Graduate School of Communication
Master’s Programme: Communication Science
Track: Corporate Communication
Master’s Thesis
Impact of Frames, Evidences and Emotional
Aspects of the Message on the Number of
‘Likes’, Shares and Comments of Facebook
Posts of Non-Profit Organisations
Inna Iljina
Student №: 10841784
Supervisor: Theo Araujo
University of Amsterdam
Abstract
Message frame, evidence and emotions included in the message have been broadly studied in marketing and advertising areas, but the effects of these aspects have not been examined on Facebook activity of non-profit organisations (NPOs). At the same time ‘likes’, comments and shares are good indicators of publics engagement, sympathy towards the message and desire to promote it. Therefore the knowledge of how to motivate the intentions to like, share or comments would be beneficial for NPOs. Especially in the times when they are struggling in finding volunteers and raising money (as it was proved, people who are engaged with an organisation online, also more likely join its initiatives in real life). This content analysis study assessed the effect of different message evidences, frames and emotional aspects of the message on number of ‘likes’, shares and comments in case of Facebook pages of NPOs, who aim to help refugees. It was found that messages with statistical evidence are shared more than messages with narrative or no evidence. Negatively framed messages and messages that contain negative emotions are liked and shared more than positively framed messages and messages that contain positive emotions. Besides, messages with any emotions receive more comments and likes than non-emotional messages. The findings of this study could help PR practitioner of non-profit organisations to better engage with their publics and increase promotion of their messages on Facebook.
Keywords: message frame, message evidence, emotions, face expression, Facebook,
Impact of Frames, Evidences and Emotional Aspects of the Message on the Number of ‘Likes’, Shares and Comments of Facebook Posts of Non-Profit Organisations
In 2011, 60% of non-profit organisations (NPOs) had active Twitter and Facebook accounts. NPOs mostly use these accounts to raise funds and to improve the ability to communicate with their stakeholders and engage them with their activities, such as
fundraising, signing petitions, joining events and others (Waters, 2007). However, most of the charities (organization that were set up to provide help and raise money for those in need (Oxford Dictionary)) have recently struggled to raise money through social media and to successfully engage with their publics (Flandez, 2011). Charities are also considered a type of NPOs, because they have the same objectives and stakeholders as NPOs. It is important for NPOs to be successful in their external communications on social media, because they face a lot of challenges nowadays as for example to attract and engage volunteers and donors. Every year, 10,000 new organisations are joining the non-profit sector (Hankinson, 2000).
Therefore, NPOs may find it beneficial to use social media in order to reach their target audience, to engage with it, as well as to spread their message more effectively and
efficiently. It was proven that people who engage with an organisation online, are also more likely to choose its products and support its activities in a real life (in case of NPOs: to make donations and support NPOs’ initiatives).
Regarding the engagement on social media and Facebook in particular, it was found that a ‘Like’ represent the overall positive attitude towards the message, a comment indicates the level of engagement with organisation and shares signify individual’s desire to promote the message (Saxton & Waters, 2014). When person ‘likes’, comments on or shares the organisation’s post, they publicly expresses their opinion and voluntarily shows their brand preference. In this way electronic word-of-mouth (eWOM), defined as “any positive or negative statement made by potential, actual or former customer about a product or company, which was made available to a multitude of people via the Internet” (Hennig-Thurau et al., 2004, p. 39) occurs. Individual’s preferences are also automatically shown in person’s friends’ newsfeeds, which makes the organisation noticed by more social media users (including actual and potential external stakeholders of the organisation) as well as message to be spread further (Coulter & Roggeveen, 2012).
Both public engagement with the message and its dissemination are crucial for NPOs, because they increase public involvement and spread messages. Therefore, it is useful to
understand how to increase the effectiveness of social media messages of NPOs in terms of receiving ‘likes’, shares and comments. Broadly used by NPOs in their external
communications message aspects, such as frame (positive (gain) frame of making actions or negative (loss) frame of not acting) and message evidence (statistical (presenting many cases) or on narrative (presenting one case)) have been widely studied (mostly in experimental studies) in advertising and marketing research in terms of effects they have on persuasion and behavioral intentions (Rothman et al., 2006; Allen & Preis, 2009). Several studies have also examined emotional aspect of the message: the presence of either positive or negative emotions, or any emotion; portrayal of the person in terms of the emotion of his/her face expression (Dobele et al., 2011; Lin et al., 2006; Small & Verrochi, 2009). However, the aforementioned message aspects were not studied on Facebook context of non-profit organisations, even though it was multiple times shown that these aspects indeed influence receiver’s attitudes towards the message. Social media’s environment differs from advertising environment, as well as goals of communication of non-profit organisations are different from for-profit organisations. Therefore the current study examined the effect of the message aspects on the number of Facebook ‘likes’, shares and comments in case of non-profit organisations.
However, this research was conducted not on Facebook posts of any NPOs, but specifically on NPOs, that aim to help refugees. Refugees were one of the most important issues in 2015. In spring 2015, the number of refugees coming to Europe increased
dramatically. NPOs started actively help refugees and increased awareness about this problem and therefore they produced a decent amount of refugees-concerned posts on Facebook. Then, at the beginning of autumn 2015, Europe experienced a second wave of refugee arrival, when thousands of refugees were crossing the Mediterranean Sea and were also dying on their way to Europe. In December 2015, the situation concerning the refugees was still actual and NPOs were actively using Facebook. During the times of problematic economic situation and
different attitudes towards this issue in the society, it is crucial for NPOs, whose aim is to help refugees in particularly, to receive as much support and promotion from general public as possible. Therefore the research question is:
RQ: How do different frames, evidences and emotional aspects of the message influence the number of ‘likes’, shares and comments that Facebook posts receive in case of non-profit organisations who aim to help refugees?
