, Mahajan
and Muller 1998). In all cases, as newer technolo-
gies become available, the diffusion curve eventually
plateaus and starts declining. The point at which it
plateaus has been assumed to be the point of market
saturation, but researchers have focused less on this
saturation point compared to the takeoff point.
We argue that the same characteristics of a bubble
that lead to an accelerated takeoff are also associated
with a higher market saturation point. A lower inno-
vation cost obtained from the parent firm’s ability to
raise cheap capital will not only lead to a faster takeoff
but also to a broader adoption of the innovation. Thus,
we expect that the magnitude of the bubble and the capital
raised during the bubble are positively related to the visibil-
ity that the innovation achieves after takeoff. This visibility
is proxied by the maximum Google Ngram value that
the innovation attains in its lifetime.
4. Data and Methods
4.1. Selection of Innovations Included in
Our Sample
Assembling a comprehensive data set of innovations
introduced during the last two centuries is a challeng-
ing task. We start with the largest set of radical inno-
vations documented in the literature, a list of 64 new
products introduced between 1864 and 1998 provided
in Chandy and Tellis (2000). The authors used the his-
torical approach to data collection (Golder 2000) and
consulted “more than 250 books and 500 articles in peri-
odicals” to construct their sample (Chandy and Tellis
2000, p. 5). Therefore, we believe that this list consti-
tutes a good initial sampling frame for our study. Of the
64 innovations mentioned in Chandy and Tellis (2000),
we were able to identify and collect data for 43 innova-
tions commercialized between 1886 and 1998.6
In our review of the literature on technological rev-
olutions, we also encountered a few additional inno-
vations that have been singled out by economists
as being notable in their ability to spur economic
growth. For instance, Shiller (2015, p. 125) mentions
the steam engine train, motion pictures, and the Inter-
net (or, more specifically, the World Wide Web) as
major innovations that ushered what he calls “a new
era.” Through this archival process, we add eight addi-
tional innovations to the Chandy and Tellis (2000) list:
the steam engine train, the telegraph, motion pictures,
rayon, the airplane, the Internet, the GPS receiver, and
the smartphone. Table WA-B3 in Web Appendix B
details the sources used to identify these innovations.
Our final sample consists of 51 innovations.
4.2. Identifications of Parent Firms and of
Financial Data Used for the Bubble Tests
After assembling our sample of innovations, we iden-
tify the date when each innovation was commercial-
ized, as well as the parent firms involved in the commer-
cialization process. To do so we use extensive archival
searches encompassing a large number of books and
articles on the history of technology, major U.S. and UK
journals that date back to the 19th century (such as the
Wall Street Journal and the London Times), and academic
articles that examine the diffusion of innovations (e.g.,
Agarwal and Bayus 2002, Golder and Tellis 1997).
From the set of parent firms, we select those with
available stock and dividend data. For most innova-
tions in the 20th century, company-specific data are
available from the Center for Research in Security
Prices (CRSP) beginning with the year 1925. For ear-
lier innovations, we turn to the Cowles Foundation
database (https://som.yale.edu/faculty-research/our
-centers-initiatives/international-center-finance/data/
historical-cowles), which covers the period from 1871
to 1938. Unlike the CRSP, Cowles provides data on
industry indices, not on individual companies. If we are
able to identify a clear industry index that corresponds
to a particular innovation (for example, “automobile”),
we use the Cowles price and dividend data for that
index, and the set of parent firms for that particular
innovation is deemed to be the set that the Cowles
researchers had chosen to include in their index.
If the Cowles index is too broad (for example, “appli-
ances” instead of “refrigerator”), we identify parent
Sorescu et al.: Two Centuries of Innovations and Stock Market Bubbles
8Marketing Science, Articles in Advance, pp. 1–23, ©2018 INFORMS
Table 1. Sources of Stock and Dividend Data
Data source Description
CRSP data set Monthly, stock-level data are available from the Center for Research in Security Prices at the
University of Chicago from December 1925 to the present. Stock-level data are used to construct a
price index and monthly dividend series for each innovation. The monthly price index is
constructed using the value-weighted average return (excluding dividends) of the stocks included
in the index. The monthly dividend series for each index is constructed by first computing each
stock’s monthly dividend yield as the difference between its total monthly return and its monthly
return excluding dividends. The monthly dividend yield of the index is then computed as the
value-weighted average of the component stocks’ dividend yields. Monthly dividends in terms of
the price index are computed as the product of the index’s dividend yield and the value of the price
index at the end of the prior month.
London Stock Exchange (Old LSE
data set)
The Old LSE data set was produced by compiling hard copy entries from the Investors Monthly Manual,
a record of the London Stock Exchange for the period from 1871 to 1930. Common stock prices and
dividends from this data set are used to construct the required price indexes and dividend series in
a manner consistent with those that are based on CRSP data.
The Cowles Commission for
Research in Economics
The Cowles data set, produced from the Cowles Commission’s third monograph (1939), contains
industry-level common stock data series for the period 1871 to 1938. We use two price indexes from
these data: a price index that excludes dividends and a total return index that includes both
dividends and price appreciation. As we do with the CRSP data, we use the difference in the
returns of these two indexes to construct a monthly, industry-level dividend series denominated in
terms of the price index.
Compustat Global Monthly, stock-level data for foreign stocks are available from Compustat Global from January 1985 to
the present. Variables constructed from this data set are computed in the same manner as described
above for the CRSP data. This data set was used to collect stock price and dividend data for
Japanese firms that did not trade in the United States during the time period of interest.
New York Stock Exchange (Old
NYSE data set)
The Old NYSE data set was produced from data used in Goetzmann et al. (2001). Common stock
prices and dividends from this data set are used to construct the required price indexes and
dividend series in a manner consistent with those that are based on the CRSP data.
Note. This table presents the data sources used to collect stock and dividend data for the bubble tests.
companies from historical sources and collect data on
individual stocks from the Historical London Stock
Exchange and Historical New York Stock Exchange da-
tabases, both of which are available at Yale University
in electronic format. Finally, data not available in elec-
tronic format are hand collected from contemporane-
ous newspapers in the United States and the United
Kingdom. One exception is the data set on the UK rail-
road industry from the 19th century, which was pro-
vided to us by Gareth Campbell (Campbell 2012).
Table 1describes our data sources, and Table 2pro-
vides a list of the 51 innovations in our sample, along
with the years of commercialization and the corre-
sponding portfolios of parent firms. A full descrip-
tion of the history of these innovations is available on
request from the authors. For innovations not covered
by the CRSP or Cowles, we occasionally have to confine
our analysis to a single firm, because of data availabil-
ity. For example, for the telegraph industry, we study
only the stock of Western Union, as we are unable to
identify data for other companies. Our results are not
likely to be biased by these single-firm portfolios.7
Our portfolios do not contain the complete set of
early entrants in each industry. While this is a limi-
tation caused by data availability, our goal here is to
create clean portfolios for the measurement of bub-
bles, rather than to extensively study the entry process
into new industries. As a result, we are primarily con-
cerned with building portfolios that have significant
exposure to the focal innovation (and little exposure
to everything else), rather than to make the portfo-
lios as comprehensive as possible. To accomplish this
goal, we endeavor as much as possible to include in
our test portfolios only firms that are pure plays, in
the sense that their product portfolio includes primar-
ily the innovation and not much else. To the extent
to which our test portfolios may occasionally include
firms that are not pure plays (such as multiproduct
firms that commercialize other products in addition
to the focal innovation), the inclusion of such firms
would merely bias our results toward zero, because
these firms would have a smaller exposure to the eco-
nomic forces that ignite the bubble.8
4.3. Characteristics of Innovations That Are More
Likely to Be Associated with Bubbles
Previously, we identified three innovation characteris-
tics that increase the probability of a bubble: the inno-
vation’s degree of radicalness, its potential to generate
network effects, and the extent of its public visibility.
Ideally, we would like to collect contemporaneous data
for each of these three characteristics to minimize the
risk of a retrospective bias. However, we could only do
so in the case of visibility. We measure contempora-
neous visibility using the Google Ngram data, which
Sorescu et al.: Two Centuries of Innovations and Stock Market Bubbles
Marketing Science, Articles in Advance, pp. 1–23, ©2018 INFORMS 9
Table 2. Firms Included in the Bubble Analysis of Each Innovation
Innovation Comm. year Firms (and initial year) included in index
Steam engine train 1825 Campbell (2000) Railway Index
Telegraph 1845 Western Union (1865)
Incandescent vacuum lamp 1880 Cowles Electrical Equipment Index components: Edison General Electric/General Electric
(1890)
Automobile 1886 Cowles Automobiles and Trucks Index components: General Motors (1912), Chrysler (1914),
Studebaker (1912), Willys-Overland (1915), Chandler Motor Car (1916), Saxon Motor Car
(1916), Stutz Motor Car (1916), White Motor (1916), Pierce-Arrow Motor Car (1917), Fisher
Body (1918), Hupp Motor (1918), Mack Trucks (1920), Packard Motor Car (1920)
Portable camera 1888 Eastman Kodak (1904)
Disk phonograph 1894 Gramophone company (1904)
Motion pictures 1898 Cowles Theater and Motion Picture Index components: Famous Players-Lasky (1919, later
renamed to Paramount Publix), Loew’s (1920), Radio-Keith-Orpheum (1920)
Photoelectric scanning fax 1907 Cowles Utilities—Telephone and Telegraph Index Components: American Telephone and
Telegraph (1907), New York and New Jersey Telephone (1907), Central South America
Telegraph (1907), Mackay Companies (1907), Pacific Telephone and Telegraph (1909), All
America Cables (1918)
Electric clothes washer 1908 Cowles Electrical Equipment Index components: General Electric (1908), Electric Storage
Battery Co. (1915), National Conduit and Cable (1917), Westinghouse Electric and
Manufacturing (1918)
Electric percolator 1908 Cowles Electrical Equipment Index components: General Electric (1908), Electric Storage
Battery Co. (1915), National Conduit and Cable (1917), Westinghouse Electric and
Manufacturing (1918)
Electric toaster 1908 Cowles Electrical Equipment Index components: General Electric (1908), Electric Storage
Battery Co. (1915), National Conduit and Cable (1917), Westinghouse Electric and
Manufacturing (1918)
Rayon 1910 Cowles index: Industrial Rayon Corp. (1926), Tubize Artificial Silk Co. of America (1926),
Snia Viscosa Co. (1926), Celanese Corp. of America (1927), Tubize Chatillon Corp. (1930)
Refrigerator 1916 General Electric (1925), Kelvinator (1926)
Airplane 1919 Bendix Aviation (1925), Curtiss Aeroplane (1927), Wright Aeronautical (1925), Curtiss
Wright Corporation (1929), National Air Transport (1929), United Aircraft and Transport
(1929)
Radio 1919 RCA (1925), AT&T (1925), Westinghouse Electric and Manufacturing (1925)
Electric typewriter 1920 IBM (1928)
Electric blanketa1930 General Electric (1928)
Electric dishwashera1930 General Electric (1928)
Electric shaver 1930 Remington Rand (1928)
Electric garbage disposer 1935 General Electric (1928)
TV 1936 Admiral (1945), Crosley (1940), Emerson (1945), Farnsworth Television and Radio (1943),
General Electric (1940), Galvin Manufacturing/Motorola (1946), Philco (1940), RCA
(1940), Westinghouse Electric (1940), Zenith (1940)
Fluorescent light bulb 1938 General Electric (1936), Westinghouse Electric and Mfg (1936)
FM radio 1940 RCA (1935), Westinghouse Electric and Mfg (1935), Zenith Radio (1935), Philco (1940),
Galvin Manufacturing/Motorola (1946)
Ballpoint pen 1945 Eversharp (1946)
Magnetic tape player 1947 Minnesota, Mining and Mfg (1946)
Microwave oven 1947 Raytheon (1958), Litton (1958)
NTSC color TV 1954 General Electric (1952), Philco (1952), RCA (1952), Raytheon (1952), Hoffman Electronics
(1955)
Electric can opener 1956 Westinghouse Electric (1950)
Videocassette recorder 1956 North American Phillips (1950), RCA (1950), Magnavox (1950), Ampex (1959)
Electronic watch 1957 Bulova Watch (1955), Hamilton Watch (1955), Elgin National Watch (1955)
Cassette 1964 North American Phillips (1962), Memorex (1968), Sony (1970), Fuji (1972), Hitachi (1972),
TDK Electronics (1975)
Electronic desktop calculator 1964 Texas Instruments (1959), Hewlett-Packard (1959), Friden (1959), Wang Laboratories (1968)
Electronic pocket calculator 1968 Texas Instruments (1965), Hewlett-Packard (1965), Canon (1965), Bowmar Instrument
(1965), Craig (1969), Toshiba (1972)
Dot-matrix printer 1970 Centronix Data Computer (1972)
Single-player video game 1972 Magnavox (1969), General Instrument (1969), Centuri (1972), Bally Manufacturing (1972),
Williams Electronics (1981)
Digital LED watch 1975 Texas Instruments (1973)
Disposable shaver 1975 Gillette (1975), Schick (1975), Bic Pen (1975)
Sorescu et al.: Two Centuries of Innovations and Stock Market Bubbles
10 Marketing Science, Articles in Advance, pp. 1–23, ©2018 INFORMS
Table 2. (Continued)
Innovation Comm. year Firms (and initial year) included in index
Personal computer 1975 Apple (1980), Commodore (1978), IBM (1978), Intel (1978), RadioShack (1978), Lotus
Development (1983), Ashton Tate (1983), Compaq (1983)
Laser printer 1976 Hewlett-Packard (1974), IBM (1974), Xerox (1974), Canon (1974)
Laser disc player 1978 MCA (1976), Pioneer (1976), Sony (1976), Phillips (1976)
Mobile phone 1979 AT&T (1980), Motorola (1980), Ericsson (1981), Interdigital (1981), SBC (1984), Nokia (1988),
Qualcomm (1991)
Camcorder 1983 Hitatchi (1983), JVC (1983), Panasonic (1983), Sony (1983)
Compact disc player 1983 Panasonic (1981), Philips (1981), Sony (1981), Toshiba (1981)
Portable computer 1983 IBM (1975), Radioshack (1975), Hewlett-Packard (1975)
Digital camera 1989 Fuji (1986), Nikon (1986), Casio (1986), Olympus (1986)
Internetb1991 List of Internet IPOs included in Loughran and Ritter (2004)
Minidisc player 1992 Sony (1990)
Digital video disc player 1997 Sony (1990)
Digital high-definition TV 1998 Zenith (1990), Sony (1990), Panasonic (1990), Hitatchi (1990), Technicolor (1999)
GPS receiver 2000 Garmin (2000), Lowrance (2000), Trimble (2000)
Smartphone 2000 Apple (2002), Ericsson (2002), Handspring (2002), Nokia (2002), Research in Motion (2002),
Google (2004)
Notes. This table presents the names of the firm(s) included in the test portfolio of each innovation. The year of each firm’s entry into the
portfolio is included in parentheses. For the Internet industry, we provide the source of the index employed. The first column shows the name
of each innovation and the second column shows the year of commercialization.
