This paper was presented at the Association for Education in Journalism and
Mass Communication in San Antonio, Texas August 2005.
         If you have questions about this paper, please contact the author
directly. If you have questions about the archives, email
rakyat [ at ] eparker.org. For an explanation of the subject line, 
send email to
[log in to unmask] with just the four words, "get help info aejmc," in the
body (drop the "").

(Feb 2006)
Thank you.
Elliott Parker
====================================================================

A Content Analysis of Newspaper Coverage on the Sept 11 Attacks, Bali 
Bombing and Madrid Train Bombing

By Tan Li Hoeng Joann, Yang Yanni and Jamaliah Bte Othman,
Nanyang Technological University, Singapore
School of Communication & Information

Contact number: +65 9628 0440
Email: [log in to unmask]
Fax number: +65 6220 4133

Student research paper for the International Communication Division 
of AEJMC 2005 Annual Conference

Abstract
A content analysis of 312 articles from 13 newspapers on the Sept. 11 
attacks, Bali bombing and Madrid train bombing was conducted. The 
study examined which news frames were dominant, and integrated 
international news flow and framing analysis. Across all events, 
"threat to political climate" and "coverage of victims and potential 
victims" were dominant. For Sept. 11, trade and casualty count were 
significant predictors of "threat to economy" and "coverage of 
victims and potential victims" frames respectively.

A Content Analysis of Newspaper Coverage on the Sept 11 Attacks, 
Bali Bombing and Madrid Train Bombing

Student research paper for the International Communication Division 
of AEJMC 2005 Annual Conference

Abstract
A content analysis of 312 articles from 13 newspapers on the Sept. 11 
attacks, Bali bombing and Madrid train bombing was conducted. The 
study examined which news frames were dominant, and integrated 
international news flow and framing analysis. Across all events, 
"threat to political climate" and "coverage of victims and potential 
victims" were dominant. For Sept. 11, trade and casualty count were 
significant predictors of "threat to economy" and "coverage of 
victims and potential victims" frames respectively.
A Content Analysis of Newspaper Coverage on the Sept 11 Attacks, 
Bali Bombing and Madrid Train Bombing

Introduction
Terrorism, a widely discussed and debated issue in the news media 
today, became prevalent as a result of three major attacks: Sept. 11 
in 2001, Bali bombing in 2002, and Madrid train bombing in 2004. The 
first attack was on an international financial hub, the second, a 
tourist haunt, and the third, a place of daily commute. These three 
acts of calculated violence against civilians had immediate and 
overwhelming effects on the international community.
A handful of people in the media saw symbolic meanings and trends 
behind the attacks. An online New York Post article ("Chilling '911' 
Murder Link", March 2004) pinpointed that exactly 911 days stood 
between the events of Sept. 11 and the Madrid train bombing. The 
Straits Times from Singapore, among other newspapers, also pointed 
out that the Bali bombing occurred one year, one month and one day 
after Sept. 11 ("Bali's tourism targets dashed by attacks", October 
2002). In addition, radical Islamist group Al-Qaeda and its Asian 
affiliate Jemaah Islamiah were said to be responsible for the attacks.
The similarities and patterns that link each event provide an 
opportunity for comparisons in the selection of news frames in the 
coverage of the attacks. Yet, the attacks remain separate enough for 
newspapers to approach each event independently. Hence the questions 
of how and what frames were selected in the news coverage?
Past research has suggested that a variety of determinants - ranging 
from journalists' backgrounds to intrinsic characteristics of news 
events to external forces driving journalism – could contribute to 
international news coverage and presentation, albeit their 
inconsistency in predictability across different topics with varied 
nature, scale and operationalisation of concepts. However, much of 
this effort has been devoted to examining the amount or type of news 
story concerning intended nations, instead of how reported issues 
were being framed. The current study attempts to offer additional 
effort to the handful of past investigations that have tried to close 
this research gap.
Such research has been either pure theoretical (i.e. Scheufele, 1999) 
or empirical (i.e. Chang & Chang, 2003; Chang et al., 2004) but with 
limited scope for cross-national comparisons. With increased emphasis 
on terrorism at the turn of the 21st century and the string of high 
profile terror attacks globally, this study aims to investigate how 
the three aforementioned terror events were covered in the 
international press in terms of framing and its processes—namely how 
determinants of international news might have affected framing of the 
three events.
The present research content analysed 312 news articles from the 
world press. With a much larger scale and scope of sample, the study 
aspires to offer some insight to improve the understanding of the 
theoretical and empirical complexity of international communication, 
framing research and, above all, their interactions – an important 
area lacking in research.
	Terrorism is a controversial term with multiple definitions. 
Anderson and Sloan (2002) stated that terrorism evokes an image of 
fear: "This frightening imagery of terrorism in turn has often 
provoked an emotional response equally in the lay observer, the 
scholar, or the policy maker who cannot ignore their gut reaction to 
the threat and acts of bloodshed" (p.1).  Wilkinson (2000) offered 
the definition that "terrorism is the systematic use of coercive 
intimidation, usually to service political ends" (p.12), and is used 
for creation of climate of fear, publicity for a cause, and coercion 
of a target to submit to terrorists' demands.
	The manner of reporting terrorism has since grown in importance due 
to the possibility of manipulation of the media by terrorists. Picard 
(1993) stated that media and media coverage can be considered "modern 
tools of terrorists" (p. 6), that by increasing audience exposure to 
the consequences of the activities, concerns and anxiety about the 
government and its institutions will also increase. In previous 
years, acts of terrorisms were merely ignored or unreported in the 
American media (Kelly and Mitchell, 1981; Crelisten, 1987). However, 
with increasing attention placed on terrorism, the relationship 
between terrorism and the mass media deserves closer scrutiny. It is 
crucial to devise methods to report fairly on such activities without 
diminishing the media's responsibility to the public and affecting 
the duties of law enforcement agencies.	
	This link was further explored by researchers such as Graber (1989), 
who pointed out that media coverage of terrorism involves several 
stages: First, concentration on the dissemination of facts on the 
disaster and victims, and information for people about safety and 
government response; and second, organisation of facts to present a 
logical explanation of the event so as to enforce the view that 
everything is under control and that normalcy measures are in place. 
Competition between media agencies has possibly guided these stages 
(Hoffman, 2002). Having broken the story and captured viewers' 
attention, the priority of the media is to hold that attention with 
equally gripping follow-up reports. The media's focus thus invariably 
shifts from the limited quantity of hard news to more human-interest 
type feature stories (p.138).
	Studies have been done on coverage of the Sept. 11 attacks (Traugott 
and Brader, 2002; Moore and Pan, 2002). Traugott and Brader (2002) 
found that there was an initial emphasis on the details of the 
incidents. This was followed by government-initiated reports of their 
response after two days, with an eventual third phase of background 
reporting focusing on "explaining" the event. Moore and Pan's (2002) 
study found that terrorists did alter the emphasis of the U.S. print 
media, with greater status given to the Palestinian and Muslim 
causes. Apparently, certain factors have led to changes—or framing—in 
news content in media coverage of such world events.

