Download as pdf or txt
Download as pdf or txt
You are on page 1of 18

Article

Journal of International Marketing


1-18
Narrowband Influencers and Global Icons: ª American Marketing Association 2020
Article reuse guidelines:

Universality and Media Compatibility in the sagepub.com/journals-permissions


DOI: 10.1177/1069031X19897893
journals.sagepub.com/home/jig
Communication Patterns of Political
Leaders Worldwide

Renana Peres, Sunali Talwar, Liav Alter, Michal Elhanan,


and Yuval Friedmann

Abstract
This article analyzes how political leaders communicate with their target audiences and examines whether they adopt a country-
specific communication persona, or react to the global media-intensive environment by offering more universal communication.
Politicians communicate through presentational (e.g., social media) and representational (e.g., press) outlets, and the compatibility
between these outlets represents the leader’s effectiveness in transmitting the desired messages to the audience. The authors of
this study suggest a theoretical framework that classifies public figures’ communication along two dimensions: universality
(particular–universal) and media compatibility (low–high). The authors used language processing tools to study the sentiment,
topic mixture, and use of pronouns by 61 global world leaders in more than 300,000 messages from the leaders’ Twitter accounts
and press articles. The results show a high level of universality across political leaders in sentiment, topic mixture, and pronoun
usage. The media compatibility is high, with Twitter being slightly more positive. Most leaders fall within the categories of
Cosmopolitan Antagonist (high universality, low media compatibility) and Global Icon (high universality, high media compatibility).
Overall, the sentiment of their communications is positive. Popular topics include diplomacy, economy, corruption, and the Arab
world. No significant relationship was found between the sentiment or communication topics and country characteristics.

Keywords
political marketing, social media, press, politicians, presentational media, representational media, universality, media compatibility,
globalization, content analysis, topic analysis, LDA, sentiment analysis, communication of political leaders

Much of the brand building of political leaders is done through country and people. On the other hand, many current topics of
communication with their audiences (Bennet and Iyengar interest are related to global issues. Moreover, politicians oper-
2008; Dahlgren 2009). The current age of global communica- ate in a multichannel, media-intensive environment, which, as
tion platforms, access to information, and global press cover- suggested by recent trends in communication theory, is a strong
age poses new challenges to political leaders who want to driver in enhancing globalization (Krotz 2007) and influences
design, develop, manage, and control their public persona. public figures (Van Aelst et al. 2008) to take a more global
We examine these challenges using two communication view of their media identity and adopt a more universal com-
dimensions. The first dimension is the level of universality. munication style.
On one hand, the political leader of a country can be perceived
as a salient, visible national icon. In branding terminology, a
country’s leader, at least while in office, can be viewed as a Renana Peres is an Associate Professor of Marketing, Hebrew University of
distinct national brand or as a recognizable brand element of Jerusalem, Israel (email: renana.peres@mail.huji.ac.il). Sunali Talwar is an
the country, similar to the flag, major monuments, and famous independent researcher working in Germany (email: sunalitalwar@gmail.
citizens (French and Smith 2010). A leader’s communication is com). Liav Alter is a research student, Hebrew University of Jerusalem,
Israel (email: liav.alter@mail.huji.ac.il). Michal Elhanan is a research student,
mainly with the local population. Therefore, one would expect
Hebrew University of Jerusalem, Israel (email: michal.elhanan@mail.huji.ac.il).
country leaders to adopt a country-specific communication Yuval Friedmann is a research student, Hebrew University of Jerusalem, Israel
persona that reflects the interests and characteristics of their (email: yuval.friedmann@mail.huji.ac.il).
2 Journal of International Marketing XX(X)

 The Cosmopolitan Antagonist talks about diverse


topics resembling the topics generally discussed, but the
media’s view of these topics is incompatible with the
leader’s. Examples are Kim Jong Un, who as a head of
Cosmopolitan
Peculiar Notorious state speaks about common topics such as economy,
Low

Antagonist
e.g. Bill Cosby diplomacy, and military, but the media (outside North
Madia compability

e.g. Kim Jong Un, Rush Limbaugh


Korea) is critical of his opinions and claims, and Rush
Limbaugh, who expresses strong right-wing opinions on
many topics and is not regarded positively by many
other media outlets.
Narrowband
Global icon
 The Global Icon communicates about topics that are
Influencer commonly discussed by others, and the media’s opinion
High

e.g. Michelle Obama, Emma Watson


e.g. Kate Middleton, Greta Thunberg
and perception of the leader are highly aligned with the
leader’s opinion. Examples are Michelle Obama, who as
the first lady of the United States advocated for poverty
Parcular Universal awareness, education, and healthy lifestyles and
Universality received positive coverage from the media, and Emma
Watson, an actress who is also a women’s rights activist
Figure 1. Four prototypical patterns of public communication. and was named a UN Women Goodwill Ambassador.

Studying these two dimensions is interesting for three rea-


The other dimension is the extent to which the leaders sons. First, brand researchers have long wondered how globali-
can communicate the identity they selected. Politicians com- zation processes affect brand perceptions (Cayla and Arnould
municate with their audiences through a mix of presentational 2008; Diamantopoulos et al. 2019; Gürhan-Canli, Sarıal-Abi,
channels (e.g., social media), where the leader has the full and Hayran 2018; Steenkamp 2019). Because political leaders
freedom of self-presentation, and representational channels are salient national brands, their globalization process might
(e.g., press, television), which are based on journalism and are provide early insights as to what one can expect for commercial
supposed to be more objective (Hepp, Breiter, and Hasebrink brands. Second, the citizens of the world’s nations are increas-
2018). High compatibility in information and sentiment ingly facing challenges of a global nature, such as immigration,
between these two types of channels implies that the represen- environmental issues, and the future of the European Union and
tational media generally reflects the leader’s self-presentational other regional treaties. More than ever, these issues influence the
goals. Low media compatibility indicates a high level of inde- lives of individual citizens. Coping with these challenges
pendence and investigative motivation by the press. requires world leaders to create a communication infrastructure
We organize these two dimensions into a 2  2 matrix that goes beyond national boundaries, and therefore it is of inter-
(Figure 1) to suggest the following prototypical communication est to determine whether such infrastructure has already been
patterns. In this article we focus on political leaders, but the created. Third, one of the outcomes of intensive, multichannel
classification can apply to any public figure. media usage is the continuous spillover of contents between
media platforms. Specifically, the presentational and repre-
 The Peculiar Notorious communicates about specific sentational channels increasingly overlap. Still, politicians
topics that deviate from the average distribution of cur- often complain about media bias (Brants et al. 2010; D’Ales-
rent affairs. The media coverage of the leader is not sio and Allen 2000; Geis 1987). Hence, studying the compat-
compatible with the leader’s own discussion of these ibility in communication styles between presentational and
topics on social media. An example (from the nonpoli- representational communication outlets is important to under-
tical world) is Bill Cosby, who is connected to the spe- stand the real variety and objectivity in the information that
cific subject of TV shows and has received negative citizens receive on their leaders (Shapiro and Hemphill 2017).
media coverage recently (i.e., large gap between his From an international branding perspective, political leaders
representation and self-presentation). are an interesting case of personal branding in a media-
 The Narrowband Influencer is interested in uncom- intensive environment. Similar to commercial brands, they
mon, specialized subjects, and the press coverage of cope with the interplay of their local and international identi-
his/her ideas is highly compatible with his/her opinion. ties. However, they are people, and as personal brands, they use
Thus, the leader controls the message and leads public communication in ways that are much more elaborate than the
opinion in the subject of interest. Examples are Kate communication methods of commercial brands. In addition,
Middleton, whose media coverage deals mostly with their ability to communicate their identity depends not only
fashion and royal family activity and is generally posi- on their own marketing communications, but also strongly on
tive, and Greta Thunberg, a climate activist whose their press coverage. As personal branding gains an increasing
media coverage usually agrees with her opinions. importance in art, entertainment, sports (Chadwick and Burton
Peres et al. 3

