Frame analysis of nonprofit organizations on microblog : a comparison of individualistic and collectivistic cultures

(1)

Graduate School of Communication

Master’s Programme Communication Science

MASTER’S THESIS

FRAME ANALYSIS OF NONPROFIT

ORGANIZATIONS ON MICROBLOG

A Comparison of Individualistic and Collectivistic Cultures

By Yujiao Pang

Student ID 10919074

Supervisor Theo Araujo

Date of Completion 24.06.2016

(2)

Abstract

Microblog has become a vital tool for NPOs to communicate with their stakeholders, but less is known about how different frames used by NPOs influence engagement with online users across cultures. Based on a content analysis of 1500 tweets sent by 30 NPOs on Twitter and Sina Weibo platforms across five countries, this research examined one type of message framing (promotion/prevention message), and two aspects of visual framing of people (number of main characters in a picture and intimacy level of human bodies) used by NPOs on microblog across individualistic and collectivistic cultures. To measure engagement on microblog, we use number of likes and number of retweets as dependent variables. Results show that, users from individualistic cultures are more likely to like and retweet a tweet that contains a picture shows fewer main characters, and the more main characters shown in the picture of a tweet, the more likes and retweets will be given by users from collectivistic cultures.

(3)

Introduction

The advent of social media, especially the rapid diffusion of microblog (e.g. Twitter, Sina Weibo), gives a great chance for nonprofit organizations to interact and communicate with stakeholders (Lovejoy & Saxton, 2012). This interaction mainly refers to customer

engagement which defined as “the level of a customer’s cognitive, emotional and behavioral investment in specific brand interaction (Hollebeek, 2011, p. 555).” In other words,

engagement aims at building a two-way conversation and an emotional bond between organizations and their customers. In the case of NPOs, their stakeholders are equivalent to customers for for-profit organizations. To measure the engagement on microblog, a “Like” that expresses positivity towards the content of the tweet, and the retweet function that allows one user to repost a tweet from another user can be used (Dugan, 2015; Boyd et al., 2010).

Everyone has a reason on liking or retweeting a tweet, but it cannot be denied that content is a determining factor. How to frame the content directly affect people’s reaction. “To frame is to select some aspects of a perceived reality and make them more salient in a communicating text, in such a way as to promote a particular problem definition, causal interpretation, moral evaluation, and/or treatment recommendation for the item described” (Entman, 1993, p.52). Frames are essential structures that explain how people’s mind thinks differently to similar event due to different factors. To understand people’s presumption, their backgrounds will be heavily considered.

Previous studies have found that cultural difference is one of the central drivers of people’s attitudes towards brands or advertising on social media. For example, Goodrich and De Mooij (2013) indicated that the use of information sources that influence online purchase

(4)

decisions strongly varies by different cultures. Messages can be particularly effective when they recognize and reinforce a particular group’s values (Dutta, 2007). That is why

organizations have been consistently creating culturally tailored messages to reach diverse audiences from various ethnic groups. This research followed the cultural sensitivity approach and examined one aspect of culture – individualism and collectivism (IDV/COL) defined by Hofstede (2001) – and its role in interacting NPOs with their stakeholders on microblog. In this study, we explored the cultural appeals of individualism and collectivism, and their impact when combined with promotion/prevention messages based on loss and gain frames and two aspects of visual framing of people, namely number of main characters and intimacy level shown in a picture.

In the context of globalization, organizations keep strengthening their global

communication strategies. NPOs are not exception, but they face more challenges due to their limited finance and manpower resources. To expand their influence, they need to connect with global and local audiences like for-profits do. Although previous articles have developed communication strategies for NPOs using social media, few of them discuss this topic from a cross-cultural level. Even less is known about international differences in framing usage of microblogging of NPOs and how these frames influence online engagement with users. To help address the research gap, this research will develop the research question: How will promotion and prevention messages and visual framing of people used by NPOs on microblog receive more “Likes” and retweets in individualistic/collectivistic cultures? From an individualistic vs. collectivistic perspective, this paper aims at identifying opportunities for NPOs to strengthen their strategies on promoting engagement with stakeholders on

(5)

microblog.

Theoretical background Engagement on Social Media

In the last few years, organizations emphasized and made every effort to enhance engagement with customers. But what does “engagement” mean? Many practitioners and scholars are working on ways to define and measure one specific type of engagement: customer engagement. Van Doorn et al. (2010) state that customer engagement behaviors go beyond transactions, and may be specifically defined as a customer’s behavioral

manifestations that have a brand or firm focus, beyond purchase, resulting from motivational driver. The motivational driver actually refers to an emotional bond between customers and organizations, which means that to engage customers is not just satisfied or loyal. The customers are emotionally attached to the organization’s brands or services (Gallup Consulting, 2010).

With the rise of social media, people rely on the Internet more and more in their everyday life. This change also influences the relationship between customers and

organizations. In the past, customers had to go to great lengths to get ahold of the brand to be heard, but now organizations are getting better at making those complicated interactions a thing of the past by creating conversations with customers through social media. The biggest contribution of social media is that it converts mass media from traditional, one-way

transmission into a more open and interactive two-way conversation. By using social media, organizations can develop relationships with existing as well as new followers and build communities that interactively collaborate to identify and understand problems and provide

(6)

solutions for them (Sashi, 2012).

Microblog is one type of social media in which brief posts (typically 140-200 characters) can be written or received with a variety of computing devices, including cell phones. The short messages of microblogging including short sentences, images, videos and links, which makes microblog a popular social media. Twitter is just one popular microblogging services that has been used by 310 million monthly active users (“Twitter company facts”, 2016), while as one of the most influential social media platforms in China, Sina Weibo (hereafter referred to as ‘Weibo’) has reached 222 million monthly active users as of September 2015 (“Weibo search users insights 2015”, 2015).

Each social media platform defines engagement differently based on their features and functionality. There are several forms of engagement used on microblog like Twitter and Weibo: “Like”, comment, click on link, the use of the”@” symbol to mention or reply appointed users, the use of retweet or share that allows one user to repost a tweet from another user, and the use of hashtags (#) which denotes that a message is relevant to a particular topic (e.g. Lovejoy, Waters & Saxton, 2012). To measure the engagement on microblog, the function of “Like” and retweet can be important. In 2015, Twitter changed their star icon for favorites to a heart and named it as “like”. While on Weibo, a “thumb up” icon represents “like”. Both the heart and “thumb up” are universal symbols that resonates across languages and cultures (Dugan, 2015).These icons which are internationally recognized as symbols for positivity are more expressive. They enable people to convey their positive emotions. Therefore “Like” something on microblog express positivity towards the content of a tweet (Dugan, 2015).

