Carl E. Enomoto, Kiana Douglas
ISSN 2071-789X
RECENT ISSUES IN SOCIOLOGICAL RESEARCH
Economics & Sociology, Vol. 12, No. 1, 2019
233
DO INTERNET SEARCHES FOR
ISLAMIST PROPAGANDA PRECEDE
OR FOLLOW ISLAMIST TERRORIST
ATTACKS?
Carl E. Enomoto,
New Mexico State University, Las
Cruces, NM, U.S.A,
E-mail: cenomoto@nmsu.edu
Kiana Douglas,
New Mexico State University, Las
Cruces, NM, U.S.A,
E-mail: kianaldouglas@gmail.com
Received: October, 2018
1st Revision: December, 2018
Accepted: February, 2019
DOI: 10.14254/2071-
789X.2019/12-1/13
ABSTRACT. Using a Vector-Autoregressive (VAR) model, this
paper analyzes the relationship between Islamist terrorist
attacks and Internet searches for the phrases such as “join
Jihad” or “join ISIS.” It was found that Internet searches for
“join Jihad” and “taghut” (Arabic word meaning “to rebel”)
preceded the Islamist terrorist attacks by three weeks over the
period January 2014 to December 2016. Internet searches for
“kufar” (the derogatory Arabic word for non-Muslims)
preceded the attacks that resulted in deaths from the Islamist
terrorist groups. Casualties, including those injured and killed
by the Islamist groups, were also found to precede Internet
searches for “join Jihad” and “ISIS websites.” Countermeasures
to the usage of social media for terrorist activity are also
discussed. As an example, if Internet searches for specific terms
can be identified that precede a terrorist attack, authorities can
be on alert to possibly stop an impending attack. Chat rooms
and online discussion groups can also be used to disseminate
information to argue against terrorist propaganda that is being
released.
JEL Classification
: F52, F51,
C01
Keywords
: Islamist terrorist attack, propaganda, Internet search,
causality, VAR model.
Introduction
Islamist terrorist attacks are now becoming frequent, worldwide events. According to
CNN (2018), “Since declaring its caliphate in June 2014, the self-proclaimed Islamic State has
conducted or inspired more than 140 terrorist attacks in 29 countries other than Iraq and Syria,
where its carnage has taken a much deadlier toll. Those attacks have killed at least 2,043 people
and injured thousands more” (p.1). Some have attributed these attacks to social media and the
radicalization that follows. The examples include the shooting at the Fort Lauderdale Airport
in Florida on January 6, 2017. The shooter reportedly spoke with other “…jihadis in chatrooms
and websites inspired by the Islamic State, the terrorist group also known as ISIS or ISIL”
(Sheth, 2017). After the Orlando nightclub shooting on June 12, 2016, President Obama stated
that there was no clear link between the shooter and ISIS, but the shooter may have been
influenced by what he saw online (BBC News, 2016). On London’s Westminster Bridge, an
attacker drove into pedestrians and stabbed a policeman on March 22, 2017. Authorities were
Enomoto, C. E., & Douglas, K. (2019). Do Internet searches for Islamist
propaganda precede or follow Islamist terrorist attacks?. Economics and Sociology,
12(1), 233-247. doi:10.14254/2071-789X.2019/12-1/13
Carl E. Enomoto, Kiana Douglas
ISSN 2071-789X
RECENT ISSUES IN SOCIOLOGICAL RESEARCH
Economics & Sociology, Vol. 12, No. 1, 2019
234
considering the possibility that his actions were due to what he saw online (Holden and
Hosenball, 2017).
It may be, however, that it is not the searching for jihadist websites that leads to terrorist
attacks, but it is the terrorist attacks that are leading to the Internet searches for jihadist websites.
A portion of terrorist communication is not found on the regular world-wide-web where users
can find it using search engines like Google, but rather it is found on the Dark Web (Weimann,
2016). According to Weimann, the Dark Web is only accessible using special browsers or
special software. In this case, terrorist attacks are drawing in the attention of those individuals
who have sympathy for the beliefs of the attackers (or are just curious) and are causing those
individuals to search the Internet for Islamist propaganda. No radicalization or attacks may be
taking place because of the searches for terrorist propaganda.
The purpose of this paper is to determine the direction of causality between the Internet
searches for Islamist phrases such as “Join Jihad” or “Isis Websites,” and the number of actual
Islamist terrorist attacks worldwide. If the direction of causality is from the Internet searches
for terrorist propaganda to terrorist attacks, that does provide authorities with useful
information. However, there could be many external factors at play. For example, economic
factors may be giving rise to the Internet searches for terrorist propaganda in the first place.
Military conflicts in the states may have a correlation with terrorist attacks, independent of the
Internet searches for propaganda. Global meetings, official speeches and resolutions also affect
terrorist attacks, independent of or in conjunction with the Internet searches for terrorist
propaganda.
The purpose of this paper is not to consider all the factors affecting terrorist attacks
worldwide, but to focus on one aspect of terrorism specifically -- the timing between Internet
searches for terrorist propaganda and terrorist attacks.
Conway (2017) stated,
It is impossible to adequately answer the question of why the Internet is playing a greater
role in contemporary violent extremism and terrorism absent prior knowledge of what
role, if any, the Internet is playing in the latter. Unfortunately, basic descriptive research
is largely missing from this field, along with more complex theory-informed approaches
seeking to show causal connections. This is pretty astounding given the treasure trove
of data now available online.
Aaron Zelin observed as recently as 2013 that:
More than 11 years after the attacks of 9/11 and nearly a decade since the rise of popular
online jihadi Internet forums, there is strikingly little empirical research on the manner
in which jihadi activists use the Web to propagate their cause. Whereas researchers and
policy analysts have systematically collected and analyzed the primary source material
produced by al-Qaeda and its allies, very little work has been done on the conduits
through which that information is distributed—and even to what extent, anyone is
accessing that propaganda other than counterterrorism analysts (pp. 78-79).
The purpose of this paper is to fill this gap in the literature.
