The Rise of New Communication Technologies Questions

How do new communication technologies affect television industries in Arab countries?

International Journal of Communication 13(2019), 653–673
1932–8036/20190005
Selective Exposure and Perceived Identification With Characters
in Transnational Arabic Television
TAMARA KHARROUB
Independent Researcher
ANDREW J. WEAVER
Indiana University Bloomington, USA
Given the transnational nature of the Arabic satellite television industry and the cultural
differences among Arab regions, this experiment examined the role of social identity in
Arab viewers’ media choices and perceived identification with characters. Overall, Arab
viewers identified more with in-group characters (gender and cultural region) and were
more interested in watching programs that feature in-group lead characters. The
perceived intended audience of the television program mediated the relationship
between cultural identity on one hand and perceived identification with characters and
selective exposure on the other. The findings show that social identity plays an
important role in Arab viewers’ media consumption, and are discussed in light of social
identity theory.
Keywords: social identity, transnational media, identification with characters, selective
exposure, Arab television
The Arabic version of the TV reality show Big Brother (Al-Ra’is) received heavy criticisms and was
cancelled in 2004 only days after it started airing on Middle East Broadcasting Corporation’s satellite
television channel (Kraidy, 2010). Meanwhile, the Arabic adaptation of the Dutch music contest reality
show Star Academy persisted to air on Lebanese Broadcasting Corporation’s satellite television channel for
several years with the most recent (11th) season airing its final episode in 2016. Both shows and other
reality programs on pan-Arab television were heavily criticized by ultraconservative Saudi clerics for their
gender dynamics, particularly for showing unmarried male and female contestants sharing living spaces.
Even though both Star Academy and Al-Ra’is generated controversies in Saudi Arabia, Star Academy was
based in Lebanon and hosted more participants from the Levant Arab states, and Al-Ra’is was filmed in
the Saudi neighbor Bahrain and hosted more participants including women from the Arabian Gulf. This
difference in the cultural settings of the two programs may explain the shutdown of Al-Ra’is compared
with that of Star Academy.
Tamara Kharroub: tamara.kh@gmail.com
Andrew J. Weaver: weaveraj@indiana.edu
Date submitted: 2017‒01‒28
Copyright © 2019 (Tamara Kharroub and Andrew J. Weaver). Licensed under the Creative Commons
Attribution Non-commercial No Derivatives (by-nc-nd). Available at http://ijoc.org.
654 Tamara Kharroub and Andrew J. Weaver
International Journal of Communication 13(2019)
Whereas entertainment programming in the Arab world is usually produced in the more liberal
Lebanon or Egypt, sponsors of Al-Ra’is wanted to attract the wealthy Saudi viewers by setting up the
reality show in a cultural setting close to that of Saudi Arabia (Kraidy, 2010). Because filming such
programs in the Saudi kingdom itself is not permitted, Bahrain was chosen. This decision was based on
the belief that Saudi viewers would be more likely to watch and identify with a show filmed in a country
geographically close and culturally similar to Saudi Arabia and that featured participants from the Arabian
Gulf similar to Saudi viewers. For the same reason, Saudi clerics believed that the cultural similarity
between the Saudi viewers and the show would lead to more favorable responses toward the media
personalities and thus greater influence by their behaviors and actions. Although a complex combination
of political, economic, and religious issues contributed to the Al-Ra’is controversy, the role of social and
cultural identity in this pan-Arab context cannot be ignored. The cultural identity of Al-Ra’is was too close
to home for the Saudi clerics, and most likely led to its shutdown. In this particular case, Bahrain’s status
as a vassal state made it easier for Saudi Arabia to exert influence and pressure to close production.
As the above example demonstrates, the cultural proximity between the participants in Al-Ra’is
and the Saudi viewers triggered a more pronounced reaction to this particular show than other television
programs that feature non-Gulf Arab participants. Although this example dates back to 2004, it is the
most prominent instance revealing the role of regional Arab cultural identity in the consumption of
transnational Arab media content. More recent examples point to similar patterns of identity-based
phenomena. For example, the popularity of Turkish television drama serials (dubbed into Arabic in Syrian
dialect) over the last decade has also raised controversy. Whereas American television imports are
subtitled into Arabic and the imported Mexican telenovelas are dubbed into literary Arabic, thus
maintaining a social distance between the characters and the viewers, Turkish series are dubbed into
colloquial Arabic and the characters’ names are Arabized (Buccianti, 2010). This cultural proximity has
generated public concern regarding the potential effects of Turkish dramas on the Arab viewers. It is
believed that such Arabized context and the sharing of a religious identity (Turkish characters being
Muslim) might increase the degree to which viewers connect and engage with the characters.
In the context of the Arabic-language transnational satellite television industry, social identity
theory (Tajfel, 1978) suggests that a shared social identity between the viewers and the media
personalities (characters or participants) would lead to an increased desire to view the programs and
greater identification with media personalities. The present study examined the role of social identity
(cultural and gender) in Arab viewers’ selective exposure to in-group and out-group media content and
their perceived identification with in-group and out-group characters in the transnational Arab media
context.
The Transnational Arab Television Industry
Television development in the Arab world went through several phases before evolving into the
current transnational industry. Ayish (2011) identifies three primary phases of development in the Arab
television industry; the formative phase between the 1950s and 1970s that ultimately led to government
control of the industry, the national expansion phase up to the 1990s of growth in production and
coverage, and the global phase following globalization trends during the 1990s including technological
International Journal of Communication 13(2019)
Selective Exposure and Identification 655
advancements, literacy, and privatization, which led to relaxed government control and the launching of
commercial operations.
With the global trends and the advent of satellite technology in the early 1990s, the Arabic
television market achieved a transnational reach beyond the limited national terrestrial radiuses, broke
away from political control, and evolved to include more diverse content. Globalization allowed
marginalized and oppositional groups to join the market, thus providing ideological variety as well as
choice and opportunity in Arabic television programming (Tawil-Souri, 2008). However, while political
elites loosened their grip on the media market, the economic elites took control of the industry. With the
commercialization aspect of globalization came the phenomenon of what some call the “Saudi media
empire” (Sakr, 2006). The structure of the new transnational market gave advantage to the business
elites in the region and the rich Gulf states to launch media conglomerates. The dominant financial
position of Gulf countries in this industry gave them a majority of the market share and a greater sphere
of influence in the region.
The rapidly growing Arab satellite television industry had reached 975 fully functional free-to-air
channels by the end of 2017 (Arab Advisors Group, 2017). Available in all 22 Arab countries and serving a
population of more than 406 million with at least 80% television penetration rates (Kraidy & Khalil, 2009),
satellite television remains the most popular mass communication medium across the Arab world.
Although more than 406 million inhabitants of the Arab world share a common language, they
vary greatly in their cultural tastes and sensibilities and their social, cultural, political, and legal systems
(Kalliny, Dagher, Minor, & De Los Santos, 2008). Some homogeneity among the Arab countries does exist
as a result of historical circumstances, but numerous differences among the Arab nations exist especially
with regard to wealth and economic standards, dialects of Arabic language, cultural traditions, religion and
religiosity, social conservativeness, and gender norms (Kalliny et al., 2008; Kraidy & Khalil, 2009). For
example, Saudi Arabia operates under strict Islamic law, does not hold elections, and places severe
restrictions on women’s rights and freedoms. On the other hand, Arab countries in the Levant (e.g.,
Lebanon) are considered more liberal and have relatively more gender equality.
Kalliny and colleagues (2008) capture this diversity by placing Arab countries on a liberal–
conservative continuum based on cultural differences. In general, they categorize Arab countries into
three cultural groups based on geographical locations: Asian non-Gulf countries (Levant), Arabian Gulf
countries, and African countries (Kalliny et al., 2008). They propose that countries in the Levant region
are on the liberal end, whereas Arab Gulf states are the most conservative, and African Arab countries lie
in the middle of the continuum. This categorization demonstrates the existence of three distinct cultural
regions within the Arab world. Along similar lines, the three categories used by other media scholars to
divide the region are the Arab Maghreb, which includes the western Arab countries in North Africa; the
Arab Gulf; and the Arab Mashreq, which consists of the Arab Middle East nations excluding the Gulf (e.g.,
Mellor, 2011). Most classifications place Egypt with the African Arab countries in the middle of the liberal–
conservative spectrum.
656 Tamara Kharroub and Andrew J. Weaver
International Journal of Communication 13(2019)
In the transnational Arab television market with such high levels of diversity, questions emerge
regarding viewers’ preferences and media consumption choices. In terms of content diversity, media
globalization has resulted in a paradox in Arab media. On one hand, commercial and consumerist trends
produced “Westernized” Arabic programing with images of seminude pop stars. On the other hand, there
arose an ethnoreligious populist reaction to globalization as a way to counter what is perceived as
imperialist influence, resulting in Islamic TV channels and religious and conservative content (Tawil-Souri,
2008). To take women’s representation as an example, a content analysis of Arabic drama serials
(Kharroub & Weaver, 2014) revealed that Arab Gulf programs had more female characters in female
stereotypical roles and activities than programs from the more liberal Levant countries. In this regard, the
financial power of conservative Arab Gulf states may translate into cultural influence through transnational
television networks, and is therefore worthy of investigation.
Audience’s media choices and perceived identification with characters in television programs are
especially important as they have been shown to play a role in the effects of media content on the
viewers’ attitudes and beliefs (e.g., Kaufman & Libby, 2012). In the pan-Arab television industry, media
programs are produced in several Arab countries and are consumed by viewers throughout the Arab
world. The transnational nature of this media market raises questions about the effects of media content
produced in one region and reflecting its cultural and social norms on the viewers’ attitudes and beliefs in
other regions. Given the important role of identity in the processes of identification and effects and the
differences among the Arab countries, the present study used the social identity framework to examine
whether Arab viewers are interested in watching television programs produced in different Arab cultural
regions and whether they identify with characters from Arab cultural regions different from their own.