(7)

users. Research illustrated that those who are trying to engage in conversations or share information are more likely to retweet (Boyd et al., 2010). There are diverse motivations for retweeting, for example to amplify or spread tweets to new audiences; to begin a conversation; or to express opinions publicly etc. (Boyd et al., 2010). No matter what reasons are behind, the content people retweet is inextricably tied to the goals they have related to self-image and self-promotion, supporting conversation and building community (Boyd et al., 2010).

Organizations have realized the importance of microblogging and how microblogging can be harnessed to build stronger relationship with publics. In order to expand their influence, nonprofits keep expanding their market and trying to connect stakeholders worldwide.

However, according to Waters et al. (2009), NPOs failed to take advantage of the interactive nature of social media. The reasons behind this lie in a lack of time and resources being put into organizations’ social media accounts, or lack of knowledge on how to make the best of this technology (Lovejoy, Waters & Saxton, 2012). Apart from these reasons, this study suggests that cultural differences of media practitioners and online users should also be taken into consideration.

Individualistic/Collectivistic Cultural Dimension

Culture refers to a set of structures and institutions, values, traditions and norms which can be transmitted through generations in a certain period and place (e.g., Shweder & LeVine, 1984). An increasing interest has been seen in the consequences of culture for global marketing and advertising. Because of the growth of global business, technology and the Internet,

cross-cultural communication has become strategically important to organizations. To understand how people from different cultures speak, communicate and perceive the world

(8)

around them, dimensional models can be used that delineate national cultural values and help explain and predict consumer’s behavior (Hofstede, 2001). Several models about national culture have been developed in the past decades, but Hofstede’s model is used most widely in marketing studies (Goodrich & De Mooij, 2013). This model contains five dimensions of national culture labeled Power Distance, Individualism/Collectivism, Masculinity/Femininity, Uncertainty Avoidance, and Long-/Short-Term Orientation. Among these dimensions, Individualism(IDV)/Collectivism (COL) has been central to cross-cultural research and has attracted substantial attention from scholars including customer psychology, cultural psychology, marketing, and communication (e.g. Aaker, 2006; Oyserman, Coon & Kemmelmeier, 2002; Triandis, 1995).

Individualism can be defined as people looking after themselves and their immediate family only, while collectivism can be defined as people belonging to in-groups who look after each other in exchange for loyalty. In individualistic cultures, people are “I”-conscious and their identity is in the person. Whereas in collectivism cultures, people are “we”-conscious and their identity is based on the social system to which they belong. In particular, individualism is mostly seen in the cultures of Western Europe and North America, whereas collectivism is mostly seen in the cultures of Asia, Africa, and parts of Europe and Latin America (Triandis, 1993; Nelson & Fuvish, 2004). Past research has shown that the individualistic and

collectivistic frameworks have important implications for the persuasiveness and processing of advertisements, namely persuasion versus creating trust (De Mooij & Hofstede, 2010; Goodrich & De Mooij, 2013; De Mooij & Hofstede, 2011). That is because individualistic cultures are low-context communication cultures in which people communicate in an explicit

(9)

and directly way. Collectivism cultures are high-context communication cultures with an indirect style of communication, thus it is necessary to build a relationship and trust between parties.

Message Framing: Promotion and Prevention Message

Goffman defined frames as “the principles of organization which govern [social] events” (1981, p.63). Entman proposed that “to frame is to select some aspect of a perceived reality and make them salient in a communicating text” (1993, p.52). That means in the light of an issue, media makers create a frame package, which contains a definition, an explanation, and an evaluation of the event and ultimately results in a number of logical conclusions. This process is not only delivering information, but also building meaning on certain messages. Thus, frames can shape people’s thoughts and expectations and try to make them feel the way the presenter wants them to. “Frames are a central part of a culture and are

institutionalized in various ways,” as Goffman stated. As a part of culture, its essence is in social interaction which happens between the textual level, the cognitive level, the extra medial level and the stock of frames that is available in a given culture (Van Gorp, 2007). In other words, frames are essential structures that explain how people’s mind thinks differently to similar event due to different factors. Whether media makers or audiences, in the process, they all act in accordance with their cultures. That is why messages can be particularly effective when they recognize and adapt to a particular group’s values (Dutta, 2007).

Higgins’s (1997) regulatory focus theory refers to the extension of the basic hedonic principle of approach and avoidance to allow for distinct self-regulatory strategies and needs. Specifically, self-regulations toward any specific goal focused on two distinct types of desired

(10)

goals – promotion and prevention. Individuals with a promotion focus regulate their behavior toward positive outcomes, and those with a prevention focus direct their behavior away from negative outcomes (Sung & Choi, 2011).

Research have suggested that the independence or interdependence of an individual’s currently accessible self-construal is an antecedent of regulatory focus. These distinct self views have primarily been explored through cross-cultural comparison. Independent self-construal, commons to members of individualistic cultures, emphasizes autonomy and individuality and their attributes is constructed by making the individual separate and unique from others. The independent goal of being positively distinct, with its emphasis on

achievement and autonomy, may be more consistent with a promotion focus. Interdependent self-construal, on the other hand, common to members of collectivistic cultures, defines individuals by social relationships and group memberships with an emphasis on social harmony and norms (Markus & Kitayama, 1991; Triandis, 1989). The goal of harmoniously fitting in with others, with its emphasis on fulfilling various social roles and maintaining connections with others, may be more consistent with prevention focus. Therefore, previous research found that individuals with a dominant independent self prefer to focus on

promotional versus prevention based information regarding themselves, whereas the

converse is true for individuals with a dominant interdependent self (e.g., Aaker and Lee 2001; Lee, Aaker, and Gardner 2000). As individuals from individualistic and collectivistic cultures have difference in self-construal, we assume that their attitudes toward a tweet will also be different in terms of the two types of messages. Thus we proposed our first two hypotheses:

(11)

tweet by an NPO when it uses a promotion frame than when it does not use this frame. H2: Users from collectivistic cultures will be more likely to (a) like and (b) retweet a tweet by an NPO when it uses a prevention frame than when it does not use this frame.

Visual Framing of People

Text is not the only communication form of content on microblog, organizations also convey information through visual messages such as photographs and infographics. The visual components are powerful framing tools due to the fact that viewers see them as representations of reality (Messaris & Abraham, 2001), and they have power to create stronger emotional and immediate cues (Rodriguez & Dimitrova, 2011). Visuals can readily communicate a complex idea with a greater level of experience and familiarity to a viewer than can be achieved with text alone.

