ELERAP 219
No. of Pages 11, Model 5+
ARTICLE IN PRESS
27 October 2006 Disk Used
1
Electronic Commerce Research and Applications xxx (2006) xxx–xxx
www.elsevier.com/locate/ecra
4
Eric Walden *, Glenn J. Browne
5
Texas Tech University, ISQS, Box 42101, Lubbock, TX 79409, United States
6
7
Received 7 April 2006; received in revised form 29 September 2006; accepted 29 September 2006
OO
PR
8 Abstract
ED
This work proposes that information cascade theory can help to explain the formation of the Internet bubble. We propose that the
bubble existed because a lack of good information about the potential value of electronic commerce led investors to rely on other investors’ private valuations of electronic commerce. We use the event study methodology to estimate returns to company announcements of
electronic commerce initiatives in 1999 and 2000. We find that after controlling for network externalities and time trends, investors’ valuations of the returns to electronic commerce initiatives were significantly influenced by the market return from prior periods. Moreover,
the relative weight placed on prior periods’ returns decreased as the variance of the prior periods’ returns increased. Both of the results
are consistent with the behavior predicted by information cascade theory.
Ó 2006 Published by Elsevier B.V.
CT
9
10
11
12
13
14
15
16
F
3
Rational fads in investor reactions to electronic
commerce announcements: An explanation of the Internet bubble
2
RR
E
17 Keywords: DotCom bubble; Electronic commerce; Empirical research; Event study; Information cascade theory; Investor psychology; Rational fads;
18 Stock market
19
20 1. Introduction
CO
Technology is subject to fads – situations in which many
agents adopt a technology with exaggerated zeal [1–3].
Recently, we witnessed an impressive technology fad –
the Internet bubble. From 1998 to 2000 the popular press
hailed the beginning of a new economy, and investments
in electronic commerce activities were believed to be the
best use of one’s money. Then, by 2002, we returned to
the old economy and electronic commerce investment
was no longer seen as being particularly better or worse
than any other kind of business investment. Businesses certainly needed the electronic channel, but they also needed
many other things. During this time the chosen vehicle
for investing in electronic commerce was the stock market.
For example, in the 24 months from January 1998 to January 2000, the AMEX Internet stock index increased
nearly 600%. However, by January 2002 it had almost
UN
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
*
Corresponding author. Tel.: +1 806 742 1925; fax: +1 806 742 3193.
E-mail address: ewalden@ba.ttu.edu (E. Walden).
returned to its 1998 level. Even today, in 2006, it is only
about 50% higher than its 1998 level.
Why did investors believe that the returns on electronic
commerce initiatives should be so large? Mills [4] cites four
key reasons given to explain the Internet bubble: It was an
accident, it was engineered by the incentive structures of
financial services firms, it was due to inexperienced investors entering the market, and it was an example of ‘‘the
madness of crowds’’. To this list we add another reason.
This work proposes that investors’ estimates of electronic
commerce returns were the best guess of rationally-motivated individuals making decisions with great uncertainty.
Specifically, we argue that each individual was very uncertain about how large returns on e-commerce investments
should have been, and therefore supplemented his or her
own information by relying on information contained in
the investment behavior of others. Although there were
clearly other causes of the bubble, our data suggest that
as much as 25% of the value investors placed in an electronic commerce investment was based on the value others
placed on prior electronic commerce investments.
1567-4223/$ - see front matter Ó 2006 Published by Elsevier B.V.
doi:10.1016/j.elerap.2006.09.001
Please cite this article in press as: E. Walden, G.J. Browne, Rational fads in investor reactions to electronic ..., Electron. Comm. Res.
Appl. (2006), doi:10.1016/j.elerap.2006.09.001
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
ELERAP 219
27 October 2006 Disk Used
72 2. Background literature and relevant theory
CO
RR
E
CT
ED
In the present research we develop an explanation of the
development of the DotCom bubble using a rational explanation of fad-like behavior. The most popular rational
explanation of fad-like behavior is network externality theory [5–8]. Communication networks often have the characteristic that their value increases as the number of agents on
the network increases. Thus, as more agents join a network, the value of the network increases, causing more
agents to want to join the network. This positive feedback
loop leads to rapid growth. However, this theory also suffers from three shortcomings in the present context. First,
not all technologies are characterized by network effects
[9–11]. Thus, network externality theory is not general
enough to explain the full range of technology decisions.
Second, network externality theory does not consider the
information available to decision makers, and thus misses
an important aspect of the decision maker’s environment.
Third, network externality theory does not offer a good
explanation of failed or mistaken fads. Agents always
choose the most popular technology, even though it may
not be the most advanced, because the size of the network
outweighs the benefits of advanced features.
To overcome these difficulties, we take the perspective of
information cascade theory [12,13]. Information cascade
theory is based on sequential decisions in an informationpoor environment. In this context, decision makers are
faced with several courses of action and must make the best
choice. Unfortunately, no single decision maker has complete information; rather, each has some imperfect private
information. Decision makers can observe the actions, but
not the private information, of prior decision makers.
Observing the actions allows them to make inferences
about the prior decision makers’ private information. A
rational decision maker then incorporates those inferences
with his or her own private information to increase the likelihood of making a correct decision, which is a reasonable
response to an information-poor environment. However,
because decisions occur in a sequence, a mistake made by
a particular decision maker propagates to all future decision makers. If a decision maker has poor information that
UN
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
leads him to make an inappropriate decision, then all
future decision makers will incorporate that poor information into their information set. It is rational for each individual decision maker to make use of the information
implicit in the actions of others. However, as more individuals make the same decision, there is increasing pressure
for other decision makers to make the same decision,
regardless of their private information. Thus, we experience
a situation in which everyone begins making similar decisions and a fad for a poor product or service is begun.
Information cascades have been validated in laboratory
experiments [14–16], and behaviors consistent with information cascades have been observed in natural phenomena
[13].
Past research has studied this type of sequential decision
making in strings of individual decision makers. In the
present research, we extend this idea to explain how individuals infer group-level information in prior periods. We
posit that the members of a group of decision makers individually infer information about the group by examining
other group members’ prior decisions. Specifically, we propose that stock markets react simultaneously to events, but
that when another similar event occurs, individual investors infer information about the value of the event by the
market’s cumulative reaction to prior, similar events.