Literature Review
Research on international news flow has been varied, dealing with 
issues such as the vast disparity in the amount of foreign news in 
various countries, or differences in news origin. One commonly 
examined aspect involves determinants of international news coverage.
Determinants of foreign news coverage have been categorised into 
either event-oriented or context-oriented factors (i.e. Hester, 1973; 
Chang, Shoemaker and Brendlinger, 1987; Chang, 1998). Event-oriented 
determinants refer to the inherent characteristics of the news event, 
while context-oriented determinants refer to factors external to the 
event. Some determinants have been shown consistently to be good 
predictors of news coverage, while others have produced mixed results.
One possible determinants of international news flow is distance, of 
which includes geographic distance and cultural distance. Geographic 
distance refers to the physical distance between the news origin and 
the reporting nation (Kariel and Rosenvall, 1984). While some studies 
concluded that geographic distance is not a significant determinant 
of international news flow (i.e. Hicks and Gordon, 1974; Chang et 
al., 1987; Chang and Lee, 1992), others have found it to be 
significant (i.e. Van Belle, 2000; Kariel and Rosenvall, 1984; Wu, 
2003) and are in line with Rosengren's (1970) hypothesis that "the 
more distant the event, the more unpredictable and less important it 
seems" (p.79).
The inconsistency may have resulted from a number of reasons. First, 
the inclusion of the U.S. (Kariel and Rosenvall, 1984) may have 
contributed to the difference. Second, different dependent variables 
were used in the studies. Hicks et al. (1974) tested geographic 
distance against prominence of coverage (which is an index of number 
of items, amount of space allocated, and the size headlines used); 
and second, amount of straight, negative, and elite-oriented 
coverage. Other content analyses tested external factors against mere 
publication of events. A third reason is the different 
operationalisation of independent variables; for instance, measured 
in thousands of kilometers (Van Belle, 2000), coded dichotomously 
(Chang et al., 1987), using Air Force distance charts or perceived 
distance. Finally, a variation in sample selection may have 
contributed to the differences. Van Belle (2000) concentrated on 
disaster coverage instead of foreign news coverage, thus removing 
another common platform for comparison between studies.
Cultural distance is measured in terms of language affinity (Pei, 
1960) between countries (i.e. Hester, 1973; Chang et al., 1987; Chang 
et al. 1992). Language affinity is a possible factor because firstly, 
language difference could be a technical impediment to news 
transmission; and secondly, language is an integral aspect of 
culture, thus countries sharing the same family of languages would 
generally be closer to one another culturally.
  Besides distance, trade is another possible determinant of 
international news coverage. Hester (1973) argued that strong trade 
links would purportedly mean an increased flow of information between 
the countries. However, results from past studies vary: from finding 
trade to be one of the most important factors influencing news 
coverage (Wu, 1998; 2003), to being a minor, albeit significant, 
factor (Kariel and Rosenvall, 1984; Golan and Wanta, 2003), and to 
being irrelevant as a factor of international news coverage (Chang et 
al., 1987; Chang and Lee, 1992; Robinson and Sparkes, 1976).
These differences could be due to variance in measurement of trade in 
the above studies. For example, Chang et al. (1987) used ordinal 
ratings of 1 to 3 to represent similarity of economic systems between 
countries; U.S. trade relations were merely a factor of perception 
for the news editors in the survey by Chang and Lee (1992); trade was 
measured at the ratio level in the study of Kariel and Rosenvall 
(1984); and Golan and Wanta (2003) measured economic ties as the 
ranking of the nation on the U.S. list of trading partners. Given 
these disparities, a direct comparison of results cannot be made. In 
addition, a time difference of 30 years exists between earlier 
studies (i.e. Hester, 1973) and more recent ones (i.e. Wu, 2003). 
With rapid globalisation within the last decades, it is plausible 
that the importance of world trade has changed greatly in various 
countries over this period.
Another factor to be considered is the gross national product (GNP) 
per capita of countries. GNP per capita could be an influence on news 
coverage because of its close links with the development and wealth 
of a country. Hester (1973) and Chang (1998) argued that more 
information would flow from powerful to less powerful nations.
Findings regarding the influence of GNP per capita on international 
news flow vary, as has been in the case of trade. It has been found 
to be an important determinant among a list of possible factors (Kim 
and Barnett, 1996), or to have a significant yet minor and limited 
influence on international news flow (Kariel and Rosenvall, 1984; Wu, 
2003). Other studies, however, concluded that GDP per capita are 
insignificant or irrelevant as a determinant of international news 
coverage (Van Belle, 2000; Chang and Lee, 1992).
The differing results could be due to various reasons, such as 
differences in countries studied (the U.S. versus Canada), data 
source (World Development Report versus The Penn World Table versus 
International Bank of Reconstruction and Development), and content 
(disaster news versus foreign news).
In addition to the abovementioned contextual factors, 
content-oriented factors, or characteristics of the event itself, 
such as conflict, timeliness, or impact, can make it newsworthy.
Impact can be defined in terms of casualty level. Itule and Anderson 
(2003) stated that "readers are always interested in stories that 
have considerable impact on their communities" (p.15). In support, 
Van Belle (2000) found that the number of people killed is 
statistically significant in influencing coverage, as did Chang et 
al. (1987) who concluded that U.S. editors considered loss of lives 
and properties to be important determinants. This is because an event 
with a higher casualty level has greater impact, and hence greater news value.
According to Goffman (1974), people actively organise and categorise 
life experiences to make sense of them. Frames are the individual's 
mentally stored clusters of information or the "schemata of 
interpretation" (p. 21) that guide individuals in the construction of 
information. Drawing upon that psychological definition, Goffman 
(1974) elaborated on how frames are "the principles of organization 
which govern events – at least social ones - and our subjective 
involvement in them." (p. 10-11). Later research may have offered 
other definitions of frames, but the idea of frames as means of 
organising information was shared among others (Tuchman, 1978; 
Gamson, 1989; Entman, 1993; Norris, 1995). Entman (1993) defined 
framing as "to select some aspects of a perceived reality and make 
them more salient in a communicating text", so as to fulfill four 
main functions of "problem definition, causal interpretation, moral 
evaluation, and treatment recommendation for the item described" (p. 
52). However, he qualified that sentences in the content may perform 
more than one of the four framing functions, and in some instances, a 
frame may not even include any of the four functions.
With regards to the production of news, Gitlin (1980) employed 
Goffman's (1974) psychological concept of frames in his study of the 
relationship between the news media and the New Left movement. He 
defined frames as the "persistent selection, emphasis, and exclusion" 
of news and that frames "enable journalists to process large amounts 
of information quickly and routinely package the information for 
efficient relay to their audiences" (p. 7).
Researchers such as Gamson (1989) indicated that facts alone are 
meaningless. Journalists and news workers employ frames to give facts 
intrinsic meaning by putting them into a context that provides 
organisation and meaning. To Gamson, a frame is "a central organising 
idea or story line that provides meaning" (Gamson and Modigliani, 
1987, p. 143).
Tuchman (1978) talked about how "frames organise 'strips' of everyday 
life and transform them into defined events" (p.7). She argued that 
news organisations are complex organisations subject to certain 
inevitable processes. Journalists thus have to make decisions as 
professionally as possible based on newsroom needs. Tuchman (1978) 
thus linked Goffman's (1974) study to the ideological and structural 
processes of journalism, news organisations and their sources, who 
help structure "social reality".
Relevant to the current study are investigations of framing research 
applied to terrorism, as conducted by Berkowitz and Gavrilo (2001), 
Baden (2001), Chang and Chang (2003), and Eckstein (2003). Berkowitz 
and Gavrilo's (2001) empirical study of the American framing of 
terrorism in Israel stated that news was shaped by three news 
cultures – that of the journalism profession, news organisation, and 
society at large. They supported their argument through an analysis 
of news articles in The New York Times and USA Today, and found that 
news stories tend to reinforce current cultural foundations of 
society, reaffirming what the public has come to stereotype. 
Eckstein's (2003) analytical study of the news frames on Sept. 11 
attacks and 1995 Oklahoma City bombing concluded that for both 
attacks, the events were framed more episodically than thematically.
Similarly, Baden (2001) suggested that The New York Times tend to 
frame terrorism coverage with a more U.S. official or 
military-positive tone, while the Guardian of London framed its 
coverage with a more foreign official-negative tone. Chang and 
Chang's (2003) study on world editorials on Sept. 11 terrorist 
attacks showed that most world press framed its editorials to promote 
"the importance of an international coalition to stand by the United 
States" and "to share the responsibility to counter terrorism" (p. 56).
	As earlier studies (i.e. McCombs and Shaw, 1972) have shown, the 
news media play an important role in the news consumers' setting of a 
political agenda. In the realm of terrorism, the majority of the 
public has little access to information regarding terrorist 
developments and dealings other than through the media. Hence, media 
scholars argue that news media have the power to shape public opinion 
and understanding on topics about which they are ignorant.
	An examination of media frames in the coverage of the three separate 
terror attacks by newspapers in 13 countries provides a good 
opportunity to observe how different newspapers framed the issue 
within their respective political, economic and social milieus. In 
the study, framing will be used as a dependent variable (Scheufele, 1999).	
	Despite the little research that has been done to examine the 
possible link between framing analysis and the influence of factors 
on news coverage, there has been some inkling of how such a 
connection might be established in recent studies (Scheufele, 1999; 
Chang and Chang, 2003; Chang, et al., 2004).
	Scheufele (1999) outlined five factors that suggest an influence on 
media framing: social norms and values, organisational pressures and 
constraints, pressures of interest groups, journalistic routines, and 
ideological or political orientations of the journalist. An example 
is that population, in terms of racial makeup of a community, can 
influence the framing of news (Goshorn and Gandy, 1995; Gandy, et al., 1997).
	Chang and Chang (2003) integrated international news flow analysis 
with framing analysis in their study of the Sept. 11 terrorist 
attacks. They examined if context-oriented factors such as trade, 
diplomatic and military ties would affect news frames employed by 
international newspaper editorials focusing on the attacks. Results 
showed that religion was an important determinant in how different 
countries' newspapers frame international editorials on the terror 
attacks. In countries where Christianity or Hinduism was dominant, 
editorials tend to adopt the frame of "calling for an international 
coalition to counter terrorism". The study also indicated that 
factors like press and political freedom of countries were likely 
predictors of expressed kinship with the United States.
	A study by Chang et al. (2004), which examined the influence of 
contextual factors on the selection of news frames for cross-national 
news coverage of Severe Acute Respiratory Syndrome (SARS), found a 
positive correlation between trade and framing of SARS as a "threat 
to economy". However, other contextual factors like political factors 
and foreign direct investment did not have any significant effect on 
news framing.
In the same vein, this study attempts to link international news flow 
theory with framing analysis. Factors such as trade, geographic 
distance, number of casualties, proportion of Muslims in country, 
number of terror attacks in the past, and gross domestic product per 
capita will function as independent variables to determine the 
possible differences in framing of the three terror attacks in 
international newspapers.