Theoretical Background
The aspects of the message examined in the current research were widely studied in marketing or advertising research, as well as in the charity research area. The main and the most relevant for the current research findings will be presented and discussed further, followed by the formulation of the hypotheses.
Message Evidence
One of the strategies often employed in the communication of non-profit organisations is an inclusion of message evidence. Message evidence, defined as “a supporting material (or a proof), that asks the message receiver to accept the conclusions of the communicator” (Allen & Preis, 1997, p. 125), is one of the strategies to affect the perceived value of a particular problem (Reynolds & Reynolds, 2002). Through message evidence the charity organisation may demonstrate the importance and relevance of its goal (Das et al., 2008).
Message evidence can be presented either by giving statistical information or
describing one narrative case (Das et al., 2008). In other words, the communicator can either include one case story or statistical summary of large number of cases in the message to demonstrate that information or conclusion presented in the message is valid and objective (Allen & Preis, 2009). For example, the message “Some more positive news to end the week - Mahin, one of our Big Night In refugee chefs, just got her first full-time permanent job since being forced to flee. So pleased for her.” could be considered as a message with narrative evidence, while message “20 refugee families from Syria got their new homes in England
thanks for Your support.” could be considered as a message with statistical evidence.
Research on NPOs with charitable focus indicates that message evidence enhances the effectiveness of persuasive messages (Morman, 2000, Reynolds & Reynolds, 2002), because evidence could make potential donors more aware of the situation (Morgan & Miller, 2002). However, it is not clear which type of message evidence is more persuasive. One group of studies argue that providing narrative case stories is more effective than abstract, statistical information, as narrative case stories induce stronger mental imagery, possess stronger intuitive appeal and reduce counterarguments (e.g., Green, 2006; Green & Brock, 2000; Green et al., 2004; Reinard, 1988; Rook, 1987). Other researchers on the contrary, came to opposite conclusions: that statistical evidence is more persuasive than the use of narratives (Allen & Preis, 1997; Baesler & Burgoon, 1994). Allen & Preis (1997) explained it with an
assumption that, when large number of cases is used, it brings a sense of objectivity in the analysis of the text.
Within the research area regarding NPOs with charitable intentions, Kopfman et al. (1998) examined the effects of narrative versus statistical evidence on cognitive and
emotional reactions to organ donation messages. The results showed that cognitive reactions are affected more strongly by statistical evidence, whereas emotional reactions are affected more strongly by narrative evidence. Applying these findings to our research, it should be first mentioned that cognitive reactions mean intentions or wish to do something (an action). Social media activities such as sharing and commenting involve actions, because comments represent people’s engagement with organization and shares show that people promote or support the message (Saxton & Waters, 2014). Both engagement and promotion are active actions, therefore commenting and sharing can be considered as cognitive reactions. It was found that cognitive reactions in their turn are affected more by statistical evidence.
Therefore, it can be argued, that statistical evidence of the message will lead to more shares and more comments than narrative evidence of the message. This leads us to propose the following hypotheses:
H1: Posts that contain statistical evidence will have more shares than posts that contain narrative evidence or no evidence.1
H2: Posts that contain statistical evidence will have more comments than posts that contain narrative evidence or no evidence.
On the contrary, ‘Like’ can be considered as a rather emotional than cognitive reaction, as it represents attitude, feeling of sympathy (Saxton & Waters, 2014) and requires less action. As emotional reactions are affected more strongly by narrative evidence, it can be stated that messages that contain narrative evidence will lead to more ‘likes’ than messages with statistical evidence. Therefore the following hypothesis is proposed:
H3: Posts that contain narrative evidence will have more ‘likes’ than posts that contain statistical evidence or no evidence.
Message Frame
Message can be presented using various frames. Framing itself can be defined in numerous ways, but in the context of a present research message framing refers to “the
presentation of one of two equivalent value outcomes to different decision makers, where one outcome is presented in positive or gain terms, and the other is presented in negative or loss terms” (Das et al., 2008, p. 163). Labeling a glass of water ‘half empty’ or ‘half full’
represents an illustration of message framing, because each label presents only one side of the information (Martin, 1995).
A message can be framed in either negative or positive terms, otherwise known as loss versus gain frames (Detweiler et al., 1999; Rimer & Kreuter, 2006). When using a gain-frame, the benefits of taking actions are highlighted (Rothman et al., 2006). For example, the message “Thanks to your incredible support over the last few months, we've been able to launch a brand new project. It's going to help around 300 asylum seekers and refugees facing imminent hunger.” is considered as positively framed, while when applying a loss-frame the consequences of not-taking actions are emphasized (Rothman et al., 2006). For example, “If you won’t make a donation, the needs of families won’t be met and their already not brilliant condition will worsen”.
Previous findings regarding the effectiveness of positive and negative frames were mixed. Some studies suggested that positive frames are more persuasive than negative ones (Levin & Gaeth, 1988), while other studies stated the opposite (Ahluwalia et al., 2000; Davis, 1995; Herr et al., 1991). The study of Ditto and Lopez (1992) demonstrated that negatively framed information better grabs attention in general and receives more scrutiny than
positively framed information. At the same time, Baker and Petty (1994) found that negative framing has bigger impact in a decision-making process than positive framing. Facebook sharing is considered a rather high engagement activity (Cho et al., 2014), because before sharing a person really needs to make a decision to spread further the post. Taking into account these facts and what Baker and Patty (1994) have found, it can be stated, that negatively framed messages will be shared more than positively framed messages.
H4: Negatively framed posts will receive more shares than posts with positive or neutral frames.
