Online interest in radical Islam and terrorist attacks

Nicolini, Marcella; Sabatini, Fabio; Fantazzini, Dean Article Online interest in radical Islam and terrorist attacks Peace Economics, Peace Science and Public Policy (PEPS) Provided in Cooperation with: De Gruyter Brill Suggested Citation: Nicolini, Marcella; Sabatini, Fabio; Fantazzini, Dean (2025) : Online interest in radical Islam and terrorist attacks, Peace Economics, Peace Science and Public Policy (PEPS), ISSN 1554-8597, De Gruyter, Berlin, Vol. 31, Iss. 2, pp. 161-192, https://doi.org/10.1515/peps-2024-0054 This Version is available at: https://hdl.handle.net/10419/333342 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. https://creativecommons.org/licenses/by/4.0/ Marcella Nicolini*, Fabio Sabatini and Dean Fantazzini Online Interest in Radical Islam and Terrorist Attacks https://doi.org/10.1515/peps-2024-0054 Received December 6, 2024; accepted February 13, 2025; published online March 6, 2025 Abstract: This study investigates the link between terrorist attacks and public interest in radical Islam and violent extremism, using monthly data from 150 countries between 2004 and 2015. Employing a dynamic common correlated effects (DCCE) estimator to account for potential cross-country correlations, our analysis reveals that attacks carried out in the name of Islam significantly drive online searches for sensitive keywords. Specifically, terms suggesting violent actions, such as “beheadings,”and explicitly jihad-related terms show stronger correlations, indicating heightened interest in the terrorists’actions and messages. Our findings suggest that terrorist attacks not only occur in areas where public attention to terrorism is already heightened but also intensify interest in violent extremism within those regions. This amplification may contribute to terrorists’objectives by increasing the public visibility they seek. Keywords: islamist extremism; terrorism; Google searches; dynamic common correlated effects estimator (DCCE) JEL Classification: C33; F50; H56 We thank Elena Esposito, Domenico Giannone, Robert Hefner, Michael Jetter, Daniele Massacci, Jeremy Menchik, Daniele Paserman, Paolo Pinotti, Andrea Pozzi and seminar participants at the META 2018 workshop in Nizhny Novgorod for useful comments and suggestions. Luca Biglieri, Emma Cambieri, Antonella Cozzolino, Simona Fabris, Emanuela Padellaro, Vittoria Senzalari, and Vittorio Spinelli provided excellent research assistance. Marcella Nicolini thanks the Institute for the Study of Muslim Societies & Civilizations at Boston University for hospitality. All errors and omissions are our own. *Corresponding author: Marcella Nicolini, Department of Economics and Management, University of Pavia, Via San Felice 5 –27100 Pavia (PV), Italy, E-mail: [email protected]. https://orcid.org/00000001-9034-2562 Fabio Sabatini, Sapienza University of Rome, Rome, Italy; and IZA, Bonn, Germany, E-mail: [email protected]. https://orcid.org/0000-0002-1228-3322 Dean Fantazzini, Moscow School of Economics, Moscow State University, Moscow, Russia, E-mail: [email protected]. https://orcid.org/0000-0002-1481-3382 Peace Econ. Peace Sci. Pub. Pol. 2025; 31(2): 161–192 Open Access. © 2025 the author(s), published by De Gruyter. This work is licensed under the Creative Commons Attribution 4.0 International License. 1 Introduction Terrorists strive to capture public attention to recruit new followers and disseminate their message, ultimately bolstering their capacity to inflict harm (Nelson and Scott 1992). Understanding how public interest in violent ideologies relates to terrorist attacks is crucial to preventing the escalation of radicalization into violence. However, gauging the popularity of terrorists and their ideologies is challenging for multiple reasons. News coverage of terrorism does not reliably indicate the public’s active interest or opinions, which can be heavily shaped by biased media sources (Gentzkow and Shapiro 2004). Furthermore, curiosity about and support for extremist ideologies are often understated and considered taboo, complicating efforts to measure these sentiments through surveys or experimental methods (Stephens-Davidowitz 2014). In this paper, we examine the relationship between public interest in violent extremism and terrorist attacks, with a focus on Islamist terrorism. This specific focus allows us to identify a range of distinctive keywords associated with radical Islam. Building on the classification of keywords developed by Ahmed and Lloyd George (2016), we selected 47 sensitive online queries and gathered their monthly Google search volumes globally from 2004 to 2015. These keywords include terms indicative of curiosity about radical Islam, expressions of sympathy towards terrorist ideologies, and phrases explicitly related to violent jihadist actions, such as the killing of infidels. Data on terrorist attacks were sourced from the Global Terrorism Database (GTD). Since the GTD does not specify the origins of the attacks, we undertook additional efforts to document the backgrounds and motivations of the terrorist groups involved. Our comprehensive database comprises detailed information on 21,710 attacks carried out worldwide by self-proclaimed Islamist groups in the name of Islam. We provide evidence of a positive correlation between online searches for sensitive keywords and Islamist attacks. The association is robust to controlling for cross-country dependence, which is especially crucial in this context as a globally common shock, such as the 9/11 attack on the Twin Towers, might spillover and have varying impacts across different countries. Upon looking into specific keywords, we find that attacks are more significantly and sizably associated with “extreme” queries, particularly those relating to violence, such as “beheadings”, or those explicitly related to the jihad, such as “Abdullah Azzam”, a Sunni scholar and founding member of Al Qaeda, also known as the “Father of Global Jihad”(Edwards 2017). This result survives several robustness checks and placebo tests. When we extend the analysis to incorporate spatial lags, our results indicate that Google searches in neighboring countries are positively and significantly correlated with 162 M. Nicolini et al. attacks. This suggests that when an Islamist attack occurs in a given country, individuals in surrounding countries tend to increase their searches for sensitive keywords. Interestingly, while the spatial lag for extreme keywords is not statistically significant in the