Research and Hypotheses
Based on the literature review on international news flow and framing 
analysis, several factors of foreign news coverage were selected for 
investigation as to how they might influence framing of the three 
terror attacks. The following research question and hypotheses were 
thus proposed.
	RQ1: What is the dominant news frame in the articles on the three 
terror attacks?
While previous studies on foreign news coverage and terror attacks 
have focused on the amount of coverage received in different 
countries, few have examined the types of frames dominating coverage 
of the attacks. This study aims to find out what is the dominant news 
frame used in the coverage of the three terror events.
H1: 	The larger the trade volume, the more dominant the "threat to 
economy" frame will be.
	Earlier studies have revealed varied results about the influence of 
trade volume between countries on the amount of media coverage. The 
hypothesis proposes that when trade volume between countries is high, 
a terror attack in one country is likely to have greater economic 
repercussions on its trading partners when market operations are disrupted.
H2: 	The higher the GDP per capita of the country, the more dominant 
the "threat to economy" frame will be.
	The gross domestic product per capita of a country is a measure of 
its wealth and economic development. A terror attack can have 
international ramifications, hence a stronger economy rather than a 
weaker one, is more likely to be affected by a global event.
H3: 	The higher the number of casualties from the country in the 
attack, the more dominant the "coverage of victims and potential 
victims" frame will be.
	Literature review has shown that loss of lives and property are 
significant determinants of news coverage. This study attempts to 
take a deeper look at the influence of casualty level on news 
coverage by proposing that a country with a higher casualty rate from 
a terror attack is more likely to cover the event from the victims' 
perspectives. As the impact of the event is greater on the country, 
newspapers are expected to carry updates of the victims to their 
families, friends, and fellow citizens back at home.
H4: 	The closer the country is to the victim country, the more 
dominant the "coverage of victims and potential victims" frame will be.
Countries in close proximity to each other are more likely to feel 
closer psychologically to their neighbours, than to countries further 
away. This can be due to greater travel and movement between 
neighbouring countries, or due to the more similar cultures between 
them. Thus, they are likely to include more coverage on victims.
H5:	The higher the proportion of Muslim population in the country, 
the more dominant the "coverage of perpetrators" frame will be.
	In this study, perpetrators of the three terror attacks are radical 
Muslims of Al-Qaeda and Jemaah Islamiah. Even as the attacks are 
relevant to the rest of the world, they are undoubtedly more relevant 
directly to the Muslim population than those of other religious 
groups. Thus, it is proposed that countries with a higher proportion 
of Muslims in the country are more likely to cover the attacks by 
shedding more information on the perpetrators, who are differentiated 
from typical Muslims as the extremists in the religious group.
H6:	The larger the number of terror attacks that has occurred in the 
country, the more dominant the "call for aggressive international 
action against terrorists" frame will be.
	At least one previous study has indicated that social distance is a 
significant factor influencing coverage of major disasters. It is 
likely that a country that has experienced terror attacks before can 
better empathise and understand the consequences of the terror attack 
in the victim country, and thus is socially closer than one that has 
never been through such attacks. Therefore, it is likely that the 
former is more likely to call for punitive actions against the 
perpetrators to address the injustice.

Method
	A purposive sample of the following 13 newspapers was selected: The 
Straits Times (Singapore), The New Straits Times (Malaysia), The 
Jakarta Post (Indonesia), The Bangkok Post (Thailand), China Daily 
(China), The Daily Yomiuri (Japan), Sydney Morning Herald 
(Australia), Globe and Mail (Canada), The New York Times (U.S.), The 
London Times and Sunday Times (U.K.), The Jerusalem Post (Israel), 
Frankfurter Allgemeine Zeitung (Germany) and El Pais (Spain). No 
articles were taken from The New York Times, The Jakarta Post and El 
Pais for Sept. 11 attacks, Bali bombing and Madrid train bombing 
respectively, as the study is looking at influences on and framing of 
foreign events.
The sample was chosen because they constitute a good mix of 
newspapers from different regions. Importantly, they are among the 
leading daily newspapers in their countries of origin in terms of 
circulation and influence and written in the English-language with 
the exception of the newspapers from Spain and Germany. This is 
because the leading and most influential newspapers in these two 
countries are in their native tongue, thus the use of the 
German-language newspaper Frankfurter Allgemeine Zeitung and the 
Spanish-language daily El Pais.
	News articles published in the first seven days after the terror 
attack made up the sample frame. This time frame allowed for 
sufficiently significant coverage of the event and was also long 
enough for the newspapers to explore the different story angles of 
the attacks. Therefore, the time periods from which the news articles 
were taken are: Sept. 12 - 18, 2001 (for Sept. 11 attacks); Oct. 13 - 
19, 2002 (for Bali bombing); and March 12 - 18, 2004 (for Madrid 
train bombing). This resulted in a total of 21 days for content 
collection and analysis.
	The news sample was retrieved using Factiva®[1], and the search 
terms used were as follows: for Sept. 11 attacks, "terror" or "World 
Trade Center"; For the Bali bombing, "terror" or "Bali" were used; 
and for Madrid train bombing, "terror" or "Madrid". This preliminary 
search resulted in 3,409 articles.
Irrelevant articles such as editorials, interviews, and paid notices 
among many others, were filtered out, resulting in a sample list of 
1,449 relevant articles. Further sifting of articles includes 
selecting only articles where the attack or consequences arising from 
it is mentioned in the first five paragraphs. According to Fico and 
Cote (1999), there is an assumption that material higher up in 
articles will have a greater likelihood of being read. In their 
study, they used this assumption to create an index of how equally 
assertions by political candidates were made from the first through 
the fifth paragraphs.
 From the final sample, disproportionate, stratified, random sampling 
was conducted to obtain 312 newspaper articles for analysis, using a 
computer randomiser. This approach was taken to ensure that all 
newspapers would be represented in the final sample. The amount of 
coverage of the attacks in the different newspapers varied greatly, 
and to use proportionate sampling would mean newspapers with very 
little coverage would not be represented at all. Of all, China Daily 
had the smallest number of articles (2) and The London Times or 
Sunday Times had the largest number of articles (82).
Intercoder reliability testing was conducted between the three 
coders. Using paragraph as the coding unit and article as the unit of 
analysis, 60 English-language articles were randomly selected from 
the final sample of 312 articles for the test. All three coders coded 
the 60 articles. Each of their results was then correlated against 
the other two coders, using Pearson's Product Moments. This resulted 
in three sets of correlations, each set testing a possible pairing 
between the three coders, which on average yielded an acceptable 
range of 86.7 to 100 percent.
Non-English language articles were excluded from test sample due to 
the lack of proficiency of the three coders. Two interpreters were 
employed to assist in the coding of the non-English language articles 
– one proficient in Spanish and another in German. While each 
translated the articles verbally by paragraphs, one of the three 
initial coders was simultaneously coding. This was done over 
face-to-face meetings to ensure that all doubts were clarified.




Factors influencing coverage (independent variables)
	Distance .Distance is defined as the geographic distance between the 
capitals of each victim and reporting country. It is measured in 
kilometers. Figures were derived from the website www.indo.com/distance.
	GDP per capita. Gross domestic product per capita is defined as the 
market value of all the goods and services produced within the 
borders of a nation during a specified period, per person. It is 
measured in U.S. dollars, and the figures were derived from the 
Global Market Information Database, published by Euromonitor at 
http://www.euromonitor.com/gmid/default.asp. Three sets of data were 
used as the three terror attacks occurred in different years. GDP 
figures are generally tabulated at the end of the year; hence for a 
more accurate measure of the country's economy before it was affected 
by the terror attacks, the previous year's results were used. 
Therefore, year 2000, 2001 and 2003 figures were used for the Sept. 
11 (2001), Bali bombing (2002) and Madrid train bombing (2004), respectively.
Trade. Trade is defined as the volume of trade between each victim 
country and reporting country. As is the case with GDP per capita, 
three sets of data were derived from the Global Market Information 
Database, published by Euromonitor at 
http://www.euromonitor.com/gmid/default.asp. Using data of trade 
volume prior to the attacks are important, so that the figures 
reflect trade relations more accurately before countries are affected 
by the attacks. For the current study, trade figures are given in 
U.S. dollars, in billions.
	Number of terror attacks. Number of terror attacks is defined as the 
total count of terror attacks known to have occurred in the 
countries, up to the point of each terror attack. Information was 
derived from the Terrorism Research Center at www.terrorism.com. 
Three sets of data were used in the analysis of each event: number of 
terror attacks in a country up till 10 Sept. 2001 (for Sept. 11), up 
till 11 October 2002 (for Bali bombing), and up till 10 March 2004 
(for Madrid train bombing).
	Proportion of Muslim population. Proportion of Muslim population is 
defined as the number of Muslims in each country as a percentage of 
the total population. It is thus a ratio measure and figures were 
derived from Muslim Population Worldwide at 
http://www.islamicpopulation.com/index.html.
The major sources of this website are the Population Reference 
Bureau, the CIA Fact Sheet, the U.S. State Department, various 
country reports and news agencies.
	Casualty level. Casualty level is defined as the number of citizens 
of each country killed in each attack. Figures were derived from 
Wikipedia, The Free Encyclopedia, at http://en.wikipedia.org. and 
http://www.Sept.11victims.com/Sept.11victims/COUNTRY_CITIZENSHIP.htm.