The context for our investigation is firms’ adoption of
electronic commerce technologies during the Internet stock
bubble. We argue that the Internet stock bubble was, to
some degree, a rational response by investors to the early
success of Internet firms. However, this response was made
in an environment with poor information about the true
costs and benefits of electronic commerce. A few early
apparent successes (e.g., Amazon, eBay, Dell) caused
informed investors to value electronic commerce very
highly, which drove up the returns of electronic commerce
initiatives [17]. Because the market responded well to these
early events, investors inferred high values for similar
events, which they incorporated into their own private valuations. This caused investors to value the events higher
than if they had relied solely on their own private information. The cumulative effect thereby drove valuations even
higher, and the abnormal valuations of early events propagated through future events.
It could be argued that the observed behavior indicating
the rise of DotCom stocks is just a variation on the ‘‘hothand phenomenon’’ (we thank an anonymous reviewer
for this excellent point; see also [18]). The ‘‘hot hand’’
refers to the belief among many people that ‘‘streaks’’ that
occur (such as a basketball player who has made many baskets consecutively) are likely to continue (and thus other
players should pass the basketball to him so he can take
additional shots). Although most associated with sports
activities, the belief applies to many superstitious behaviors
(such as people saying to a person who has had several
good outcomes recently that he should take a trip to Las
Vegas). (See [19] for a detailed explanation of this
phenomenon.)
F
In this paper, we investigate empirically the idea that
technology fads are a consequence of agents inferring
information about the benefits of a technology from the
observable actions of other agents. We use stock return
data from the ‘‘DotCom’’ era to investigate this question.
In particular, we investigate agents’ willingness to pay for
technology based on other agents’ willingness to pay for
similar technologies.
The paper proceeds as follows. In the next section we
review literature explaining technology fads. We then
develop a theoretical explanation of fads as information
processes. We then describe the data and conduct our analyses. We conclude with implications of our findings for
research and practice.
OO
58
59
60
61
62
63
64
65
66
67
68
69
70
71
E. Walden, G.J. Browne / Electronic Commerce Research and Applications xxx (2006) xxx–xxx
PR
2
No. of Pages 11, Model 5+
ARTICLE IN PRESS
Please cite this article in press as: E. Walden, G.J. Browne, Rational fads in investor reactions to electronic ..., Electron. Comm. Res.
Appl. (2006), doi:10.1016/j.elerap.2006.09.001
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
ELERAP 219
No. of Pages 11, Model 5+
ARTICLE IN PRESS
27 October 2006 Disk Used
E. Walden, G.J. Browne / Electronic Commerce Research and Applications xxx (2006) xxx–xxx
219
220
221
222
3. Theory
223
We now turn to developing a new model of information
cascades appropriate to investor decision making. First, we
introduce our assumptions and discuss their impact. Next,
we define the individual and market perceptions of value.
Then, we add the effects of information cascades on perceptions of value. Finally, we add network externalities and
time as control variables.
We make several assumptions to help formalize our
model. First, we assume that financial markets quickly
incorporate the information contained in an event into
the price of the stock (a standard assumption in much
stock market research). Second, we assume that all investors have access to the same public information (e.g., analyst recommendations, media reports, economic
indicators). These assumptions together allow us to control
for the effects of public information, which is interesting
but well studied. The release of public information is an
event, and if it is quickly incorporated into the price of a
stock, then when we observe the effects of events the price
of the stock will have already been adjusted for the
information.
The last critical assumption is that investors have some
private information that allows them to generate a private
valuation for an event. Every investor has private information simply by virtue of his unique experience. For example, the event of a firm developing a new type of laser
will have different meanings for an optics engineer, a financial analyst, and an attorney. The optics engineer may be
aware of a need for that type of laser for a new technology,
or may be aware of an even better laser that is about to be
announced. A financial analyst may know a great deal of
financial information about the firm. An attorney may be
aware of a lawsuit about to be filed by someone who was
blinded by a laser. Every individual is unique in the knowledge he uses to make sense of an event, and, hence, uses to
form an estimate of the value of the event. This means that
when investors observe an electronic commerce event they
will all form different perceptions of its worth. Damodaran
captures this notion perfectly when he says, ‘‘I have no
doubt that you will disagree with me on some of the inputs
I have used, and the values that you assign these firms will
be different from mine’’ [21, p. xvi].
Stock market reactions to events yield a unique opportunity to observe willingness-to-pay for IT implementations. When a firm makes an announcement of an event,
each individual in the market must evaluate her own willingness-to-pay to be part of that event. If the event in question is an IT implementation, then investors must form an
opinion about how much that implementation will increase
the value of a firm, and hence how much more (or less) they
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
PR
OO
F
initiatives were the consensus estimates of many knowledgeable individuals, each of whom had some private
information. Thus, information about the future could be
gained from the past.
CO
RR
E
CT
ED
Similarly, it could be argued that the fall of DotCom
stocks is due to a behavioral phenomenon known as the
‘‘gambler’s fallacy’’ [18]. The gambler’s fallacy refers to
the belief that relatively short sequences of events should
be representative of the event’s overall distribution. For
example, if a coin is flipped five times and lands on ‘‘heads’’
every time, most people will be willing to bet that a sixth
flip will land on ‘‘tails’’. This reflects the belief that six flips
should be roughly representative of the overall event distribution (50% each for heads and tails). However, the likelihood of the sixth flip landing on tails is only .50, and it is
thus irrational to bet more on tails than on heads for that
flip. In the context of the DotCom crash, it could be argued
that people simply began to believe that too many ‘‘heads’’
(i.e., too many rising technology stocks) had occurred in a
row and quit buying stocks (i.e., bet in the other direction).
However, there are several key differences between situations in which the hot hand and gambler’s fallacy explanations prevail and the current context. In contexts such as
‘‘hot’’ shooters in basketball games, the information environment is arguably quite rich. People observe the behavior directly and draw inferences based on their direct
observations. Second, in such situations, there is a lack of
apparent causal explanations for the occurrences (e.g., a
player making six baskets in a row or red occurring on
eight consecutive spins of a roulette wheel). Thus, people
invent causal explanations such as the hot hand, when in
fact statistical analyses reveal that the occurrences are simply random occurrences in known data distributions [20].
Decision errors occur due to the hot hand belief and gambler’s fallacy because the sequences are truly random and
no information about the future can be gained from the
past.