In the context of cultural differences, visual frames are also expected to be applied in different ways. One of the common types of photo content shared on microblog is people portrayal. A people portrayal is a photo of people or the one with human faces in them. Hofstede (2001) stated that comparison of the number of people shown in advertisements is not a measure of individualism/collectivism. This study will instead focus on the main

characters, those who have a direct relation to the screen or occupy the main part of the picture, shown in the images. Human faces are readily distinguishable which are shown to be powerful visual tool that used in human no-verbal communication. Based on the research of Bakhshi et al. (2014), the existence of a face in photo increases the likelihood of receiving “like” and comments. Therefore, we assume that individualistic/collectivistic people have differences in liking the images with different number of main characters.

(12)

H3: The fewer main characters shown in the picture of a tweet sent by an NPO, the more (a) likes and (b) retweets it will get from users from individualistic culture.

H4: The more main characters shown in the picture of a tweet sent by an NPO, the more (a) likes and (b) retweets it will get from users from collectivistic culture.

Social distance is another concept used as criteria to analyze images. Social distance is related to Hall’s (1966) concept of proxemics or the psychology of people’s use of space. Six values can be assigned to social distance based on how the human subjects’ bodies are represented in the frame: intimate distance, close personal distance, far personal distance, close social distance, far social distance and public distance (Hall, 1966, Rodriguez & Dimitrova, 2011). According to Hall (1966), perception of the levels of intimacy of space is culturally determined. Collectivists have a general tendency to be more expressive of emotions, touching, and prefer close social distance, whereas individualists are less

emotionally expressive, generally more socially interactive, and prefer more social distance (Hall, 1966).

H5: The lower the social distance (higher intimacy) shown in the image of a tweet, the more (a) “Likes” and (b) retweets the tweet could receive from users from collectivistic culture.

H6: The higher the social distance (lower intimacy) shown in the image of a tweet, the more (a) “Likes” and (b) retweets the tweet could receive from users from individualistic culture.

In general, to answer our research questions, we come up with five hypotheses from two aspects: message framing and visual framing of people portrayal. However, when a tweet includes an image, it often will be accompanied by the corresponding text description. That

(13)

means online users’ attitudes may not be affected by an image only or the text only, but by a combination of text and visuals. The interactive effect of text framing and visual framing is an underexplored and less predictable phenomenon relevant to new media (e.g., Fahmy et al., 2014). Paivio (1991) have proposed that learning and memory is improved when media messages are presented in both visual and verbal modalities. Although Powell et al. (2015) have found that when images and text are presented together in a typical news report, the text frames influences opinions regardless of the accompanying image, whereas the frames carried by the image drives behavioral intentions irrespective of the linked text. We still want to explore the influence of a combined frames on microblog. Therefore we propose a

sub-research question: combined promotion/prevention message and visual framing of people, will they influence users’ attitudes to like or retweet a tweet sent by NPOs?

Methodology

For this study, the hypotheses’ testing is based on a quantitative content analysis of microblog messages on two microblog service platforms, namely Twitter and Sina Weibo.

Sample

The sample accounts of microblog were selected in three steps. First of all, based on the measure of individualism as proposed by Hofstede (2001) 1(high scores on this factor were anchored at individualism, low scores at collectivism), relating to our language limitation, five countries were finally selected: China, Japan, South Africa, United Kingdom and United States. Secondly, a ranking of Top 100 Nonprofits on the web (2015)2 was used to select

1

The actual scores were retrieved from the website https://geert-hofstede.com/national-culture.html 2

The ranking this research used was the 2015 version, see the websitehttps://topnonprofits.com/lists/best-nonprofits-on-the-web/

(14)

global organizations. This ranking selected 500 organizations from thousands of nonprofit organizations and collected data for each nonprofit from all 7 measured criteria (e.g.,

Facebook Likes, Twitter Followers). Finally they released a ranking of Top 100 Nonprofits on the web. From all the 100 nonprofits, this study chose three global organizations that have Twitter or Weibo accounts for the five selected countries: United Nations Children’s Fund (UNICEF), Save the Children, and Doctors without Borders.

Thirdly we also choose three local nonprofits for each country. That is because local accounts focus more on local cultures and thus the effect of IDV/COL can be enhanced. They were selected according to two criteria. First, the selected organizations must have more than 2,000 posts on Twitter or Weibo so that enough messages can be coded. Second, these microblog accounts must have at least 2,000 followers on Twitter or Weibo. Based on these, 15 NPOs’ microblog accounts were selected. There are many type of NPOs around the world, to better fit our hypotheses, NPOs that aim to help people, including education, healthcare or human rights oriented, are objects of concern as human portrayals are the study units of this research. Because of their human-based values, they may be present and describe more people portrayals and their situations on their microblogging accounts. Then 30 NPO accounts are selected finally (See Appendix 1).

The time frame for sampling of post was from May. 1st, 2015 to May. 1st, 2016. Within this time period, 50 posts were randomly selected from all the posts presented on each NPO accounts’ page. This research collected 1500 tweets totally from 30 microblog accounts, among which 25 posts contained more than 1 images within a single post. For posts that have more than one pictures, it is hard to distinguish the difference between several pictures as

(15)

they all belong to one post and receive the same number of likes and retweets. Moreover, on both microblog service platforms, the first picture always takes the biggest space and thus attracts more attention compared with other pictures. As a result, only the first picture shown in the post was chosen to present the post which has more than one pictures and be

analyzed.

Original posts, retweets and replies are three types of post on Twitter, while Weibo doesn’t have the reply function. In terms of retweets, number of likes or retweets of a retweet is only presented on Weibo but not on Twitter. Therefore, this research will focus on original microblog messages posted by NPOs to balance the different functions of Twitter and Weibo. Thus, to do the data preparation, we first removed all other types of tweets (n = 172) and only kept original post. Because the selected microblog accounts were different in their size of organization and number of followers, which we assumed will influence their number of likes and retweets. In order to ensure the accuracy of results, 8 posts were found had higher scores than 1000 on the two dependent variables and therefore were removed from the dataset. Finally, the sample was composed of 1320 posts.

Measures

The codebook (See Appendix) of this research analyzed three aspects of each post: basic information, promotion/prevention message and visual framing. Based on the research of promotion/prevention message of Aaker and Lee (2001) and research of Rodriguez & Dimitrova (2011), we developed our codebook. To ensure informant quality and validity, we first did a pilot test of 150 tweets to test whether the codebook could be used for further research. The pretest posts were not included in the final sample. The codebook was initially

(16)

constructed in English. After the pretest, we translated each variables into Chinese to help our second coder better understand.

Basic information category. The basic information of every post including, for example,

name of NPO, the country of the account, date and time of each post, type of post and form of post, and number of likes and retweets, among which the number of likes (M = 35.95, SD =61.73 ) and retweets (M = 31.64, SD = 61.73) are considered as dependent variables of this research. Number of likes or retweets of an original post are shown on both of the microblog service platforms.