2008), high tech (Anderson 2013), and the organizational envi- Conceptual Background
ronment (Ulrich and Smallwood 2007), insights on communi-
cation patterns are of importance. Politicians as Brands
The goal of this article is to provide a first data-intensive, Research in political marketing relates to political candidates
systematic, and comparative investigation of universality in as one component in a trinity consisting of the party as the
the communication styles of the world’s major political brand, the politician as its tangible characteristics, and policy
leaders. We classify the leaders into the four quadrants of as core service offerings (Henneberg and O’Shaughnessy
the universality–compatibility matrix and ask how univers- 2007). Recently, voters’ consideration has shifted from focus-
ality and media compatibility relate to the leader’s country ing on parties to focusing on the candidates (Guzman and
characteristics. To do so, we collected a massive data set of Sierra 2009), and, consequentially, researchers have started to
the communications of 61 major political leaders. We study political candidates as standalone brands (Guzman and
extracted the entire Twitter communication from their Twit- Sierra 2009; Hockett 2005; Phipps, Brace-Govan, and Jevons
ter accounts, and retrieved from the LexisNexis database the 2010; Smith 2009). This focus shift is in line with the increas-
press articles in which they were mentioned between Janu- ing interest in personal branding in other domains, such as art,
ary 2018 and July 2019. Thus, we captured the leader’s entertainment, social media, and high tech (Anderson 2013;
communication in a presentational social media outlet Chadwick and Burton 2008; Ulrich and Smallwood 2007).
(Twitter) and in a representational media outlet (printed The expansion and proliferation of communication outlets
press), which is supposed to be a more objective, in the past decades has created profound changes in the way
journalism-based medium. We used text processing tools, people, especially politicians, manage their personal brand.
such as translation, elimination of stop words, and cleaning The main concept in this discussion is “mediatization” (Coul-
to preprocess the data, and we implemented automated lan- dry and Hepp 2017). The term relates to the omnipresence of
guage processing tools to analyze the sentiment, topic mix- multiple communication channels, which interact with each
ture, and usage of “I,” “we,” “you,” and “they” messages. other and collectively transform (i.e., “mediatize”) many
We tested the dependence of these elements on country aspects of our everyday lives (Hjarvard 2013; Krotz and Hepp
characteristics, such as the gross domestic product (GDP) 2013). In this article, we discuss two challenges faced by per-
per capita, the continent, and Hofstede’s cultural dimensions sonal brands in general, and political leaders in particular, in a
(Hofstede 2011). mediatized world.
Our results indicate a striking universality among political
leaders. Leaders are similar to each other in their sentiment,
topic mixture, and usage of pronouns. Overall, the sentiment is
Particularism Versus Universalism
positive, and the discussed topics are mainly diplomacy, econ- The increased complexity of the communication landscape in
omy, corruption, and the Arab world. “We” messages are used the past decades is both an outcome and a driver of globaliza-
the most, followed by “I” messages. The results are robust tion. In one direction, globalization facilitates the creation of
across communication channels, with Twitter being slightly global communication infrastructures and digital communica-
more positive on sentiment score than the press articles. With tion platforms. For example, the availability of internet infra-
a few exceptions, we saw almost no significant dependence on structures, usage of the English language, and communalities in
country characteristics. Relative to Africa, leaders from Asia, consumer tastes facilitate the operation of social media chan-
Europe, and Central America use first-person messages more nels, such as Instagram and Twitter. However, a recent stream
than second- and third-person messages while tweeting. In of research focuses on the other direction to argue that the
addition, leaders from countries with shorter-term orientation variety and ubiquity of communication channels is not only
were found to use first-person pronouns, relative to second- or an outcome but also an important driver of globalization. The
third-person pronouns, more often on Twitter than leaders from main argument of this research stream is that mediatization is a
countries with longer-term orientations do, but these effects are core driver of globalization (Krotz 2007): the exposure to mul-
small. Of the four quadrants on the universality–compatibility tiple communication channels helps connect people, creates
matrix, most leaders were classified into the rightmost quad- ties that cross national boundaries, and helps to create a global
rants: Cosmopolitan Antagonists (topics similar to those of the culture.
average leader, with relatively low media compatibility) and Political leaders depend heavily on communication with
Global Icons (topics similar to those of the average leader, with their citizens. Therefore, mediatization processes have deeply
high media compatibility). affected the political arena (Enli and Skogerbø 2013; Howard
The rest of the article is organized as follows. First, we and Parks 2012). Marshall (2015), shows how people change
present the conceptual background. Then we describe the many of their everyday behaviors to fit with their online per-
data, explain the preprocessing, and describe the sentiment sona; Strömbäck and Van Aelst (2013) demonstrate how polit-
and content analysis procedures. Finally, we discuss the ical parties adapt their structure and organization to respond to
results of the sentiment, topic mixture, and pronoun fre- the increasing complexity of communication; Van Aelst et al.
quency analysis and present the leaders’ classification on (2008) claim, using the case study of the 2003 election cam-
the communication matrix. paign in Belgium, that the media tend to promote certain
4 Journal of International Marketing XX(X)

politicians, and therefore politicians adapt their actions and American conservatives founded watchdog groups to focus
personality to follow the “media logic” in order to receive a attention on what they perceived as biased reporting by the
higher level of media attention. American press (Geis 1987), although proof of such bias is
The fact that political leaders operate in a global, mediatized difficult to find (D’Alessio and Allen 2000).
world raises an intriguing question pertaining to their commu- One of the outcomes of mediatization is the constant inter-
nication styles. On one hand, the political leader of a country is action and spillovers across communication channels and
a brand that is strongly associated with the country and is media environments (Hepp, Breiter, and Hasebrink 2018). Tra-
assumed to have a strong country-of-origin effect. Also, much ditionally, presentational media, meaning communication ini-
of their target audience is local. Therefore, we can expect tiated directly by the transmitting party (e.g., social media,
strong local flavor in the leader’s communication and, conse- announcements, press releases), was distinct from representa-
quently, large differences in the communication styles of lead- tional media, defined as reporting done by a third party (e.g.,
ers. On the other hand, mediatization calls for globalization: the television or press). Presentational media typically differs from
intensive usage of communication outlets of all types, together representational media in that it is more interactive, is inexpen-
with the global nature of many of these outlets and the exposure sive, and is generated by ordinary users instead of professionals
to new audiences, might drive leaders to adopt a more global (e.g., journalists; Hepp, Hjarvard, and Lundby 2015).
communication persona and therefore employ a more homo- The massive growth of social media has reduced depen-
genous communication style that depends less on local dency on the press for information (Shang, Wu, and Li 2017)
characteristics. and has enabled ongoing direct communication between pub-
Commercial brands also cope with the particularism–uni- lic figures and their audiences. Realizing the opportunities
versalism balance. Brand researchers have intensively studied offered by new media settings, politicians across the world
how globalization influences brand perceptions (Cayla and have eagerly adopted social media (Bennett and Iyengar 2008;
Arnould 2008; Diamantopoulos et al. 2019; Gürhan-Canli, Bos, Van der Brug, and De Vreese 2011; Wattal et al. 2010)
Sarıal-Abi, and Hayran 2018; Steenkamp 2019). This influence and are using a mix of presentational and representational
is intricate. On one hand, brands use global positioning to media platforms to communicate with their citizens and vot-
appeal to their consumers (Diamantopoulos et al. 2019; Nijssen ers (Chadwick 2006).
and Douglas 2011), and consumers develop global orientation In mediatized environments, presentational and representa-
that influences their brand perception (Guo 2013). However, on tional media are becoming increasingly interwoven. For exam-
the other hand, the country of origin of the brand has still an ple, Twitter affects political news reporting (Shapiro and
effect on attitudes, preferences, and purchasing decisions Hemphill 2017), and, at the same time, user-generated commu-
related to the brand (Herz and Diamantopoulos 2017; Kock, nicative patterns on social media may involve references to
Josiassen, and Assaf 2019), and cultural dimensions and coun- news articles. Strömbäck and Aelst (2013) claim that social
try characteristics affect consumer-brand interactions (Cayla media affects actual political behaviors, which, in turn, are
and Arnould 2008; Özsomer 2012). reported by the press media. Thus, the boundaries between
This intensive research done for commercial brands has not media environments are blurred. Hence, a question of interest
yet extended to the study of personal brands. Specifically, is the extent to which the two media types differ and whether
research on communication in political science is mostly con- leaders are able to communicate their identity consistently
ceptual (e.g., Krotz and Hepp 2013) or uses case studies (e.g., across presentational and representational outlets.
Enli and Skogerbø 2013), and researchers have not system-
atically monitored leaders’ communication to estimate the Integrating the Dimensions: Four Communication Types
level of universality. Personal brands differ from commercial
brands in, among other things, the complexity of their com- This paper aims to explore how political leaders operate in a
munication with the target audience. Their usage of commu- mediatized word, with respect to the two dimensions discussed
nication is much more elaborate than that of commercial previously. As described in Figure 1, public figures (in our
brands, and their ability to communicate their identity case, political leaders) can be classified on a 2  2 matrix in
depends not only on their own marketing communication but which the horizontal dimension is the level of similarity of the
also strongly on their press coverage. person to the “average” public figure in the same domain (par-
ticular–universal) and the vertical dimension is the level of
compatibility between the presentational and representational
Compatibility of Media Outlets media concerning the leader (low–high).
Political leaders often complain that their ability to communi- In what follows, we describe the data collection and the
cate their identity, goals, and ideas is oppressed by the press, analysis.
which they often claim is biased against them. Brants et al.
(2010) surveyed politicians in the Netherlands and found that
politicians regard journalists as too sensationalist, too event
Data
driven in their coverage, and too focused on power struggles We collected data on the presentational and representational
rather than substance. In the United States in the 1980s, communication of 61 political leaders from January 2018
Table 1. Leaders’ Data.

Twitter Total Tweets Total Press


Months Followers January 2018– January 2018– GDP Per
Name Country Age in Position (Millions) July 2019 July 2019 Capita (USD) Population Democracy PD IND MAS UA LT IND