Das et al. (2008) found that in the narrative evidence condition charity’s work is perceived more relevant when the message is framed positively rather than negatively. The relation is reversed in the statistical evidence conditions. It can be assumed that, when charity’s work is perceived as more relevant, public also have more positive attitude towards it. In Facebook context, ‘likes’ represent public’s overall positive attitude towards the post
(Saxton & Waters, 2014). Therefore, it can be argued that positively framed messages could lead to more ‘likes’ than negatively framed messages, when narrative evidence is used. The effect could be opposite in the statistical evidence conditions. Thus, the following hypotheses derive:
H5: Positively framed posts will lead to more ‘likes’ than negatively framed posts, when narrative evidence is used in the posts.
H6: Negatively framed posts will lead to more ‘likes’ than positively framed posts, when statistical evidence is used in the posts.
Emotional Messages
Several studies have demonstrated that general emotional appeals are effective in persuasion. Forgas (2006) stated that emotions “appear to influence what we notice, what we learn, what we remember, and ultimately the kinds of judgments and decisions we make” (p. 273). Previous studies found that in written communication emotional stimuli (expressed in emotional words or emotional framing) of messages may provoke extensive cognitive processes (Bayer et al., 2012; Kissler et al., 2007; Smith & Petty, 1996). In its turn, an increased level of cognitive involvement may lead to a higher possibility of behavioral responses to emotional stimuli in terms of information sharing (Heath, 1996; Peters et al., 2009; Rime, 2009). Assuming that commenting and sharing are considered to be cognitive processes, it can be argued, that emotional messages 3 will receive more shares and more comments than non-emotional messages. This leads us to the following hypotheses
H7: Emotional posts will have more shares than non-emotional posts.
H8: Emotional posts will have more comments than non-emotional posts.
Types of Emotions
Regarding the types of emotions expressed in the message, it should be first mentioned that message can contain either negative or positive emotions. Messages with negative
emotions consist from questions, statements or words which aim to trigger fear, sadness, frustration, worriedness, miserableness or angriness. Messages with positive emotions contain questions, statements or words which aim to trigger happiness, cheerfulness, gladness,
The results of the study of Lin et al. (2006) showed that types of emotions in online content influence information sharing intentions (in the case of forwarding emails). When the content of the email made people feel positive emotions, it was forwarded more often. The authors explained their findings by social psychology theory. It suggests that when people share message that makes them to feel emotions, people allow others to feel similar emotions (Seta et al., 1994). The study of Bergman and Milkman (2012) on Times online articles also found that positive content is more viral than negative (it was shared more often). However, Bergman and Milkman (2012) also stated that the wish to share content also depends on the level of content’s arousal: content that evokes high-arousal positive (awe) or negative (anger or anxiety) emotions is shared more than content that evokes low-arousal, or deactivating, emotions (for example, sadness).
It can be stated that the motivation behind forwarding emails might be similar to social media’s sharing, as both activities include passing the message further to online friends. Besides, it can be assumed that Facebook post that contains positive emotion will also make people feel rather positive than the post with negative emotions Considering the findings of two studies discussed above, it can be argued, that posts that contain positive emotions will motivate people to share them more than the posts that contain negative emotions. Therefore the following hypothesis derives:
H9: Posts that contain positive emotions will have more shares than posts that contain negative emotions.
Face Expression
Small and Verrochi (2009) examined the effect of facial emotional expression in charity advertisements. One of the results showed that sad face expression increased donations, comparing to happy or neutral face expressions. Sharing was awarded by high levels of engagement (Cho et al., 2014) and it can be assumed that donating also involves a rather high level of engagement. Therefore it can be argued that the number of shares will increase when the person on the picture of the post has a sad face expression, rather than the when with neutral or happy face expressions. However, another two studies (Bergman & Milkman, (2012) and Lin et al. (2006)) found that positive content is more viral and that content that makes people feel positive emotions tends to be forwarded more often. Considering that sad face expression could lead to sad rather than happy feelings, our first argument contradicts with these findings. It forces us to propose two competing hypotheses:
H10a: Posts with photos of people (refugees)4 with sad face expression will have more shares than posts with photos of people (refugees) with happy or neutral face expressions.
H10b: Posts with photos of people (refugees)4 with happy or neutral face expression will have more shares than posts with photos of people (refugees) with sad face expressions.
The study of Small and Verrochi (2009) also showed that a sad face of the person at advert leads to the expression of sympathy. In both photos (the ones attached to the Facebook posts and photos of the adverts), the same type of people (people for whom charities are trying to get support) are depicted. Besides, ‘likes’ mean sympathy of the person. Therefore it can be argued that sad emotional expression of the refugee will lead to more ‘likes’ than happy or neutral face expressions.
H11: Posts with photos of people (refugees)4 with sad face expression will have more ‘likes’ than posts with photos of people (refugees) with happy or neutral face expressions.
In order to avoid contradictions with the previous hypotheses, it is important to mention, that ‘likes’ and shares are different actions/re-action: one is cognitive, another – emotional reaction.
Non-Profit Organisations and Social Media
Further it will be covered what was researched in the area of non-profit organisations regarding their use of social media and Facebook in particular.
Only few studies were conducted regarding NPOs’ or charities’ activity on social media. Lovejoy’s and Saxton (2012) provided a classification of the main types of messages that non-profit organisations use in their updates on Twitter. It was concluded that NPOs use Twitter for three major functions: ‘Information’, ‘Community-building’, and ‘Action’. Wherein ‘Information’ messages consist from information about the organisation, its activities or anything of its fan page’s interest; ‘Community-building’ messages foster relationships, create networks, and build communities, promoting interactivity and dialogue. Finally ‘Action’ messages are aimed at getting followers and fans to ‘do something’ for the organisation (donate, attend an event, join a movement, share the statement ect.).
