Nicolini, Ma cella; Saba ini, Fabio; Fan azzini, Dean
A icle
Online in e es in adical Islam and e o is a acks
Peace Economics, Peace Science and Public Policy (PEPS)
P o ided in Coope a ion wi h:
De G uy e B ill
Sugges ed Ci a ion: Nicolini, Ma cella; Saba ini, Fabio; Fan azzini, Dean (2025) : Online in e es in
adical Islam and e o is a acks, Peace Economics, Peace Science and Public Policy (PEPS), ISSN
1554-8597, De G uy e , Be lin, Vol. 31, Iss. 2, pp. 161-192,
h ps://doi.o g/10.1515/peps-2024-0054
This Ve sion is a ailable a :
h ps://hdl.handle.ne /10419/333342
S anda d-Nu zungsbedingungen:
Die Dokumen e au EconS o dü en zu eigenen wissenscha lichen
Zwecken und zum P i a geb auch gespeiche und kopie we den.
Sie dü en die Dokumen e nich ü ö en liche ode komme zielle
Zwecke e iel äl igen, ö en lich auss ellen, ö en lich zugänglich
machen, e eiben ode ande wei ig nu zen.
So e n die Ve asse die Dokumen e un e Open-Con en -Lizenzen
(insbesonde e CC-Lizenzen) zu Ve ügung ges ell haben soll en,
gel en abweichend on diesen Nu zungsbedingungen die in de do
genann en Lizenz gewäh en Nu zungs ech e.
Te ms o use:
Documen s in EconS o may be sa ed and copied o you pe sonal
and schola ly pu poses.
You a e no o copy documen s o public o comme cial pu poses, o
exhibi he documen s publicly, o make hem publicly a ailable on he
in e ne , o o dis ibu e o o he wise use he documen s in public.
I he documen s ha e been made a ailable unde an Open Con en
Licence (especially C ea i e Commons Licences), you may exe cise
u he usage igh s as speci ied in he indica ed licence.
h ps://c ea i ecommons.o g/licenses/by/4.0/
Ma cella Nicolini*, Fabio Saba ini and Dean Fan azzini
Online In e es in Radical Islam and Te o is
A acks
h ps://doi.o g/10.1515/peps-2024-0054
Recei ed Decembe 6, 2024; accep ed Feb ua y 13, 2025; published online Ma ch 6, 2025
Abs ac : This s udy in es iga es he link be ween e o is a acks and public in-
e es in adical Islam and iolen ex emism, using mon hly da a om 150 coun ies
be ween 2004 and 2015. Employing a dynamic common co ela ed effec s (DCCE)
es ima o o accoun o po en ial c oss-coun y co ela ions, ou analysis e eals
ha a acks ca ied ou in he name o Islam significan ly d i e online sea ches o
sensi i e keywo ds. Specifically, e ms sugges ing iolen ac ions, such as “behead-
ings,”and explici ly jihad- ela ed e ms show s onge co ela ions, indica ing
heigh ened in e es in he e o is s’ac ions and messages. Ou findings sugges ha
e o is a acks no only occu in a eas whe e public a en ion o e o ism is
al eady heigh ened bu also in ensi y in e es in iolen ex emism wi hin hose
egions. This amplifica ion may con ibu e o e o is s’objec i es by inc easing he
public isibili y hey seek.
Keywo ds: islamis ex emism; e o ism; Google sea ches; dynamic common
co ela ed effec s es ima o (DCCE)
JEL Classifica ion: C33; F50; H56
We hank Elena Esposi o, Domenico Giannone, Robe He ne , Michael Je e , Daniele Massacci, Je emy
Menchik, Daniele Pase man, Paolo Pino i, And ea Pozzi and semina pa icipan s a he META 2018
wo kshop in Nizhny No go od o use ul commen s and sugges ions. Luca Biglie i, Emma Cambie i,
An onella Cozzolino, Simona Fab is, Emanuela Padella o, Vi o ia Senzala i, and Vi o io Spinelli p o ided
excellen esea ch assis ance. Ma cella Nicolini hanks he Ins i u e o he S udy o Muslim Socie ies &
Ci iliza ions a Bos on Uni e si y o hospi ali y. All e o s and omissions a e ou own.
*Co esponding au ho : Ma cella Nicolini, Depa men o Economics and Managemen , Uni e si y o
Pa ia, Via San Felice 5 –27100 Pa ia (PV), I aly, E-mail: [email p o ec ed]. h ps://o cid.o g/0000-
0001-9034-2562
Fabio Saba ini, Sapienza Uni e si y o Rome, Rome, I aly; and IZA, Bonn, Ge many,
E-mail: [email p o ec ed]. h ps://o cid.o g/0000-0002-1228-3322
Dean Fan azzini, Moscow School o Economics, Moscow S a e Uni e si y, Moscow, Russia,
E-mail: [email p o ec ed]. h ps://o cid.o g/0000-0002-1481-3382
Peace Econ. Peace Sci. Pub. Pol. 2025; 31(2): 161–192
Open Access. © 2025 he au ho (s), published by De G uy e . This wo k is licensed unde he
C ea i e Commons A ibu ion 4.0 In e na ional License.
1 In oduc ion
Te o is s s i e o cap u e public a en ion o ec ui new ollowe s and dissemina e
hei message, ul ima ely bols e ing hei capaci y o inflic ha m (Nelson and Sco
1992). Unde s anding how public in e es in iolen ideologies ela es o e o is
a acks is c ucial o p e en ing he escala ion o adicaliza ion in o iolence. How-
e e , gauging he popula i y o e o is s and hei ideologies is challenging o
mul iple easons. News co e age o e o ism does no eliably indica e he public’s
ac i e in e es o opinions, which can be hea ily shaped by biased media sou ces
(Gen zkow and Shapi o 2004). Fu he mo e, cu iosi y abou and suppo o
ex emis ideologies a e o en unde s a ed and conside ed aboo, complica ing
effo s o measu e hese sen imen s h ough su eys o expe imen al me hods
(S ephens-Da idowi z 2014).
In his pape , we examine he ela ionship be ween public in e es in iolen
ex emism and e o is a acks, wi h a ocus on Islamis e o ism. This specific
ocus allows us o iden i y a ange o dis inc i e keywo ds associa ed wi h adical
Islam. Building on he classifica ion o keywo ds de eloped by Ahmed and Lloyd
Geo ge (2016), we selec ed 47 sensi i e online que ies and ga he ed hei mon hly
Google sea ch olumes globally om 2004 o 2015. These keywo ds include e ms
indica i e o cu iosi y abou adical Islam, exp essions o sympa hy owa ds
e o is ideologies, and ph ases explici ly ela ed o iolen jihadis ac ions, such as
he killing o infidels. Da a on e o is a acks we e sou ced om he Global
Te o ism Da abase (GTD). Since he GTD does no speci y he o igins o he a acks,
we unde ook addi ional effo s o documen he backg ounds and mo i a ions o he
e o is g oups in ol ed. Ou comp ehensi e da abase comp ises de ailed in o -
ma ion on 21,710 a acks ca ied ou wo ldwide by sel -p oclaimed Islamis g oups in
he name o Islam.