aBecause the electric blanket and the electric dishwasher were introduced in the same year (1930), and because the test portfolio for both
products is limited to General Electric, we can only perform a joint bubble test for these two products. If a bubble is detected, it could be
attributed to either one of the two innovations or to both.
bWe use the term Internet to refer to the World Wide Web (WWW) enhancement to the Internet technology. Tim Berners-Lee invented the
WWW in 1990, and the concept was first commercialized in 1991. Berners-Lee’s invention consisted of three new technologies that remain
in place today: HTML, URL, and HTTP. These technologies allowed for a user-friendly communication among computers, which made the
Internet accessible to the general public.
shows the frequency with which a search term appears
in published books through time. To measure radical-
ness and the potential for network effects, we use rat-
ings obtained from participants in a Mechanical Turk
(MTurk) study. While these ratings are provided post
hoc (and could be affected by memory bias), they pro-
vide a consistent measure across our diverse sample of
innovations. We describe below the MTurk and Google
Ngram measures and the two data sources used to
obtain them.
4.3.1. MTurk Study. We asked 100 participants in an
MTurk study to rate the 51 innovations included in our
sample on several dimensions relevant to our research.
We requested that the participants be U.S. based, have
a Human Intelligence Task (HIT) approval rate of 95%
or higher, and have “masters” qualifications in MTurk.
Participants were each paid $5 and took an average of
52 minutes to complete the survey.
At the beginning of the survey, we provided spe-
cific definitions for our characteristics of interest, along
with examples of innovations that were not part of the
list to be evaluated. We provided the most commonly
referenced definitions of radical innovation, direct and
indirect network effects, and disruptive innovation. We
present these definitions in Web Appendix C. Dis-
ruptive innovations are initially inferior to existing
products. However, through time, they improve and
become the dominant players in the market (Sood and
Tellis 2011). The “disruptive” dimension has not been
studied previously in relation to stock prices, and is
not part of our theoretical framework that links inno-
vation with bubbles. We include this dimension in the
MTurk survey to control for the possibility that any new
product that is fundamentally different from the status
quo might be associated with a bubble, not just radical
innovations. Disruptive innovations, when introduced,
do not provide higher consumer benefits than exist-
ing alternatives (unlike radical innovations). Therefore,
comparing the effects of disruptive and radical innova-
tion in relation to bubbles can yield interesting insights
on how investors value products that provide signifi-
cant increases in consumer benefits.
After providing the definitions and examples of rad-
ical, disruptive, direct network effect, and indirect net-
work effect, we instructed participants to evaluate each
innovation in relation to the existing technology at
the time of its commercialization. For instance, we
told participants that when they evaluate the color TV
on the radical dimension, they should do so in relation
to the black and white TV that was the standard when
the color TV was commercialized. To facilitate this task,
we listed innovations in the order of commercializa-
tion. For each characteristic, we presented respondents
with our list of innovations and asked them to rate their
agreement on a one-to-five scale (where 1 is “strongly
disagree” and 5 is “strongly agree”) that each innova-
tion has that particular characteristic. We also allowed
Sorescu et al.: Two Centuries of Innovations and Stock Market Bubbles
Marketing Science, Articles in Advance, pp. 1–23, ©2018 INFORMS 11
respondents to choose the following option: “I am not
able to make an assessment.” The survey included four
reading checks. Only respondents who correctly com-
pleted all reading checks were able to finish the survey
and obtain payment.
We use the average ratings across the 100 partici-
pants to measure the characteristics of the 51 innova-
tions in our sample. The sample average of these
average ratings is as follows: 4.10 for the radical di-
mension, 3.51 for disruptive, 3.44 for indirect net-
work effects, and 3.25 for direct network effects.9We
acknowledge that these post hoc ratings might not fully
capture the consumer perceptions of these products
at the time when they were introduced. However, we
hope that this preliminary effort, which allows us to
present some cross-sectional evidence on the associa-
tion between innovation characteristics and stock mar-
ket bubbles, will stimulate additional research on the
valuation of distinct types of innovations.
4.3.2. Google Ngram Data. Given the span and scope
of our sample, data on a single measure of diffusion—
such as product-level sales—would be very difficult
to obtain. Instead, to gain preliminary insight into the
adoption of each innovation, we measure its contem-
poraneous visibility in the public domain using the
annual frequency with which the innovation is men-
tioned in printed books. The numerator of this fre-
quency measure is the number of times that a given
word appears in print in all books digitized by Google
in a given year. The denominator includes the total
number of words in all books during the same year.
For example, the Ngram value for the word “radio” in
1929 is 1.92 ×105. Thus, the word “radio” represents
one in every 52,035 words across all books published
that year.
We use Google Ngram data to compute three mea-
sures of visibility of relevance to our study.10 First, to
test the intrinsic bubble model of Froot and Obstfeld
(1991), we use Ngram frequencies measured during
the year when the bubble is detected as a proxy for
contemporaneous cash flows. Second, to measure the
magnitude of postbubble visibility, we use the maxi-
mum value of Ngram frequencies achieved during the
lifetime of the innovation, regardless of date, which
is a proxy for the market saturation point. Third, the
rate of increase in visibility is measured as the distance
in time between the commercialization date and the
inflection point in the time plot of Ngram frequencies.
The inflection point, a proxy for takeoff, is the year with
the largest positive change in the Ngram frequencies.11
To evaluate the extent to which visibility is an appro-
priate proxy for diffusion, we collect product adop-
tion data for a subsample of 25 innovations. Adop-
tion data for 24 of these innovations are obtained from
the Statistical Abstract of the United States, published
annually by the Bureau of the Census. In addition, we
obtain adoption data for the Internet from the World
Bank, which publishes annual measures of Internet
usage per 100 inhabitants. We then compare the adop-
tion data with the visibility data previously obtained
from Google Ngram. The results are presented in
Web Appendix D and in Table WA-D1. Panel A of
Table WA-D1 shows our data sources and the partic-
ular measure of product adoption obtained for each
of these 25 innovations. Panel B of Table WA-D1 com-
pares the takeoff dates computed from visibility data
with the takeoff dates computed from product adop-
tion data. The two measures correlate very highly.
Overall, the results presented in Web Appendix D sug-
gest that the visibility measure computed from Google
Ngram data is a good proxy for product adoption.
4.4. Detecting Stock Market Bubbles
We use the statistical method developed by West (1987)
to detect the presence of bubbles in the time series
of stock prices. For each innovation, we form a single
test portfolio containing all firms shown in the right-
most column of Table 2. For each test portfolio, we
compute a monthly value-weighted price index and,
separately, a monthly value-weighted dividend index,
using as weights the market capitalization of each stock
at the end of the previous month. The Cowles data set
already comes in the form of value-weighted test port-
folios. In the end, this process produces 102 different
time series, two series (dividends and prices) for each
of the 51 innovations.
We perform a bubble test for each innovation using
the price and dividend time series of its respective test
portfolio. We require a minimum of 24 months to esti-
mate the parameters of the model. After a bubble is
detected, we continue the process until the model indi-
cates that a bubble is no longer present. Our empirical
specification is motivated by the model of West (1987),
which begins with the no-arbitrage condition that the
current-period stock price is equal to the discounted
value of the next-period stock price and dividends
PtρtEt[Pt+1+dt+1]+t,(1)
where Etis the expectation operator, ρt1/(1+kt),
and ktis the discount rate for the stock at time t. West’s
(1987) model further assumes that the aggregate divi-
dend of each test portfolio follows a zero-mean, AR(1)
process
dtϕtdt1+νt.(2)
From Equations (1) and (2), we obtain F
t, the funda-
mental value of the stock
F
tδtdt,(3)
where
δtρtϕt
1ρtϕt
.(4)
Sorescu et al.: Two Centuries of Innovations and Stock Market Bubbles
12 Marketing Science, Articles in Advance, pp. 1–23, ©2018 INFORMS
If no bubble is present, the price of the stock, Pt, equals
its fundamental value, F
t. Thus,
Ptδtdt.(5)
According to Equation (5), in the absence of bubbles,
the observed stock price, Pt, should equal to a multi-
ple δtof the current dividend dt, where δtdepends,
as shown in Equation (4), only on the fundamental
parameters that govern the stock’s law of motion: ρt
and ϕt.12 However, if bubbles are present, the observed
stock price Ptcould be significantly higher. The bubble
test consists of comparing δtestimated in Equation (5)
with its intrinsic value estimated in Equation (4). If δt
exceeds this intrinsic value, the stock price Ptis higher
than the present value of its future dividends, implying
that a bubble is present.