With much attention given to media ownership and content, little is known about the viewers’
media choices in the Arab transnational television market. Surveys show a tendency by Arab viewers to
favor programs that feature their own country. Tunisian, Lebanese, Moroccan, and Egyptian viewers prefer
local programs produced in their own countries (Arab Advisors Group, 2005, 2006, 2008, 2010b). In
contrast, the majority of Jordanian TV viewers prefer watching Syrian and Egyptian drama series than
Jordanian dramas (Arab Advisors Group, 2010a). These findings show some differences in media
preferences among viewers in different Arab countries, but they also suggest that local programming is
still preferred. However, not much is known about the processes that facilitate these phenomena. Little
work is available on audience-centered approaches to the study of Arab media, and no empirical studies
have examined the role of identity in media choices and identification with characters among Arab TV
audiences. The present study examined the role of Arab viewers’ cultural and gender identities in their
media choices and their perceived identification with media characters in the transnational Arab television
environment.
Selective Exposure and Social Identity
People are known to favor programs that feature members of their social groups. For example,
studies have shown that viewers prefer programs produced in their home country (Waisbord, 2004) and
programs that feature characters that share their gender (Trepte, 2004) and race (Weaver, 2011b). Social
International Journal of Communication 13(2019)
Selective Exposure and Identification 657
identity theory and self-categorization theory have been used to explain how media choices can be
determined by social group membership (Trepte, 2006).
Social identity theory (Tajfel, 1978) posits that people categorize themselves as belonging to
different social groups such that one part of the self is defined by belonging to one social group. Henri
Tajfel (1978) defines social identity as “that part of an individual’s self-concept which derives from his
knowledge of his membership of a group together with the value and emotional significance attached to
that membership” (p. 63). Interest in entertainment content that features in-group members can be
explained by the desire to learn about one’s social group and define group norms (Hogg, 2010). As such,
based on social identity theory, we expected Arab viewers to show more interest in programs that feature
characters that share their social groups, particularly their gender and cultural background.
In terms of cultural background, although culture is a difficult term to delineate, it is commonly
defined as a set of interrelated qualities of observable artifacts, values, and underlying assumptions (such
as customs, morals, knowledge, etc.) that are shared by and specific to one social group (e.g., SpencerOatey, 2012). Several cultural groups exist in the Arab world. For the purpose of the present study and
given the structure and financing of the transnational Arab television industry, the regional cultural groups
within the Arab world are most relevant here. Based on the three regional cultural categories in the Arab
world (Levant, North Africa, and Gulf), we aimed to understand Arab viewers’ media choices and perceived
identification with characters.
In this context, Straubhaar’s (2003) cultural proximity concept can best explain media
consumption choices. According to Straubhaar, consumers and media audiences prefer products from
“one’s own culture or the most similar possible culture” (p. 85). Selection and enjoyment of media content
increase when it exists in the same “cultural linguistic” sphere as the viewers (Straubhaar, 2003). In
addition to language, other related factors such as dress, lifestyle, gestures, and traditions have been
discussed as determinants of cultural proximity (La Pastina & Straubhaar, 2005). In the transnational Arab
context, regional cultural categories (North Africa, Levant, and Gulf) have been shown to exhibit distinct
cultural cues that are visible in the media, such as dialect, dress, traditions, and gestures. The media
choices of Arab viewers in relation to the three cultural groups are herein examined.
Looking at quantitative studies of media preference for programs that feature characters sharing
the same cultural social group as the viewers (in-group members), the findings are inconclusive. As shown
earlier, studies of Arab audiences show some preference for local programming along national lines, but
differences among Arab countries exist (Arab Advisors Group, 2005, 2006, 2008, 2010a, 2010b). Looking
at other populations and national social groups, Trepte (2004) found no difference in U.S. viewers’
preference for German or American programs. Through two international quasi-experimental studies
conducted in the United States and Germany and in Great Britain and Germany, Trepte and
Krämer (2007) found no evidence for preference of national media products, even when the participants’
national identity was made salient. Trepte and Krämer suggest that, unlike sports, the salience of national
categories in entertainment programs is not high.
658 Tamara Kharroub and Andrew J. Weaver
International Journal of Communication 13(2019)
However, regional cultural identity in the Arab world is salient in entertainment programs,
especially through the observable cultural cues that distinguish among the three Arab regional regions.
Therefore, the present study examined Arab viewers’ media choices in relation to television programs
produced in different Arab cultural regions. The following research question guided this inquiry:
RQ1:
Do Arab viewers prefer TV programs produced in their own cultural region over programs
produced in other Arab cultural regions?
The second component of social identity examined in the present study was gender. Studies have
consistently found that females prefer female characters, whereas males prefer male characters. For
example, Trepte (2004) manipulated the gender of the protagonist in a description of entertainment
programs, and found that participants in both Germany and the United States preferred television series
that featured same-sex protagonists. Similarly, the present study hypothesized that Arab viewers would
prefer to watch television programs of same-sex lead characters:
H1a:
Arab female viewers will express more interest in watching programs that feature female lead
characters than male lead characters.
H1b:
Arab male viewers will express more interest in watching programs that feature male lead
characters than female lead characters.
Perceived Identification With Characters and Social Identity
The second process influenced by social identity and examined in the present study was
identification with characters. The process of identification is understood differently depending on the
genre of the medium, such that several concepts of viewers’ reactions to media have been discussed as
identification
including
liking,
similarity,
empathy,
wishful
identification,
fandom,
and
parasocial
interactions (Cohen, 2001; Hoffner & Buchanan, 2005; Hoffner & Cantor, 1991; Liebes & Katz, 1990).
Identification with characters in fictional narrative genres has been shown to be an important element of
narrative persuasion (De Graaf, Hoeken, Sanders, & Beentjes, 2011) and is therefore considered herein.
As such, the present study examined Arab viewers’ identification with characters in Arabic fictional drama
serials (Musalsalat).
Identification with characters is defined as a connection between an individual and another
person, “such that the individual adopts traits, attitudes, or behaviors of the other person, or incorporates
the other’s characteristics into his or her sense of self” (Hoffner & Buchanan, 2005, p. 326). In his seminal
paper, Cohen (2001) distinguishes between identification and other responses to media content based on
the internalization of media messages. He defines identification with fictional media characters in narrative
texts as an “imaginative process” (p. 253) that occurs in response to exposure to media characters; it is a
process of imagining oneself inside a textual reality where viewers feel with the character rather than
about the character.
International Journal of Communication 13(2019)
Selective Exposure and Identification 659
In addition to the temporary process of identification described by Cohen (2001) and others,
Hoffner and Buchanan (2005) discuss the different yet equally important notion of “wishful identification”:
the desire to be like the characters in media texts. According to Hoffner and Buchanan, wishful
identification is similar to the durable process of “long-term identification” proposed by Rosengren,
Windahl, Hakansson, and Johnsson-Smaragdi (1976), which extends beyond the viewing situation. In the
present study, Cohen’s conceptualization of identification and the long-term process of wishful
identification are both relevant and were therefore investigated.
Identification with media characters is a central process in media consumption experiences, as it
is believed to influence audience responses to media texts (Hoffner & Buchanan, 2005) and the role of
media in the socialization process (Bandura, 1986). One of the factors that motivates viewers to identify
with certain media characters and emulate their behavior is perceived similarity (Bandura, 1986). Social
cognitive theory predicts that people are more likely to pay attention to and model behaviors performed
by similar others (Bandura, 2001). Hoffner and Cantor (1991) argue “some degree of similarity to media
characters seems to promote a desire to be like them, possibly because certain similarities signal that it is
both possible and appropriate for the viewer to become like the character in additional ways” (p. 87).
Studies have found that viewers feel similar to characters who are like themselves in terms of
demographic characteristics such as gender (e.g., Reeves & Miller, 1978) and race (e.g., Appiah, 2001).
This similarity based on social group membership leads to higher identification with the media characters.
For example, Kaufman and Libby (2012) found that identification with a character (what they call
“experience-taking”) is higher when the participant and the character share a relevant group membership.
Identification, in turn, increases the effects of media on the viewers (e.g., Liebes & Katz, 1990). For
example, Kaufman and Libby (2012) found that 65% of participants in the character in-group condition
reported intentions to behave similarly to the character (e.g., going to vote), compared with only 29% of
participants in the character out-group condition.
When considering gender as a social identity, research on gender-based identification with media
characters shows that viewers identify more with same-sex characters than opposite-sex characters
(Hoffner, 1996; Reeves & Miller, 1978). In the present study, we hypothesized that Arab viewers would
identify more with programs that feature same-sex lead characters (i.e., when the viewer and the
character share the same sex group membership, or gender in-group):
H2a:
Arab female viewers will identify more with female lead characters than with male lead
characters.
H2b:
Arab male viewers will identify more with male lead characters than with female lead characters.
Similarly, in terms of regional cultural groups, we predicted that identification with media
characters would be higher in programs from the same Arab cultural region as the viewers (i.e., cultural
in-group):
H3:
Arab viewers will identity more with cultural in-group characters (from their cultural region) than
with cultural out-group characters (from other cultural regions).
660 Tamara Kharroub and Andrew J. Weaver
International Journal of Communication 13(2019)
Perceived Relevance and Intended Audience
Several researchers have been trying to understand the processes by which social identity affects
media choices and the level of connection with media personalities. Two distinct processes have been
examined: the perceived relevance of the plot to the viewers’ lives and the perception of being part of the
intended audience of the program.
The perception of the relevance of the storyline (i.e., events and issues) by the viewers has been
found to influence interest in the program. For example, viewers prefer films that have ethnic orientations
(i.e., cultural traditions) that are congruent with the viewers’ ethnic background (Grier, Brumbaugh, &
Thornton, 2006). Similarly, Cohen and Ribak (2003) showed that women liked Ally McBeal and found it to
be more relevant than men did, because relevance (through realism) allows viewers to engage with the
text and lose themselves in the plot. Weaver (2011a) also found that perceived relevance of a movie
theme was related to the viewers’ intent to see the movie.
In addition to influencing interest in the program, text relevance is also believed to impact
identification with the characters. For example, the relevance of the themes in the film Thelma & Louise to
women increased female audience’s identification with the female protagonists (Cooper, 1999). The
present study explored the role of plot relevance in mediating the effect of social identity on interest in the
show and on perceived identification with the lead characters. With limited research literature on the topic,
the current study expands evidence of the role of plot relevance in media choices and consumption:
RQ2a: Will the perceived relevance of the plot account for the effect of social identity on media choices?
RQ2b: Will the perceived relevance of the plot account for the effect of social identity on perceived
identification with characters?