Abdel Fattah al-Sisi Egypt 64 272 2.61 2,500 3,774 2,549.1 98,423,598 1 70 25 45 80 7 4
Abdullah II Jordan 57 246 1.52 115 2,420 4,247.8 9,965,318 0 70 30 45 65 16 43
Abdullah bin Zayed Al Nahyan United Arab Emirates 47 162 4.51 3,200 1,808 43,004.9 9,630,959 0 — 90 25 50 80 —
Adel al-Jubeir Saudi Arabia 57 9 2.83 447 3,184 23,219.1 33,702,756 0 95 25 60 80 36 52
Aleksandar Vučić Serbia 49 26 .299 2,571 2,444 7,234 8,802,754 1 86 25 43 92 52 28
Andrés Manuel López Mexico 65 8 5.91 3,242 1,848 9,698.1 126,190,788 1 81 30 69 82 24 97
Obrador
Arif Alvi Pakistan 69 11 2.03 3,198 1,642 1,472.9 212,228,286 1 55 14 50 70 50 0
Benjamin Netanyahu Israel 69 125 1.66 3,202 1,670 41,614 8,381,516 1 13 54 47 81 38 —
Danilo Medina Dominican Republic 67 84 .694 3,238 365 7,650.1 10,627,141 1 65 30 65 45 13 54
Dmitry Medvedev Russia 53 87 4.83 1,591 3,534 11,288.9 145,734,038 1 93 39 36 95 81 20
Donald Trump United States 73 31 63.4 3,194 1,646 62,641 327,096,265 1 40 91 62 46 26 68
Donald Tusk European Union 62 246 1.07 543 1,630 36,531.7 — 1 — — — — — —
Edgars Rinkēvičs Latvia 45 94 .002455 3,248 926 18,088.9 1,928,459 1 44 70 9 63 69 13
Emmanuel Macron France 41 27 4.07 3,248 1,527 41,463.6 64,990,511 1 68 71 43 86 63 48
Ernesto Araújo Brazil 52 7 .366 124 986 8,920.8 209,469,323 1 69 38 49 76 44 59
Evo Morales Bolivia 59 163 .572 3,240 3,040 3,548.6 11,353,142 1 — — — — — —
Giuseppe Conte Italy 55 14 .261 455 2,681 34,318.4 60,627,291 1 50 76 70 75 61 30
Hillary Clinton United States 71 — 25.2 3,204 1,593 62,641 327,096,265 1 40 91 62 46 26 68
Imran Khan Pakistan 66 12 10.2 3,198 1,678 1,472.9 212,228,286 1 55 14 50 70 50 0
Iván Duque Colombia 43 13 .924 3,243 1,399 6,651.3 49,661,048 1 67 13 64 80 13 83
Jair Bolsonaro Brazil 64 7 4.93 3,176 1,870 8,920.8 209,469,323 1 69 38 49 76 44 59
Javad Zarif Iran 59 72 1.22 604 1900 5627.7 81,800,188 0 58 41 43 59 14 40
Joko Widodo Indonesia 58 58 11.9 1,793 3,205 3,893.6 267,670,543 1 78 14 46 48 62 38
Jorge Arreaza Montserrat Venezuela 46 24 1.6 3,189 1,488 16,054.5 28,887,118 1 81 12 73 76 16 100
Juan Orlando Hernández Honduras 50 67 .376 3,179 1,752 2,482.7 9,587,522 1 80 20 40 50 — —
Justin Trudeau Canada 47 45 4.55 3,225 976 46,124.7 37,074,562 1 39 80 52 48 36 68
Khalid bin Ahmed Bahrain 59 167 .525 3,190 1,325 24,050.8 1,569,446 0 — — — — — —
Kono Taro Japan 56 24 .752 3,227 2,323 39,286.7 127,202,192 1 54 46 95 92 88 42
Lee Nak-yeon South Korea 67 26 .128 3,202 1,449 31,362.8 51,171,706 1 60 18 39 85 100 29
Mahathir Mohamad Malaysia 94 15 1.07 3,198 2,016 11,239 31,528,033 0 104 26 50 36 41 57
Mahmood Qureshi Pakistan 63 12 2.62 1,527 1,348 1,472.9 212,228,286 1 55 14 50 70 50 0
Marcelo Ebrard Mexico 59 8 1.62 3,214 1,129 9,698.1 97,338,657 1 81 30 69 82 24 97
Martı́n Vizcarra Peru 56 17 .607 1,642 1,401 6,947.3 31,989,260 1 64 16 42 87 25 46
Mauricio Macri Argentina 60 44 4.91 3,240 3,675 11,652.6 44,361,150 1 49 46 56 86 20 62
Mevlüt Çavuşoğlu Turkey 51 45 1.35 3,228 1,913 9,311.4 82,340,088 1 66 37 45 85 46 49
Miguel Dı́az-Canel Cuba 59 16 .156 776 2,000 8,541.2 11,338,134 0 — — — — — —
Mike Pompeo United States 55 16 .766 942 1,687 62,641 327,096,265 1 40 91 62 46 26 68
Mohammed Al Maktoum United Arab Emirates 70 162 9.77 3,239 1,897 43,004.9 9,630,959 0 — 90 25 50 80 —
Mohammed bin Zayed Al United Arab Emirates 58 177 2.56 3,209 4,137 43,004.9 9,630,959 0 — 90 25 50 80 —
Nahyan
Moon Jae-in South Korea 66 27 1.77 3,074 1,738 31,362.8 51,171,706 1 60 18 39 85 100 29
(continued)

5
6
Table 1. (continued)

Twitter Total Tweets Total Press


Months Followers January 2018– January 2018– GDP Per
Name Country Age in Position (Millions) July 2019 July 2019 Capita (USD) Population Democracy PD IND MAS UA LT IND

Muhammadu Buhari Nigeria 76 51 2.22 3,237 1,953 2,028.2 195,874,683 1 80 30 60 55 13 84


Narendra Modi India 68 63 49.5 3,214 1,900 2,015.6 1,352,642,280 1 77 48 56 40 51 26
Nicolás Maduro Venezuela 56 76 3.65 3,224 1,637 16,054.5 28,887,118 1 81 12 73 76 16 100
Paul Kagame Rwanda 61 232 1.51 2,786 3,320 773 12,301,970 1 — — — — 18 37
Pedro Sánchez Spain 47 14 1.07 3,248 989 30,523.9 46,692,858 1 57 51 42 86 48 44
Petro Poroshenko Ukraine 53 58 1.2 3,248 1,825 3,095.2 44,246,156 1 92 25 27 95 55 15
Pope Francis Vatican 82 77 18.1 2,007 1,767 — 801 0 — — — — — —
Rania Al Abdullah Jordan 48 246 10.5 2,126 822 4,247.8 9,965,318 0 70 30 45 65 16 43
Recep Tayyip Erdoğan Turkey 65 60 — 2,627 1,780 9,311.4 82,340,088 1 66 37 45 85 46 49
Ricardo Rosselló Puerto Rico 40 31 .161 3,211 2,853 31,651.3 3,039,596 1 68 27 56 38 19 99
Saad Hariri Lebanon 49 32 1.54 3,231 2,898 8,269.8 6,859,408 1 75 40 65 50 14 25
Salman bin Abdulaziz Al Saud Saudi Arabia 83 55 7.65 330 3,249 23,219.1 33,702,756 0 95 25 60 80 36 52
Sebastian Kurz Austria 32 17 .347 3,219 2,774 51,512.9 8,891,388 1 11 55 79 70 60 63
Sebastián Piñera Chile 69 17 2.24 3,220 1,778 15,923.4 18,729,160 1 63 23 28 86 31 68
Shinzo Abe Japan 64 79 1.5 1,656 1,743 39,286.7 127,202,192 1 54 46 95 92 88 42
Steffen Seibert Germany 59 108 .922 2,895 1,554 48,195.6 83,124,418 1 35 67 66 65 83 40
Tamim bin Hamad Qatar 39 26 .58 101 3,205 69,026.5 2,781,682 0 93 25 55 80 — —
Teodoro Locsin Philippines 70 10 .65 3,203 649 3,102.7 106,651,394 1 94 32 64 44 27 42
Theresa May United Kingdom 62 36 .892 1,614 1,614 42,491.4 67,141,684 1 35 89 66 35 51 69
Uhuru Kenyatta Kenya 57 76 .00554 16 1,843 1,710.5 51,392,565 1 70 25 60 50 -
Vladimir Putin Russia 66 87 .73 3,208 1,512 11,288.9 145,734,038 1 93 39 36 95 81 20
Notes: PD ¼ Hofstede power distance; IDV ¼ Hofstede individualism; MAS ¼ Hofstede masculinity; UA ¼ Hofstede uncertainty avoidance; LT ¼ Hofstede long-term orientation; IND ¼ Hofstede indulgence.
Peres et al. 7

through July 2019. We constructed the list of leaders from entire data set needs to be translated. Second, while the press
listings in the popular press and Twitter reports (e.g., Twiplo- articles are usually edited before publication and use standard
macy; https://twiplomacy.com/ranking/50-effective-world- language, Twitter is characterized by a low degree of regula-
leaders/#). As illustrated in Table 1, the list contains leaders tion, and tweets use spoken language (Crystal 2006). Process-
from Europe, Asia, North, Central and South America, Africa, ing Twitter data requires coping with contracted auxiliaries
and the Middle East, all of whom either held an active political (e.g., “I’ll,” “won’t,” “isn’t,” “ain’t,” “gotta,” “wanna”), slang
position or were politically active (e.g., Hillary Clinton) during expressions (e.g., “lol”), exclamations (e.g., “oh,” “yay,”
the data collection period. For each political leader, we col- “aww”), terms coined explicitly by Twitter (e.g., “retweet” or
lected two types of communication data: “RT”), use of the “@” and “#” symbols to mention other users
and topics, and errors of spelling, capitalization, and
1. Social media: Twitter is a commonly used social media punctuation.
communication outlet for political leaders (Golbeck To prepare the data, all tweets were decoded to natural
et al. 2010) and has been recently used in research on language (some of them were coded as bytes) and were trans-
political leaders (Bovet and Makse 2019; Bovet, Mor- lated to English using Google translation libraries. All linguis-
one, and Makse 2018). We used Twitter API to down- tic characteristics of a tweet (i.e., @ usernames, hashtags,
load the tweets from each political leader’s public retweet abbreviations) were retained. LexisNexis preproces-
account. We included everything posted under the lead- sing included cleaning all the metadata (e.g., length, load date,
er’s Twitter handle, including original tweets as well as language, publication type, journal code, section, byline). Non-
retweets, with the exception of Angela Merkel, for English publications were translated to English as part of the
whom we used her spokesman’s Twitter account data corpus. The entire text from both sources was converted to
because her own Twitter account was not sufficiently lowercase, and extra blank spaces were removed.
active. Altogether, we collected a total of 150,696 Twit-
ter posts. Note that although the political leaders often
employ people who do some of the actual tweeting on Content Analysis
their behalf, they mostly are personally involved in their
We conducted two types of content analysis on the communi-
Twitter account, and the leaders’ tweets are regarded as
cation data:
the closest approximation to direct communication with
their audience (Draper 2018). 1. Sentiment analysis: We applied sentiment analysis to
2. Press articles: We mined the LexisNexis database measure the level of positivity or negativity of the
(using its API) to collect the entire press coverage of tweets and the news articles (Pang and Lee 2004; Zhao,
the leaders on our list, resulting in a total of 120,689 Qin, and Liu 2010). We used two sentiment analysis
press articles. The data set contains most of the world’s tools. The first is the polarity measure of the TextBlob
press publications, including online and offline daily Python library, which returns a sentiment scaled from
newspapers and magazine articles as well as press 1 to 1. The second sentiment score (VADER of
releases and newswire articles. Python) returns four different scores: positive (0 to 1),
neutral (0 to 1), negative (0 to 1), and compound (1 to
In addition, because we wanted to examine the communi-
1). The first three are scores of positivity, neutrality, and
cation style with respect to country and leader characteristics,
negativity, respectively, and the compound score is a
we collected, for each leader, a list of personal characteristics:
weighting of all three. We used the weighted score
age, role, and tenure in position. For each country, we collected
(compound) as an alternate measure for the sentiment
data on the country’s GDP, social inequality (Gini index), and
analysis. Both measures (hereafter termed “polarity”
its political regime. We also used, for each country, the latest
and “sentiment”) intend to capture the same construct
measures of Hofstede’s cultural dimensions (Hofstede 2011),
and are highly correlated; we use them both for robust-
namely, uncertainty avoidance, masculinity versus femininity,
ness purposes.
long-term orientation versus short-term orientation, individu-
For the Twitter data, we used the entire text of the
alism versus collectivism, indulgence versus restraint, and
post for the sentiment analysis. The news articles are
power distance.
longer and in many cases only partially relate to the
political leader. Therefore, the overall sentiment of the
article might not represent the sentiment attached to the
Preprocessing of Communication Data political leader. To overcome this problem, we con-
Conducting content analysis on Twitter and on press articles is ducted the sentiment analysis of news articles only on
challenging, with each data source bearing its own challenges: paragraphs that include the name of political leader.
First, the Twitter accounts of most global leaders from non- 2. Topic analysis: We used latent Dirichlet allocation
English-speaking countries are not in English, and press arti- (LDA) to elicit the topics of the Twitter and news posts.
cles also come in many languages. Content analysis algorithms, This method is regarded as an effective way to identify
however, are mostly trained on English dictionaries; hence, the topics (i.e., subjects discussed in one or more
8