We p o ide e idence o a posi i e co ela ion be ween online sea ches o
sensi i e keywo ds and Islamis a acks. The associa ion is obus o con olling o
c oss-coun y dependence, which is especially c ucial in his con ex as a globally
common shock, such as he 9/11 a ack on he Twin Towe s, migh spillo e and ha e
a ying impac s ac oss diffe en coun ies. Upon looking in o specific keywo ds, we
find ha a acks a e mo e significan ly and sizably associa ed wi h “ex eme”
que ies, pa icula ly hose ela ing o iolence, such as “beheadings”, o hose
explici ly ela ed o he jihad, such as “Abdullah Azzam”, a Sunni schola and
ounding membe o Al Qaeda, also known as he “Fa he o Global Jihad”(Edwa ds
2017). This esul su i es se e al obus ness checks and placebo es s. When we
ex end he analysis o inco po a e spa ial lags, ou esul s indica e ha Google
sea ches in neighbo ing coun ies a e posi i ely and significan ly co ela ed wi h
162 M. Nicolini e al.
a acks. This sugges s ha when an Islamis a ack occu s in a gi en coun y, in-
di iduals in su ounding coun ies end o inc ease hei sea ches o sensi i e
keywo ds. In e es ingly, while he spa ial lag o ex eme keywo ds is no s a is i-
cally significan in he global sample, i becomes bo h s a is ically significan and
la ge in magni ude when es ic ing he analysis o coun ies ha ha e expe ienced
an Islamis a ack. O e all, hese findings highligh he ole o geog aphic p oximi y
in shaping online sea ch beha io , pa icula ly in egions wi h a his o y o Islamis
e o ism. We also expanded ou analysis by including empo al lags o Google
sea ches: he coefficien o Google sea ches in he p e ious mon h is posi i e and
significan , sugges ing ha pas sea ch ac i i y is posi i ely associa ed wi h subse-
quen Islamis a acks.
While we canno claim causali y in in e p e ing ou esul s, he empi ical
analysis con ibu es o he unde s anding o he ela ionship be ween e o ism and
he ac i e in e es in iolen ex emism in se e al ways. P e ious s udies sugges ha
he media co e age o e o is a acks eeds iolen ex emism, as he a en ion o
he public encou ages e o is s o plan new a acks. Ou app oach offe s a way o
measu e he public’s a en ion o ex emis ideas and documen s i s ela ionship
wi h e o is s’missions. The link be ween he occu ence o he a acks and
sea ches o iolen and ex eme keywo ds is s iking. This sugges s ha e o is s
choose o a ack in places whe e he e is a high le el o in e es and conside a ion o
hei claims and ope a ions. In u n, hese a acks nu u e public in e es in iolen
ex emism in hose egions po en ially aiding e o is s in ga ne ing he public
a en ion hey seek.
This pape b idges wo s ands o li e a u e. The fi s s and deals wi h he
ela ionship be ween media and e o ism (El Ouadghi i and Peillex 2018; Gen zkow
and Shapi o 2004; Je e 2017, 2019; Rohne and F ey 2007). Ou s udy builds on his
body o esea ch by linking i o he b oade empi ical li e a u e on he de e minan s
o e o ism, aiming o unco e he unde lying d i e s o e o is ac i i ies. P e-
ious esea ch has explo ed he ela ionship be ween e o ism and unde de el-
opmen (Fe idun and Sezgin 2008), po e y (Ca uso and Schneide 2013), educa ion
(Basse i, Ca uso, and Schneide 2018), declining social capi al (Hel s ein 2014), na -
u al esou ce en s (Ajide and Alimi 2020), fiscal decen aliza ion (A zaghi and
Gaibulloe 2024), immig a ion (Bo e and Böhmel 2016), and poli ical exclusion (Choi
and Piazza 2016). Ou wo k con ibu es o his body o li e a u e by using Google
sea ches o add ess he challenge o de ec ing he popula i y and celeb i y s a us o
e o is s. Mo eo e , we show ha he public’s ac i e a en ion o e o is s’claims,
such as in en ional sea ches o specific keywo ds, a he han passi e news con-
sump ion, is s ongly co ela ed wi h e o is a acks. This esul aligns wi h ecen
e idence ha e o is a acks p o oke endu ing public eac ions (Bo e, E hy ou-
lou, and Picka d 2024). These esponses can mani es no only as heigh ened in e es
Online In e es in Radical Islam and Te o ism 163
in e o is messaging and ea o u he a acks bu also as ein o ced na ional
iden i y and inc eased suppo o na ional and sup ana ional ins i u ions (E hy-
oulou, Picka d, and Bo e 2024). Addi ionally, ou finding ha he se e i y o a acks,
measu ed by he numbe o casual ies and inju ies, gene a es s onge public in-
e es in e o is ac i i ies u he co obo a es he insigh s om Bo e, E hy oulou,
and Picka d 2024; E hy oulou, Picka d, and Bo e 2024, who documen ha a acks
wi h a highe numbe o ic ims elici s onge and mo e p olonged public eac ions.
The second s and encompasses s udies using online sea ches o ack hidden o
difficul o measu e social phenomena such as acism (Doleac and S ein 2013; Giu-
lie i, Tonin, and Vlassopoulos 2019; S ephens-Da idowi z 2014), con acep i e use
and abo ion (Kea ney and Le ine 2015), eligiosi y (Ben zen 2019), po nog aphy
(MacInnis and Hodson 2015), and p osocial a i udes (Gu ie and Melniko 2016). We
connec o his li e a u e by in oducing he use o online sea ches in he s udy o
adicaliza ion and e o ism. Mo eo e , we con ibu e by adop ing an econome ic
app oach ha accoun s o c oss-sec ional dependence, which add esses some o he
c i icisms o using Google sea ches in esea ch.
The es o he pape p oceeds as ollows: in he nex sec ion, we desc ibe ou
da a on online sea ches o sensi i e keywo ds and e o is episodes. We hen
explain ou econome ic app oach in Sec ion 3. Sec ion 4 p esen s he analysis o he
ela ionship be ween he online in e es in Islamis ex emism and e o is a acks
and he obus ness checks. In Sec ion 5, we b iefly summa ize and discuss ou main
findings. Sec ion 6 concludes.
2 Da a
2.1 Measu ing In e es in Radical Islam and Violen Ex emism
As o Janua y 2019, almos 4.4 billion people we e ac i e In e ne use s, oughly
co esponding o 59 % o he global popula ion.
1
Wi h a ound 90 % ma ke sha e
wo ldwide be ween 2010 and 2015, Google is by a he wo ld’s leading sea ch en-
gine.
2
S a ing om Janua y 2004, he company has made a ailable in o ma ion on
he que ies ha people sea ch on he engine wi h a ool called Google T ends
(he ea e GT).
3
The empi ical e idence shows ha GT consis en ly and s ongly
1See www.s a is a.com.
2Da a sou ced om www.s a is a.com/s a is ics/216573/wo ldwide-ma ke -sha e-o -sea ch-engines/
3Fo any specific keywo d, GT p o ide an index o he olume o que ies en e ed in o Google in a
gi en geog aphic a ea. The que y index is he a io be ween he o al que y olume o he keywo d
in ques ion o e a ce ain pe iod and he o al numbe o que ies in he same pe iod. A he coun y
le el, Google p o ides sea ch ends wi h a mon hly equency. In any selec ed pe iod, he maximum
164 M. Nicolini e al.
de ec many phenomena o which con en ional da a a e publicly a ailable wi h a
delay, o a a lowe equency. Fo example, que ies consis en ly de ec economic
beha io s such as job sea ches (Bake and F adkin 2017), he issuance o bad checks
(Eksi, Gu dal, and O man 2017), deposi ealloca ion (Fech , Thum, and Webe 2019),
and he use o Ube se ices (Be ge , Chen, and F ey 2018), o name a ew.