Another factor that has been found to influence interest in a media product and identification with
characters is the perception of the intended audience of a program. For example, consumer behavior
research has shown that people who perceive themselves to be part of the intended target audience of an
advertisement are more likely to show interest in watching and more likely to express positive attitudes
toward the ad (Grier & Brumbaugh, 1999). In films, Weaver (2011a) found that shared social identity
between the viewer and the character (race in this case) influences perceptions of the intended audience
of the film, which in turn is strongly related to interest in the media product and intentions to see the film
(r = .68). In fact, the perception of the intended audience fully mediated the effect of actor’s race on
behavioral intentions to see the film. Thus, in the present study, we investigated the applicability of this
phenomenon in the transnational Arab context, and examined Arab viewers’ perceptions of being part of
the intended audience as a mediator of the effect of social identity on interest in the show and on
perceived identification with the lead characters:
RQ3a:
Will the perception of the intended audience of the program account for the effect of social
identity on interest in the program?
International Journal of Communication 13(2019)
RQ3b:
Selective Exposure and Identification 661
Will the perception of the intended audience of the program account for the effect of social
identity on perceived identification with the characters?
Method
Participants
Participants were college students recruited through universities in Palestine, Egypt, and Syria
and through student-centered Facebook pages. The participants (N = 123) were between the ages of 18
and 31 years (M = 25.3 years). There were 80 women and 43 men, of which 84 were Palestinian, 13 were
Syrian, 12 from Egypt, four from Lebanon, and the remaining from other countries. However, only 83
participants completed the questions about same-sex selective exposure and perceived identification with
characters and were included in this analysis, of whom 57 were women and 26 were men. Moreover, using
the three categories of Arab cultural regions, Palestinian, Syrian, and Lebanese participants only were
included in the within-subjects analyses for media choices and perceived identification based on cultural
identity (N = 70), as participants from these three countries represent the same Arab Levant cultural
region.
Design and Procedure
This experiment used a 3 (producing country: Syria, Egypt, or Kuwait) ´ 2 (sex of lead character)
within-subjects design, resulting in six experimental conditions. The participants were sent the link to a
Web page where they were told that they would be participating in a study part of a research project
examining “media use in different cultures.” Participants were then randomly directed to a page containing
all six conditions but employing different combinations of story synopses. The participants were first asked
to provide demographic information including age, sex, and country of origin, followed by viewing six
synopses of Arab television fictional drama programs (one for each condition) and completing measures
for selective exposure, perceived identification with characters, plot relevance, and perception of the
intended audience for each condition. The introductory Web page and all materials and measures were
translated and presented to participants in Arabic.
Materials
Twelve unique synopses of television drama serials (Musalsalat) were created based on existing
American and Arab television drama plots. Each synopsis consisted of the title and a one-paragraph
description of the plot. The producing country of the program (to delineate regional cultural identity) and
the sex of the lead character were manipulated to create six conditions: Syria and female, Egypt and
female, Kuwait and female, Syria and male, Egypt and male, and Kuwait and male. These three countries
were chosen because the majority of drama shows on Arab television come from those countries
(Chahine, El Sharkawy, & Mahmoud, 2007) and because they represent different points on the Arab
cultural spectrum (Kalliny et al., 2008), where Syria (as part of the Levant regional cultural category) lies
on the moderate end, Kuwait (as part of the Gulf regional cultural category) on the conservative end, and
Egypt in the middle. The data were collected before the start of the Syrian civil war. Twelve synopses were
662 Tamara Kharroub and Andrew J. Weaver
International Journal of Communication 13(2019)
used for the six conditions to increase generalizability. The synopses and conditions were counterbalanced
such that each synopsis was paired with each condition an equal number of times, and every synopsis–
condition combination was seen by an equal number of participants.
Measures
Selective Exposure
Three items adapted from Weaver (2011b) were used to measure participants’ interest in
watching the show: “Based on the synopsis, how interested are you in watching this show on television?”;
“Based on the synopsis, how interested are you in watching this show on DVD?”; and “Based on the
synopsis, how interested are you in watching this show on the Internet?” The questions were scored on a
5-point scale: 1 = not at all, 2 = a little, 3 = somewhat, 4 = very, and 5 = a lot. The average of the
responses to the three questions was used as a selective exposure measure. Cronbach’s alphas for this
scale for each of the six conditions ranged from .90 to .94.
Perceived Identification With Characters
Because both identification and wishful identification are relevant and important to assess the
level of connection with characters in this study, we used four items to assess the level of identification
with the lead character in each show, two of which were used by Cohen and Ribak (2003) to measure
identification: “I feel like I could identify with the characters in this program” and “I think there are many
points of similarity between the lead characters and myself,” and two measured the wishful identification
dimension (Hoffner, 1996; Rosengren et al., 1976): “I would like to do the kinds of things the lead
characters do in this program” and “I want to be like the characters in this program.” As identification with
characters was measured by asking participants to report their identification and reactions to the
characters, it was considered a measure of perceived identification. Participants indicated their agreement
with the four statements using a 5-point scale: 1 = strongly disagree, 2 = disagree a little, 3 = n either
agree nor disagree, 4 = agree a little, and 5 = strongly agree. Scores were averaged to create a measure
of identification with the lead character. Cronbach’s alphas for this scale for each of the six conditions
ranged from .89 to .93.
Relevance of the Plot
Three items were used by Cohen and Ribak (2003) to measure the perceived thematic relevance
of the plot by the participants, and were therefore used in the present study: “The issues dealt with in the
plot are realistic,” “The plot of this TV show is relevant to me,” and “The plot of this TV show reminds me
of my life and the lives of those around me.” Participants indicated their agreement with the statements
using a 5-point scale: 1 = strongly disagree, 2 = disagree a little, 3 = n either agree nor disagree, 4 =
agree a little, and 5 = strongly agree. Scores were averaged to create a measure of perceived relevance
of the plot. Cronbach’s alphas for this scale for each of the six conditions ranged from .76 to .84.
International Journal of Communication 13(2019)
Selective Exposure and Identification 663
Perceived Intended Audience
Three items, adapted from Weaver (2011a), were used to measure the perceived intended
audience of the program: “I believe I am part of the intended audience for this TV show,” “I believe most
of my friends would enjoy this TV show,” and “I believe this TV show was made for me.” Participants
indicated their agreement with the statements using a 5-point scale: 1 = strongly disagree, 2 = disagree
a little, 3 = n either agree nor disagree, 4 = agree a little, and 5 = strongly agree. Scores were averaged
to create a measure of perceived intended audience. Cronbach’s alphas for this scale for each of the six
conditions ranged from .71 to .83.
Statistical Power
The G*Power 3.1 statistical program (Faul, Erdfelder, Lang, & Buchner, 2007) was used to
calculate the power for the analyses. For the within-subjects factors (F tests for repeated-measures
analysis of variance [ANOVA] within factors) with six within-subjects conditions (Syrian male, Syrian
female, Egyptian male, Egyptian female, Kuwaiti male, and Kuwaiti female) and two between-subjects
groups (male and female participants) and a sample size of N = 83, the post hoc power to find a medium
effect size (f = 0.25) and p = .05 was 0.99. For the between-subjects factor (F tests for repeatedmeasures ANOVA between factors) with two groups (male and female participants) and a sample size of N
= 70, the post hoc power to find a medium effect size (f = 0.25) and p = .05 was 0.84. For the within–
between interactions (F tests for repeated-measures ANOVA within–between interactions) and a sample
size of N = 70, the post hoc power to find a medium effect size (f = 0.25) and p = .05 was 0.99.
Results
The hypotheses and research questions were tested using repeated-measures ANOVA. A 3
(country of the plot) ´ 2 (sex of lead character) ´ 2 (sex of participant) factorial design was used, where
sex of the participants was a between-subjects factor and the plot’s country and lead character’s sex were
within-subjects factors.