global sample, it becomes both statistically significant and larger in magnitude when restricting the analysis to countries that have experienced an Islamist attack. Overall, these findings highlight the role of geographic proximity in shaping online search behavior, particularly in regions with a history of Islamist terrorism. We also expanded our analysis by including temporal lags of Google searches: the coefficient for Google searches in the previous month is positive and significant, suggesting that past search activity is positively associated with subsequent Islamist attacks. While we cannot claim causality in interpreting our results, the empirical analysis contributes to the understanding of the relationship between terrorism and the active interest in violent extremism in several ways. Previous studies suggest that the media coverage of terrorist attacks feeds violent extremism, as the attention of the public encourages terrorists to plan new attacks. Our approach offers a way to measure the public’s attention to extremist ideas and documents its relationship with terrorists’missions. The link between the occurrence of the attacks and searches for violent and extreme keywords is striking. This suggests that terrorists choose to attack in places where there is a high level of interest and consideration for their claims and operations. In turn, these attacks nurture public interest in violent extremism in those regions potentially aiding terrorists in garnering the public attention they seek. This paper bridges two strands of literature. The first strand deals with the relationship between media and terrorism (El Ouadghiri and Peillex 2018; Gentzkow and Shapiro 2004; Jetter 2017, 2019; Rohner and Frey 2007). Our study builds on this body of research by linking it to the broader empirical literature on the determinants of terrorism, aiming to uncover the underlying drivers of terrorist activities. Previous research has explored the relationship between terrorism and underdevelopment (Feridun and Sezgin 2008), poverty (Caruso and Schneider 2013), education (Bassetti, Caruso, and Schneider 2018), declining social capital (Helfstein 2014), natural resource rents (Ajide and Alimi 2020), fiscal decentralization (Arzaghi and Gaibulloev 2024), immigration (Bove and Böhmelt 2016), and political exclusion (Choi and Piazza 2016). Our work contributes to this body of literature by using Google searches to address the challenge of detecting the popularity and celebrity status of terrorists. Moreover, we show that the public’s active attention to terrorists’claims, such as intentional searches of specific keywords, rather than passive news consumption, is strongly correlated with terrorist attacks. This result aligns with recent evidence that terrorist attacks provoke enduring public reactions (Bove, Efthyvoulou, and Pickard 2024). These responses can manifest not only as heightened interest Online Interest in Radical Islam and Terrorism 163 in terrorist messaging and fear of further attacks but also as reinforced national identity and increased support for national and supranational institutions (Efthyvoulou, Pickard, and Bove 2024). Additionally, our finding that the severity of attacks, measured by the number of casualties and injuries, generates stronger public interest in terrorist activities further corroborates the insights from Bove, Efthyvoulou, and Pickard 2024; Efthyvoulou, Pickard, and Bove 2024, who document that attacks with a higher number of victims elicit stronger and more prolonged public reactions. The second strand encompasses studies using online searches to track hidden or difficult to measure social phenomena such as racism (Doleac and Stein 2013; Giulietti, Tonin, and Vlassopoulos 2019; Stephens-Davidowitz 2014), contraceptive use and abortion (Kearney and Levine 2015), religiosity (Bentzen 2019), pornography (MacInnis and Hodson 2015), and prosocial attitudes (Guriev and Melnikov 2016). We connect to this literature by introducing the use of online searches in the study of radicalization and terrorism. Moreover, we contribute by adopting an econometric approach that accounts for cross-sectional dependence, which addresses some of the criticisms of using Google searches in research. The rest of the paper proceeds as follows: in the next section, we describe our data on online searches of sensitive keywords and terrorist episodes. We then explain our econometric approach in Section 3. Section 4 presents the analysis of the relationship between the online interest in Islamist extremism and terrorist attacks and the robustness checks. In Section 5, we briefly summarize and discuss our main findings. Section 6 concludes. 2 Data 2.1 Measuring Interest in Radical Islam and Violent Extremism As of January 2019, almost 4.4 billion people were active Internet users, roughly corresponding to 59 % of the global population. 1 With around 90 % market share worldwide between 2010 and 2015, Google is by far the world’s leading search engine. 2 Starting from January 2004, the company has made available information on the queries that people search on the engine with a tool called Google Trends (hereafter GT). 3 The empirical evidence shows that GT consistently and strongly 1See www.statista.com. 2Data sourced from www.statista.com/statistics/216573/worldwide-market-share-of-search-engines/ 3For any specific keyword, GT provide an index of the volume of queries entered into Google in a given geographic area. The query index is the ratio between the total query volume for the keyword in question over a certain period and the total number of queries in the same period. At the country level, Google provides search trends with a monthly frequency. In any selected period, the maximum 164 M. Nicolini et al. detect many phenomena for which conventional data are publicly available with a delay, or at a lower frequency. For example, queries consistently detect economic behaviors such as job searches (Baker and Fradkin 2017), the issuance of bad checks (Eksi, Gurdal, and Orman 2017), deposit reallocation (Fecht, Thum, and Weber 2019), and the use of Uber services (Berger, Chen, and Frey 2018), to name a few. However, it is in the detection of usually undercover interests and behaviors that Google searches play a crucial role. Online searches are generally conducted in private, elude social control, and allow people to express socially taboo, anti-social, or even criminal interests and attitudes that they would hardly reveal in public. For example, MacInnis and