Frames (dependent variables)
	A set of 12 frames were determined and coded accordingly in the 
news. They are "threat to economy", "threat to political climate", 
"threat to religious or racial harmony", "coverage of victims and 
potential victims", "coverage of perpetrators", "show of sympathy or 
empathy", "call for aggressive action against terrorists", "call for 
non-aggressive or undecided action against terrorists", 
"international aid to victim countries", "aid within victim 
countries", "others" and "Not Applicable".
	The coding unit is a paragraph. Not all frames were used as 
dependent variables. Those used were namely "threat to economy", 
"coverage of victims and potential victims", "coverage of 
perpetrators", and "call for aggressive action against terrorists", 
"Call for non-aggressive or undecided action against terrorists".

Statistical procedure
	Descriptive statistics were used to answer the research question. To 
test the hypotheses, factors influencing coverage (independent 
variables) were run against selected news frames (dependent 
variables), using multivariate regression analysis. Hypotheses 1 and 
2 share the same dependent variable ("threat to economy" frame), as 
do Hypotheses 3 and 4 ("coverage of victims and potential victims" 
frame). Hence, for each event, four multivariate regressions were 
run. Analysis of the three terror attacks thus resulted in a total of 
12 multivariate regression analyses.
This study examines possible factors affecting the dominance of 
selected news frames. The concept of dominance is measured using an 
index made up of the following constructs: (1) frequency of the frame 
used, (2) placement of the frame by paragraph, and (3) placement of 
the frame of direct quotes, also by paragraph.
Before the index is created, variables measuring placement had to be 
recoded into a continuous variable. This is done by applying the 
equation [N – (n – 1)]/N, where N refers to the total number of 
paragraphs in each article and n refers to the paragraph number in 
which each of the 12 frames was used for the first time (Chang, 
2003). Scores greater than 1, which is derived when n=0, are 
converted to 0. Therefore, the converted scores range from 0 to 1; 
the earlier in the article a frame first appears, the greater the score.
Reliability analysis was then conducted on the standardised data set, 
with the variables being correlated against other variables forming 
the index of dominance. All variables for a possible index had to 
have acceptable correlations of greater than 0.70 and less than 0.95.
The results were as follows: Threat to economy (0.908), threat to 
political climate (0.884), threat to religious and racial harmony 
(0.916), coverage of victims and potential victims (0.881), coverage 
of perpetrators (0.851), show of sympathy or empathy (0.883), call 
for aggressive action against terrorists (0.893), call for 
non-aggressive or undecided action against terrorists (0.916), 
international aid to victim countries (0.843), aid within victim 
countries (0.896), others (0.872), non-applicable (0.828). The lowest 
alpha was for the "N.A." frame, while two frames "Threat to religious 
or racial harmony" and "Call for non-aggressive or undecided action 
against terrorists" scored the highest alpha. These results show that 
there is no issue of multicollinearity.
As all correlations fit the criteria, each of the three variables of 
each frame were aggregated then averaged to form a new variable – 
dominance of each frame. This variable would be used in the later 
statistical procedures.