Selective Exposure
Hypothesis 1a predicted that female Arab viewers would express more interest in shows that
feature female lead characters than shows with male lead characters and Hypothesis 1b predicted that
male Arab viewers would express more interest in shows that feature male lead characters than female
lead characters. Sex of participants had a significant main effect, F(1, 81) = 8.36, p < .01, partial ƞ2 = .09, where women expressed more interest in the shows (M = 2.18) than men (M = 1.58). Hypothesis 1 was supported because the gender of the lead characters interacted with the sex of participants and significantly affected the desire to watch the show, F(1, 81) = 13.25, p < .001, partial ƞ2 = .14. Table 1 shows the means and confidence intervals. 664 Tamara Kharroub and Andrew J. Weaver International Journal of Communication 13(2019) Table 1. Viewers’ Selective Exposure to Television Programs. Male lead characters Female lead characters Participant M 95% CI M 95% CI Male (n = 26) 1.67 [1.33, 2.01] 1.48 [1.10, 1.89] Female (n = 57) 2.03 [1.80, 2.26] 2.33 [2.07, 2.60] Note. p < .01. Research Question 1 asked whether Arab viewers express more desire to watch programs produced in the same cultural region as them (i.e., that feature cultural in-group lead characters) than those with cultural out-group characters. The results show a significant difference in the Levant viewers’ preferences for Syrian, Egyptian, and Kuwaiti programs, F(2, 67) = 14.62, p < .001, partial ƞ2 = .30. Bonferroni post hoc tests revealed that participants who were from the Levant region in the Arab world (Palestine, Syria, and Lebanon) were significantly more interested in watching Syrian shows (M = 2.20, 95% CI = [1.93, 2.47]) than Kuwaiti shows (M = 1.69, 95% CI = [1.45, 1.94]), p < .001, and more interested in watching Egyptian shows (M = 2.03, 95% CI = [1.78, 2.28]) than Kuwaiti shows (M = 1.69, 95% CI = [1.45, 1.94]), p < .001. Arab Levant viewers expressed more interest in watching Syrian programs over Egyptian programs; however, this finding was not statistically significant (p = .18). Perceived Identification With Characters Hypothesis 2a predicted that Arab female viewers would identify more with characters in shows that feature female lead characters and Hypothesis 2b predicted that Arab male viewers would identify more with characters in shows that feature male lead characters than female lead characters. Hypothesis 2 was supported, as the gender of the lead characters interacted with the sex of participants and significantly affected perceived identification with characters in the show, F(1, 81) = 5.64, p < .05, partial ƞ2 = .07. Table 2 shows the means and confidence intervals. There was no statistically significant difference in female (M = 1.76) and male (M = 1.75) participants’ perceived identification with characters in programs featuring male lead characters. Moreover, there was no main effect of sex on identification with characters overall (p = .28). Table 2. Viewers’ Perceived Identification With Characters. Male lead characters Female lead characters Participant M 95% CI M 95% CI Male (n = 26) 1.75 [1.56, 2.04] 1.59 [1.29, 1.90] Female (n = 57) 1.67 [1.56, 1.96] 1.93 [1.73, 2.14] Note. p < .05 Hypothesis 3 predicted that Arab viewers would identify more with cultural in-group lead characters than cultural out-group lead characters. There was a significant difference in participants’ perceived identification with characters in Syrian, Egyptian, and Kuwaiti programs, F(2, 67) = 9.22, p < .001, partial ƞ2 = .22. Bonferroni post hoc tests revealed that participants from the Levant region in the Arab world identified significantly more with the lead characters when they were told that the characters were Syrian (M = 1.96, 95% CI = [1.74, 2.18]) than with the characters presented as Kuwaiti (M = 1.59, International Journal of Communication 13(2019) Selective Exposure and Identification 665 95% CI = [1.39, 1.79]), p < .001, even though the synopses were exactly the same. Participants also identified significantly more with characters when they were told that the characters were Egyptian (M = 1.83, 95% CI = [1.63, 2.03]) than when they were told that the characters were Kuwaiti (M = 1.59, 95% CI = [1.39, 1.79]), p < .05. Arab Levant viewers identified more with characters when they were presented as Syrian than when they were presented as Egyptian, but this difference was not statistically significant (p = .38). Perceived Relevance and Intended Audience To test the role of perceived relevance and perceived intended audience in mediating the effect of in-group/out-group dynamics on selective exposure and identification, we used a path model analysis. For this purpose, the data were rearranged such that every combination of participant and condition became a unique row of data. Moreover, the cultural identity of the characters was transformed into a cultural distance variable, where the higher value indicated greater cultural distance between the Levant viewers and the characters (1 = Syrian characters, 2 = Egyptian characters, 3 = Kuwaiti characters). Similarly, the combination of the sex of the character and the sex of the participant was transformed into one variable with two levels, where the same-sex condition was in-group and the opposite-sex condition was out-group (0 = in-group, 1 = out-group). The path analyses were performed for the participants from the Levant countries only (Palestine, Syria, and Lebanon). The model was tested using AMOS 7.0, with 200 bootstrap samples, 95% confidence interval. Model fit was evaluated using the chi-square test, the comparative fit index (CFI), the normed fit index (NFI), and the root mean square error of approximation (RMSEA), as recommended by Thompson (2004). For the overall model, the chi square was not significant, c2(1, N = 419) = 0.016, p = .899. The NFI was 1.000, the CFI was 1.000, and the RMSEA was 0.0000, all of which indicate an acceptable goodness of fit. Using methods for model trimming recommended by Kline (2004), we sequentially removed the nonsignificant paths from the model. A procedure of removing the path with the smallest critical value and refitting the model was repeated until no nonsignificant paths remained. The resulting model demonstrated good fit, c2(8, N = 419) = 9.115, p = .333. The NFI was 0.992, the CFI was 0.999, and the RMSEA was 0.018. This model is shown in Figure 1. 666 Tamara Kharroub and Andrew J. Weaver International Journal of Communication 13(2019) Figure 1. Path model showing the significant relationships (standardized regression weights) between social identity and selective exposure and identification with characters, through plot relevance and perception of the intended audience. *p < .05. **p < .01. ***p < .001. There were no significant direct effects of sex and regional cultural identity on either selective exposure or perceived identification with characters. Interestingly, regional cultural identity also did not directly affect the perceived relevance of the plot, and sex identity did not directly affect the perception of the intended audience. However, there were direct effects of cultural distance on the perception of being part of the intended audience (r = −.21, p < .001) and of sex identity on perceived relevance of the plot (r = .08, p < .05). See Figure 1 for all direct effects. Although sex and cultural identity did not directly affect selective exposure and perceived identification, they had significant indirect effects on both. In terms of regional cultural identity, increased cultural distance between the viewers and the characters decreased interest in watching the programs, that is, selective exposure (r = −.13, p < .01) and decreased the level of perceived identification with characters (r = −.15, p < .05), via perceived relevance of the plot and the perception of the intended audience. Moreover, there was a significant indirect effect of cultural distance on perceived relevance of the plot (r = −.15, p < .01) via perception of the intended audience. With regard to viewers’ sharing the International Journal of Communication 13(2019) Selective Exposure and Identification 667 same-sex identity with the characters, there were significant indirect effects of sex on identification with characters (r = .05, p < .05) and on selective exposure (r = .03, p < .05) via perceived relevance of the plot. These results, showing no significant direct effect when the mediators were included but only indirect effects, indicate that relevance of the plot and perception of the intended audience together mediated the effects of sex and cultural identity on selective exposure to the programs and identification with characters. Regarding Research Question 2, perceived relevance of the plot mediated the influence of sex identity (but not the influence of cultural identity) on selective exposure to the programs and perceived identification with characters. Regarding Research Question 3, the perception of the intended audience mediated the influence of cultural identity (but not the influence of sex identity) on selective exposure and perceived identification with characters. However, the relationships between sex identity and the two dependent variables did not follow the expected pattern when relevance of the plot was added to the model. The characters’ out-group membership with regard to the viewers’ sex identity increased interest in the program and perceived identification with the character. In conclusion, accounting for plot relevance and perception of the intended audience eliminated the direct effects of identity on selective exposure and perceived identification with characters. In particular, perceived relevance of the plot explained the effect of in-group sex identity on selective exposure and perceived identification with characters, and perception of being part of the intended audience explained the effect of cultural distance on selective exposure and perceived identification with characters. Overall, the effects were stronger for cultural identity than sex identity, and the strongest mediator for these effects was the perception of being part of the intended audience. Discussion The present experiment provides an empirical examination of Arab viewers’ social identity as a driver of selective exposure to media content and perceived identification with television characters in the transnational Arabic media industry. The findings suggest that viewers in the Levant cultural region of the Arab world are less interested and might identify less with characters in programs produced in the more conservative Arabian Gulf, which are dominant in the television industry. The findings may confirm the Saudi clerics’ fear about Saudi viewers’ interest in and identification with the contestants in Al-Ra’is. Saudi viewers are expected to identify more with characters from the Arabian Gulf cultural region and be more influenced by this in-group content. In this study, consistent with existing findings (e.g., Hoffner, 1996; Reeves & Miller, 1978; Trepte, 2004), female viewers expressed higher identification with female characters and more interest in programs that feature female lead characters, whereas male viewers identified more with male lead characters and were more interested in watching programs with male lead characters. The findings regarding cultural identity were also in line with these results. Participants from the Levant region in the Arab world expressed most interest in watching television programs produced in Syria and least interest in Kuwaiti programs. Similarly, the participants identified more with Syrian and Egyptian characters than with Kuwaiti characters. These findings are particularly compelling because the very subtle manipulations 668 Tamara Kharroub and Andrew J. Weaver International Journal of Communication 13(2019) (changing the gender of the lead character and the country name, while keeping the exact same synopses) produced significant effects. The cultural identity findings give credence to two of the three commonly referenced cultural regions (the Levant and the Arabian Gulf), suggesting a salience of regional cultural differences in transnational Arab media products (e.g., through visual cultural cues such as dialect and dress). However, the results for Egypt were not significant. Participants did not express more interest in and did not identify more with characters in Syrian programs than Egyptian programs. This is likely