Empirically, the bubble test examines whether ˆ
δt
[ˆ
ρtˆ
ϕt/(1ˆ
ρtˆ
ϕt)]. This test requires that we compute
ˆ
δt,ˆ
ρt, and ˆ
ϕt, the empirical estimates of parameters δt,
ρt, and ϕt, and that we do so for all values of t. We
estimate ˆ
ρtfrom Equation (1), ˆ
ϕtfrom Equation (2),
and ˆ
δtfrom Equation (4). The empirical specification
for each of these equations includes an intercept and
an error term.
We define Ht, a test statistic for detecting bubbles, as
follows:
Htˆ
δtˆ
ρtˆ
ϕt/(1ˆ
ρtˆ
ϕt)ˆ
σt,(6)
where ˆ
σt, the standard error of the Htstatistic, is com-
puted using the heteroskedasticity-adjusted method of
Newey and West (1987). This test statistic follows a chi-
square distribution with two degrees of freedom (West
1987). We estimate Hteach month, t, for each innova-
tion, beginning with the 25th month in the time series
of returns and dividends.
Detecting the Beginning and End of the Bubble. A bub-
ble is deemed to form in month tif the Htstatistic
is significantly positive at the 10% level (two-tail test)
during months t,t+1, and t+2. The bubble is deemed
to continue beyond month t+2so long as the Ht
statistic remains significantly positive. Once a bubble
is detected, we allow up to two consecutive months
of insignificant Hstatistics before concluding that the
bubble has ended. The bubble is deemed to end when
the Hstatistic is no longer significantly positive and
remains so for at least two months.
Measuring the Magnitude of a Bubble. We define mis-
pricing as the difference between the observed price of
the test portfolio and the fundamental value predicted
by Equation (3). For each innovation, we compute mis-
pricing for each month, t, during the bubble’s exis-
tence. The peak of the bubble is deemed to occur when
mispricing is highest. We measure the magnitude of
the bubble as the ratio of mispricing to the fundamen-
tal value of the test portfolio, computed at the peak of
the bubble. Its statistical significance is the p-value of
the contemporaneous Htstatistic.
5. Results
We begin by performing a bubble test for each of the 51
innovations in our sample. The results are presented
in Table 3. Our method identifies bubbles in approx-
imately 73% of the innovations studied (37 out of 51
innovations). The results corroborate the presence of
bubbles in a number of cases that had been conjec-
tured in books and in the popular press. For example,
we identify bubbles in the rail, telegraph, automobile,
radio, airplane, and Internet industries. We also iden-
tify bubbles in innovations that had not been men-
tioned previously in the literature, such as the personal
computer or the microwave oven. Finally, we find no
evidence of bubbles in the remaining 14 innovations,
which include the color TV and the electronic watch.
5.1. Economic Value Added
To measure the economic value added of each innova-
tion, we compute the buy-and-hold abnormal returns
(BHARs) of each innovation’s test portfolio from the
beginning to the end of the bubble. BHARs are obtained
by subtracting the buy-and-hold returns of the stock
market from the buy-and-hold returns of the test port-
folio.13 We compute this measure for each innovation.
The statistical significance is obtained from the monthly
time series variation using a t-test of paired differences.
The results are presented in Table 4. The first column
in Table 4lists all 37 innovations for which we have
previously identified bubbles, and the second column
provides the year of commercialization. The third col-
umn in Table 4shows the buy-and-hold returns of par-
ent firms from the beginning to the end of the bubble.
The fourth column shows the buy-and-hold returns of
the market during that same period. The fifth column
shows the BHARs, computed as the difference between
the previous two columns. The last column shows the
p-value for all positive values of BHAR.
Despite the presence of bubbles, we find evidence
of positive economic value added for innovations in
our sample. On average, firms in our test portfolios did
well during the bubbles, both in absolute terms and
by comparison to the market. A hypothetical investor
who purchased a typical test portfolio immediately
prior to the start of the bubble and remained invested
throughout the bubble would have earned, at the end
of the bubble, an average return of 65.01%, compared
to an average market return of only 20.42%. The aver-
age difference of 44.58% is statistically larger than zero
(p0.0054). In approximately 76% of the cases studied
(28 out of 37), the test portfolio was worth more (in
market-adjusted terms) at the end of the bubble than
it was at the beginning. This fraction is statistically dif-
ferent from 1
2(p0.0233).
5.2. Do Bubbles Lead or Lag Visibility
We compare the start date of each innovation bubble
with the takeoff in visibility (computed as the inflection
Sorescu et al.: Two Centuries of Innovations and Stock Market Bubbles
Marketing Science, Articles in Advance, pp. 1–23, ©2018 INFORMS 13
Table 3. Bubbles Associated with the Stock Prices of Innovating Companies
Bubble magnitude
Bubble Bubble Bubble Bubble
Innovation Comm. year detected? start date peak date Value (%) p-value end date
Steam engine train 1825 Yes 8/1843 3/1844 31.10.0001 2/1845
Telegraph 1845 Yes 12/1870 2/1873 75.60.0123 9/1873
Incandescent vacuum lamp 1880 Yes 6/1892 10/1892 16.30.0410 4/1893
Automobile 1886 Yes 3/1919 11/1919 119.50.0166 9/1920
Portable camera 1888 Yes 7/1918 4/1921 101.10.0753 4/1921
Disk phonograph 1894 Yes 7/1906 4/1907 102.60.0014 1/1908
Motion pictures 1898 Yes 9/1921 9/1922 49.80.0002 5/1923
Photoelectric scanning fax 1907 No
Electric clothes washer 1908 No
Electric percolator 1908 No
Electric toaster 1908 No
Rayon 1910 Yes 2/1928 4/1928 24.40.0001 7/1928
Refrigerator 1916 Yes 4/1928 8/1929 369.30.0009 9/1930
Airplane 1919 Yes 1/1928 2/1929 183.90.0024 9/1929
Radio 1919 Yes 1/1928 8/1929 86.50.0252 11/1930
Electric typewriter 1920 Yes 4/1933 7/1935 51.80.0063 12/1935
Electric blanket 1930 Yesa11/1938 11/1938 41.20.0972 10/1939
Electric dishwasher 1930 Yesa11/1938 11/1938 41.20.0972 10/1939
Electric shaver 1930 No
Electric garbage disposer 1935 Yes 11/1944 5/1945 12.00.0931 7/1945
TV 1936 Yes 2/1943 1/1946 50.30.0012 9/1946
Fluorescent light bulb 1938 No
FM radio 1940 Yes 11/1947 5/1948 26.40.0871 3/1949
Ballpoint pen 1945 Yes 1/1955 3/1956 47.70.0001 9/1957
Magnetic tape player 1947 Yes 2/1960 6/1960 62.90.0001 8/1960
Microwave oven 1947 Yes 8/1965 2/1966 59.20.0002 4/1966
NTSC color TV 1954 No
Electric can opener 1956 No
Videocassette recorder 1956 Yes 12/1965 4/1966 79.10.0065 6/1966
Electronic watch 1957 No
Cassette 1964 Yes 8/1969 11/1969 72.90.0001 3/1970
Electronic desktop calculator 1964 Yes 12/1971 10/1973 50.30.0001 6/1974
Electronic pocket calculator 1968 Yes 12/1971 1/1973 46.90.0001 11/1973
Dot-matrix printer 1970 No
Single-player video game 1972 Yes 3/1979 11/1981 171.20.0029 1/1984
Digital LED watch 1975 No
Disposable shaver 1975 Yes 6/1979 8/1979 20.10.0001 8/1979
Personal computer 1975 Yes 7/1982 5/1983 86.20.0031 8/1983
Laser printer 1976 No
Laser disc player 1978 Yes 9/1983 12/1989 186.10.0454 6/1990
Mobile phone 1979 Yes 7/1986 12/1989 101.90.0001 8/1991
Camcorder 1983 Yes 1/1988 7/1988 50.70.0001 6/1990
Compact disc player 1983 Yes 10/1985 6/1987 93.50.0711 6/1987
Portable computer 1983 No
Digital camera 1989 No
Internet 1991 Yes 2/1998 2/2000 54.10.0001 3/2000
Minidisc player 1992 Yes 4/1992 1/1994 43.60.0010 8/1994
Digital video disc player 1997 Yes 12/1997 12/1997 58.50.0665 7/1998
Digital high-definition TV 1998 Yes 3/1999 6/1999 35.70.0936 6/1999
GPS receiver 2000 Yes 4/2007 9/2007 96.70.0001 12/2007
Smartphone 2000 Yes 8/2004 9/2005 134.90.0275 9/2005
Notes. The table shows the extent to which firms associated with each innovation experience stock price bubbles. Each bubble is identified
using West’s (1987) model, described in this paper. When the model detects a bubble, we show the start, peak, and end dates (year/month)
of each bubble. The bubble peak date is the month with the largest difference between the observed price index of the test portfolio and its
predicted price index from the model. At each peak date, we show the bubble magnitude as the percentage difference between observed prices
and predicted values, and the corresponding p-value of the H-statistic from the bubble test.
aBecause the electric blanket and the electric dishwasher were introduced in the same year (1930), and because the test portfolios for both
products consist of the same firm (General Electric), we are only able to perform a joint bubble test for these two innovations. The bubble
detected here could be attributed to either one of the two innovations or to both. We chose to attribute the bubble in the analysis to both
innovations, but the results remain substantively the same if we attribute it to either one of the two innovations.
Sorescu et al.: Two Centuries of Innovations and Stock Market Bubbles
14 Marketing Science, Articles in Advance, pp. 1–23, ©2018 INFORMS
Table 4. Buy-and-Hold Returns for the Beginning to End of the Bubble Period
Buy-and-hold returns from beginning to end of bubble (%)
Innovation Comm. year Innov. stocks Market Difference (BHAR) p-value
Steam engine train 1825 45.20 13.20 32.00 0.0135
Telegraph 1845 54.70 22.90 77.60 0.0037
Incandescent vacuum lamp 1880 4.33 4.69 0.36 0.4599
Automobile 1886 38.80 0.00 38.80 0.1275
Portable camera 1888 88.10 3.70 91.80 0.0144
Disk phonograph 1894 80.20 2.20 78.00 0.0264
Motion pictures 1898 35.40 34.50 0.90 0.2734
Rayon 1910 14.60 7.90 22.50
Refrigerator 1916 54.30 13.30 67.60 0.0321
Airplane 1919 128.90 54.30 74.60 0.1757
Radio 1919 18.50 18.80 37.30 0.0232
Electric typewriter 1920 145.00 140.50 4.50 0.4992
Electric blanket 1930 14.21 0.78 14.99
Electric dishwasher 1930 14.21 0.78 14.99
Electric garbage disposer 1935 12.62 14.66 2.04
TV 1936 18.90 54.00 35.10
FM radio 1940 0.16 4.65 4.49 0.3254
Ballpoint pen 1945 1.80 15.30 13.50
Magnetic tape player 1947 34.47 2.94 31.52 0.1003
Microwave oven 1947 75.60 9.50 66.10 0.0216
Videocassette recorder 1956 3.02 5.90 8.92 0.2777
Cassette 1964 10.10 1.90 12.00 0.3073
Electronic desktop calculator 1964 53.04 15.40 68.44 0.0065
Electronic pocket calculator 1968 97.21 2.79 100.00 0.0021
Single-player video game 1972 36.84 80.74 43.90
Disposable shaver 1975 16.04 11.77 4.27 0.1837
Personal computer 1975 105.20 54.00 51.20 0.0797
Laser disc player 1978 264.44 92.06 172.38 0.0649
Mobile phone 1979 55.10 43.80 11.20 0.3858
Camcorder 1983 24.38 40.27 15.89
Compact disc player 1983 97.50 58.70 38.80 0.2398
Internet 1991 570.60 54.60 516.00 0.0018
Minidisc player 1992 50.86 18.22 32.65 0.2105
Digital video disc player 1997 12.75 13.43 0.68
Digital high-definition TV 1998 42.63 11.52 31.11 0.1258
GPS receiver 2000 64.80 3.90 60.90 0.0707
Smartphone 2000 115.70 15.90 99.70 0.0157
Average 65.01 20.42 44.58 0.0054
Notes. For each innovation where a bubble has been detected, the table shows the buy-and-hold returns realized by firms associated with
the innovation, the buy-and-hold returns of the stock market during the same period, and the buy-and-hold abnormal returns (BHARs),
consisting of the difference between the previous two variables. These abnormal returns provide a measure of the economic value added by
the innovation during the bubble period. The last column provides the p-value of a test that each BHAR is significantly higher than zero (for
positive BHARs). The last row in the table shows the sample mean of all variables, as well as the p-value of a test that the sample mean BHAR
is equal to zero.
point in the Google Ngram data). On average, bubbles
lead the takeoff in visibility, which is what we would
expect if market investors are forward looking.