Howe e , i is in he de ec ion o usually unde co e in e es s and beha io s ha
Google sea ches play a c ucial ole. Online sea ches a e gene ally conduc ed in
p i a e, elude social con ol, and allow people o exp ess socially aboo, an i-social, o
e en c iminal in e es s and a i udes ha hey would ha dly e eal in public. Fo
example, MacInnis and Hodson (2015) use GT o s udy how in e es in po nog aphy
ela es o eligiosi y and conse a ism ac oss Ame ican s a es. Google-based mea-
su es o acism s ongly p edic black mo ali y (Chae e al. 2015), black homeown-
e ship (Ha is and Yelowi z 2018), access o public se ices (Giulie i, Tonin, and
Vlassopoulos 2019), and he pe o mance o Ba ak Obama in 2008 and 2012 U.S.
p esiden ial elec ions (S ephens-Da idowi z 2014). Google da a ha e also p o ed
help ul in de ec ing o p edic ing con acep i e use and abo ion (Kea ney and
Le ine 2015), suicides (Ma-Kellams e al. 2015), d ug abuse (Pe due, Hawdon, and
Thames 2018), and he p opaga ion o sexually ansmi ed diseases such as syphilis
(Young e al. 2018) and HIV (Young and Zhang 2018).
Along hese lines, he in e es in an i-sys em g oups is a ypically hidden social
pa e n ha Google sea ches can help unco e . Resea ch in o he ela ionship be-
ween he In e ne and e o ism has so a ocused on he ole o social ne wo king
si es (Aly e al. 2016; Weimann 2015). T acking jihadi p opaganda on social media is
use ul o de ec ing ex emis s’communica ion and enac ing coun e - e o ism
measu es (Zei zoff2017). Howe e , adicalized indi iduals’ac i i ies on pla o ms
like Facebook and Twi e do no necessa ily cap u e he a en ion o he gene al
public in Islamis e o ism. People likely adicalize as a esul o he sp eading o
eelings sympa he ic o ex emis s ac oss socie y, a phenomenon ha eludes de ec-
ion h ough social media (Ahmed and Lloyd Geo ge 2016). Sea ch engines, by
con as , can help o de ec he ac i e a en ion o he public o ex emis ideas (Je e
2019).
In ou empi ical analysis, we use a lis o keywo ds ha , acco ding o he Cen e
on Religion & Geopoli ics (CRG), a e likely o lead o ex emis con en i en e ed in o
a sea ch engine. The CRG fi s de eloped a se o wo ds and ph ases ha migh be
used in esea ch on adical Islam o in delibe a e a emp s o access iolen ex emis
que y sha e is no malized o be 100, and he que y sha e a he ini ial da e is no malized o be ze o
(Choi and Va ian 2012). By agg ega ing millions o sea ches wi h a mon hly equency o e a long
pe iod ac oss a la ge sample o coun ies, GT allow cons uc ing eliable indica o s o a b oad a ay
o beha io al a iables.
Online In e es in Radical Islam and Te o ism 165
con en . The lis was hen employed as a basis o asce ain u he po en ially
associa ed keywo ds h ough specific ools such as Google Keywo d Planne , Google
Sugges ph ases, and Google- ela ed sea ches. The final lis comp ised 143 keywo ds.
Based on he sea ch equency da a and a quali a i e assessmen , he CRG hen
selec ed a se o 47 keywo ds. The se is he sum o wo subg oups. The fi s includes
popula keywo ds ha we e sea ched on a e age a a minimum equency o 500
imes pe mon h. The second g oup includes keywo ds explici ly ela ed o iolen
ex emism. Ahmed and Lloyd Geo ge (2016) conduc ed a sea ch engine esul page
(SERP) analysis o he keywo ds o asce ain he ex en o which hey e u n
“ex eme”con en . Though being non- iolen and mos ly ela ed o poli ical Islam,
keywo ds belonging o he fi s g oup e u n ex emis con en in a ema kable
sha e o cases, an icipa ing he isk o unin en ional exposu e o jihadi p opaganda.
The second se comp ises keywo ds ha , hough being less popula , ca y a highe
po en ial o isk. O e all, 44 % o he ma e ial was explici ly iolen . In con as ,
coun e -na a i e con en ou pe o med ex emis ma e ial in only 11 % o he
esul s gene a ed.
4
Se e al si es aiming o p esen legi ima e Islamic cul u e we e also ound o hos
ex emis ma e ial. Fo example, Kalamullah.com and Wo ldo Islam.in o –which
can be e ie ed by en e ing ou selec ion o keywo ds in o a sea ch engine –con ain
esou ces o o dina y Muslims and hose in e es ed in Islam, such as online e -
sions o he Qu an, collec ions o Hadi h li e a u e, and a a ie y o Islamic books.
Howe e , hese si es also p o ide jihadi manuals and audio lec u es by known
jihadis. O e all, he Google sea ches o he chosen se o keywo ds e u ned a as
a ay o undamen alis ma e ial spanning om non- iolen ex emism o explici ly
iolen jihadism.
We use he GT o he 47 keywo ds o measu e he ac i e in e es in adical Islam
in 149 coun ies o e he pe iod 2004–2015. Figu e 1 shows on a map he no malized
olume o sea ches o he keywo ds wo ldwide. While Google is he p edominan
sea ch engine in mos coun ies, he e a e some no able excep ions, e.g., China. This
could go in he di ec ion o weakening he link we expec o de ec be ween online
sea ches, measu ed h ough GT, and e o is a acks.
4Ahmed and Lloyd Geo ge (2016) looked a he fi s wo pages o esul s, as 92 % o Google afficis
on page one, wi h page wo only ecei ing 5 % o he affic and page h ee 1 % (Chi ika Insigh s 2013).
The au ho s spli he ex eme con en in o h ee ca ego ies: iolen , non- iolen , and poli ical
Islamis . Websi es we e ca ego ized as iolen i hey con ained ei he images o g aphic iolence o
exho a ions o iolence. The poli ical Islamis ca ego y included con en exp essing a specific a -
fini y o a pa icula Islamis g oup. The non-ex eme con en ega ded coun e -na a i e o
“neu al”websi es, such as news websi es o po als a ge ing esea che s and s uden s. Websi es
deemed o be ex eme bu non- iolen we e hose exp essing an i-Semi ic, homophobic, acis , o
sec a ian iews wi hou a depic ion o , o inci emen o, iolence.
166 M. Nicolini e al.
A a ie y o people may be sea ching o keywo ds ha e u n ex emis con-
en , om academics o jou nalis s. We do no claim o be measu ing he online
sea ch ac i i y o pe spec i e e o is s. Ins ead, we assume ha Google sea ches
measu e he in e es o he public in adical Islam and iolen ex emism. P e ious
s udies employing Google sea ches o assess he hypo he ical ou comes o social
phenomena all sha e he assump ion ha ends o specific keywo ds can p oxy he
incidence o hidden social pa e ns in he popula ion (e.g., S ephens-Da idowi z 2014;
Kea ney and Le ine 2015; Gu ie and Melniko 2016).
Table 1 epo s he lis o keywo ds. We p o ide a glossa y wi h a b ie
desc ip ion o hei meaning in Appendix A.1.