Level of individualism and collectivism. A recalculated scale from 0 to 100 that

measures individualism and collectivism, as proposed by Hofstede (2001)3, was used by this research to select five countries with different scores. The higher the score, the higher level of individualism it represents (M = 63.54, SD = 26.71): China (20), Japan (46), South Africa (65), United Kingdom (89) and United States (91). To test H2, H3 and H4, a new variable “Level of collectivism” (M = 36.46, SD = 26.71) was computed by using 100 minus the value of level of individualism.

Message framing category. The message framing section investigated the two types

of messages: promotion message (M = 0.18, SD = 0.39) and prevention message (M = 0.10, SD = 0.30). Two master students coded message of each post individually for a presence or absence of the type. The coders were encouraged to pick the type that best fit the post. Coders coded for promotion messages when the content of a tweet demonstrated obvious positive outcomes, to view events framed in terms of winning being more important. To the

3

(17)

extent that promotion message focused on things that would be gained by engaging in a behavior. On the other side, prevention messages were set up to prevent negative outcomes, to view events framing in terms of losing as being more important. This kind of message focused on things that were lost by not engaging in a behavior.

Visual framing category. The last category aimed at analyzing images of posts from

two aspects of visual messages: number of main characters and level of intimacy. In this research, main characters were defined as those people who have a direct relation to the camera or occupy the main part of the picture. Based on this rule, researchers were directly counting the number of main characters from 1 to 5 or more. (31.9% of the messages had one main character, 19.5% had two main characters, 6.9% had 3 main characters, 4.7% had 4 main characters and 14.7% had 5 or more main characters.)

Six values can be assigned to the variable level of intimacy based on how the human subjects’ bodies are represented in the frame. As one picture can only show one level of intimacy, we recoded these six binary variable into a 6-point scale to measure the intimacy of human bodies in the image, the lower the score, the closer the intimacy space (M = 3.41, SD = 1.58) (Hall, 1966, Rodriguez & Dimitrova, 2011). However, H4 aimed at exploring the relationship between dependent variable and collectivism, therefore level of intimacy was recoded as another 6-point scale, namely social space: the lower the score, the closer the social distance (M = 3.58, SD = 1.57).

Inter-coder Reliability

Due to language limitation, this study invited a second coder who was responsible for analyzing 300 Japanese posts. However as the second coder is not an English speaking

(18)

people, in order to ensure reliability and validity of the codes and avoid subjectivity, we found a third coder to help with checking reliability. 75 Chinese and 75 English posts among the total posts, which took 10% of the 1500 posts, were given to first and second coders to enable reliability testing. Inter-coder reliability was checked automatically (Freelon, 2010). The result showed that there was a strong inter-coder reliability for each variable, from the lowest variable promotion/prevention message (Krippendorff’s Alpha (interval) = 0.89) to highest variable numbers of like (Krippendorff’s Alpha (interval) = 0.99).

Results Main Analysis

H1 and H2 aimed at analyzing the effect of relationship between IDV/COL level and promotion/prevention message on number of likes and retweets. To test the hypotheses, four multiple regressions were conducted, following we will describe results of each model.

The first multiple regression with number of likes as dependent variable, showed that the model as a whole was significant, F (5, 1313) = 32.11, p<.000. The regression model can therefore by used to predict the variation in number of likes, but the strength of the prediction is moderate: 33% of the variation in number of likes can be predicted on the basis of the dependent variables (R2=.11). The results indicated that the interaction between level of individualism and promotion message (p>.05), had no significant effect on the dependent variable. On the other hand, the regression model with number of retweets as dependent variable was also significant, F (5, 1313) = 18.12, p<.000. The independent variables

predicted 25% of the variance in number of retweets (R2=.07). Number of followers (p<.000) significantly predicted number of retweets for each post, while level of individualism (p=.001)

(19)

had a weak association. However, other predictors had no significant effect on number of retweets. H1 was not supported (See Table 1).

Table 1 – Regression Analysis Summary for Variables Predicting Number of Likes and Number of

Retweets (N = 1319)

Number of likes Number of retweets

b SE B b* b SE B b*

Constant 5.92 5.64 7.63 4.98

Level of individualism 0.16 0.07 .07* 0.19 0.06 .10**

Promotion message 4.73 11.6 .03 -2.87 10.24 -.02

Level of individualism x Promotion

message -0.06 0.17 -.03 0.06 0.15 .03

Number of followers 6.16 0.00 .30*** 3.54 0.00 .20***

Number of images 7.59 3.64 .06* 4.43 3.21 .04

R2 0.33 0.07

F 32.11*** 18.12***

Note. *p<0.05, **p<0.01, ***p<0.00a. B, unstandardized regression coefficient; SE, unstandardized standard error; b*, standardized beta.

To test H2, two multiple regressions were conducted with level of collectivism, prevention message and their interaction as independent variable and number of followers and number of images as control variables. The regression model with number of likes as independent variable was significant, F (5, 1314) = 32.68, p<.000 (See Table 1). The independent variables predicted 33% of the variance in number of likes (R2=.11). Based on the results, interaction between collectivism level and prevention message failed to predict number of likes. Another model with number of retweets as independent variable as a whole was significant as well, F (5, 1314) = 18.14, p<.000. The independent variables predicted 25% of the variance in number of likes (R2 = .07). Collectivism level (p<.000), and number of followers (p<.000) had a significant, moderately strong association with number of retweet. However the other three dependent variables had no significant influence on number of retweets. Thus, our H2 were not supported either (See Table 2).

(20)

Retweets (N = 1320)

b SE B b* b SE B b*

Constant 22.79 4.48 27.59 3.96

Level of collectivism -0.16 0.07 -.07* -0.21 0.06 -.11***

Prevention message -21.81 14.42 -.11 -1.59 12.75 -.01

Level of collectivism x Prevention

message 0.30 0.26 .08 0.07 0.23 .02

Number of images 7.27 0.05 .05 4.34 3.20 .04

R2 0.11 0.07

F 32.68*** 18.14***

A multiple regression is conducted to see if individualism level, number of main

characters and their interaction, and number of followers and number of image predicted the number of likes of a tweet. The model was significant, F (5, 1312) = 33.78, p<.000 (See Table 3). The independent variables predicts 34% of the variance in number of likes (R2 = .11). All independent variables had a significant association with number of likes (See Table 3).