Regarding the effects of Facebook messages specifically, Cho et al., (2014) examined the number of ‘likes’, shares and comments for the four different message types, divided by four models of public relation: public information model, two-way asymmetry, two-way
symmetry and press-agency model. Different amount of comments were found for the models. People were more likely to comment on the messages of two-way symmetric model. On top of that, different levels of engagement were prescribed to different users actions. Commenting activity received the highest, sharing activity –medium, and liking activity – the lowest level of engagement.
Saxton & Waters (2014) examined users’ reaction on the message types by their functions. Namely they looked at the volume of ‘likes’, shares and comments that
Informational, Promotional (including Call-to-action messages), and Community-building messages of charities receive on Facebook. It was found that people tend to like and comment more on Call-to-action messages and Community-building messages than on Informational messages. However, individuals were more likely to share Informational messages than Call-to-action and Community-building messages (ibid).
Even though two of the studies mentioned above evaluated the effect of different types of messages on the number of ‘likes’, shares, comments, knowing the effect of only messages different by function and two or one-way communication is not sufficient in predicting a success of people’s engagement with the message. Besides, considering that different message types receive different numbers of ‘likes’, shares, comments (Cho et al., 2014; Saxton & Waters, 2014), it is reasonable to examine each group of message function regarding the effect of the aspects of the message on the numbers of ‘likes’, shares and comments. However, in the present research message classification of Lovejoy and Saxton (2012) will be used, because all social media messages of NPOs can be broadly defined by one of these three categories, and they are used by NPOs in the real life. Therefore a sub-research question is:
Sub-RQ: How do different frames, evidences and emotional aspects of the message influence the number of ‘likes’, shares and comments that Facebook posts receive in case of non-profit organisations who aim to help refugees, regarding each group of charities’
Facebook messages: ‘Information’, ‘Community-building’ and ‘Action’ messages?
Methods
Sample
Sample was composed from Facebook posts of the following organisations: Refugee
selected for several reasons. Firstly, the selected non-profit organisations are one of the biggest and top NPOs (that deal with refugees), and that therefore have rather big amount of followers and fans on Facebook. Secondly, the selected organisations also produce rather large amount of social media content, which is required for the proper analysis. Besides, similar studies on the Facebook use of charities/NPOs selected top 100 charities for their analysis. However, due to the time limitations of the research and the research scope, this study is limited by top three refugees’ NPOs of UK. Thirdly, organisations from the same country were selected in order to avoid the influence of the differences in the countries on the behavior responses to the messages. Finally, mentioned above organizations were selected because of the language of their Facebook posts (English), as for the correct analysis a totally clear understanding of the posts is required.
Research Design
Content analysis of Facebook pages of the organisations listed above was conducted. Each sampled/coded post was measured by message frame, message evidence, emotions presented in the message, face expression of refugees on the photos of the message, the number of ‘likes’, shares, comments and the type of the function(s) of the message, as well as for the existence of embedded photo, video and link. It should be noted, that only posts about any refugees or refugee’s related information of each of the sampled organisations were coded. Posts from 1st March 2015 to 2nd December were sampled. This time frame was
selected, because it covers the main peaks of the refugee issue (all ‘waves’ of refugees’ arrival to EU shores). There was no period during the time frame when there were no posts
published. Approximately same amount if posts were posted by the organisations each month. At the beginning of the coding, open coding was applied in order to test the codebook. Intracoder reliability was conducted for all independent variables and resulted in rather high Krippendorff's Alpha values for most of the variables (see table 1). Ten percent of the sample was re-coded by the same coder after four weeks from the first coding. Then the results were put in the same Microsoft Excel document, which was later uploaded to a special website (http://dfreelon.org/utils/recalfront/recal-oir/), which calculated Krippendorff’s Alphas for
each variable. Intracoder reliability was not conducted for the dependent variables, as they were counted in numbers, therefore their variables were considered as straight-to-define and therefore already reliable.
Table 1
Krippendorff’s Alphas’ Values for the Main Independent Variables
Variable Krippendorff’s Alpha Percent Agreement (%) Positive frame 0.9 95.5 Negative frame 0.9 95.5 Statistical evidence 0.76 81 Narrative evidence 0.94 97.7 Positive emotions 0.81 90 Negative emotions 1 100 Dominant function 0.96 97.7 Face expression 1 100 Measures Independent Variables
Message evidence. Message evidence was operationalized by looking at what type of
example (case story of one refugee/refugee family or statistical information/summary information across a larger number of cases) was provided in the message.
Message frame. Message frame was operationalized by looking at what frame (loss or gain)
was used/emphasized in the message. When the message contained information about refugees that emphasizes the benefits of taking action or positive consequences of taking action (for example, donating money), positive frame was assigned. When the message mentioned and emphasized the costs of failing to take actions, or was focused on the negative consequences of not taking actions (for example, not supporting refugees), negative frame was assigned (Das et al, 2008; Rothman et al, 2006).
It should be noted, that when the message contained neither positive nor negative frames, no frame was assigned; when both frames were presented in the message, the frame which occupied more space in the message was assigned. The same procedures were applied when assigning message evidence.
Emotions included in a message. Message was defined as containing negative emotions,
of the following emotions: fear, sadness, frustration, worriedness, miserableness, angriness. Message was defined as containing positive emotions, when it consisted from questions, statements or words which aimed to trigger happiness, cheerfulness, gladness, warmness, vigorousness or enthusiasm (Lin et al., 2006). Later in the analysis these two variables were used to form variable ‘emotions presented’. Messages that contained either negative or positive emotions were defined as ‘emotional’ messages, while messages which contained neither positive nor negative emotions were labeled as ‘non-emotional’ messages.
