The Value of Social Media for Predicting
Stock Returns
Michael Nofer
The Value of Social
Media for Predicting
Stock Returns
Preconditions, Instruments
and Performance Analysis
With a Foreword by Prof. Dr. Oliver Hinz
Michael Nofer
Darmstadt, Germany
Dissertation, TU Darmstadt, Germany, 2014
ISBN 978-3-658-09507-9
ISBN 978-3-658-09508-6 (eBook)
DOI 10.1007/978-3-658-09508-6
Library of Congress Control Number: 2015935424
Springer Vieweg
© Springer Fachmedien Wiesbaden 2015
This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or
part of the material is concerned, speci¿ cally the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on micro¿lms or in any other physical way, and
transmission or information storage and retrieval, electronic adaptation, computer software, or by
similar or dissimilar methodology now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this
publication does not imply, even in the absence of a speci¿c statement, that such names are
exempt from the relevant protective laws and regulations and therefore free for general use.
The publisher, the authors and the editors are safe to assume that the advice and information in this
book are believed to be true and accurate at the date of publication. Neither the publisher nor the
authors or the editors give a warranty, express or implied, with respect to the material contained
herein or for any errors or omissions that may have been made.
Printed on acid-free paper
Springer Vieweg is a brand of Springer Fachmedien Wiesbaden
Springer Fachmedien Wiesbaden is part of Springer Science+Business Media
(www.springer.com)
Forew d
Firms
impact
user
conta s,
"Like"
their
cbines
tha
rev al
.s enisub-e
like
Tecbnis h
special.
the
can
Univers t
Darmst d
Inter
poten ial
ap ro ch
in
question
and
si not
.tuo
an
easy
stock
work.
wil
al
es ay
in
depth
at
whic
dna
and
the
domain
bo k
the
undertaki g
practi oners
this
such
methods.
thoug f l
of
the
tries
with
of
set
hand
.syad
Big
Dat
a gold
rush.
are
also
valu ot a s e
the
the
precondiI t is o n e
gains.
a great
care,
examin s
phenom
a dif er nt
is t h e
res a ch
This
car ied
both,
other
poten ial-
dat
and
perfo mance
reto ns
the
of
dis erta on
stock
in
Nofer's
mid le
!bat
this
on
rof
ther by
thes
the
Nofer's
the
valu b e
can
at r c ive
beli v
,stnirp o f
se m
and
in
madat
dig tal
dna
os
like
consumer
ac es ibl
dleif
on
tha
as et
leav
cur ently
social
cli k ng
agre
uniq e
tha
this
are
man ge
by
tangible
ot
networks
i s fre ly
dev lopm nts
used
to
bil on
Many
their
user
one
tas e
recom ndati s
hand
be
base
ac es
thes
of
over
websit .
their
of
social
a lot o f thes
dat
ot f o r e c a s t
fut re
monetiz d.
This
poten ial
makes
t o b e the n e w
oil a n d
we
Res arch
and
practi oners
alike
beating
eht
stock
market.
Michael
o f social
media
rof
predict ng
instrume
and
final y
as e
tsrif
dis erta on
!bat
examin s
such
a huge
dat
base
and
with
o f this
bo k
A com n
them
the
importan
and
timely
examin s
t h e is ue
a t hand.
empir cal
work
!bat
has
be n
I hig ly
recom end
this
bo k
ot
i n predict ng
eht
dev lopm nt
o f the
mileston
ot b e o n e
hig ly
ben fit
from
Michael
this
publicat on
and
I beli v
tha
dat
one
make
many
by
on
User
or
user
their
justif ed
mainly
number
has
com uni ate
on
be
a large
their
to
av il b e
can ot
firms
to
exampl
al owing
and
base
pref nc s
ac es
for
servic
nowad ys
but
thes
have
Facebo k
conte
capit l z on
their
can
onli e
are
inve tor s,
makes
authors
nice
!bat
aspect .
share
!bat
or
eht
Pinter s
many
their
and
market
rof
in
rof
post
but ons
ident fy g
inter s d
poten ial
sre
with
life
ot
nO
the
the
or
daily
regist red
thes
ly be
is a s u m e d
ble
value
tions,
of
with
Facebo k
our
I laud
res a ch
who
market.
The
and
in m y
I wish
the
b e a huge
Prof.
bo k
are
the
sah
opin
the
author
al
readthe
best
!s ec u
.rD
Oliver
Hinz
Acknowledg ments
This
dis erta ion
was
t o take
this
op rtunity
a t eht
Chair
o f Information
I would
particular y
giv ng
me
the
chance
a n inter sting
field
to be among
the
first
Hinz
ful y
de icates
joyed
working
in
ag in.
It was
also
thank
Prof.
tional y
Electronic
Services,
of
One
article
mano
(University
very
schaft).
I am
notable
res arche s.
I would
also
ported
each
other
sup ort.
The
advice
valu ble
to me.
I
by
def ats
snfIer d
have
met
talented
program ing
task .
Final y,
I want
ac ompanyi g
me
ble
without
the
encouragem nts
me
over
t h e years.
Darmstadt
ac epted
by
to
TU
thank
Darmstadt
those
I
Systems
like
conduct
to
of
to
he
time
to
like
to
the
writ en
thank
my
those
Jan
Munter-
We
our
my
such
have
sup-
especial y
me
I am
dat
of
would
with
interp sonal
was
comf rting
SC.
terms
who
and
conju ction
with
for
esp cial y
people
Dr.
Darmstad .
Karls uher
and
dis erta ion
This
of
in
TU
in
I ad iSystems
col qniums
me
friends
journey.
work
Franz
sup orted
dna
years.
(Frauoh fer-Ges l -
scient fic
club
Prof.
I en-
ag in
the
dis erta ion.
Prof.
but
Markos
fort nate
ins ghts
RoBnagel
at
for
such
day
Information
ko wledg ,
the
I was
over
alongside
Heiko
Hinz,
Every
my
col eagues
who
this
grow
of
to
fo tbal
time
in
201 ,
responsib lit es,
new
team
(Chair
during
Oliver
ne d .
co-supervis ng
Dr.
thank
to
on
in
gain g
the
technical
favorite
Chair
when ver
se
my
I rec ived
Dr.
numerous
op rtunity
with
like
my
condit ons
the
Ben1ia
for
was
particularly
Prof.
his
and
I would
during
outs aoding
founde
atmospher ,
to
thank
only
my
stodents
supervisor,
under
Despite
stodents
his
2014.
me
Markets.
my
When
sucb
a pleas nt
a great
pleasure
Dr.
Alexander
TU
Darmstad )
this
dis erta ion
o f GOt inge )
grateful
for
not
Electronic
thank
stodents.
November
sup orted
who
res arch
stody.
PhD
his
in
people
wonderful
not
after
also
the
grateful
to
col ection
and
family
have
be n
unco dit n a1ly
for
pos i-
sup orted
Michael
Nofer
Table
of
Forew d
stnem gd lwo kcA
Conte s
.
V
.
o f serugiF
o f selbaT
o f snoita verb A
tsiL
tsiL
tsiL
1
.