The outline of this paper is as follows. In the next section, a brief overview of the
literature on Islamist terrorist groups and social media will be given. Section 2 discusses the
data collected for this study followed by the model used to estimate the relationships between
internet searches for Islamist propaganda and Islamist terrorist attacks throughout the world.
The final section contains conclusions and a discussion of the findings.
Carl E. Enomoto, Kiana Douglas
ISSN 2071-789X
RECENT ISSUES IN SOCIOLOGICAL RESEARCH
Economics & Sociology, Vol. 12, No. 1, 2019
235
1. Literature review
1.1. Islamist Groups
According to the Constitutional Rights Foundation (2018), Islamist groups have
different goals but many common beliefs.
Although their goals may differ, Islamist groups generally want to set up states based
on Islamic fundamentalism, or literal interpretation of the Koran, the holy scripture of
Islam, and the Hadith, a collection of sayings of the Prophet Muhammad. They believe
that government based on Sharia, or Islamic law, is superior to any government based
on secular laws, democracy in which multiple political views are represented, or any
religion other than fundamentalist Islam.
Many Islamist terrorists, often called jihadist terrorists, view themselves as following
Muhammad’s example. Muhammad in A.D. 622 had to flee from Mecca with a small
band of followers. Yet in 630, he returned with an army of followers to conquer Mecca
and then spread Islam throughout the Arabian Peninsula. Terrorist groups often see
themselves as small bands that will similarly lead Islam to victory.
The overwhelming majority of Muslims deplore terrorist attacks and view them as
violating the Koran. Most fundamentalist Muslims also believe terrorism violates
Islamic law. Nonetheless, the Islamic State and other jihadist groups draw their
supporters from the ranks of Islamic fundamentalists (p. 1).
There are several Jihadist groups (Islamist terrorist groups) throughout the world.
Gardner (2014) lists the following. 1) Al-Qaeda in the Arabian Peninsula, 2) Islamic State in
Iraq and the Levant (ISIS), 3) Al-Qaeda in the Islamic Maghreb, 4) Boko Haram, 5) Al-Shabab,
6) Taliban, 7) Ansal al-Sharia in Libya, 8) Ansar al-Sharia in Tunisia, 9) Jemaah Islamiah, 10)
Abu Sayyaf, and 11) Ansar Bayt Al-Maqdis. Gardner stated that these groups have continued
to flourish in many countries due to corrupt governments that abuse their power and people.
While many papers in the literature have discussed the beliefs and activities of Islamist
terrorist groups, none of them that the authors are aware of, have statistically measured the
direction of causality between internet searches for Islamist propaganda and Islamist terrorist
attacks. If internet searches for specific terms are preceding terrorist attacks, countermeasures
can be developed. These countermeasures may involve closely monitoring internet activity for
specific searches, identifying individuals and organizations in the communication network, and
ascertaining plans of these entities. If, on the other hand, Islamist terrorist attacks are leading
to internet searches for Islamist propaganda but not vice versa, then internet searches for
Islamist propaganda may have been done by curiosity seekers or others who may not intend to
commit a terrorist attack.
1.2. Social Media and Islamist Terrorism
Twitter has been used by Islamist extremists to spread their propaganda and recruit
future members to their cause. According to Ullah (2017, p.10), “…as of August 2017, the
Sunni Islamist militia Jabhat al-Nusra had over 300,000 followers on affiliated sites, and
@ikhwanweb, the Muslim Brotherhood’s English-language account, had 145,000.”
Furthermore, the extremists are constantly changing their twitter accounts to avoid being
blocked by the regulators of Twitter. “One of the most notorious is the al-Shabaab press office’s
Carl E. Enomoto, Kiana Douglas
ISSN 2071-789X
RECENT ISSUES IN SOCIOLOGICAL RESEARCH
Economics & Sociology, Vol. 12, No. 1, 2019
236
English-language Twitter account, originally set up in 2011, which continually posts extremist
content, is continually blocked, and continually regenerates under slightly different names.
During Al-Shabaab's attack on a Kenyan shopping mall in 2013, for instance, the group live-
tweeted the attack’s progress despite Twitter’s best efforts to stop them” (ibid, p. 11-12). Ullah
provides further examples of how Islamist extremists have used social media such as Facebook
and YouTube, to further their cause and recruit new members. Examples include showing
videos where others had insulted the Prophet Muhammad and the burning of the Koran by a
Christian pastor in the U.S.
Parekh, Amarasingam, Dawson and Ruths (2018) have proposed a new way of
collecting data from social media like Twitter to track jihadist terrorist plans, operations and
movements. By accounting for the relations and interactions of social media users, a “social
graph” can be modeled which excludes irrelevant information.
Hassan (2018) has recognized that propaganda from the Islamic State continues to be
put on the internet and authorities have a difficult time controlling it. The propaganda focuses
on generating support against non-Muslim armies in Muslim territory. Hassan also highlights
how the Islamic State tries to persuade youth to go against parental consent and join jihad.
However, it is pointed out that extremist Islamist groups do not have the theological justification
for doing this.
Soliev (2018) discussed the threats made by the Islamic State against the 2018 World
Cup games in Russia, using online propaganda. This threat may have been brought on by
Russian involvement in Syria and Iraq. Soliev traces the time line of propaganda against the
World Cup by looking at specific events. “In October 2017, the pro-IS Wafa’ Media Foundation
released digitally altered images of football superstars like Neymar, Lionel Messi and Cristiano
Ronaldo being executed in its propaganda materials. IS supporters and sympathizers in different
parts of the world have subsequently joined the pro-IS online extremist group in their
propaganda campaign (p. 17).“ Soliev argues that the propaganda campaign by the Islamic State
could bring about attacks from other sources as well.