Results and Discussion
The research question sought to examine which news frames were most 
dominant for each event. Results show that each terror attack appears 
to have different dominant news frames. However, a closer examination 
reveals that two frames are consistently found in the top three 
dominant news frame throughout all three attacks.
Results from Table 1 (see Appendix A) showed that the most dominant 
frame for the Sept. 11 attacks was the "threat to economy" frame 
(mean = 1.322). This was followed by "coverage of victims and 
potential victims" (mean = 0.959), and "threat to political climate" 
(mean = 0.828).
The Sept. 11 terrorist attacks struck the heart of a bustling 
commercial hub which was home to almost 500 U.S. and foreign 
companies in 2001. Much commercial activities and businesses were at 
stake, and the consequences of the attack were economically crippling 
worldwide. Thus, it is not surprising that an extensive amount of 
international media coverage would be devoted to the economic effects 
of the attack, and hence the use of the "threat to economy" frame. It 
was noted by Graber (1998) that the media often disseminated facts on 
the disaster and victims in the first stage of terrorism coverage, 
and by Hoffman (2002) that during crisis reporting, the media's focus 
invariably shifts from hard news to more human-interest type feature 
stories. With more than 3,000 casualties involved in the Sept.11 
attacks, the media would place a large emphasis on priming the 
"coverage of victims and potential victims" frame – hence accounting 
for its ranking as the second-most dominant frame. The "threat to 
political climate" frame ranked third. Its dominance is also 
justified by Graber (1998) who mentioned that information about 
safety and government response is also included in the first stage of 
media coverage of terrorism. The dominance of the "threat to 
political climate" news frame can also be explained through the 
observations of Traugott and Brader (2002) who noted that 
government-initiated reports of their response towards the terrorists 
would surface two days after the attack.
For the Bali bombing, the "coverage of victims and potential 
victims", with a mean of 1.698, was the most dominant frame. This was 
followed by the "threat to political climate" frame (mean = 1.032), 
and the "threat to economy" frame (mean = 0.719).
The dominance of "coverage of victims and potential victims" frame 
could be expected since almost 200 casualties from 22 nations were 
involved in the explosions at the nightclubs - both popular tourist 
haunts. Based on Graber (1998) and Hoffman's (2002) discussions on 
human-interest news, newspapers would seek to contrast the holiday 
intentions of the tourists with the aftermath of the bombing. Reports 
of the attacks could have also focused on victims' accounts of horror 
and panic inside and outside the bars, and the acts of individual 
heroism. The "threat to political climate" is the second most 
dominant news frame. A possible explanation is the demands from 
foreign governments on their Indonesian counterpart to intensify 
crackdowns on terrorists (i.e. The Jakarta Post, "Japan turns up 
pressure on Mega to fight terror", October, 2002). Additional reports 
on government investigations of terrorists may have increased 
dominance of the "threat to political climate" news frame. The third 
most dominant frame is the "threat to economy" possibly because the 
bombing had a major impact on Indonesian's tourism industry as it 
occurred at the favourite tourist haunt of Bali. Bali accounts for 
one-third of Indonesia's tourism earnings and the attack had not only 
resulted in worries over the nation's economy but also Indonesia's 
trade ties with other countries. (i.e. The Jakarta Post, "Analysts 
cut GDP growth predictions after Bali terror", October, 2002). Hence, 
it could be expected that other countries and foreign media would be 
anxious about the economic effects of the attacks.
For the Madrid train bombing, the most dominant frame was "threat to 
political climate" with a mean of 1.425. This was followed by 
"coverage of perpetrators" (mean = 1.003), and "coverage of victims 
and potential victims" (mean = 0.530).
The rationale for the "threat to political climate" being the most 
dominant news frame could be due to the ousting of the then incumbent 
Spanish government in 2004 - the most immediate consequence of the 
train attacks. After the train bombing, the Spanish people were upset 
with the government: first, for misleading the country into thinking 
that ETA—the country's separatist group—was responsible for the 
attack, and second, for believing that the government's involvement 
in the Iraq war with the U.S. made Spain the target of terrorists. 
Consequently, the then incumbent party lost the country's election to 
Spain's Socialist Party. This change in political climate was labeled 
as "bin Laden's first regime change" (The Jerusalem Post, "London 
attack 'inevitable'", March, 2004). With much focus on the elections 
and resultant change in ruling party, "threat to political climate" 
is thus the most dominant frame of coverage. The second most dominant 
frame is "coverage of perpetrators". As mentioned, the Spanish people 
felt the government had deceived the people by putting the blame on 
ETA initially and yet evidence subsequently pointed the finger at 
Al-Qaeda instead. Thus, there was much news coverage on ETA and 
Al-Qaeda as suspects for the attacks, making "coverage of 
perpetrators" one of the top three dominant frames. The "coverage of 
victims and potential victims" ranks third in frame-dominance for the 
Madrid train bombing. Journalism practice and its criteria for 
newsworthiness is a possible explanation for its prevalence. Human 
interest is one of many criteria of newsworthiness because putting a 
face to news descriptions draws readers' interest and concern in 
events. In major events, as in the three terror attacks examined in 
this study, the "coverage of victims and potential victims" would no 
doubt be an important frame.
Overall, the results indicated that the "threat to political climate" 
and the "coverage of victims and potential victims" were consistently 
among the top three dominant frames for all three terrorist attacks. 
The dominance of the "threat to political climate" frame could be 
attributed to the need for government assistance and response right 
after the occurrence of disasters. Since only government and related 
organisations would have the means and resources for large-scale 
assistance, citizens would naturally look towards them for help. 
Another possible immediate response by the public would be to blame 
the government for the attacks by questioning national security and 
the policies of the government. Given these responses, coverage of 
the attacks within the first week can be expected to be predominantly 
on government relief efforts, as well as government assurances of 
national security and policies.
As with journalism guidelines that human-interest news is newsworthy 
and attention grabbing, most media would choose this form of soft 
news as follow-up coverage after dissemination of facts on the 
attack. Thus, it is a logical finding that the "coverage of victims 
and potential victims" frame is one of the most prevalent news angles.
Hypothesis 1 (H1) predicted that the trade volume of a country would 
be positively correlated with the "threat to economy" frame, and 
hypothesis 2 (H2) predicted that the GDP per capita of a country 
would be positively correlated with the "threat to economy" frame.
	Results from Table 2 (see Appendix A) showed that for Sept. 11 
attacks, H1 was supported (p = 0.023, B = 0.008), but not H2 (p = 
0.393, B = -8.420E-06). The negative B-coefficient of GDP per capita 
in H2 also meant that the smaller the GDP per capita of the country, 
the more dominant the "threat to economy" frame will be. For Bali 
bombing, H1 was not supported (p = 0.144, B = -0.107). In fact for 
H1, trade would be negatively correlated with the dependent variable 
if it was significant because of its negative B-coefficient. H2 was 
not supported either (p = 0.078, B = 7.301E-05). Distance was the 
only significant predictor of the news frame (p = 0.007, B = 
-1.570E-04), which meant that the closer the country is to Indonesia, 
the more dominant the "threat to economy" frame will be. For the 
Madrid train bombing, H1 was not supported (p = 0.214, B = 0.007). 
However, H2 was supported (p = 0.022, B = 6.075E-05). The proportion 
of Muslim population was also a significant predictor of the 
dominance of the news frame (p = 0.009, B = 0.028), which implied 
that a larger proportion of Muslim population in a country would lead 
to a more dominant "threat to economy" frame.
For the Sept. 11, H1 was supported, which is in line with previous 
international news flow studies (i.e. Wu, 1998; 2003; Kariel and 
Rosenvall, 1984) that found trade to be a primary predictor of news 
coverage. An attack on the financial capital of a global economic 
superpower has repercussions on international economies. It is likely 
that countries with large import and export trade volumes with the 
U.S. would be greatly concerned about how the attacks would affect 
them, thus devoting a hefty amount of coverage to the "threat to 
economy" frame.
	On the other hand, H1 was not supported for Bali bombing and Madrid 
train bombing. An overview of previous studies had suggested that 
trade is an inconsistent factor influencing news flow (Robinson and 
Sparkes, 1976). Differences in the measurement of trade could also 
account for the inconsistency. A study by Chang et al. (2004), which 
found a positive correlation between the size of a country's economy 
and the frame "threat to economy" had used a combined index of trade, 
economic development and foreign direct investment figures as a 
measurement of the size of the economy. The present research however, 
uses a country's import and export figures with the victim countries 
as a measurement of trade, and thus a direct comparison cannot be 
made with the former study.
H2 was supported for the Madrid train bombing, but not for the Sept. 
11 attacks and Bali bombing. This finding is similar to conflicting 
conclusions by previous researchers (Van Belle, 2000; Wu, 2003). 
Although insignificant, GDP per capita's negative correlation with 
the news frame for the Sept. 11 attacks could be due to China and 
Indonesia's low GDP per capita data. The exclusion of U.S., with a 
high GDP per capita data, from the Sept. 11 study, coupled with China 
and Indonesia's low GDP per capita figures, could have decreased the 
overall GDP per capita for the 12 countries dramatically, hence 
leading to a negative correlation between GDP per capita and "threat 
to economy" frame.
For Bali bombing, distance was a significant predictor of the "threat 
to economy" news frame. This is probably because countries closer to 
Indonesia, especially those in ASEAN, tend to have closer economic 
ties; hence their newspapers would devote more coverage to this news 
frame. Another possible reason for the results is the tendency for 
Southeast Asian governments to place heavy emphasis on the economic 
interests of their countries over other concerns. A terror attack on 
an important trading partner would most likely trigger huge concerns 
about how affected their economy would be.
For the Madrid train bombing, the proportion of Muslim population was 
also a significant predictor of the dominance of "threat to economy" 
news frame. An explanation for this result could be due to the 
assumption that a country with a larger proportion of Muslim 
population would harbour a larger number of Islamic fundamentalists 
and sympathisers. The presence of such "radicals" could lead to 
copycat terror attacks, which would have devastating effects on the 
economy. As a result, countries with large proportions of Muslim 
populations fear that foreign companies would hold such an assumption 
and withdraw valuable investments, thereby affecting their economy. 
Thus media coverage could be devoted to how these situations would be 
a "threat to economy".
Hypothesis 3 (H3) predicted that the number of casualties would be 
positively correlated with the "coverage of victims and potential 
victims" frame, and hypothesis 4 (H4) predicted that distance would 
be negatively correlated with the "coverage of victims and potential 
victims" frame.	
Results from Table 3 (see Appendix A) showed that for the Sept. 11 
attacks, H3 was supported (p = 0.007, B = 0.022), but not H4 (p<0.01, 
B = 4.521E-04), as results indicated that the further the country is 
from the U.S., the more dominant the "coverage of victims and 
potential victims" frame. Table 3 also showed that trade was 
positively correlated with the dependent variable (p = 0.005, B = 
0.010). Proportion of Muslim population was also a significant 
predictor but the negative B-coefficient suggests that a smaller 
proportion of Muslim population in the country would lead to more 
dominance of the news frame for Sept. 11 attacks (p<0.01, B = 
-0.076). For Bali bombing, neither H3 (p = 0.094, B = 0.014) nor H4 
(p<0.01, B = 3.909E-04) were supported for the Bali bombing. In the 
case of H4, distance was significant but ran in opposite direction to 
the hypothesis. Table 3 also showed that trade volume was a 
significant predictor (p = 0.002, B = 0.442), which meant that the 
larger the trade volume, the more dominant the "coverage of victims 
and potential victims" frame will be. For Madrid bombing, H3 cannot 
be applied because the casualties involved did not include citizens 
of the 12 other countries in this study. However, H4 was supported (p 
= 0.024, B = -1.390E-04).
That H3 was supported only for the Sept. 11 attacks could be due to 
the fact that while more than 3,000 people from 36 nations died in 
the Sept. 11 attacks, the casualty count and nations involved in the 
Bali bombing (almost 200 deaths from 22 nations) and the Madrid train 
bombing (191 people from 14 nations) were comparatively lower. The 
findings correspond to the research by Van Belle (2000) who found 
that the number of people killed is statistically significant in 
influencing coverage, and Chang et al. (1987) who concluded that U.S. 
editors considered loss of lives and properties to be important determinants.
H4 was supported only for the Madrid train bombing. While distance 
was a significant predictor for both Sept. 11 attacks (p<0.01, B = 
4.521E-04) and Bali bombing (p<0.01, B = 3.909E-04), their positive 
B-coefficient values indicated an opposite direction proposed by the 
hypothesis. This seems in line with the mixed results from previous 
research using distance as a determinant. Van Belle (2000), Kariel 
and Rosenvall (1984), and Wu (2003) had shown that geographic 
distance had an influence on news coverage although with varied 
effects while other research (Hicks and Gordon, 1974; Chang et al., 
1987; Chang and Lee, 1992) showed otherwise.
Trade was a significant predictor for the "coverage of victims and 
potential victims" frame for the Sept. 11 attacks and Bali bombing. 
As an international financial hub, the victims who died in the World 
Trade Center were predominantly office workers such as bankers, 
financial analysts and businessmen. The tenants in the World Trade 
Center Twin Towers were also of foreign and local mix; hence a 
country that has assets in the building is most likely to have 
trading ties with the U.S.  A country with a larger trading volume 
with the U.S. would probably have a larger number of citizens working 
in the building. For example, U.K., which had a trade volume of $8.5 
billion with the U.S. in 2000, also had the largest number of foreign 
citizens (67) who were casualties in the attacks. In Bali's case, 
Japan, Australia and Germany had large trade volume with Indonesia in 
2001. The number of casualties involved from each of the three 
countries was also amongst the largest. Hence this can be an 
explanation as to why trade volume is positively correlated to the news frame.
Significant predictors in opposite direction of the dependent 
variable for the Sept. 11 attacks also included GPD per capita and 
proportion of Muslim population. The "coverage of victims and 
potential victims" frame increased as the GDP per capita decreased. 
Once again, the low GDP per capita figures for China and Indonesia 
may have skewed the dominance of the news frame. The inconsistency of 
GDP per capita on coverage may also have contributed to this result. 
Victims and potential victims involved in the attacks were mainly 
non-Muslims. As mentioned in previous studies, the population makeup 
of the community can influence framing of news (Goshorn and Gandy, 
1995; Gandy et al.,1997). Hence countries with smaller proportions of 
Muslim population would see the relevance in devoting coverage to the 
news frame because of race, religion or cultural affinity.
Hypothesis 5 (H5) predicted that the proportion of Muslim population 
in the country would be positively correlated with the "coverage of 
perpetrators" frame.
	As seen from Table 4 (see Appendix A), for the Sept. 11 attacks, H5 
was not supported (p = 0.398, B = -0.003.) It was also not supported 
for Bali bombing (p = 0.365, B = 0.007), and Madrid train bombing (p 
= 0.254, B = -0.016). For the latter, a negative B-coefficient, 
similar to the results of Sept. 11 attacks indicated that a smaller 
the proportion of Muslim population in a country would lead to more 
dominant use of "coverage of perpetrators" frame.
As H5 was not supported across all three terrorist attacks, the 
proportion of Muslim population has no significant influence on the 
"coverage of perpetrators" frame at all in this study. In addition to 
its insignificance, the direction of the proportion of Muslim 
population variable for Sept. 11 attacks (B = -0.003) and the Madrid 
train bombing (B = -0.016) go against the hypothesis. One reason for 
this is that the attacks took place in U.S. and Spain respectively 
where the proportion of Muslim population is small, as compared to 
Indonesia. For the Bali bombing, countries with a predominantly 
Muslim population (e.g. Malaysia) may be more concerned with how the 
Indonesian government was conducting investigations against people of 
the same "religion" as them. However for the Sept. 11 attacks and 
Madrid train bombing, countries with smaller proportions of Muslim 
population may not be so concerned about offending the minorities and 
would devote more to the "coverage of perpetrators" frame.
Hypothesis 6 (H6) predicted that the number of terror attacks would 
be positively correlated with the "call for aggressive action against 
terrorists" frame.
 From Table 5 (see Appendix A), H6 was not supported for the Sept. 11 
attacks (p = 0.441, B = 0.001). For the Bali bombing, the hypothesis 
was supported (p = 0.012, B = 7.358E-04). Lastly, H6 was not 
supported for the Madrid train bombing (p = 0.024, B = -0.002), and 
the negative B-coefficient meant that the lower the number of terror 
attacks that has occurred in the country, the more dominant the "call 
for aggressive action against the terrorists" frame will be. For 
Madrid, other factors were significant predictors of frame dominance. 
Trade was significant and ran in the opposite direction of the news 
frame (p = 0.024, B = -0.004), suggesting that a smaller trade volume 
would lead to greater dominance of the dependent variable. The 
results of distance (p = 0.032, B= -1.390E-05) showed that the closer 
the country is to Spain, the more dominant the "call for aggressive 
action against terrorist" frame will be.
That H6 is only supported for Bali bombing event and not the other 
events is in line with the study by Chang and Chang (2003), which 
found that terrorism experience is not a significant indicator of 
editorial coverage. For the Sept. 11 attacks, given the scale of the 
attacks in terms of number of victims and impact to global economy, 
it is possible that demands or suggestions for military responses 
might be considered premature in the first week after the attacks. 
During this period, newspapers would be more concerned about the 
casualties, economic impact, or rescue work. Moreover, the Sept. 11 
attacks can be considered the watershed of global terrorism, being 
the most major terror attack on the world's most powerful nation. It 
is after Sept. 11 that nations become more wary of terrorism, both 
globally and within their countries. Thus, the first week after Sept. 
11 may have been too early for nations to issue calls for aggressive 
action. However, Bali bombing occurred about a year after the Sept. 
11 attacks. Reactions were stronger as the world had been exposed to 
increasing coverage of radical terrorist groups and consequences of 
their actions. This has led to increasing awareness of terror attacks 
in their own countries and thus, increasing the demands, suggestions 
or requests for punitive, military actions as suggested from the 
hypothesis, H6.
Results for the Madrid train bombing are significantly different from 
the other two events. Trade, number of terror attacks, and distance 
are significant in predicting the dominance of the frame; however, 
they are negatively correlated to the dependent variable. It is 
possible that countries that engage in trade with Spain are less 
likely to call for aggression, which would have negative 
repercussions on trade itself. Trade with Spain could be more stable 
if the country does not engage in warfare.
In terms of terror attacks, U.K., Israel and Germany had the highest 
number of such attacks. These same three countries also had a higher 
amount of coverage for the Madrid train bombing as compared to other 
countries. It is possible that such countries which have had more 
experience in dealing with terror attacks would have less media 
coverage on the "call for aggressive action against terrorists" frame 
because fighting terrorism has long been a national concern to them. 
The Madrid train bombing by itself, would not trigger media coverage 
on action against terrorism because the media has been desensitised 
by the frequency of terror attacks. It is also possible that such 
countries, which have had many terror attacks in the past, have 
realised that aggressive action against suspected perpetrators was 
not an effective method to combat terrorism.
Finally, distance was negatively correlated with the dependent 
variable. This might be linked to the fact that immediately after the 
attacks, the Spanish government accused the ETA of masterminding the 
attacks. Countries geographically closer to Spain could thus feel 
more threatened by the ETA than countries further away, and thus call 
for aggressive action to suppress the group.
Across the three events for all 6 hypotheses, the R2 values are very 
small, with most values ranging from 0.024 to 0.191. Thus significant 
factors, when examined individually, can only predict the dominance 
of the frame to a very small extent. For H3 and H4 though, the Bali 
bombing had a relatively large R2 value of 0.299. This means that 
these independent variables contribute to nearly a third of the 
dominance of the frame. However, given the small B-values of the 
individual factors, once again they only predict the dominance of the 
frame to a small extent.
Overall, these small R2 values suggest that there are many other 
independent variables not examined in this current study, which could 
have contributed to the dominance of the news frames. Such factors 
could include tourism, press and political freedom of a country, 
number of diplomats, eliteness of nations and the use of news sources.