Hodson (2015) use GT to study how interest in pornography relates to religiosity and conservatism across American states. Google-based measures of racism strongly predict black mortality (Chae et al. 2015), black homeownership (Harris and Yelowitz 2018), access to public services (Giulietti, Tonin, and Vlassopoulos 2019), and the performance of Barak Obama in 2008 and 2012 U.S. presidential elections (Stephens-Davidowitz 2014). Google data have also proved helpful in detecting or predicting contraceptive use and abortion (Kearney and Levine 2015), suicides (Ma-Kellams et al. 2015), drug abuse (Perdue, Hawdon, and Thames 2018), and the propagation of sexually transmitted diseases such as syphilis (Young et al. 2018) and HIV (Young and Zhang 2018). Along these lines, the interest in anti-system groups is a typically hidden social pattern that Google searches can help uncover. Research into the relationship between the Internet and terrorism has so far focused on the role of social networking sites (Aly et al. 2016; Weimann 2015). Tracking jihadi propaganda on social media is useful for detecting extremists’communication and enacting counter-terrorism measures (Zeitzoff2017). However, radicalized individuals’activities on platforms like Facebook and Twitter do not necessarily capture the attention of the general public in Islamist terrorism. People likely radicalize as a result of the spreading of feelings sympathetic to extremists across society, a phenomenon that eludes detection through social media (Ahmed and Lloyd George 2016). Search engines, by contrast, can help to detect the active attention of the public to extremist ideas (Jetter 2019). In our empirical analysis, we use a list of keywords that, according to the Centre on Religion & Geopolitics (CRG), are likely to lead to extremist content if entered into a search engine. The CRG first developed a set of words and phrases that might be used in research on radical Islam or in deliberate attempts to access violent extremist query share is normalized to be 100, and the query share at the initial date is normalized to be zero (Choi and Varian 2012). By aggregating millions of searches with a monthly frequency over a long period across a large sample of countries, GT allow constructing reliable indicators of a broad array of behavioral variables. Online Interest in Radical Islam and Terrorism 165 content. The list was then employed as a basis to ascertain further potentially associated keywords through specific tools such as Google Keyword Planner, Google Suggest phrases, and Google-related searches. The final list comprised 143 keywords. Based on the search frequency data and a qualitative assessment, the CRG then selected a set of 47 keywords. The set is the sum of two subgroups. The first includes popular keywords that were searched on average at a minimum frequency of 500 times per month. The second group includes keywords explicitly related to violent extremism. Ahmed and Lloyd George (2016) conducted a search engine result page (SERP) analysis of the keywords to ascertain the extent to which they return “extreme”content. Though being non-violent and mostly related to political Islam, keywords belonging to the first group return extremist content in a remarkable share of cases, anticipating the risk of unintentional exposure to jihadi propaganda. The second set comprises keywords that, though being less popular, carry a higher potential for risk. Overall, 44 % of the material was explicitly violent. In contrast, counter-narrative content outperformed extremist material in only 11 % of the results generated. 4 Several sites aiming to present legitimate Islamic culture were also found to host extremist material. For example, Kalamullah.com and WorldofIslam.info –which can be retrieved by entering our selection of keywords into a search engine –contain resources for ordinary Muslims and those interested in Islam, such as online versions of the Quran, collections of Hadith literature, and a variety of Islamic books. However, these sites also provide jihadi manuals and audio lectures by known jihadis. Overall, the Google searches for the chosen set of keywords returned a vast array of fundamentalist material spanning from non-violent extremism to explicitly violent jihadism. We use the GT for the 47 keywords to measure the active interest in radical Islam in 149 countries over the period 2004–2015. Figure 1 shows on a map the normalized volume of searches for the keywords worldwide. While Google is the predominant search engine in most countries, there are some notable exceptions, e.g., China. This could go in the direction of weakening the link we expect to detect between online searches, measured through GT, and terrorist attacks. 4Ahmed and Lloyd George (2016) looked at the first two pages of results, as 92 % of Google trafficis on page one, with page two only receiving 5 % of the traffic and page three 1 % (Chitika Insights 2013). The authors split the extreme content into three categories: violent, non-violent, and political Islamist. Websites were categorized as violent if they contained either images of graphic violence or exhortations to violence. The political Islamist category included content expressing a specific affinity to a particular Islamist group. The non-extreme content regarded counter-narrative or “neutral”websites, such as news websites or portals targeting researchers and students. Websites deemed to be extreme but non-violent were those expressing anti-Semitic, homophobic, racist, or sectarian views without a depiction of, or incitement to, violence. 166 M. Nicolini et al. A variety of people may be searching for keywords that return extremist content, from academics to journalists. We do not claim to be measuring the online search activity of perspective terrorists. Instead, we assume that Google searches measure the interest of the public in radical Islam and violent extremism. Previous studies employing Google searches to assess the hypothetical outcomes of social phenomena all share the assumption that trends for specific keywords can proxy the incidence of hidden social patterns in the population (e.g., Stephens-Davidowitz 2014; Kearney and Levine 2015; Guriev and Melnikov 2016). Table 1 reports the list of keywords. We provide a glossary with a brief description of their meaning in Appendix A.1. 