IX
.
XV
.
noitcud r I
1.
2.1
3.1
2
IV
....
sy pon S
hcraes R
1.2 1
2. 1
3.2 1
4.2 1
erutc S
tekraM
Abstrac
1.2
2.
3.2
4.2
I VX
1
stxe noC
.
tekraM
3
ycnei f E
.
o f sdworC
Wisdom
do M
ycavirP
3
.
si ylanA
4
.
dna
o f eht
.
noitcud r I
suoiverP
1.2
2.
lacir pmE
1.3 2
2.3 sevitp rc D
3. 2
nois uc D
1.4 2
2.4
1
.
5
ytiruceS
.
.
Dis erta on
seilamonA
no
5
6
de iS-owT
noitcuA
Platforms
•.•
1
1
.
I
hcraes R
.
stekraM
de iS-owT
tneic f E
31
.
stekraM
ydutS
dna
tekraM
seilamonA
31
41
61
71
71
.
..
Platform
noitp rcseD
.
.
si ylanA
.
.
snoita m L
noisulc C
dna
.
erut F
hcraes R
.
02
32
52
52
X
elbaT
3
Are
Crowds
- The
Abstrac
1.3
3.2
3.
3.4
4
Case
Using
the
Mo d
3.4
4.
5.4
6.4
7.4
of
Introduc i
Previous
1.2 3
2. 3
puteS
1.3
2.3
stlu eR
1.4 3
3.6
Abstrac
1.4
2.4
the
Inter
Wiser
a Stock
than
Predict on
Conte s
Experts?
Com unity
27
..
..
2.4 3
Discu on
1.5 3
2.5 3
Ap endix
5.3
on
of
72
..
72
Res arch
.
Domain
Theor tical
30
Background
.
30
dnuorgkcaB
o f Empir cal
Dat
Col ecti n
Dat
Analysi
o f Empircal
Comparis n
Analyst
Divers ty
..
snoitac lpmI
Sun ary
..
.
Study
30
.
63
.
63
.
24
Study
o f Forecast
the C r o w d
Indep c
and
and
.
4
Ac ura y
betw n
Profes i nal
.
4
84
15
15
25
5
.
.
and
Twit er
to
.ko ltUO
.
Predict
Ef ect?
the
Stock
Market:
Wher
si
....
36
..
Introduc i
Previous
1.2 4
2. 4
3.2 4
Empircal
1.3 4
stlu eR
1.4
2.4
3.4
Trading
Conclusi
Ap endix
36
..
Res arch
laroiv heB
ecn ulf I
evitc d rP
Fina ce
o f Mo d
eulaV
.
Study
Dat
36
56
.
.
of
Col ecti n
..
evitp rcs D
pihsno taleR
pihsno taleR
and
56
on
and
scit a S
erahS
laicoS
the
Stock
Method
..
..
76
.
.
.
laicoS
Fol wer-W ight d
Market
.
.
Media
betw n
betw n
Stra egy
snruteR
.
Mo d
and
eht
kcotS
laicoS
Market
.
96
17
17
67
67
7
Mo d
7
80
28
58
elbaT
5
of
The
and
Abstrac
1.5
5.2
5.3
5.4
5.
5.6
7.5
6
XI
Conte s
Security
Impact
Breach s
of
Privacy
- A
Violat ns
Labor t y
Experim nt
.
98
..........
Introduction
Relat d
Theor tical
1.3 5
5.3 2
5.3
Res arch
Labor t y
1.5
5. 2
5. 3
Discu ion
1.6 5
5.6 2
xidnep A
Litera u
Econ mi
89
.........
........
Work
89
Background
Privacy
Security
Trust
19
29
29
39
59
69
9
9
......
........
........
.........
Model
........
Experim nt
Method
Result
Robustne
.......
........
........
Check
.........
yram uS
Lim ta ons
201
401
501
501
601
801
......
..........
and
Fut re
Res arch
....
......
..
901
List
of
Figures
Figure
:1-2
ycneuq rF
o f snoitcasn rT
per
Figure
:2-
ycneuq rF
o f snoitca rT
per
Figure
:1-3
tohsne rcS
o f na
Analyst'
Figure
:2-3
tnempol v D
o f dra n tS
Figure
:3-
tnempol v D
o f eht
o f redn G
Figure
:4-3
tnempol v D
Figure
:1-4
lacitero hT
Figure
:2-4
IMS
Figure
:3-4
L&P
rebm voN
Figure
:1-5
lautpecnoC
yadke W
htnoM
.
..
o f egA
egA
revo
ytisrev D
revo
emiT
revo
emiT
73
emiT
..
5
...
65
...
65
96
revo
seig tar S
3102
....
91
02
...
egar vA
o f gnidarT
krowema F
o f eht
noita veD
seulaV
,03
yaD
dna
....
IMSW
Chart
htnoM
noitad em oc R
krowema F
dna
,raeY
emiT
...
ne wt b
58
enuJ
,1 3102
dna
68
....
79
List
of
Tables
Table
1- :
Dis erta ion
Table
2-1:
The
Ef ect
of
Season i ty
on
Number
Table
2- :
The
Ef ect
of
Season i ty
on
Stand r ize
Table
3-1:
Operation zation
Table
3-2:
Result
from
Table
3- :
Result
of
Table
3-4:
Comparison
Table
3-5:
Result
from
Regr s ion
Analysi
...
50
Table
3-6:
Result
from
Regr s ion
Analysi
...
50
Table
3-7:
Result
from
Regr s ion
Analysi
...
75
Table
3-8:
Result
from
Regr s ion
Analysi
...
75
Table
3-9:
Result
from
Regr s ion
Analysi
...
85
Table
3-10:
Result
from
Regr s ion
Analysi
...
85
Table
3-1 :
Result
from
Regr s ion
Analysi
...
59
Table
3-12:
Result
from
Regr s ion
Analysi
...
59
Table
3-1 :
Result
from
Regr s ion
Analysi
...
60
Table
3-14:
Result
from
Regr s ion
Analysi
...
60
Table
3-15:
Result
from
Regr s ion
Analysi
...
16
Table
3-16:
Result
from
Regr s ion
Analysi
...
16
Table
4-1:
Depr s iv
Table
4-2:
Influe c
ofSMI
Table
4-3:
Influe c
210 /21(
ofWSMI
- )3102 50
Articles
.....
9
Sum ary
Probit
4- :
Result
ofVARmodel
Table
4-5:
Trading
Stra egy
Sales
Prices
.
2
..
24
83
Regr s ion
....
Score
Agil ty
4
Matching
...
.45
....
Mo d
Table
Daily
....
Propensity
of
of
47
Sta es
Deriv d
on
the
on
........
Stock
eht
WAST
Market
Stock
..
1 021 0(
37
- 03/2 1 )
.
7
Market
87
(Dec mb r
.....
by
, 1 201
- May
,13
2013)
.