because of Egypt’s geography between the Levant and North Africa and the Arab viewers’ familiarity with Egyptian media products. The findings in general support social identity theory (Tajfel, 1978), in that our Arab viewers favored programs that feature their gender and cultural in-groups and identified more with characters that share their social identities. These findings have important implications for media effects. Social cognitive theory suggests that increased exposure and identification with media characters influence the impact of media content on Arab viewers, where learning is directly related to the observation of media models and behaviors (Bandura, 1988). Moreover, social cognitive theory predicts that people are more likely to pay attention to and model behaviors performed by similar others (Bandura, 2001). Therefore, the results suggest that Arab viewers are more likely to be influenced by messages and model behaviors presented by in-group characters (same-sex and proximate cultural identity). In addition, the study found significant mediators of the relationships between identity and media choices, especially with regard to cultural identity. Consistent with Weaver’s (2011a) findings, perception of being part of the intended audience explained the effect of cultural distance on interest in watching the programs and perceived identification with characters. In the Arab context, the results show that a shared cultural identity between the viewer and the character influences perceptions of the intended audience, which in turn influences the viewers’ interest in television programs and their identification with media characters, that is, the effect of cultural identity on selective exposure and identification is fully mediated by the perception that the program was made for them. The unexpected findings for the mediation analyses regarding sex identity might be due to differences between male and female viewers’ responses to the sex of the characters. The setup of the mediation path model, in which sex identity was treated as in-group versus out-group, did not account for gender differences. In fact, in the first part of the data analysis, we found that male and female participants did not differ in their level of identification with male lead characters. This pattern is consistent with existing research (Hoffner, 1996; Reeves & Miller, 1978), which found that boys identify more with same-sex characters than girls. Overall, the present study reveals important findings about the role of social identity in the processes of media consumption among Arab viewers. Consistent with other findings (Kaufman & Libby, 2012; Trepte, 2004), the study shows that Arab viewers are more interested in in-group content and are more likely to identify more with characters similar to themselves, suggesting that such in-group media content might reinforce their existing attitudes and beliefs, especially about gender roles. Moreover, even International Journal of Communication 13(2019) Selective Exposure and Identification 669 when Arab viewers do consume out-group media programs, the impact of out-group content on their attitudes and beliefs may be lessened. Although it provides a replication and expansion of the research on social identity and identification to the Arab media industry and delivers important empirical data of Arab viewers’ media choices and exposure in transnational contexts, the present study is not without limitations. The relatively small sample, the majority of which was female, and the focus of the study on Levant participants present the primary limitations to the study. This limitation is due to the extreme difficulty in collecting data from Arab viewers in different countries. A second limitation is presented by the measure of perceived identification with characters through reading a synopsis, which is also due to logistical challenges in data collection from this understudied population. However, this method of examining identification with characters using textual narratives has been used in several studies (e.g., De Graaf et al., 2011; Weaver, 2011b). Despite these limitations, the study provides important contributions to this area of research and evidence of media choices in the Arab context. Future research should provide comparison with participants from the Arabian Gulf (e.g., Kuwait and Saudi Arabia) and North Africa. For example, viewers from conservative countries in the Gulf might be interested in programs from more liberal Arab countries as a way of escapism, which would be an additional interesting interaction with characters to explore. Moreover, as previous research has found that identification with media characters can increase the effects of media content on the viewers’ attitudes and beliefs (e.g., Kaufman & Libby, 2012), this impact should also be directly investigated in the Arab context. Examination of regional conflicts and political polarization in Arab responses to media messages and message source, including on social media, would be important to assess in the current political climate. Finally, further investigation of the nature and salience of cultural identity in transnational Arab media content could continue to develop our understanding of consumption behavior and effects in this context. References Appiah, O. (2001). Black, White, Hispanic, and Asian American adolescents’ responses to culturally embedded ads. The Howard Journal of Communications, 12, 29–48. doi:10.1080/10646170117577 Arab Advisors Group. (2005). 48% of households in Cairo use the Internet and 46% have satellite TV [Press release]. Retrieved from http://www.arabadvisors.com/Pressers/presser-260105.htm Arab Advisors Group. (2006). Terrestrial TV is still going strong in Morocco [Press release]. Retrieved from http://www.arabadvisors.com/Pressers/presser-100706.htm Arab Advisors Group. (2008). Lebanese TV stations—LBC and Future TV—are the most popular amongst TV viewers in Lebanon [Press release]. Retrieved from http://www.arabadvisors.com/Pressers/presser-040208.htm 670 Tamara Kharroub and Andrew J. Weaver International Journal of Communication 13(2019) Arab Advisors Group. (2010a). Jordanian TV viewers prefer Syrian and Egyptian drama series over Jordanian ones [Press release]. Retrieved from http://www.arabadvisors.com/Pressers/presser190910.htm Arab Advisors Group. (2010b). Tunisian FTA channels top the charts amongst Tunisian TV viewers. Pan Arab channels follow [Press release]. Retrieved from http://www.arabadvisors.com/Pressers/presser-251110.htm Arab Advisors Group. (2017). 53.6% of FTA satellite TV channels broadcast in the Arab world have online presence [Press release]. Retrieved from http://www.arabadvisors.com/536-fta-satellite-tvchannels-broadcast-arab-world-have-online-presence Ayish, M. (2011). Television broadcasting in the Arab world: Political democratization and cultural revitalism. In N. Mellor, M. Ayish, N. Dajani, & K. Rinnawi (Eds.), Arab media: Globalization and emerging media industries (pp. 85‒102). Cambridge, UK: Polity Press. Bandura, A. (1986). Social foundations of thought and action: A social cognitive theory. Englewood Cliffs, NJ: Prentice Hall. Bandura, A. (1988). Organizational application of social cognitive theory. Australian Journal of Management, 13(2), 275–302. Bandura, A. (2001). Social cognitive theory of mass communication. Media Psychology, 3, 265–299. Buccianti, A. (2010). Turkish soap operas in the Arab world: Social liberation or cultural alienation? Arab Media & Society, 10(Spring). Retrieved from https://www.arabmediasociety.com/turkish-soapoperas-in-the-arab-world-social-liberation-or-cultural-alienation/ Chahine, G., El Sharkawy, A., & Mahmoud, H. (2007). Trends in Middle Eastern Arabic television series production: Opportunities for broadcasters and producers. Beirut, Lebanon: Booz Allen Hamilton. Cohen, J. (2001). Defining identification: A theoretical look at the identification of audiences with media characters. Mass Communication & Society, 4(3), 245–264. doi:10.1207/S15327825MCS0403_01 Cohen, J., & Ribak, R. (2003). Gender differences in pleasure from television texts: The case of Ally McBeal. Women’s Studies in Communication, 26, 118–134. doi:10.1080/07491409.2003.10162454 Cooper, B. (1999). The relevancy of gender identity in spectators’ interpretations of Thelma & Louise. Critical Studies in Mass Communication, 16, 20–41. doi:10.1080/15295039909367070 International Journal of Communication 13(2019) Selective Exposure and Identification 671 De Graaf, A., Hoeken, H., Sanders, J., & Beentjes, J. (2011). Identification as a mechanism of narrative persuasion. Communication Research, 39(6), 802–823. Faul, F., Erdfelder, E., Lang, A.-G., & Buchner, A. (2007). G*Power 3: A flexible statistical power analysis program for the social, behavioral, and biomedical sciences. Behavior Research Methods, 39, 175‒191. doi:10.3758/BF03193146 Grier, S. A., & Brumbaugh, A. M. (1999). Noticing cultural differences: Ad meanings created by target and non-target markets. Journal of Advertising, 28(1), 79–93. Grier, S. A., Brumbaugh, A. M., & Thornton, C. G. (2006). Crossover dreams: Consumer responses to ethnic-oriented products. Journal of Marketing, 70, 35–51. Hoffner, C. (1996). Children’s wishful identification and parasocial interaction with favorite television characters. Journal of Broadcasting & Electronic Media, 40, 389–402. Hoffner, C., & Buchanan, M. (2005). Young adults’ wishful identification with television characters: The role of perceived similarity and character attributes. Media Psychology, 7, 325–351. doi:10.1207/S1532785XMEP0704_2 Hoffner, C., & Cantor, J. (1991). Perceiving and responding to mass media characters. In J. Bryant & D. Zillmann (Eds.), Responding to the screen: Reception and reaction processes (pp. 63–103). Hillsdale, NJ: Erlbaum. Hogg, M. A. (2010). Influence and leadership. In S. T. Fiske, D. T. Gilbert, & G. Lindzey (Eds.), Handbook of social psychology (5th ed., Vol. 2, pp. 1166–1207). Hoboken, NJ: Wiley. Kalliny, M., Dagher, G., Minor, M. S., & De Los Santos, G. (2008). Television advertising in the Arab world: A status report. Journal of Advertising Research, 48(2), 215–223. Kaufman, G. F., & Libby, L. K. (2012). Changing beliefs and behavior through experience-taking. Journal of Personality and Social Psychology. 103(1), 1–19. doi:10.1037/a0027525 Kharroub, T., & Weaver, A. J. (2014). Portrayals of women in transnational Arab television drama series. Journal of Broadcasting & Electronic Media, 58(2), 179–195. doi:10.1080/08838151.2014.906434 Kline, R. B. (2004). Beyond significance testing. Washington, DC: American Psychological Association. Kraidy, M. (2010). Reality television and Arab politics. New York, NY: Cambridge University Press. Kraidy, M., & Khalil, J. (2009). Arab television industries. London, UK: Palgrave Macmillan. 