2.2 Terrorist Attacks Data on terrorist attacks are drawn from the Global Terrorism Database (GTD) maintained by the National Consortium for the Study of Terrorism and Responses to Terrorism (START) at the University of Maryland. The database collects information on the attacks perpetrated around the world from 1970 through 2018. Including more than 190,000 cases over the whole period, and 80,694 over our period of analysis (2004–2015), the GTD is the most comprehensive unclassified database on terrorist attacks. 5 Differently from other databases on terrorism, the GTD includes systematic data on domestic as well as transnational and international terrorist incidents. For each attack, the database provides information on several variables,amongwhichthedateandlocation, Figure 1: Google trends monthly averages for the mean of the 47 search terms. Notes: The intervals plotted correspond to the quartiles of the distribution of the monthly GTs for the average of the 47 keywords. 5See https://www.start.umd.edu/gtd/about/ and LaFree and Dugan (2007) for more information on the database. Online Interest in Radical Islam and Terrorism 167 the weapons used, the nature of the target, the number of casualties (distinguishing victims into deaths and wounded), and, when available, the group or individual responsible for the attack. We make an additional effort to collect further information on the 999 terrorist groups identified in the GTD that have been active in the 2004–2015 period. We classify groups according to their ideology or intent. Inspecting each terrorist event allows us to enrich the GTD database by further categorizing the attacks according to the ideology, religious background, and motivation of the perpetrating group. Out of 80,694 events, it is possible to identify the group’s background in 34,907 cases (43.3 % of cases). Religious fundamentalist organizations have caused two-thirds of these events, and 97.8 % of the attacks with a religious identity have been perpetrated by self-proclaiming Islamist groups, corresponding to 21,710 registered attacks. We Table :List of the  keywords. Most popular Most extreme Apostasy Abdullah Azzam Apostate Amaq Agency Apostates Apostate Islam Ayman al-Zawahiri Apostates in Islam Caliphate Beheadings Crusader Crusader Army Crusaders Crusaders against Islam Dabiq Dabiq Pdf Dabiq Magazine How to do Jihad Ibn Taymiyyah Ibn Taymiyyah Jihad Islamic State Inspire Magazine Jihad Jewish Coalition Jihad meaning Jihad for Ummah Kafir Jihad in the Quran Khalifah Meaning Khalifah Khilafah Khilafah Syria Kuffar Killing Apostates Martyr Killing Infidels Martyrdom in Islam Killing Kuffar Martyrs Mujahid Mujahideen Preparing for Jihad Shahada Rafidah Suicide vest Soldiers of the Caliphate Taghut 168 M. Nicolini et al. corresponds to a change in the Islamist attacks equivalent to approximately 12.4 % of its mean, indicating a moderate effect size in relative terms. This number increases to 14.3 % in the subsample of countries hit by Islamist attacks. 9 We present our main results using alternative measures of the volume of Google searches in Table 3. 10 Column 1 in Panel A reports our preferred specification, with the mean value of the 47 keywords. In column 2, we present the results with the first factor of a principal component analysis (PCA) of the 47 different keywords. We observe a positive and statistically significant coefficient for GT. The lagged value of the dependent variable, IA, also displays a positive and significant coefficient, as expected: while investigating the relationship between Islamist attacks and Google searches, we cannot neglect the existence of attacks in the previous period. To explore the relationship between Google searches and the attacks in greater depth, we differentiate between keywords. Our goal here is to distinctly focus on terms that may indicate a genuine interest in radical ideology instead of merely reflecting a heightened attention to terrorism news and events. First, we adopt the classification of “popular”and “extreme”keywords as proposed by Ahmed and Lloyd George (2016), which is reported in Table 1. We compute the average volume of Google searches for the two groups. Both “popular”and “extreme”keywords significantly and positively relate to the attacks, with extreme keywords displaying a larger coefficient. Panel B of Table 3 shows the results on the subsample of 63 countries that experienced at least one Islamist attack in the period under consideration. Across different specifications, results reported in Panel B confirm those on the full sample: the sign and statistical significance of coefficients do not change, but we notice a slight increase in their size. As countries that did not face attacks perpetrated in the name of Islam are now excluded from the sample, the specification in Panel B likely captures the relationship between the interest of the public to Islamist extremism and Islamist attacks more neatly. Not surprisingly, the estimated coefficients are slightly larger in this subsample. 9In the sample of countries affected by any terrorist attack (Panel A), the effect of a one-standarddeviation increase in Google searches is given by β×σ GT , where β= 0.0485 and σ GT = 2.607, yielding 0. 0485 ×2.607 = 0.1264. Expressed as a percentage of the dependent variable’s mean (IA =1.016), this corresponds to 12.4 %. For the subsample of countries affected by Islamist attacks (Panel B), the effect is β= 0.1079 and σ GT = 3.175, resulting in 0.1079 ×3.175 = 0.3425. Relative to the mean of the dependent variable (IA =2.402), this corresponds to 14.3 %. 10 Diagnostic tests support the choice of the DCCE estimator. The combined evidence from the Fisher-type and the IPS panel unit root tests confirms that the series are stationary. The Pesaran (2004) test for cross-sectional dependence suggests significant cross-sectional correlations for all our variables. The Pesaran (2007) panel unit root test in the presence of cross-sectional dependence also rejects the null hypothesis of homogenous non-stationarity. Online Interest in Radical Islam and Terrorism 175 To better understand the positive correlation between Google searches and terrorist attacks perpetrated by self-proclaiming Islamist organizations, we distinguish the keywords into six groups according to their meaning. The first two categories include keywords referring to ISIS and Al Qaeda, respectively. In the third group, we focus on apostasy: in radical Islam, apostasy is not necessarily understood as the personal choice of renouncing a religious belief. Instead, it is often considered a public act of political secession from the Muslim community. This view makes apostates one of the favorite targets of jihadi propaganda (Torres Soriano 2010). The fourth group comprises keywords evoking violent actions, such as “beheadings”. The fifth and sixth categories contain terms related to religious doctrines and the history of political Islam, respectively. 