Face expression. Refugees’ face expression was measured by looking at their face expression
and emotional activity (on the photo attached to the post). When the refugee on the photo was smiling, laughing or looked happy in general, happy face expression was assigned. When the refugee on the photo was crying, looked lost, shocked, vulnerable or sad, sad face expression was determined. When refugee’s face did not express either positive or negative emotions or feelings, neutral face expression was assigned (Small & Verrochi, 2009).
Message function. Classification of the messages regarding their function was taken from
Lovejoy’s and Saxton’s (2012) study. ‘Information’ function was assigned to the message if the message contained information about the organization, its activities or anything of
potential interest to fans and followers of organization’s page. ‘Community-building’ function was assigned to the message, if message fostered relationships, created networks or promoted interactivity and dialogue (‘bonding’ messages, such as ‘thank you’ and acknowledgement messages were also indicators of this function). ‘Action’ function was assigned to the
message if it was aiming at getting followers and fans to ‘do something’ for the organization, whether to donate, buy a product, attend an event, join a movement, launch a protest, or share link or message.
Dependent Variables
To measure public reactions to different types of organisation’s messages, three measures were employed: the number of ‘likes’, the number of comments, and the number of shares associated with each Facebook post.
Control Variables
Message function. The research of Saxton and Waters (2014) showed that regarding
charities’ Facebook posts people tend to ‘like’ and comment more on Call-to-action messages and Community-building messages than on Informational messages. However, individuals are more likely to share Informational messages than Call-to-action and Community-building messages. Therefore it was decided to control for message types regarding their functions. Classification of the messages regarding their function and how they were defined were taken again from Lovejoy’s and Saxton’s (2012) study (described earlier).
Photo/video/link. It was proved that information richness theory has impact on information
dissemination (Ngwenyama & Lee, 1997; Daft & Lengel, 1986; Markus, 1994). Information richness indicates that receivers understand rich information more quickly than lean
information, and that the medium, that consists from more expressions, gestures and tones, is more powerful (Daft & Lengel, 1986). Based on information richness theory, it can be argued that posts with video or photo may increase Facebook users sharing, ‘liking’ or commenting intentions (because of their richer content). The same can be stated regarding the existence of link in the post. Therefore each post was controlled for the existence of link, photo and video.
Media attention. It was decided to control for media attention, because refugee theme is a
sensitive issue, and audience’s attention for in on social media (and Facebook in particularly) may vary depending on the amount of the attention to this issue and discussions about
refugees in the traditional media.
Media attention was operationalized by the number of articles published about
refugees (where refugees were mentioned) in four most readable UK newspapers: Daily Mail;
The Guardian, Daily Mirror and Daily Telegraph (Hollander, 2013) for the each month of the
sample’s time frame period (1st
March -2nd December 2015).
Fan ‘likes’. Facebook pages of the sampled organisations had different number of fan ‘likes’.
It could have affected the volume of ‘likes’, shares and comments. Therefore it was decided to control for the organisation by including the number of fan ‘likes’ of the organisation’s
Results
In total 454 posts (units of analysis) were used for the statistical tests. The percentages and exact numbers of the posts by each message aspects can be found in table 2.
Table 2
Numbers of Posts by Message Aspects
Message Aspect Type of the Aspect % N
Frame Positive 33.7 153 Negative 25.3 115 No 41 186 Evidence Statistical 17.2 78 Narrative 28.2 128 No 54.6 284
Emotions presented Yes* 49.2 223
No 50.8 230
*Type of emotion Positive 60.1 134 Negative 39.9 89
Regarding the face expression, only a third (33.3 %) of all posts contained photos where refugee was depicted and where it was possible to define their emotion. Among them 40.3 % (62 posts) contained photos of refugees with happy face expression, 31.2 percent (48 posts) - photos of refugees with sad face expression and 28.6 percent (44 posts) - refugees with neutral face expression.
Regarding the division of the messages by dominant function, most of the messages (71.1 % or 323 posts) were with ‘Information’ as a dominant function, almost quarter of the messages (19.8 % or 90 posts) had ‘Action’ function as a dominant function, and only nine percent (41posts) of the messages were ‘Community-building’ messages.
To answer the main research question various one way Multivariate Analysis of Covariance (MANCOVA) were conducted, where numbers of ‘likes’, shares and comments were dependent variables, media attention, existence of link, photo, video, number of ‘likes’ on the organisations’ Facebook page, and message functions (‘Information’, ‘Community-
building’ and ‘Action’ separately) were control variables. To answer the sub-question split-file option was used, so the results for each category (message functions) were displayed. It is important to note, that in order to run MANCOVAS variables ‘Likes’, Shares and Comments were log-transformed, because MANOVA requires all dependent variables to be normally distributed (and dependent variables were not normally distributed at the beginning). When running statistical tests for sub-question, sample was divided by groups using dominant function of the message, because almost each message contained features of several message functions.
There was a statistically significant difference in the number of
‘likes’/shares/comments based on message evidence, F (6,870)= 4.59, p < .0005, Wilk’s Λ = 0.94, partial η2
= .031.Message evidence had a statistically significant small effect on the number of shares F (2,437) = 6.70, p = .001, partial η2 = .03 and statistically significant even smaller effect on ‘likes’ F (2,437) = 4.34, p = .013, partial η2
= .02. No statistically significant effect of message evidence was found on the number of comments: F (2,437) = 2.15, p = .117, partial η2
= .01. Therefore second hypothesis that posts that contain statistical evidence will have more comments than posts that contain narrative evidence or no evidence was rejected. Post-hoc test indicated the significant difference between statistical and narrative evidence messages (Mdifference = .34, p = .001) and between messages with no evidence and
messages with statistical evidence (Mdifference = -.25, p = .011) regarding shares. Messages
with statistical evidence got more shares (M = 1.09, SD = 0.74) than messages with narrative evidence (M = 0.76, SD = 0.77) and than messages with no evidence (M = 0.90, SD = 0.74). Therefore first hypothesis that posts that contain statistical evidence will have more shares than posts that contain narrative evidence or no evidence was supported.