2.2 Te o is A acks
Da a on e o is a acks a e d awn om he Global Te o ism Da abase (GTD) main-
ained by he Na ional Conso ium o he S udy o Te o ism and Responses o
Te o ism (START) a he Uni e si y o Ma yland. The da abase collec s in o ma ion on
he a acks pe pe a ed a ound he wo ld om 1970 h ough 2018. Including mo e han
190,000 cases o e he whole pe iod, and 80,694 o e ou pe iod o analysis (2004–2015),
he GTD is he mos comp ehensi e unclassified da abase on e o is a acks.
5
Diffe -
en ly om o he da abases on e o ism, he GTD includes sys ema ic da a on domes ic
as well as ansna ional and in e na ional e o is inciden s. Fo each a ack, he
da abase p o ides in o ma ion on se e al a iables,amongwhich heda eandloca ion,
Figu e 1: Google ends mon hly a e ages o he mean o he 47 sea ch e ms. No es: The in e als
plo ed co espond o he qua iles o he dis ibu ion o he mon hly GTs o he a e age o he 47
keywo ds.
5See h ps://www.s a .umd.edu/g d/abou / and LaF ee and Dugan (2007) o mo e in o ma ion on
he da abase.
Online In e es in Radical Islam and Te o ism 167
he weapons used, he na u e o he a ge , he numbe o casual ies (dis inguishing
ic ims in o dea hs and wounded), and, when a ailable, he g oup o indi idual
esponsible o he a ack.
We make an addi ional effo o collec u he in o ma ion on he 999 e o is
g oups iden ified in he GTD ha ha e been ac i e in he 2004–2015 pe iod. We
classi y g oups acco ding o hei ideology o in en . Inspec ing each e o is e en
allows us o en ich he GTD da abase by u he ca ego izing he a acks acco ding o
he ideology, eligious backg ound, and mo i a ion o he pe pe a ing g oup. Ou o
80,694 e en s, i is possible o iden i y he g oup’s backg ound in 34,907 cases (43.3 %
o cases). Religious undamen alis o ganiza ions ha e caused wo- hi ds o hese
e en s, and 97.8 % o he a acks wi h a eligious iden i y ha e been pe pe a ed by
sel -p oclaiming Islamis g oups, co esponding o 21,710 egis e ed a acks. We
Table :Lis o he keywo ds.
Mos popula Mos ex eme
Apos asy Abdullah Azzam
Apos a e Amaq Agency
Apos a es Apos a e Islam
Ayman al-Zawahi i Apos a es in Islam
Calipha e Beheadings
C usade C usade A my
C usade s C usade s agains Islam
Dabiq Dabiq Pd
Dabiq Magazine How o do Jihad
Ibn Taymiyyah Ibn Taymiyyah Jihad
Islamic S a e Inspi e Magazine
Jihad Jewish Coali ion
Jihad meaning Jihad o Ummah
Kafi Jihad in he Qu an
Khali ah Meaning Khali ah
Khila ah Khila ah Sy ia
Kuffa Killing Apos a es
Ma y Killing Infidels
Ma y dom in Islam Killing Kuffa
Ma y s Mujahid
Mujahideen P epa ing o Jihad
Shahada Rafidah
Suicide es Soldie s o he Calipha e
Taghu
168 M. Nicolini e al.
co esponds o a change in he Islamis a acks equi alen o app oxima ely 12.4 % o
i s mean, indica ing a mode a e effec size in ela i e e ms. This numbe inc eases o
14.3 % in he subsample o coun ies hi by Islamis a acks.
9
We p esen ou main esul s using al e na i e measu es o he olume o Google
sea ches in Table 3.
10
Column 1 in Panel A epo s ou p e e ed specifica ion, wi h
he mean alue o he 47 keywo ds. In column 2, we p esen he esul s wi h he fi s
ac o o a p incipal componen analysis (PCA) o he 47 diffe en keywo ds. We
obse e a posi i e and s a is ically significan coefficien o GT. The lagged alue o
he dependen a iable, IA, also displays a posi i e and significan coefficien , as
expec ed: while in es iga ing he ela ionship be ween Islamis a acks and Google
sea ches, we canno neglec he exis ence o a acks in he p e ious pe iod.
To explo e he ela ionship be ween Google sea ches and he a acks in g ea e
dep h, we diffe en ia e be ween keywo ds. Ou goal he e is o dis inc ly ocus on
e ms ha may indica e a genuine in e es in adical ideology ins ead o me ely
eflec ing a heigh ened a en ion o e o ism news and e en s. Fi s , we adop he
classifica ion o “popula ”and “ex eme”keywo ds as p oposed by Ahmed and Lloyd
Geo ge (2016), which is epo ed in Table 1. We compu e he a e age olume o
Google sea ches o he wo g oups. Bo h “popula ”and “ex eme”keywo ds
significan ly and posi i ely ela e o he a acks, wi h ex eme keywo ds displaying a
la ge coefficien .
Panel B o Table 3 shows he esul s on he subsample o 63 coun ies ha
expe ienced a leas one Islamis a ack in he pe iod unde conside a ion. Ac oss
diffe en specifica ions, esul s epo ed in Panel B confi m hose on he ull sample:
he sign and s a is ical significance o coefficien s do no change, bu we no ice a
sligh inc ease in hei size. As coun ies ha did no ace a acks pe pe a ed in he
name o Islam a e now excluded om he sample, he specifica ion in Panel B likely
cap u es he ela ionship be ween he in e es o he public o Islamis ex emism
and Islamis a acks mo e nea ly. No su p isingly, he es ima ed coefficien s a e
sligh ly la ge in his subsample.
9In he sample o coun ies affec ed by any e o is a ack (Panel A), he effec o a one-s anda d-
de ia ion inc ease in Google sea ches is gi en by β×σ
GT
, whe e β= 0.0485 and σ
GT
= 2.607, yielding 0.
0485 ×2.607 = 0.1264. Exp essed as a pe cen age o he dependen a iable’s mean (IA =1.016), his
co esponds o 12.4 %. Fo he subsample o coun ies affec ed by Islamis a acks (Panel B), he effec
is β= 0.1079 and σ
GT
= 3.175, esul ing in 0.1079 ×3.175 = 0.3425. Rela i e o he mean o he dependen
a iable (IA =2.402), his co esponds o 14.3 %.
10 Diagnos ic es s suppo he choice o he DCCE es ima o . The combined e idence om he
Fishe - ype and he IPS panel uni oo es s confi ms ha he se ies a e s a iona y. The Pesa an
(2004) es o c oss-sec ional dependence sugges s significan c oss-sec ional co ela ions o all ou
a iables. The Pesa an (2007) panel uni oo es in he p esence o c oss-sec ional dependence also
ejec s he null hypo hesis o homogenous non-s a iona i y.
Online In e es in Radical Islam and Te o ism 175
To be e unde s and he posi i e co ela ion be ween Google sea ches and
e o is a acks pe pe a ed by sel -p oclaiming Islamis o ganiza ions, we dis in-
guish he keywo ds in o six g oups acco ding o hei meaning. The fi s wo ca e-
go ies include keywo ds e e ing o ISIS and Al Qaeda, espec i ely. In he hi d
g oup, we ocus on apos asy: in adical Islam, apos asy is no necessa ily unde s ood
as he pe sonal choice o enouncing a eligious belie . Ins ead, i is o en conside ed
a public ac o poli ical secession om he Muslim communi y. This iew makes
apos a es one o he a o i e a ge s o jihadi p opaganda (To es So iano 2010). The
ou h g oup comp ises keywo ds e oking iolen ac ions, such as “beheadings”. The
fi h and six h ca ego ies con ain e ms ela ed o eligious doc ines and he his o y
o poli ical Islam, espec i ely.