97
18
tsiL
XVI
Table
:6-4
Table
:7-4
Table
Table
:8-4
:9-4
Influe c
ofSMI
Influe c
of
1 02/ (
- )2102130
ofWSMI
(12/ 0
- )3102/SO
of
- )3102/SO
(12/ 0
Table
Table
:1-5
:2-5
SMI
Influe c
Influe c
Simulation
Labor t y
Impact
Investm n .
on
the
withou
Market
(12/ 0
Anger
on
-05/2013)
Share
86
.
Returns
..
78
.
withou
Anger
on
Share
Returns
.
SMI
withou
Security
Anger
on
Share
Returns
8
dna
Security
in
Breach
the
......
and
.......
78
.
o f Privacy
Experim nt
of
Stock
o f selbaT
Privacy
59
Incide ts
on
Trust
and
301
List
AlC
AIM
API
AST
BIC
BAN
ETF
FWB
IS N
POMS
SAD
SMI
UGC
VIF
WoC
WSMI
of
Ab reviat ons
Akaike
Af ect
Ap lication
Aktuel
Bayesian
European
Exchange-trade
Frankfurte
Inter ational
Profile
Season l
Social
User
Vari nce
Wisdom
Weighted
infon ation
criterion
infusion
model
prog am ing
interface
Stim ungs kal
information
article
crite on
number
fund
Wertpa ierb i se
Securit es
of
Identif cation
mo d
sta e
af ective
disorde
Mo d
Index
gen rated
conte
inflation
factor
of
Crowds
Social
Mo d
Index
Number
1
noitcud r nI
1.
Synops
laicoS
Media
large y
sknaht
the
using
5 .2
snoitac lp
.tenr I
Media
laicoS
Media
sdlrow
of
and
Haenl i
of
beW
,0.2
and
ohw
al ow
blogs
dna
noita res d
enif d
builds
laicoS
eht
the
ot
of
,)ebuT oY
sihT
on
grow
,)+elgo G
,rkcilF
.)tfarc W
build
tha
ot
selpmax
Facebo k,
,)01 2(
tha
era
promine t
,.g e(
of
ot
ediw lro
si e x p c t d
most
,.g e(
World
ac es
peo l
number
ehT
com unit es
ap lic t ons
Media
ideol g ca
sa
and
creation
and
a
-onbcet
egnabcx
o f User
Conte ."
laicoS
Media
sniat oc
a vast
produce
by
of
gniu m
sti
routinely
i n informat
,.e i(
sedivorp
erutan
laicoS
Media
laicoS
In
Media
For
sa
particul ,
tha
peo l
.la
be
sela
.)90 2
using
,ylra im S
Acording
ot
be
a long-sta di
used
etab d
the
ef ic nt market
informat
detsac rof
at
on
stnev
and
beginu
and
busine
sevl m ht
elgo G
of
dev lop-
an
search
outhreak
res a ch
evah
and
Chang
20 9),
Skiera
,)30 2
and
.)60 2
user-g n at d
stock
conte
.snrute
This
it
can
si
on
invest ga o
prices
erahs
hypot esi ,
dne
by
and
n apS(
whet r
and
siht
to
about
using
(Dhar
Mayzlin
predict
ref
seiuapmoc
behavior,
whet r
to
to
creat d
rof
the
egnabcxE
seniurax
can
scimeda A
directly
MySpace
and
the
the
htiw
dat "
source
consumer
using
Stock
dis erta on
ot
ot
sela
of
an lyzed
.tenr I
erut f
be
terms
.)01 2
and
valu b e
can
(Chevali r
smroftalp
esnop r
tsacerof
Hol yw d
siht
Media
post
sa
' big
dat
resp ct
Amazon
the
inter s g
detal r-u f
With
the
on
wel
Frank
esarhp
,elba i v
ot
scimed p
cisum
using
and
eht
si
much
sa
public y
used
rof
detsac rof
sela
Bifet
whic
purs ing
,erom ht uF
ret i v d
,.g e(
an
ind v als
,tne oc
slanoi ef rp
than
.)210
be
pres nt
,tsav
ulf
hcraes
et
employ
can
,ecnats i
Tel is
can
dna
techuiq s
increas
user-g n at d
rathe
dna
av il b e
dat
of
ind v uals
(Tirun l a
public y
ot
amount
private
inters
si
euD
.ton
Kapl n
siht
skrowten
,efiL
Inter -bas d
snoitad u f
prima ly
laicrem o
tne oc
help
industry profes i nal
expon tial
cial
social
conte
Second
work
and
,sraey
easi r
bil on
.)3102
,)ret iwT
,.g e(
2014,
rec nt
,retsaf
79.1
(eMark t
in
peo l
,trope
include
,.g e(
of
eMark t
7102
seti b w
eht
an
ni
year
microbl gs
virtual
upon
"group
lacigo
Gen rat d
,)sre u
.srehc a
to
.stnem
,seir uq
grebsniG(
suc e f l y
eivom
bo k
the
ecnatropmi
al owing
snoitac lp
by
increas g
conetis
ot
laicoS
gained
broad n
Acording
bil on
and
evah
ot
-oS
si
be
elbis opm
M. Nofer, The Value of Social Media for Predicting Stock Returns,
DOI 10.1007/978-3-658-09508-6_1, © Springer Fachmedien Wiesbaden 2015
detsac rof
a
or
to
predict
1 noitcudortnI
2
share
1970).
market
stance,
day
condit on
prices
since
In
last
for
Overal ,
to
market
whet r
are
of
the
the
of
of
tha
we k.
share
For
retu ns
are
markets
four
Social
markets
hig er
ther fo
are
published
Media
ef ic nt.
are
transformed
to
is
(Fam
exposed
fina c l markets.
stock
Inef ic nt
comprise
informati n
in
rep atedly
in-
on
Mona pre-
prices.
on
is
factored
have
ef ic n y
found
of
days
im ed at ly
articles.
platforms
The
and
Res arch
the
nec s ary
the
market
judgin
the
ques-
ef ic n y
Understandi g
for
article
with
on
Inter .
first
deals
in
how
of
ef ic n y
Inter
the
afore-
markets.
second
and
with
third
to
rega d
ap lic t ons,
in
the
for
articles
dev lopm nts.
of
third
the
and
article
deals
predict ve
fe lings
from
to
sa
importan
Media
platforms
and
in
flow
While
article
on
networks
invest
turn,
in
serv
stock
sa
the
from
aims
sa
' Wis-
com unity,
col e t d
user ,
such
forecast
to
an lysi
Inter
microbl gs
or
to
ref s
Mo d
among
and
Tel is
predict on
for
investor '
risk
ap eti
e t . l a 201 ;
Kar bul t
o f privacy
fourth
article
invest ga es
the
influe c
consumer
o r investor
behavior.
Privacy
dua
security
precondit ons
for Inter
users'
wil ng es
platforms
(Fogel
dua
Nehmad
20 9).
Inter
b y dat
theft
o r hacker
at cks,
can
b e shaken
over
the
last
years
(Kel y
2013;
Silve ra
201 ).
practi oners
ne d
ong i
ac es
to c o n t e
from
orde
to u s e
this
informati n
for
the
predict on
of
of
nec s ary
t o stody
people's
behavior
in c a s e
whic
pos ibly
threat n
people's
wil ng es
t o com unicate
o f info ation.
the
first
thre
articles
use
dat
from
Social
Media
o f a labor t y
exp riment
describ
the
outc mes
a univers ty
campus.