In contrast, in the context of investing in e-commerce
companies, the environment was information-poor because
investors were not able to observe technological innovations directly. Thus, they had to base their inferences about
the value of innovations and e-commerce companies selling
them on (at best) secondary information. Second, investors
were aware of many possible explanations for the value of
e-commerce stocks, but they had no good basis for determining causation (i.e., no good methods for performing
valuations) because of the scarcity of information available
to them. Thus, because investors’ private information was
poor and the resulting ability to make inferences about
causation (valuation methods for e-commerce stocks) was
also poor, our thesis is that investors chose to follow the
behaviors of other investors who (they assumed) had better
knowledge.1 Finally, prior realized values of e-commerce
UN
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
1
We do not completely discount the perspective of the hot hand and the
gambler’s fallacy as contributing explanations for stock market bubbles.
Since both problems arise as a result of people’s misconceptions about
random events, they provide an important perspective on stock market
‘‘runs’’, as suggested in recent research [18]. However, for reasons stated in
the text, we believe that our hypothesized reasons are stronger potential
explanations of bubbles.
3
Please cite this article in press as: E. Walden, G.J. Browne, Rational fads in investor reactions to electronic ..., Electron. Comm. Res.
Appl. (2006), doi:10.1016/j.elerap.2006.09.001
ELERAP 219
27 October 2006 Disk Used
si;t ¼ v þ ei;t ;
The variable v represents the true value of the technology
and e is an independent, mean-zero error term with constant variance across time and individuals.
Each agent evaluates his own private signal and submits
a bid for participation in the technology to a market maker
whose goal is to maximize transaction volume. This simplification allows us to model the daily behavior of equity
markets without having to consider the dynamic trading
considerations. The market is set up to maximize transaction volume (this is a reasonable assumption, since many
equities are handled by market makers whose job it is to
insure that all of the trades that can be made are made).
The market clears at v plus the median of the distribution
of e, because the median is the point at which exactly half
the people are willing to sell and half the people are willing
to buy. For simplicity, assume the expected value of the
median to be zero. The expected market-clearing return,
then, is v. This can be written as
310 Rt ¼ Rt ðst Þ ¼ v þ medianðet Þ;
where E½Rt ¼ v:
ð2Þ
CO
Here Rt is a function of the vector st = {s1,t, s2,t, . . . , sn,t}
and the median is a function of the vector et =
{e1,t, e2,t, . . . , en,t}.
For each technology initiative, the expected marketclearing return is the value of the technology. While there
will be variations because of the stochastic nature of the
private signals, there will be no consistent biases over time.
UN
311
312
313
314
315
316
317
ð1Þ
for all i and t:
318 3.1. Adding information cascades to the operational model
319
320
321
322
323
324
325
326
F
ED
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
VAR½ei;t ¼ r
and
CT
290
where E½ei;t ¼ 0
2
estimates than an individual [22]. We assume that agents
cannot observe other agents’ private signals, but can
observe the market-clearing return in prior periods.
We also assume that only one prior period is informative. There are at least three reasons why this is a reasonable assumption. First, and most importantly, there is no
compelling theoretical reason to believe that the return
from two periods prior would have a direct effect on the
return in the current period. Note that the return at time
t 2 does have an effect on the return in time t through
its effect on the return at time t 1. To include two periods
would give the return at t 2 both a direct effect and an
indirect effect, and this is not theoretically reasonable. Second, using two periods complicates the model considerably
by making it a polynomial in the lags. Thus, parsimony
suggests using only one lag. Third, when a second lag is
included and the model is estimated, the coefficient on
the first lag is unchanged and the coefficient on the second
lag is non-significant. Thus, theoretically and empirically, it
does not seem that a second lag is needed. Therefore, we
consider only one lag.
Finally, in addition to the current private signal and the
prior period’s observed return, we define a weighting term
a that measures the relative weight the decision maker
places on these two variables. Each agent’s private valuation of the technology becomes fi,t = f(si,t, Rt1,a), where
f is increasing in both si,t and Rt1. This indicates that
errors in the market-clearing return of the prior initiative
would tend to continue into the current period. Thus, if
for some reason the market overvalued (undervalued) a
technology in the earlier period, it would tend to stay overvalued (undervalued) in the current period.
Without loss of generality, we can normalize such that
0 6 a 6 1, where a = 1 ) f = f(si,t) and a = 0 ) f =
f(Rt1). The optimal (i.e., variance minimizing) value of a
will depend upon the relative information contained in s
and Rt1, or, in other words, on the inverse of the relative
standard error of s and pt1. Define rs and rp as the standard error of s and pt1, respectively. The weighting term
can be defined as a = a(rs, rp), where a is decreasing in
rs and increasing in rp. Thus, as the variance of the current
period’s private signal increases, the weight assigned to the
current period’s private signal decreases, and as the variance of the prior period’s return increases, the weight
assigned to the current period’s private signal increases.
To make the example concrete, assume that f(si,t, Rt1, a)
is a convex combination of si,t and Rt1. Thus, agents are
rational in the sense that they update their beliefs in a
Bayesian manner. Then each agent’s valuation of the technology is
OO
are willing to pay to own the stock in question. The individuals then rapidly buy and sell until the price reaches a
new equilibrium. The difference between the old price and
the new price represents the return to the electronic commerce initiative. Thus, we are able to observe the consensus
opinion about the information content of an event, which
is backed by investors’ real money and represents the
potential for real gain or loss. This is unique because investors must commit resources to the point of equilibrium.
Thus, they are motivated to make use of all the information available to them, even if it is sparse.
To operationalize the theory, assume there is a market
of x risk-neutral agents. Each agent i receives a private signal about the value of a technology initiative at time t of
the form
RR
E
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
E. Walden, G.J. Browne / Electronic Commerce Research and Applications xxx (2006) xxx–xxx
PR
4
No. of Pages 11, Model 5+
ARTICLE IN PRESS
The model above assumes that agents ignore the information contained in the realized market-clearing return
of prior initiatives. In general, for well-behaved distributions on e, the market-clearing return will have lower variance than the individual private signals and thus be quite
informative. From a human psychology perspective, most
people know they do not know everything and realize that
a group of independent decision makers can make better
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
f ðsi;t ; Rt1 Þ ¼ asi;t þ ð1 aÞRt1
¼ av þ ð1 aÞRt1 þ aei;t ;
ð3Þ 378
where 0 6 a 6 1.
379
To form a measure for estimation, it is necessary to 380
determine the overall market return, which is observable. 381
Please cite this article in press as: E. Walden, G.J. Browne, Rational fads in investor reactions to electronic ..., Electron. Comm. Res.