Especially the interaction between individualism level and number of main characters (p=.005) strongly predicted number of likes. The other model with number of retweets was also

significant, F (5, 1312) = 24.50, p<.000. 29% of the variation in number of retweets can be predicted on the basis of the dependent variables (R2=.08). Level of individualism (p<.000), number of main characters (p<.000), and their interaction (p<.000), and number of followers (p<.000), had a strongly significant influence on number of retweets. Thus, our H3 was supported: The fewer main characters shown in the picture of a tweet sent by an NPO, the more likes and retweets it will get from users from individualistic culture.

Retweets (N = 1318)

(21)

Constant -5.55 7.05 -14.10 6.17

Level of individualism 0.35 0.10 .15*** 0.56 0.09 .28***

Number of main characters 5.66 2.43 .15* 10.41 2.13 .33***

Level of individualism x Number of

main characters -0.10 0.04 -.20** -0.17 0.03 -.39***

Number of images 8.26 3.90 .06* 4.41 3.41 .04

R2 0.11 0.09

F 33.78*** 24.50***

On the other hand, we also found statistical proof for our H4 (See Table4). The multiple regression model with number of likes as dependent variable, collectivism level, number of main characters and their interaction, and number of followers and number of image as independent variables was also significant, F (5, 1312) = 33.78, p<.000. The independent variables predicts 34% of the variance in number of likes (R2 = .11). Level of collectivism (p<.000), number of main character (p=.01), number of followers (p<.000), number of images (p=.03), have a significant association with number of likes (See Table 3). Significant influence was found on number of likes predicted by the interaction of collectivism level and number of main characters (p=.005). Another multiple regression with number of retweets as

independent variable was significant, F (5, 1312) = 24.50, p<.000. The independent variables predicts 29% of the variance in number of retweets (R2=.09). Level of collectivism (p<.000), number of main character (p<.000), the interaction of these two variables (p<.000), number of followers (p<.000), have a significant moderately strong association with number of retweets, while number of images (p>.05), doesn’t have any influence on the independent variable. H4 was supported, which means that the more main characters shown in the picture of a tweet, the more likes and retweets it will receive from users from collectivistic cultures.

(22)

Retweets (N = 1318)

b SE B b* b SE B b*

Constant 29.82 5.28 41.60 4.62

Level of collectivism -0.35 0.10 -.15*** -0.56 0.09 -.28***

Number of main characters -4.47 1.74 -.12* -6.93 1.53 -.22***

Level of collectivism x Number of

main characters 0.10 0.04 .16** 0.17 0.03 .32***

Number of images 8.26 3.90 .06* 4.41 3.41 .04

R2 0.11 0.09

F 33.78*** 24.50***

Our last hypotheses were not supported, since neither of the interaction between IDV level and intimacy level nor the interaction between COL level and social distance significantly predicted number of likes and retweets (See Table 5 and Table 6). Number of followers (p<.000), as a control variable, was the most significant predictor and it took most of the influence of number of likes or retweets for all the four models. Although we found that IDV (p=.034 vs. p=0.37) also had a significant influence on these two dependent variables in two models which took IDV level, intimacy level, the interaction between them, number of

followers and number of images as independent variables, we failed to support H5 and H6.

Retweets (N = 1019)

b SE B b* b SE B b*

Constant 2.34 12.23 -3.60 10.87

Level of individualism 0.34 0.16 .15* 0.29 0.14 .15*

Intimacy space 1.33 2.83 .03 2.35 2.52 .07

Level of individualism x Intimacy

space -0.05 0.04 -.10 -0.03 0.04 -.09

Number of images 4.63 6.35 .02 7.09 5.65 .04

(23)

F 30.05*** 15.36*** Note. *p<0.05, **p<0.01, ***p<0.00a. B, unstandardized regression coefficient; SE, unstandardized standard error; b*, standardized beta.

Retweets (N = 1018)

b SE B b* b SE B b*

Constant 13.80 976 27.59 3.96

Level of collectivism -0.02 0.16 -.01 0.94 1.72 -.03

Social Distance 3.20 1.94 .08 0.94 1.72 .03

Level of collectivism x Social

Distance -0.05 0.41 -.09 -0.03 0.04 -.08

Number of images 4.47 6.39 .02 7.12 5.69 .04

R2 0.13 0.07

F 29.97*** 15.32***

To extend the results, we also analyzed the relationship between two dependent

variables and the combined independent variables. The first regression model with number of likes as dependent variable, variables that related to individualism as independent variables was significant (See Table 7). The independent variables predicts 39% of the variance in number of retweets (R2=.09). We didn’t found any other prediction on number of likes from other predictors. However, in the next significant regression model with number of retweets as dependent variable, F (13, 1005) = 10.68, p<.000, we found that the interaction between IDV and number of main characters (p=.002), interaction between IDV and intimacy level (p=.02), and interaction among IDV, promotion message and intimacy level (p=.005) were significantly predicted number of retweets. Thus, when every variables were controlled, we found that the lower level of intimacy, the more retweets will receive by people from individualistic cultures; the fewer number of main characters shown in the pictures, the more retweets will be given by

(24)

users from individualistic cultures; a tweet with promotion message and lower level intimacy shown in the pictures will receive more retweets from users from individualistic cultures.

Retweets (N = 1018)

b SE B b* b SE B b*

Constant -7.77 12.50 -16.25 10.94

Individualism level 0.41 0.21 .18* 0.43 0.18 .22*

Promotion message -4.30 13.14 -.03 -7.06 11.50 -.05

Number of main characters 13.45 3.25 .33*** 17.36 2.85 .50***

Intimacy -5.26 3.22 -.14 -6.13 2.82 -.18*

Individualism level x Promotion

message -0.20 0.37 -.08 -0.52 0.33 -.25

Individualism level x Number of

main characters -0.16 0.08 -.30 -0.22 0.07 -.49**

Individualism level x Intimacy 0.08 0.06 .20 0.12 0.05 .32*

Individualism level x Promotion message x Number of main characters

-0.03 0.13 -.03 0.08 0.12 .11

Individualism level x Promotion

message x Intimacy 0.10 0.10 .16 0.25 0.09 .44**

Individualism level x Intimacy x

Number of main characters -0.01 0.01 -.14 -0.01 0.01 -.15

Individualism level x Promotion message x Number of main characters x Intimacy -0.01 0.03 -.03 -0.04 0.03 -.26 Number of followers 6.06 0.00 .30*** 3.23 0.00 .19*** Number of images 4.53 6.32 .02 6.72 5.53 .04 R2 F 0.15 13.66*** 0.12 10.68*** Note. *p<0.05, **p<0.01, ***p<0.00a. B, unstandardized regression coefficient; SE, unstandardized standard error; b*, standardized beta.