.00
.05
.10
.15
.20
.25
.30

−.05
−.40
−.20
.00
.20
.40
.60
.80

Abdel Fattah al-Sisi Abdel Fattah al-Sisi


Abdullah bin Zayed Al Nahyan Abdullah bin Zayed Al Nahyan
Adel Aljubeir Adel Aljubeir
Aleksandar Vucic Aleksandar Vucic
Andrés Manuel López Obrador Andrés Manuel López Obrador
Steffen Seibert Steffen Seibert
Arif Alvi Arif Alvi
Benjamin Netanyahu Benjamin Netanyahu
Danilo Medina Danilo Medina
Dmitry Medvedev Dmitry Medvedev
Donald Trump Donald Trump

Twitter polarity
Twitter sentiment

Donald Tusk Donald Tusk


Mahathir Mohamad Mahathir Mohamad
Edgars Rinkevics Edgars Rinkevics
Emir Tamim bin Hamad Emir Tamim bin Hamad
Emmanuel Macron Emmanuel Macron
Ernesto Araújo Ernesto Araújo
Evo Morales Evo Morales
Giuseppe Conte Giuseppe Conte
Hillary Clinton Hillary Clinton
Imran Khan Imran Khan
Iván Duque Iván Duque
Jair Bolsonaro Jair Bolsonaro
Javad Zarif Javad Zarif
Joko Widodo Joko Widodo

Press articles polarity


Jorge Arreaza Montserrat Jorge Arreaza Montserrat
Press articles sentiment

Juan Orlando Hernández Juan Orlando Hernández

generated content (e.g., Büschken and Allenby 2016;


applied in marketing research to analyze user-
collections (Krippendorff 2004). It has been recently
documents) and topic categories (i.e., groups of topics
belonging to a common subject area) in large document
Justin Trudeau Justin Trudeau
Khalid bin Ahmed Khalid bin Ahmed
King Abdullah II King Abdullah II
King Salman bin Abdulaziz Al Saud King Salman bin Abdulaziz Al Saud

B: Polarity Per Leader: Twitter Versus Press Articles


Kono Taro Kono Taro
Lee Nak-yeon Lee Nak-yeon
A: Sentiment Per Leader: Twitter Versus Press Articles

Marcelo Ebrard Marcelo Ebrard


Martín Vizcarra Martín Vizcarra
Mauricio Macri Mauricio Macri
Mevlüt Çavuşoğlu Mevlüt Çavuşoğlu
Miguel Díaz-Canel Miguel Díaz-Canel
Mike Pompeo Mike Pompeo
Mohammed bin Zayed Al Nahyan Mohammed bin Zayed Al Nahyan
Moon Jae-in Moon Jae-in
Muhammadu Buhari Muhammadu Buhari
Narendra Modi Narendra Modi
Nicolás Maduro Nicolás Maduro
Paul Kagame Paul Kagame
Pedro Sánchez Pedro Sánchez
Petro Poroshenko Petro Poroshenko
Pope Francis Pope Francis

Figure 2. Compound sentiment (Panel A) and polarity (Panel B) per leader for Twitter and press articles.
Rania Al Abdullah Rania Al Abdullah
Recep Tayyip Erdoğan Recep Tayyip Erdoğan
Ricardo Rosselló Ricardo Rosselló
Saad Hariri Saad Hariri
Sebastian Kurz Sebastian Kurz
Sebastián Piñera Sebastián Piñera
Shah Mahmood Qureshi Shah Mahmood Qureshi
Sheikh Mohammed Al Maktoum Sheikh Mohammed Al Maktoum
Shinzo Abe Shinzo Abe
Teodoro Locsin Teodoro Locsin
Theresa May Theresa May
Uhuru Kenyatta Uhuru Kenyatta
Vladimir Putin Vladimir Putin
Journal of International Marketing XX(X)

a single document. In a standard preprocessing stage for


Tirunillai and Tellis 2014). We trained the LDA model

the LDA, we removed exclamation marks, emoticons,


on the Twitter data, where each tweet was considered as

common slang words, URLs, hashtags, and Twitter-


Peres et al. 9

specific notation (i.e., “b’,” “RT,” “b’RT”), although Table 2. Regression of the Average Sentiment and Polarity of a
we used them for the sentiment analysis. We also con- Leader on Country Characteristics for Twitter and Press Articles.
ducted stemming (using NLTK Porter Stemmer) to exo- Press Press
genously force words with similar meanings, such as Twitter Twitter Articles Articles
“great,” “greatly,” “greatest,” and “greater,” to be Sentiment Polarity Sentiment Polarity
treated as a bundle. We filtered out tokens that appear
in less than 15 documents or in more than 50% of the logGDP .045 .014 .014 .006
Continent
documents (a token is a word or a short word combina- Asia .116 .073 .073 .002
tion). From the tokens left, we kept the 100,000 most Europe .035 .165 .165 .019
common speech tokens. The algorithm yielded its best Central America .144 .059 .059 .016
results for 15 topics (we manually looked through South America .118 .001 .001 .019
results of different numbers of subjects to finalize the North America .098 .122 .122 .010
number), and the different topics were manually named. Democracy .000 .001 .001 .000
Then the chosen LDA model was applied on the Twitter Hofstede Dimensions
Power distance .066 .039 .039 .018
data and the LexisNexis data, with each document
Individualism .002 .001 .001 .000
being all the tweets or all the reports (respectively) Masculinity .000 .001 .001 .000
connected to each single leader. In this way, we Uncertainty avoidance .002 .003 .003 .000
obtained the distribution of topics discussed in by each Long-term orientation .000 .003 .003 .000
political leader on Twitter, and in connection with the Indulgence .001 .000 .000 .000
leader in the press.

freedom of press, and social inequality. None of these variables


Results were found to be significant.
The results show a clear universality across leaders and also
Universality and Media Compatibility in Sentiment similarity across communication types. The major political
Figure 2 illustrates the values of the sentiment/polarity mea- leaders in our data set are similar in the sentiment of their
sures per leader for Twitter and for press articles. In general, communication, and their sentiment, measured in terms of both
the sentiment/polarity of leaders is neutral to positive, with the polarity and compound sentiment, does not show a clear depen-
Pakistani leader Shah Mahmood Qureshi being the most posi- dence on the country characteristics. This universality is also
tive (.63, .231, .667, and .127 for Twitter sentiment, Twitter expressed in the similarity of sentiment of Twitter and press
polarity, press sentiment, and press polarity, respectively), and articles. The traditional notion of self-presentation as portray-
politicians such as the British leader Theresa May and Abdul- ing the individual in a positive manner and representation in
lah bin Zayed Al Nahyan from the United Arab Emirates being journalism as being more objective and therefore less posi-
tively biased has faint support in our data. Sentiment is gener-
at the bottom for positivity (note the relatively negative press
ally neutral to positive, and Twitter is slightly more positive in
coverage of the latter). The average sentiment (using the com-
polarity, but this difference is not robust across the other senti-
pound sentiment measure described previously) is .279 (SD ¼
ment measure we used.
.134) for Twitter and .272 (SD ¼ .196) for the press articles,
meaning that they are not significantly different from each
other (p ¼ .62). The average polarity of the leaders in our data
set is .13 (SD ¼ .054) for Twitter and .064 (SD ¼ .03) for the Universality and Media Compatibility in Communication
press articles. These numbers indicate that Twitter is signifi- Topics
cantly more positive than the press articles, but the difference is Tables 3 through 5 present the results of the content analysis.
small and is not robust across different sentiment measurement The LDA procedure identified 15 topics that span the leaders’
methods. communication: (1) the Arab world, (2) condolences, (3) con-
We further tested to what extent the sentiment of the polit- gratulations, (4) corruption, (5) diplomacy, (6) economy,
ical leaders’ communication depends on their country and per- (7) education, (8) elections, (9) fighting terrorism, (10) global
sonal characteristics. Table 2 describes the results of four relationships, (11) humanitarian crisis, (12) immigration,
regression models, where the dependent variables are the senti- (13) national holiday, (14) official thank-you messages, and
ment and polarity for Twitter and press articles and the depen- (15) parliamentary issues. Table 3 describes, for each topic,
dent variables are the country characteristics. None of the the 15 words with the highest likelihood. Note that LDA clas-
variables were found to be significant. That is, the sentiment sifies all the words on all the topics, providing for each word
and the polarity of communication do not depend on the lead- the likelihood that it belongs to the topic. Table 4 presents the
ers’ individual country characteristics. We repeated this anal- average weights of the topic mixture over all the leaders, for
ysis with the leaders’ personal characteristics, as well as with Twitter and for the press articles, and compares them using a
other country characteristics, such as level of happiness, series of t-tests. The table indicates that diplomacy (i.e., words
10
Table 3. LDA Results: The Top 15 Words with the Highest Likelihood for the 15 Topics.