In a recent U.S. Senate hearing, opinions were voiced on “Isis Online: Countering
Terrorist Radicalization and Recruitment on the Internet and Social Media” (Committee on
Homeland Security and Governmental Affairs, U.S. Senate, 2016). Senator Portman, in his
opening comments, reported on how ISIS had used online propaganda as a weapon. “The
damage wrought by that weapon is considerable: Orlando, 49 dead; San Bernardino, 14; Fort
Hood, 13 dead; the Boston Marathon, 3 dead and hundreds wounded. Each of these killers was
reportedly radicalized to some degree by online jihadist content (p. 2).” The subcommittee also
noted that ISIS posts online instructions to those in English-speaking countries, on how to make
their trip to Syria (p. 71).
Charlie Winter (2016) stated that most Islamist propaganda is in Arabic, however, “there
are official propaganda channels operating in no less than nine languages…” (p. 9). The
propaganda discusses life in the caliphate, paints a picture of abundance and friendship, and
presents an alternative lifestyle which attracts recruits. The final step in the recruitment process
involves an enlister. “This is the point at which the process usually enters encrypted
communications and becomes clandestine (p. 11).” Von Behr, Reding, Edwards, and Gribbon
(2013) analyzed how the internet was used in the radicalization of terrorists in the U.K. They
reported that a Google search for “how to make a bomb” gave 1,830,000 results. A search for
“Salafi publications” gave 46,200 results and a search for “beheading video” gave 257,000
results (p. 3). The authors concluded that the internet increased the opportunity and speed of
radicalization, based on interviews with radicalized individuals.
Bertram and Ellison (2014) reported that 112 websites were found that were associated
with Sub Saharan African terrorist groups (p. 9). Most of the terrorist web activity took place
in the Eastern, Western and Southern regions of Africa.
Carl E. Enomoto, Kiana Douglas
ISSN 2071-789X
RECENT ISSUES IN SOCIOLOGICAL RESEARCH
Economics & Sociology, Vol. 12, No. 1, 2019
237
Veilleux-Lepage (2016) discussed the evolution of Islamist propaganda in moving
Muslims to jihad. Starting with amateur videos from 2004 to 2007 and the first “authentic
jihadist website,” azzam.com, extremists soon moved into YouTube (p. 37). The author stated,
“Aided by the popularity of such mediums and the fact that no Arabic language skills or high
level of Internet literacy were now required to locate jihadist content, YouTube rapidly became
a significant platform for jihadist groups and their supporters, fostering a thriving subculture
which used it to communicate and share propaganda worldwide” (p. 39). Twitter accounts and
e-Magazines have also been used by terrorist groups.
Thompson (2011) described why social media like Facebook, Twitter, and YouTube,
are well suited to meet the needs of radicals. By uploading videos or sending out a tweet, one
individual can influence many. Radicalization and recruitment are more easily accomplished
from those countries with access to the internet.
Conway (2006) examined how terrorists use the internet. These uses include, (1)
information provision and publicity, (2) financing through the sale of books and videos, (3)
exploitation of e-commerce tools and entities such as credit card fraud and the use of internet
businesses to fund terrorist activities, (4) networking to allow greater communication between
different terrorist groups, (5) recruitment, (6) data mining to gather sensitive information about
other nations, and (7) sharing information such as how to make bombs.
Conway (2012) gave several examples of how the internet had been used by those in
favor of violent jihad to spread their message and plan and carry out attacks. Online
radicalization could have played a role in the cases of the London bombers in 2005, the shooting
of two U.S. airmen in the Frankfurt Airport in 2010, the Fort Hood shootings in 2009, and
others (p. 1). Furthermore, the internet and social media have made the messages of the violent
jihadist more readily available to those who do not speak Arabic as these messages are now
translated into many languages.
Weimann (2004) analyzed how terrorist groups use the internet. He found that the
websites of terrorists contained information about the group, their political beliefs, their
accomplishments, leaders, and goals. The terrorist groups also wanted to spread disinformation
and fear and gain publicity for themselves. They also want to use their websites to justify the
violence they use against enemies.
The Homeland Security Institute (2009) outlined how the internet was being used by
terrorists to attract young people to their organizations. By using games, cartoons, videos and
special designs on their websites, terrorist groups can introduce themselves to young people
whom they would not otherwise reach. Younger people are also more likely to be in close
communication with their peers, and thus terrorist messages are more easily transmitted to a
larger audience.
The United Nations Office on Drugs and Crime (2012) reported on the use of the internet
by terrorist groups. Some of the reasons these groups used the internet were to promote their
propaganda, promote violence, radicalize groups of young people, recruit new members,
finance their organizations, and plan an attack. Ways the internet could be used to counter
terrorist activity were also given.
A paper by Beckmann, Dewenter, and Thomas (2017) and another paper by Pfeiffer
(2012) looked at the relationship between news media coverage and terrorist attacks. Using
Granger Causality tests, they found a bidirectional relationship between news articles on
terrorism and terrorist attacks.
In the next section, the data used in this study to examine the relationship between
internet searches for Islamist propaganda and Islamist terrorist attacks will be described.
Carl E. Enomoto, Kiana Douglas
ISSN 2071-789X
RECENT ISSUES IN SOCIOLOGICAL RESEARCH
Economics & Sociology, Vol. 12, No. 1, 2019
238
2. Methodological approach
Weekly worldwide data from 2014 to 2016 was collected from the website, “The
Religion of Peace,” for the number of Islamist terrorist attacks, the number of casualties from
such attacks (killed and injured), and the number killed from such attacks. Table 1 provides
descriptive statistics for these variables.
Table 1. Descriptive Statistics of Variables
Variable Mean std. dev.
number of attacks 53.86 10.91
total number of casualties 1,036.15 515.14
killed 522.97 391.1
Number of weekly internet searches for phrases such as “Join Jihad,” “Join ISIS,” “ISIS
chat rooms,” “ISIS recruitment”, “ISIS websites,” “ISIS twitter,” “khilafa,” (successor
(Fernandez, 2015)), “jizya,” (poll tax to humiliate Christians and Jews (ibid)), “taghut,”
(jihadist term for “others” associated with idolatry and unbelief (ibid)), and “kufar,” (which
refers to infidels and is a term used for the jihadist movement (ibid)), were taken from Google
Trends. Furthermore, “taqiyya” (concealing one’s true identity or denying one’s religion (1389
Blog-Counterjihad, 2013)), was translated into its Arabic representation and a measure of
weekly internet searches for this term was found using Arabic Google Trends.