11
Fo each g oup, we compu e he a e age olume o
Google sea ches and assess i s ela ionship wi h e o is a acks pe pe a ed in he
name o Islam.
Panel A o Table 4 epo s esul s in all coun ies ha aced an a ack. The se s o
keywo ds ela ed o he wo mos p ominen Islamis e o is g oups, Al Qaeda and
ISIS, a e no significan . Que ies e e ing o he concep o apos asy in Islam, such as
“kafi ”and “kuffa ”, significan ly and posi i ely co ela e wi h he a acks. The index
o iolen keywo ds, including que ies like “beheadings”and “killing infidels”,
significan ly and posi i ely ela es o he a acks. The agg ega ions o keywo ds
comp ising he doc inal and his o ical e ms a e ins ead no s a is ically signifi-
can .
12
Panel B o Table 4 epo s he esul s on he subsample o coun ies affec ed by
Islamis a acks in he pe iod unde conside a ion. Again we obse e an inc ease in
he size o he es ima ed coefficien s.
We ex end ou analysis by inco po a ing spa ial lags o ou a iables o in e es .
Specifically, we enhance ou model by including spa ial lags o he Islamis a acks
a iable and Google sea ches o examine how e o is a acks and sea ch ac i i y in
neighbo ing coun ies influence each ocal coun y. Using a con igui y ma ix, we
cons uc spa ially lagged e sions o bo h he dependen and independen a iables
and in eg a e hese addi ional eg esso s in o he DCCE es ima ion. The esul s a e
p esen ed in Table 5.
This analysis e eals wo key insigh s. Fi s , we find no s a is ically significan
ela ionship be ween Islamis a acks in neighbo ing coun ies and he occu ence o
an a ack in a gi en ocal coun y. In o he wo ds, he likelihood o expe iencing an
Islamis a ack in coun y idoes no appea o be influenced by a acks in con iguous
coun ies.
11 Table A.2 in he Appendix A.2 defines he six ca ego ies in de ail.
12 The doc inal e ms a e no specific o Islam and could gene ically e e also o o he Ab ahamic
eligions, such as Ch is iani y. Resul s do no change i we ex end he Religion index by including
“Shahada”, a doc inal e m specific o Islam.
176 M. Nicolini e al.
Howe e , his null esul con as s sha ply wi h he effec o Google sea ch
ac i i y in neighbo ing coun ies. The spa ial lag o Google sea ches is posi i e and
s a is ically significan , sugges ing ha when an Islamis a ack occu s in coun y i,
indi iduals in su ounding coun ies end o inc ease hei sea ches o sensi i e
keywo ds. Mo e ele an is he finding ha i we conside popula and ex eme
keywo ds sepa a ely, he spa ial lag o ex eme wo ds is no s a is ically significan
in he global sample, bu i becomes s a is ically significan and la ge in magni ude i
we ocus on he subsample o coun ies ha expe ienced an Islamis a ack. O e all,
hese findings indica e ha geog aphic p oximi y plays a c ucial ole in shaping
online sea ch beha io , pa icula ly in egions whe e Islamis a acks a e mo e
p e alen .
Ano he way in which we can expand ou analysis is by including empo al lags
o he Google sea ches. The esul s epo ed in Table 6 show ha including Google
sea ches a ime −1 does no subs an ially al e ou main findings ega ding he
con empo aneous Google sea ches and he lag o he dependen a iable. The co-
efficien o Google sea ches in he p e ious mon h is posi i e and significan ,
Table :Al e na i e agg ega ions o GTs.
Panel A ()()()()()()
Coun ies hi by any e o is a ack ISIS Al Qaeda Infidels Violen Religion His o y
GT
−. . .*.*** . .
(.)(.)(.)(.)(.)(.)
IA
−
.*** .*** .*** .*** .*** .***
(.)(.)(.)(.)(.)(.)
R. . . . . .
# Obse a ions , , , , , ,
# Coun ies
Panel A ()()()()()()
Coun ies hi by Islamis a acks ISIS Al Qaeda Infidels Violen Religion His o y
GT −. . .*.** . .
(.)(.)(.)(.)(.)(.)
IA
−
.*** .*** .*** .*** .*** .***
(.)(.)(.)(.)(.)(.)
R. . . . . .
# Obse a ions , , , , , ,
# Coun ies
DCCE es ima es. The dependen a iable is he numbe o Islamis a acks. Fo he defini ion o he diffe en g oupings,
see Table A.. S anda d e o s in pa en heses, ***, **, and * deno e, espec i ely, %, %, and % significance le els.
Online In e es in Radical Islam and Te o ism 177
sugges ing ha pas sea ch ac i i y is posi i ely associa ed wi h subsequen Islamis
a acks. Howe e , when we diffe en ia e be ween popula and ex eme keywo ds,
we do no find any s a is ically significan effec s. No ably, he significance o
con empo aneous Google sea ches emains obus , pa icula ly o ex eme
keywo ds.
Finally, we explo e how Google sea ches o ou sensi i e keywo ds ela e o
indica o s o he ou comes o e o is missions. To his end, we epea he analysis
using he numbe o killed, wounded, and o e all affec ed people as dependen
Table :In oducing spa ial lags.
Panel A ()()()
Coun ies hi by any e o is a ack Mean Popula Ex eme
GT
.*** .** .***
(.)(.)(.)
GT
,s−
.*.*.
(.)(.)(.)
IA
−
.*** .*** .***
(.)(.)(.)
IA
,s−
. . .
(.)(.)(.)
R. . .
#Obse a ions , , ,
# Coun ies
Panel B ()()()
Coun ies hi by Islamis a acks Mean Popula Ex eme
GT
.*** .** .***
(.)(.)(.)
GT
,s−
.** .** .**
(.)(.)(.)
IA
−
.*** .*** .***
(.)(.)(.)
IA
,s−
. . .
(.)(.)(.)
R. . .
#Obse a ions , , ,
# Coun ies
DCCE es ima es. The dependen a iable is he numbe o islamis a acks. Mean is he unweigh ed mean o he GTs o
he keywo ds, o he Popula and Ex eme g oupings, see Table . S anda d e o s in pa en heses, ***, **, and
* deno e, espec i ely, %, %, and % significance le els.
178 M. Nicolini e al.
a iables.
13
We epo he esul s in Table 7: GT a e posi i e and significan when
conside ing he numbe o wounded and affec ed people, bo h on he ull se o
coun ies (Panel A) and in he subse o coun ies affec ed by an Islamis a ack (Panel
B). This sugges s ha la ge a acks, ha in ol e mo e ic ims, igge mo e in e es
on he web.
As a u he obus ness check, we discuss wo placebo es s o he main findings
epo ed in column 1 o Table 3. Fi s , we es whe he Google sea ches a e signifi-
can ly and posi i ely co ela ed wi h non-Islamis a acks. In Column 1 o Table 8, we
epo esul s using all kinds o a acks as dependen a iable, including hose
Table :In oducing empo al lags o Google sea ches.
Panel A ()()()
Coun ies hi by any e o is a ack Mean Popula Ex eme
GT
.*.*.**
(.)(.)(.)
GT
−
.*. .
(.)(.)(.)
IA
−
.*** .*** .***
(.)(.)(.)
R. . .
#Obse a ions , , ,
# Coun ies
Panel B ()()()
Coun ies hi by Islamis a acks Mean Popula Ex eme
GT
.** .*.**
(.)(.)(.)
GT
−
.*. .
(.)(.)(.)
