The framing of 9/11 in American, French, and Dutch national newspapers (2001–2015): An inductive approach to studying events

In this article, the authors employ work on events, framing, and social problems to develop an inductive approach to studying events. The article’s central premise is that tracing direct references made to events offers a fruitful strategy to investigate their framing cross-nationally and over time. The authors apply their approach to the case of 9/11 in American, French, and Dutch national newspapers (2001–2015). By combining word counting, topic modeling, and content analysis they examine the amount of attention given to 9/11, the issues to which it has been linked, and the implications attributed to it. The results indicate that the framing of 9/11 in the above three countries has been stable and uniform regarding foreign issues. There are however enduring, marked cross-national differences with respect to domestic issues. In France, 9/11 has barely been related to such issues; in the United States, the event has been connected principally to national security; and in the Netherlands, it has mainly prompted the problematization of Muslim immigrants. Because 9/11 has been a significant event many years after its occurrence, albeit differently so in each country of study, the findings point to the relevance of studying the framing of events cross-nationally and over extended periods of time.


Introduction
The terrorist attacks on the Twin Towers (New York) and the Pentagon (Washington, DC) on 11 September 2001 are generally considered important catalysts for policy shifts and societal transformations, including the introduction of the Patriot Act, the War in Afghanistan, the Iraq War, and the rise of anti-Muslim sentiments. It is patently clear that 9/11 has been a significant event, yet identifying its specific transformative effects is a challenge. This is true for events in general: whereas we know that they are important and consequential, it is difficult to understand their meanings and impacts. Several scholars have in recent years made strides towards opening this 'black box' of events (Berezin, 2009;Sewell, 2005;Wagner-Pacifici, 2010. However, this body of work still leaves certain questions unanswered. What is an appropriate strategy for studying event framings? How do they differ among social contexts, and in what ways do they change over time? This article aims to answer these questions by introducing and employing an inductive approach to studying events. The central premise of this approach is that tracing the direct references made to events constitutes a fruitful strategy for studying the framing of events in different countries and over time. Employing work on framing and social problems (especially, Entman, 1993;Hilgartner and Bosk, 1988;Snow and Benford, 1988), we derive three core event-framing parameters: the amount of attention events receive, the issues to which they are related, and the implications attributed to them. We then use a mixed-methods combination of word counting, topic modeling, and manual quantitative and qualitative content analysis to study these three parameters for the case of 9/11 in American, French, and Dutch national newspapers over the period of 2001-2015. Our investigations reveal, in short, that 9/11 has remained significant for a long time, although diversely so, in each of these three contexts: implications have been attributed to it in each country many years after its occurrence; however, these implications concern very different issues. These findings indicate the relevance of investigating the framing of events over extended periods of time and in different country contexts.

Theory: Events, framing, and social problems
Why and how are events studied?
Scholars such as Berezin (2009), Sewell (2005, and Wagner-Pacifici (2010, 2017) make a strong argument why studying events is important: to understand social change. Whereas social life usually occurs with a high degree of continuity and transformations often come about in a fairly gradual way, events offer opportunities for swift alterations. Most occurrences people experience during their lives are 'ordinary happenings' that take place without significant change (Sewell, 2005: 226-228). Events, in contrast, initiate periods of transformative disruption, which Sewell (2005) calls 'breaks with normal life ' and Swidler (1986) and Zolberg (1972) typify as 'unsettled times' and 'moments of madness,' respectively.