Appl. (2006), doi:10.1016/j.elerap.2006.09.001
ELERAP 219
No. of Pages 11, Model 5+
ARTICLE IN PRESS
27 October 2006 Disk Used
E. Walden, G.J. Browne / Electronic Commerce Research and Applications xxx (2006) xxx–xxx
The reader will recognize Eq. (5) as the standard form of an
autoregressive process. Such a process can be estimated by
ordinary least squares. Note that the model uses only one
lag because the latest estimate of value is contained solely
in that lag. There is, however, carryover from earlier lags
because the value estimate in any period t 1 is a function
of the value in period t 2. Thus, the behavior of prior
periods is accounted for in the model although we only estimate the direct impact of one lag.
This discussion leads to a new definition of an information cascade for the continuous version of the theory presented here.
403
404
405
406
407
Information Cascade: An information cascade occurs,
under the preceding assumptions, when a market’s valuation of the current period’s adoption decision is significantly dependent upon the market’s valuation of the
prior period’s adoption decision.
408
409
410
411
412
413
414
415
416
Thus, to test for the presence of an information cascade, b1 from Eq. (5) is estimated. An information cascade indicates that individual market participants are
not simply applying their own private signals about the
value of a technology, but are also applying public information about other participants’ valuations as inferred
from the emergent market return in prior periods. Thus,
a positive coefficient for b1 indicates information cascade
behavior.
ED
391
392
393
394
395
396
397
398
399
400
401
402
v ¼ vðtÞ ¼ k 2 þ k2 t;
UN
CO
We have assumed that the true underlying value, v, of
the technology is constant. However, this may not be the
case. As noted above, in e-commerce research network
externalities [5,7,8,23,24] have been invoked to help explain
adoption of e-commerce technologies. For example, the
more users on the Internet, the more value to any potential
adopter because more users leads to a greater potential
range of information available to additional users. However, while the Internet as a communication medium is certainly characterized by network externalities, it does not
necessarily follow that the Internet as a business environment is characterized by those same externalities. This is
an empirical question, and one that must be addressed
given the operationalization of our model of information
cascades.
To incorporate the possibility of network externalities,
we redefine v as v(n), to recognize that value is a function
of the number of participants in the network. This model
will occur when
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
ð7Þ 462
where k2 is the value of the technology at the beginning of
the sample time frame, t is the time since the beginning of
the sample period, and k2 is a constant to be estimated.
Again, this assumes that the change over time is linear.
Lastly, it is possible that both effects occur simultaneously, so that v is a function, v(n, t), of both time and
the number of other participants in the market. It is defined
as
CT
RR
E
417 3.2. Including the possibility of changing technology value
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
F
ð5Þ
where k1 is some inherent value of being on the network, n
is the number of other firms on the network, and d is a constant to be estimated. As a proxy for n we use the cumulative number of adoptions in our data set at each point in
time. This implicitly assumes that our observation of the level of adoption is proportional to the actual level of adoption. It also assumes that the rate of change of value is
linear. Research in rational expectations supports this
proxy. The notion is that ‘‘. . . due to network externalities,
it is in the best interest of each firm within a sub-group sharing similar characteristics and/or serving similar markets to
adopt simultaneously’’ [25, p. 5]. Thus, it does not necessarily matter whether the participants are using the network to
communicate with one another. Even if the firms are only
building a network that others will use, they should still
adopt together.
Another possibility is that the value does not depend on
the number of adopters but still changes over time. For
example, if technology improves over time, it is reasonable
to assume that the value of adopting that technology will
increase. This leads to a reformulation of value as a function of time. This model is defined when
OO
387 This can be rewritten as
388
390 Rt ¼ b0 þ b1 Rt1 þ nt :
ð6Þ 438
vðnÞ ¼ k 1 þ d1 n;
PR
382 The market realization is the median of all individual
383 valuations,
384
386 Rt ¼ av þ ð1 aÞRt1 þ medianðaei;t Þ:
ð4Þ
5
v ¼ vðn; tÞ ¼ k 3 þ d3 n þ k3 t:
463
464
465
466
467
468
469
470
471
ð8Þ 473
Given these models, Eq. (4) becomes
Rt ¼ aðk þ dn þ ktÞ þ ð1 aÞRt1 þ medianðaei;t Þ:
This can be rewritten as
Rt ¼ c0 þ c1 Rt1 þ c2 n þ c3 t þ nt ;
474
ð9Þ 476
477
478
ð10Þ 480
where c2 = 0 if j = 2 and c3 = 0 if j = 1.
481
3.3. Including the possibility of changes in the relative
weighting
482
483
As discussed above, the optimal weighting is a function
of the noise in the private signal and the noise in the prior
period’s return so that a ¼ aðr2s ; r2Rt1 Þ. While it is conceptually difficult to measure the noise in the private signal for
the same reasons it is difficult to measure the private signal
itself, it is possible to estimate the noise in the prior period’s return. The estimate of the noise in the prior period’s
return is given by
2
^2Rt1 ¼ ^nt1 ;
r
ð11Þ
484
485
486
487
488
489
490
491
Please cite this article in press as: E. Walden, G.J. Browne, Rational fads in investor reactions to electronic ..., Electron. Comm. Res.
Appl. (2006), doi:10.1016/j.elerap.2006.09.001
493
ELERAP 219
27 October 2006 Disk Used
E. Walden, G.J. Browne / Electronic Commerce Research and Applications xxx (2006) xxx–xxx
516 This can, in turn, be rewritten as
517
2
2
Rt ¼ u0 þ u1 ½^
nt1 þ u2 Rt1 þ u3 ½^
nt1 Rt1 þ nt :
519
The secondary hypothesis of interest concerns the relative weighting of private information and information contained in the prior period’s return. This is captured by the
coefficient x6. The weight investors apply to the prior period’s return should decrease as the prior period’s return
becomes noisier. Thus, x7 should be negative and significant if investors are placing less weight on the prior period’s return as the variance of the prior period’s return
increases. This leads to our second hypothesis:
RR
E
522 3.4. Combining all models
CO
523
To derive the fullest estimate we can combine Eqs. (8)
524 and (13) as
h
i
2
Rt ¼ a1 þ a2 ½^
nt1 ½k 3 þ d3 n þ k3 t
h
i
2
^
þ
1
a
þ
a
½
n
ð15Þ
Rt1 þ nt :
1
2
t1
526
UN
527 This can be rewritten as
528
2
2
Rt ¼ x0 þ x1 n þ x2 t þ x3 ½^
nt1 þ x4 n ½^
nt1 2
2
^
^
t
½
n
R
þ
x
½
n
þ
x
þ
x
Rt1 þ nt :
5
t1
6
t1
7
t1
530
ð16Þ
We can now propose hypotheses based on this equation.