The Arab world god bin moham zai bless king merci jordan martyr al camp arab putin holi ramadan
.0362 .0312 .0303 .0259 .0183 .0111 .0110 .0104 .0103 .0101 .0100 .0098 .0098 .0097 .0093
Condolences famili peopl victim amp attack todai death live friend condol condemn lose affect year thought
.0395 .0171 .0155 .0135 .0132 .0128 .0105 .0104 .0102 .0096 .0092 .0092 .0091 .0078 .0072
Congratulations congratul great happi peopl world love wish dai countri women nation best proud peac year
.0262 .0192 .0189 .0186 .0177 .0167 .0142 .0140 .0127 .0115 .0112 .0107 .0106 .0106 .0101
Corruption govern ukrain work public amp forc todai reform power nation implement need peopl tax corrupt
.0287 .0173 .0138 .0131 .0098 .0089 .0087 .0077 .0075 .0074 .0069 .0069 .0069 .0068 .0066
Diplomacy meet minist presid prime foreign discuss confer council austria issu visit relat receiv cooper todai
.0577 .0544 .0427 .0286 .0230 .0187 .0140 .0136 .0132 .0130 .0124 .0120 .0115 .0105 .0103
Economy develop invest project job educ work creat improv new year million economi plan countri opportun
.0184 .0158 .0148 .0137 .0118 .0117 .0115 .0112 .0109 .0107 .0106 .0101 .0096 .0095 .0094
Education school gt lt read new vote student content open door photo amp talent year pti
.0196 .0192 .0160 .0144 .0142 .0135 .0127 .0122 .0115 .0080 .0076 .0074 .0070 .0067 .0065
Elections elect live parti vote watch candid todai peopl nation come like congress campaign speech time
.0328 .0231 .0176 .0159 .0159 .0143 .0130 .0108 .0091 .0088 .0083 .0082 .0081 .0079 .0078
Fighting terrorism fight terror know iran want peopl act pai regim war time that attack terrorist world
.0128 .0113 .0105 .0099 .0098 .0086 .0085 .0083 .0079 .0079 .0072 .0068 .0066 .0061 .0061
Global relationships good chile eu china countri meet amp europ todai world cooper presid south trade korea
.0255 .0201 .0198 .0185 .0165 .0158 .0156 .0148 .0137 .0134 .0134 .0127 .0116 .0113 .0112
Humanitarian crisis right lebanon venezuela secur protect state peopl law countri human peac support defend israel american
.0314 .0197 .0164 .0143 .0143 .0143 .0139 .0136 .0125 .0118 .0108 .0093 .0091 .0089 .0087
Immigration hariri brexit new center todai meet product migrat technolog industri innov open nation inaugur presid
.0423 .0195 .0168 .0133 .0123 .0117 .0110 .0106 .0104 .0101 .0088 .0088 .0081 .0080 .0078
National holiday year chang dai todai th anniversari independ commit democraci peopl ago histori movement nation freedom
.0234 .0227 .0199 .0177 .0136 .0135 .0119 .0113 .0106 .0095 .0085 .0084 .0072 .0070 .0066
Official thank-you thank work support forward look todai peopl problem countri continu nato welcom commit year dai
.1086 .0392 .0142 .0131 .0123 .0114 .0110 .0096 .0094 .0093 .0086 .0083 .0079 .0078 .0076
Parliamentary amp pm interview conserv turkei senat eu presid trump uk pension negoti time moon sai
.0474 .0302 .0140 .0128 .0123 .0118 .0116 .0115 .0113 .0112 .0111 .0104 .0102 .0101 .0098

Notes: All words were converted into lowercase letters. Words might be truncated because of the stemming process (e.g., “migrat” stands for “migration,” “immigrants,” and other words). The numbers are the
likelihood that the word belongs to the topic. *p < .05. In this case, no variable was found significant.
Peres et al. 11

Table 4. The Average Topic Distribution over Leaders, for Twitter characteristics of the leaders and country characteristics, such
and Press Articles. as level of happiness, freedom of press, and social inequality,
p-Value Twitter Press Articles
which did not lead to any results of interest.
Topic of t-Test Average (LexisNexis) Average What additional insights can we gain from the differences
between the topics discussed on Twitter and in the press
The Arab world .7951 .0736 .0790
Condolences .0000* .0624 .0372
articles? As discussed, Table 4 indicates that although many
Congratulations .0000* .1031 .0262 of the topics are discussed on average equally in both media,
Corruption .0050* .0768 .0981 some topics (e.g., congratulations, condolences, thank-you
Diplomacy .0004* .1005 .1545 messages) are more typical on Twitter, whereas topics such
Economy .0345* .0816 .0615 as diplomacy and parliamentary issues are discussed more in
Education .0686 .0491 .0595
Elections .0125* .0617 .0792
press articles. We wanted to analyze the differences between
Fighting terrorism .2328 .0698 .0625 the outlets at the individual leader’s level (in addition to the
Global relationships .5576 .0681 .0629 aggregate comparison presented previously) and ask to what
Humanitarian crisis .8009 .0719 .0740 extent a leader’s communication on Twitter is similar to or
Immigration .4044 .0465 .0503 different from the leader’s press coverage. To do so, we calcu-
National holiday .3443 .0596 .0555
Official thank-you .0000* .0528 .0320
lated the cosine similarity of the two topic mixture distribution
Parliamentary .0021* .0483 .0752 vectors for each leader. Cosine similarity treats the two topic
mixtures of a leader’s media as two vectors in a 15-dimensional
Notes: Each t-test compares the distance between the two average values, space and calculates their scalar (dot) product. Both the press
over all the leaders (N ¼ 61). *p < .05. articles and Twitter topic distribution vectors of each leader
were normalized so that the sum of squares is 1 before the dot
related to meetings, conferences, discussions, etc.) is the most product of the two vectors was calculated, so the similarity is
popular topic for both communication outlets. Press articles between 0 (lowest possible similarity; vectors are orthogonal)
focus significantly more than Twitter on corruption, diplo- and 1 (highest possible similarity). The cosine similarity values
macy, elections, and parliamentary issues, whereas Twitter, of the leaders on our list range from .14 (Emir Tamim bin
relative to the press articles, deals more with personal mes- Hamad) to .29 (Pope Francis), with an average of .22 (SD ¼
sages, such as condolences, congratulations, and thank-you .03). Although these numbers may seem low, they represent
messages, as well as economy. Many of the topics, such as the relatively high similarity, given the different structure and typ-
Arab world, education, fighting terrorism, and global relation- ical grammatical nature of the texts in these outlets. Hence, for
ships, receive similar emphasis both on Twitter and in the political leaders, the differences between Twitter and press
press. articles are no different than the typical difference between the
To what extent do communication topics depend on the contents of these two channels, given their fundamentally dif-
ferent characteristics. In other words, we see no evidence that
country characteristics? We tested, similarly to the sentiment
differences in topic mixture between outlets are a result of an
analysis, how the topic mixture depends on the country char-
outlet-specialized communication strategy.
acteristics. For each topic, we ran a regression model over the
We further checked whether the level of similarity depends
61 leaders, in which the dependent variable is the weight of the
on the country characteristics. We ran a regression model in
topic in the leader’s communication (for either Twitter or the
which the dependent variable was the cosine similarity score
press articles) and the independent variables are the individual
and the independent variables were the country characteristics
and country characteristics. The results are presented in Table 5
used in the previous regressions. The results, described in
(Panel A for Twitter and Panel B for press articles). The results
Table 6, indicate that the higher the power distance and indi-
indicate that overall, the topic mixture does not seem to vidualism are, the higher is the topic similarity between Twitter
strongly depend on the individual and country characteristics. and press articles. Although the coefficients are significant, it is
On Twitter, immigration is a prevalent topic for democracies hard to interpret them as representing a theoretically based,
and for countries with high GDP. Interestingly, immigration fundamental mechanism.
has more presence in the tweets of African leaders relative to Similarly to the sentiment analysis, the content analysis
leaders from other continents. Immigration is also discussed in implies universality across leaders in communication topics.
countries with lower uncertainty avoidance, shorter-term orien- These results, grounded in the theoretical perspective of
tation, and lower indulgence. The Arab world is discussed less the “mediatized communication as a meta process of
in democracies (probably because it is discussed by Arab lead- globalization” (Krotz 2007), imply that despite the distinct
ers, who are mostly not democratic rulers). Being a democracy differences between cultures, leaders generate a communica-
is positively correlated with a country’s leader’s tweets on tion mix that is weakly dependent on the country characteris-
corruption and education, and less with tweets on diplomacy. tics. In branding terminology, although political leaders are
For press articles, results are similar (although fewer variables brands that are strongly identified with their country, their
are significant). In addition to these variables, we tried a large communication style does not reflect that. This is the case both
set of models with different variables, including individual for their own generated content on Twitter and for the press
12
Table 5. Regressions of the Topic Mixture of a Leader on Country Characteristics for Twitter and Press Articles.