The
remainder
o f this
introduc i n
can,
on
reto ns.
Previous
201 )
20 4)
a stock
sta e
mo d
stock
and
social
in
whet r
for
from
prac-
201 ).
Frank
specif al y
dat
Social
and
e t .la
and
Tirun l a
and
article
use
with
value
emoti ns,
and
extrac ed
dat
second
phenomeno
have
(e.g ,
from
res a chers
Bol en
(Antweil r
The
Crowds"
for
revi ws
boards
user-g nerat d
directly
(e.g ,
consumer
of
col ect d
forecasting
mes ag
value
was
pos ib l t es
price
used
predict ve
Dat
inter s ing
share
stock
the
reto ns.
of er
exampl ,
market
examin
stock
whic
of
are
on
The
fourth
ducte
is
res a chers
(1985)
Fina ce
The
dents
fied
Social
Media
increasingly
ers
forms
ther fo
inc de ts,
the
info nation
howev r,
contradict
West rfield
to o t h e r
the predict on
the
dis erta ion
anom lies
el ctroni
proces
of
deca s,
tha
and
compared
of
piec
few
anom lies
Jaf e
ref s
tion
the
user
mentio ed
conte
Media
ti oners
studie ,
discu ons
stock
dom
while
user
people's
be
This
wil ngues
ev ry
the
Twit er
to
such
Facebo k
det n i e
info nation
can
or
proxy
Twit er.
thus
dua
(Bol en
their
201 ).
and
security
inc -
have
be n
to
share
ident conte
user '
on
trus
in
whic
Social
oc ur ed
Howev r,
res a chSocial
Media
share
platIt
reto ns.
privacy
and
is
security
and
thus
ap lic t ons,
the
tha
was
is d e v o t d
conto
2.1
hcraes R
stxetnoC
first
sum ariz ng
3
elucidat ng
the
The
four
I)
Michael
/ Hinz,
Michael
contex s,
Market
and
Anomalies
Confer c
/ Hinz,
The
-
on
Oliver
Case
Econ mi s,
Nofer,
Market:
ne ri g,
/ Hinz,
is the
Wher
forthc ming.
Nofer,
The
borat y
coming.
Michael
(3),
then
on
Information
Two-Side
System
(ECIS),
Oliver
(2014).
Oliver
Using
Inter
Wiser
Joumal
Twit er
to
&
Busine
Violat ns
&
Predict
of
the
Information
/ ,leganBoR
and
Heiko
Security
Informati n
Stock
Engi-
System
Jan
Privacy
Busine
Res arch
the
Com unity.
/ Munterma ,
of
Experim nt.
on
Predict on
Ef ect?
Impact
Crowds
Are
30 - 8.
Mo d
/ Hinz,
Econ mic
(2014).
a Stock
of
84
Michael
(2014).
- A
Breach s
System
La-
Engi e r ng,
forth-
Contex s
1.2
Market
ycnei f E
to
the
a way
(Farn
ef ic nt
tha
market
ev ry
have
of
price
Fields
(e.g ,
such
sa
the
the
share
price
sa
factored
said
into
information
to
fina c l
markets
whet r
such
anom lies
be
aver g s
empir cal y
in
in
observ d
include
Jaf e
ef ct"
I f markets
cur ent
achiev
ext nds
i t is importan
are
ef ic nt,
share
prices,
superio
retu ns.
investors'
also
exist
market
to
long
achiev
run.
Howev-
markets.
Prom-
anom lies,
such
West rfield
1985),
and
market
in
in
fina c l
(Jegad sh
of
ef ic ent
unable
the
calend r
and
as umption
are
refl cted
should
market
;139
forecasting,
a predicto .
markets
im ed at ly
investor
mean
distor n
"mo entum
contradic
the
be n
thes
ef ct"
For
is
then,
to
anom lies
fina c l
informati n
Theor tical y,
examples
used
of
compared
sev ral
hypot esi ,
piec
.)0791
retu ns
in
res a ch
.niapS
Busine
be
mediat ly
use
lies
unclear
(201 ).
European
Experts?
2.1
Oliver
Platfon s.
than
4)
and
:era
Nofer,
3)
background
article.
articles
Nofer,
Auction
Barcelona,
2)
Ac ording
such
prices
superio
,re
ine t
"we k nd
anom lies,
anom lies
theor tical
each
and
Titman
as
the
techni al
19 3).
Thes
ef ic n y.
to
know
whet r
then
whic
should
entails
The
tha
powers,
Inter
platforms.
be
can ot
investor
of
pres nce
predict ve
on
informati n
informati n
market
anom but
This
i t remains
dis erta-
can
im-
1 noitcudortnI
4
tion
betw n
thus
smia
2. 1
The
and
Surowieck
bet r
world
whic
20 4).
Inter
Wik pedia's
the
201 ).
the
solving
close
this
and
in
an
on
eht
litera u
onli e
by
auction
invest ga in
trans c io
platform.
a/Crowds
Wisdom
of
(W
Crowds
practi oners
since
decis on
s i the
ref s
The
can
oc)
the
(20 4).
eht
be
used
points
ot
entri s
quality
I t s i hardly
an
tha
(Surowieck
are
ot
crowd
exampl ,
for
the
tha
ot
et .la
Raj gop l n
"crowds urcing"
with
as i t
on
compar ble
20 5;
companies
ind v uals
oftline
a Mil ona re",
ac ur y
(Giles
then,
de icat d
(Howe
pos e
make
the
case
show,
ency lop dias
surp i ng,
al owing
Be
al
a large
Studies
and
print
to
of
when
task :
hig
in
can
in
Wants
p=ent
insta ce
complex
of
are
' Who
senraJ
crowd
exampl
09
toward
solve
entri s
Inter ,
of
over
author
a large
promine t
life n
res a ch
scien
tha
One
in
from
po nlar
conte ds
answer
also
at en io
by
ind v uals.
cor e t
litera o
increas g
int oduce
basic l y
a wef
Audienc "
eht
ot
gained
was
idea
than
"Ask
has
term
This
exp rt-w i en
task
task
like
to
problem
20 8).
the
of
are
Inter .
fina ce,
The
with
ther
divers
of
share
boards
tha
ot
an lyze
stock
predict on
dat ,
includ g
stock
the
wise
crowds
crowds
have
be n
divers ty
and
set ing.
sa
the
case
knowledg ,
influe c d
other
group
divers ty,
how
ev ry
me b rs.
and
the
crowd
crowd
with
in
resp ct
or
gend r,
of
pre-
changi
divers
,ega
crowd,
eht
condit s
of
are
importance
Inter
by
stody
perfo mance
cnlture,
eht
much
four
clarify
me b rs
the
of
to
world,
wil
eht
In
col e t
motiva n,
crowd
hig l ted
si
oftline
influe c
also
goal
the
-la
techniqu s.
ot
dis erta on
tha
such
has
being
This
Media
predict v
platforms,
ratings.