These ideas have been used to investigate a wide variety of events. For instance, in his seminal study of the French Revolution, Sewell (2005: 225-244) shows how the taking of the Bastille triggered a redefinition of the very concept of politics. Berezin (2009) analyzes whether the Maastricht Treaty's ratification in 1993 played a role in the National Front's 2002 electoral breakthrough; Geiß et al. (2016) examine the extent to which the framing of economic policies in Germany changed in the wake of Lehman Brothers' bankruptcy; and Woods and Arthur (2014) research possible shifts in American public discourse about immigrants in response to 9/11. Scholars have also used a wide variety of data and methods to study events. However, most share one methodological element: they employ pre/post measures to investigate the impacts of events. Analyses of 9/11 illustrate this (for other events, see Álvarez-Benjumea and Winter, 2020;Berezin, 2009;de Vries, 2017;Kriesi and Pappas, 2015;Larsen et al., 2020;Legewie, 2013;Semetko et al., 2003). Fetzer and Soper (2003) use two surveys -one conducted beforehand (July 2001) and one afterward (April 2002) -to study shifts in British, French, and German public attitudes towards state accommodation of Islamic practices in schools. They find a decrease in support in all three countries (albeit not a significant one in France), and conclude that these data 'partially confirm' (Fetzer and Soper, 2003: 255) their hypothesis that 9/11 affected these attitudes. Sander and Putnam (2010: 9) claim that there is a 'post-9/11 split' in the trend of declining civic engagement among Americans. They show that in the years after 2001, percentages of political and societal involvement increased among young people in the US, indicating a '9/11 generation' (Sander and Putnam, 2010: 10) which, through experiencing the event and its aftermath at an early age, became more engaged in civic affairs. Woods and Arthur (2014) compare articles about immigration from The New York Times andThe Wall Street Journal during 1998-2004. They find that the percentage of articles negatively representing immigrants in the US was higher in 2001-2004 than in 1998-2001, and conclude that American political discourse has taken an 'authoritarian turn' (Woods and Arthur, 2014: 435) in response to 9/11. The idea that a change in public attitudes or discourse in the period following a significant rupture must have something to do with the rupture's occurrence is intuitively appealing. However, it is not a direct measurement of an event's impact. Other significant events and developments can co-occur, which makes it difficult to identify the effects of one specific event by only focusing on differences before and after the fact (Muñoz et al., 2020: 189-191).
This means of measuring becomes even less reliable when one wants to study event effects over a long period of time. Most of the event scholars mentioned above research events' transformative effects during relatively short periods of time, in the first days or weeks after an event occurs. This is apparent in the terminology they use: 'breaks with normal life' (Sewell, 2005), 'unsettled times' (Swidler, 1986), and 'moments of madness' (Zolberg, 1972). However, Wagner-Pacifici (2010 proposes that the significance of events should be studied over long periods of time. She contends that events have an inherently 'restless' character: their relevance comes and goes in different 'flows' (e.g., they may suddenly gain importance after being ignored for a long time) and they take on distinctive 'forms,' when related to varying issues, discussions, and historical narratives (Wagner-Pacifici, 2017: 10-15). Such a historical perspective calls for a methodology that facilitates tracking events' effects over long periods of time: how can effects be related to the original occurrence 10 or 20 years after the fact, given all the other things that happen in between?

Towards an inductive approach to studying events: Attention, issue attachments, and implications
To address this issue, we introduce an inductive approach to studying events. This approach begins from the assumption that a fruitful strategy for studying events is to investigate the frames that actors directly relate to them. Focusing on instances in which actors refer to an event enables close measurement of its framing processes, over long periods of time, in differing social contexts. Various scholars employ these reasonings to study diverse framing processes (Matthes, 2009;Roggeband and Vliegenthart, 2007;Törnberg and Törnberg, 2016;van Gorp and Vercruysse, 2012).
The next question is: how can we study the public significance of events inductively? The work on events we have discussed above does not offer much in the way of concrete tools to pursue this end. Yet, research on framing and social problems, which both deal with meaning-making processes in public spheres, proposes several useful concepts: 1. The amount of attention an event receives. Hilgartner and Bosk (1988) famously argued that social problems compete with one another for attention in public arenas (see also Lawrence and Birkland, 2004;Maratea, 2008;Ungar, 1992). The more attention paid to a problem, the higher the chances citizens will express worry about it and that politicians will be forced to act. Similarly, we can assume that events also compete with one another for attention, and the more attention an event receives, the higher its political and social significance will be. Hilgartner and Bosk (1988: 71) claim that the allocation of attention is dynamic over time, creating possible shifts in the public importance of social problems over the course of years (see also Nisbet and Huge, 2006;Ungar, 1992). A similar argument may be made for events, which supports our aim to follow developments in their public significance over long periods of time. 2. The issues attached to events. Research on framing and social problems claims that issue attachment is a central feature in the construction of social problems. How problems are defined (Entman, 1993: 52) or diagnosed (Snow and Benford, 1988) matters a great deal with regard to how politicians are forced to respond to them and the policies formulated to tackle them. Benson and Saguy's (2005) analysis of the framing of the social problems of sexual harassment and immigration in American and French newspapers provides a concrete example. Concerning sexual harassment, one of their main empirical conclusions is that, whereas in the US this social problem is often framed as sex-based discrimination, in France it is generally considered an instance of power abuse (Benson and Saguy, 2005: 250). Such cross-nationally diverse definitions (Entman, 1993) or diagnoses (Snow and Benford, 1988) of the same social problem pave the way for varying country-specific political responses. Many other analyses indicate that issue attachment to social problems or events can differ significantly among national contexts (see, for instance, de Vreese et al., 2001;Ferree et al., 2002;Koopmans et al., 2005;, which calls for the inclusion of this concept in a country-comparative approach to events.