Continent Hofstede Dimensions

Central South North Power Uncertainty Long-Term


Independent Variable logGDP Asia Europe America America America Democracy Distance Individualism Masculinity Avoidance Orientation Indulgence
A: Regression of Twitter Topic Mixture of a Leader on Country Characteristics
The Arab world .0160 .0000 .0199 .0390 .0718 .0135 .1441* .0004 .0007 .0006 .0014 .0013 .0005
Condolences .0026 .0824* .0445 .0230 .0266 .0536 .0507* .0004 .0000 .0002 .0004 .0004 .0003
Congratulations .0070 .1206* .0822 .1010 .0288 .0378 .0255 .0001 .0007 .0010 .0013 .0012 .0011
Corruption .0019 .0279 .0143 .0693 .0205 .0050 .0669* .0004 .0001 .0002 .0001 .0002 .0001
Diplomacy .0265 .1458 .0276 .1133 .0413 .1256 .1121* .0013 .0004 .0022* .0014 .0008 .0016
Economy .0066 .0056 .0260 .1014 .0011 .0102 .0518 .0000 .0003 .0012 .0006 .0002 .0007
Education .0044 .0358 .0238 .0038 .0225 .0179 .0502* .0003 .0005 .0004 .0002 .0000 .0001
Elections .0028 .0259 .0661 .0159 .0318 .0071 .0072 .0004 .0004 .0007 .0000 .001* .0000
Fighting terrorism .0023 .0564 .0520 .0051 .0209 .0566 .0140 .0003 .0002 .0003 .0006 .0007 .0002
Global relationships .0044 .0107 .0176 .0367 .0244 .0346 .0091 .0002 .0008 .0000 .0004 .0007* .0002
Humanitarian crisis .0083 .0039 .0500 .0523 .0460 .0299 .0087 .0003 .0004 .0002 .0003 .0008 .0001
Immigration .0264* .0886* .1026* .1124* .0747* .1355* .0401* .0000 .0005 .0002 .0007* .0006* .0009*
National holiday .0015 .0103 .0057 .0203 .0375* .0010 .0301* .0002 .0003 .0002 .0000 .0002 .0003
Official thank-you .0029 .0096 .0128 .0223 .0147 .0107 .0167 .0002 .0004 .0001 .0003 .0001 .0002
Parliamentary .0091 .0642 .0451 .0215 .0323 .0869 .0406 .0003 .0006 .0001 .0004 .0006 .0007
B: Regression of Press Articles (LexisNexis) Topic Mixture of a Leader on Country Characteristics.
The Arab world .0131 .0103 .0189 .0075 .0406 .0221 .1222* .0006 .0003 .0003 .0010 .0009 .0006
Condolences .0010 .0405 .0161 .0057 .0124 .0078 .0303* .0002 .0001 .0006 .0002 .0001 .0009
Congratulations .0015 .0303 .0116 .0014 .0054 .0104 .0132 .0002 .0000 .0001 .0001 .0002 .0001
Corruption .0009 .0153 .0084 .0583 .0382 .0049 .0265 .0011* .0003 .0001 .0002 .0001 .0004
Diplomacy .0276 .1907 .1671 .0108 .0003 .0604 .0609 .0006 .0011 .0001 .0015 .0009 .0018
Economy .0053 .0282 .0164 .2064* .0268 .0147 .0178 .0004 .0002 .0006 .0000 .0001 .0009
Education .0117 .0132 .0080 .0142 .0067 .0242 .0444* .0000 .0005 .0001 .0001 .0007 .0008
Elections .0182 .0386 .0321 .1064* .0305 .0594 .0584* .0001 .0007 .0003 .0008 .0000 .0000
Fighting terrorism .0007 .0260 .0176 .0027 .0065 .0002 .0056 .0000 .0002 .0005 .0006 .0001 .0006
Global relationships .0079 .0288 .0281 .0084 .0164 .0476 .0229 .0005 .0004 .0003 .0003 .0013* .0000
Humanitarian crisis .0126 .0015 .0141 .0620 .0326 .0255 .0201 .0002 .0007 .0002 .0003 .0006 .0000
Immigration .0175* .0552 .0540 .0339 .0423 .0966* .0244 .0001 .0001 .0001 .0003 .0007* .0007
National holiday .0058 .0200 .0071 .0100 .0332 .0234 .0104 .0003 .0002 .0000 .0004 .0002 .0001
Official thank-you .0064* .0103 .0021 .0023 .0063 .0186 .0062 .0001 .0000 .0000 .0003 .0001 .0002
Parliamentary .0120 .0244 .0021 .0134 .1260 .0589 .0004 .0008 .0001 .0001 .0006 .0008

*p < .05.
Peres et al. 13

Table 6. Average Topic Distribution over Leaders, for Twitter and describe what the leader did and therefore are less indicative
Press Articles. of the leader’s communication style.
Coef. p-Value
Pronoun usage is related to sentiment. Table 7 presents the
results of two regression models, where the independent variable
GDP 4.86E-07 .882 is the polarity for Twitter and press articles and the dependent
Continent
Asia .1955 .133 variables are the percentage of each pronoun type. We can see
Europe .1511 .271 that for both communication outlets, more positive sentiment is
Central America .1156 .406 associated with higher usage of “we” messages and lower usage
South America .1664 .203 of “they” messages. We obtain similar results for the sentiment
North America .0389 .835 measure. Note that both sentiment algorithms we use do not
Democracy .0434 .625
Hofstede Dimensions consider pronoun usage as input for determining the sentiment,
Power distance .0035* .034 so these two dimensions are not forced to be correlated.
Individualism .0033* .048 We also tested the percentage of the four types of pronouns
Masculinity .0011 .533 versus the country characteristics. The results are presented in
Uncertainty avoidance .0016 .311 Table 8. Similar to the results on sentiment and topic, the use of
Long-term orientation .0002 .922
Indulgence .0017 .348 “I,” “we,” “they,” or “you” messages seems to be independent
of country characteristics. The only significance we found is
*p < .05. that when tweeting, leaders from Asia, Europe, and Central
America use “we” messages more often than leaders from
Africa do. Also, leaders from countries with shorter-term orien-
articles, which are supposed to represent the spectrum of inter- tation were found to use first-person pronouns (“I” and “we”)
ests of the local readers. on Twitter more often than leaders from countries with longer-
The similarity exists not only among leaders, but also term orientation. The rightmost two columns of Table 8
between Twitter, which is a presentational social media chan- describe regression models in which the dependent variables
nel, and press articles, which are a form of representational are the differences between the number of “I” or “we” mes-
media. The differences in sentiment between Twitter and press sages and the number of “they” or “you” messages. This anal-
articles are small. Although the two outlets differ in some ysis indicates the use of first-person relative to second- and
topics (Table 3), their topic mixtures have many similarities, third-person pronouns (the left columns relate to the overall
and the differences seem to result from the inherent nature of proportion of each type, irrespective of the others). In line with
these outlets, not from channel-specific communication strate- the finding described earlier, relative to African leaders, leaders
gies. In this sense, our results are in line with previous findings from Asia, Europe, and Central America use first-person mes-
showing that journalists rely on social media as a source of sages more than second- and third-person messages when
information while constructing their news articles (Shapiro and tweeting. Also, leaders from countries with shorter-term orien-
Hemphill 2017). tation were found to use first-person pronouns more often than
second- and third-person pronouns on Twitter.
Universality and Media Compatibility in “I,” “We,”
“You,” and “They” Messages Mapping on the Universality–Compatibility Matrix
As an additional analysis, we tested another linguistic charac- We combined the two dimensions to classify the leaders in our
teristic of the leaders’ communications: some language models sample on the universality–compatibility matrix. The matrix is
(Packard, Moore, and McFerran 2018) emphasized the use of conceptual, and therefore, we had several options to operationa-
“I” or “we” versus “you” and “they” as a possible measure of lize it. We chose the horizontal axis to represent the extent to
the amount of self-focus versus orientation toward others in which the topic distribution of the focal leader is similar to the
one’s communication. average topic distribution across leaders. Our measure of simi-
We used the content analysis algorithm to count the pro- larity is the cosine similarity (of the normalized vectors) between
portion of first-person, second-person, and third-person words the topic distribution vector of the focal leader’s tweets and the
for Twitter and for the press articles. Words with similar average distribution vector of all leaders on Twitter. Recall that
meanings were grouped together (e.g., the words “I,” “me,” cosine similarity ranges from 0 to 1, with 1 indicating identical
and “myself” were all counted as “I”). Figure 3 presents the vectors. The vertical axis, representing media compatibility,
percentage of these pronouns. For both outlets, “we” mes- indicates the difference in average sentiment between the lead-
sages are predominant, followed by “I” messages. While er’s Twitter account and the leader’s press articles. Since in our
“they” messages are more prevalent than “you” messages in sample the Twitter sentiment was always higher than the senti-
press articles, the opposite is true for Twitter. This finding is ment of the news articles, low compatibility means more nega-
understandable because of the more descriptive nature of tive press coverage, and high compatibility means that the press
press items. Note that we did not count the use of “he” or articles are similar in sentiment to the Twitter account. Note that
“she,” as these words are commonly used in press articles to we could have chosen this axis to be the topic similarity or a
14

1.0
1.2

.0
.2
.4
.6
.8
.01
.02
.03
.04
.05
.06
.07

0
Abdel Fattah al-Sisi
Abdullah bin Zayed Al Nahyan Abdel Fattah al-Sisi
Adel Aljubeir Abdullah bin Zayed Al Nahyan
Aleksandar Vucic Adel Aljubeir

I
Aleksandar Vucic
I

Andrés Manuel López Obrador


Steffen Seibert Andrés Manuel López Obrador
Steffen Seibert

priate for testing press bias.


Arif Alvi Arif Alvi
Benjamin Netanyahu Benjamin Netanyahu
Danilo Medina

We
We

Dmitry Medvedev Danilo Medina


Donald Trump Dmitry Medvedev
Donald Tusk Donald Trump
Mahathir Mohamad Donald Tusk
Edgars Rinkevics Mahathir Mohamad
A: Twitter Pronoun Usage

They
Edgars Rinkevics
They

Emir Tamim bin Hamad Emir Tamim bin Hamad


Emmanuel Macron Emmanuel Macron
Ernesto Araújo Ernesto Araújo
Evo Morales Evo Morales
Giuseppe Conte

B: Press Articles Pronoun Usage

You
You

Hillary Clinton Giuseppe Conte


Hillary Clinton
Imran Khan Imran Khan
Iván Duque Iván Duque
Jair Messias Bolsonaro Jair Messias Bolsonaro
Javad Zarif Javad Zarif
Joko Widodo Joko Widodo
Jorge Arreaza Montserrat Jorge Arreaza Montserrat
Juan Orlando Hernández Juan Orlando Hernández

Global Icons (26 leaders), who have communication topics


classification in Figure 1, we see that the largest group is the
middle point of the value range for each axis. Following the
the intersection between the four quadrants arbitrarily as the
both press articles and Twitter data in our sample. We defined
Figure 4 presents the mapping of the 54 leaders who had
combined similarity, but sentiment seems to be the most appro-
Justin Trudeau Justin Trudeau
Khalid bin Ahmed Khalid bin Ahmed
King Abdullah II King Abdullah II
Salman bin Abdulaziz Al Saud Salman bin Abdulaziz Al Saud
Kono Taro Kono Taro
Lee Nak-yeon Lee Nak-yeon
Marcelo Ebrard Marcelo Ebrard
Martín Vizcarra Martín Vizcarra
Mauricio Macri Mauricio Macri
Mevlüt Çavuşoğlu Mevlüt Çavuşoğlu
Miguel Díaz-Canel Miguel Díaz-Canel
Mike Pompeo Mike Pompeo
Mohammed bin Zayed Al Nahyan Mohammed bin Zayed Al Nahyan
Moon Jae-in Moon Jae-in
Muhammadu Buhari Muhammadu Buhari
Narendra Modi Narendra Modi