knowledg ,
201 ).
means
char te is c
In
In
studie
min g
l es
Ant-
earli
pos e
tex
major
Inter .
indep c
Divers ty
decis on .
buy
ident f :
e t .la
res a ch
on
the
Thes
retu ns
from
(e.g ,
res a ch
another
on
(Sim ons
withou
based
an lysi ,
dat
Fina ce
of
help
of er
and
dis erta on
used
Social
the
This
compares
The
!o haY
sophi t ca ed
many
.skcot
and
osla
20 1).
stock
now
perfo manc
rof
of
of
with
com unit es
picks
or
about
exist nc
discu
.tekram
Whitelaw
mes ag
sa
com unity
aver g
Bul
discu on
the
ot
one
on
platforms,
who
Ragin
and
eht
befor
had
thes
res a ch ,
sa
Tumarkin
onli e
the
earli
such
20 4;
demonstra ed
and
of
com unit es
on
me t
of
find gs
discu on
Howev r,
ap rent
recom ndati s
an lyst
the
Frank
predict on
is
profes i ual
ext nds
and
so-cal edstock
Crowds
backgrounds conti ual y
eht
those
exist
of
Wisdom
from
util zes
Besid
condit s
for
wise
indep c
degr s
a n Inter
t o various
Previous
pend t
decis on
gap
sel r
modsiW
In
the
ind v uals
dis erta on
them
ther fo
fina c l
weil r
ready
value.
authors
contras ,
more
ot
buyers
education.
me b rs'
partic pant
indecan
make
a
2.1
hcraes R
stxetnoC
3.2 1
5
Mo d
The
litera o
ple's
,suhT
tors
peo l
Analysi
relationsh p
betwe n
(e.g ,
risk
ap eti
are
more
A
in
bad
of
in
mo d
the
role
of
winter
sle ping
fmding
in turn
the
Yagil
Today,
20
Social
real-tim .
air
market
One
wil ng es
search
ouline
20 ;
net
cy
formati n.
at cks
Twit er
Privacy
also
miK
deci
to
ref s
sta e
Kamstr
et
measure
mo d
extrac ed
al.
Twit er
ext nd
or
previous
struc o e,
Experim ntal
find-
whic
might
play
res arch
(Bon
and
bave
the
dat
on
20 6;
Sy
emoti nal
(Guil ory
Inter t
public
has
mes
observ d
on
sta e
from
wil
conet
from
the
Twit er,
might
Intem t
.la
this
af ect
be avior.
people's
of
thef s.
be n
their
For
hacked,
the
Kel y
to
Nehmad
usage
of
(Gef n
when
their
tha
notes
InterPriva-
persoual
such
repo ts
(201 )
trust
20 9).
inc de ts
(2013)
Silve ra
influenc
role
dna
and
re-
tha
importan
(Fogel
include
is the
Previous
factors
an
col ection
while
ap lic t ons
others.
main
relationsh p
close
play
also
meanwhile,
insta ce,
Media
with
two
networks
over
Social
as
social
contr l
breach s,
bave
security
con erns
for
of
opin s
their
dna
Privacy
up
sign
dat
sbare
because
20 8).
to
other
ac ounts
mo d
20 7;
to
mo d
existenc
to
privacy
Security
or
the
user
esp cial y
user
.la
res a chers
tha
for
Inter t
e t .la
Sport
investors'
et
al ow
ind cates
retu ns.
coudit on
bas
behavior,
lower
invest aut mn
ytiruceS
fundame t l
of
identif ed
days
(20 3)
during
disor e .
influenc
are
the
to
cloudy
.la
af ectiv
(Edmans
ap licat ons
and
et
are
Facebo k
can
predict
stock
This
dis erta ion
ings
i n the
litera u e
b y consider ng
the
com unity
an
importan
role
when
measuring
mo d
sta e .
shown
that
mo d
sta e
can
spread
among
ind v uals
e t .la 20 5).
Relat d y,
a number
o f rec nt
studie
tagion
oc urting
via
tex -based
com unicat on
201 ;
Hanco k
e t .la 2 0 8 ;
Kramer
201 ).
Using
dis erta ion
wil
test
whet r
emoti ual
contagion
people's
ofl ine
be avior,
particular y
their
investm n
4.2 1
on
than
retu ns
can
stock
ref s
.)1
Media
res arch
Init al
days
sea ou l
pol ution
if
that
ef ct
Kamstra
from
is l o w e r
anom lies
weather
sun y
sbare
suf er
af ect
and
tha
ind v uals
dna
market
on
19 3).
inves-
ap eti
prices.
the
hig er
1983).
since
risk
uncover d
Saunders
the
peo-
Tversky
esa rcni
the
erahs
insta ce,
are
20 3;
depr s ion,
many
ba its,
when
and
Levy
retu ns
to
band,
bave
For
Shumway
other
in
bias,
and
dnet
decr asing
fina ce
established
(Johns
prices
the
to
sta e .
erahs
tha
dna
On
is wel
misat ribut on
the
sta e
erahs
leads
studies
people's
to
mo d
risk .
whic
retoms
stock
mo d,
take
mo d,
number
their
a go d
to
and
Ac ording
20 3).
on
in
wil ng
by
al.
dep n a
era
are
sta e
et
i f investor
driven
phenomeno
(Hirshleif r
gated
and
ev nts,
sta e
20 ;
in
mo d
Kamstra
over
a sim lar
inas
hacker
250,
breach
1 noitcudortnI
6
for
5.6
formati n
search
ne d
a
i s ther fo
rity
breach s.
This
peo l
mainly
and
security
litera o
method l gy
dis erta ion
tion
to
3.1
Overal ,
res a ch
the
articles
Chapter
have
ter
gen rated
dat
(article
and
marized.
mil on
ac ounts.
be
or
investor
from
aim
who
ac es
to
on
util ze
capit l
Acquist
et
on
focus
and
security
Structure
consi t
The
res a ch
io
chapter
2)
and
conte
two
)4
Social
.•
In
Market
article
the
the
platform
pa er
the
buyers,
market
20 5
sa
and
consumer
nO
20 9.
el ctroui s,
is
on
buyers
the
sum-
E-Com erc
io
observ d
to
can
be
the
Inter .
gen ral y
empir cal
clas i-
an lysi ,
and
the
on
sel r
platform,
jew l ery,
briefly
an
res a ch
and
betw en
5
chapter
consumer
on
frequ ntly
of
the
usiog
Platforms
are
hypot esi
user-
i s done
article
stream
For
of
value
anom lies
fundame t l.
tha
Chap-
this
Auction
this
sta u .
anom lies
Afterwa ds,
inc de ts
tha
four
platforms.
security
res a ch
market
of
the
publicat on
predict ve
prices;
share
one
of
Inter
Two-Side
of
published
cover
market
the
distor n
trans c io
such
this
reac-
four
and
on
each
ext nds
or
dat se
and
price
ef ic ent
betw en
to
contras ,
over i w
ap lic t ons.