3. The implications inferred from events. People may pay attention to an event and may attach it to issues. Yet, the impact of an event only becomes truly concrete once actors start pushing for political or societal transformations in response.
Researchers who study framing refer to this process as suggesting remedies (Entman, 1993: 52) or prognostic framing (Snow and Benford, 1988). Many scholars use one of these concepts to investigate courses of action formulated to tackle social problems (e.g., Eilders and Lüter, 2000;Johnston et al., 2014;Ryan, 2004;Snow and Byrd, 2007). We propose the concept of implications to study similar processes pertaining to events, because we consider the term neutral. Both 'remedies' and 'prognostic framing' imply resolutions to negative situations (treating a problem), whereas many calls for change can be inferred from events that are conceived positively by a significant number of people (e.g., the Fall of the Berlin Wall).

Cases
We have chosen the United States, France, and the Netherlands as cases because they represent an array of sociological similarities and differences that make for an interesting cross-national comparison of 9/11's framing. On the one hand, these are all Western liberal democracies that are considerably connected politically, economically, and culturally. One would therefore not expect the framing of 9/11 to be vastly different between the three national contexts. On the other hand, the countries vary in several factors that might cause distinct cross-national research outcomes. For instance, the fact that the attacks took place in the United States and mostly killed Americans is likely to have resulted in a higher level of event identification for Americans (cf. Berezin, 2009: 93). Dutch political and cultural elites generally have a more positive stance towards the United States than their French counterparts (Bowen, 2006;Kooijman and Kuipers, 2008), which could make them respond differently to the event. In line with our inductive approach, we do not offer specific hypotheses for our respective cases, but use inductive methods to reveal salient similarities and differences in the framing of 9/11.

Corpus
Our study focuses on national newspapers, since they are central fora for political and cultural elites -politicians, journalists, public intellectuals, and scholars -to share their views on matters of societal importance and thus shape the public framing of events (cf. Benson and Saguy, 2005;. To construct our corpus, we used the LexisNexis database to select newspaper articles that mention 'September 11' ('11 septembre' in French and '11 september' in Dutch). We decided on this term after a process of trial and error, in which other search terms turned out to be less fruitful. 1 Since we aim for analyzing a wide variety of framing practices among national elites, we included all types of newspaper articles with references to the event (opinion articles, interviews, news stories, background reports, etc.).
The selected newspapers were different for each of the three research steps. To measure the amount of attention given to 9/11, it was essential to include comparable newspapers. We therefore examined only elite newspapers, two from each country: To analyze issues and implications, of importance was not how many articles mentioning 9/11 were published, but rather their contents. Thus, for these two research steps, we selected articles from all the national newspapers available in LexisNexis: Le Monde, Le Figaro, La Croix, and L'Humanité (France); De Volkskrant, NRC Handelsblad, Trouw, Algemeen Dagblad, and De Telegraaf (the Netherlands); The New York Times, The Washington Post, USA Today, and the New York Daily News (the United States). New York Daily News is a local newspaper, though it also reports on national and international affairs; we included it in order to cover the tabloid press for the American case. The Wall Street Journal (US) and Liberation (France) are not part of our corpus, since they are not available in LexisNexis. Our entire corpus consists of 141,454 articles (80,013 American, 41,220 French, and 20,221 Dutch).

Operationalization
Attention. To chart cycles of attention for 9/11, we used basic word-counting, a method frequently employed in framing and social problems research (e.g., Nisbet and Huge, 2006: 19-20;Shih et al., 2008: 152). We measured how many newspaper articles per year from each country mentioned the event in the period 2001-2015, and then standardized these numbers according to the newspapers' respective sizes (cf. Alimi and Maney, 2018: 765-767).
Issues. To analyze the issues related to 9/11, we used the topic-modeling technique Latent Direchlet Allocation (LDA). 2 Topic modeling yields clusters of word associations -topics -that often appear together in text corpora (Blei, 2012;Grimmer and Stewart, 2013). Being an inductive research technique, it aligns with our aim to investigate the framing of 9/11 as openly as possible in different country contexts, over an extended time period. Although making analyses for a long period of time is not the most common way to perform topic modeling, some scholars do this (Hall et al., 2008;Marshall, 2013).