Post-hoc test also showed that there was a very small significant difference between no evidence and statistical evidence (Mdifference = -.14, p = .025) regarding ‘likes’, however
there was no statistically significant difference between narrative evidence and statistical evidence. Therefore the third hypothesis that posts that contain narrative evidence will have more ‘likes’ than posts that contain statistical evidence or no evidence was not supported.
It should be noted that the assumption of equal variances in the population was not violated for ‘likes’ (Levene’s F (2,445) = 2.64, p = .073), but was violated for shares.
However, the significant variance for shares was still in the population, because the variance ratio for comments’ data was less than two (Field, 2012, p. 6).2
There was no statistically significant difference in the number of
‘likes’/shares/comments based on message frame’s and evidence’s interaction effect, F (2,425) = 1.23, p = .293, Wilk’s Λ = .99 partial η2 = .006. Therefore both hypothesis five and six (about the differences in the numbers of ‘likes’ regarding different frames in different evidence conditions) were rejected. However, there was a statistically significant difference in the number of ‘likes’/shares/comments based only on message frame, F (6,870) = 11.9, p < .0005, Wilk’s Λ = 0.85, partial η2
= .076.
Message frame had a statistically significant small effect on the number of shares F (2,437) = 10.02, p < .0005, partial η2 = .044, statistically significant medium effect on the number of ‘likes’ F (2,437) = 18.39, p < .0005, partial η2
= .078, and also significant small effect on the number of comments: F (2,437) = 4.05, p = .018, partial η2 = .018. Post-hoc test, however indicated the significant difference only between the messages with no frame and negatively framed messages (Mdifference = - .36, p < .0005) and between positively frames and
negatively framed messages (Mdifference = - .28, p = .002) regarding shares. Negatively framed
messages (M = 1.24, SD = 0.67) were shared more than positively framed messages (M = 0.81, SD = 0.72). Thus the fourth hypothesis that negatively framed posts will receive more shares than positively framed posts and post with neutral frame was confirmed.
Significant difference was also found for the number of ‘likes’ between messages with no frame and negatively framed messages (Mdifference = .18, p = .001), as well as between
messages with no frame and positively framed messages (Mdifference = -.26, p < .0005).
Post-hoc test also revealed significant difference between messages without any frame and
negatively framed ones (Mdifference = .12, p = .031) regarding the number of comments message
receive. The values of means and standard deviation of different message frames can be found in Table 3.
Table 3
Means and Standard Deviations for Different Message Frames Regarding ‘Likes’, Shares, Comments Dependent Variable Frame M SD N Likes Lg10 No 1.38 0.59 184 Negative 1.61 0.44 113 Positive 1.63 0.56 151
Shares Lg10 No 0.74 0.77 184 Negative 1.24 0.67 113 Positive 0.81 0.72 151 Comments Lg10 No 0.22 0.37 184 Negative 0.37 0.38 113
It should be noted that the assumption of equal variances in the population was not violated for comments (Levene’s F (2,445) = 1.60, p = .203), but was violated for ‘likes’ (Levene’s F (2,445) = 7.08, p = .001) and shares (Levene’s F (2.445) = 4.48, p = .012). However, as the variance ratio for ‘likes’ and shares data was less than two, there still were equal variances in the population (Field, 2012, p. 6)2.
There was a statistically significant difference in the number of
‘likes’/shares/comments based on the presence of the emotions in the message, F (3,435) = 3.41, p = .018, Wilk’s Λ = 0.98, partial η2 = .023. Emotional presence had a statistically significant weak effect on the number of ‘likes’ F (1,437) = 9.20, p = .003, partial η2 = .021 and weak statistically significant effect on the number of comments F (1,437) = 5.15, p = .024, partial η2
= .01. No statistically significant effect of message evidence was found on the number of shares: F (1,437) = 3.01, p = .083, partial η2 = .007. Therefore the hypothesis seven (about emotional and non-emotional posts regarding the number of shares) was not supported. Post-hoc test indicated the significant difference between emotional and non-emotional messages (Mdifference = .12, p = .003) regarding ‘likes’, and regarding comments (Mdifference =
.08, p = .024). Emotional messages received more ‘likes’ (M = 1.69, SD = 0.47) than non-emotional messages (M = 1.35, SD = 0.59), at the same time non-emotional messages received more comments (M = 0.36, SD = 0.40) than non-emotional messages (M = 0.22, SD = 0.35). Thus hypothesis eight that emotional posts will have more comments than non-emotional posts was supported.
It should be noted that the assumption of equal variances in the population was not violated for ‘likes’ (Levene’s F (1,455) = 0.23, p = .63) and shares (Levene’s F (1,445) = 1.39, p= .24), but was violated for comments (Levene’s F (1,445) = 15.77, p < .0005). However, the variance ratio for comments data was less than two, indicating that there was equal variance in the population (Field, 2012, p. 6)2.
There was a statistically significant difference in numbers of ‘likes’/shares/comments based on the types of emotions included in the message, F (3,211) = 16.07, p = <.0005, Wilk’s Λ = 0.807, partial η2
= .193.Types of emotions included in the message had a statistically significant medium effect on the number of shares F (1,212) = 17.79, p < .0005, partial η2
= .071, but no statistically significant effect on number of ‘likes’: F (1,212) = 2.63,
p= .106, partial η2 = .012 and comments F (1,212) = 2.24, p = .136, partial η2 = .010. Post-hoc test indicated the significant difference between messages that contain positive emotions and messages that contain negative emotions (Mdifference = -.43, p < .0005) for the number of
shares. Messages that contain positive emotions (M = 0.83, SD = 0.76) were shared less than messages that contain negative emotions (M = 1.32, SD = 0.61). Therefore hypothesis nine that posts that contain positive emotions will have more shares than posts that contain negative emotions was rejected, as the opposite result was obtained.