672 Tamara Kharroub and Andrew J. Weaver International Journal of Communication 13(2019) La Pastina, A., & Straubhaar, J. (2005). Multiple proximities between television genres and audiences. Gazette, 67, 271–288. Liebes, T., & Katz, E. (1990). The export of meaning: Cross-cultural readings of Dallas. New York, NY: Oxford University Press. Mellor, N. (2011). Introduction. In N. Mellor, M. Ayish, N. Dajani, & K. Rinnawi (Eds.), Arab media: Globalization and emerging media industries (pp. 1‒11). Cambridge, UK: Polity Press. Reeves, B., & Miller, M. M. (1978). A multidimensional measure of children’s identification with television characters. Journal of Broadcasting & Electronic Media, 22(1), 71–86. Rosengren, K. E., Windahl, S., Hakansson, P. A., & Johnsson-Smaragdi, U. (1976). Adolescents’ TV relations: Three scales. Communication Research, 3, 347–366. doi:10.1177/009365027600300401 Sakr, N. (2006). The impact of commercial interests on Arab media content. In Emirates Center for Strategic Studies and Research (Ed.), Arab media in information age (pp. 61–88). Abu Dhabi, UAE: Emirates Center for Strategic Studies and Research. Spencer-Oatey, H. (2012). What is culture? A compilation of quotations. GlobalPAD Core Concepts. Retrieved from http://go.warwick.ac.uk/globalpadintercultural Straubhaar, J. D. (2003). Choosing national TV: Cultural capital, language, and cultural proximity in Brazil. In M. G. Elasmar (Ed.), The impact of international television: A paradigm shift (pp. 77– 110). Mahwah, NJ: Erlbaum. Tajfel, H. (1978). Differentiation between social groups. London, UK: Academic Press. Tawil-Souri, H. (2008). Arab television in academic scholarship. Sociology Compass, 2(5), 1400–1415. Thompson, B. (2004). Exploratory and confirmatory factor analysis: Understanding concepts and applications. Washington, DC: American Psychological Association. Trepte, S. (2004). Soziale dentitat und Medienwahl. Eine binationale Studie zum Einfluss von GenderIdentitat and Nationaler Identitat auf die Auswahl unterhaltender Medieninhalte [Social identity and media choices: The influence of gender identity and national identity on selective exposure]. Medien & Kommunikationswissenschaft, 52(2), 230–249. Trepte, S. (2006). Social identity theory. In J. Bryant & P. Vorderer (Eds.), Psychology of entertainment (pp. 225–274). Mahwah, NJ: Erlbaum. International Journal of Communication 13(2019) Selective Exposure and Identification 673 Trepte, S., & Krämer, N. (2007). Expanding social identity theory for research in media effects: Two international studies and a theoretical model. Hamburger Forschungsbericht zur Sozialpsychologie [Hamburg Social Psychology Research Papers], 78, 1-37. Hamburg, Germany: Universität Hamburg, Arbeitsbereich Sozialpsychologie. Waisbord, S. (2004). McTV: Understanding the global popularity of television formats. Television and New Media, 5(4), 359–383. doi:10.1177/1527476404268922 Weaver, A. J. (2011a, November). An examination of the potential mediators of the relationship between actors’ race and intention to view movies. Paper presented at the annual meeting of the National Communication Association, New Orleans, LA. Weaver, A. J. (2011b). The role of actors’ race in White audiences’ selective exposure to movies. Journal of Communication, 61, 369–385. doi:10.1111/j.1460-2466.2011.01544.x Copyright of International Journal of Communication (19328036) is the property of University of Southern California, USC Annenberg Press and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. INTERNATIONAL JOURNAL ON MEDIA MANAGEMENT 2019, VOL. 21, NOS. 3–4, 177–192 https://doi.org/10.1080/14241277.2019.1695257 Prevalence of Business Models in Global OTT Video Services: A Cluster Analysis Eun-A Park Western Colorado University, USA ABSTRACT This paper is an empirical analysis of the prevalence of business models in the Over The Top (OTT) video content distribution sector. From a review of the literature on taxonomies of OTT business models, it identifies five attributes of an OTT content distribution platform: ownership, vertical integration with content producers, platform/multiplatform compatibility, service type, and revenue model. Applying cluster analysis to SNL Kagan’s global database of 798 OTT distribution networks, the most five commonly occurring combinations of attributes are identified. The paper concludes by discussing the characteristics of these business models. ARTICLE HISTORY Received 18 February 2019 Revised 21 August 2019 Accepted 10 November 2019 Introduction The increasing bandwidth of broadband networks has created opportunities for OTT services to enter and erode traditional broadcasting markets. In the United States, subscriptions to traditional cable television services have shrunk to 94 million households in 2018 (or 74% of the estimated 127 million US households) (Spangler, 2018). This trend is especially strong among the younger generation (18–34), who are much more likely to opt for alternative video delivery services. In Europe, national pay-TV penetration levels vary between 24 and 97 percent, and declines have not reached the dramatic US levels: however, most new revenue growth is coming from non-traditional Subscription Video on Demand (SVOD) services (Ene, 2019). As the traditional cable television providers continue to lose customers, the future seems bright for OTT video providers. Sensing opportunity in the OTT video services market, a number of different entities, including traditional broadcasters, new content providers, and telecom companies have deployed a variety of competitive offerings. The growth in OTT video services is exemplified by the rise of Netflix in the United States. Initially launched as a DVD mailing service, it launched its OTT video streaming service in 2007 (Voigt, Buliga, & Michl, 2016). By 2010, Netflix was drawing a larger share of subscribership and revenue from online streaming CONTACT Eun-A Park epark@western.edu Communication Arts, Language & Literature, Western Colorado University, 1 Western Way, Gunnison, CO 81231, USA © 2019 Institute for Media and Communications Management 178 E.-A. PARK (Netflix, 2010), demonstrating the viability of an OTT-based business strategy. Since then, Netflix has increased its subscriber base and expanded internationally, reaching an overall subscriber count of 104 million in the second quarter of 2017 (Statista, 2017). Yet, Netflix is largely an exception in the global OTT market, where most new entrants are struggling to find traction (Agnese, 2016), and even for Netflix, subscriber growth might have already plateaued in its major markets (Newman, 2016), and revenues are no longer growing exponentially (Kim, 2016). Data show that, as of late 2016, there were more than 800 OTT video providers worldwide (SNL Kagan, 2017). The emergence of these providers has shaken up the complex vertically integrated broadcast television and cable industries in many countries (Skot, 2014), and led to phenomena such as “cord-cutting,” and “cord-shaving” (Arolovitch, 2015; Banerjee, Rappoport, & Alleman, 2014).1 Due to these paradigm shifts, scholarly studies have examined the business models and strategic positioning of TV players (Abreu, Nogueira, Becker, & Cardoso, 2017; Aidi et al., 2013; Ross & Erasmus, 2013). Other insights are provided by consulting firm reports or the trade press (ABI Research, 2014; Song, 2013, 2014). These scholarly studies and industry reports point to a diversity of business models that have been deployed, including subscription-based models, advertising-supported models, and multiscreen strategies. But none of these models have proved to be consistent performers regarding revenue generation and profitability. Thus, despite the promise of OTT services and the threat they pose to traditional broadcasters, there is no single dominant business model in the OTT video distribution sector. Instead, a wide diversity of options exist within each attribute of an OTT platform: for example, platform capability (PC/ Mac, smartphone, tablet, connected TV, game console, Internet streaming players, pay-TV set-top box), revenue models (free/ad-supported, transactional, subscription, app fee, premium content cost), etc. Therefore, the objective of this paper is to investigate the most commonly occurring attributes of the OTT business models. It seeks to find out if there are common features or attributes to OTT business models, and if yes, what their relative frequencies are and how correlated they are with other features/ attributes. Finally, it asks what combinations of attributes are most common among business models in OTT video service. This paper proceeds as follows. In the next section, an overview of the literature on OTT business models is provided, focusing on the features or attributes of models that have been the focus of studies. We then explicate the concept of a “business model” and use the literature to identify the key attributes of OTT video service business models, to be investigated in the data analysis. We then present our data and describe our methodology. The third section presents the analysis, which leads to the concluding chapter. INTERNATIONAL JOURNAL ON MEDIA MANAGEMENT 179 Literature review With the rising popularity of OTT services, providers have experimented with a wide variety of platforms, content sources, revenue models, and multiscreen strategies (Qin & Wei, 2014). Ad-supported models like YouTube have capitalized on significant amounts of user-generated content, while subscription-based services such as Netflix and Amazon Prime Video have invested aggressively in original content (Castillo, 2016). Facebook’s video distribution is growing, and Apple is planning to add video to its music service. Traditional video providers have added more features to their service package to compete with OTT: for example, CBS distributes original TV series exclusively over SVOD networks. Satellite TV operators are transforming themselves into internet MVPDs, such as Viasat to Viaplay, and DISH to Sling. Platforms for the distribution of content are proliferating: personal video recorders, transaction VOD, and subscription VOD, and set-top boxes are deploying many new capabilities such as Roku TV that can pause and catch up live broadcasts. Given the wide diversity of OTT models available, researchers have proposed various taxonomic classifications (Abreu et al., 2017; Crawford, 2015). Abreu et al. (2017) surveyed video content offerings from 62 countries and proposed a classification based on a 2 × 2 matrix, with the provider (managed operator network or OTT) as one dimension and type of content (linear or non-linear) as the other dimension. Therefore, the four categories of viewing are linear offerings from managed operator networks (traditional television), non-linear offerings from managed operator networks (time-shifted and Video on Demand, VOD services), linear OTT services (streaming content from traditional TV stations) and the market-disrupting non-linear OTT services (Netflix et al.). Abreu et al. (2017) organize business models of TV viewing (time-shifted TV, VOD, personal video recorders, catch-up TV, subscription VOD, app-based distribution, etc.) along with this 2 × 2 classification.2 The results show that 34 countries had one or more operators offering Catch-up TV services. Most operators with Catch-Up TV services also offered short and long-form online TV. Video revenue is made up of subscription fees and advertising revenue, as well as electronic sell-through retail and on-demand revenue from online services delivering TV and video content. Similarly, Crawford (2015) divided OTT video distributors into four categories: OTT aggregators of original and licensed content;3 traditional pay-TV distributors deploying multiscreen, online-offline, “TV everywhere” strategies; individual content owners, especially sports leagues (MLB, NBA); and device manufacturers who distribute content either on their own or in partnership with other groups.4 While Abreu et al. focus primarily on the provider and the type of viewing experience offered, others focus on the revenue models (Martínez, NavioMarco, & Perez-Leal, 2017; Waterman & Sherman, 2016). Martínez et al. 180 E.-A. PARK (2017) differentiate between three models: fixed payments, pay-per-view, and subscription. Waterman and Sherman (2016), in their analysis of the economics of online video, focus more on the revenue models. They identify five basic online video revenue models: à la carte rentals and purchases, or VOD; subscription; ad-supported professional content; ad-supported user-generated content; and verification-dependent, bundled content. VOD systems allow users to select and view from a menu of choices on a per play basis, while subscription services charge a periodic (monthly, annual, etc.) flat fee for access to an entire library of content. Ad-supported networks, whether based on professional or user-generated content, do not charge the user directly for access but finance themselves through advertising. Hybrid models, involving both subscription and advertising, or “freemium” models in which users willing to pay a fee are exposed to no or less advertising are also possible. Finally, verificationdependent, bundled content models have been used by traditional broadcasters who allow authenticated offline subscribers to access streamed content online. This is common with broadcasters’ multiscreen, “TV everywhere” strategies (Waterman & Sherman, 2016). Several authors have also focused on the content of OTT video (Martínez et al., 2017). A primary differentiation is between short-form and long-form content. The short-form video, roughly less than 10 minutes in length, includes much user-generated content, as well as excerpts from television shows, music videos, trailers, and teasers. The long-form video includes programming such as full-length television shows, documentaries, and movies. There is an association between the type of video and revenue models, with the short-form video being predominantly advertising-supported, and long-form content distributed through