Conclusion
	The major contribution of this study is the linking of two theories 
– international news flow and framing analysis, and the comparison of 
how the they affect three seemingly similar terrorist attacks. The 
results of the current study have shown that three attacks were 
different in terms of its dominant frames and the predictability of 
factors affecting coverage.
"Threat to political climate" and "coverage of victims and potential 
victims" were always among the top three frames used in all three 
events. Given the heavy responsibility of governments in responding 
to these crises and reassuring the public, it would be expected that 
the "threat to political climate" frame would be dominant. Research 
on what is newsworthy has also shown that stories with a 
human-interest angle allow readers to relate better to the event, 
hence it is also understandable why "coverage of victims and 
potential victims" is among the most dominant frames.
In terms of H1 which links trade to the "threat to economy" frame, 
the hypothesis was only supported for the Sept. 11 attacks. Previous 
studies that use trade or economy as predictors of coverage had 
offered varying results, possibly due to the differences in the 
measurement of trade in various studies. As for H2 which predicts 
that GDP per capita would be positively correlated to the "threat to 
economy" news frame, it was only supported for the Madrid train 
bombing. Once again, studies that used GDP per capita as a predictor 
of news coverage had given conflicting results. China and Indonesia's 
low GDP per capita could also have affected the overall GDP per 
capita, resulting in a negative correlation between independent and 
dependent variables for Sept. 11.
H3, which links casualty rates to the frame "coverage of victims and 
potential victims" is only supported for the Sept. 11 attacks. The 
number of nations which suffered casualties in the other two terror 
attacks (Bali: 202 from 22 nations, Madrid: 191 from 14 nations) were 
comparatively lower than for the Sept. 11 attacks (more than 3,000 
from 36 nations). This is in line with Van Belle's (2002) research on 
the significance of casualty rates, and could account for why H3 was 
supported for the Sept. 11 attacks but not the Bali or Madrid train bombing.
As for H4 which predicted distance as a determinant of the frame 
"coverage of victims and potential victims", it is only supported for 
the Madrid train bombing. Based on previous research, it has been 
shown that even when geographic distance affects coverage, the effect 
varies and its influence could be minimal. Van Belle (2000) found 
that distance's effect is generalisable, Kariel and Rosenvall (1984) 
found that geographic distance accounted for only a small proportion 
of explained variance, while Wu (2003) found this factor to be 
significant, although only in developing countries.
For H5, which links the proportion of Muslim population to the 
"coverage of perpetrators" frame, the hypothesis is not supported 
across all events. This could be due to differences in what the media 
in each country see as relevant news. For example, countries with 
predominantly Muslim population may be more concerned with how the 
Indonesian government was conducting investigations against 
perpetrators who claim to be of the same religion. On the other hand, 
countries with smaller proportion of Muslim population may not be so 
concerned with offending their minorities by reporting more on 
perpetrators for the Sept. 11 and Madrid train bombing.
Lastly, H6 which ties the number of terror attacks to the "call for 
aggressive action against terrorists" frame is only supported for the 
Bali bombing. The lack of support for the other two events is in line 
with Chang and Chang's (2003) findings that terrorism experience is 
not a significant factor.
Based on past knowledge and literature review, a research 
incorporating the theories of international news flow and framing 
analysis with a cross-national media coverage of three separate 
terror attacks for a week is unprecedented. Some hypotheses proposed 
have been supported for certain events, illustrating how the two 
theories relate to each other empirically. It has thus achieved the 
goal of the study – of investigating how factors of international 
news flow might affect not merely the amount of foreign media 
coverage, but the framing of events in the foreign media as well.
The current study analysed only how two aspects of news output relate 
to each other. That no hypothesis was supported across all three 
events is probably due to the many other aspects which influence the 
eventual news output (such as sourcing patterns, journalists' 
backgrounds, political orientations of newspapers, etc), but were not 
investigated in the study. A challenge for future studies will be to 
integrate all possible aspects of influence on news output, and find 
out how they interact with one another to influence how foreign 
events are eventually covered
Overall, despite the lack of a shared hypothesis supported across all 
three events, the present study has nonetheless offered some insight 
into improving the understanding of the complexity of international 
news flow communication, framing research, and their relationships
Limitations and recommendations
Apart from time and budget constraints, this study is limited by a 
number of other constraints. First, the results and discussion of 
this study can only be generalised to the 13 newspapers and the time 
frame of one week after the attacks. The inclusion of more newspapers 
and a longer time period might offer more news angles and a different 
set of dominant frames.
There is also the dilemma of breadth versus depth - using a larger 
number of newspapers with fewer articles from each newspaper or using 
fewer newspapers but with more articles from each. The current study 
could be improved by exploring the option of depth for a more 
comprehensive study on international news coverage.
The inconsistent finding of the study is also due the lack of 
consistent coverage of the three events for all newspapers. This is 
especially so for the Madrid train bombing in which the following 
newspapers had no coverage of the event: The New Straits Times, China 
Daily, The Bangkok Post, and The Daily Yomiuri. Even among newspapers 
that covered these events, some provided only scant or minimal 
coverage, thus resulting in examination of a mere couple of articles 
as representations of the newspapers' coverage.
Another limitation is the use of interpreters during coding. During 
verbal translation, some meaning of the articles may be lost, and 
thus affecting the coding process. Although the hired interpreters 
were highly proficient in Spanish and German, they were not natives 
of the country, hence they may not have the full contextual 
understanding of the Spanish and German articles. A recommendation 
for overcoming this limitation is to engage natives of foreign 
countries as coders (Peter and Lauf, 2002).
A recommendation for further research is to examine coverage of 
terror attacks more similar in nature. That is, a comparative study 
of terror bombing on tourist spots or a comparison of terrorist 
attacks on financial hubs in different regions.
As shown from the small R2  values, the six independent variables 
that were chosen in the study accounted for only a small percentage 
of the dominance of the frame. More factors, such as tourism figures 
and the political orientations of newspapers, can be included for 
further research.
Finally, this study can also be improved if there was a comprehensive 
list of the number of people injured in each terrorist attack. By 
adding the number of injured to the casualty rate, it could have 
painted a more accurate picture of number of victims of the terror attacks.
APPENDIX A
Table 1
Summary of Frequency Tests for News Frames
RQ1: What is the dominant news frame in the articles on the three 
terror attacks?