According to Fernandez (2015), “There certainly is a jihadist ideology that must be
defeated, beyond crushing ISIS on the battlefield in Syria and Iraq, but in the volatile world of
social media, what often marks jihadism is not so much an ideology fully formed and
understood but a semi-digested revolutionary argot of highly symbolic words and phrases –
jihadist shorthand mostly divorced from history and context, dumbed down for the zealous
convert, and used as an ideological blunt instrument” (p. 1). Fernandez stated that “khilafa,” for
example, was revived by ISIS to remind people of the caliphate. “Jizya,” was revised by ISIS
to let Christians and Jews know that if they don’t convert, they will be subject to discrimination
and taxes. “Taghut,” has a complicated history, according to Fernandez, but can refer to the
overthrow of anti-jihadist regimes. “Taqiyya,” comes from the Quran which states, “Let
believers not make friends with infidels in preference to the faithful…,” according to 1389
Blog-Counterjihad (2013).
The different phrases described above were put into Google Trends which gives search
volume indexes over time for the different phrases. As an example, if a search is done for “ISIS
websites,” and the geographical area selected for searches in the world, then the search volume
index (SVI) will be a number between 0 and 100 for different weeks in a year. The SVI for a
week is calculated by taking the number of worldwide searches for the phrase “ISIS websites”
and dividing by the total number of worldwide Google searches. This proportion is then
normalized by the week with the highest proportion. As an example, suppose for the first week
of May 2017, Google searches for the phrase “ISIS websites” as a proportion of all searches
was 0.30, and on the first week of September 2016, the proportion was 0.90. Also, suppose that
the week with the highest proportion of searches was the first week of September 2016. All
weekly proportions are then normalized to 0.90. Thus, the search volume index for the phrase
“ISIS websites,” for the first week of May 2017 would be 0.30/0.90 which would be 0.33 or 33
after multiplying by 100. A higher search volume index for a given week shows greater interest
in a particular phrase relative to a week with a lower SVI.
Google Trends search volume indices have been used by many researchers. Ricketts and
Silva (2017) used Google Trends to analyze morbidity and mortality based on specific Google
Carl E. Enomoto, Kiana Douglas
ISSN 2071-789X
RECENT ISSUES IN SOCIOLOGICAL RESEARCH
Economics & Sociology, Vol. 12, No. 1, 2019
239
search terms. Vicens-Feliberty and Ricketts (2016) used Google Trends to analyze how Google
searches in Puerto Rico for terms involving “moving to the U.S.,” affected actual migration.
Enomoto, Noor, and Widner (2017) used Google Trends to see how searches for different dating
sites and apps such as Tinder, Zoosk, and Eharmony, were related to cases of STDs in different
U.S. states. Other papers that have used Google Trends include Choi and Varian (2012), Vosen
and Schmidt (2011), and Wu and Brynjolfsson (2013).
All variables and number of searches for the various terms and phrases (as measured by
Google Trends search volume indices) used in this study were collected each week from
January 2014 to December 2016 which resulted in 156 observations. Furthermore, all variables
and number of searches for the different phrases were found to be stationary using the
Augmented Dickey-Fuller test.
In the next section, the Vector Autoregressive (VAR) model used in this study is
described and the results are presented.
3. Conducting research and results
VAR models (vector-autoregressive models) have been used by many researchers to
determine the direction of causality between variables such as advertising and sales, monetary
policy and interest rates, economic development and political democracy, and many other
variables. The advantage of using VAR models is that once the optimal lag structure for the
variables is found using different statistics such as the AIC (Akaike information criterion) or
SC (Schwarz information criterion) or other established statistics, Granger causality tests can
be applied to see which variable affects the others. In this study, a two-equation, two-variable
VAR (vector autoregressive) model was estimated such as the one below, to determine if
internet searches for Islamist propaganda preceded or followed Islamist terrorist attacks.
0 1 1 1 1
... ...
t t k t k t k t k t
Yβ β Y β Y α X α X u
(1)
0 1 1 1 1
... ...
t t k t k t k t k t
Xδ δ Y δ Y γ X γ X v
. (2)
In equations (1) and (2), the variable
t
Y
represents a weekly measure of Islamist
terrorist activity. Such measures include (a) number of Islamist terrorist attacks worldwide, (b)
a total number of casualties from Islamist terrorist attacks worldwide including number killed
and the number injured, and (c) total number killed from Islamist terrorist attacks worldwide.
The variable
t
X
represents weekly worldwide internet searches for Islamist propaganda as
measured by Google Trends search volume indices. Examples of
t
X
include search phrases
such as “join jihad,” “Isis websites,” and “taghut” which is jihadist shorthand for “the other,”
for idolatry and unbelief” (Fernadez, 2015).
In the VAR model, there are two hypotheses of interest:
1
0 1 2
: ... 0Hαα
and
2
0 2 3
: ... 0.Hδδ
One of four conclusions can be reached: (a) We fail to reject
1
0
H
and
2
0
H
, in which case there is no Granger causality between X and Y. (b) We fail to reject
1
0
H
but
reject
2
0
H
, in which case Y is said to Granger-cause X. (c) We reject
1
0
H
but fail to reject , in
which case X is said to Granger-cause Y. (d) We reject
1
0
H
and
2
0
H
, in which case there is
bidirectional causality between X and Y.