Of the 37 innovation bubbles in our sample, 28 start
no later than the takeoff in visibility. For a formal statis-
tical test, we calculate the difference (in years) between
the year when the bubble starts and the year when
the visibility takes off. We then average this difference
across the 37 innovation bubbles in our sample. On
average, bubbles lead visibility by approximately 4.5
years, and this difference is statistically different from
zero (t2.90).
5.3. Determinants of Bubbles
We now turn to examining the determinants of bub-
bles. Table 5presents the results. The dependent vari-
able in Table 5is the magnitude of the bubble for
innovations where bubbles were detected. For the
remaining innovations, the dependent variable is set
to zero. We use a Tobit model to account for the fact
that the magnitude of the bubble is left-censored to
zero. The first four independent variables in Table 5
are the characteristics of innovations previously men-
tioned: radical, disruptive, indirect network effects, and
direct network effects. These variables represent the
Sorescu et al.: Two Centuries of Innovations and Stock Market Bubbles
Marketing Science, Articles in Advance, pp. 1–23, ©2018 INFORMS 15
Table 5. Factors Related to the Magnitude of Innovation Bubbles
Dependent variable: Bubble_Mag (Magnitude of the Innovation Bubble)
Independent variables Model 1 Model 2 Model 3 Model 4 Model 5 Model 6
Radical 82.92 83.60 86.82 87.43
2.562∗∗ 2.322∗∗ 2.826∗∗∗ 2.862∗∗∗
Disruptive 86.92 119.877.04 76.42
2.097∗∗ 2.476∗∗ 1.9541.951
Indirect Network Effects 55.33 87.93 58.96 58.20
1.8002.243∗∗ 2.014∗∗ 2.002∗∗
Direct Network Effects 21.06 44.14 37.91 38.53
0.859 1.388 1.533 1.57
Ngram_Start 3.70 4.14
2.057∗∗ 2.161∗∗
Ngram_Peak 3.93 4.37
2.182∗∗ 2.275∗∗
Intercept 112.742.79 140.2142.4 20.98 18.74
0.945 0.316 1.234 1.261 1.314 1.152
AIC 457.02 366.59 454.98 454.49 459.71 459.25
No. obs. 51 43 51 51 51 51
Notes. The table shows the results of six models that relate the magnitude of the innovation bubble (Bubble_Mag) to two different categories
of independent variables. The first category evaluates the extent to which the innovation (i) is Radical, (ii) is Disruptive, (iii) has Direct Network
Effects, and (iv) has Indirect Network Effects. This evaluation is performed through an MTurk survey of 100 qualified respondents. The second
category is the extent to which the innovation has been featured in books—as captured by Google Ngram frequencies—(i) up to the time when
a bubble is detected (Ngram_Start) and (ii) up to the year of the peak of the bubble (Ngram_Peak). When no bubbles are detected, we compute
estimated start and peak dates for the bubble based on the year of product commercialization. In such cases, Ngram_Start and Ngram_Peak
are computed using Ngram frequencies based on the estimated (rather than actual) start and peak years of the bubble. Models 1 and 3–6 are
estimated using the entire sample of 51 innovations. Model 2 is confined to the sample of radical innovations included in Chandy and Tellis
(2000). The dependent variable is the Bubble_Mag when a bubble is detected. When a bubble is not detected, Bubble_Mag is set to zero. All
models are estimated using a Tobit specification, which accounts for the magnitude of the bubble being censored at zero. The t-statistics are
shown in italics under each coefficient.
,∗∗, and ∗∗∗ denote coefficients that are statistically significant at the 10%, 5%, and 1% levels, respectively.
average ratings provided by MTurk respondents on
each characteristic. The remaining independent vari-
ables are two alternative proxies for the contempora-
neous visibility of the innovation. They are obtained
using the Ngram frequencies at the start of the bubble,
and, alternatively, using the Ngram frequencies at the
peak of the bubble.
In Model 1 we show that the magnitude of the bubble
is positively related to the radical innovation measure
(β82.92,p<0.05) and negatively related to the dis-
ruptive innovation measure (β86.92,p<0.05). The
negative effect of the disruptive characteristic suggests
that it is difficult for investors to forecast whether a new
technology that is initially inferior to existing alterna-
tives might one day improve and surpass these alterna-
tives in performance. In regard to network effects, we
find a significant positive association between the mag-
nitude of the bubble and the presence of indirect net-
work effects (β55.33,p<0.10). However, the relation
with direct network effects is insignificant. This could
be due in part to the high correlation between indirect
and direct network effects, but it is also possible that
indirect network effects might be more closely aligned
with the concept of exploding expectations mentioned
in Section 2. Investor enthusiasm will be heightened
if an innovation not only has the potential to gener-
ate significant growth in its own industry, but could
also gain additional functionality through complemen-
tary products that would enhance its consumer ben-
efits. Accurately determining all categories in which
complementary products may be introduced, as well
as the demand for these complementary products,
could be even more difficult than assessing the primary
demand for the focal innovation; optimism toward the
innovation’s potential could be compounded by opti-
mism about complementary products. In Model 2 we
show that these results also hold for the subset of
radical innovations listed in Chandy and Tellis (2000),
which constitute a more conservative subsample for
our study.
The Ngram visibility variables are introduced in Mo-
dels 3 through 6. As conjectured previously, we find a
significantly positive relation between the bubble mag-
nitude and the visibility of the innovation measured
at the beginning of the bubble (β3.70,p<0.05 in
Model 3 and β4.14,p<0.05 in Model 5) and at the
peak of the bubble (β3.93,p<0.05 in Model 4 and
β4.37,p<0.05 in Model 6). These results are con-
sistent with the intrinsic bubble theory of Froot and
Obstfeld (1991).14
5.4. Capital Raised and Postbubble
Visibility of Innovations
In our next set of tests, we measure the amount of
new equity capital raised during the bubble period and
Sorescu et al.: Two Centuries of Innovations and Stock Market Bubbles
16 Marketing Science, Articles in Advance, pp. 1–23, ©2018 INFORMS
Table 6. Analysis of Net Capital Raised and Postbubble Visibility
Panel A: Univariate statistics
Variable
Name Description Number of obs. Mean value t-stat. p-value
DCapRaised Market-adjusted capital raised during bubble 38 0.1857 1.65 0.1064
DCapRaised2 Market-adjusted capital raised before and during
bubble
37 0.2795 1.94 0.0602
Ngram_Max Highest value of Ngram for innovations with
bubbles minus that of innovations without
bubbles
38 2.1622 2.33 0.0256∗∗
Ngram_Inflex Time to inflection point for innovations with
bubbles minus that of innovations without
bubbles; time to inflection point is the difference,
in years, between the commercialization date
and the date when variable Ngram reaches its
inflection point through time
38 4.5929 1.11 0.2727
Panel B: Cross-sectional analysis
Dependent
variable CapRaised DCapRaised Ngram_Max Ngram_Inflex
Indep. Model Model Model Model Model Model Model Model Model Model Model Model
variables 7 8 9 10 11 12 13 14 15 16 17 18
Bubble_Mag 0.00523 0.00575 0.0224 0.0168 0.0169 0.0399 0.0212 0.0210
2.48∗∗ 2.82∗∗∗ 2.46∗∗ 1.911.801.28 0.79 0.73
CapRaised 1.5275 1.0576 4.1631 3.5719
5.50∗∗∗ 3.14∗∗∗ 2.26∗∗ 2.22∗∗
DCapRaised 1.4806 0.9467 3.9418 3.2794
5.68∗∗∗ 2.89∗∗∗ 2.19∗∗ 2.04∗∗
Intercept 0.1240 0.1237 5.5365 5.5018 4.5729 4.7040 4.6900 12.9706 13.0479 14.4609 14.0179 14.0552
0.80 0.82 13.53∗∗∗ 13.09∗∗∗ 8.82∗∗∗ 9.62∗∗∗ 9.47∗∗∗ 7.02∗∗∗ 6.97∗∗∗ 5.52∗∗∗ 5.24∗∗∗ 5.19∗∗∗
R-square 0.1458 0.1813 0.1614 0.1475 0.1847 0.2508 0.2341 0.0676 0.0590 0.0331 0.0756 0.0665
Number of obs. 38 38 38 38 38 38 38 38 38 38 38 38
Notes. The table shows an analysis of net capital raised and postbubble visibility for the 38 innovations in our sample that were commercialized
in 1919 and later, beginning with the radio and the airplane. CapRaised is the percentage of capital raised during the bubble period divided by
the market value of equity before the bubble. DCapRaised is the difference between CapRaised and the percentage capital raised by the entire
stock market during that same period. DCapRaised2 is the same as DCapRaised except that it is measured beginning with the commercialization
date through the end of the bubble. All three values are winsorized at the 99th percentile to mitigate the effect of outliers. The postbubble
visibility is measured with Ngram_Inflex (the difference, in years, between the commercialization date and the date when the Ngram time
series reaches its inflection point) and with Ngram_Max (the maximum value of the Google Ngram frequency during the lifetime of an
innovation). Bubble_Mag is the magnitude of the innovation bubble; when a bubble is not detected, Bubble_Mag is set to zero. Panel A shows
univariate descriptive statistics, and panel B presents cross-sectional regressions. Standard errors in Models 9 through 18 are corrected for
heteroskedasticity using White’s method (1980). The t-statistics are shown in italics in panel B.
,∗∗, and ∗∗∗ denote statistical significance at the 10%, 5%, and 1% levels, respectively (two-tailed test); denotes a coefficient that is marginally
insignificant (p0.1064).
examine how this new capital relates to the postbubble
visibility of innovations. The results are presented in
Table 6.
Panel A of Table 6provides univariate statistics for
the dependent variables. The first two variables in panel
A are measures of market-adjusted equity raised by
parent firms around the bubble period. The first of
these variables, DCapRaised, measures equity raised
during the bubble period only. The second variable,
DCapRaised2, measures the equity raised between the
commercialization date and the end of the bubble. Both
variables are computed as the amount of new equity
raised by parent firms divided by the market value of
equity of these parent firms at the beginning of the
bubble, minus the equivalent ratio of equity capital
raised by the broader stock market during the same
period.15 On average, firms in our sample are able to
raise new equity capital during the bubble correspond-
ing to 18.57% of these firms’ market value of equity at
the beginning of the bubble. The result is borderline
insignificant (p0.1064), perhaps because of the low
power of the small sample size (N38). Better statistical
significance is achieved when the new equity raised is
measured over a longer period—from the commercial-
ization date until the end of the bubble. In this case, the
average equity capital raised is 27.95% (p0.0602).16
Sorescu et al.: Two Centuries of Innovations and Stock Market Bubbles
Marketing Science, Articles in Advance, pp. 1–23, ©2018 INFORMS 17
Turning now to the postbubble visibility of the inno-
vation, the third variable in panel A, Ngram_Max,
shows the difference between the average of the highest
value of Ngram frequencies achieved by innovations
with bubbles and the same average achieved by inno-
vations without bubbles. As conjectured, this differ-
ence is positive and statistically significant (p0.0256),
suggesting that innovations with bubbles become sig-
nificantly more visible during the postbubble period.