The first hypothesis is simply that investors rely on the
information contained in prior realizations of electronic
commerce returns when formulating their current evaluation of an electronic commerce return. This suggests that
2
For our particular data, when at is defined in this manner, it ranges
from 0.243 to 0.756 and thus satisfies the assumption that it be between
zero and one. However, for other data the estimation might not proceed so
conveniently and it may be necessary to constrain the function to the zero
to one interval. A logistic or sigmoidal transformation would accomplish
this. We thank an anonymous reviewer for this useful suggestion.
542
543
544
545
546
547
548
549
550
551
Hypothesis 2 (Relative Weighting hypothesis): Increas- 552
ing the variance of the prior period’s return decreases 553
the effect of prior returns on the current return.
554
We also include network externalities and time as control variables. We offer no hypotheses about these variables,
but they both provide alternative explanations of the market’s behavior that could, in the absence of control, generate similar behavior. For example, if the value of an
electronic commerce initiative increases as the number of
adopters increases, then we will see value increasing consistently over time. Each new electronic commerce initiative
will increase the value of future electronic commerce events.
If we do not provide this control, then even in the absence
of the process we described there will be a significant effect
of prior period returns on current period returns. Similarly,
if the return to electronic commerce were simply rising over
time due to a wealth effect of investors or improving technology or some other factor, we would expect prior returns
to predict current returns. Thus, we control for the number
of prior adoptions and a linear time trend.
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
4. Data
572
The data for this study come from firms’ adoptions of
electronic commerce technologies. This is a particularly
good environment for consideration because e-commerce
is characterized by both high uncertainty and poor observability of private signals. Each data point is a firm’s public
announcement of an e-commerce initiative in the media.
We collected the data from a full text search of company
announcements related to e-commerce in the period
between January 1, 1999 and December 31, 2000, using
two leading news sources: PR Newswire and Business
Wire. Following prior literature [26,27], we searched
Lexis/Nexis for announcements containing the words
launch or announce within the same sentence as the words
online or commerce and .com and NYSE, NASDAQ, or
AMEX. The search yielded 4744 potential announcements
– 2170 in 1999 and 2574 in 2000.
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
CT
ð14Þ
520 Given the specification of the model, we would expect
521 /2 > 0 and /3 < 0.
531
532
533
534
535
538
Hypothesis 1 (Information Cascades hypothesis): The 539
effect of prior returns on the current return is positive 540
and significant.
541
ED
511 This allows at to be a linear function of [nt1]2.2 Eq. (4) can
512 then be rewritten as
513
h
i
2
Rt ¼ a1 þ a2 ½^
nt1 v
h
i
2
^
þ
a
½
n
ð13Þ
þ
1
a
Rt1 þ nt :
1
2
t1
515
the coefficient on return should be positive and significant. 536
Thus, our first hypothesis is:
537
F
where nt1 is the error term from the prior period, and the
Ù indicates an estimate. Thus, the prior period’s estimated
variance depends on the distance from the predicted value
to the last period’s return.
This implies that the weighting scheme changes over
time in response to the noise in the market so that
at = a([nt1]2) and a is increasing in [nt1]2. This means
that, as suggested earlier, the weight applied to the private
signal increases as the noise in the prior period’s return
increases and that the weight applied to the prior period’s
return decreases as the noise in the prior period’s return
increases. At a high level, this means that investors rely less
on the prior period’s return when it is further away from its
expected value.
To operationalize this, let
^ 2
ð12Þ
510 at ¼ a1 þ a2 ½nt1 :
OO
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
PR
6
No. of Pages 11, Model 5+
ARTICLE IN PRESS
Please cite this article in press as: E. Walden, G.J. Browne, Rational fads in investor reactions to electronic ..., Electron. Comm. Res.
Appl. (2006), doi:10.1016/j.elerap.2006.09.001
ELERAP 219
No. of Pages 11, Model 5+
ARTICLE IN PRESS
27 October 2006 Disk Used
E. Walden, G.J. Browne / Electronic Commerce Research and Applications xxx (2006) xxx–xxx
Table 2
Independent variables
Variable
Description
Rt
Average abnormal return on electronic commerce initiatives on
day t
The number of days since December 31, 1998
The cumulative number of electronic commerce events
The square of the estimated error of R in time t 1, that is,
(actual Rt1 predicted Rt1)2
F
Time
N
r2t1
635
636
637
638
639
5. Results
640
PR
OO
the data are presented in Table 1, where price and volume
are the average price and volume of the stock over the twoyear period.
We use four variables in our estimation. These are
described in Table 2.
To correct for overall market movements, it is necessary 641
to calculate an abnormal return according to the following 642
formula.
643
AR ¼ Rs;t ðds þ cs Rm;t Þ:
ð17Þ 645
ED
We use Subramani and Walden’s definition of an event:
‘‘The criteria we used to identify an announcement as an
event was that the news item be an announcement of a new
electronic commerce-related initiative or the extension or
expansion of an existing initiative. Miscellaneous announcements such as estimates of expected earnings, news about
personnel changes, and site traffic volumes, etc. were discarded’’ [27, p. 141]. Electronic commerce is an activity
related to the trading of goods and services facilitated by
Internet technologies.
To insure the consistency and accuracy of the coding,
two coders – one of the authors and a graduate student
unfamiliar with the hypotheses of the study – worked independently. Each coder performed his own analysis of the
data. The data were then matched, and one of the coders
reconsidered any disagreements. In this phase, the coder
could change his and only his original coding if on the second examination he agreed with the other coder. If the
coder did not agree, he wrote his own comments as to
why he believed his original coding was correct. The data
were then given to the other coder, who had a chance to
change his original coding based on the first coder’s comments. Any disagreements that could not be resolved by
this two-step process were then decided in a face-to-face
meeting between the two coders.
To focus attention on the task at hand, we excluded several different types of announcements from the coding.