persi tenc
techni al,
78,06
the
study
consumer
coutaio
whet r
of
fol wing,
are
The
col e t d
In
direct
an
iovest ga
on
anom lies
contradic
in
ev nt
each
oc ur
predict on
Media
o f privacy
Anomalies
iovest ga es
Market
markets.
calend r,
3)
the
iufl enc
behavio
1:
privacy
made
20 4).
provides
also
to
the
iovestm n
.la
dis erta ion
type,
with
deals
4 (article
wel -know
this
I-I
study
markets
resp ct
examin s
of
be n
whic
of
goal,
fina c l
with
from
chapters
Table
I, w h i c
article
observ d
have
iufl enc
the
evif
articles.
the
2 discu e
how
res a chers
using
et
chapters
res a ch
be n
3 (article
sa
of
remaiu ng
published
Cavusoglu
secu-
iovest ga
,raf
breach s
ef ects-namely,
It
or
Dis erta ion
dis erta iou
articles:
iocludi g
20 6;
privacy
to
such
user .
breach s.
o f the
the
.la
first-o de
to
the
have
to
reto ns
Inter
rega d
at emp s
iore-
time,
stock
by
as e ing
reactions
same
predict ng
So
when
personal
the
exp riment
various
market's
(e.g ,
for
is gen rated
with
breach s.
ef cts
iostance,
at
Media
whic
a labor t y
security
and
For
wil
Social
behavior
second- r e
show
privacy
use
how
platforms;
dat ,
privacy
breach s.
to
to
consumer
wil
fol wing
hig l ht
Media
public
study
dis erta ion
react
focused
examples
Social
to
nec s ary
four
Thes
stolen
conti u s
Article
This
platform.
fina c l
Anomalies
fied
thors
auction
products
LinkedI
might
sel r
or
a German
of er
cosmeti .
various
au-
3.1
o f eht
erutc rtS
One
na ci l
persi ten
prices
the
Novemb r
rather
platforms
central
of
tha
turn-of the
lower
the
featur
both
month
on
we k nd.
to
in
the
on
2:
of
of
be n
perfo m
predict on
com unity
of
the
WoC
on
stock
predict on
stock
and
This
as ign
struc e
makes
predict ons.
The
and
crowd
virtual y
dif er
The
anyo e
with
resp ct
aims
study
with
to
those
wel
as
of
the
profes i nal
DAX
.xedni
from
tha
the
crowd
crowd
me b rs
per
points
res arch
year
companies.
outperfo med
the
platform
the
provider
authors
tha
to
the
daily
decis on
study
return
are
made
of
influe c
understandi g
of
an
Inter
crowd.
-if
is a
insta ce,
activ es
3.7
to
perc nt
in
lower
sign f ca t,
the
study
but
show
tha
auction
inef c ies.
Internet
Wiser
than
Prediction
Experts?
-
Com unity
phenomeno
than
For
their
sta i cal y
of
market
to
ther
shift
are
are
result
the
sim lar
platform,
ef ct.
to
ef cts
on the
a Stock
bet r
in
observ d
largest
of
the
buyers
product
same
the
describ
can
the
margin l
Crowds
peo le
has
,mus
are
On
day-of the-w k
Thes
Crowds
Case
platforms
spur
contai
Are
The
auction
as
for
In
Inter
tha
anom lies.
wel
Janu ry.
magnitude.
Wisdom
is
market
prices
compared
smal
study
sa
whic might
Sundays,
Furthe mo ,
The
lowed
show
when
sign f ca t
a bet r
mance
find g
in
are
teurs
sa
mendatio s
show
perc ntage
and
sional
7
markets
Article
The
groups
WoC
stock
standi g
the
cus
tions.
of
tions
bers
noita res iD
wher
a few
task
this
large
The
ind v als.
in
many
dif er nt
the
sa a n e x a m p l ,
pa er
the
Inter .
Dat
was
col ect d
com unit es
i n Europe.
Memb rs
buy
and
sel
ratings
ac ording
it pos ible
to a g r e t
and
i s divers
t o the
ext n
tha
o n the
platform.
can
regist r
to k n o w l e d g ,
cult re,
age,
gend r,
compare
the
stock predict ons
an lyst
(who
are
also
Overal ,
the
authors
col ect d
May
20 7
and
betw en
are
able
t o achiev
a retu n
hig er
compared
to t h e p r o f e s i n a l
Furthe mo ,
both
the
crowd
DAX
index
during
the
four-yea
has
made
sev ral
changes
o f divers ty
and
the
ef cts
o f the
crowd's
share
recom ndatio s
indep nt
from
each
other.
divers ty
o n daily
retu ns.
I n sum,
of how
divers ty
and
indep nce
and
divers
power
omine
of
world.
contribu es
a
a bet r
from
to
their
ther
underone
Swewise,
o f Sharewis
market
a large
no a c e s
an lyze
are
exp ctanumber
lim tacrowd
of
pres nt
of
dis-
Ther fo ,
and
the
Taking
me -
education.
thes
de icat d
am -
on
the
10, 46
platform)
share
Aug st
recom-
201 .
that
The
,si
on
result
aver g ,
0.59
an lyst
of
me b rs
banks
the
dna
profes-
period.
over
the
years,
indep c .
whic
od
Howev r,
al-
The
result
improve
ther
the
article
influe c
is
contribu es
the
perfo -
no
1 noitcudortnI
8
Article
3:
Using
Mo d
in
Various
studies
b y their
driven
labor t y
exp riments
weath r,
time
dna
mo d
sta e
Kamstr
e t .la
their
artif c al ty.
t o measure
mo d
alre dy
iod cated
might
bave
predict ve
er, i t s i u n c l e a r
the
qnality
o f the
lory
e t .la 2 0 1 ;
spread
due
to
.smrof
The
article
10
mil on
German
study
also
iotegra s
provid og
a first- me
stock
retu ns.
The
tive
value
for s t o c k
e s b y 3.
basi
that
ag re at
predict
stock
retu ns.
contagion
o n the
iognes
to i o v e s t
Article
Sev ral
of
breach s
suc e .
when
search s
Cavusoglu
orde
Twit er
Ef ect?
fina ce
and
sta e .
with
changeov r,
turn
the
20 ).
In
contrs ,
sta e
io
tha
Social
with
and
result .
to
what
rec nt
et
.la
articles
2014)
tex -based
this
of
tha
the
of
number
iovestiga on
result
rev al
retu ns.
In
poi ts
on
Twit er
next
mo d
can
dat
trust
res arch
predic-
iodex
iocreas-
si
no
evid nce
numbers)
can
tha
io
emoti nal
particular
the
has
of
20 4).
are
contras
aims
as
focused
and
to
after
thes
show
wil -
threat
ac eptance
security
to
capit l
compaoies'
market
reactions
Typical y,
reev nts
their
consumer
the
and
cor espondi g
and
direct
io
Security
Privacy
breach s.
the
studies
the
factor
security
prices
to
and
a key
20 9).
a serious
on
ther fo e
privacy
share
In
article
Violations
Nehmad
and
primarily
iofluenc
decr asiog
this
DAX
iod cation
Privacy
Experiment
protec ion
dna
Fogel
consumers'
the
of
Laboratory
identif ed
(e.g ,
e t .la
on
have
fol wer
first
behavior,
Impact
- A
have
observ
contagion
ther
withou
the
of i ne
Economic
platforms
studying
the
The
thus
stock .