On the other hand, no significant effects predicted by interaction among COL, number of main characters and social distance level on number of likes nor number of retweets, even though both of the models as a whole were significant (p<.000 vs. p<.000) (See Table 8). However, when we control all the variables related to COL, the interaction among COL level, number of main characters and social distance level (p=.05) had a weak significant influence

(25)

on number of likes. The results indicated that in the case of controlling all these variables, we found: if a tweet contains a picture shows more main characters as well as shows higher level of intimacy, it will receive more likes from users from collectivistic cultures.

Retweets (N = 1018)

b SE B b* b SE B b*

Constant 54.62 14.96 69.88 13.17

Collectivism level -0.68 0.33 -.31* -1.11 0.29 -.58***

Prevention message -23.61 18.21 -.11 -4.63 16.03 -.03

Number of main characters -8.92 2.57 -.22** -11.48 2.26 -.33***

Social Distance -1.68 2.41 -.04 -5.35 2.13 -.16*

Collectivism level x Prevention

message 0.38 0.62 .10 0.35 0.55 .11

Collectivism level x Number of

main characters 0.13 0.08 .21 0.22 0.07 .42**

Collectivism level x Social distance -0.02 0.07 -.04 0.05 0.06 .12 Collectivism level x Prevention

message x Number of main characters

-0.02 0.17 -.01 -0.04 0.15 -.04

Collectivism level x Prevention

message x Social distance -0.01 0.15 -.01 -0.04 0.13 -.05

Collectivism level x Number of

main characters x Social distance 0.03 0.02 .17* 0.03 0.01 .16 Collectivism level x Prevention

message x Number of main characters x Social distance

0.00 0.05 .01 0.01 0.04 .02 Number of followers 5.98 .00 .30*** 3.31 0.00 .19*** Number of images 4.26 6.36 .02 7.64 5.60 .04 R2 F 0.15 13.90*** 0.11 9.90*** Note. *p<0.05, **p<0.01, ***p<0.00a. B, unstandardized regression coefficient; SE, unstandardized standard error; b*, standardized beta.

Discussion and Implication

The present study sheds light on the effect of promotion/prevention message framing, visual framing of people on number of likes and retweets of a tweet sent by NPOs across

(26)

nonprofit organizations in our sample over a one-year period, we analyzed their texts and first image by using promotion and prevention message framing and two types of visual framing: number of main characters and intimacy space of human bodies shown in the image.

An important finding is about the influence of number of main characters on people’s engagement behaviors on microblog across individualistic and collectivistic cultures. Photos are becoming prominent means of online communication. Previous literature have examined the pervasive presence of photos in the online world. For example, in a research of Instagram, Bakhshi, Shamma and Gilbert (2014) found that the existence of a face in a photo significantly affects its social engagement. To develop how people interact and engage with the content of image, we analyze the influence of number of main characters shown in the picture on

number of likes and retweets toward a tweet across individualism and collectivism. The results indicated that number of main characters shown in the image on tweets do have a significant effect on users’ attitudes. The fewer main characters shown in the picture of a tweet sent by an NPO, the more likes and retweets it will get from users from individualistic culture. On the other hand, the more main characters shown in the picture, the more likes and retweets are given by people from collectivistic culture.

Surprisingly, we failed to confirm our hypotheses that relate to promotion and prevention message. Aaker and Lee (2001) demonstrated that the effectiveness of a promotion or

prevention framed appeal depends on one’s interdependent self or independent self and perceived situation. As we mentioned before, relating to IDV/COL, individualists tend to have an independent self as they are encouraged to emphasize individual separate and unique from others, while collectivists tend to have an interdependent self as they emphasized more

(27)

on social harmony and norms and groups (Markus & Kitayama, 1991; Triandis, 1989).

However, in the case of NPOs’ use of microblogging, promotion and prevention messages do not affect people’s decisions of liking or retweeting a tweet. A possible reason behind this is that by the impact of globalization, culture is viewed as moving (Zhang, 2010). In the context of globalization, different cultures are inter-connected across geographical boundaries, and individuals in a particular locality can demonstrate elements of multiple cultures (Craig and Douglas 2006; Hermans and Kempen 1998; Hong et al. 2000). Empirical studies have also suggested that Chinese generation-X consumers are becoming bicultural because they have incorporated both individualistic and collectivistic values (Zhang, 2009). As a result, even people from different countries are different in fundamental cultural roots, the fusion of individualistic and collectivistic cultures has become a norm for people of the new era. Thus, their reactions toward promotion and prevention messages are not as obvious as we

expected.

Our final analysis of combined variables also discovered influence on liking or

retweeting a tweet. For users from individualistic cultures, if the content of a tweet contains promotion message and also with a picture shown lower intimacy, they are more likely to retweet the tweet. While if a tweet has pictures that show more main characters and higher intimacy of human bodies, it has great possibility to receive likes and retweets from users from collectivistic cultures.

From our results, we also observed that the number of likes and number of retweets of a same tweet is different. This will also lead to different results for the influence of frames. People like a tweet to show their positivity toward the content of the tweet and the

(28)

organizations. However, when they retweet a tweet, it not only expresses their endorsement of it, but also through retweeting to share or spread it to their audiences and express their opinions publicly (Boyd et al., 2010). Future research could develop more on whether cross culture dimension is one of the reason for people liking or retweeting a tweet.

Managerial Implication

Microblogging service provides a great chance for NPOs to engage with their stakeholders across time and space, and become more interactive than using their websites alone. Using social media is a trendy thing to do for organizations, however, what contents they can post on social media are much more influential on customer engagement.

Our findings provide NPO practitioners with managerial implications in the usage of visual framing. Managers of NPOs should not simply be concerned about sending images on their microblog pages. It seems more important that these images actually influence users’ intention to engage with them. For users from individualistic cultures, NPOs may post images with single or fewer main characters. Whereas for users from collectivistic cultures, NPOs may post images with more main characters. As those users may consider a group of people as a social unit, they prefer to share and enhance in-group relationships.

Moreover, in the era of social media, in the case of microblog, information is presented by using more than one medium. In terms of the form of content of a tweet, most of them are either photo-illustrated, or videos, and contain no text-only “status” type posts. Therefore, we suggest that managers of NPOs should also remember to use combined frames within one tweet to enhance engagement with their stakeholders. NPOs, who are facing to audiences from individualistic cultures, could try to combine a promotion message with a picture that

(29)

shows higher level of intimacy of human bodies in a tweet. On the other side, for those people from collectivistic cultures, NPOs could focus more on the use of images. Specifically, they could try to post pictures with a group of main characters and show their bodies in a lower intimacy.

Limitations and Future Research

The study has some limitations. First, due to the language limitation, our choice of countries may be limited. Stratified sampling could be taken to extract samples of countries from the IDV scale in the future research to ensure the validity of samples. Second, this research only focus on one type of NPOs, which are NPOs aimed to helping people. Most selected organizations of our sample are children’s and health NPOs, which makes the results not able to extend to any type of NPOs.

