The value of social media for predicting stock returns preconditions, instruments and performance analysis

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