IA
−
.*** .*** .***
(.)(.)(.)
R. . .
#Obse a ions , , ,
# Coun ies
DCCE es ima es. The dependen a iable is he numbe o islamis a acks. Mean is he unweigh ed mean o he GTs o
he keywo ds, o he Popula and Ex eme g oupings, see Table . S anda d e o s in pa en heses, ***, **, and
* deno e, espec i ely, %, %, and % significance le els.
13 We compu e he o e all numbe o affec ed indi iduals as he sum o he killed and inju ed
people.
Online In e es in Radical Islam and Te o ism 179
pe pe a ed by non-Islamis e o is s. In columns 2, we emo e Islamis a acks
om he pool. We do no find any significan co ela ion be ween hese a acks and
sea ches o ou sensi i e keywo ds.
As a second exe cise, we conside a e m ha is comple ely un ela ed o he
Islamis a acks. We ollow D’Amu i and Ma cucci (2017) and choose he e m
“dos”,whichisanac onym o Disk Ope a ing Sys em.
14
We expec Google
sea ches o he e m “dos” o be un ela ed o Islamis a acks. As epo ed in
column 3 o Table 8, he coefficien is no s a is ically significan . We also epea
heexe cisewi h heGT o “google”finding again ha i is no co ela ed wi h
he occu ence o he a acks. Panel B shows ha he esul s o he placebo es s do
no change when only conside ing he subsample o coun ies affec ed by Islamis
a acks.
Table :Robus ness wi h diffe en measu es o iolence.
Panel A ()()()
Coun ies hi by any e o is a ack y= killed y= wounded y= ic ims
GT
. .*.*
(.)(.)(.)
y
−
.*** .*** .***
(.)(.)(.)
R. . .
# Obse a ions , , ,
# Coun ies
Panel B ()()()
Coun ies hi by Islamis a acks y=killed y=wounded y= ic ims
GT . .*.*
(.)(.)(.)
y
−
.*** .*** .***
(.)(.)(.)
R. . .
# Obse a ions , , ,
# Coun ies
DCCE es ima es. The dependen a iable is espec i ely, he log o killed pe sons, he log o wounded pe sons, he log o
o al ic ims (equal o he sum o killed and wounded). S anda d e o s in pa en heses, ***, **, and * deno e,
espec i ely, %, %, and % significance le els.
14 I also s ands o “Depa men o S a e”in he U.S., “Demok a ska opozicija S bije”, a poli ical
pa y in Se bia, and he IATA code o he Dios ai po in Papua New Guinea.
180 M. Nicolini e al.
5 Discussion
Ou panel analysis e eals a sys ema ic ela ionship be ween Islamis a acks and
online sea ches o a se o sensi i e keywo ds o e he pe iod 2004–2015 wo ldwide:
online sea ches a e significan ly and posi i ely co ela ed o a acks. Diffe en kinds
o use s may sea ch o some o he keywo ds we employ in he empi ical analysis,
Table :Placebo es s.
Panel A ()()()()
Coun ies hi by any
e o is a ack
y= all a acks y= non-Islamis
a acks
y= Islamis a acks
GT
. −.
(.)(.)
dos
.
(.)
google
−.
(.)
y
−
.*** .*** .*** .***
(.)(.)(.)(.)
R. . . .
# Obse a ions , , , ,
# Coun ies
Panel B ()()()()
Coun ies hi by
Islamis a acks
y=all a acks y=non-Islamis
a acks
y=Islamis a acks
GT
. .
(.)(.)
dos
−.
(.)
google
−.
(.)
y
−
.*** .*** .*** .***
(.)(.)(.)(.)
R. . . .
# Obse a ions , , , ,
# Coun ies
DCCE es ima es. Columns –p esen placebo es s adop ing al e na i e dependen a iables. Columns and p esen
placebo es s adop ing diffe en explana o y a iables. The dependen a iable is he o al numbe o e o is a acks
(columns –), he numbe o non-Islamis a acks (columns –), while in columns and is he s anda d measu e
adop ed h ough he empi ical analysis, he numbe o Islamis a acks. S anda d e o s in pa en heses, ***, **, and
* deno e, espec i ely, %, %, and % significance le els.
Online In e es in Radical Islam and Te o ism 181
om s uden s who wan o deepen hei knowledge o poli ical Islam o jou nalis s
and schola s in e es ed in esea ching iolen ex emism. Online sea ches may also
de ec he celeb i y o e o is o ganiza ions. Howe e , we also use keywo ds
explici ly deno ing an in e es in ana icism and iolence ha migh be sea ched by
sympa hize s o he jihad.
Dis inguishing be ween keywo ds allows us o unde s and be e he posi i e
ela ionship be ween Google sea ches and he a acks. We di ide ou se o 47 key-
wo ds in o wo g oups. The fi s g oup con ains keywo ds wi h an a e age mon hly
equency highe han 500 sea ches in he UK. Such sea ches a e no explici ly
iolen and mos ly ega d poli ical Islam. The SERP analysis conduc ed by Ahmed
and Lloyd Geo ge (2016), howe e , shows ha hese seemingly non ex emis
sea ches o en e u n ex eme con en aimed a explici ly p omo ing adicaliza ion
among In e ne use s. The second g oup con ains mo e ex eme que ies, which a e
mos ly ela ed o iolen ex emism and sys ema ically lead o undamen alis ,
mili an , and explici ly iolen con en (Ahmed and Lloyd Geo ge 2016).
We find ha he online sea ches o bo h g oups o que ies a e significan ly
associa ed wi h he a acks. The coefficien o he popula keywo ds signals ha he
celeb i y o e o is s among he public is significan ly and posi i ely co ela ed wi h
he a acks.
When we inco po a e spa ial lags o examine how e o is a acks and sea ch
ac i i y in neighbo ing coun ies influence a acks wi hin each coun y, we find no
s a is ically significan ela ionship be ween Islamis a acks in neighbo ing coun-
ies and he occu ence o an a ack in a gi en ocal coun y. Howe e , he spa ial lag
o Google sea ches is posi i e and s a is ically significan , sugges ing ha when an
Islamis a ack occu s in a ce ain coun y, indi iduals in su ounding coun ies end
o inc ease hei sea ches o sensi i e keywo ds. Mo e no ably, when dis inguishing
be ween popula and ex eme keywo ds, we find ha he spa ial lag o ex eme
wo ds is no s a is ically significan in he global sample bu becomes s a is ically
significan and la ge in magni ude when es ic ing he analysis o he subsample o
coun ies ha ha e p e iously expe ienced an Islamis a ack. O e all, hese findings
highligh he ole o geog aphic p oximi y in shaping online sea ch beha io ,
pa icula ly in egions wi h a his o y o Islamis e o ism.
When we expand ou analysis, including empo al lags o he Google sea ches a
ime −1, we find ha his does no subs an ially al e ou main findings ega ding
con empo aneous Google sea ches and he lag o he dependen a iable. The co-
efficien o Google sea ches in he p e ious mon h is posi i e and significan , sug-
ges ing ha pas sea ch ac i i y is posi i ely associa ed wi h subsequen Islamis
a acks. Howe e , when we diffe en ia e be ween popula and ex eme keywo ds,
we do no find any s a is ically significan effec s. No ably, he significance o
182 M. Nicolini e al.
con empo aneous Google sea ches emains obus , pa icula ly o ex eme
keywo ds.