11 For each group, we compute the average volume of Google searches and assess its relationship with terrorist attacks perpetrated in the name of Islam. Panel A of Table 4 reports results in all countries that faced an attack. The sets of keywords related to the two most prominent Islamist terrorist groups, Al Qaeda and ISIS, are not significant. Queries referring to the concept of apostasy in Islam, such as “kafir”and “kuffar”, significantly and positively correlate with the attacks. The index of violent keywords, including queries like “beheadings”and “killing infidels”, significantly and positively relates to the attacks. The aggregations of keywords comprising the doctrinal and historical terms are instead not statistically significant. 12 Panel B of Table 4 reports the results on the subsample of countries affected by Islamist attacks in the period under consideration. Again we observe an increase in the size of the estimated coefficients. We extend our analysis by incorporating spatial lags of our variables of interest. Specifically, we enhance our model by including spatial lags of the Islamist attacks variable and Google searches to examine how terrorist attacks and search activity in neighboring countries influence each focal country. Using a contiguity matrix, we construct spatially lagged versions of both the dependent and independent variables and integrate these additional regressors into the DCCE estimation. The results are presented in Table 5. This analysis reveals two key insights. First, we find no statistically significant relationship between Islamist attacks in neighboring countries and the occurrence of an attack in a given focal country. In other words, the likelihood of experiencing an Islamist attack in country idoes not appear to be influenced by attacks in contiguous countries. 11 Table A.2 in the Appendix A.2 defines the six categories in detail. 12 The doctrinal terms are not specific to Islam and could generically refer also to other Abrahamic religions, such as Christianity. Results do not change if we extend the Religion index by including “Shahada”, a doctrinal term specific to Islam. 176 M. Nicolini et al. However, this null result contrasts sharply with the effect of Google search activity in neighboring countries. The spatial lag of Google searches is positive and statistically significant, suggesting that when an Islamist attack occurs in country i, individuals in surrounding countries tend to increase their searches for sensitive keywords. More relevant is the finding that if we consider popular and extreme keywords separately, the spatial lag for extreme words is not statistically significant in the global sample, but it becomes statistically significant and larger in magnitude if we focus on the subsample of countries that experienced an Islamist attack. Overall, these findings indicate that geographic proximity plays a crucial role in shaping online search behavior, particularly in regions where Islamist attacks are more prevalent. Another way in which we can expand our analysis is by including temporal lags of the Google searches. The results reported in Table 6 show that including Google searches at time t−1 does not substantially alter our main findings regarding the contemporaneous Google searches and the lag of the dependent variable. The coefficient for Google searches in the previous month is positive and significant, Table :Alternative aggregations of GTs. Panel A ()()()()()() Countries hit by any terrorist attack ISIS Al Qaeda Infidels Violent Religion History GT t −. . .*.*** . . (.)(.)(.)(.)(.)(.) IA t− .*** .*** .*** .*** .*** .*** (.)(.)(.)(.)(.)(.) R. . . . . . # Observations , , , , , , # Countries       Panel A ()()()()()() Countries hit by Islamist attacks ISIS Al Qaeda Infidels Violent Religion History GT −. . .*.** . . (.)(.)(.)(.)(.)(.) IA t− .*** .*** .*** .*** .*** .*** (.)(.)(.)(.)(.)(.) R. . . . . . # Observations , , , , , , # Countries       DCCE estimates. The dependent variable is the number of Islamist attacks. For the definition of the different groupings, see Table A.. Standard errors in parentheses, ***, **, and * denote, respectively, %, %, and  % significance levels. Online Interest in Radical Islam and Terrorism 177 suggesting that past search activity is positively associated with subsequent Islamist attacks. However, when we differentiate between popular and extreme keywords, we do not find any statistically significant effects. Notably, the significance of contemporaneous Google searches remains robust, particularly for extreme keywords. Finally, we explore how Google searches for our sensitive keywords relate to indicators of the outcomes of terrorist missions. To this end, we repeat the analysis using the number of killed, wounded, and overall affected people as dependent Table :Introducing spatial lags. Panel A ()()() Countries hit by any terrorist attack Mean Popular Extreme GT t .*** .** .*** (.)(.)(.) GT t,s− .*.*. (.)(.)(.) IA t− .*** .*** .*** (.)(.)(.) IA t,s− . . . (.)(.)(.) R. . . #Observations , , , # Countries    Panel B ()()() Countries hit by Islamist attacks Mean Popular Extreme GT t .*** .** .*** (.)(.)(.) GT t,s− .** .** .** (.)(.)(.) IA t− .*** .*** .*** (.)(.)(.) IA t,s− . . . (.)(.)(.) R. . . #Observations , , , # Countries    DCCE estimates. The dependent variable is the number of islamist attacks. Mean is the unweighted mean of the GTs for the  keywords, for the Popular and Extreme groupings, see Table . Standard errors in parentheses, ***, **, and * denote, respectively, %, %, and  % significance levels. 178 M. Nicolini et al. variables. 13 We report the results in Table 7: GT are positive and significant when considering the number of wounded and affected people, both on the full set of countries (Panel A) and in the subset of countries affected by an Islamist attack (Panel B). This suggests that larger attacks, that involve more victims, trigger more interest on the web. As a further robustness check, we discuss two placebo tests for the main findings reported in column 1 of Table 3. First, we test whether Google searches are significantly and positively correlated with non-Islamist attacks. In Column 1 of Table 8, we report results using all kinds of attacks as dependent variable, including those Table :Introducing temporal lags of Google searches. Panel A ()()() Countries hit by any terrorist attack Mean Popular Extreme GT t .