Carl E. Enomoto, Kiana Douglas
ISSN 2071-789X
RECENT ISSUES IN SOCIOLOGICAL RESEARCH
Economics & Sociology, Vol. 12, No. 1, 2019
240
3.1. Application of multiple linear regression analysis
In the first model that was estimated,
t
Y
was the weekly worldwide number of Islamist
terrorist attacks. The variable
t
X
was a measure of the weekly worldwide number of internet
searches for the phrase “join jihad.” As previously discussed, the Google trends search volume
index for “join jihad” was used for this measure. To determine the lag length used in the VAR
model represented by equations (1) and (2), five statistics were examined: (1) the LR sequential
modified LR test statistic, (2) the FPE (the final prediction error statistic), (3) the AIC (Akaike
information criterion), (4) the SC (Schwarz information criterion), and (5) the HQ (Hannan-
Quinn information criterion). The lag length that was identified as the best from the majority of
the five statistics was used. In this case, that lag length was three weeks. Furthermore, all of the
inverse roots of the characteristic polynomial were within the unit circle indicating a stable
VAR model. Tests for autocorrelation (Portmanteau autocorrelation test) and heteroscedasticity
(White’s heteroscedasticity test using levels and squares of the regressors) indicated no
autocorrelation or heteroscedasticity present. The results of the Granger causality test are given
below in Table 2.
Table 2. VAR Granger Causality Tests Between Number of Islamist Terrorist Attacks
and Internet Searches for "Join Jihad."
Dependent Variable: Number of Islamist Terrorist Attacks
Variable Chi-square DF Prob
Join Jihad 9.718* 3 0.0211
Dependent Variable: "Join Jihad"
Variable Chi-square DF Prob
Number of Islamist
Terrorist Attacks 1.8 3 0.6149
*Indicates significance at the 5% level
In this case, internet searches for “Join Jihad” are preceding Islamist terrorist attacks by
three weeks, given the significant chi-square statistic of 9.718. However, the number of Islamist
terrorist attacks is not leading to internet searches for “Join Jihad,” given the insignificant chi-
square statistic of 1.8. These results are reinforced when examining the estimates of equations
(1) and (2) of the VAR model.
1 2 3 1 2 3
(2.68 )* ( 3.30) (2.75 ) (3.07 ) ( 0.50 ) ( 0.71) ( 2.33)
10.99 0.26 0.22 0.24 0.03 0.04 0.12
t t t t t t t
Y Y Y Y X X X
(1a)
R-squared = 0.45, AIC = 7.12, n = 153, *t-statistics in parentheses
1 2 3 1 2 3
(1.16)* (1.30) ( 0.65) ( 0.51) (4.93 ) (2.16) (1.67 )
7.20 0.16 0.08 0.06 0.40 0.18 0.14
t t t t t t t
X Y Y Y X X X
(2a)
R-squared = 0.39, AIC = 7.95, n = 153, *t-statistics in parentheses
Carl E. Enomoto, Kiana Douglas
ISSN 2071-789X
RECENT ISSUES IN SOCIOLOGICAL RESEARCH
Economics & Sociology, Vol. 12, No. 1, 2019
241
With Y = number of attacks and X = internet searches for “Join Jihad,” equation (1a)
and the results from Table 2, show that the past values of X jointly affect the current value of
Y after controlling for past values of Y. More specifically, the coefficient
3t
X
is positive and
significant indicating that internet searches for the term “Join Jihad” three weeks ago had a
positive effect on the current number of Islamist terrorist attacks worldwide. Equation (2a) and
the results from Table 2, on the other hand, show no individual effects or group effects of past
terrorist attacks (Y) on current internet searches for “Join Jihad” (X).
In the above model,
t
Y
was used to represent the weekly worldwide number of Islamist
terrorist attacks. It may be, however, that other measures of Islamist terrorist activity either lead
to internet searches for Islamist propaganda or are led by internet searches. In a second model,
t
Y
was defined to be a weekly worldwide number of casualties (killed and injured) from Islamist
terrorism. Thus, it may be that it is the actual number killed and injured from Islamist terrorism,
rather than simply the number of incidents or Islamist terrorist attacks, that affects or is affected
by internet searches for the phrase “join Jihad.” Equations (1) and (2) above were again
estimated with
t
Y
taking on this new definition. The best lag length was determined to be two
weeks. The VAR model was also identified as stable with no autocorrelation or
heteroscedasticity detected. The results of the Granger causality test are given in Table 3.
Table 3. VAR Granger Causality Tests Between Number of Casualties from Islamist
Terrorist Attacks and Internet Searches for "Join Jihad."
Dependent Variable: Number of Casualties from Islamist Terrorist Attacks
Variable Chi-square DF Prob
Join Jihad 3.855 2 0.1455
Dependent Variable: "Join Jihad"
Variable Chi-square DF Prob
Number of Casualties from
Islamist Terrorist Attacks 7.223* 2 0.027
*Indicates significance at the 5% level
The results from Table 3 indicate that searches for the phrase “join Jihad,” did not
precede or Granger cause number of casualties from Islamist terrorist attacks. However, a
number of casualties did precede or Granger causes internet searches for “join Jihad.” While
the number of Islamist terrorist attacks per week did not give rise to internet searches for “join
Jihad” from the previous model, a number of casualties has had a greater impact on individuals
and their internet search behavior. This may in part, be due to the news coverage is given these
types of events. If terrorist attacks are unsuccessful for the terrorists and no one is injured or
killed, there may be little news coverage. If there are casualties, there is more news coverage
and more internet searches for Islamist propaganda.
The results from Table 3 are reinforced by looking at the estimates of equations (1) and
(2) when
t
Y
is a number of weekly-worldwide casualties. These estimates are presented in
equations (1b) and (2b).