Variable
Sept.11 attacks
Bali bombing
Madrid train bombing
Threat to economy
1.322 (1)
0.719 (3)
0.285
Threat to political climate
0.828 (3)
1.032 (2)
1.425 (1)
Threat to religious or racial harmony
0.141
0.066
0.160
Coverage of victims and potential victims
0.959 (2)
1.698 (1)
0.530 (3)
Coverage of perpetrators
0.492
0.546
1.003 (2)
Show of sympathy or empathy
0.259
0.206
0.240
Call for aggressive action against terrorists
0.233
0.005
0.023
Call for non-aggressive or undecided action against terrorists
0.273
0.087
0.192
International aid to victim countries
0.068
0.206
0.056
Aid within victim countries
0.212
0.038
0.096
Others
0.417
0.380
0.275
N.A.
0.378
0.515
0.517
Note. Numbers within brackets indicate ranking order of dominance. 
For example, (1) means the most dominant frame, and (2) means the 
second most dominant frame.
Table 2

Summary of Multivariate Regression Analysis for Independent Variables 
Predicting Dominance of "Threat to Economy" Frame (N=312)

H1: The larger the trade volume, the more dominant the frame "threat 
to economy" will be.
H2: The higher the GDP per capita of the country, the more dominant 
the frame "threat to economy" will be.

Event
Independent variable
_
_
Standard error
p
R2
Sept.11 Attacks
(N = 185)
Trade
0.008
0.567
0.004
0.023
0.185
GDP
-8.420E-06
-0.029
*
0.393
Terror attacks
0.010
0.120
0.014
0.237
Distance
2.928E-05
0.067
*
0.402
Casualties
0.005
0.066
0.009
0.302
Muslim pro.
0.020
0.162
0.015
0.089
Bali Bombing
(N = 83)
Trade
-0.107
-0.205
0.099
0.144
0.182
GDP
7.301E-05
0.306
0.011
0.078
Terror attacks
-0.005
-0.068
0.011
0.335
Distance
-1.570E-04
-0.460
*
0.007
Casualties
-0.011
-0.262
0.007
0.065
Muslim pro.
0.003
0.026
0.025
0.446
Madrid Train Bombing
(N = 44)
Trade
0.007
0.177
*
0.214
0.146
GDP
6.075E-05
0.652
*
0.022
Terror attacks
0.004
0.156
0.005
0.233
Distance
2.173E-05
0.155
*
0.277
Casualties
-
-
-
-
Muslim pro.
0.028
0.681
0.011
0.009
Note. *value <0.001

Table 3

Summary of Multivariate Regression Analysis for Independent Variables 
Predicting Dominance of "Coverage of Victims and Potential Victims" 
Frame (N=312)

H3: The higher the casualty rate, the more dominant the frame 
"coverage of victims and potential victims" will be.
H4: The closer the country is to the victim country, the more 
dominant the frame "coverage of victims and potential victims" will be.

Event
Independent variable
_
_
Standard error
p
R2
Sept.11 Attacks
(N = 185)
Trade
0.010
0.741
0.004
0.005
0.191
GDP
-1.180E-04
-0.425
*
0.000
Terror attacks
0.013
0.163
0.013
0.165
Distance
4.521E-04
1.076
*
0.000
Casualties
0.022
0.315
0.009
0.007
Muslim pro.
-0.076
-0.632
0.014
0.000
Bali Bombing
(N = 83)
Trade
0.442
0.531
0.147
0.002
0.299
GDP
-6.060E-05
-0.158
*
0.212
Terror attacks
-0.020
-0.190
0.016
0.099
Distance
3.909E-04
0.716
*
0.000
Casualties
0.014
0.211
0.011
0.094
Muslim pro.
0.011
0.053
0.037
0.382
Madrid Train Bombing
(N = 44)
Trade
-0.024
-0.314
0.021
0.086
0.106
GDP
-6.450E-05
-0.379
*
0.122
Terror attacks
-0.011
-0.247
0.009
0.130
Distance
-1.390E-04
-0.542
*
0.024
Casualties
-
-
-
-
Muslim pro.
-0.017
-0.227
0.021
0.214
Note. *value _ 0.001

Table 4

Summary of Multivariate Regression Analysis for Independent Variables 
Predicting Dominance of "Coverage of perpetrators" Frame (N=312)


H5: The higher the proportion of Muslim population in the country, 
the more dominant the "coverage of perpetrators" frame will be.

Event
Independent variable
_
_
Standard error
p
R2
Sept.11 Attacks
(N = 185)
Trade
0.003
0.257
0.003
0.201
0.036
GDP
-2.520E-06
-0.012
*
0.460
Terror attacks
0.004
0.065
0.011
0.362
Distance
2.151E-05
0.066
*
0.411
Casualties
0.004
0.067
0.008
0.313
Muslim pro.
-0.003
-0.034
0.012
0.398
Bali Bombing
(N = 83)
Trade
-0.117
-0.318
0.076
0.065
0.040
GDP
5.920E-05
0.351
*
0.066
Terror attacks
0.005
0.100
0.008
0.281
Distance
-4.910E-05
-0.204
*
0.150
Casualties
-6.580E-04
-0.022
0.006
0.454
Muslim pro.
0.007
0.072
0.019
0.365
Madrid Train Bombing
(N = 44)
Trade
-0.022
-0.258
0.019
0.129
0.107
GDP
7.791E-06
0.040
*
0.450
Terror attacks
-0.009
-0.173
0.011
0.213
Distance
-1.750E-05
-0.060
*
0.411
Casualties
-
-
-
-
Muslim pro.
-0.016
-0.189
0.024
0.254
Note. *value _ 0.001
Table 5

Summary of Multivariate Regression Analysis for Independent Variables 
Predicting Dominance of "Call for aggressive action against 
terrorists" Frame (N=312)

H6: The larger the number of terror attacks that has occurred in the 
country, the more dominant the frame "call for aggressive action 
against terrorists".