*.*.** (.)(.)(.) GT t− .*. . (.)(.)(.) IA t− .*** .*** .*** (.)(.)(.) R. . . #Observations , , , # Countries    Panel B ()()() Countries hit by Islamist attacks Mean Popular Extreme GT t .** .*.** (.)(.)(.) GT t− .*. . (.)(.)(.) IA t− .*** .*** .*** (.)(.)(.) R. . . #Observations , , , # Countries    DCCE estimates. The dependent variable is the number of islamist attacks. Mean is the unweighted mean of the GTs for the  keywords, for the Popular and Extreme groupings, see Table . Standard errors in parentheses, ***, **, and * denote, respectively, %, %, and  % significance levels. 13 We compute the overall number of affected individuals as the sum of the killed and injured people. Online Interest in Radical Islam and Terrorism 179 perpetrated by non-Islamist terrorists. In columns 2, we remove Islamist attacks from the pool. We do not find any significant correlation between these attacks and searches for our sensitive keywords. As a second exercise, we consider a term that is completely unrelated to the Islamist attacks. We follow D’Amuri and Marcucci (2017) and choose the term “dos”,whichisanacronymforDisk Operating System. 14 We expect Google searches for the term “dos”to be unrelated to Islamist attacks. As reported in column 3 of Table 8, the coefficient is not statistically significant. We also repeat theexercisewiththeGTfor“google”finding again that it is not correlated with the occurrence of the attacks. Panel B shows that the results of the placebo tests do not change when only considering the subsample of countries affected by Islamist attacks. Table :Robustness with different measures of violence. Panel A ()()() Countries hit by any terrorist attack y= killed y= wounded y= victims GT t . .*.* (.)(.)(.) y t− .*** .*** .*** (.)(.)(.) R. . . # Observations , , , # Countries    Panel B ()()() Countries hit by Islamist attacks y=killed y=wounded y=victims GT . .*.* (.)(.)(.) y t− .*** .*** .*** (.)(.)(.) R. . . # Observations , , , # Countries    DCCE estimates. The dependent variable is respectively, the log of killed persons, the log of wounded persons, the log of total victims (equal to the sum of killed and wounded). Standard errors in parentheses, ***, **, and * denote, respectively, %, %, and  % significance levels. 14 It also stands for “Department of State”in the U.S., “Demokratska opozicija Srbije”, a political party in Serbia, and the IATA code for the Dios airport in Papua New Guinea. 180 M. Nicolini et al. 5 Discussion Our panel analysis reveals a systematic relationship between Islamist attacks and online searches for a set of sensitive keywords over the period 2004–2015 worldwide: online searches are significantly and positively correlated to attacks. Different kinds of users may search for some of the keywords we employ in the empirical analysis, Table :Placebo tests. Panel A ()()()() Countries hit by any terrorist attack y= all attacks y= non-Islamist attacks y= Islamist attacks GT t . −. (.)(.) dos t . (.) google t −. (.) y t− .*** .*** .*** .*** (.)(.)(.)(.) R. . . . # Observations , , , , # Countries     Panel B ()()()() Countries hit by Islamist attacks y=all attacks y=non-Islamist attacks y=Islamist attacks GT t . . (.)(.) dos t −. (.) google t −. (.) y t− .*** .*** .*** .*** (.)(.)(.)(.) R. . . . # Observations , , , , # Countries     DCCE estimates. Columns –present placebo tests adopting alternative dependent variables. Columns and present placebo tests adopting different explanatory variables. The dependent variable is the total number of terrorist attacks (columns –), the number of non-Islamist attacks (columns –), while in columns and is the standard measure adopted throught the empirical analysis, the number of Islamist attacks. Standard errors in parentheses, ***, **, and * denote, respectively, %, %, and  % significance levels. Online Interest in Radical Islam and Terrorism 181 from students who want to deepen their knowledge of political Islam to journalists and scholars interested in researching violent extremism. Online searches may also detect the celebrity of terrorist organizations. However, we also use keywords explicitly denoting an interest in fanaticism and violence that might be searched by sympathizers of the jihad. Distinguishing between keywords allows us to understand better the positive relationship between Google searches and the attacks. We divide our set of 47 keywords into two groups. The first group contains keywords with an average monthly frequency higher than 500 searches in the UK. Such searches are not explicitly violent and mostly regard political Islam. The SERP analysis conducted by Ahmed and Lloyd George (2016), however, shows that these seemingly non extremist searches often return extreme content aimed at explicitly promoting radicalization among Internet users. The second group contains more extreme queries, which are mostly related to violent extremism and systematically lead to fundamentalist, militant, and explicitly violent content (Ahmed and Lloyd George 2016). We find that the online searches for both groups of queries are significantly associated with the attacks. The coefficient for the popular keywords signals that the celebrity of terrorists among the public is significantly and positively correlated with the attacks. When we incorporate spatial lags to examine how terrorist attacks and search activity in neighboring countries influence attacks within each country, we find no statistically significant relationship between Islamist attacks in neighboring countries and the occurrence of an attack in a given focal country. However, the spatial lag of Google searches is positive and statistically significant, suggesting that when an Islamist attack occurs in a certain country, individuals in surrounding countries tend to increase their searches for sensitive keywords. More notably, when distinguishing between popular and extreme keywords, we find that the spatial lag for extreme words is not statistically significant in the global sample but becomes statistically significant and larger in magnitude when restricting the analysis to the subsample of countries that have previously experienced an Islamist attack. Overall, these findings highlight the role of geographic proximity in shaping online search behavior, particularly in regions with a history of Islamist terrorism. When we expand our analysis, including temporal lags of the Google searches at time t−1, we find