Coders excluded marketing announcements or news of customer acquisition and minor, temporary promotions such
as Christmas or Super Bowl specials. Coders also removed
earnings announcements and management changes by
firms. Coders eliminated announcements of minor Web
Site redesign, unless the redesign developed new capabilities. Coders also dropped announcements pertaining to
mergers and acquisitions. To eliminate the bias introduced
by thinly-traded stocks that might be illiquid and low-value
stocks that are not representative of the broad market, we
removed firms with an average share price of less than one
dollar and firms with an average trading volume of less
than 50,000 shares per day. This procedure is consistent
with prior literature [27]. Of the 4744 potential announcements, 2097 were coded as announcements of e-commerce
initiatives and 1273 contained enough data to estimate
abnormal returns. When there were multiple announcements in a single trading day the average return was used,
resulting in a total of 402 data points. The demographics of
The term in parentheses is the expected return on the stock,
where d and c are estimates from day 251 to day 2 before
the announcement (see [28] for detailed description of the
technique). The R terms are returns, with the subscript t
denoting time, the subscript s referring to a specific stock,
and the subscript m referring to the market return, in this
case the return on the S&P 500.
Finally, to allow for announcements that occurred
before the market opened or after the market closed, the
three-day cumulative abnormal return was used. This
return measure is the summation of abnormal returns starting on the day before the event and ending on the day after
the event [29,30]. The models in Eqs. (10), (14) and (16)
were then estimated using the cumulative abnormal return
data. The results are shown in detail in Table 3.
As the table shows, the empirical testing supported the
information cascade hypothesis in all specifications. In fact,
in all specifications, the t-statistic for the effect of the prior
period’s return was the greatest. Moreover, the coefficient
was not sensitive to the inclusion of the cumulative number
of adoptions or the time of adoption.
The Relative Weighting hypothesis (H2) was supported
in the absence of a control for time and network effects.
While it has the correct sign in the full model, it fails to
achieve significance. Thus, support is mixed for the hypothesis that investors reduce their reliance on prior period
returns when those returns are more extreme.
When the information content of the prior period was
included in the form of the estimated standard error from
the prior period, the influence of the prior period increased,
as suggested by information cascade theory. The coefficient
on the interaction between prior period return and prior
period information content was negative, as predicted,
and this coefficient was significantly negative in the
CT
RR
E
CO
Table 1
Data demographics
Price
Volume
Average
cumulative
abnormal
return by
day
UN
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
7
Mean
Median
SE
Min
Max
$29.96
3,021,852
0.06%
$24.56
591,744
0.05%
$23.79
6,412,472
8.62%
$1.02
50,256
35.48%
$158.57
33,475,710
65.47%
Please cite this article in press as: E. Walden, G.J. Browne, Rational fads in investor reactions to electronic ..., Electron. Comm. Res.
Appl. (2006), doi:10.1016/j.elerap.2006.09.001
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
8
Intercept
Rt1
AR(1)
Time
RR
AR(1) + time
0.01
1.42
0.13***
2.61
3.00E 5
1.56
EC
D
Rt1*r2t1
Time* r2t1
r2t1
R2
Adjusted R2
White’s heteroskedasticity test (p value)
*
**
***
At the 0.10 level.
At the 0.05 level, and.
Indicates significance at the 0.01 level.
0.018
0.016
0.458
0.024
0.020
0.003***
0.023
0.018
0.007***
AR(1) + variance interaction
0.04**
2.31
0.12**
2.43
3.60E 4**
1.99
1.70E 4*
1.82
TE
r2t1
N
AR(1) + time + N
8.89E 3
1.17
0.13***
2.65
1.00E 5
1.34
N
*
AR(1) + N
0.032
0.025
0.003***
0.01
0.34
1.27***
2.12
PR
Full model
0.04**
2.47
0.24
3.64
3.90E 4**
2.03
1.82E 4*
1.84
.28
0.40
1.18
1.21
1.80E 4
0.02
3.11E 4
0.07
0.069
0.053
0.241
2.03E 3
0.45
0.26
4.25
0.052
0.045
0.861
OO
No. of Pages 11, Model 5+
F
ARTICLE IN PRESS
4.88E 4
0.11
0.14***
2.76
ELERAP 219
Parameter
27 October 2006 Disk Used
CO
Table 3
Autoregressive results (t-statistics on bottom)
E. Walden, G.J. Browne / Electronic Commerce Research and Applications xxx (2006) xxx–xxx
Please cite this article in press as: E. Walden, G.J. Browne, Rational fads in investor reactions to electronic ..., Electron. Comm. Res.
Appl. (2006), doi:10.1016/j.elerap.2006.09.001
UN
ELERAP 219
No. of Pages 11, Model 5+
ARTICLE IN PRESS
27 October 2006 Disk Used
E. Walden, G.J. Browne / Electronic Commerce Research and Applications xxx (2006) xxx–xxx
6.2. Limitations
763
As with any research effort, there are certain limitations
to the work performed in this paper. A central limitation of
the study is the period examined. Undoubtedly, the time
period under consideration was not a typical period in
stock market history. We chose it specifically to examine
behavior in fad-like conditions. Therefore, it is not clear
that we can generalize the results to other periods. On
the other hand, information cascade theory suggests that
some fads are rational specifically because relying on the
behavior of others improves one’s chance of making a good
decision. In fact, information cascade theory shows that
people will, on average, do better by relying on the signals
of others. The theory mainly focuses on what happens
when things go awry because that is more interesting, but
following the model we suggest does in fact result in better
estimates over time than relying on personal information
alone. Thus, the results may be generalizable, with this period simply representing an extreme outcome.
There are also several statistical limitations that are
worth noting. First, on many days there were multiple
announcements, and we used the average abnormal return
to all announcements on a day. However, this is only one
way to deal with the problem that many firms may pursue
a particular course of action on a given day. We chose to
look at day-to-day changes rather than event-to-event
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
PR
OO
F
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
CT
713 6. Conclusion
714 6.1. Contribution
CO
We have proposed a theoretical explanation of the DotCom fad of 1999–2000. Specifically, we posited that the
willingness of investors to commit capital to successive
electronic commerce initiatives stemmed from the lack of
information in the environment. To supplement their
own limited information investors used the pooled information present in the market’s previous reactions to e-commerce initiatives. Each successive positive market reaction
modified investor perceptions of electronic commerce in a
positive manner, and each successive negative reaction
modified investor perceptions in a negative manner, thus
creating reinforcing effects that caused a rapid rise, then
fall, in perceptions of electronic commerce value. The primary implication of this research is that IT will continue
to be characterized by fads. IT is complex and requires
complementary organizational changes to realize value.
This means that the real benefits of IT are not known when
firms have to make decisions about them. One mechanism
rational individuals can use to help them value novel IT is
UN
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
to observe how others value the same IT. While this helps
each decision maker individually, it leads to fads.