Ou finding ha la ge a acks– hose esul ing in a highe numbe o casual ies
and inju ies–gene a e g ea e online in e es aligns wi h ecen e idence on he
a ying in ensi y o public esponses o he se e i y o e o is a acks (Bo e,
E hy oulou, and Picka d 2024; E hy oulou, Picka d, and Bo e 2024). Specifically,
Bo e, E hy oulou, and Picka d 2024 ha e p o ided e idence ha la ge-scale a acks
elici mo e p olonged public eac ions, whe eas he impac o smalle -scale inciden s
ends o ade wi hin a mon h. Expanding on his, E hy oulou, Picka d, and Bo e 2024
ha e documen ed ha he emo ional esponse o e o ism ex ends beyond imme-
dia e ea , ein o cing na ional iden i y–pa icula ly when a acks in ol e a highe
numbe o ic ims and a e pe pe a ed by Islamic ex emis s–s eng hening he
sense o na ional belonging in he UK.
A key limi a ion o his analysis–sha ed by all s udies ha ely on Google
sea ches o ack hidden beha io s–is he inabili y o pe o m sen imen analysis.
Mo e b oadly, ob aining indi idual–le el insigh s in o he mo i a ions behind online
sea ches o de e mining which specific websi es use s isi a e ecei ing Google’s
sugges ed esul s is impossible. Ou app oach allows us o ack sea ches o sensi-
i e keywo ds, such as “jihad,”bu mo e explici que ies–like “how o join jihad,”
“jihad is good,”o “figh agains jihad”and “ e enge agains jihad”–gene a e sea ch
olumes oo low o be eliably analyzed.
Consequen ly, ou in e p e a ion o he posi i e co ela ion be ween Google
sea ches and e o is a acks is inhe en ly cons ained and equi es a deg ee o
specula i e easoning. While we p ima ily in e p e he inc ease in sea ch olume as
a eflec ion o heigh ened in e es in e o is ac i i ies, we canno disen angle his
effec om al e na i e explana ions–such as he amplifica ion o ea , as docu-
men ed by Bo e, E hy oulou, and Picka d (2024), o he ac i a ion o nega i e
ecip oci y mechanisms (Feh and Ga che 2010).
Fo ins ance, indi iduals may be mo i a ed o sea ch o e o ism- ela ed
con en ou o a desi e o e alia ion, po en ially seeking ways o coun e ac e o is
in e es s h ough hei ac ions. Un o una ely, he unde lying sen imen s d i ing
hese sea ches emain inaccessible and canno be inco po a ed in o any analysis
based on Google T ends da a.
A p omising di ec ion o u u e esea ch could be o in es iga e he ela ionship
be ween e o is a acks and beha io s indica i e o nega i e ecip oci y, offe ing
insigh s in o whe he he public exhibi s a desi e o e alia ion in he a e ma h o
such e en s. While online sea ches explici ly linked o e alia o y in en a e unlikely
o gene a e sea ch olumes la ge enough o sys ema ic analysis, his line o inqui y
could be ex ended o examine he po en ial impac o exposu e o a acks on o ing
beha io . Specifically, u u e esea ch could explo e whe he e o is a acks lead o
Online In e es in Radical Islam and Te o ism 183
inc eased suppo o poli ical pa ies ad oca ing s onge coun e e o ism mea-
su es–sugges ing ha elec o al p e e ences may, a leas in pa , be shaped by a
pe cei ed need o e ibu ion.
To da e, e idence sugges s ha Is aeli o e s exposed o ocke a acks om he
Gaza S ip end o inc ease hei suppo o igh –wing pa ies (e.g., Ge manski and
Zei zoff2014). Howe e , sys ema ic esea ch on whe he o e s exhibi a simila
e alia o y esponse o pe cei ed h ea s in b oade con ex s–such as Wes e n
Eu ope and he U.S.– emains limi ed.
Ano he po en ial limi a ion is ha Google may no be he p ima y sea ch engine
in some egions. Howe e , inco po a ing his ac o in o ou empi ical analysis
p esen s significan me hodological challenges ha would comp omise he eli-
abili y o ou es ima es. The main issue is he lack o consis en and eliable s a is ics
on he use o al e na i e sea ch engines ac oss coun ies du ing ou s udy pe iod
(2004–2015). While a ailable ma ke sha e da a indica e ha Google was he domi-
nan pla o m in mos coun ies, we lack p ecise in o ma ion on sea ch engine
p e e ences in egions whe e Google was less p e alen due o cul u al o poli ical
ac o s, such as China and Russia.
Addi ionally, while Google sea ch da a a e publicly accessible, he same does no
apply o o he pla o ms like Baidu o Yandex, making i impossible o assess
whe he sea ch olumes o specific keywo ds ollow simila pa e ns ac oss
diffe en sea ch engines. As a esul , any analysis ha in e ac s online sea ches wi h
an index o Google usage in each coun y would be inhe en ly un eliable. Add essing
his issue in u u e esea ch will depend on da a a ailabili y, pa icula ly om
pla o ms ope a ing in coun ies whe e go e nmen con ol o e media and online
in o ma ion emains s ingen .
While being awa e o hese limi a ions, which p e en u he in e p e a ion o
he co ela ions we find, we belie e ou choice o using Google sea ches has some
ad an ages. Fi s , Google sea ches do ack gene al in e es in some keywo ds, wi h
Google being he fi s landing page o almos all use s when sea ching he web.
O he pla o ms, e.g., Twi e , would no be as ep esen a i e. Second, he SERP
analysis by Ahmed and Lloyd Geo ge (2016) on he keywo ds we conside in he
p esen analysis ensu es ha hese e ms a e sensi i e, as use s can end up inad-
e en ly on ex emis online pages. Keeping in mind his ca ea , he posi i e co -
ela ion we find is consis en wi h p e ious e idence sugges ing ha media co e age
and celeb i y encou age e o is s o plan new missions (F ey and Os e loh 2018;
Je e 2017, 2019; Je e and Walke 2018). The significan and posi i e coefficien o
he que ies o he ex eme keywo ds, on he o he hand, suppo s he in e p e a ion
ha online sea ches may de ec no only he celeb i y, bu also he popula i y o he
ex emis message. The sp eading o cu iosi y and sympa hy o adical Islam and
iolen ex emism may signal he o ma ion o a e ile g ound o adicaliza ion.
184 M. Nicolini e al.
Chudik, Alexande , and M. Hashem Pesa an. 2015. “Common Co ela ed Effec s Es ima ion o
He e ogeneous Dynamic Panel Da a Models wi h Weakly Exogenous Reg esso s.”Jou nal o
Econome ics 188 (2): 393–420.
D’Amu i, F ancesco, and Ju i Ma cucci. 2017. “The P edic i e Powe o Google Sea ches in Fo ecas ing US
Unemploymen .”In e na ional Jou nal o Fo ecas ing 33 (4): 801–16.
Dickey, D. A., and W. A. Fulle . 1979. “Dis ibu ion o he Es ima o s o Au o eg essi e Time Se ies wi h a
Uni Roo .”Jou nal o he Ame ican S a is ical Associa ion 74 (366a): 427–31.
Doleac, J. L., and L. C. D. S ein. 2013. “The Visible Hand: Race and Online Ma ke Ou comes.”The Economic
Jou nal 123 (572): F469–F492.
Edwa ds, D. B. 2017. Ca a an o Ma y s. Sac ifice and Suicide Bombing in A ghanis an. Oakland, Cali o nia:
Uni e si y o Cali o nia P ess.