Third, IDV (vs. COL) is the only cultural dimension that we used to test framing use of NPOs. Future research could incorporate other cultural variables that may have a potential impact on microblogging users’ behaviors. For example, with a similar level of COL, people with long-term orientation prefer a more modest self-presentation, whereas people with short-term orientation prefer to enhance the self in their presentation (Goodrich and De Mooij, 2013). These difference on presenting themselves may also affect their perceived impression on a tweet sent by organizations.

Lastly, the images we used to operationalize our variables are all the first pictures of each tweet. Although the first picture of a tweet usually takes the greatest space and draws attention more easily, we still cannot overlook the influence of other images. Future research

(30)

could take all images of a tweet into analysis and examine the visual framing used by NPOs across cultures from a more comprehensive perspective.

References

Aaker, J. L. (2006). Delineating culture. Journal of Consumer Psychology, 16, 343–347. Alper, M. (2013). War on Instagram: Framing conflict photojournalism with mobile

photography apps. New Media & Society, 1461444813504265.

Boyd, D., Golder, S., & Lotan, G. (2010, January). Tweet, tweet, and retweet: Conversational aspects of retweeting on twitter. In System Sciences (HICSS), 2010 43rd Hawaii

International Conference on (pp. 1-10). IEEE.

Briones, R. L., Kuch, B., Liu, B. F., & Jin, Y. (2011). Keeping up with the digital age: How the American Red Cross uses social media to build relationships. Public relations review, 37(1), 37-43.

Chen, Y. R. R., & Fu, J. S. (2015). How to be heard on microblogs? Nonprofit organizations' follower networks and post features for information diffusion in China. Information, Communication & Society, 1-16.

Curtis, L., Edwards, C., Fraser, K. L., Gudelsky, S., Holmquist, J., Thornton, K., & Sweetser, K. D. (2010). Adoption of social media for public relations by nonprofit organizations. Public Relations Review, 36(1), 90-92.

De Mooij, M., & Hofstede, G. (2010). The Hofstede model: Applications to global branding and advertising strategy and research. International Journal of Advertising, 29(1), 85-110. De Mooij, M., & Hofstede, G. (2011). Cross-cultural consumer behavior: A review of research

(31)

Dutta, M. J. (2007). Communicating about culture and health: Theorizing culture-centered and cultural sensitivity approach. Communication Theory, 17, 304–328.

Edman, H. (2010). Twittering to the top: A content analysis of corporate tweets to measure organization-public relationships (Doctoral dissertation, Faculty of the Louisiana State University and Agricultural and Mechanical College In partial fulfillment of the

Requirements of the degree of Master of Mass Communication in The Manship School of Mass Communication by Haley Edman BA, Louisiana State University).

Entman, R. M. (1993). Framing: Toward clarification of a fractured paradigm.Journal of communication, 43(4), 51-58.

Falls, Jason. (2012). Defining engagement: is it capturing attention or more? Retrieved from https://www.socialmediaexplorer.com/content-sections/news-and-noise/defining-engage ment/.

Fahmy, S., Bock, M. A., &Wanta,W. (2014). Visual communication theory and research: A mass communication perspective. New York: Palgrave Macmillan.

Freelon, D. (2010). ReCal: Intercoder reliability calculation as a web service. International Journal of Internet Science, 5(1), 20-33.

Gao, Q., Abel, F., Houben, G., & Yu, Y. (2012). A comparative study of users’ microblogging behavior on Sina Weibo and Twitter. User Modeling, Adaptation, and Personalization: Lecture Notes in Computer Science, 7379, 88–101.

Geert Hofstede, Gert Jan Hofstede, Michael Minkov, Cultures and Organizations: Software of the Mind. Revised and Expanded 3rd Edition. New York: McGraw-Hill USA, 2010

(32)

comparison of online and offline purchase decision influences.Journal of marketing communications, 20(1-2), 103-116.

Hofstede, G. 2001. Culture’s Consequences. 2nd ed. Thousand Oaks, CA: Sage. Irani, L. (2015). The cultural work of microwork. New Media & Society, 17(5), 720-739. Kreuter, M. W., Lukwago, S. N., Bucholtz, D. C., Clark, E. M., & Sanders-Thompson, V.

(2002). Achieving cultural appropriateness in health promotion programs: Targeted and tailored approaches. Health Education and Behavior, 30, 133–146.

Luo, Z., Osborne, M., Tang, J., & Wang, T. (2013). Who will retweet me? Finding retweeters in Twitter. Proceedings of the 36th International ACM SIGIR Conference on Research and Development in Information Retrieval, Dublin, Ireland.

Lovejoy, K., Waters, R. D., & Saxton, G. D. (2012). Engaging stakeholders through Twitter: How nonprofit organizations are getting more out of 140 characters or less. Public Relations Review, 38(2), 313-318.

Mills, J., & Clark, M. S. (1982). Exchange and communal relationships. In L. Wheeler (Ed.), Review of personality and social psychology (pp. 91–127). Beverly Hills, CA: Sage.

Nagarajan, M., Purohit, H., & Sheth, A. P. (2010, May). A qualitative examination of topical tweet and retweet practices. Proceedings of the 4th International AAAI Conference on Weblogs and Social Media, Washington, DC, USA.

Oyserman, D., Coon, H. M., & Kemmelmeier, M. (2002). Rethinking Individualism and Collectivism: Evaluation of theoretical assumptions and meta-analyses. Psychological Bulletin, 128, 3–72.

(33)

Paivio, A. (1991). Images in mind. New York: Harvester Wheatsheaf.

Powell, T. E., Boomgaarden, H. G., De Swert, K., & de Vreese, C. H. (2015). A clearer picture: The contribution of visuals and text to framing effects. Journal of Communication, 65(6), 997-1017.

Singer, J. B. (2014). User-generated visibility: Secondary gatekeeping in a shared media space. New Media & Society, 16(1), 55-73.

Shweder, R. A., & LeVine, R. A. (1984). Culture theory: Essays on mind, self and emotion. Cambridge University Press.

Triandis, H. C. (1995). Individualism & collectivism. Boulder, CO: Westview Press. Triandis, H. C. (2001). Individualism and collectivism: Past, present, and future. In D.

Matsumoto (Ed.), The handbook of culture and psychology (pp. 35–50). New York, NY: Oxford University Press.

Twitter company facts, (2016, March 31). Retrieved from https://about.twitter.com/zh-hans/company

Waters, R. D., Burnett, E., Lamm, A., & Lucas, J. (2009). Engaging stakeholders through social networking: How nonprofit organizations are using Facebook. Public relations review, 35(2), 102-106.