subscription-based models. Authors have also focused on less commonly observed attributes of OTT platforms. Kim, Kim, Hwang, Kim, and Kim (2017) identified recommendation systems, resolution, and viewing options as essential product attributes of OTT. In their comparison of Chinese and Korean OTT users, they found that Chinese consumers were most persuaded by resolution, followed by the recommendation system and viewing options; Korean consumers ranked the recommendation system as the most valuable attribute, followed by viewing options and resolution. Though these characteristics are important in a competitive sense and influence user adoption of OTT services, there is less attention paid to them in the literature. Business models Much of the literature discussed above on the taxonomy of OTT video providers focus on identifying distinctive qualities of each service provider; specifically, the literature has focused on identifying business models for OTT video services, as a combination of these distinctive qualities. Simply INTERNATIONAL JOURNAL ON MEDIA MANAGEMENT 181 put, a business model represents the “underlying economic logic that explains how we can deliver value to customers at an appropriate cost” (Magretta, 2002, online). Another definition states that a business model is “a “coherent framework that takes technological characteristics and potentials as inputs and converts them into customers and markets into economic outputs” (Chesbrough & Rosenbloom, 2002, quoted in Ghezzi, Cortimiglia, & Frank, 2015, p. 348). Essentially, a business model combines the distinct but interrelated activities of producing a good or service (innovating the product, obtaining raw materials, components or parts, manufacturing, etc.) and distributing it (identifying and reaching customers, marketing and distributing the good or service and delivering after-sale service). Ghezzi et al. (2015) identify four dimensions of business models: value proposition, value creation, value delivery, and value appropriation. The value proposition is the unique product or service that creates value for the customer; in the case of OTT, this is the audiovisual software, with the added functionalities of timeflexible viewing, alternative screen, or multiscreen capability, or mobile portability differentiating it from traditional video offerings. The value creation refers to the organizational characteristics of the provider and includes the supply chains, resources, and production modes enabling the creation of the value proposition. The value delivery implicates the key partnerships, distribution channels and customer relationship strategies that convey the product or service to the customer. Finally, value appropriation refers to the revenue generation, pricing and sharing mechanisms that enable the realization of profit and its allocation to various partners along the value chain. Steffan and Richard (2014) have comprehensively reviewed classification schemes for business models by focusing on their attributes. Commonly used factors to differentiate business models include revenue schemes (sales versus rental versus licensing), relationships with customers (sales personnel versus selfservice versus automated delivery), distribution channels (retail outlets versus online), etc. (see Table 3, p. 25 for a full listing of business model attributes). To identify the business models for OTT video services, we first consulted the literature to locate the most commonly used attributes of OTT video services. From the taxonomies of OTT video service reviewed in the section above, three attributes may be identified: organizational mode (content originators, aggregators, redistributors, etc.), distribution channels (wired/mobile, devices), and revenue models (advertising-funded, subscription, hybrid, etc.). However, several authors also highlighted the importance of the firm’s organizational structure (Crawford, 2015; Waterman & Sherman, 2016). Accordingly, we add two more attributes to this list: ownership (broadcasters, telecommunications companies, Internet-based companies) and vertical integration (integrated or non-integrated). “Vertical integration” refers to the combination of production activities at different stages of the chain of production within one company. Vertical 182 E.-A. PARK integration with a program production company may enable an OTT video provider to have easy access to a library of content. However, current research has not examined how prevalent each type is within an attribute (e.g., what percentage of OTT video providers are advertising funded?), and how the attributes relate to each other (e.g., what is the most common revenue model associated with mobile content delivery?). The sections below attempt to answer these questions, but first, we describe our data. Data analysis The primary data for this analysis is SNL Kagan’s Global OTT provider database. SNL Kagan is a leading firm that provides data and analysis services for the media, entertainment, and communications industries. It was created through the 2007 acquisition of Kagan Research,5 a Monterey, CA-based media research firm founded in 1969, by Charlottesville, VA-based SNL Financial, a business intelligence provider (Qualtrough, 2007). SNL Kagan provides access to a broad range of information and databases on the media and entertainment industries to its clients, including universities. SNL Kagan’s Global OTT provider database lists 798 OTT providers from 71 different countries, and several of their attributes, including platform capabilities, revenue models, and programming sources. The database identifies the name of the OTT service, the service “parent” (the organization offering the service), the geographical region within which the service is offered (e.g., Asia, Western Europe etc.), the national markets in which the service is available, the service type (OTT aggregator, TV everywhere, catch-up TV, etc.), the distribution devices supported (connected TV, game console, smartphone, tablet, etc.), the revenue models (advertising, authenticated, rental), notes to revenue models, and size of the archive. For the data analysis, each OTT service was coded along several dimensions. First, the service “parent” was coded into six categories, based on their primary business: telecommunications company (TEL), broadcaster (BRO), pay-TV provider (PAYTV), media content producer (MEDIA), web company (WEB), and other or unclassified (OTHER). To aid the classification, additional information about companies was collected from their websites when needed. The categorization was exclusive: each service “parent” was assigned only to one category. Second, the service type was coded as reported by SNL Kagan: as a content aggregator (AGGRE), TV everywhere (TV_EVERY), and catch-up TV (CATCHUP). Service types too were coded exclusively. In some cases, it was observed that OTT services offered by the same company, but with different service types were branded with different names; each name, therefore, can be associated exclusively with a service type. It may be noted that it is not the owner of an OTT service that is coded, but the OTT service itself. SNL Kagan’s data includes multiple offerings from the same parent company, INTERNATIONAL JOURNAL ON MEDIA MANAGEMENT 183 with separate codes for each of the OTT services. Therefore, for example, multiple offerings from Disney, such as Hulu and ESPN+ are separately coded for the business model followed by each. Third, the platforms on which the service is distributed was coded. While listing all platforms exhaustively, seven platforms could be identified: connected TV; game console; Internet streaming player; pay-TV set-top box; PC/ Mac; smartphone; and tablet. Since many services were available for multiple platforms, the coding for “platform” was done non-exclusively. Fourth, the revenue model was coded as reported by SNL Kagan into five categories: advertising, authenticated, rental, subscription, and purchase. Fifth and finally, a variable “vertical integration” was defined as present if the parent company of the OTT video service was also a content producer: information about the parent company was researched online to code for this variable. The percentage distribution of attributes for all the categories is provided in Table 1. For example, the parent company of 18 percent of all OTT video services was a telecommunications company, and broadcasters owned 23 percent. Note that the percentage distribution by the platform is more than 100 percent, since ‘platforms’ was coded non-exclusively: an OTT video service may be made available to multiple platforms. A review of the data in Table 1 provides useful insights into the research objectives of this paper. Worldwide, web companies are the entities most often deploying OTT video, with 26 percent, followed by broadcasters (23%) and telecom companies (18%). Only 44 percent of all OTT video providers are vertically integrated with content producers: the rest aggregate content from a variety of sources (56%). Vertically integrated companies are more likely to offer multiscreen TV everywhere strategies (25% of all OTT services, or 57% of vertically integrated OTT services), while 19 percent of all OTT services pursue a catch-up TV strategy (43% of vertically integrated OTTs). An overwhelming majority of OTT services are accessible over PC/MAC platforms (96%), tablets (90%), and smartphones (89%). A little over half (53%) of all OTT services are accessible on connected TV sets, while streaming media players (35%) and game consoles (29%) are also popular. Perhaps not surprisingly, only 11 percent of all OTT services are accessible over payTV set-top boxes, since these devices are controlled by cable/broadband providers, whose video services compete with OTT services. In terms of revenue models, OTT services are split almost evenly between authenticated content providers, subscription services, and advertising-based models, with each accounting for a quarter of the global market. The remaining OTT video services follow a rental or purchase based model. To investigate further the relationships among the variables, bivariate correlations were calculated and reported in Table 2. Significant correlations were observed between several of the variables: telecom providers were likely to deploy TV Everywhere strategies using an authenticated content revenue model. 184 E.-A. PARK Table 1. Percentage distribution by category. Variable Ownership Telecom company Broadcasters Pay TV providers Media companies Web companies Others Vertical Integration Service Type Aggregators TV Everywhere Catch up TV Platforms Connected TV Game Consoles Screaming media players Pay TV set top box PC/MAC Smartphone Tablets Revenue Model Authenticated Subscription Advertising Rental Purchase ID Percentage TEL BRO PAY TV MEDIA WEB OTHERS VERTICAL .18 .23 .14 .17 .26 .03 .44 AGGRE TV_EVERY CATCHUP .56 .25 .19 CONNECTV GAMECONS STREAMIN SETTOP PCMAC SMARTP TABLET .53 .29 .35 .11 .96 .89 .90 AUTHEN SUBSCR ADVERT RENTAL PURCHA .25 .26 .25 .15 .09 Note: N = 798 for ownership, vertical integration, service type, and platform; N = 797 for revenue model (1 missing data) Broadcasters were strongly likely to deploy OTT services like catch-up TV, which was also likely to be advertising supported. Meanwhile, web companies did not display a clear pattern regarding revenue models, though they were somewhat more likely to use subscription-based, rental or purchase models, and somewhat less likely to use authenticated content models. A potential problem was the perfect negative correlation between vertical integration and an aggregation service type: this is examined more in the cluster analysis section below. Interesting patterns were also observed in terms of devices. The strongest correlation was observed between smartphones and tablets: OTT services available over one were almost certainly available over the other as well. Weaker positive correlations were also observed between connected TVs, streaming media players, and game consoles. To further investigate these patterns, cluster analysis was conducted. Cluster analysis To identify the various OTT business models, a hierarchical cluster analysis followed by a k-means analysis was performed using SPSS. Each subject’s relative standing on each of the variables was estimated by computing factor -.191** -.204** -.280** -.076* .269** PAY_TV MEDIA WEB OTHERS VERTICAL .514** -.229** CATCHUP -.237** -.098** -.103** 