E hy oulou, G., H. Picka d, and V. Bo e. 2024. “Te o is Violence and he Fuzzy F on ie : Na ional and
Sup ana ional Iden i ies in B i ain.”Jou nal o Law, Economics and O ganiza ion 1–27. h ps://doi.o g/
10.1093/jleo/ewae003.
Eksi, O., M. Y. Gu dal, and C. O man. 2017. “Fines e sus P ison o he Issuance o Bad Checks: E idence
om a Policy Shi in Tu key.”Jou nal o Economic Beha io and O ganiza ion 143: 9–27.
El Ouadghi i, I., and J. Peillex. 2018. “Public A en ion o “Islamic Te o ism”and S ock Ma ke Re u ns.”
Jou nal o Compa a i e Economics 46 (4): 936–46.
Fech , F., S. Thum, and P. Webe . 2019. “Fea , Deposi Insu ance Schemes, and Deposi Realloca ion in he
Ge man Banking Sys em.”Jou nal o Banking and Finance 105: 151–65.
Feh , E., and S. Ga che . 2010. “Fai ness and Re alia ion: The Economics o Recip oci y.”Jou nal o Economic
Pe spec i es 3 (14): 159–81.
Fe idun, M., and S. Sezgin. 2008. “Regional Unde de elopmen and Te o ism: The Case o Sou h Eas e n
Tu key.”De ence and Peace Economics 19 (3): 225–33.
F ey, B. S., and M. Os e loh. 2018. “S a egies o Deal wi h Te o ism.”CESi o Economic S udies 64 (4):
698–711.
Gen zkow, M. A., and J. M. Shapi o. 2004. “Media, Educa ion and An i-ame icanism in he Muslim Wo ld.”
Jou nal o Economic Pe spec i es 16 (3): 117–33.
Ge manski, A., and T. Zei zoff. 2014. “E hnic G oups, Poli ical Exclusion and Domes ic Te o ism.”Ame ican
Poli ical Science Re iew 3 (108): 588–604.
Giulie i, C., M. Tonin, and M. Vlassopoulos. 2019. “Racial Disc imina ion in Local Public Se ices: A Field
Expe imen in he Uni ed S a es.”Jou nal o he Eu opean Economic Associa ion 17 (1): 165–204.
Gu ie , S., and N. Melniko . 2016. “Wa , Infla ion, and Social Capi al.”Ame ican Economic Re iew 106 (5):
230–5.
Ha is, T. F., and A. Yelowi z. 2018. “Racial Clima e and Homeowne ship.”Jou nal o Housing Economics 40:
41–72.
Hel s ein, S. 2014. “Social Capi al and Te o ism.”De ence and Peace Economics 25 (4): 363–80.
Im, Kyung So, M. Hashem Pesa an, and Yongcheol Shin. 2003. “Tes ing o Uni Roo s in He e ogeneous
Panels.”Jou nal o Econome ics 115 (1): 53–74.
Je e , M. 2017. “The Effec o Media A en ion on Te o ism.”Jou nal o Public Economics 153: 32–48.
Je e , M., and J. K. Walke . 2018. The Effec o Media Co e age on Mass Shoo ings. Bonn, Ge many: IZA
Discussion Pape No. 11900.
Je e , M. 2019. “The Inad e en Consequences o Al-Qaeda News Co e age.”Eu opean Economic Re iew
119 (IZA Discussion Pape No. 10708): 391–410.
Kea ney, M. S., and P. B. Le ine. 2015. “Media Influences on Social Ou comes: The Impac o MTV’s 16 and
P egnan on Teen Childbea ing.”Ame ican Economic Re iew 105 (12): 3597–632.
Online In e es in Radical Islam and Te o ism 191
LaF ee, G., and L. Dugan. 2007. “In oducing he Global Te o ism Da abase.”Poli ical Violence and
Te o ism 19 (2): 181–204.
Ma-Kellams, C., F. O , J. H. Baek, and I. Kawachi. 2015. “Re hinking Suicide Su eillance: Google Sea ch
Da a and Sel -Repo ed Suicidali y Diffe en ially Es ima e Comple ed Suicide Risk.”Clinical
Psychological Science 4 (3): 1–5.
MacInnis, C. C., and Go don Hodson. 2015. “Do Ame ican S a es wi h Mo e Religious o Conse a i e
Popula ions Sea ch Mo e o Sexual Con en on Google?”A chi es o Sexual Beha io 44 (1): 137–47.
Maliach, A. 2010. “Abdullah Azzam, Al-Qaeda, and Hamas: Concep s o Jihad and Is ishhad.”Mili a y and
S a egic Affai s 2 (2): 79–95.
Nelson, P. S., and J. L. Sco . 1992. “Te o ism and he Media: An Empi ical Analysis.”De ence and Peace
Economics 3 (4): 329–39.
Pe due, R. T., J. Hawdon, and K. M. Thames. 2018. “Can Big Da a P edic he Rise o No el D ug Abuse?”
Jou nal o D ug Issues 48 (4): 508–18.
Pesa an, M. Hashem. 2004. Gene al Diagnos ic Tes s o C oss Sec ion Dependence in Panelss. Bonn,
Ge many: IZA Discussion Pape No. 1240.
Pesa an, M. Hashem. 2006. “Es ima ion and In e ence in La ge He e ogeneous Panels wi h a Mul i ac o
E o S uc u e.”Econome ica 74 (4): 967–1012.
Pesa an, M. Hashem. 2007. “A Simple Panel Uni Roo Tes in he P esence o C oss-Sec ion Dependence.”
Jou nal o Applied Econome ics 22 (2): 265–312.
Pesa an, M. Hashem, and Ron Smi h. 1995. “Es ima ing Long-Run Rela ionships om Dynamic
He e ogeneous Panels.”Jou nal o Econome ics 68 (1): 79–113.
Pesa an, M. Hashem, Yongcheol Shin, and Ron P. Smi h. 1999. “Pooled Mean G oup Es ima ion o Dynamic
He e ogeneous Panels.”Jou nal o he Ame ican S a is ical Associa ion 94 (446): 621–34.
Rohne , D., and B. S. F ey. 2007. “Blood and Ink! The Common-In e es -Game be ween Te o is s and he
Media.”Public Choice 133 (1–2): 129–45.
S ephens-Da idowi z, S. 2014. “The Cos o Racial Animus on a Black Candida e: E idence Using Google
Sea ch Da a.”Jou nal o Public Economics 118: 26–40.
To es So iano, M. R. 2010. “The Road o Media Jihad: The P opaganda Ac ions o Al Qaeda in he Islamic
Magh eb.”Te o ism and Poli ical Violence 23 (1): 72–88.
Weimann, G. 2015. Te o ism in Cybe space: The Nex Gene a ion. New Yo k, NY: Wood ow Wilson Cen e
P ess wi h Columbia Uni e si y P ess.
Woold idge, J. M. 2002. Econome ic Analysis o C oss Sec ion and Panel Da a. Camb idge, Massachusse s,
London, UK: MIT P ess.
Young, S. D., and Q. P. Zhang. 2018. “Using Sea ch Engine Big Da a o P edic ing New HIV Diagnoses.”
PLoS One 13 (7): e0199527.
Young, S. D., E. A. To one, J. U a a, and S. O. A al. 2018. “Using Sea ch Engine Da a as a Tool o P edic
Syphilis.”Epidemiology 29 (4): 574–8.
Zei zoff, T. 2017. “How Social Media Is Changing Conflic .”Jou nal o Conflic Resolu ion 61 (9): 1970–91.
192 M. Nicolini e al.