Figure 3. Pronoun usage frequency per leader for Twitter (Panel A) and press articles (Panel B).
Nicolás Maduro Nicolás Maduro
Paul Kagame Paul Kagame
Pedro Sánchez Pedro Sánchez
Petro Poroshenko Petro Poroshenko
Pope Francis Pope Francis
Rania Al Abdullah Rania Al Abdullah
Recep Tayyip Erdoğan Recep Tayyip Erdoğan
Ricardo Rosselló Ricardo Rosselló
Saad Hariri Saad Hariri
Sebastian Kurz Sebastian Kurz
Sebastián Piñera Sebastián Piñera
Shah Mahmood Qureshi Shah Mahmood Qureshi
Sheikh Mohammed Al Maktoum Sheikh Mohammed Al Maktoum
Shinzo Abe Shinzo Abe
Teodoro Locsin Teodoro Locsin
Theresa May Theresa May
Uhuru Kenyatta Uhuru Kenyatta
Vladimir Putin Vladimir Putin

similar to the average and have high media compatibility


Journal of International Marketing XX(X)

includes Pope Francis, King Salman of Saudi Arabia, and


ibility but a more peculiar topic distribution. This group
Narrowband Influencers (13 leaders), with high media compat-

Mohammed bin Zayed from the United Arab Emirates. The


nuel Macron from France. The second largest group is the
Medvedev, Donald Trump from the United States, and Emma-
(i.e., low differences in sentiment), such as the Russian Dmitry
Peres et al. 15

Table 7. Average Leader Polarity for Twitter and Press Articles Cosmopolitan Antagonists (nine leaders), with generic topic
Versus Usage of Pronouns. distribution and lower media compatibility, included Benjamin
Twitter Polarity Press Articles Polarity Netanyahu from Israel and the Canadian leader Justin Trudeau.
Finally, the Peculiar Notorious are the least frequent in our
Intercept .0748* .0715*
sample, with only six leaders, most of them near the borders
% of “I” 1.6114 9.2876
% of “We” 1.759* 4.0929* of the quadrant. Their salient member is King Abdullah II from
% of “They” 9.8802* 7.5499* Jordan. To examine whether the four quadrants represent any
% of “You” 2.7287 18.9278 profiling with respect to country characteristics, we ran a set of
ANOVA tests on all the country characteristics used in the
*p < .05.

Table 8. Average Leader Usage of Pronouns Versus Country Characteristics.


“I,” “We,” “They,” “You,” (“I” þ “We”)  (“I” þ “We”) 
“I,” “We,” “They,” “You,” Press Press Press Press (“They” þ “You”), (“They” þ “You”),
Twitter Twitter Twitter Twitter Articles Articles Articles Articles Twitter Press Articles
logGDP .00005 .00033 .00006 .00040 .00029 .00032 .00055 .00029 .00006 .00022
Continent
Asia .01186 .0241* .00261 .00369 .00181 .00122 .00072 .00051 .0002* .00001
Europe .00750 .0268* .00199 .00008 .00036 .00058 .00052 .00062 .0001* .00007
Central America .01525 .0276* .00093 .00452 .00002 .00214 .00007 .00000 .0004* .00002
South America .00494 .01031 .00153 .00037 .00063 .00251 .00107 .00024 .00012 .00002
North America .00533 .01261 .00176 .00309 .00044 .00035 .00046 .00014 .00019 .00006
Democracy .00586 .0141* .00157 .00081 .00098 .00019 .00047 .00038 .0002* .00004
Hofstede Dimensions
Power distance .00002 .00016 .00000 .00003 .00002 .00002 .00000 .00000 .01759 .00032
Individualism .00006 .00008 .00002 .00007 .00000 .00001 .00003 .00001 .02966 .00180
Masculinity .00003 .00015 .00001 .00004 .00003 .00003 .00001 .00001 .03742 .00205
Uncertainty avoidance .00005 .00019 .00002 .00003 .00000 .00001 .00001 .00000 .03239 .00136
Long-term orientation .0001* .0004* .00003 .00004 .00001 .00001 .00001 .00000 .0131* .00050
Indulgence .00002 .00018 .00001 .00001 .00004 .00004 .00001 .00001 .01335 .00444

*p < .05.

.25

Peculiar Notorious Cosmopolitan


Antagonist
T. Locsin

.20 M. Qureshi
J. Hernández
Abdullah II

S. Seibert

.15
itter and Press

J. Trudeau S. Piñera
P. Sanchez
( igh →

N. Maduro
B. Netanyahu
S. Kurz

K. bin Ahmed
P. Kagame N. Modi
.10 A. Aljubeir L. Nak-yeon
D. Medina
Narrowband J. Widodo Global
Media Compatibility:

J. Zarif A. Vucic
P. Poroshenko D.Tusk
Influencer S. Abe
E. Araújo
E. Morales G. Conte Icon
H. ClintonE. Rinkevics
M. bin Salman I. Duque D. Trump
Pope Francis I. Khan S. Hariri J. Bolsonaro
M. Al Maktoum M. Díaz-Canel D. Medvedev
.05 R. Rosselló
R. Al Abdullah M. Macri
M. Pompeo M. Buhari

K. Taro M. Vizcarra R. Erdoğan


M. bin Zayed A. Alvi J. Montserrat E. Macron
M. Mohamad Moon Jae-in
T. bin Hamad
.00
.50 .55 .60 .65 .70 .75 .80 .85 .90 .95 1.00

Universality: Topic Similarity to Other Political Leaders (Particular → Universal)

Figure 4. Mapping of the political leaders on the universality–compatibility matrix.


16 Journal of International Marketing XX(X)

previous regressions and did not see a significant difference. Classifying the leaders in the universality–compatibility
We also tried to cluster the data using k-means clustering and space, we saw that most of them fall in the universal–high
did not obtain meaningful insights. The results confirm the compatibility quadrant.
pattern we observed in the previous analysis of the leaders’ This research is not free of limitations. First, our selection of
communication: despite being strong national brands, their leaders is biased toward leaders with high visibility in the glo-
topic distribution shows a generic, universal pattern, and they bal arena. Local politicians, who deal less with international
have a relatively high level of media compatibility. However, affairs, might demonstrate more country-specific communica-
some of them talk about topics that deviate from the general tion patterns. Second, our choice of outlets is limited to press
distribution, and some enjoy press coverage that reflects their articles and Twitter. However, political leaders are active on
presentational media (which, interestingly, does not show other social platforms (e.g., Facebook, Instagram) and other
meaningful dependence on the political system of the country). representational media, such as television, which were not
studied here and could demonstrate different patterns. Third,
universality in communication can be measured in additional
ways, such as by the proportion of communication in English,
Discussion and Conclusions The number of international press conferences initiated by the
The focus of this article is to study universality and media leader, and the leader’s choice of clothing. Fourth, at the tech-
compatibility in the communication of political leaders. One nical level, language processing, although continuously
would expect that leading politicians, being salient icons of improving, does not yet deal reliably with multilingual data
their countries, will adopt a country-specific communication sets, nonstandard language, grammatical complexities, and
persona, and hence their communication will demonstrate meaning extraction.
country-related characteristics. On the other hand, these leaders Our results add to the understanding of globalization in
operate in a mediatized environment. This environment is con- brands that are typically strongly associated with their country
sidered a driver of globalization and suggests the value of more of origin. The universalism we observe might add to the body
homogenous, universal communication. of results in brand research, showing the advantages of global
Heavy mediatization also affects the relationship between positioning (Diamantopoulos et al. 2019; Guo 2013; Nijssen
presentational and representational communication outlets, and Douglas 2011). Methodologically, we show how these
which operate simultaneously and influence each other. While questions can be addressed quantitatively and systematically,
variety suggests large differences between outlets, the mutual for a large number of countries, using publicly available data.
influence might lead to high media compatibility. While these approaches are commonly used in marketing, our
Our study is the first, to the best of our knowledge, to study paper suggests that they could be implemented to gain insights
in communication research and political studies.
these two dimensions systematically and quantitatively, using
Is the universalism we observed good or bad for the citizens
large-scale data. We collected the communication data of 61
of the world? As stated in the introduction, the citizens of the
global world leaders from their Twitter accounts and press
world’s nations are increasingly facing challenges of a global
coverage, and used language processing tools to study the sen-
nature, such as immigration, environmental issues, and the future
timent, topic mixture, and use of pronouns.
of the European Union and other regional treaties. Coping with
Our results show a high level of universality and media
these challenges requires world leaders to create communication
compatibility across political leaders. Leaders are similar to
infrastructures that go beyond national boundaries. Our results
each other in their sentiment, topic mixture, and use of mes-
provide an optimistic view that such infrastructures might exist.
sages. Overall, the sentiment is positive, and the discussed Regarding media compatibility, the high compatibility level
topics are mainly diplomacy, economy, corruption, and the observed between Twitter and the press articles is a reminder of
Arab world. Of the pronouns, “we” messages are used the most, the costs that come with the mediatization that enables this
followed by “I” messages. The results are robust across com- universality. Our results indicate that the continuous spillover
munication channels, with Twitter being slightly more positive of content between presentational channels (e.g., social media)
than the press articles. Press articles focus significantly more and representational channels (e.g., press, television) makes
than Twitter on corruption, diplomacy, elections, and parlia- them similar to each other and raises concerns about the actual
mentary issues, whereas Twitter, relative to the press articles, variety of information sources available to citizens and voters,
deals more with personal messages, such as condolences, con- the objectivity of these sources, and their reliability.
gratulations, and thank-you messages, as well as economy.
With a few exceptions, we saw almost no significant depen-
Associate Editor
dence on country characteristics. Relative to Africa, leaders from
Asia, Europe, and Central America use first-person messages Kelley Hewett
more than second- and third-person messages when tweeting.
Also, leaders from countries with shorter-term orientation were Declaration of Conflicting Interests
found to use first-person pronouns, relative to second- or third- The author(s) declared no potential conflicts of interest with respect to
person pronouns, more often on Twitter, but the effects are small. the research, authorship, and/or publication of this article.
Peres et al. 17