0.006 0.014 STREAMIN SETTOP PCMAC SMARTP TABLET -.210** -.089* -.098** ADVERT RENTAL PURCHA -.091* -.109** .439** 0.014 -.307** -.123** -.158** -.185** -.090* .486** 0.03 0.006 0.032 0.028 -0.061 -.086* -.200** -0.029 -.166** -.156** .484** -.300** .300** -0.064 -.237** -.180** 1 PAY_TV .087* .166** -0.057 -0.028 -0.03 .750** -.307** -.322** .322** -.087* -.319** -.242** -.218** 1 BRO .104** .171** -0.045 0.034 -.197** -.090* -.117** -0.024 0.064 0.02 -0.045 -0.037 -.112** -.198** .261** -.261** -.072* -.264** 1 MEDIA .127** .139** -0.036 .173** -.335** 0.042 .084* 0.066 -.084* .365** .235** .349** -.276** -.336** .511** -.511** -.095** 1 WEB .152** 0.068 -0.057 -0.004 -.093** -0.05 -0.044 0.031 .070* .116** .072* .087* -.077* -.093** .142** -.142** 1 OTHERS -.225** -.349** .173** -.392** .654** -0.004 0.021 -.092** -.092** -.388** -.207** -.359** .543** .655** -1.000** 1 VERTICAL .225** .349** -.173** .392** -.654** 0.004 -0.021 .092** .092** .388** .207** .359** -.543** -.655** 1 AGGRE -.177** -.242** -.338** -.343** 1.000** -0.001 0.015 -.187** -.207** -.346** -.179** -.339** -.279** 1 TV_EVERY -.090* -.175** .595** -.117** -.279** -0.004 0.009 .092** .112** -.109** -0.065 -.080* 1 CATCHUP .080* .168** 0.001 .144** -.337** .156** .101** .148** .193** .430** .454** 1 CONNECTV 0.039 0.015 -0.043 .182** -.178** .148** .109** .093** .265** .418** 1 GAMECONS **Correlation is significant at the 0.01 level (2-tailed). *Correlation is significant at the 0.05 level (2-tailed). .511** -.162** AUTHEN SUBSCR REV MODEL -.213** -.194** CONNECTV GAMECONS PLATFORM -.269** AGGRE TV_EVERY SERVICE 1 -.257** BRO TEL TEL OWNERSHIP Table 2. Correlations. .205** .079* -0.043 .188** -.345** .184** .173** .099** .125** 1 STREAMIN 0.005 .084* 0.046 .087* -.207** -.112** -.148** 0.068 1 SETTOP 0.058 0.061 0.065 0.035 -.188** 0.024 -0.003 1 PCMAC -0.034 -.086* -0.012 .083* 0.021 .891** 1 SMARTP -0.03 -0.02 -0.039 0.069 0.005 1 TABLET -.177** -.242** -.338** -.343** 1 AUTHEN -.181** -.248** -.346** 1 SUBSCR -.179** -.245** 1 ADVERT -.128** 1 RENTAL INTERNATIONAL JOURNAL ON MEDIA MANAGEMENT 185 Cluster 1 2 3 4 5 Cluster 1 2 3 4 5 N 429 199 146 11 4 PAY TV .04(.20) .43(.50) .02(.14) .27(.47) .00(.00) SETTOP .13(.34) .00(.00) .18(.39) .18(.41) .75(.50) Platforms STREAMIN .53(.50) .07(.25) .25(.44) .00(.00) .75(.50) BRO .10(.31) .01(.07) .88(.32) .00(.00) 1.00(.00) GAMECONS .38(.49) .16(.36) .23(.42) .09(.30) 1.00(.00) TEL .08(.27) .53(.50) .00(.00) .73(.47) .00(.00) CONNECTV .70(.46) .24(.43) .45(.50) .09(.30) 1.00(.00) N 429 199 146 11 4 PCMAC .98(.14) .90(.30) 1.00(.00) .91(.30) 1.00(.00) MEDIA .26(.44) .04(.20) .08(.30) .00(.00) .00(.00) Ownership Table 3. Means and standard deviations of variables by cluster group. SMARTP .88(.32) .90(.30) .90(.29) 1.00(.00) .75(.50) TABLET .90(.30) .90(.30) .90(.30) 1.00(.00) 1.00(.00) WEB .48(.50) .00(.00) .01(.08) .00(.00) .00(.00) AUTHEN .00(.00) 1.00(.00) .00(.00) .00(.00) .00(.00) OTHERS .04(.19) .00(.00) .00(.00) .00(.00) .00(.00) SUBSCR .43(.50) .00(.00) .14(.35) .00(.00) .00(.00) TV_EVERY .00(.00) 1.00(.00) .00(.00) .00(.00) .00(.00) ADVERT .17(.38) .00(.00) .82(.39) 1.00(.00) .00(.00) RENTAL .26(.44) .00(.00) .00(.00) .00(.00) 1.00(.00) Revenue model AGGRE 1.00(.07) .00(.00) .01(.08) 1.00(.00) 1.00(.00) Service Type PURCHAS .14(.35) .00(.00) .04(.20) .00(.00) .00(.00) CATCHUP .00(.07) .00(.00) .99(.08) .00(.00) .00(.00) 186 E.-A. PARK INTERNATIONAL JOURNAL ON MEDIA MANAGEMENT 187 scores, which were then used as input variables for clustering (Cook, 2005). The more strongly the values (i.e., factor scores) are situated in the negative range, the more the company’s factors are below the cluster average, while positive values indicate a rating above the cluster average. Distances between the clusters were calculated with the Euclidean distance measure, and aggregation of clusters was performed with Ward’s procedure. To reflect the true structure of the data set, the elbow criterion was used to decide on the number of clusters, which resulted in choosing a five-cluster solution as the most appropriate representation of the data (Yim & Ramdeen, 2015). A potential problem with the clustering procedure was the perfect negative correlation between vertical integration and an aggregation service type. If there is a strong correlation between variables, they are not sufficiently unique to identify distinct segments. If highly correlated variables are used for cluster analysis, specific aspects addressed by these variables will be overrepresented in the clustering solution. In this regard, absolute correlations above 0.90 are always problematic (Sarstedt & Mooi, 2014). Thus, the vertical integration variable (VERTICAL) was dropped because it was highly correlated with the aggregator service type. Means and standard deviations of each variable for the five cluster groups are presented in Table 3. To detect differences in scores across the different variables between the different clusters, ANOVA was performed followed by a Tukey test. Below, the attributes of the five clusters are discussed. Cluster 1 (N = 429). This cluster is mostly comprised of web companies, though some media companies are also represented. Regarding content, these OTT providers are exclusively aggregators, and the primary revenue model is subscriptions, though a majority does not favor it. A substantial percentage also use rental models, advertising, and purchase. The primary distribution platforms are PC/Mac, smartphones, and tablets, though connected TV and streaming media are also used. In terms of the operations of their business model, companies in this group are non-traditional video providers without access to in-house program production facilities or business alliances with program providers. Therefore, their service type is aggregation – putting together a roster of programming from diverse sources. Also, as new entrants, companies in this group opportunistically choose revenue models with some – those with a more extensive audience base – favoring advertising and others opting for rental or subscription models. Finally, these companies without access to owned distribution platforms such as cable broadband or DSL networks have to target new video consumption devices such as computers, smartphones, and tablets. Cluster 2 (N = 199). This cluster is dominated by telecom and pay-TV companies with very little participation by other types of owners. The group is defined by its exclusive use of “TV everywhere” strategies and authenticated 188 E.-A. PARK content. The model relies strongly on distribution via PC/Mac, smartphones, and tablets, with even less use of connected TV and other modes than Cluster 1. A large part of this cluster is comprised of non-traditional video providers (telecom companies) who, however, have the benefit of access to DSL infrastructures. A second category is pay-TV companies formerly affiliated with cable/broadband networks. Both groups have seen their audiences under attack due to cord-cutting and falling pay-TV subscription rates. Therefore, both groups appear to aim with a “TV everywhere” business model to migrate with their customers to new viewing platforms such as PC/Mac, smartphones and tablets. Cluster 3 (N = 146). This group is comprised overwhelmingly of broadcasters, though some media companies are also represented. The service type is almost exclusively “catch-up TV,” funded primarily through advertising, with a little subscription. As in the case of the other clusters, Cluster 3 too tends to distribute over PC/Mac, smartphones, and tablets, though it also has a substantial presence on other media, including connected TV, gaming consoles, and streaming media players. The major strength of companies in this cluster, primarily comprised of broadcasters, is expertise in the production of original programming and access to large programming libraries. However, these companies have seen a steady erosion of their live audience ratings, especially in the United States, but also in many leading markets, with negative consequences for advertising revenues. Launching a “catch-up TV” business model may fulfill two objectives: one, to increase the appeal of their programming by combining it with offerings from multiple broadcast stations; and two, to increase total viewership and stanch the erosion of advertising revenue. Cluster 4 (N = 11). A relatively small cluster is comprised of telecom companies and some pay-TV companies, as in the case of Cluster 2. It may, therefore, be considered an alternative business model for telecom and payTV companies, mainly differentiated from the latter by its reliance on an aggregation service model (whereas Cluster 2 used a “TV everywhere” strategy) and an advertising-based revenue model (contrasted with Cluster 2’s reliance on authenticated content). As in the case of all clusters, Cluster 4 relies on distribution over PC/Mac, smartphone, and tablet, but unlike Cluster 2, there is practically no presence on any other platform. The contrast between Clusters 2 and 4 hints at the different strategies adopted by telecom companies, that form a large part of both clusters, as they pursue video audiences. Telecom companies are not traditional video distributors, but have over the past expertise and experience in video program delivery through acquisitions and/or market participation. However, those that have acquired greater expertise in video programming appear to follow a “TV everywhere” strategy, while those that have not have opted for an aggregation strategy. INTERNATIONAL JOURNAL ON MEDIA MANAGEMENT 189 Cluster 5 (N = 4). The smallest cluster, it is comprised exclusively of broadcasters; it may be an alternative model for the broadcasters in Cluster 3. These aggregators rely exclusively on a rental based revenue model. Of all the clusters, they are the most “platform agnostic,” equally present on all distribution platforms: connected TV, game consoles, streaming media, payTV set-top boxes, PC/MAC, smartphones and tablets. But uniquely among the clusters, Cluster 5 uses set-top boxes heavily, but perhaps this is an artifact of small numbers. In the next section, we return to our research questions and present our conclusions. Discussion and conclusion While numerous studies have examined business models in the new OTT video services sector, few quantitative analyses have been aimed at analyzing the relative prevalence of business models, or the elucidation of the most common attributes of these business models. To fill this gap in the literature, the objective of this paper was to investigate whether there are common features or attributes to OTT business models, and if yes, what their relative frequencies are and how correlated they are with other features/attributes. Finally, it asks what combinations of attributes are most common among business models in OTT video service. Using SNL Kagan’s global database of OTT video service providers, this paper coded five attributes of business models identified based on the literature: ownership, service type, vertical integration, platforms, and revenue models and compared their relative frequencies and correlations across dimensions. Cluster analysis was performed to identify the most common combinations of attributes in business models for OTT video services. The first takeaway from the cluster analysis is that ownership is not fully coincidental with business models. Though clusters are dominated by one or the other form of ownership, diverse owners are represented within each cluster. Media companies are split between subscription-based aggregators (Cluster 1) and advertising-funded catch-up TV (Cluster 3). Similarly, telecom and pay-TV companies favor either TV everywhere with authenticated content (Cluster 2) or advertising-based aggregators (Cluster 4). Broadcasters are mostly advertising-funded “catch-up TV providers (Cluster 3), though rental based aggregation might be an alternative business model for them. In other words, telecommunications companies, broadcasters, media firms, and web businesses offer OTT video services using business models not common within their ownership category. A second takeaway from the cluster analysis was that service types and revenue models are more tightly coupled than ownership types. Depending on the service type, specific revenue models appear to be preferred: TV 190 E.-A. PARK everywhere is provided exclusively with authenticated content, while catchup TV is predominantly advertising-funded. Aggre...