that this does not substantially alter our main findings regarding contemporaneous Google searches and the lag of the dependent variable. The coefficient for Google searches in the previous month is positive and significant, suggesting that past search activity is positively associated with subsequent Islamist attacks. However, when we differentiate between popular and extreme keywords, we do not find any statistically significant effects. Notably, the significance of 182 M. Nicolini et al. contemporaneous Google searches remains robust, particularly for extreme keywords. Our finding that larger attacks–those resulting in a higher number of casualties and injuries–generate greater online interest aligns with recent evidence on the varying intensity of public responses to the severity of terrorist attacks (Bove, Efthyvoulou, and Pickard 2024; Efthyvoulou, Pickard, and Bove 2024). Specifically, Bove, Efthyvoulou, and Pickard 2024 have provided evidence that large-scale attacks elicit more prolonged public reactions, whereas the impact of smaller-scale incidents tends to fade within a month. Expanding on this, Efthyvoulou, Pickard, and Bove 2024 have documented that the emotional response to terrorism extends beyond immediate fear, reinforcing national identity–particularly when attacks involve a higher number of victims and are perpetrated by Islamic extremists–strengthening the sense of national belonging in the UK. A key limitation of this analysis–shared by all studies that rely on Google searches to track hidden behaviors–is the inability to perform sentiment analysis. More broadly, obtaining individual–level insights into the motivations behind online searches or determining which specific websites users visit after receiving Google’s suggested results is impossible. Our approach allows us to track searches for sensitive keywords, such as “jihad,”but more explicit queries–like “how to join jihad,” “jihad is good,”or “fight against jihad”and “revenge against jihad”–generate search volumes too low to be reliably analyzed. Consequently, our interpretation of the positive correlation between Google searches and terrorist attacks is inherently constrained and requires a degree of speculative reasoning. While we primarily interpret the increase in search volume as areflection of heightened interest in terrorist activities, we cannot disentangle this effect from alternative explanations–such as the amplification of fear, as documented by Bove, Efthyvoulou, and Pickard (2024), or the activation of negative reciprocity mechanisms (Fehr and Gatcher 2010). For instance, individuals may be motivated to search for terrorism-related content out of a desire for retaliation, potentially seeking ways to counteract terrorist interests through their actions. Unfortunately, the underlying sentiments driving these searches remain inaccessible and cannot be incorporated into any analysis based on Google Trends data. A promising direction for future research could be to investigate the relationship between terrorist attacks and behaviors indicative of negative reciprocity, offering insights into whether the public exhibits a desire for retaliation in the aftermath of such events. While online searches explicitly linked to retaliatory intent are unlikely to generate search volumes large enough for systematic analysis, this line of inquiry could be extended to examine the potential impact of exposure to attacks on voting behavior. Specifically, future research could explore whether terrorist attacks lead to Online Interest in Radical Islam and Terrorism 183 increased support for political parties advocating stronger counterterrorism measures–suggesting that electoral preferences may, at least in part, be shaped by a perceived need for retribution. To date, evidence suggests that Israeli voters exposed to rocket attacks from the Gaza Strip tend to increase their support for right–wing parties (e.g., Getmanski and Zeitzoff2014). However, systematic research on whether voters exhibit a similar retaliatory response to perceived threats in broader contexts–such as Western Europe and the U.S.–remains limited. Another potential limitation is that Google may not be the primary search engine in some regions. However, incorporating this factor into our empirical analysis presents significant methodological challenges that would compromise the reliability of our estimates. The main issue is the lack of consistent and reliable statistics on the use of alternative search engines across countries during our study period (2004–2015). While available market share data indicate that Google was the dominant platform in most countries, we lack precise information on search engine preferences in regions where Google was less prevalent due to cultural or political factors, such as China and Russia. Additionally, while Google search data are publicly accessible, the same does not apply to other platforms like Baidu or Yandex, making it impossible to assess whether search volumes for specific keywords follow similar patterns across different search engines. As a result, any analysis that interacts online searches with an index of Google usage in each country would be inherently unreliable. Addressing this issue in future research will depend on data availability, particularly from platforms operating in countries where government control over media and online information remains stringent. While being aware of these limitations, which prevent further interpretation of the correlations we find, we believe our choice of using Google searches has some advantages. First, Google searches do track general interest in some keywords, with Google being the first landing page for almost all users when searching the web. Other platforms, e.g., Twitter, would not be as representative. Second, the SERP analysis by Ahmed and Lloyd George (2016) on the keywords we consider in the present analysis ensures that these terms are sensitive, as users can end up inadvertently on extremist online pages. Keeping in mind this caveat, the positive correlation we find is consistent with previous evidence suggesting that media coverage and celebrity encourage terrorists to plan new missions (Frey and Osterloh 2018; Jetter 2017, 2019; Jetter and Walker 2018). The significant and positive coefficient of the queries for the extreme keywords, on the other hand, supports the interpretation that online searches may detect not only the celebrity, but also the popularity of the extremist message. The spreading of curiosity and sympathy for radical Islam and violent extremism may signal the formation of a fertile ground for radicalization. 