In order to make cross-national and historical comparisons, we split the corpus into 42 separate subcorpora, one for each country for each of the 14 years from 2001 to 2015. Before running the topic models, we cleaned the data by deleting stop words and commonly used phrases. As is customary in topic modeling, we experimented with the number of topics. We settled on 10 topics for subcorpora of up to 1000 articles, and 20 topics for ones with more than 1000 articles (for similar results with a topic-modeling analysis of newspaper articles, see Marshall, 2013). This yielded a database of 240 topics for the US and 170 each for both France and the Netherlands.
To be able to find meaningful patterns in the database, the topics had to be interpreted. Two researchers -the article's first author and another sociologist -coded the topics independently by looking at the 15 most frequent terms. When a topic did not make sense on first sight, both coders used a topic-modeling tool that indicates for each text (newspaper article) how representative it is for a topic. Then they read the five most representative articles. If this procedure did not yield a clear understanding of a topic, it was coded as 'unclear.' The coding's Krippendorff alpha was 0.93, which is a very acceptable intercoder reliability score (Krippendorff, 2012).
Each topic was coded for two characteristics. The first was for the issue it concerned. The coders did not begin with predefined categories, but gave each topic a short description that they thought fit best. Second, topics were coded as relating to foreign or domestic affairs (meaning, whether or not the topic referred to developments outside or inside the country for which the code was given). The idea to include this characteristic came up as a result of the inductive research strategy: it emerged during the analysis process that significant differences regarding this parameter were present among the three countries.
Coding for issues provided a list of more than 100 topic codes, many of which were very specific (for instance, a Dutch one in 2002 about safety issues during the Winter Olympic Games). To facilitate comparison, the coders followed the example of other researchers who inductively analyze issue attachment (Snow et al., 2007: 392-293;van Gorp and Vercruysse, 2012: 1275-1276 and reduced this list to a manageable number of 12 aggregated topics: Memories of 9/11; Islam; Iraq War; War in Afghanistan; Safety; Terrorism; Economy; Media and Arts; International Relations; National Politics; Air Traffic; and American Politics/Society. 3 They were able to assign 98% of the specific topics to these 12 aggregated topics; the remaining 2% were coded as 'Other.' As a last step in the topic-modeling analysis, for each subcorpus the percentage of content that a single topic added to its total size was investigated. These percentages were then used to measure the relative sizes of the various aggregated topics and of the distribution between foreign and domestic topics in all the subcorpora.
Implications. We searched for event implications by looking for sentences or paragraphs containing claims that certain ideas, policies, or institutions should be supported, rejected, or transformed due to 9/11. To include the full array of voices present among national elites, we made no distinction between implications that were expounded by politicians, journalists, academics, or any other professional. Because we aimed for generating comparable results, we used the list of aggregated topics and counted how many implications were related to each of them.
Additionally, to get a better understanding of the discursive nature of the associations and arguments put forward by the actors referenced in the newspapers, we also engaged in close reading of the implications. In the results section, we provide quotations to illustrate our findings of this qualitative analysis. As earlier studies of public meaning-making processes indicate (e.g., Bail, 2014;Benson and Saguy, 2005), a mixed-methods design is a fruitful research strategy, which yields a combination of general trends (through quantitative measurements) and depth of analysis (via qualitative investigations).
Since it was not feasible to undertake the implications analysis for all articles in the corpus, we conducted it for five periods. Three of them included the two weeks immediately following a terrorist attack: 9/11, the train attacks in Madrid in March 2004, and the Charlie Hebdo attack in Paris in January 2015. We analyzed the latter two attacks to examine if and how framings of 9/11 changed after the occurrence of new events. To find newspaper articles about these cases, we added 'Madrid' plus 'train,' and 'Charlie Hebdo,' respectively, as search terms. The other two periods we examined were two weeks of 'settled times' (Swidler, 1986): weeks during which there were no events (e.g., terrorist attacks, military invasions, national elections) that could easily be associated with 9/11. One period was in the fifth year after the attacks (in 2006), the other during the tenth year (in 2011). These periods were included to see what 'role' 9/11 (still) played during these settled times.