These fads can move in both directions, with adopters
overvaluing IT or undervaluing IT. Therefore, both
researchers and organizations must be careful in evaluating
new IT. This means that decision makers should collect as
much private information about the value of IT as possible
before incorporating the behaviors of others. However, the
behavior of others should be incorporated because it does
lead to better decisions, but it should be discounted
appropriately.
An important implication for practitioners is that theory
should be used when valuing IT. Theory that bases its
arguments on well-tested logic can facilitate realistic valuations of IT. For example, while a common perception at
the time of the DotCom bubble in the popular press was
that the Internet was a new economy with new rules, and
that firms like eToys and Pets.com would replace firms like
Toys R Us and Petsmart, researchers used the resourcebased view of the firm to argue, ‘‘. . .the initial disadvantages
of non-net firms from being on the learning curve with respect
to Internet technologies and the novel e-commerce context
are likely to be largely offset by the considerable advantages
derived from the migration of existing firm competencies to
e-commerce operations’’ [33, p. 195]. Use of theory provides
a foundation for decision making, which can lead to the
avoidance of irrational fads. This is particularly important
in domains such as IT, in which the artifacts being evaluated change frequently.
ED
return-only model. Introducing network externalities (as
measured by cumulative adoption) and time did not affect
the coefficient of prior period return, but did render the
interaction coefficient non-significant. This pattern is indicative of information cascades.
The models above are robust to autocorrelation because
autocorrelation is captured by the model; however, three of
the models show signs of heteroskedasticity. Re-estimating
the models using White’s heteroskedasticity consistent
covariance matrix changes the significance of the effect on
the lagged variable slightly, dropping the t-statistics from
2.61, 2.65, and 2.43 to 1.73, 1.75, and 1.62, respectively.
This reduces the significance levels to 0.085, 0.081, and
0.106. However, this does not change the overall pattern
of the results. Lagged values are still the most important
determinant of current values across all models. Moreover,
the full model does not have heteroskedasticity problems,
nor does the lagged-value-only model, which suggests that
the impact of lagged values is robust.
Another point to note is that the R2 for each model is
lower than we usually see. This is typical of event analysis
because the variance of returns is very high. R2 measures
the ratio of explanation to noise, and the stock market is
very noisy. Every day millions of things occur that move
stock prices. As can be seen from Table 1, the mean return
in this study was 0.06%, while the standard error was
8.62%, which is more that 100 times larger. The return
on an electronic commerce initiative depends on a huge
number of variables, including factors about the initiative,
who is undertaking the initiative, and the environment in
which it occurs. There are certainly other systematic effects
that we have not captured, but our goal is to explain the
causal mechanisms of one effect that has significant impact.
RR
E
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
9
Please cite this article in press as: E. Walden, G.J. Browne, Rational fads in investor reactions to electronic ..., Electron. Comm. Res.
Appl. (2006), doi:10.1016/j.elerap.2006.09.001
ELERAP 219
27 October 2006 Disk Used
References
858
[1] E. Abrahamson, Managerial fads and fashions: the diffusion and
rejection of innovations, Academy of Management Review 16
(3(July)) (1991) 586–612.
[2] R.G. Fichman, The diffusion and assimilation of information
technology innovations, in: R.W. Zmud (Ed.), Framing the
Domains of IT Management: Projecting the Future. . .Through
the Past, Pinnaflex Educational Resources, Cincinnati, OH, 2000,
pp. 105–128.
[3] R.J. Kauffman, X. Li. Payoff externalities, informational cascades
and managerial incentives: a theoretical framework for it adoption
herding, in: 2003 INFORMS Conference on IS and Technology,
Atlanta, GA, 2003.
[4] D.Q. Mills, Buy, Lie and Sell High: How Investors Lost Out on
Enron and the Internet Bubble, Financial Times Prentice Hall, Upper
Saddle River, NJ, 2002.
[5] E. Brynjolfsson, C.F. Kemerer, Network externalities in microcomputer software: an econometric analysis of the spreadsheet market,
Management Science 42 (12) (1996) 1627–1647.
[6] N. Economides, C. Himmelberg, Critical mass and network evolution
in telecommunications, in: G. Brock (Ed.), Toward a Competitive
Telecommunication Industry, Lawrence Erlbaum Publishers, Mahwah, NJ, 1995, pp. 47–63.
[7] J.M. Gallaugher, Y.M. Wang, Network externalities and the provision of composite IT goods supporting the e-commerce infrastructure,
Electronic Markets 9 (1) (1999) 14–19.
[8] R.J. Kauffman, J. McAndrews, Y.M. Wang, Opening the ‘Black Box’
of network externalities in network adoption, Information Systems
Research 11 (1) (2000) 61–82.
[9] R.J. Kauffman, E.A. Walden, Economics and electronic commerce:
survey and directions for research, International Journal of Electronic
Commerce 5 (4) (2001) 5–116.
[10] S.J. Liebowitz, S.E. Margolis, Network externality: an uncommon
tragedy, The Journal of Economic Perspectives 8 (2) (1994) 133–150.
[11] S.J. Liebowitz, Re-thinking the Networked Economy: The Real
Forces that Drive the Digital Marketplace, Amacom Press, New
York, NY, 2002.
[12] S. Bikhchandani, D. Hirshleifer, I. Welch, A theory of fads, fashion,
custom and cultural change as informational cascades, Journal of
Political Economy 100 (5) (1992) 992–1026.
[13] S. Bikhchandani, D. Hirshleifer, I. Welch, Learning from the
behavior of others: conformity, fads, and informational cascades,
The Journal of Economic Perspectives 12 (3) (1998) 151–170.
[14] L.R. Anderson, C.A. Holt, Classroom games: information cascades,
The Journal of Economic Perspectives 10 (4) (1996) 187–193.
[15] L.R. Anderson, C.A. Holt, Information cascades in the laboratory,
The American Economic Review 87 (5) (1997) 847–862.
[16] L.R. Anderson, Payoff effects in information cascade experiments,
Economic Inquiry 39 (4) (2001) 609–615.
CO
RR
E
CT
An area for future research is a consideration of the
scope of e-commerce initiatives that lead to fads. Our unit
of analysis is e-commerce initiatives generally, but other
researchers have delineated different categories of e-commerce initiatives [9,31,32]. Future research could examine
which categories of electronic commerce activities are most
subject to fads and what sorts of spillovers one category
has on another. For example, consumer portal investors
may react more strongly to market information than online
travel services investors. Similarly, content provider investors may react more strongly to the information contained
in the market’s reaction to initiatives by online retailers
than to the information contained in market reactions to
business-to-business exchanges. Elucidating theory about
who receives information from whom about the value of
IT initiatives has great potential value.