The fourth variable in panel A, Ngram_Inflex, mea-
sures the length of time between commercialization
and the Ngram inflection point computed for innova-
tions with bubbles, minus that for innovations with-
out bubbles. As conjectured, innovations with bubbles
have a higher rate of increase in postbubble visibility:
the Ngram inflection point is achieved 4.59 years ear-
lier, on average, when compared to innovations with-
out bubbles. This result, however, is not statistically
different from zero, perhaps because of the low power
of the test. Another possibility is that the capital raised
during the bubble may be the factor that leads to a
faster takeoff rather than the magnitude of the bubble
itself. We examine this possibility in our cross-sectional
analysis.
The results of the cross-sectional analysis are shown
in panel B of Table 6. The first two models in panel B
examine the relation between new equity capital raised
during the bubble (the dependent variable) and the
magnitude of the bubble. In Model 7, the dependent
variable is the actual amount of equity capital raised
(as a fraction of prebubble capital) without any mar-
ket adjustment. In Model 8, the capital raised is market
adjusted. The independent variable in both models is
the magnitude of the bubble. As discussed previously,
the magnitude of the bubble could be inversely pro-
portional to the cost of equity during the bubble. The
results in Models 7 and 8 show that, on average, firms
raise more equity capital when the magnitude of the
bubble is higher (β0.00523,p<0.05 in Model 7 and
β0.00575,p<0.01 in Model 8), perhaps taking advan-
tage of temporarily lower funding costs.
In Models 9 through 13 of panel B, the dependent
variable is the maximum level of visibility achieved by
each innovation subsequent to the bubble (measured
using Ngram frequencies). In Models 14 through 18,
the dependent variable is the time distance between
commercialization and the inflection point of the
Ngram time series. This time distance is the inverse of
the rate of increase in postbubble visibility. The inde-
pendent variables in Models 9 through 18 are the mag-
nitude of the bubble and the two measures of equity
capital that had been previously used as dependent
variables (in Models 7 and 8).
The results from Models 9 through 13 show that the
maximum postbubble visibility is positively related to
the new capital raised during the bubble (β1.5275,
p<0.01 in Model 9 and β1.4806,p<0.01 in Model 10)
and to the magnitude of the bubble (β0.0224,p<0.05
in Model 11). When both independent variables are
included (Models 12 and 13), the coefficients remain
statistically significant.
The results from Models 14 through 18 show that the
time lag between commercialization and the Ngram
inflection point is negatively related to the amount of
new capital raised during the bubble, but not to the
magnitude of the bubble. The coefficient of the new
capital raised is β4.1631,p<0.05, in Model 14; β
3.9418,p<0.05, in Model 15; β3.5719,p<0.05,
in Model 17; and β3.2794,p<0.05, in Model 18.
This suggests that the effect of the bubble magnitude
on the rate of increase in visibility is channeled through
the new capital raised during the bubble, rather than
through the bubble itself.
Overall, the results in panel B of Table 6suggest that
parent firms faced with larger stock market bubbles
raise more capital, on average. In turn, this capital is
positively related to the maximum visibility achieved
by each innovation (postbubble) and to the rate of
increase in this visibility.
5.5. Macroeconomic Factors Impacting Bubbles
A common hypothesis in the popular press is that
bubbles are fueled by the availability of cheap credit
(Kindleberger and Aliber 2005, Wood 2006) and by the
collective frenzy afflicting investors during good eco-
nomic times (Perez 2002). We find no support for this
hypothesis. We use two different approaches to deter-
mine whether macroeconomic conditions explain the
occurrence of innovation bubbles. First, we estimate a
negative binomial model in which the dependent vari-
able is the number of innovation bubbles encountered
each year during the period from 1800 to 2010.17 Inde-
pendent variables include the level of interest rates
that year, the growth rate in gross domestic product
(GDP) over the previous year, the ratio of national debt
to GDP, and the ratio of the monetary base to GDP.
None of these independent variables are statistically
significant (the results are available from the authors
on request).
Second, we evaluate the possibility that innovation
bubbles in our sample might be caused by contempo-
raneous marketwide bubbles. To do so, we use West’s
(1987) model to determine whether marketwide bub-
bles occur during the same time periods as the inno-
vation bubbles detected in Table 3. The results are pre-
sented in Table 7. We detect simultaneous stock market
bubbles in 50% of the innovations studied (18 out of
36).18 This fraction is significantly different from unity
(p<0.0001). Moreover, in 83% of the cases (15 out of
18), the magnitude of the marketwide bubble is smaller
than the magnitude of the innovation bubble. This
fraction is significantly different from 1
2(p0.0365).
Sorescu et al.: Two Centuries of Innovations and Stock Market Bubbles
18 Marketing Science, Articles in Advance, pp. 1–23, ©2018 INFORMS
Table 7. Contemporaneous Market Bubbles
Bubble magnitude (%)
Innovation Comm. year Market bubble detected? Innov. stocks Market Difference
Steam engine train 1825 Yes 31.10 15.04 16.06
Telegraph 1845 Unknown 75.60 n/a n/a
Incandescent vacuum lamp 1880 No 16.25 16.25
Automobile 1886 No 119.50 119.50
Portable camera 1888 No 101.10 101.10
Disk phonograph 1894 No 102.60 102.60
Motion pictures 1898 Yes 49.80 24.93 24.87
Rayon 1910 No 24.40 24.40
Refrigerator 1916 No 369.30 369.30
Airplane 1919 Yes 183.90 19.38 164.52
Radio 1919 Yes 86.50 19.38 67.12
Electric typewriter 1920 Yes 51.80 291.30 239.50
Electric dishwasher 1930 Yes 41.24 7.10 34.14
Electric blanket 1930 Yes 41.24 7.10 34.14
Electric garbage disposer 1935 Yes 11.97 29.80 17.83
TV 1936 Yes 50.30 29.76 20.54
FM radio 1940 No 26.41 26.41
Ballpoint pen 1945 Yes 47.70 27.86 19.84
Microwave oven 1947 No 59.20 59.20
Magnetic tape player 1947 No 62.90 62.90
Videocassette recorder 1956 No 79.14 79.14
Cassette 1964 No 72.90 72.90
Electronic desktop calculator 1964 Yes 50.31 20.85 29.46
Electronic pocket calculator 1968 Yes 46.90 20.85 26.05
Single-player video game 1972 No 171.16 171.16
Personal computer 1975 Yes 86.20 24.46 61.74
Disposable shaver 1975 No 20.10 20.10
Laser disc player 1978 No 186.11 186.11
Mobile phone 1979 Yes 101.90 20.86 81.04
Compact disc player 1983 Yes 93.50 15.35 78.15
Camcorder 1983 Yes 50.69 91.35 40.65
Internet 1991 Yes 54.10 8.27 45.83
Minidisc player 1992 No 43.60 43.60
Digital video disc player 1997 No 58.50 58.50
Digital high-definition TV 1998 No 35.68 35.68
GPS receiver 2000 Yes 96.70 9.96 86.74
Smartphone 2000 No 134.90 134.90
Average 79.33 18.99 60.44
Notes. For each innovation where a bubble has been detected, the table shows the result of West’s (1987) bubble test applied to the entire stock
market at the time of the innovation bubble. When a marketwide bubble is detected, the table provides the magnitude of the market bubble
(in percentage points). We are unable to conduct a market bubble test for the telegraph because of the absence of market dividend data during
that period. The last column of the table shows the difference between the magnitude of the innovation bubble and the magnitude of the
marketwide bubble. When no marketwide bubble is detected, the magnitude of the marketwide bubble is assumed to be zero for the purpose
of computing this difference. The last row in the table shows the sample mean of the following bubble magnitudes: (i) all innovation bubbles,
(ii) all market bubbles (including the values of zero previously mentioned), and (iii) the difference between the innovation bubble and the
market bubble.
In addition, the difference in magnitude between mar-
ketwide bubbles and innovation bubbles is signifi-
cantly negative (p0.0699). Finally, innovation bub-
bles tend to lead market bubbles, on average. Of the 18
market bubble cases, the innovation bubble leads the
market bubble in 10 cases and lags the market bubble
in 4 cases. In the remaining four cases, the innovation
bubble begins at the same time as the market bub-
ble. These results suggest that innovations are the ones
spurring bubbles in the overall market, not the other
way around.19
In short, we find no evidence that innovation bub-
bles are driven by low interest rates, by expansion in
the monetary base, by high GDP growth, or by bub-
bles in the broader stock market. While it is entirely
possible that other types of bubbles could be ignited
by such macroeconomic conditions, innovation bub-
bles appear to be ignited by the innovation per se. This
conclusion is consistent with the findings of Frehen
et al. (2013), who show that the Dutch wind trade and
South Sea Bubbles of 1720 were driven by innovation.
The conclusion is also corroborated by the fact that we
Sorescu et al.: Two Centuries of Innovations and Stock Market Bubbles
Marketing Science, Articles in Advance, pp. 1–23, ©2018 INFORMS 19
are detecting two innovation bubbles during the Great
Depression of the 1930s, a period known for tight mon-
etary policy and low economic growth.
6. Discussion and Conclusion
We explore a new facet of the interaction between the
stock market and the consumer product market—the
association between innovation and stock market bub-
bles. We examine 51 major innovations that were com-
mercialized during the 19th and 20th centuries. In 37 of
these 51 cases, we detect the presence of bubbles in the
stock price of parent firms. We show that these bub-
bles are more likely to occur for innovations that are
radical and that have the potential to generate indirect
network effects. Moreover, the magnitude of these bub-
bles is proportional to the visibility of the underlying
innovation at the time of the bubble, as measured by
the contemporaneous frequency with which the inno-
vation appears in books, obtained from Google Ngram.
Parent firms that experience bubbles are shown to raise
additional equity capital during the bubble period. In
turn, this new equity capital is positively related to the
magnitude and to the rate of increase in the visibility of
the innovation after the bubble. Below we summarize
our paper’s contribution to theory, practice, and pol-
icy making; discuss the limitations of our research; and
provide a set of implications and directions for future
research.
6.1. Contributions to Theory, Practice, and
Policy Making
To our knowledge, our study presents the most com-
prehensive evidence of a systematic association be-
tween innovations and bubbles. The popular press has
long conjectured that technological revolutions and
bubbles seem to go hand in hand, often attributing the
cause of these bubbles to investor irrationality. How-
ever, there is insufficient academic evidence to sup-
port this conjecture. While a few studies in financial
economics evaluate the presence of bubbles around
innovations, these studies tend to focus on one or two
industries and generally treat innovation as an undif-
ferentiated output of the aggregate production func-
tion. Moreover, most studies do not provide formal
statistical tests for bubbles, and there is still a signifi-
cant degree of disagreement among authors about the
causes and economic consequences of bubbles.
On the other hand, marketing studies do a better
job at recognizing heterogeneity among various types
of innovation and their impact on shareholder value
(e.g., Pauwels et al. 2004, Sood and Tellis 2009, Sorescu
and Spanjol 2008). However, the marketing literature
does not explore the association between innovation,
bubbles, and capital raised, or, more generally, the
manner in which the new product market interacts
with the stock market. Our paper seeks to bridge the
gap between the innovation literatures in marketing
and in financial economics by illustrating one way in
which the marketing and finance functions come to-
gether to create and support economically significant
innovation.
The financial consequences we document in this
paper are noteworthy additions to the marketing litera-
ture and provide managers with actionable guidelines
when major innovations appear on their radar screens.
While these innovations are expected to be associated
with stock price bubbles, they are also expected to
add significant value to the parent firm after bubbles
deflate. Thus, investing in the commercialization of
innovations appears to be worthwhile, especially since
bubbles seem to create a favorable environment for
firms to raise cheap equity capital without sending
negative signals to the market.
Our results challenge some of the views expressed
in the popular press about the association of innova-
tions with bubbles. We show that innovation bubbles
are not confined to “technological revolutions.” Table 3
shows a number of bubbles surrounding radical inno-
vations (such as the microwave oven) in time periods
that are not normally associated with technological
revolutions. Moreover, we find no evidence that bub-
bles are fueled by strong GDP growth or by loose mon-
etary policy. For example, we document two bubbles
during the 1930s, a period of declining GDP and tight
monetary policy. In short, innovation bubbles appear
to be intrinsic to the innovation per se, and seem to be
unrelated to aggregate macroeconomic conditions.