It should be noted that the assumption of equal variances in the population were not violated for comments (Levene’s F (1,220) = 1.5, p = .220), but were violated for shares (Levene’s F (1,220) = 2.9, p = <.0005) and for ‘likes’ (Levene’s F (1,220) = 3.9, p = .048). However, the variance ratio for shares and ‘likes’ data was less than two, indicating that there were equal variances in the population (Field, 2012, p. 6)2.
There was no statistically significant difference in numbers of ‘likes’/shares/comments based on the face expression of the refugees on the photos of the posts, F (9,1056) = 1.24, p = .27, Wilk’s Λ = 0.97, partial η2
= .008. Therefore the hypotheses concerning the face expression of the refugees depicted in the photos of the posts (H10 and H11) were not supported.
Results for ‘Information’, ‘Community-Building’ and ‘Action’ Function Messages
Further only the main findings regarding the message groups divided by the dominant functions will be presented. The detailed results with all statistical values can be found in the appendix.
Results alike all the hypotheses of the main research question were found for the messages with ‘Information’ as a dominant function. Only regarding the effect of the message evidence, there was no difference in the number of ‘likes’ for different message evidence, while there was a difference when examining all the messages together. Interesting, that the group analysis also detected the difference in ‘likes’ depending on the type of the emotions
used in the message for messages with ‘Information’ as a dominant function, while it was not found when analysing all the messages together. However, the p-value for this difference was marginal (critical) (p = .048) and the difference between means was very small (Mdifference =
.14). Therefore, it can be argued that this finding (difference) cannot be considered as fully significant and this difference can be omitted.
Regarding the results for the messages with ‘Action’ as a dominant function, the significant difference was found only for the effect of message frame. For both, number of shares and number of ‘likes’, the same differences were detected as when analysing all messages together. Only for messages with ‘Community-building’ function as a dominant no statistically significant differences in the number of ‘likes’/comments/shares based on any independent variable (message aspect) were found. The absence of statistically significant findings for both groups (messages with ‘Action’ and messages with ‘Community-building’ as dominant functions) could be potentially explained by rather small amount of cases (91 and 40 cases), comparing with the number of cases for ‘Information’ messages (318 cases). Therefore it cannot be stated that message aspects have no effect on the number of ‘likes’/shares/comments for the messages with ‘Action’ or ‘Community-building’ as dominant functions.
Conclusion and Discussion
The aim of this research was to investigate the effect of different types of frame, evidence, emotions of the message (and face expression of the refugees on the post’s photo) on the number of ‘likes’, shares and comments of the Facebook posts of non-profit
organisations, focused to help refugees. Besides, the same effects were examined separately for the messages with three types of dominant functions. The analysis revealed several important results, which are discussed below.
One of the key findings of this study is that messages with statistical evidence are shared more than messages with narrative and no evidences. This finding is in line with the results of Kopfman’s et al. (1998) study on the evidence’s influence on the cognitive and emotional reactions towards organ donation messages. In this study was found that statistical evidence has more strong effect on cognitive intentions. Assuming that sharing activity is a cognitive reaction, it can be stated that cognitive responses are indeed better affected by messages with statistical evidence (thus the current study supports this assumptions). The
current study also extended the theory, as it showed that statistical evidence motivates not only donation, but also sharing intentions.
However, no statistical difference was found between the messages with statistical and narrative evidences regarding the number of ‘likes’, even though the messages with statistical evidence received a bit more likes than the messages with no evidence. These finding
potentially contradicts with the results of the same Kopfman’s et al. (1998) study, as it
showed that emotional reactions (‘likes’, in this case) are affected more by statistical evidence (than messages with no evidence) and no difference in effect was revealed for narrative or statistical evidence. The study of Kopfman et al. stated the opposite.
Another key finding of this study, that negatively framed messages are shared and liked more than positively framed messages or messages without any frame, supports the findings of Baker and Petty (1994) that negative framing has bigger impact in decision-making process. Our research also contributes to those studies that argue that negative frames are more persuasive than positive ones (Ahluwalia et al., 2000; Davis, 1995; Herr et al., 1991). Besides, our findings extend the theory about message framing: for the Facebook activity of NPOs, negatively framed messages are more effective in achieving engagement, promotion of the message and overall positive attitude towards it.
However, our results do not correspond to the previous results of the effect of frames in different evidence condition (Das et al. 2008), as for Facebook posts of NPOs no difference between message frames in different evidence conditions was found. Therefore it can be assumed that message frame and message evidence are not connected in case of ‘liking’, commenting and sharing intentions for NPOs, even though they were connecting regarding the relevance of the charity’s work (Das et al., 2008). From this it can be concluded that people’s perception of the relevance of NPOs’ work and their intentions to ‘like’ NPOs’ messages are two different entities. Even if people consider NPOs’ work relevant when appropriate message frame is used in the appropriate evidence condition, it does not mean they will ‘like’ or share NPOs’ message.
The main findings about emotions showed that emotional messages (the ones that contain any type of emotions, both positive and negative), get more ‘likes’ and more
comments than non-emotional messages. While when comparing emotion types, it was found that the messages that contain negative emotions are shared more than the messages with positive emotions. These findings correspond to the previous conclusions about the
connection of emotional stimuli to cognitive processes. Namely, they support the notion that the messages with any emotions indeed received more ‘likes’ and comments.