The
Breaches
lower
Previous
(i.e ,
be
iofluen
mo d
contras ,
roughly
an lysi ,
sta e
mo d,
might
of
2013.
the
of
In
alone
result
Inter t
iofluenc
day.
can
plat-
and
iot
improving
Guil-
sample
1 02
mo d
trsdiog
sta e
Thes
io
of
case
the
a uniq e
fol wers
the
2014;
Inter
Media
Social
betwe n
iot
fol wer- ighted
tha
influenc
e t .la
the
on
with
Twit er
have
platforms
Howev-
by
col ect d
by
res a chers
201 )
might
emoti ns
res arch
wer
20 3;
lim ted
dev lopments.
facil t ed
stream
twe s
iofluenc
e t .la
other
and
contagion
Coviel
tha
com unicat on
ext nds
Bol en
(e.g ,
show
whic
Shumway
al ow
market
io
(e.g ,
are
Twit er
stock
emoti nal
ext n
Sev ral
Kramer
the
from
rega d
to
of
studie
ap lic t ons
(e.g ,
be
factors
and
Media
studies
few
can
shown
ext rnal
previous
extrac ed
markets
be n
has
al
thes
A
stock
Hirshleif r
valu b e,
sta e
value
tha
pol ution),
(e.g ,
si t h e
Where
relationsh p
dif er nt
air
real-tim .
mo d
whet r
shown
of
market
stil
today's
Market:
This
result ,
stock
While
Stock
a number
sport
io
the
have
mo d
and
authors
ef cts,
Predict
psychol gy
partic pants'
4:
Inter t
to
emphasi
(e.g ,
on
reaction
secondvia
first-
3.1
o f eht
erutc rtS
orde
univers ty
investm
,spuorg
formed
contr l
noita res iD
9
.stcef oT
siht
,dne
.supmac
product
of
of
a security
was
security
wil ng es
on
both
and
ot
the
trus
consider g
rof
Table
Con-
" s" a
DCE'i?iC
,.
5'
6
influe c
of
toge h r
privacy
-ni
one
privacy nd
refl ct
is v e r y
consumer '
importan
ind v als
ehT
ohw
rof
eulav
.nois ced
stody
want
ot
of
Crowds
build ng
security
more
ther fo
use
icm
;a
;a
1
Social
sah
impl -
Media
ap lic -
z: u
,,
(d~
'c~l
~ )lC
5' °
~H
1' "
! I:!l.0~
" E
~~1S.
.5
g
DC
1ii",~
gl)
>fIl~
"'0
5. E O
E
~
.-"0
§:g~
8. .l!: ~
0;1
~1.lz
,
§·E
il"ll.
~..c:
I
1!6a.
.s~CIO
: ;.!I
.
,,
DC-"
Du C
~"E
"E~
,,~
1
&
Security
0
'i?ol<. " 0
,,~
z: u
Privacy
Analysi
,,~
i~
W"
Mo d
-?i'~
,,
0"
!&
,ydots
' i ? - ,,~
"om
z: u
0
~
the
whic
investm
res a ch
WIsdom
,"
:::J~~
, " . D' C
1aotca
Market
~1:~
.0:
measur
Howev r,
and
:!l.~'-
0
was
also
.se opru
ef ic ency
lC
ther
Dis erta onA cl
tex
i1
siht
.ynapmoc
practi one s
trea m n
other
was
.knab
of
the
gnitsacerof
:1-
the
owt
the
an
informat .
ot
~·E:~8:5~
~m~
0)_
3!0 "o~_I::sa .1: e .
m
~
6
m~
01: 0
)lCP if
i ! l " - . :! ~
~-g= - eCl)~o~.:=
!l~:!l."O
L.
~~
E
m.e!
g>8C1)~:5~~
O"DC
)lC )lC
=8~o>>o"O
O)E·:=I: .c
.8
!:5
sa
O">g
E
EI!!5i~~i.g
Jr O C O J 0 : : J z . 5"
Di· ;.u Di· &
:I
'; ,"
O"
; . u J:
J
_.
m
I-
U1E
2
_
.
~~ : 1 E
I:
J:
,
;
,! .
~ 5.-:2
~'B
~:~
.- E "
" '"
"
:I
)lC
:5~=
i81-
.
a
into
wer
past;
ad it on l
elba
money
while
rec nt
amount,
result
in
eht
any
the
on
own
Ther
breach
in
investm
with
.knab
a privacy
bank
given
exp rim nt
their
a lanoitc f
the
er w
interac
Ac ording
rof
not
authors
trus
a labor t y
invest d
of
of
saw
the
ot
consumer '
when
cations
tions
by
info ed
breach
,yaw
condu te
stnedo
of er d
whic
whic
this
authors
exp rim nt,
being
eno
group,
In
the
eht
In
0)01:0
1: .-
jJ§CI)co
0·>
.2' Q
-
l i
0. 0
l! > a s ! 1 : ' s i " ' "
. 5>. 0'"
>
.
.4
.2
.3
.1
Using
Stock
do M
yrot
curity
ehT
ycavirP
ytinum oC
Are
Wiser
case
tekraM
de iS
tnemirepxE
cimon cE
Crowds
of
Breach s
Twit er
Mark t:
Ef ect?
than
noitcuA
snoitaloiV
seilamonA
a Stock
elcitrA
Experts?
no
to
Wher
sn oftalP
tcapmI
the
- A
Predict
Predict on
dna
no
Intem
Labor -
of
-eS
the
s i the
- The
-owT
detc l oC
ytisrev
Public
throug
dna
no
ytinum
of
erahS
snoitcasn rT
sreyub
na neG
kcots
Intem
lanois eforp
supmac
twe s
esiwerahS
snoitadnem ocer
dna
Twit er
noitc derp
noitcua
at d
user
Dat
srel s
ne wteb
col e t d
morf
API
(=crowd)
st ylan
mroftalp
no
-moc
-inu
a
gnimochtroF
noitamrofnI
gnimochtroF
noitamrofnI
.B
dehsilbuP
s enisuB
.210
ne ri g.
ne ri g.
detnes rP
ecner fnoC
84(3): 0 - .
System
Publicat on
BHV
BHV
BHV
scimon cE
,)S/CE(
:ni
:ta
smetsyS
smetsyS
Ranki g
:ni
:ni
folamu J
Ranki g
Ranki g
naeporuE
no noitamrofnI
BHV
s enisuB
s enisuB
Barcelon ,
Sta us
.B
-ignE
-ignE
Ranki g
.B
.B
&
&
4.
.3
.2
.1
Experts?
tnemi
- A
Crowds
cimon cE
Labor t y
ytiruceS
Stock
a Stock
fo ycavirP
dna
ehT
the
Wher
?te f
gnisU
ytinum oC
fo
Are
tem tnI
tekraM
de iS-owT
sn oftalP
ret iwT
Article
resiW
is t h e
seilamonA
snoitaloiV
:tekraM
- The
Predict on
noitcuA
on
sehcaerB
do M
the
tcapmI
Exper-
ot tciderP
naht
case
no
Ef-
~!