While more research is needed to fully understand
the relation between bubbles, issuance of new equity
capital, and the diffusion of innovations, the evidence
presented in this paper suggests that bubbles could
actually foster economic growth by facilitating the
development of new infrastructure. This finding is con-
sistent with the predictions of Olivier (2000). In spite
of bubbles, innovating firms have higher stock market
gains at the end of a bubble compared to the gains real-
ized by the stock market as a whole during the same
period. The negative consequences of bubbles are most
likely confined to new investors who buy overvalued
stocks close to the height of the bubble. These find-
ings suggest that economic policies designed to sup-
press bubbles might be misguided. These implications
are limited to bubbles that are spurred by important
technological breakthroughs and cannot be extended
to other types of bubbles, such as bubbles on intrinsi-
cally useless assets (e.g., tulip bulbs), bubbles caused
by fraud (e.g., Enron), or bubbles possibly fueled by
loose monetary policy.
6.2. Limitations
6.2.1. A Caveat on the Econometric Measurement of
Bubbles. All econometric models that attempt to detect
Sorescu et al.: Two Centuries of Innovations and Stock Market Bubbles
20 Marketing Science, Articles in Advance, pp. 1–23, ©2018 INFORMS
the presence of bubbles (including West 1987) are test-
ing a joint null hypothesis that (i) no bubble is present
and (ii) the correct dividend expectation model is used.
Rejecting the null hypothesis implies either that a
bubble is present or that we are using the wrong div-
idend expectations model. While this problem does
not have an econometric resolution, we offer a con-
ceptual argument in favor of the bubble interpreta-
tion: the alternative interpretation—the wrong divi-
dend model—carries the embedded assumption that
the 37 innovations for which we detect bubbles could
have generated even higher cash flows, but did not
do so because of bad luck; that is, from an ex ante
perspective, these innovations must have offered a
nonzero probability of a subsequent increase in earn-
ings at a rate significantly higher than what was actu-
ally observed in reality. Yet, by bad luck, this possibility
never materialized. By contrast, the bubble explanation
does not require this type of assumption and appears,
at least to us, to be more credible. We acknowledge,
however, that despite the weight of the evidence pre-
sented, one can never be certain that the price patterns
shown in this paper are true bubbles as opposed to
bubble-free valuations in anticipation of extraordinary
events that never materialized.
6.2.2. Discussion of Causality. We do not claim that
we can unequivocally establish a causal relation be-
tween innovation and bubbles. We mainly aim to pro-
vide formal statistical tests that document stock market
bubbles across the largest set of innovations identi-
fied to date. It is difficult to establish a causal relation
between innovation and bubbles because the nature
of our data—rare historical events—is not conducive
to a formal statistical test of causality. Thus, we can-
not exclude the alternative interpretation that bubbles
and innovations might be jointly caused by a third,
unknown factor, such as periodic changes in social
norms that could foster both creativity in the product
market and optimism in the stock market.
We note, however, that our paper’s managerial impli-
cations do not change under this alternative hypothe-
sis. To the extent that firms can take advantage of bub-
bles to raise equity capital on favorable terms, and to
the extent that this new capital can facilitate the dif-
fusion of the innovation, it does not matter whether
bubbles are caused by innovation or whether both are
jointly caused by a third factor, so long as both phe-
nomena occur simultaneously. In both cases, our rec-
ommendations to policy makers and managers remain
the same.
6.2.3. Data Limitations. Despite the significant effort
invested into assembling our data, the sample is not
without limitations. First, our sampling scheme con-
sists of a set of commercially successful innovations.
Over the time span of our sample, there could have
been other new products introduced that had the
potential to be radical and generate network effects but
eventually failed in the marketplace. However, we note
that this type of selection bias is a limitation that is
common in the radical innovation and diffusion liter-
atures, whereby most authors take a post hoc view in
assembling the sample. Second, we were not able to
control for the sales of each innovation as a percentage
of the parent firm sales. We used archival data to quali-
tatively determine the extent to which each innovation
was an important part of the firm’s product portfolio at
the time of commercialization. We endeavored to limit
our test portfolios, as much as possible, to pure-play
firms. However, we recognize that, despite our best
efforts, there will be some variation in our sample in
terms of the importance that the focal innovation plays
in the product portfolio of its parent firm. This vari-
ation leads to an error in measurement problem that
attenuates the coefficients in Table 6toward zero. Thus,
the results presented in Table 6should be viewed as
conservative.
6.3. Directions for Future Research
Our paper suggests five promising avenues for future
research. First, our results imply that publicly traded
firms might have a resource advantage when it comes
to commercializing innovation because of the ease of
access to new equity capital, particularly during bubble
periods. If so, are privately held corporations at a dis-
advantage? Does the pattern of venture capital invest-
ment in private firms follow the patterns observed in
the public equity markets?
Second, while we provide first-order evidence of the
association between bubbles, external financing, and
innovation, additional studies are needed to under-
stand specifically how this new capital is actually de-
ployed and how it affects the diffusion of the new
product.
Third, the theory on intrinsic bubbles suggests that
marketing metrics may play a critical role in deter-
mining the market value of firms in new industries.
An important implication of this theory is that in the
absence of accurate forecasts about long-term cash
flows, investors are likely to anchor their valuation on
observable short-term metrics of performance, and val-
uations are likely to exceed intrinsic values for a period
of time. These short-term metrics most likely come
from the product market, as opposed to the financial
market, because information from the product market
is typically the first to become available to investors.
This is evidenced by examples in the dot.com era,
where valuations were multiples of website traffic, or
from more recent examples in the social media domain,
where valuations are multiples of the number of users.
Therefore, marketing metrics could play a critical role
Sorescu et al.: Two Centuries of Innovations and Stock Market Bubbles
Marketing Science, Articles in Advance, pp. 1–23, ©2018 INFORMS 21
in determining the market value of companies in new
industries. While our findings are certainly consistent
with this interpretation, additional research is needed
to provide a more direct link between marketing met-
rics and the valuation of companies that pursue a sig-
nificant innovation strategy.
Fourth, the extent to which firms can appropriate
gains from innovations associated with bubbles re-
quires additional study. While we show in our paper
that these innovations have a positive economic value
added, we have not controlled for variations that may
arise from firms’ differential ability to protect their
innovations through patents. Should a firm be able,
through a strong patenting effort, to prevent competi-
tors from commercializing the same technology, the
firm might obtain monopolistic profits, but might also
miss out on the network effects that arise when com-
petitors leverage the same technology. These trade-offs
are a fruitful area of future research.
Fifth, the social benefits of cheap equity financing
provided by innovation bubbles must be evaluated
against their social costs. The magnitude of these social
costs depends, in part, on how these bubbles affect
the aggregate disparity of wealth in the society. Shiller
(2015) and Perez (2002) both hypothesize that bubbles
have contributed to a significant increase in the dispar-
ity of wealth observed during the past 100 years, and
Piketty (2013) presents evidence consistent with this
prediction. If true, there is perhaps a significant hid-
den social cost of innovation bubbles, as no stable soci-
ety appears to have survived over the long term when
the levels of wealth became extremely concentrated.20
Future research is needed to understand the relation
between innovation bubbles and wealth inequality.
We hope that our research provides an impetus for
others to further study the effects of innovation at the
macro level, including their relation to the stock mar-
ket, to economic growth, and to economic policy. At
the micro level, we hope that our research can help
managers better navigate through the perils of major
shifts in technological paradigms and even learn how
to profit from periods of temporary overvaluation that
might provide windows of opportunity to obtain exter-
nal financing on favorable terms.
Acknowledgments
The authors thank Rajesh Chandy and participants at
the Marketing Strategy Meets Wall Street conference at
Singapore Management University, the 2013 Strategic Man-
agement Society Lake Geneva Special Conference on Big
Bang Innovation, and the 2012 INFORMS Marketing Science
Conference, as well as participants in research seminars at
Bocconi University, Tilburg University, the University of Ari-
zona, and the University of Washington, for their valuable
feedback. The authors also thank Gareth Campbell for stock
prices and dividend data for all firms listed in the Railway
Times.
Endnotes
1As we describe in detail in Section 4, we do not measure diffusion
directly; rather, we use a contemporaneous measure of the public
visibility of innovations that is based on Google Books Ngram data.
For a subsample of innovations, we show in Section 4that the Ngram
visibility data correlate highly with the actual diffusion data based
on product adoption.
2The nine innovations mentioned by Shiller (2015) are the phono-
graph, electricity, trains, the automobile, radio, electrification, mo-
tion pictures, TV, and the Internet. However, Shiller (2015) does not
conduct formal statistical tests to document the existence of the bub-
bles associated with these nine innovations.
3We do not focus here on the exact mechanism that might trig-
ger a rational bubble; we only study boundary conditions that are
conducive to the emergence of rational bubbles. The literature in
financial economics discusses several trigger mechanisms for ratio-
nal bubbles (e.g., Allen et al. 1993, Zeira 1999). Common among these
papers is the idea that when the cost of entry into the stock mar-
ket is low and there is asymmetric information among investors, an
excessive number of investors will enter the stock market, driving
the price of the stock temporarily above its intrinsic value. We note a
very interesting parallel between this mechanism and the one that in
the marketing literature has been shown to trigger excessive entry of
firms into a new industry. Specifically, Shen and Villas-Boas (2010)
and Shen (2014) show that under similar conditions of low entry
costs into the product market and information asymmetry among
firms, an excessive number of firms will enter a new industry, driv-
ing economic rents to become negative. Interestingly, both actions
are perfectly rational.
4For example, The Economist (2000) observed that “technological rev-
olutions and financial bubbles seem to go hand in hand,” while
Gross (2009) is featured on CNN Money under the title “The bub-
bles that built America” (see http://money.cnn.com/galleries/2007/
news/0705/gallery.bubbles/index.html.).
5Even when several firms start commercializing a major innovation
around the same time, their stocks are likely to be perceived as hav-
ing a similar level of risk. Therefore, risk cannot be diversified away
by investing in a portfolio of firms that manufacture the same radical
innovation.
6We excluded 21 innovations from Chandy and Tellis (2000) for the
following reasons: 12 innovations were excluded because none of
the parent firms were publicly traded at the time of commercializa-
tion. Four additional innovations were excluded because dividend
or price data were not available. Three innovations were considered
duplicates of innovations already in our sample. One innovation was
excluded because none of the parent firms were pure plays. The final
innovation (telephone) was excluded because of the lack of data for
Bell at the time of commercialization and because Bell’s successor
(AT&T) became a regulated utility in 1910. Firms in regulated indus-
tries typically face a cap on revenues that rules out the bubble solu-
tion in the Blanchard and Fischer (1989) model. A full explanation
for each of these exclusions is presented in Web Appendix B.
7All of our bubble tests are performed on single time series, even
those that are based on a large portfolio of firms. Specifically, if an
innovation portfolio contains two or more firms, we begin by aggre-
gating their prices and dividends into single industry-specific time
series and perform the tests on these two aggregate time series as
if the entire portfolio were a single firm. Moreover, the theory that
underlies the formation of bubbles posits that firms commercializ-
ing the same innovation have the same probability of experiencing
a bubble, as the bubble arises exogenously to firm characteristics,
being driven instead by the characteristics of the innovation that
meets the four conditions described in Section 2.1. Consequently, our
results should not be affected by the number of firms included in
each portfolio.