184 M. Nicolini et al. Chudik, Alexander, and M. Hashem Pesaran. 2015. “Common Correlated Effects Estimation of Heterogeneous Dynamic Panel Data Models with Weakly Exogenous Regressors.”Journal of Econometrics 188 (2): 393–420. D’Amuri, Francesco, and Juri Marcucci. 2017. “The Predictive Power of Google Searches in Forecasting US Unemployment.”International Journal of Forecasting 33 (4): 801–16. Dickey, D. A., and W. A. Fuller. 1979. “Distribution of the Estimators for Autoregressive Time Series with a Unit Root.”Journal of the American Statistical Association 74 (366a): 427–31. Doleac, J. L., and L. C. D. Stein. 2013. “The Visible Hand: Race and Online Market Outcomes.”The Economic Journal 123 (572): F469–F492. Edwards, D. B. 2017. Caravan of Martyrs. Sacrifice and Suicide Bombing in Afghanistan. Oakland, California: University of California Press. Efthyvoulou, G., H. Pickard, and V. Bove. 2024. “Terrorist Violence and the Fuzzy Frontier: National and Supranational Identities in Britain.”Journal of Law, Economics and Organization 1–27. https://doi.org/ 10.1093/jleo/ewae003. Eksi, O., M. Y. Gurdal, and C. Orman. 2017. “Fines versus Prison for the Issuance of Bad Checks: Evidence from a Policy Shift in Turkey.”Journal of Economic Behavior and Organization 143: 9–27. El Ouadghiri, I., and J. Peillex. 2018. “Public Attention to “Islamic Terrorism”and Stock Market Returns.” Journal of Comparative Economics 46 (4): 936–46. Fecht, F., S. Thum, and P. Weber. 2019. “Fear, Deposit Insurance Schemes, and Deposit Reallocation in the German Banking System.”Journal of Banking and Finance 105: 151–65. Fehr, E., and S. Gatcher. 2010. “Fairness and Retaliation: The Economics of Reciprocity.”Journal of Economic Perspectives 3 (14): 159–81. Feridun, M., and S. Sezgin. 2008. “Regional Underdevelopment and Terrorism: The Case of South Eastern Turkey.”Defence and Peace Economics 19 (3): 225–33. Frey, B. S., and M. Osterloh. 2018. “Strategies to Deal with Terrorism.”CESifo Economic Studies 64 (4): 698–711. Gentzkow, M. A., and J. M. Shapiro. 2004. “Media, Education and Anti-americanism in the Muslim World.” Journal of Economic Perspectives 16 (3): 117–33. Getmanski, A., and T. Zeitzoff. 2014. “Ethnic Groups, Political Exclusion and Domestic Terrorism.”American Political Science Review 3 (108): 588–604. Giulietti, C., M. Tonin, and M. Vlassopoulos. 2019. “Racial Discrimination in Local Public Services: A Field Experiment in the United States.”Journal of the European Economic Association 17 (1): 165–204. Guriev, S., and N. Melnikov. 2016. “War, Inflation, and Social Capital.”American Economic Review 106 (5): 230–5. Harris, T. F., and A. Yelowitz. 2018. “Racial Climate and Homeownership.”Journal of Housing Economics 40: 41–72. Helfstein, S. 2014. “Social Capital and Terrorism.”Defence and Peace Economics 25 (4): 363–80. Im, Kyung So, M. Hashem Pesaran, and Yongcheol Shin. 2003. “Testing for Unit Roots in Heterogeneous Panels.”Journal of Econometrics 115 (1): 53–74. Jetter, M. 2017. “The Effect of Media Attention on Terrorism.”Journal of Public Economics 153: 32–48. Jetter, M., and J. K. Walker. 2018. The Effect of Media Coverage on Mass Shootings. Bonn, Germany: IZA Discussion Paper No. 11900. Jetter, M. 2019. “The Inadvertent Consequences of Al-Qaeda News Coverage.”European Economic Review 119 (IZA Discussion Paper No. 10708): 391–410. Kearney, M. S., and P. B. Levine. 2015. “Media Influences on Social Outcomes: The Impact of MTV’s 16 and Pregnant on Teen Childbearing.”American Economic Review 105 (12): 3597–632. Online Interest in Radical Islam and Terrorism 191 LaFree, G., and L. Dugan. 2007. “Introducing the Global Terrorism Database.”Political Violence and Terrorism 19 (2): 181–204. Ma-Kellams, C., F. Or, J. H. Baek, and I. Kawachi. 2015. “Rethinking Suicide Surveillance: Google Search Data and Self-Reported Suicidality Differentially Estimate Completed Suicide Risk.”Clinical Psychological Science 4 (3): 1–5. MacInnis, C. C., and Gordon Hodson. 2015. “Do American States with More Religious or Conservative Populations Search More for Sexual Content on Google?”Archives of Sexual Behavior 44 (1): 137–47. Maliach, A. 2010. “Abdullah Azzam, Al-Qaeda, and Hamas: Concepts of Jihad and Istishhad.”Military and Strategic Affairs 2 (2): 79–95. Nelson, P. S., and J. L. Scott. 1992. “Terrorism and the Media: An Empirical Analysis.”Defence and Peace Economics 3 (4): 329–39. Perdue, R. T., J. Hawdon, and K. M. Thames. 2018. “Can Big Data Predict the Rise of Novel Drug Abuse?” Journal of Drug Issues 48 (4): 508–18. Pesaran, M. Hashem. 2004. General Diagnostic Tests for Cross Section Dependence in Panelss. Bonn, Germany: IZA Discussion Paper No. 1240. Pesaran, M. Hashem. 2006. “Estimation and Inference in Large Heterogeneous Panels with a Multifactor Error Structure.”Econometrica 74 (4): 967–1012. Pesaran, M. Hashem. 2007. “A Simple Panel Unit Root Test in the Presence of Cross-Section Dependence.” Journal of Applied Econometrics 22 (2): 265–312. Pesaran, M. Hashem, and Ron Smith. 1995. “Estimating Long-Run Relationships from Dynamic Heterogeneous Panels.”Journal of Econometrics 68 (1): 79–113. Pesaran, M. Hashem, Yongcheol Shin, and Ron P. Smith. 1999. “Pooled Mean Group Estimation of Dynamic Heterogeneous Panels.”Journal of the American Statistical Association 94 (446): 621–34. Rohner, D., and B. S. Frey. 2007. “Blood and Ink! The Common-Interest-Game between Terrorists and the Media.”Public Choice 133 (1–2): 129–45. Stephens-Davidowitz, S. 2014. “The Cost of Racial Animus on a Black Candidate: Evidence Using Google Search Data.”Journal of Public Economics 118: 26–40. Torres Soriano, M. R. 2010. “The Road to Media Jihad: The Propaganda Actions of Al Qaeda in the Islamic Maghreb.”Terrorism and Political Violence 23 (1): 72–88. Weimann, G. 2015. Terrorism in Cyberspace: The Next Generation. New York, NY: Woodrow Wilson Center Press with Columbia University Press. Wooldridge, J. M. 2002. Econometric Analysis of Cross Section and Panel Data. Cambridge, Massachussets, London, UK: MIT Press. Young, S. D., and Q. P. Zhang. 2018. “Using Search Engine Big Data for Predicting New HIV Diagnoses.” PLoS One 13 (7): e0199527. Young, S. D., E. A. Torrone, J. Urata, and S. O. Aral. 2018. “Using Search Engine Data as a Tool to Predict Syphilis.”Epidemiology 29 (4): 574–8. Zeitzoff, T. 2017. “How Social Media Is Changing Conflict.”Journal of Conflict Resolution 61 (9): 1970–91. 192 M. Nicolini et al.