Funding Draper, Robert (2018), “The Man Behind the President’s Tweets,” The
The author(s) disclosed receipt of the following financial support for New York Times Magazine (April 16), https://www.nytimes.com/
the research, authorship, and/or publication of this article: This article 2018/04/16/magazine/dan-scavino-the-secretary-of-offense.html.
was supported by the Israeli National Foundation and the KMart Enli, Gunn Sara and Eli Skogerbø (2013), “Personalized Campaigns in
foundation. Party-Centred Politics: Twitter and Facebook as Arenas for Polit-
ical Communication,” Information, Communication & Society, 16
(5), 757–74.
References French, Alan and Gareth Smith (2010), “Measuring Political Brand
Anderson, Chris (2013), “The Shared Genius of Elon Musk and Steve Equity: A Consumer Oriented Approach,” European Journal of
Jobs,” Fortune (November 21), https://fortune.com/2013/11/21/ Marketing, 44 (3/4), 460–77.
the-shared-genius-of-elon-musk-and-steve-jobs/. Geis, Michael L. (1987), “The Problem of Bias in Political Journal-
Bennett, Lance W. and Shanto Iyengar (2008), “A New Era of Min- ism,” in The Language of Politics. New York: Springer, 58–77.
imal Effects? The Changing Foundations of Political Commu- Golbeck, Jennifer, Justin M. Grimes, and Anthony Rogers (2010),
nication,” Journal of Communication, 58 (4), 707–31. “Twitter Use by the U.S. Congress,” Journal of the American
Bos, Linda, Wouter Van der Brug, and Claes De Vreese (2011), “How Society for Information Science and Technology, 61 (8), 1612–21.
the Media Shape Perceptions of Right Wing Populist Leaders,” Guo, Xiaoling (2013), “Living in a Global World: Influence of Con-
Political Communication, 28 (2), 182–206. sumer Global Orientation on Attitudes Toward Global Brands from
Bovet, Alexandre and Hernán A. Makse (2019), “Influence of Fake Developed Versus Emerging Countries,” Journal of International
News in Twitter During the 2016 US Presidential Election,” Marketing, 21 (1), 1–22.
Nature Communications, 10 (1), 7–20. Gürhan-Canli, Zeynep, Gülen Sarıal-Abi, and Ceren Hayran (2018),
Bovet, Alexandre, Flaviano Morone, and Hernán A. Makse (2018), “Consumers and Brands Across the Globe: Research Synthesis and
“Validation of Twitter Opinion Trends with National Polling New Directions,” Journal of International Marketing, 26 (1), 96–117.
Aggregates: Hillary Clinton vs Donald Trump,” Scientific Reports, Guzman, Francisco and Vicenta Sierra (2009), “A Political Candi-
8 (1), 8673–88.
date’s Brand Image Scale: Are Political Candidates Brands?”
Brants, Kees, Claes de Vreese, Judith Möller, and Philip van Praag
Brand Management, 17 (3), 207–17.
(2010), “The Real Spiral of Cynicism? Symbiosis and Mistrust
Henneberg, Stephan C. and Nicholas J. O’Shaughnessy (2007),
Between Politicians and Journalists,” International Journal of
“Theory and Concept Development in Political Marketing: Issues
Press/Politics, 15 (1), 25–40.
and an Agenda,” Journal of Political Marketing, 6 (2/3), 5–31.
Büschken, Joachim and Greg M. Allenby (2016), “Sentence-Based
Hepp, Andreas, Andreas Breiter, and Uwe Hasebrink (2018), Commu-
Text Analysis for Customer Reviews,” Marketing Science, 35
nicative Figurations: Transforming Communications in Times of
(6), 953–75.
Deep Mediatization. London: Palgrave Macmillan.
Cayla, Julien and Eric J. Arnould (2008), “A Cultural Approach to
Hepp, Andread, Stig Hjarvard, and Knut Lundby (2015),
Branding in the Global Marketplace,” Journal of International
“Mediatization: Theorizing the Interplay Between Media, Culture
Marketing, 16 (4), 86–112.
and Society,” Media, Culture & Society, 37 (2), 314–24.
Chadwick, Andrew (2006), Internet Politics: States, Citizens, and
Herz, Marc and Adamantios Diamantopoulos (2017), “I Use It but
New Communications Technologies. New York: Oxford Univer-
Will Tell You That I Don’t: Consumers’ Country-of-Origin Cue
sity Press.
Chadwick, Simon and Nick Burton (2008), “From Beckham to Usage Denial,” Journal of International Marketing, 25 (2), 52–71.
Ronaldo—Assessing the Nature of Football Player Brands,” Jour- Hjarvard, Stig (2013), The Mediatization of Culture and Society. Lon-
nal of Sponsorship, 1 (4), 307–17. don: Routledge.
Couldry, Nick and Andreas Hepp (2017), “The Continuing Lure of the Hockett, Jeremy (2005), “Brand ‘W’ and the Marketing of an Amer-
Mediated Centre in Times of Deep Mediatization: Media Events and ican President: Or, Logos as Logos,” Westminster Papers in Com-
Its Enduring Legacy,” Media, Culture & Society, 40 (1), 114–17. munication and Culture, 2 (2), 72–95.
Crystal, David (2006), Language and the Internet. Cambridge, UK: Hofstede, Geert (2011), “Dimensionalizing Cultures: The Hofstede
Cambridge University Press. Model in Context,” Psychology and Culture, 2 (1), article 8.
Dahlgren, Peter (2009), Media and Political Engagement. Cambridge, Howard, Philip. N. and Malcolm R. Parks (2012), “Social Media and
UK: Cambridge University Press. Political Change: Capacity, Constraint, and Consequence,” Jour-
D’Alessio, Dave and Mike Allen (2000), “Media Bias in Presidential nal of Communication, 62 (2), 359–62.
Elections: A Meta-Analysis,” Journal of Communication, 50 (4), Kock, Florian, Alexander Josiassen, and George Assaf (2019),
133–56. “Toward a Universal Account of Country-Induced Predispositions:
Diamantopoulos, Adamantios, Vasileios Davvetas, Fabian Bartsch, Integrative Framework and Measurement of Country-of-Origin
Timo Mandler, Maja Arslanagic-Kalajdzic, and Martin Eisend Images and Country Emotions,” Journal of International Market-
(2019), “On the Interplay Between Consumer Dispositions and ing, 27 (3), 43–59.
Perceived Brand Globalness: Alternative Theoretical Perspectives Krippendorff, Klaus (2004),”Reliability in Content Analysis: Some
and Empirical Assessment,” Journal of International Marketing, Common Misconceptions and Recommendations,” Human Com-
27 (4), 39–57. munication Research, 30 (3), 411–33.
18 Journal of International Marketing XX(X)

Krotz, Friedrich (2007), “The Meta-Process of Mediatization as a Continuance: Do Reach and Richness Matter?” Information &
Conceptual Frame,” Global Media and Communication, 3 (3), Management, 54 (2), 241–55.
256–60. Shapiro, Matthew A. and Libby Hemphill (2017), “Politicians and the
Krotz, Friedrich and Andreas Hepp (2013), “A Concretization of Med- Policy Agenda: Does Use of Twitter by the US Congress Direct
iatization: How Mediatization Works and Why ‘Mediatized Worlds’ New York Times Content?” Policy & Internet, 9 (1), 109–32.
Are a Helpful Concept for Empirical Mediatization Research,” Eur- Smith, Gareth (2009), “Conceptualizing and Testing Brand Personal-
opean Journal for the Philosophy of Communication, 3 (2), 119–34. ity in British Politics,” Journal of Political Marketing, 8 (3),
Marshall, David P. (2015), “Monitoring Persona: Mediatized Identity 209–32.
and the Edited Public Self,” Journal of Literary Studies, 28 (1), Steenkamp, Jan-Benedict E.M. (2019), “Global Versus Local Con-
115–33. sumer Culture: Theory, Measurement, and Future Research
Nijssen, Edwin J. and Susan P. Douglas (2011), “Consumer World- Directions,” Journal of International Marketing, 27 (1), 1–19.
Mindedness and Attitudes Toward Product Positioning in Adver- Strömbäck, Jesper and Peter Van Aelst (2013), “Why Political
tising: An Examination of Global Versus Foreign Versus Local Parties Adapt to the Media: Exploring the Fourth Dimension
Positioning,” Journal of International Marketing, 19 (3), 113–33. of Mediatization,” International Communication Gazette, 75 (4),
Özsomer, Ayşegül (2012), “The Interplay Between Global and Local 341–58.
Brands: A Closer Look at Perceived Brand Globalness and Local Tirunillai, Seshadri and Gerard J. Tellis (2014), “Mining Marketing
Iconness,” Journal of International Marketing, 20 (2), 72–95. Meaning from Chatter: Strategic Brand Analysis of Big Data Using
Packard, Grant, Sarah G. Moore, and Brent McFerran (2018), “(I’m) Latent Dirichlet Allocation,” Journal of Marketing Research, 51
Happy to Help (You): The Impact of Personal Pronoun Use in (4), 463–79.
Customer–Firm Interactions,” Journal of Marketing Research, 55 Ulrich, Dave and Norm Smallwood (2007), “Building a Leadership
(4), 541–55. Brand,” Harvard Business Review, 85 (7/8), 92.
Pang, Bo and Lillian Lee (2004), “A Sentimental Education: Senti- Van Aelst, Peter, Bart Maddens, Jo Noppe, and Stefaan Fiers (2008),
ment Analysis Using Subjectivity Summarization Based on Mini- “Politicians in the News: Media or Party Logic? Media Attention
mum Cuts,” in Proceedings of the 42nd Annual Meeting of the and Electoral Success in the Belgian Election Campaign of 2003,”
Association for Computational Linguistics. Stroudsburg, PA: European Journal of Communication, 23 (2), 193–210.
Association for Computational Linguistics, 271–78. Wattal, Sunil, David Schuf, Munir Mandviwalla, and Christine B.
Phipps, Marcus, Jan Brace-Govan, and Colin Jevons (2010), “The Williams (2010), “Web 2.0 and Politics: The 2008 U.S. Presiden-
Duality of Political Brand Equity,” European Journal of Market- tial Election and an E-Politics Research Agenda,” MIS Quarterly,
ing, 44 (3/4), 496–514. 34 (4), 669–88.
Shang, Shari S.C., Ya-Ling Wu, and Eldon Y. Li (2017), “Field Zhao, YanYan, Bing Qin, and Ting Liu (2010), “Sentiment Analysis,”
Effects of Social Media Platforms on Information-Sharing Journal of Software, 21 (8), 1834–48.

You might also like