Event
Independent variable
_
_
Standard error
p
R2
Sept.11 Attacks
(N = 185)
Trade
9.024E-04
0.147
0.002
0.316
0.024
GDP
7.932E-06
0.062
*
0.298
Terror attacks
0.001
0.027
0.007
0.441
Distance
1.278E-05
0.066
*
0.412
Casualties
0.004
0.112
0.005
0.210
Muslim pro.
0.002
0.030
0.007
0.412
Bali Bombing
(N = 83)
Trade
0.001
0.090
0.003
0.325
0.120
GDP
-5.920E-07
-0.085
*
0.351
Terror attacks
7.358E-04
0.378
*
0.012
Distance
-1.680E-07
-0.017
*
0.464
Casualties
4.907E-05
0.039
*
0.413
Muslim pro.
-4.370E-05
-0.011
0.001
0.477
Madrid Train Bombing
(N = 44)
Trade
-0.004
-0.450
0.002
0.024
0.149
GDP
-9.170E-06
-0.493
*
0.061
Terror attacks
-0.002
-0.432
0.001
0.024
Distance
-1.390E-05
-0.495
*
0.032
Casualties
-
-
-
-
Muslim pro.
-0.003
-0.392
0.002
0.082
Note. *value _ 0.001
REFERENCES

Analysts cut GDP growth predictions after Bali terror. (2002, October 
17). The Jakarta
Post, p. 13
Anderson, S., & Sloan, S. (2002). Historical dictionary of terrorism. 
Lanham, MD: The
	 Scarecrow Press, Inc.
Arshad, A. (2002, October 18). Bali's tourism targets dashed by 
attacks. The Straits
Times.
Baden, J.H. (2003, October). War with Iraq: How The New York Times 
and the Guardian
of London covered the story after the first anniversary of the Sept. 
11 terrorist
attacks. Paper presented at the Association for Education in 
Journalism and Mass
Communication Conference.
Berkowitz, D., & Gavrilos, D. (2001, August). News of terrorism in 
Israel through an
	American lens. Paper presented at the annual conference of the 
Association for
	Education in Journalism and Mass Communication, Washington D.C., USA.
Chang, K. K. (2003). Media war between Israelis and Palestinians: 
Uses of news sources
in four mainstream U.S. newspapers. Paper presented to the International
Communication Division, AEJMC annual conference, Kansas City, MO.
Chang, K. K., Salmon, C., Lee, B.K., Choi, J.H., & Zeldes, G. (2004, 
August). The
influences of contextual factors on the selection of news frames: A 
cross-national approach to the news coverage of Severe Acute 
Respiratory Syndrome. Research paper presented to the International 
Communication Division, Association for Education in Journalism and 
Mass Communication, Toronto, Canada.
Chang, T. K., Shoemaker, P. J., & Brendlinger, N. (1987). 
Determinants of international
news coverage in the US media. Communication Research, 14 (4), 396-414.
Chang, T. K., & Lee, J.W.  (1992).  Factors affecting gatekeepers' 
selection of foreign
news: A national survey of newspaper editors. Journalism Quarterly, 69(3), 554-
561.
Chang, T.K. (1998). All countries not created equal to be news: World 
system and
international communication. Communication Research, 25 (5), 528-563.
Chang, Y.L., & Chang, K. K. (2003). World editorials on Sept. 11 
terrorist attacks:
Integrating news flow analysis in the international communication and framing
analysis. International Communication Bulletin, 38 (3-4), 42-60.
Crelisten, R. (1987). Power and Meaning: Terrorism as a Struggle over 
Access to the
Communication Structure. In Contemporary Research on Terrorism, edited by P.
Wilkinson. Aberdeen Scotland: University of Aberdeen Press.
Davis, D. (2004, March 17). London attack 'inevitable'. The Jerusalem 
Post, p.6.
Eckstein, J.M. (Sept., 2003). When the terrorist is American: Analyzing news
frames of the Sept. 11, 2001 attacks and the 1995 Oklahoma City bombing.
Paper presented at the Association for Education in Journalism and Mass
Communication Conference
Entman, R. M. (1993). Framing: Toward clarification of a fractured 
paradigm. Journal of 	Communication, 43 (4), 51-58.
Fico, F., & Cote, W. (1999). Fairness and balance in the structural 
characteristics of
newspaper stories of the 1996 presidential election. Journalism & Mass
Communication Quarterly, 76, 124-137.
Gamson, W.A. (1989). News as framing. American Behavioral Scientist, 
33, 157-161.
Gamson, W.A., & Modigliani, A. (1989). Media discourse and public 
opinion: A			constructionist approach. American Journal of Sociology, 85, 1-37.
Gandy, O.H, Kopp, K., Hands, T., Frazer, K., & Phillips, D. (1997) 
Race and Risk: Factors affecting the framing of stories about 
inequality, discrimination, and just plain bad luck. Public Opinion 
Quarterly, 61, 158-182.
Gitlin, T. (1980). The whole world is watching: Mass media in the 
making and unmaking		
  of the New Left. Berkeley, CA: University of California Press.
Goffman, E. (1974), Frame Analysis: An essay on the Organization of 
Experience. New York,
Harper and Row. Berkeley: University of California Press.
Golan, G., & Wanta, W. (2003) International elections on US Network News: An
	examination of factors affecting newsworthiness. Gazette. 65 (1), 25-39.
Goshorn, K. & Gandy, O.H. (1995) Race, risk and responsibility: 
Editorial constraint in the framing of inequality. Journal of 
Communication, 45 (2), 133-151.
Graber, D. (1989). Mass media and American politics (3rd ed.). Washington D.C:
	Congressional Quarterly Press.
Hoffman, B. (2002). Inside terrorism. New York: Columbia University Press.
Hester, A. (1973). International information flow. In H. Fischer & 
J.C. Merrill (Eds.),
International and intercultural communication (pp. 242-250). New York:
Hastings House.
Hicks, R.G., & Gordon, A. (1974). Foreign news content in Israel and 
US newspapers.
Journalism Quarterly, 51, 639-644.
Hopple, G. W.  (1982). International news coverage in two elite 
newspapers. Journal of 	
Communication, 32, 1-4.
Itule, B.D., & Anderson, D.A. (2003) News Writing and Reporting: For 
Today's Media. (6th ed.). New York: McGraw-Hill.
Kariel, H. G., & Rosenvall, L. A. (1984). Factors influencing 
international news flow.
Journalism Quarterly, 61, 509-516.
Kelly, M.J. and T.H. Mitchell. 1981. Transnational Terrorism and the 
Western elite Press.
Political Communication and Persuasion, 1, 269-296.
Kim, K. & Barnett, G. (1996). The determinants of international news 
flow: A network analysis. Communication Research, 23(1), 323-352.
Norris, Pippa (1995), The restless searchlight: Network news framing 
of the post Cold
War World, Political Communications, 12, pp. 357-370
McCombs, M. E., & Shaw, D. L. (1972). The agenda setting function of 
the press. Public		Opinion Quarterly, 36, 176-187.
Midgley, S. & Rice, V. (Eds.). (1984) Terrorism and The Media in the 
1980's: The
proceedings of a conference held April 14, 1983. Washington : The 
Media Institute.
Moore, J. & Pan, D. (2002). Effects of the Sept. 11, 2001 Terrorists 
Attack on U.S.
Press Coverage. Paper presented at the Association for Education in 
Journalism and Mass Communication Conference, Miami Beach, Florida.
Pei, M.A. (1960). The world's chief languages. New York: S.F. Vanni.
Peter, J., & Lauf, E. (2002). Reliability in cross-national content 
analysis. Journalism &
Mass Communication Quarterly, 79, 815-832.
Picard, R.G. (1993). Media Portrayals of Terrorism. Ames, IA : Iowa 
State University
Press.
Purba, K. (2002, October 19). Japan turns up pressure on Mega to 
fight terror. The
Jakarta Post, p.4.
Robinson, G.J. & Sparkes, V.M. (1976). International News in the Canadian and
American Press: A Comparative News Flow Study. Gazette, 22, 203-218.
Rosengren, K. E. (1970). International News: Time and type of report. 
In H.D. Fischer &
J.C. Merrill (Eds.), International Communication: Media, Channels, Functions
(1st ed., pp. 74-80). New York: Hastings House.
Scheufele, D. A. (1999). Framing as a Theory of Media Effects. Journal of
Communication, Vol 49 No. 1-2: 103–122.
Sharrock, D. & McGrory, D. (2004, March 13) Chilling '911' Murder 
Link. New York
Post.
Tuchman, G. (1978).  Making news: a study in the construction of 
reality. New York:
The Free Press.
Traugott, M.W. & Brader, T. (2002). Patterns in the American News 
Coverage of the
Sept. 11 Attacks. Paper prepared for the Harvard Symposium "Restless 
Searchlight: Terrorism, the Media & Public Life" held at the John F. 
Kennedy School of Government, Harvard University.
Van Belle, D. A. (2000). New York Times and network TV news coverage 
of foreign
disasters: the significance of insignificant variables. Journalism & Mass
Communication Quarterly, 77(1), 50-70.
Wilkinson, P. (2000). Terrorism versus democracy: The liberal state response.
	Portland, Oregon: Frank Cass Publishers.
Wu, H.D. (1998). Investigating the determinants of international news 
flow: A Meta-Analysis. Gazette: The International Journal for 
Communication Studies, 60(6), 493-512.
Wu, H. D. (2003). Homogeneity around the world?: Comparing the 
systemic determinants of international news flow between developed 
and developing countries. Gazette: The International Journal for 
Communication Studies, 65(1), 9-24.


[1]  Set up in 1999, Factiva®, a Dow Jones & Reuters Company, has a 
database of nearly 9,000 sources from 152 countries in 22 languages. 
Factiva® was used instead of Lexis-Nexis as it was able to produce 
search results for all newspapers in our study, unlike Lexis-Nexis. 
The latter was unable to produce results for Sydney Morning Herald 
and El Pais.