Carl E. Enomoto, Kiana Douglas
ISSN 2071-789X
RECENT ISSUES IN SOCIOLOGICAL RESEARCH
Economics & Sociology, Vol. 12, No. 1, 2019
242
1 2 1 2
(5.92)* (1.61) ( 0.35 ) ( 0.63) (1.90)
0.82 0.13 0.03 2.01 5.96
t t t t t
Y Y Y X X
(1b)
R-squared = 0.04, AIC = 7.12, n = 154, *t-statistics in parentheses
1 2 1 2
(1.15)* ( 2.65) (0.16) (5.11) (3.43)
3.87 0.01 0.0003 0.40 0.26
t t t t t
X Y Y X X
(2b)
R-squared = 0.39, AIC = 7.92, n = 154, *t-statistics in parentheses
In equations (1b) and (2b),
t
Y
is a number of weekly worldwide casualties from Islamist
terrorist attacks and
t
X
is number of weekly internet searches for “join Jihad,” which is
measured using the Google Trends search volume index. The results support the Granger
causality tests from Table 3, showing that in equation (1b), internet searches are not preceding
a number of casualties. However, equation (2b) shows that the number of casualties per week
is preceding or Granger causing internet searches for “join Jihad.” More specifically, equation
(2b) indicates that number or casualties in a week have a significant and positive effect on a
number of internet searches for “join Jihad” in the following week.
Equations (1) and (2) were re-estimated to consider the internet search phrase “ISIS
websites,” In this case,
t
Y
was a weekly worldwide number of casualties from Islamist terrorist
attacks and
t
X
was a weekly number of internet searches for “ISIS websites.” The lag length
was determined to be one week given the LR, FPE, AIC, SC, and HQ statistics. The VAR model
was stable with no autocorrelation or heteroscedasticity detected. Granger causality tests
indicated that casualties from Islamist terrorist attacks preceded internet searches for “ISIS
websites,” by one week. No Granger causality was found when looking at a number of internet
searches for “ISIS websites” and number of Islamist terrorist attacks.
Another VAR model was estimated to consider internet searches for “Taghut,” which
refers to disbelievers or people who worship a god other than Allah. In equations (1) and (2),
t
Y
was a number of weekly worldwide attacks by Islamist terrorist groups and
t
X
was weekly
worldwide- internet searches for the term “Taghut.” The best lag length was determined to be
three weeks, and the VAR model was stable with no indication of autocorrelation or
heteroscedasticity. The Granger causality tests are given in Table 4.
Table 4. VAR Granger Causality Tests Between Number of Islamist Terrorist Attacks
and Internet Searches for "Taghut".
Dependent Variable: Number of Islamist Terrorist Attacks
Variable Chi-square DF Prob
Taghut 8.564* 3 0.036
Dependent Variable: "Taghut"
Variable Chi-square DF Prob
Number of Attacks from
Islamist Terrorist Groups 3.075 3 0.38
*Indicates significance at the 5% level
Carl E. Enomoto, Kiana Douglas
ISSN 2071-789X
RECENT ISSUES IN SOCIOLOGICAL RESEARCH
Economics & Sociology, Vol. 12, No. 1, 2019
243
The results in Table 4 show that internet searches for “Taghut” preceded or Granger
caused a number of Islamist terrorist attacks, but attacks did not Granger cause internet searches
for “Taghut.” The internet searches may be an indication of the discontent, frustration,
resentment, and anger that extremist groups are feeling towards those they feel are infidels.
The results of Table 4 are consistent with the estimates of equations (1) and (2) of the
VAR model given below.
1 2 3 1 2 3
(2.84 )* ( 4.45) (2.25 ) (2.82 ) ( 2.26) ( 2.10) ( 0.97 )
14.18 0.36 0.19 0.22 0.11 0.11 0.05
t t t t t t t
Y Y Y Y X X X
(1c)
R-squared = 0.45, AIC = 7.12, n = 153, *t-statistics in parentheses
1 2 3 1 2 3
(3.47 )* (0.63) ( 1.00) ( 0.88) ( 2.17 ) (1.42) (2.14)
27.97 0.08 0.13 0.11 0.18 0.12 0.18
t t t t t t t
X Y Y Y X X X
(2c)
R-squared = 0.44, AIC = 8.08, n = 153, *t-statistics in parentheses
In equations (1c) and (2c),
t
Y
was a number of attacks and
t
X
was number of internet
searches for “Taghut.” In equation (1c), the coefficients of
1t
X
and
2t
X
were the same but
opposite in sign. Increases in internet searches for “Taghut” are leading to more attacks in two
weeks while decreases in internet searches for “Taghut” are leading to more attacks in one
week. It may be that once the internet searches are made two weeks before an attack, the time
for radical propaganda is over and the time for preparing for an attack has come. No Granger
causality was found between internet searches for “Taghut” and number of casualties.
Similar results to those in Table 4 were found when using the internet search term
“kufar,” which refers to infidels and the jihadist movement. Internet searches for “kufar”
preceded the number killed in Islamist terrorist attacks, while the number killed in Islamist
terrorist attacks did not Granger cause internet searches for “kufar.” No Granger causality was
found between internet searches for “kufar” and number of casualties, or between internet
searches for “kufar” and number of attacks.
The above results are summarized in Table 5. In Table 5, in the row for “join Jihad,” a
“yes” appears under the heading “Number of Attacks/week” and under the subheading
“Searches→Attacks.” This means that the number of internet searches for the term “join Jihad”
preceded or Granger caused a number of radical Islamic terrorist attacks. A “no” under the
subheading “Attacks→Searches,” means that a number of Islamist terrorist attacks did not
precede or Granger cause internet searches for “join Jihad.” The results in Table 5 indicate that
internet searches for “taghut” also preceded or Granger caused a number of Islamist terrorist
attacks and number of internet searches for “kufar” preceded or Granger caused the number
killed in Islamist terrorist attacks. Furthermore, the number of casualties from Islamist terrorist
attacks, preceded or Granger caused a number of internet searches for “join Jihad” and “ISIS
websites.”