We found that 80 was the maximum number of 9/11-related articles during each of the five periods, with the exception of the period immediately after 9/11, which had many more articles for all three countries. Therefore, we chose a random sample of 80 articles from each country for this post-9/11 period.

Attention
Unsurprisingly, in each country, 9/11 received the most attention immediately after the attacks ( Figure 1). During the whole period of study, the most attention was given to the event in the United States. For France and the Netherlands, the cycles were largely similar -with the exception of the first two years, when 9/11 received more attention in French newspapers. However, what is perhaps most remarkable is that the event continued to receive attention long after its actual occurrence. Taking the Dutch newspaper De Volkskrant as an example, we see that although it was the newspaper with the fewest articles related to 9/11 in our corpus, it still printed 13 articles mentioning the event in April 2015.

Issues
Concerning the issues linked to 9/11, we find strong similarities among the three countries with respect to foreign affairs topics ( Table 1). The topic 'Iraq War' received a considerable amount of attention in each country in the first years after 9/11 (almost always over 10% until 2007/2008). For 'War in Afghanistan,' this was the case for almost the entire period of study. This was also true for the topic 'Terrorism,' which in each country nearly every year was over 10%. These findings corroborate the widely shared scholarly view that 9/11's framing processes in the context of foreign affairs have mostly focused on these three issues (e.g., Entman, 2003;Gershkoff and Kushner, 2005;Jackson, 2005).
Cross-national differences do exist. For instance, the topic 'Economy,' mainly concerning the impact of 9/11 on the world economy, received more attention in France. In general, though, there is a high degree of uniformity in the contents of foreign affairs topics.
This differs significantly for topics regarding domestic affairs ( Table 2). The United States and the Netherlands each have one specific topic that received substantial attention during almost the entire period of research, while both topics are more or less absent in the other two countries. For the United States, this is 'Safety' (with an attention rate of at least 10% every year), and for the Netherlands it is 'Islam' (at least 10% in 11 out of 14 years). 4 Thus, in American newspapers, 9/11 has become a 'domestic safety event,' while in the Dutch ones it has turned into a 'domestic Islam event. ' The results in the Netherlands can be interpreted as a corroboration of the idea that it has become common to frame many domestic affairs through the lens of Islam there. Several scholars claim that, for many Dutch commentators, 9/11 has provided (further) proof of an inimical relationship between Islam and 'Western' or 'Dutch' values (Entzinger, 2006;Van der Veer, 2006). The Dutch topics confirm this hypothesis: many deal with the relationship between these attacks and, for instance, the radicalization of Muslim youth in the Netherlands, or Dutch Muslims' opinions of terrorism, homosexuality, or democracy.
While we know that Muslims in the United States faced increased verbal and physical assaults after 9/11 (Bakalian and Bezorgmehr, 2009), our results indicate that the explicit public stigmatization of this group was limited compared to the Netherlands. This finding aligns with Christopher Bail's (2014: 135) observation that 9/11 did not create a 'groundswell of anti-Muslim sentiment within the American public sphere' (cf. Garg et al., 2018;Woods and Arthur, 2014). Indeed, his book's bigger empirical claim is that it was not the shock of 9/11 that instantly activated Islamophobia, but 'fringe' anti-Muslim organizations, which, especially in 2004-2008, gradually expanded space for discourses critical of Islam (Bail, 2014: 67-86    It should be noted, though, that anti-terrorism discourses and policies developed in response to 9/11 have often resulted in implicit forms of stigmatization of American Muslims. Byng (2008) indicates that, although these discourses and policies were presented as 'neutral,' in practice they targeted Muslims much more often than other social groups. Identifying such subtextual or implicit forms of stigmatization is difficult using our inductive event approach, as it only focuses on explicit event associations. However, that we found hardly any explicit relationship between 9/11 and domestic Islam as a topic in the American newspapers is interesting in and of itself.
Research on France generally assumes that 9/11 has not had a great impact in this country (Berezin, 2009;Bowen, 2006). Our findings corroborate this, though only in part. Whereas American newspapers emphasized national safety affairs and Dutch ones often focused on Muslim integration, in France 9/11 was seldom related to domestic issues ( Figure 2). The annual percentages never exceeded 16%, except for the last year of the analysis, when many French topics related the event to the attack on Charlie Hebdo. This French 'non-domestication' of 9/11 does, however, not imply that the occurrence has been insignificant for its national elites. As Figure 1 indicates, they have discussed it extensively, in some years even more than their Dutch counterparts. Yet, they have done so in the context of foreign affairs.