~~
.
i"
a.
I:
~
i'
g
~
g:i
, I n: " I .
·0
it
l!:n
~-
h
S-n
'i
[
;:
2
Market
Platforms'
Anomalies
on
Two-Side
Auction
Abstrac
A
market
anomaly
fina ci l
anomalies
shown
tends
merce
such
twe n
betwe n
ef ct
ad it onal
lower
higher
com on
which
(or
market
market
that
be
could
calend r,
empir cal y
and
res arch
empir cal y
an lyzes
ruO
markets.
empir cal
1.2
fmanci l
al ow
edart
year.
ev ry
auction
some
to
are
marginal
rather
market
smal
persi ten ly
sel r
result
our
period
realize
also
while
are
bethe
buyers
period
Overal
platforms
on
persi ten
covers
tum-of the-month
Prices
Christmas
ex-
complet d
and
would
pa er
are
a persi ten
discount).
a
Com-
trans ctions
platform
that
have
Electronic
anomalies
observ
price
of
in
78,06
auction
on
Such
be n
This
platforms
an lyzes
We
and
markets.
auction
calendar
a namreG
20 9.
and
April
typical y
hypothesi .
related
for
stody
on
sel rs
May
20 5
and
and
a day-of the-w ek ef ct
surpl se
3.0(
perc nt
i n the highly
competi ve
prices
a t the
begin g
notion
that
two-side
we howev r
can
observ
buyers
distor ion
fundamental
o f set ings
two-side
whet r
to
a price
ef ic ent-market
or
a number
of
stream
is
the
techni al
in
this
inef ic en y)
contradict
to
se m
ben fit
from
sup ort
markets
ef ic ent
the
on
inef ci n es.
Introduc i
One
o f the
o f price
is t h a t
t o the
char cterist c
over
the
past
as ociated
spective
trading
auction-type
zniH
20 8;
Thes
markets
but
Nofer,
[OmlS.
Micbael!
European
key
econ mic
discovery,
proces
i.e
of
few
with
the
this
Lower
ouline
for
cost ,
mechanism
e t .la 2 0 1 ) .
new
platforms
are
compared
have
price
Oliver
to
some
on
(201 ).
noit81m f I
Market
of
to
of
e t .la
stil
System
Anomalies
Due
proces ing
cost
with
tha
zniH
with
Two-Side
Barcelon ,
M. Nofer, The Value of Social Media for Predicting Stock Returns,
DOI 10.1007/978-3-658-09508-6_2, © Springer Fachmedien Wiesbaden 2015
Span
fina c l
stock
In
on
proemploy
dna
under- searched.
(EelS),
trading
changes
interact
platforms
20 4;
pro erties
markets
in
ac ept.
drastic
reduction
importan
fmanci l
enga e
side
undergone
pos ib lit es
a number
(Bapn
discovery
share
to
Confer c
the
new
led
a sel r
market
both
has
dna
partners
and
tha
proces
menu
dif er ntia on,
Hinz,
a buyer
a price
Inter t,
years.
price
when
find g
Auction
Spain.
fina ce,
Plat-
2 tekraM
21
the
lyef ic ent".
than
public y
be n
exchanges
market
contradic
techni al
ef ic nt-market
hypot esi
This
the
aver g
av il b e
market
shown
empir cal y
and
market
inef c y)
the
ef ic nt-market
or
fundame t l
Stock
com n
buyers
of
an
gain
for
their
we k)
on
this
exp cted
oc ur
sel r
The
introduce
ef ic n y.
of
78,06
distor n
exchanges
that
sel r .
dna
wher
ef ic nt
market
hig er
surpl e
exampl
on
Mondays,
products/ hare
while
demand
Mondays
and
distor i n
is
to
cor e t
tha
for
that
can ot
investor
a particular
o f the
investm n
a t the
time
tha
this
hypot esi
anom lies
oc ur
is a price
distor i n
hypot esi .
relat d.
and
auction
markets
mechanism
brings
This
could
b e double
prices
refl ct
the
relationsh p
neith r
buyers
nor
than
market
aver g s
sup ly
would
o n this
day
o f the
would
raise
a t the
thus
competi on
amongst
i f price
nul if ed.
Thus,
this
distor n
os t h a t
no de iS-owT
fina c l
noitcuA
markets
are
consi ten ly
given
tha
(Fam
not
does
hold
for
or
be
or
to
are
of er
time
it on
buyers
.eL(
buyers
the
sel
of
try
the
buy
to
price)
until
market
persi t
lower
to
would
raise
can ot
are
day
oc ur,
they
In
sy tema ic l y
another
would
distor ns
type
sup ly.
I f prices
not
wil ng
time.
sel r
and
any
and
able
a longer
in
namely
auctions
demaod
(i.e
same
have
side ,
English
would
we k
to
Com erc
market
betw n
for
se m
calend r,
be
could
two
drop
(or
tha
Electroni
sel r
anom ly
market
toge h r
auctions
i t has
stock
like
market
anom lies
in
is
markets
A
a fina c l
Such
retu ns
Howev r,
19 0).
on
hig er
informati n
the
1970).
Arie!
(e.g ,
"information l-
achiev
risk-cla ,
smroftalP
secrof
for
are
a longer
time
or
regula y.
I f price
and
interm diary
distor ns
persi t
buyers
to
for
to ls
The
of
in
oc ur
for
the
few
and
with
our
examin
dat .
At
be
found
the
core
a t this
the
We
of
platform.
in
section
we
We
discu
we
two-side
maroutline
the
outline
markets
litera u
section
we
befor
this
sale
dna
section
on
furthe
also
in
gap
introduce
model
dat
an lyze
fol wing
wil
wil
auction
busine s
We
res arch
2.
the
is
its
seri
price
In
on
two-side
time
market
of
platform.
:swol f
focus
emphasize
and
can
sa
could
the
sy tema ic
auction
Section
study whic
platform
an lyzed
whet r
and
increas
a long
anom lies.
market
hypot esi .
empir cal
the
an lyze
to
anom lies.
ther by
stra egy.
design
to
to
wel -informed
of eringlb d ing
market
his
ef ic nt-market
market
tha
studie
anom lies
the
pa er
the
on
res a ch
anom lies
our
res a ch
introduce
only
we ks
previous
reconsider
s i ther fo
with
resp ct
and
che k
o n a two-side
proce ds
and
wil
sold
of
remainder
outline
and
12
al ow
and
asym etri s
pa er
products
would
entry
to
platform
trans c io
of
our
this
market
have
information
at eou
aim
auction
time,
their
then
that
a two-side
a longer
optim ze
would
The
wil
kets
empir cal
are
market
close
describ ng
tives
market
a market
auction
as ert
means
seilamonA
tha
with
resp ct
to
whic
we
.3
We
report
resnlt
with
descripwhet r
an lyze
the
wil
star
emos
wil
ther
in
section