Another area for future research is the decision processes of the electronic commerce firms themselves. In this
case, our unit of analysis was investors, but firm behavior is
likely to be dependent upon investor perceptions and upon
other firms’ behavior. However, the dependencies are likely
to be different, largely because firms have a great deal of
time and resources to devote to their decisions, whereas
individual investors in modern markets must often react
very quickly to information without access to such
resources. Moreover, firm decisions are much more complicated than simply deciding on a price. Of course, firms
exercise direct control over the execution of the decision
and can choose to modify their decisions in complicated
ways over time. Taken together, the firm decision is likely
to be different than the decisions of investors.
In sum, this research offers a promising new direction
for research on technological fads. Our results indicate that
investors put significant weight on prior market sentiment
UN
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
ED
809 6.3. Future research
about the value of technology when they make their own
decisions about the value of similar technology. We propose that this occurs because the market is a means of
aggregating the scarce information available about the
potential benefits of new technologies. Thus, we follow
Surowiecki in suggesting that people make use of the wisdom of crowds [22]. Moreover, people should make use
of such wisdom, as it often leads to better individual decisions. However, it can also lead to pathological outcomes,
such as electronic commerce bubbles and busts, for the system as a whole. Thus, further investigations into the behavior of individuals in valuing technological innovations in
contexts involving potential information cascades are
warranted.
F
changes. Also, we do not allow for information decay. That
is, we only count periods as days that electronic commerce
initiatives were announced. Several days may actually
elapse between periods, particularly on weekends. It is
not clear whether it makes sense to discount the effects of
a Friday event on a Monday return more than the effects
of a Tuesday event on a Wednesday return. Finally, it is
important to note that not all events are created equal.
Some electronic commerce announcements may be profound while others may be trivial. We do not classify events
into different types. In general, this would tend to increase
the error of measure, and thus reduce the power of the
tests, so the fact that the tests are significant is even more
surprising. However, it is possible that our theory holds
for some subsets of electronic commerce initiatives and
not others. In particular, it seems reasonable to argue that
those for which there is a great deal of information would
be less susceptible to the behavior of others, while those
that are particularly novel would be more susceptible to
the processes we describe.
OO
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
E. Walden, G.J. Browne / Electronic Commerce Research and Applications xxx (2006) xxx–xxx
PR
10
No. of Pages 11, Model 5+
ARTICLE IN PRESS
Please cite this article in press as: E. Walden, G.J. Browne, Rational fads in investor reactions to electronic ..., Electron. Comm. Res.
Appl. (2006), doi:10.1016/j.elerap.2006.09.001
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
ELERAP 219
No. of Pages 11, Model 5+
ARTICLE IN PRESS
27 October 2006 Disk Used
E. Walden, G.J. Browne / Electronic Commerce Research and Applications xxx (2006) xxx–xxx
PR
OO
F
[26] M.J. Dardan, A.C. Stylianou, The impact of fluctuating financial
markets on the signaling effect of e-commerce announcements, in:
Twenty-Second International Conference on Information Systems
(ICIS 2001), New Orleans, LA, 2001.
[27] M.R. Subramani, E.A. Walden, The impact of e-commerce
announcements on the market value of firms, Information System
Research 12 (2) (2001) 135–154.
[28] A. McWilliams, D. Siegel, Event studies in management research:
theoretical and empirical issues, Academy of Management Journal 40
(3) (1997) 626–657.
[29] D. Chatterjee, V.J. Richardson, R.W. Zmud, Examining the shareholder wealth effects of announcements of newly created CIO positions,
Management Information Systems Quarterly 25 (1) (2001) 43–70.
[30] S. Rajgopal, M. Venkatachalam, S. Kotha, Managerial actions, stock
returns, and earnings: the case of business-to-business internet firms,
University of Washington, 2001.
[31] M.R. Subramani, E. Walden. Economic returns to firms from
business-to-business electronic commerce initiatives: an empirical
examination, in: Twenty-First International Conference on Information Systems, Brisbane, Australia, 2000.
[32] P. Weill, M. Vitale, Place To Space: Migrating to e-Business Models,
Harvard Business School Press, Cambridge, MA, 2001.
[33] M.R. Subramani, E.A. Walden. The ‘Dot Com’ effect: the impact of
e-commerce announcements on the market value of firms, in: The
20th International Conference on Information Systems (ICIS-99),
Charlotte, NC, 1999.
CO
RR
E
CT
ED
[17] P. Elstrom, Everybody into the pool; venture capital is drawing cash
from everywhere – thanks to amazing returns, in BusinessWeek, July
26, 1999, p. EB58.
[18] J. Johnson, G.F. Tellis, Blowing bubbles: heuristics and biases in the
run-up of stock prices, Journal of the Academy of Marketing Science
33 (4) (2005) 486–503.
[19] T. Gilovich, How We Know What Isn’t So, Free Press, New York,
1993.
[20] T. Gilovich, R. Vallone, A. Tversky, The hot hand in basketball: on
the misperception of random sequences, Cognitive Psychology 17
(1985) 295–314.
[21] A. Damodaran, The Dark Side of Valuation, Financial Times
Prentice Hall, New York, 2001.
[22] J. Surowiecki, The Wisdom of Crowds: Why the Many Are
Smarter Than the Few and How Collective Wisdom Shapes
Business, Economies, Societies and Nations, Doubleday, New
York, 2004.
[23] X. Li, Informational cascades in IT adoption, Communications of the
ACM 47 (4) (2004) 93–97.
[24] F.J. Riggins, C.H. Kriebel, T. Mukhopadhyay, The growth of
interorganizational systems in the presence of network externalities,
Management Science 40 (8) (1994) 984–998.
[25] Y.A. Au, R.J. Kauffman, and F.J. Riggins. A rational expectations
theory of technology adoption: evidence from the electronic billing
industry, in: Workshop on Information Systems and Economics
(WISE 2005), Irvine, CA, 2005.
UN
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
11
Please cite this article in press as: E. Walden, G.J. Browne, Rational fads in investor reactions to electronic ..., Electron. Comm. Res.
Appl. (2006), doi:10.1016/j.elerap.2006.09.001
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959