). For instance, the American columnist Anthony Lewis wrote on 12 September 2001 in The New York Times:
The terrorist attack should inspire reflection about all of our national security policies, including the proposed missile defense system. Its critics have always said that this country will be in greater danger in the future from suitcase bombs or other terrorist devices than from missiles, and their case has now been devastatingly made. Whatever direction policy takes, we need joint action to protect the world from terrorism, nuclear or otherwise.
The intensity and magnitude of implications immediately after the event reflect a sentiment that the world has changed and action must be taken. The appearance of such an intense statement with such a clear implication is less imaginable five years later in each of the three countries. An example of a statement representative of the intensity of the implications which were inferred from 9/11 around that time comes from an opinion piece in Trouw (10 June 2006), concerning integration policies for Dutch immigrants: Why is it that the debate on integration in our country is so narrow-minded that it is only focused on the level of Dutch language that immigrants should be able to speak? Because of the 'multicultural drama' and '9/11,' the Dutch political climate has changed dramatically, and immigrants are bullied.
On the one hand, 9/11 is very much present in this citation (it has led to a change in political climate). On the other hand, the climate is already there. No new consequences are related to the event. However, in all three countries, 9/11 regained its salience during the periods immediately following the Madrid and Charlie Hebdo attacks, with numerous implications articulated in response to those events. In these cases, the memory and imagery of 9/11 were reactivated by the new attacks. For instance, in the United States, the train attacks in Madrid were often framed as a 'reminder' or 'wake-up call' that terrorism remained a big threat. This reaffirms the status of 9/11 as a foundational terrorist experience. An editorial commentary in The Washington Post the day after the Madrid attacks (12 March 2004) illustrates this: The Bush administration's clumsy diplomacy and its critics' hyperbolic charges of 'unilateralism' sometimes obscure the fact that the United States has had true and valuable allies in the war on terrorism. Yesterday one of the best of those, Spain, suffered a blow as shocking and as terrible as any the enemy has landed since Sept. 11, 2001. . . . The horror of Madrid only confirms that a broad and determined alliance is the only answer to terrorism. It reminds us that the United States neither fights, nor suffers, alone.
It turns out that most of the patterns we came across in analyzing implications are similar to those acquired with the topic-modeling analysis. In both cases, we find uniformity in the framing of foreign affairs and variation with respect to domestic affairs. In each country, most foreign affairs implications are related to the issues 'War in Afghanistan,' 'Terrorism,' and 'Safety' (Table 3). With respect to domestic affairs, the number of implications in France is considerably lower than in the other two countries (with the period after the Charlie Hebdo attack as an exception). This confirms that 9/11 is a 'non-domestic event' in France. French public actors consider it an important occurrence, but not for their domestic affairs (Table 4). Furthermore, in the United States, 'Safety' is the domestic issue with the most implications in all five periods, while in the Netherlands it is 'Islam.' Tables 3 and 4 also show that there is a high degree of overall stability in each country in terms of the issues to which these implications refer. Accordingly, both our topicmodeling and implications analysis indicate that 9/11's issue attachment did not change much over the years. Analysis of the implications furthermore offers insight into the specific national discursive dynamics behind these patterns. It turns out that they represent omnipresent trends in each country and not (just) the specific framing of 9/11 by one distinctive political group or newspaper. For instance, in the United States, the link between 9/11 and 'Safety' is made by both left-and right-wing actors from various professions, and in all kinds of newspapers. While there are discursive differences-Democrats tend to be less in favor of introducing of the Patriot Act than Republicansthey all relate 9/11 to safety affairs.
In France, public actors mostly invoke 9/11's implications at the global or European Union level. After the Madrid attacks, various French public actors requested the introduction of a European Patriot Act (duplicating the American policy response to 9/11), not a French one. The following excerpt from an opinion piece in Le Monde (14 March 2004) illustrates the general reaction in France during this period: The European Union must introduce before the end of the month, as it did most recently in July 2002, an update to its Action Plan against terrorism, which was adopted in September 2001, during the weeks after the attacks in New York and Washington.
After the attack on Charlie Hebdo, 9/11 was often mentioned in French newspapers. Yet, this does not mean that 9/11 in itself suddenly became relevant to French domestic politics. The Charlie Hebdo attack was regularly framed as the 'French 9/11,' implying that 9/11 was not 'French,' and a comparison was frequently made: should the French copy the American response to 9/11 now that they had experienced a terrorist attack by Muslim fundamentalists in their own country? Thus, the central French framings of 9/11 remained the same after the Charlie Hebdo attack: as an American and international watershed moment, and not an event of importance to France domestically.