Sorescu et al.: Two Centuries of Innovations and Stock Market Bubbles
22 Marketing Science, Articles in Advance, pp. 1–23, ©2018 INFORMS
8A multiproduct firm can be viewed as a conglomerate of smaller,
pure-play firms. Assuming that only one of the firm’s products is
innovative—and therefore subject to a potential bubble—the stock
price of the multiproduct firm is likely to experience a smaller bubble
than the stock of a pure play firm in the same industry. Therefore, the
inclusion of multiproduct firms in our test portfolios creates an error
in the measurement of the magnitude of innovation bubbles. In some
of the tests we perform in Section 4, this measurement error is incon-
sequential because the bubble magnitude is used as a dependent
variable. In other tests, where the bubble magnitude is an indepen-
dent variable, the consequence of the measurement error is to reduce
the coefficients toward zero, resulting in a more conservative test.
9Moreover, the ratings vary significantly across innovations to allow
for a meaningful cross-sectional analysis. For the radical dimension,
average ratings vary from 3.00 to 4.80; for the direct network dimen-
sion, they vary from 2.04 to 4.94; for the indirect network dimension,
they vary from 2.09 to 4.65; and for the disruptive dimension, they
vary from 2.63 to 4.35. The list of average ratings for all innovations
for each of the four characteristics is presented in Web Appendix C.
The correlations between these ratings range from 0.55 (correlation
between disruptive and direct network effects) to 0.81 (correlation
between direct and indirect network effects). The full correlation
matrix is also presented in Web Appendix C.
10 Ngram frequencies vary significantly in terms of order of mag-
nitude from 105to 109, and, in some cases the Ngram count is
identically zero. Therefore, all Ngram frequencies are adjusted using
the natural log of (1+(RawNgram ×10+9)), where RawNgram is the
actual, raw Ngram frequency downloadedfrom Google. We do this to
improveinterpretability and mitigate the effects of heteroskedasticity.
11 This is the discrete-time equivalent of setting the second derivative
equal to zero.
12 Parameter ρtis related to the discount rate for the stock, kt, as fol-
lows: ρt1/(1+kt). Parameter ϕtis related to the annual growth rate
of dividends,gt, as follows: ϕt(1+gt). This estimation is performed
each month, t, over periods of time that start 24 months after the
beginning of the time series. Progressively, each period of time is
nested within the time period used during the subsequent month.
For example, when we attempt to detect a bubble in month t30, we
estimate parameters using data from months t1to t30. The fol-
lowing month, when attempting to detect a bubble in month t31,
we reestimate parameters with data from months t1to t31,
so the first time interval is nested within the second. Thus, the test
allows for estimates of growth and discount rates to change over
time; each month, the test produces an intertemporal average esti-
mate of discount rates and growth rates, which takes into account all
information available prior to that month.
13 For example, for a bubble that starts in March 1887 and ends in
June 1888, we would compute the buy-and-hold return of the test
portfolio from March 1887 through June 1888 and subtract from it
the buy-and-hold return of the market portfolio during that same
period. If the buy-and-hold return of the test portfolio is 20% and the
buy-and-hold return of the market is 5%, we would conclude that
the focal innovation adds 15% in terms of economic value, even after
accounting for the collapse of the bubble.
14 The results remain substantively the same if Rayon is excluded
from the sample. These results are presented in Web Appendix E.
15 The amount of new equity raised is computed as the product of
the number of shares issued times the price per share, adjusted for
stock splits and stock dividends.
16 However, not all industries appear to take advantage of overvalued
prices to issue stock. Most of the new equity capital was issued by
the airplane industry, Internet firms, and smartphone firms. To a
lesser extent, firms in the radio, microwave oven, video game, laser
disc player, camcorder, and GPS sectors also did issue more stock
compared to the market.
17 We also restrict this sample to the period from 1925 to 2010 and
the results are unchanged.
18 We are not able to perform a market bubble test associated with the
telegraph bubble because of the lack of marketwide dividend data
during the telegraph bubble period (1870–1873). Therefore, market
bubble tests are conducted for only 36 of the 37 innovations for which
we detect bubbles in Table 3.
19 We also repeat the analysis presented in Table 5after subtract-
ing the magnitude of the market bubble from the magnitude of the
innovation bubble. The results remain substantively similar.
20 For example, Piketty (2013) offers wealth inequality as one of the
leading causes of the French Revolution.
References
Agarwal R, Bayus BL (2002) The market evolution and sales takeoff
of product innovations. Management Sci. 48(8):1024–1041.
Akbas F, Boehmer E, Erturk B, Sorescu S (2017) Short interest,
returns, and unfavorable fundamental information. Financial
Management 46(2):455–486.
Allen F, Morris S, Postlewaite A (1993) Finite bubbles with short
sale constraints and asymmetric information. J. Econom. Theory
61(2):206–229.
Blanchard OJ, Fischer S (1989) Lectures in Macroeconomics (MIT Press,
Cambridge, MA).
Blanchard OJ, Watson M (1982) Bubbles, Rational Expectations and
Financial Markets, in Crises in the Economic and Financial Structure
(Lexington Books, Lexington, MA).
Campbell G (2012) Myopic rationality in a mania. Explorations
Econom. Hist. 49(1):75–91.
Chandy RK, Tellis GJ (1998) Organizing for radical product innova-
tion: The overlooked role of willingness to cannibalize. J. Mar-
keting Res. 35(4):474–487.
Chandy RK, Tellis GJ (2000) The incumbent’s curse? Incumbency,
size, and radical product innovation. J. Marketing 64(3):1–17.
Cowles A III (1939) Common stock indexes. Cowles Commission
monograph no. 3.
Danielsen BR, Sorescu SM (2001) Why do option introductions
depress stock prices? A study of diminishing short sale con-
straints. J. Financial Quant. Anal. 36(4):451–484.
DeMarzo P, Kaniel R, Kremer I (2007) Technological innovation and
real investment booms and busts. J. Financial Econom. 85(3):
735–754.
Economist, The (2000) A survey of the new economy: All technolog-
ical innovations carry risks as well as rewards. (September 21),
http://www.economist.com/node/375561.
Fama E (1998) Market efficiency, long-term returns, and behavioral
finance. J. Financial Econom. 49(3):283–306.
Ferguson N (2008) The Ascent of Money:A Financial History of the World
(Penguin Press, New York).
Frehen RGP, Goetzmann WN, Rouwenhorst KG (2013) New evidence
on the first financial bubble. J. Financial Econom. 108(3):585–607.
Froot KA, Obstfeld M (1991) Intrinsic bubbles: The case of stock
prices. Amer. Econom. Rev. 81(4):1189–1214.
Goetzmann WN, Ibbotson RG, Peng L (2001) A new historical
database for the NYSE 1815 to 1925: Performance and pre-
dictability. J. Financial Markets 4(1):1–32.
Golder PN (2000) Historical method in marketing research with new
evidence on long-term market share stability. J. Marketing Res.
37(2):156–172.
Golder PN, Tellis GJ (1997) Will it ever fly? Modeling the take-
off of really new consumer durables. Marketing Sci. 16(3):
256–270.
Golder PN, Tellis GJ (2004) Growing, growing, gone: Cascades, dif-
fusion, and turning points in the product life cycle. Marketing
Sci. 23(2):207–218.
Gross D (2009) Pops! Why Bubbles Are Great for the Economy, Kindle ed
(HarperCollins).
Sorescu et al.: Two Centuries of Innovations and Stock Market Bubbles
Marketing Science, Articles in Advance, pp. 1–23, ©2018 INFORMS 23
Guha K (2008) Greenspan urges policymakers to focus on banks’
capitalization. Financial Times (May 27). https://www.ft.com/
content/c78994f0-2b65-11dd-a7fc-000077b07658.
Gupta S, Jain DC, Sawhney MS (1999) Modeling the evolution of
markets with indirect network externalities: An application to
digital television. Marketing Sci. 18(3):396–416.
Hall BH (2005) Innovation and diffusion. Fagerberg J, Mowery D,
Nelson RR, eds. The Oxford Handbook of Innovation (Oxford Uni-
versity Press, Oxford, UK), 459–485.
Hsu PH, Xuan T, Xu Y (2014) Financial development and innovation:
Cross-country evidence. J. Financial Econom. 112(1):116–135.
Katz M, Shapiro C (1986) Technology adoption in the presence of
network externalities. J. Political Econom. 94(4):822–841.
Kindleberger CP, Aliber R (2005) Manias, Panics, and Crashes: A His-
tory of Financial Crises, 5th ed. (John Wiley & Sons, Hoboken, NJ).
Lamoreaux NR, Levenstein M, Sokoloff KL (2007) Financing innova-
tion during the second industrial revolution: Cleveland, Ohio,
1870–1920. Lamoreaux NR, Sokoloff KL, eds. Financing Innova-
tion in the United States: 1870 to the Present (MIT Press, Cam-
bridge, MA), 39–84.
Lansing KJ (2008) Speculative bubbles and overreaction to technolog-
ical innovation.Economic Letter 2018-18, Federal Reserve Bank
of San Francisco, San Francisco.
Loughran T, Ritter J (2004) Why has IPO underpricing changed over
time? Financial Management 33(3):5–37.
Mahajan V, Muller E (1998) When is it worthwhile targeting the
majority instead of the innovators in a new product launch? J.
Marketing Res. 35(4):488–495.
Markovitch DG, Golder PN (2008) Using stock prices to predict mar-
ket events: Evidence on sales takeoff and long-term firm sur-
vival. Marketing Sci. 27(4):717–729.
Myers SC, Majluf NJ (1984) Corporate financing and investment deci-
sions when firms have information that investors do not have.
J. Financial Econom. 13(2):187–221.
Neal L, Davis LE (2007) Why did finance capitalism and the sec-
ond industrial revolution arise in the 1890s? Lamoreaux NR,
Sokoloff KL, eds. Financing Innovation in the United States: 1870 to
the Present (MIT Press, Cambridge, MA), 129–161.
Newey WK, West KD (1987) Hypothesis testing with efficient
method of moments estimation. Internat. Econom. Rev. 28(3):
777–787.
Nicholas T (2008) Does innovation cause stock market runups? Evi-
dence from the great crash. Amer. Econom. Rev. 98(4):1370–1396.
Ofek E, Richardson M (2003) DotCom mania: The rise and fall of
Internet stock prices. J. Finance 58(3):1113–1138.
Olivier J (2000) Growth-enhancing bubbles. Internat. Econom. Rev.
41(1):133–151.
O’Sullivan MA (2007) Funding new industries: A historical perspec-
tive on the financing role of the U.S. stock market in the 20th
century. Lamoreaux NR, Sokoloff KL, eds. Financing Innovation
in the United States: 1870 to the Present (MIT Press, Cambridge,
MA), 163–216.
Pástor L, Veronesi P (2009) Technological revolutions and stock
prices. Amer. Econom. Rev. 99(4):1451–1483.
Pauwels K, Silva-Risso J, Srinivasan D, Hanssens DM (2004) New
products, sales promotions, and firm value: The case of the auto-
mobile industry. J. Marketing 68(4):142–156.
Peres R, Muller E, Mahajan V (2010) Innovation diffusion and new
product growth models: A critical review and research direc-
tions. Internat. J. Res. Marketing 27(2):91–106.
Perez C (2002) Technological Revolutions and Financial Capital: The
Dynamics of Bubbles and Golden Ages (Edward Elgar Publishing,
Northampton, MA).
Piketty T (2013) Le Capital au XXIeSiècle (Editions du Seuil, Paris).
Rapp D (2014) Bubbles, Booms, and Busts: The Rise and Fall of Financial
Assets (Springer, New York).
Rappoport P, White EN (1993) Was there a bubble in the 1929 stock
market? J. Econom. Hist. 53(3):549–574.
Scherbina A, Schlusche B (2014) Asset price bubbles: A survey. Quant.
Finance 14(4):589–604.
Shen Q (2014) A dynamic model of entry and exit in a growing
industry. Marketing Sci. 33(4):712–724.
Shen Q, Villas-Boas JM (2010) Strategic entry before demand takes
off. Management Sci. 56(8):1259–1271.
Shiller RJ (2015) Irrational Exuberance, 3rd ed. (Princeton University
Press, Princeton, NJ).
Sood A, Tellis GJ (2009) Do innovations really pay off? Total stock
market returns to innovation. Marketing Sci. 28(3):442–456.