Weibo search users in sights, (2015, December 22). Retrieved from

http://www.chinainternetwatch.com/16366/weibo-search-users-insights-2015/

Weber, P. (2014). Discussions in the comments section: Factors influencing participation and interactivity in online newspapers’ reader comments. New Media & Society, 16(6), 941-957.

(34)

York, Alex. (2015). What is social media engagement & why should I care? Retrieved from http://sproutsocial.com/insights/what-is-social-media-engagement/.

Zhao, C., & Jiang, G. (2011, May). Cultural differences on visual self-presentation through social networking site profile images. In Proceedings of the SIGCHI Conference on Human Factors in Computing Systems (pp. 1129-1132). ACM

Appendix 1

List of NPOs in Sample

Name of NPOs Twitter ID/Weibo ID Global/Local Country

UNICEF @联合国儿童基金会 Global China

UNICEF @UNICEFinJapan Global Japan

UNICEF @UNICEF_SA Global South Africa

UNICEF @UNICEF_uk Global UK

UNICEF @unicefusa Global USA

Doctors without Borders @无国界医生 Global China

Doctors without Borders @MSFJapan Global Japan

Doctors without Borders @MSF_southafrica Global South Africa

Doctors without Borders @MSF_uk Global UK

Doctors without Borders @MSF_USA Global USA

(35)

Save the Children @scjapan Global Japan

Save the Children @SaveChildrenSA Global South Africa

Save the Children @savechildrenuk Global UK

Save the Children @Savethechildren Global USA

One Foundation @壹基金 Local China

China Foundation For Poverty Alleviation @中国扶贫基金会 Local China

Wardrobe of Love @爱心衣橱 Local China

NSPCC @NSPCC Local UK

Cancer Research UK @CR_UK Local UK

BHF @TheBHF Local UK

Feeding America @FeedingAmerica Local USA

St. Jude Children’s Research Hospital @StJude Local USA

American Cancer Society @AmericanCancer Local USA

Smile Foundation "@SmileFundSA" Local South Africa

Mothers2mothers "@m2mtweets" Local South Africa

The Children’s Hospital Trust @chtrust1 Local South Africa

Teach For Japan @TeachForJapan Local Japan

Japan Association for Refugees @ja4refugees Local Japan

Shanti Volunteer Association @sva_1984 Local Japan

Appendix 2 Codebook

This codebook includes instructions for coding microblog messages including text and images, which will help researchers identify usage of different frames and answer questions regarding NPOs’ posts.

Basic information

1. Name of NPO – Type in the name of NPO 2. Twitter ID – Type in the Twitter ID

3. Country or region – Type in the country or region of the Twitter ID 4. Microblog name – Type in Twitter or Sina Weibo

5. URL – Type in the URL of the post

(36)

Because the five countries in the sample are distributed in different parts of world, thus each of them have a different time zone. Here researchers can just follow the time shown on the website.

 Date of post (MM.DD.YY)

 Time of post (time followed by a.m. or p.m.) (8 a.m. or 8.21 p.m.) 7. Type of Post – Type in the type of post by followed number.

Original post --- 1

This identifies a post originated by the organization’s author. Retweet --- 2

This identifies a post that organizations retweet or share a post. Reply --- 3

This identifies a post that the organization @replies other users.

8. Form of post – researchers will identify whether the post contains text, image or other forms. Type “yes” or “no” for each of the forms. If one post contains other forms, then stop coding (even a link has an image, it still be coded as other forms.

“1” – Yes; “0” – No

(1) Does it have text? (0/1) (2) Does it have an image? (0/1) (3) Does it have an animated gif? (0/1) (4) Does it have a video? (0/1)

(5) Does it have a link? (0/1)

9. Number of “Likes” – Type in the number of “likes” of the post 10. Number of retweets – Type in the number of retweets of the post 11. Actual text – Type in the actual text here

Message framing

12. Promotion versus Prevention

In this procedure, researchers will only code the text in the codebook. If there is a link within a post, researchers do not need to click the link, but analysis the title and abstract shown on the post surface. Researchers will identify whether the text of a tweet fits in one of the two types of messages (promotion message and prevention message). Since the model can overlap, it is okay if more than one model fits each tweet. Try to narrow it down to one model, if possible. Answer “yes” if the tweet corresponds to the type and “no” if it does not. If a post doesn’t show neither of the types, type “no” on both of them.

 Yes = 1; No = 0

(6) Promotion message (0/1)

Message is set up to promote positive outcomes, to view events framed in terms of winning being more important. To the extent that promotion message focuses on things that will be gained by engaging in a behavior. For example: 10,000 people

(37)

can be saved from starvation with our support. (7) Prevention message (0/1)

Message is set up to prevent negative outcomes, to view events framed in terms of losing as being more important. To the extent that prevention message focuses on things that are lost by not engaging in a behavior. For example: 10,000 people will die of starvation if we do not support them.

Visual framing

13. Visual framing of people

NOTE: In this process, researchers will start coding the images shown in a post. To

emphasize, for this part, researchers will only code static pictures (videos or automatic gifs are excluded), such as photographs. For multiple pictures in a single post,

researchers will analysis all the pictures shown in a single post one by one followed by the same order: from left to right and then from top to down.

13-1 Number of image – Type in the number of image in a single post. 13-2 Type of image – Researchers will firs identify what type each image is.

1. Photograph 2. Poster 3. Cartoon 4. Infographic 5. Other 2. Poster 1. Photograph

(38)

13-3 People portray – Researchers will identify whether the image contains people portray or not. Type “yes” if the image does contain people and “no” if it does not. If no, stop coding.

 Yes = 1; No = 0 Yes

No

13-4 Major characters: Researchers will identify how many main characters in the image. Type in the number of people shown in every image.

 1  2  3  4  5 or more 5 or more 1 people 4. Infographic

(39)

13-5 Levels of intimacy of space – Researchers will identify whether the post

demonstrates the following levels of intimacy of space (intimate distance, close personal distance, far personal distance, close social distance, far social distance, and public distance). Answer “yes” if the post corresponds to the level and “no” if it does not. The levels are mutually exclusive. Followed by examples.

 Yes = 1; No = 0  Intimate space (0/1)

Image shows face or head only  Close personal distance (0/1)

Image shows the head and shoulders  Far personal distance (0/1)

Image shows person from the waist up  Close social distance (0/1)

The whole figure will show up  Far social distance (0/1)

The whole figure as well as figure with space around it can be seen  Public distance (0/1)

Torso of at least four or five people can be seen in the picture

Close personal distance

Far personal distance Intimate space

(40)

Public distance Far social distance