In the Netherlands, 9/11 has become much more domesticated than in France. As mentioned above, the central issue discussed in Dutch newspapers was what the attacks on the Twin Towers implied for the integration of Muslims into Dutch society. Participants in this discussion often stated that there was a need to re-evaluate the group's position in the Netherlands due to multiple violent events related to Islam. When someone criticized this framing, these events were frequently mentioned as a counterargument. This happened to Dutch-Moroccan writer Hassnae Bouazza, shortly after the Charlie Hebdo attacks. In an interview with De Volkskrant (12 January 2015), Bouazza stated that she did not believe that, just because Dutch Muslims belong to the same religious group as the perpetrators, they should be obliged to publicly condemn the attacks. The interviewer responded as follows: When you sum up the things that have happened in recent years -the fatwa against Salman Rushdie, the threats against Geert Wilders and Ayaan Hirsi Ali, 9/11, the attacks in Madrid and London and now Paris, the ruthlessness of the Islamic State -one might imagine there is a problem with Islam.
The presentation of such strings of events, in which no clear distinction is made between domestic and foreign occurrences, to problematize Dutch Muslims clearly distinguishes the 9/11 framings in the Netherlands from those in the United States and France.

Theoretical insights
We have outlined our results and considered how they relate to existing research on 9/11. Yet, the question remains: what can these results tell us more generally about events, framing processes, and social problems? In the introduction, we proposed employing our inductive event approach in order to make two types of comparisons: historical and cross-national.
Regarding historical comparisons, we find that attention to 9/11 in all three countries was the highest right after the event and then subsequently diminished. This indicates that, once an event's novelty is over, it will -as Hilgartner and Bosk (1988: 62) already outlined with regard to social problems -start losing in the competition for public attention compared to other news items. What is more surprising, however, is that the numbers of implications in all three cases reached new peaks after the occurrence of new terrorist attacks: Madrid (2004) andCharlie Hebdo (2015). This shows that events' transformative potential does not have to disappear after the first several weeks or months of rupture. Yet, to bring this potential alive again requires, as Hilgartner and Bosk (1988: 71) state, 'new drama.' So, it is indeed worthwhile to study events over long periods of time, instead of only focusing on the initial period of the 'break with normal life' (Sewell, 2005), 'unsettled times' (Swidler, 1986), or 'moments of madness ' (Zolberg, 1972). Our results for 9/11 are close to Wagner-Pacifici's (2010) thesis of the 'restlessness of events.' However, at the same time, they show less fluidity in event framings than this concept suggests: in all three countries, there was significant stability in issue attachment over the course of years. Thus, we find that 9/11 has had a long and stable life as an event.
Concerning the country-comparison aspect of our analysis, we see strong evidence that, as many scholars of framing and social problems state (e.g., de Vreese et al., 2001;Ferree et al., 2002;Koopmans et al., 2005;, national differences in public framing processes surely still exist in the contemporary world. The fact that, in each of three countries, 9/11 received specific national 'colorings' proves the relevance of including multiple national contexts when analyzing event framings. Also, because our analyses of topic modeling and implications yielded very similar results, we corroborate claims that diagnostic and prognostic framing processes often correspond (Futrell, 2003;Gerhards and Rucht, 1992). This is an important result, as it points to the significance of the phenomenon of issue attachment: once an event becomes related to a certain dominant issue (e.g., safety in the US or Islam in the Netherlands), this is probably also the issue around which most discursive and policy transformations will take place.
A finding that brings a new insight to the literature is the strong difference in extents of event domestication that we discovered. Thus, event framings can not only vary among national contexts, in the sense that they get attached to different issues. They can also diverge in the degree to which these issues are domestic or foreign. We know from cultural sociological research that the extent to which cultural products 'travel' can differ substantially across countries (e.g., Heilbron, 1999;Janssen et al., 2008;Kuipers, 2015). Yet, this is -to the best of our knowledge -a new finding in relation to events.