Sood A, Tellis GJ (2011) Demystifying disruption: A new model for
understanding and predicting disruptive technologies. Market-
ing Sci. 30(2):339–354.
Sorescu A, Spanjol J (2008) Innovation’s effect on firm value and
risk: Insights from consumer packaged goods. J. Marketing 72(2):
114–132.
Stremersch S, Tellis GJ, Franses PH, Binken JL (2007) Indirect network
effects in new product growth. J. Marketing 71(3):52–74.
Tellis GJ, Prabhu JC, Chandy RK (2009) Radical innovation across
nations: The preeminence of corporate culture. J. Marketing
73(1):3–23.
Warren NL, Sorescu A (2017) Interpreting the stock returns to new
product announcements: How the past shapes investors’ expec-
tations of the future. J. Marketing Res. 54(4):799–815.
West KD (1987) A specification test for speculative bubbles. Quart. J.
Econom. 102(3):553–580.
White EN (1990) The stock market boom and crash of 1929 revisited.
J. Econom. Perspect. 4(2):67–83.
White H (1980) A heteroskedasticity-consistent covariance matrix
estimator and a direct test for heteroskedasticity. Econometrica
48(4):817–838.
Wood C (2006) The Bubble Economy: Japan’s Extraordinary Speculative
Boom of the ’80s and Dramatic Bust of the ’90s(Solstice Publishing,
Jakarta, Indonesia).
Zeira J (1999) Informational overshooting, booms, and crashes.
J. Monetary Econom. 43(1):237–257.
... Other used sources are newspapers (Solomon, 2012;Xiong & Bharadwaj, 2013) and books. Sorescu et al. (2018) use the publicly available Google Books Ngram Viewer tool to identify the historical usage of certain words in published books. ...
... These studies extract different types of text information from the given data, which can broadly be divided into simple volume of documents (e.g., reviews as in Tirunillai & Tellis, 2012) or words (Sorescu et al., 2018) and more complex classification outcomes. We distinguish among sentiment classification (distinguishing between positive and negative sentiment), content classification (assigning pre-defined category labels), and topic modeling (exploratively identifying general topics; see Berger et al., 2020;Hartmann et al., 2019). ...
Article
The marketing–finance interface is an important research field in marketing, helping demonstrate the accountability of marketing within companies and building a necessary interdisciplinary bridge to finance and accounting research. Since the first comprehensive review article by Srinivasan and Hanssens (2009), the marketing–finance field has broadened considerably, as has research in finance and accounting. This updated systematic review of extant and new research integrates research in marketing, finance, and accounting into an overarching marketing–finance research framework. We discuss new methodological developments and offer solutions to recent technical debates on the event-study method and Tobin's q. Motivated in part by a survey of marketing–finance researchers, the article identifies and synthesizes four key emerging research areas: digital marketing and firm value, tradeoffs between “doing good” and “doing well,” the mechanisms of firm-value effects, and feedback effects. The article closes with a future research agenda for this dynamic research field and offers key conclusions.
... Since the very beginning of the financial market system, there have been numerous examples of market downturns around the world. Famous instances include the Dutch Tulipomania in the 17th century in Holland, the Mississippi Bubble in France in the 18th century and the savings and loan fiasco in the United States in the 1980s (Kindleberger 1996, Rajan 2010, Sorescu et al. 2018. ...
Article
The present article introduces the concept of consumer spinning. This occurs when consumers have lost track of their initial motivation to consume and engage unwittingly in hamster-like spinning in a “wheel of misfortune”, being encouraged to spend through sweetheart deals and thus incur debt without even realizing it. This article takes for example the Global Financial Crisis of 2007-2009 (GFC) during which the use of such sweeteners by sellers of predatory mortgages and advocates for lax regulations led U.S. consumers to spin and, in the process, ignore the debt trap they were falling into. We discuss the results of a functional magnetic resonance imaging ( fMRI) experiment we conducted to test the ratio of greed over risk aversion and the hypothesized phenomenon of spinning. We believe that this is the first academic attempt to examine spinning resorting to an experimental tool used regularly in neuroscience and psychology research. Our framework indicates that government regulators should take the role of marketing sweeteners into account when devising policies. Marketing managers, who strive to educate consumers and help them reduce their debt load, thus promoting better purchasing power, stand to benefit from an understanding of these possible spinning behaviors and adapt their campaigns accordingly. We believe that there should be legal consequences for encouraging consumer spinning, as this at times may be an intentional and deceitful invitation for consumers to harm themselves financially.
... Greater demand and consumption calls for more innovations to be developed in the market (Wong et al. 2005;Schumpeter 1912). Indeed, growth in the economy impacts innovation (Sorescu et al. 2018). ...
Article
Growth in emerging markets is critical as it leads global economic growth. Emerging market innovations, which we define as innovations created in an emerging market for use by consumers or customers in that market and possibly other markets, form the engine for such growth and continue to rise in importance for various stakeholders. These innovations fundamentally differ from developed market innovations, and impact consumers, firms, governments, and the society at large. Conceptual research on emerging market innovation has not focused on the differential drivers of emerging and developed market innovations. Most empirical studies have also focused on only one emerging market, not offering generalizable conclusions across emerging markets. Thus, while prior research offers a limited understanding of factors specific to emerging market innovations, not much is known about the unique and differential drivers of emerging market innovations relative to developed market innovations. Our study offers a broader framework highlighting the unique aspects of emerging market innovations, their unique and differential (relative to developed market innovations) drivers, and related practitioner takeaways. Our framework and propositions suggest that different drivers impact the demand and supply for innovations, and, in turn, the number of innovations differently in emerging and developed markets. We outline an agenda for future research.
Article
To survive in the competitive retail landscape, retailers launch service innovations designed to grant additional value to consumers. This study investigates whether and in which circumstances retail service innovations create shareholder value, using stock returns to capture investors’ point of view. An event study is used to analyze a broad, varied set of 350 service innovation announcements by publicly listed retailers. The study shows that the customer value benefit(s) aimed for by the retail service innovation (i.e., its level of convenience and engagement) has an impact on shareholder value. Moreover, this impact is contingent upon the stage of the consumer purchase process that the innovation targets, and upon the hedonic or utilitarian nature of the products offered by the retailer that initiates the innovation. The impact on retailer shareholder value is more positive for service innovations high on convenience that speed up and simplify the shopping process, when implemented at the purchase stage or by retailers offering utilitarian products. Service innovations high on engagement that focus more on non-transactional initiatives instead fare well in the post-purchase stage.
Article
Full-text available
We study the potential for rational bubbles in the innovation sector to affect long term economic growth. We show that stock market prices of research and development (R&D) firms can include a bubble component when credit constraints are present. Bubbles are self-sustained in equilibrium by a “liquidity” premium that originates when credit constraints are relaxed. Bubbles expand the borrowing and production capacities of R&D firms, stimulate innovation, and increase the growth rate. Small firms benefit more from bubbles than big firms. Bubbles are magnified by tighter credit constraints and scarce investment opportunities. Thus, financially underdeveloped countries benefit more from bubbles. Finally, we show that bubbles can create permanent reallocation effects that benefit the innovation sector over other sectors.
Article
Predator‐prey dynamics are widely used in ecology but seldom utilized in economics and marketing, despite their ability to express financial market agents’ behaviors when considered in combination with economic cycles and financial crises. This multidisciplinary paper presents a stylized framework of a market cycle that combines the notions of supply and demand and predator‐prey interactions between buyers and sellers of housing mortgages. We illustrate our framework using data from the Global Financial Crisis and a Lotka‐Volterra predator‐prey model. We find that with our framework we can capture the dynamics of the market, particularly the peak and decline in the number of sellers and sold subprime mortgages. Our framework sheds a new light on consumer behaviors, pinpointing how they can put themselves into vulnerable prey positions. This paper is one of the first of its kind to propose market phases and predator‐prey dynamics nested in economic cycles and consumer buying trends. This article is protected by copyright. All rights reserved.
Article
Full-text available
Investors routinely follow firms' communications and actions in order to form expectations about the future performance of firms. The authors propose a set of firm and industry characteristics that influence the formation of investor expectations in the context of new product announcements. Specifically, they argue that positive expectations of future innovation output should cause an ex-ante increase in stock prices and a smaller ex-post market reaction when an actual new product is announced. Using a sample of 4,865 new product announcements made by 826 publicly traded US firms, the authors show that the stock market reaction to a new product announcement measured in a five-day window around the announcement is negatively related to (1) the number of new products previously announced by the firm, (2) the average number of new products previously announced by the firm's competitors, and (3) the average sentiment of past public news issued by the firm. These same three factors are also positively related to the market value of the firm measured immediately prior to each new announcement, controlling for increases in firm value directly attributable to past new product announcements. In contrast to many papers in the marketing literature, which imply that the added value of a marketing event can be fully assessed from the stock market reaction to the announcement of the event, the authors clarify that for recurrent events or events that are part of a firm's broader strategy, this reaction reflects only an update of investors' expectations of future firm performance.
Article
Why are some firms more successful at introducing radical product innovations than others? Following Schumpeter (1942), many researchers have suggested that firm size is the key organizational predictor of radical product innovation. The authors provide an alternate view and argue that one key variable that differentiates firms with strong radical product innovation records from others is the firms’ willingness to cannibalize their own investments. The authors identify three organizational factors that drive a firm's willingness to cannibalize. Results from a survey of three high-tech industries tend to support the alternate view that willingness to cannibalize is a more powerful driver of radical product innovation than firm size is. These results suggest a need to reconsider conventional wisdom on firm size, cannibalization, and organizational synergy.
Article
Several researchers have advocated historical or longitudinal approaches to study marketing phenomena. Although some have applied this approach, more often it has been overlooked or denigrated. The author argues that historical method is capable of producing scientific knowledge that is currently useful, rather than simply a remembrance of the past. The author presents a complete description of the historical method, so researchers can use this article as a guide when applying this method. The value of the method is illustrated by examining the prevailing finding in the marketing literature that market shares are stable over time. Although this finding is considered an empirical generalization, an analysis of more than 650 brands in 100 categories raises doubts about the longevity of market share stability.
Article
Early studies find that option introductions tend to raise the price of underlying stocks. More recent research indicates that post-1980 option introductions are associated with negative abnormal returns in underlying stocks. Other studies document increased short sale activities following option listing. This paper provides evidence that the documented abnormal returns and changes in short interest around option listings are consistent with the mitigation of short sale constraints resulting from the option introduction, and that both the abnormal returns and short interest changes around listing dates can be predicted using ex ante characteristics of the underlying stock.
Article
Despite the relatively small number of Innovators in a potential market for a new product, they are often the main target of a firm's marketing efforts. Because the Innovators tend to influence the remaining potential adopters, that is, the Majority, firms tend to allocate more marketing efforts and resources toward the Innovators than toward the Majority. Using a two-stage game model with a diffusion effect, the authors analytically investigate whether this is always an optimal strategy. They identify several market and competitive conditions in which it might be optimal for firms to target the Majority rather than the Innovators. The managerial implications of these conditions for new product launch are discussed.
Article
This paper offers an informational explanation to stock markets' booms and crashes. This explanation builds on the idea of `informational overshooting': if market fundamentals change for an unknown period of time, prices experience a boom, which ends in a crash, due to informational dynamics. The paper then shows that `informational overshooting' occurs when the market expands to a new capacity, which is unknown until it is reached. The paper presents two examples of such expansions, one due to increased productivity and the other due to entry of new investors to the stock market. One implication is that financial liberalizations tend to be followed by booms and crashes.
Article
Market efficiency survives the challenge from the literature on long-term return anomalies. Consistent with the market efficiency hypothesis that the anomalies are chance results, apparent overreaction to information is about as common as underreaction, and post-event continuation of pre-event abnormal returns is about as frequent as post-event reversal. Most important, consistent with the market efficiency prediction that apparent anomalies can be due to methodology, most long-term return anomalies tend to disappear with reasonable changes in technique.