Table 5. Summary of Granger Caus ality Tes ts
Internet Number of Atta cks/week Number of Casu alties /week
Search term
Searches →Attacks
Attacks →Searches
Searches→Casualties Casualties →Searches
join Ji had yes no no yes
ISIS websi tes no no no yes
taghut yes no no no
kufar no no yes* no
*indica tes that internet sea rch precede d number kille d rathe r than number of casu alties (number kil led and injured)
Carl E. Enomoto, Kiana Douglas
ISSN 2071-789X
RECENT ISSUES IN SOCIOLOGICAL RESEARCH
Economics & Sociology, Vol. 12, No. 1, 2019
244
The remaining Google search phrases: “join ISIS,” “ISIS chatrooms,” “ISIS
recruitment,” “ISIS twitter,” “khlifa,” “jizya,” and “taqiyya” (in its Arabic form), were found
to not affect or be affected by number of terrorist attacks or number of casualties from Islamist
terrorist attacks. Finally, as a test for the validity of the VAR model used in this paper, two
random search phrases: “sandwich” and “keys,” were used to see if they affected or were
affected by a number of Islamist terrorist attacks and number of casualties from Islamist terrorist
attacks. It was found that there was no Granger causality between either “sandwich” or “keys”
and the number of attacks or number of causalities. Furthermore, using the search term “peace”,
there was no Granger causality between “peace” and number of attacks or total number of
casualties. There was also no Granger causality between searches for “harmony” and number
of attacks or total number of casualties. There was no Granger causality between the search
term “love” and total number of attacks. Searches for the term “love” did not Granger cause
casualties. However, at the 6% level of significance, number of casualties did Granger cause
internet searches for the term “love.” Perhaps the world is reaching out and in sympathy for
those hurt by terrorism.
Conclusion
Much of the existing literature on social media and terrorism focuses on how terrorist
groups are using Facebook, Twitter, YouTube, chat rooms, and websites, to spread their
propaganda and recruit new members. What has not been explored is how internet searches for
specific phrases such as “join Jihad,” “ISIS chat rooms, “taghut,” “kufar,” or others, have
affected or are affected by Islamist terrorist attacks. This paper addresses these issues. It was
found that 1) internet searches for “join Jihad,” preceded Islamist terrorist attacks by three
weeks. 2) Casualties from Islamist terrorist attacks preceded internet searches for “join Jihad”
and “ISIS websites.” 3) Internet searches for “taghut” preceded Islamist terrorist attacks by one
week and two weeks. 4) Internet searches for “kufar” preceded the number killed by Islamist
terrorist attacks. These results indicate that internet searches for certain phrases are related to
Islamist terrorist attacks.
One possible explanation for the lag length is that there are two groups of individuals:
those not yet radicalized and those already radicalized. Approximately three weeks before the
attack some eternal political/social condition results in (a) leading those not yet radicalized to
become more interested and start searching on the topic while simultaneous (b) encouraging
those already radicalized to begin planning an attack for 3 weeks out. This second already-
radicalized group may also be further encouraged to plan an attack because of the increased
search volume, increased contacts from potential recruits and increased support in the online
community. Thus, those creating high search volume are not necessarily the same people as
those who will conduct the attack, but either or both groups are influenced by outside conditions
and the searching group may be encouraging the attacking group. These steps linking internet
searches for Islamist propaganda and terrorist attacks are outlined in Figure 1.
The United Nations Office on Drugs and Crime (UNODC, 2012) came up with several
strategies for countering the use of the internet for terrorist activities. As an example, internet
activity used by terrorists can be monitored, and information can be gathered by the authorities
to identify and possibly stop terrorist attacks. Chat rooms and online discussion groups can be
used to provide opposing opinions to those of the terrorists. If more facts are presented,
radicalization may not take place. Furthermore, the internet can be used to identify those
individuals specifically involved in recruiting. Finally, the internet can be closely monitored for
spikes in specific search phrases. Appropriate action can then be taken when possible.
Carl E. Enomoto, Kiana Douglas
ISSN 2071-789X
RECENT ISSUES IN SOCIOLOGICAL RESEARCH
Economics & Sociology, Vol. 12, No. 1, 2019
245
___________________________________________________________________________
____________________
___________________________________________________________________________
Figure 1. From internet searches for Islamist propaganda to an attack
There are several directions for future researchers to take when studying the relationship
between social media and terrorism. First, it was noted in the literature review that the terrorists
are using multiple languages on websites in disseminating their propaganda. In future studies,
Google search volume indices may be collected for searches performed in various languages.
Second, the reasons for those engaging in online searches for terrorist propaganda need to be
studied. What are individuals looking for from these websites? Do these individuals see abuses
and injustices that must be addressed? Third, Google Trends provides a number of internet
searches for specific phrases for different regions, not just worldwide searches. Thus future
studies may be able to find out which regions are responsible for the majority of internet
searches for certain phrases. Answers to these questions may help establish a better
understanding of the link between radicalization and terrorism and ways to break it.
References
BBC News (2016). Orlando nightclub shooting: How the attack unfolded, Retrieved on
February 16, 2018, from http://www.bbc.com/news/world-us-canada-36511778.
Beckmann, K. B., Dewenter, R., & Thomas, T. (2017). Can news draw blood? The impact of
media coverage on the number and severity of terror attacks. Peace Economics, Peace
Science and Public Policy, 23(1).1-16.
Group A
Group B
(2a) Those not yet
radicalized start
searching radical terms.
(2b) Those already
radicalized start planning
an attack for (avg of)
three weeks out.
(3) Commits attack
(1) Some external
political/social shift or
event encourages:
B is further encouraged by
A’s increased searches/
contact
Carl E. Enomoto, Kiana Douglas
ISSN 2071-789X
RECENT ISSUES IN SOCIOLOGICAL RESEARCH
Economics & Sociology, Vol. 12, No. 1, 2019
246
Behr, I.V., Reding, A., Edwards, C., & Gribbon, L. (2013). Radicalisation in the digital era:
The use of the internet in 15 cases of terrorism and extremism. RAND Europe.
Bertram, S. & Ellison, K. (2014). Sub-Saharan African terrorist group’s use of the internet.
Journal of Terrorism Research, 5(1), 5-26.
Choi, H. & Varian, H. (2012) Predicting the present with google trends. Economic Record,
88(S1):2-9.
CNN (2018) ISIS goes global: 143 attacks in 29 countries have killed 2,043. Retrieved on
January 10, 2019, from https://www.cnn.com/2015/12/17/world/mapping-isis-attacks-
around-the-world/index.html.