Conclusions
In this article, we draw upon work about events, framing, and social problems to develop an inductive approach for studying events, which we apply to framings of 9/11 in American, French, and Dutch national newspapers during the first 14 years after its occurrence (2001)(2002)(2003)(2004)(2005)(2006)(2007)(2008)(2009)(2010)(2011)(2012)(2013)(2014)(2015). This constitutes an innovative research approach in that it enables direct measurement of event framing processes in different social contexts over a long period of time. We used this approach to study framings of 9/11 in newspaper articles, but designed it as a generic approach for studying all kinds of events, employing various data sources. For instance, this method would suit very well as a research approach to ordinary citizens' framings of events on social media or, to measure political event effects more directly, politicians' arguments in legislative debates.
Even though our approach contributes methodologically to event research, it also has limitations. First, because of its focus on references to an event, the approach can only indicate what happened with or in response to the event, and not if this is different from what happened before the event. This limitation could be addressed in future research by including a systematic comparison of the framing of specific topics before and after an event (see Kriesi and Pappas, 2015;Legewie, 2013;Woods and Arthur, 2014). Second, the approach does not study indirect impacts or implicit references. An event like 9/11 can have knock-on events which have fallen outside of the scope of our study and actors may have been significantly affected by it but not explicitly mention it. Although these are important limitations, the advantage of our inductive approach compared to existing event research (e.g., Álvarez-Benjumea and Winter, 2020;Berezin, 2009;de Vries, 2017;Fetzer and Soper, 2003;Larsen et al., 2020;Sander and Putnam, 2010;Semetko et al., 2003) is that it takes references and associations by actors as its starting point and thus reduces reliance on conjecture and inference.
Next to these methodological advantages, our analyses also indicate that the approach is a fruitful event research strategy because it can disclose interesting empirical and theoretical insights. Empirically, our analyses show that, on the one hand, framings of 9/11 have resulted in uniform and stable associations pertaining to foreign affairs (with strong links to the War in Afghanistan, the Iraq War, and terrorism). On the other, there has been cross-national variation in framings around domestic affairs: 9/11 has become a 'safety event' in the US and an 'Islam event' in the Netherlands. In France, it has not been associated with domestic affairs, and has turned into a 'non-domestic event.' The results for the US and France are especially significant. The minimal connection in the US between 9/11 and Islam as a domestic affair adds nuance to the assumption that the event catalyzed public stigmatization of American Muslims (Garg et al., 2019;Woods and Arthur, 2014). The fact that 9/11 has hardly been related to French domestic affairs contributes a new theoretical insight to event research: the extent to which an event becomes domesticated can differ substantively depending upon national context. As we find such specific national 'colorings' in event framing, our results corroborate the importance of doing country comparative research of public meaning-making processes (cf. de Vreese et al., 2001;Ferree et al., 2002;Koopmans et al., 2005;. Another insight is that 9/11 has had a long, yet stable life in all three country contexts. Even though attention to the event dropped significantly over the years, we found new peaks in its public significance around the occurrence of comparable events: the Madrid (2004) and Charlie Hebdo attacks (2015). Some event scholars (Sewell, 2005;Swidler, 1986;Zolberg, 1972) distinguish between 'moments of madness' (the period just after an event) and 'normal times' (the period years afterwards). However, our findings underscore that many implications for political and societal change can be attributed to an event long after its occurrence (Wagner-Pacifici, 2010, indicating the relevance of long-term event research. At the same time, our results suggest that, for 9/11, the framings have been more stable than the notion of 'restlessness' (Wagner-Pacifici, 2010) implies: in all three countries, the key issues to which the event was related remained constant over time.
These insights were -of course -only derived from analysis of one event in three national contexts. While 9/11 has been a significant event, to find out whether our findings are generalizable, the various phenomena we discovered -e.g., the domestication of events and their long, stable lives -should be investigated with relation to more events, in varying social contexts, over extended periods of time.
Woods J and Arthur CD (2014) The threat of terrorism and the changing public discourse on immigration after September 11. Sociological Spectrum 34 (5)

Author biographies
Thijs van Dooremalen is a political sociologist whose central research interest is to contribute to the social scientific thinking and analysis of events. In 2019, he received his PhD in Sociology from the University of Amsterdam. He is currently working as a postdoctoral research fellow at the Weatherhead Center for International Affairs (Harvard University) and as a lecturer at Leiden University's Institute of Political Science.
Justus Uitermark is a sociologist and geographer who studies how different kinds of -online and offline -environments shape political and cultural conflict. He currently works as a professor of urban geography at the University of Amsterdam and is affiliated with the UvA's Center for Urban Studies as well as the Amsterdam Institute for Social Science Research.