Committee on Homeland Security and Governmental Affairs United States Senate (2016). ISIS
online: Countering terrorist radicalization and recruitment on the internet and social media.
U.S. Government Publishing Office. Washington, D.C.
Constitutional Rights Foundation (2018). Islamist terrorism from 1945 to the rise of ISIS.
Retrieved on February 16, 2018, from http://www.crf-usa.org/america-responds-to-
terrorism/islamist-terrorism-from-1945-to--the-death-of-osama-bin-laden.html.
Conway, M. (2017). Determining the role of the internet in violent extremism and terrorism:
Six suggestions for progressing research. Studies in Conflict & Terrorism, 40(1), 77-98.
Conway, M. (2006). Terrorism and the internet: New media – new threat? Parliamentary
Affairs, 59(2), 283-298.
Conway, M. (2012). From al-Zarqawi to al-Awlaki: The emergence of the internet as a new
Form of violent radical milieu, Retrieved on February 16, 2018, from
http://www.isodarco.it/courses/andalo12/doc/Zarqawi%20to%20Awlaki_V2.pdf.
Enomoto, C., Noor, S. & Widner, B. (2017). Is social media to blame for the sharp rise in stds?
Social Sciences, 6(3), 78.
Fernandez, A.M. (2015). Islamic words that kill. MEMRI: The Middle East Media Research
Institute. Retrieved on January 11, 2019, from https://www.memri.org/reports/islamic-
words-kill.
Gardner, F. (2014). Jihadist groups around the world. BBC News. Retrieved on February 16,
2018 from: http://www.bbc.com/news/world-middle-east-27930414.
Hassan, M. H. (2018). Refuting Islamic State (IS) jihad propaganda with the story of Uwais Al-
Qarni. Counter Terrorist Trends and Analyses, 10(10), 12-16.
Holden, M., & Hosenball, M. (2017). Investigators suspect UK killer was lone wolf radicalized
on internet. Retrieved on January 11, 2019 from: https://www.reuters.com/article/us-
britain-security-masood/investigators-suspect-uk-killer-was-lone-wolf-radicalized-on-
internet-idUSKBN171242.
Homeland Security Institute (2009). The internet as a terrorist tool for recruitment and
radicalisation of youth, Retrieved on February 16, 2018 from:
http://barakaconsult.com/uploads/reports%20on%20internet_radicalization.pdf.
Parekh, D.; Amarasingam, A.; Dawson, L. & D. Ruths (2018). Studying jihadists on social
media: A critique of data collection methodologies. Perspectives on Terrorism, 12(3), 3-
21.
Pfeifer, C. (2012). Causalities and casualties: Media attention and terrorism, 1970-2010,
Diskussionspapier, Helmut-Schmidt-Universität, Fächergruppe Volkswirtschaftslehre127.
Ricketts, C.F. & C.G. Silva (2017). An analysis of morbidity and mortality using Google trends.
Journal of Human Behavior in Social Environment, 27(6), 1-12.
Sheth, S. (2017). FBI: Fort Lauderdale shooting suspect says he carried out the attack on behalf
of ISIS. Retrieved on January 22, 2018, from http://www.businessinsider.com/fbi-fort-
lauderdale-shooting-suspect-says-he-carried-out-the-attack-on-behalf-of-isis-2017-1.
Soliev, Nodirbek. (2018). Terrorist threat to the 2018 world cup in Russia. Counter Terrorist
Trends and Analyses, 10(6), 16-21.
Carl E. Enomoto, Kiana Douglas
ISSN 2071-789X
RECENT ISSUES IN SOCIOLOGICAL RESEARCH
Economics & Sociology, Vol. 12, No. 1, 2019
247
The Religion of Peace (Website), Retrieved on February 8, 2018, from:
https://www.thereligionofpeace.com/attacks/attacks.aspx?Yr=2017.
Thompson, R. (2011). Radicalization and the Use of Social Media’. Journal of Strategic
Security, 4(4), 167–190.
Ullah, H.K. (2017). Digital world war: Islamists, extremists, and the fight for cyber supremacy.
Yale University Press. New Haven and London.
United Nations Office on Drugs and Crime (UNODC) (2012). The use of the Internet for
terrorist purposes. Retrieved on January 18, 2018 from:
http://www.unodc.org/documents/frontpage/Use_of_Internet_for_Terrorist_Purposes.pdf.
Vicens-Feliberty, M.A., & Ricketts, C.F. (2016). An analysis of Puerto Rican interest to migrate
to the United States using Google Trends. The Journal of Developing Areas, 50(2), 411-
430.
Veilleux-Lepage, Y. (2016). Paradigmatic shifts in jihadism in cyberspace: The emerging role
of unaffiliated sympathizers in Islamic State’s social media strategy. Journal of Terrorism
Research, 7(1), 36-51.
Vosen, S., & Schmidt, T. (2011). Forecasting private consumption: Survey-based indicators vs.
Google Trends. Journal of Forecasting, 30(6), 565-578.
Weimann, G. (2004). www.terror.net: How Modern Terrorism Uses the Internet. United States
Institute of Peace. Retrieved on January 18, 2018, from:
https://www.usip.org/sites/default/files/sr116.pdf.
Winter, C. (2016). Special report: An integrated approach to Islamic State recruitment.
Australian Strategic Policy Institute: Counter-Terrorism Policy Centre.
Wu, L., & Brynjolfsson, E. (2013). The future of prediction: How Google searches foreshadow
housing prices and sales. Available from: Economics of Digitization, University of
Chicago Press.
Zelin, A. (2013). The State of Global Jihad Online: A Qualitative, Quantitative, and Cross -
Lingual Analysis. New America Foundation. Washington, D.C.
1389 Blog-Counterjihad (2013). Four Arabic words every infidel must know. Retrieved on
January 25, 2018 from: http://1389blog.com/2013/06/23/four-arabic-words-every-infidel-
must-know/.