However, the finding about shares and types of emotions contradicts with the previous research. Namely, it contradicts with the study of Lin et al., (2006), that found that people are more willing to share the information that makes them feel positive emotions, and with Bergman’s and Milkman’s (2012) study that stated that positive content is more viral. The potential explanation can be the following. People (social media users) like to share negative news/messages, because they possibly want to attract attention to these news, which are seen more as problems, because they contain something negative (negative information or negative emotions).Our finding thought corresponds to (and can be explained through) another result from Bergman and Milkman (2012). Namely, they found that content that evokes high arousal emotions (both positive and negative) is shared more (ibid.). In our case, it could be that among the messages with negative emotions there were also ones that provoke anger or anxiety. These emotions are considered as high-arousal emotions, therefore the messages that contained them were also shared more.
Perhaps for the same reason negatively framed messages are liked and shared more. People are spreading negatively framed messages further, as they already follow NPO and they possibly want to support the NPO through helping it to raise awareness about something bad that could happen, if not enough support will be received. This information evokes high arousal emotions (such as anger and anxiety) in the readers.
Despite the expectations and the other study’s results of the effect of different facial emotional expressions on sympathy in charity advertisements (Small & Verrochi, 2009), no significant difference was found for the different face expression of refugees on the volume of ‘likes’ or shares. Thus, our study contradicts with the existed research area and also extends it. It showed that not all aspects that have impact in advertising still have it in social media scene (even for the same type of organisations). These results can be potentially explained by the assumption, that when the message is presented in the post, but not the photo occupies most of the space (as in the print advertisements), the text of the message is more important to the reader than the photo attached to it. However, it is also possible that no significant results were found because of rather small number of cases. Therefore the assumptions regarding the facial expression of the people on the photos attached to the posts cannot be taken for granted,
until the effect of face expression on number of ‘likes’, shares and comments was not tested on the bigger number of cases.
The present study was also aiming to shed a light on the effects of the message aspects on the Facebook post types, most widely used by non-profit organisations. However, the same results, as when analysing all posts together, were found regarding the messages with
‘Information’ function as a dominant. The results were not significant for the most message aspects regarding posts with ‘Action’ as a dominant function and no significant results were found for the messages with ‘Community-building’ function as a dominant. The results for last two groups might not be valid because of rather small volume of cases. Therefore it cannot be argued yet that post’s dominant function influence or change effects of the message aspects.
Managerial Implications
The study has indicated that different aspects of the message influence different dimensions of public reaction on social media. Taking into account that ‘likes’, shares and comments represent different types of public reactions (Saxton & Waters, 2014), social media practitioners of non-profit organisations should first realize what they want to achieve with their social media messages, and then act accordingly. Namely, to increase sharing of the posts (or the promotion of their messages), large number of cases or statistical information should be included in the message. In order to increase a volume of all social media reactions, PR professionals should include information about the consequences of not taking required actions. However, in case of messages with ‘Information’ or ‘Action’ as a dominant function, the use of both frames in the message (positive and negative) could potentially lead to more ‘likes’ (or an increase the overall sympathy towards the messages). Besides, the inclusion of any emotions into the message could increase not only the number of ‘likes’, but also comments (or make people more engaged with the organisation). However, to increase public’s desire to promote the message in particular (increase number of shares), words and phrases that can provoke and trigger negative emotions should be applied.
Referring to the findings about the effect of the face expressions, for PR practitioners of NPOs, it might indicate that the content of the message of the post potentially has more power than the photo. Therefore, they might rather pay more attention to the content of the message (and again consider what aspects of the message to include), than solely rely on a ‘power’ of the photo attached to the post.
Limitations and Recommendations for Further Research
Even though the current study has contributed to the sphere of social media NPOs’ research and produced findings that could help NPOs to increase their electronic word-of-mouth and improve the engagement with their publics, it has several limitations.
Due to the size of the study and time constraint, only two categories of the emotions were examined. Thus, further research can take more categories into consideration. Due to the same time limit, organisations that specialize only on helping refugees were examined. Even though the results of this study can be generalized for NPOs with helping intentions and the results are externally valid, further research could examine the effect of different aspects of the message on other types of NPOs.
When answering a sub-research question, not a pure message function was used in the analysis, but a dominant one. However, the message also contained aspects of other functions, thus it might have had an effect on the results. Therefore further study could use messages where only one function is presented. Besides, future research can investigate the effect of message aspects not only on three broad categories of the posts (‘Information’, ‘Community-building’ and ‘Action’), but on more in details divided categories.
It was found that most of the message aspects do not influence the number of ‘likes’, shares and comments for posts with ‘Community-building’ and ‘Action’ as dominant
functions. The reason for this could be not big enough number of cases for each of these two functions. Therefore it is recommended to test the message aspect on larger sample, as it is important to know which aspects of the message drive engagement and promotion intentions for such important types of the Facebook posts (especially ‘Action’ posts). Not significant results for the effects of facial expressions could be also explained by not sufficient number of cases. Hence, the effects could be further examined on a larger sample, as photos are used extensively both on social media and by NPOs. Thus, knowing if the aspects of the photos of the people, whom NPO is trying to help, have any effect on Facebook public could be useful.
References
Ahluwalia, R., Burnkrant, R. E., & Unnava, R. H. (2000). Consumer responses to negative publicity: The moderating role of commitment. Journal of Marketing Research, 37,
203-214.
Allen, M., & Preis, R. W. (1997). Comparing the persuasiveness of narrative and statistical evidence using meta-analysis. Communication Research Reports, 14, 125-131. Baesler, E. J., & Burgoon, J. K. (1994). The temporal effects of story and statistical evidence
on belief change. Communication Research, 21, 582-602.
Baker, S. M., & Petty, R. E. (1992). Minority influence: Integration of comparison and validation processes. Contemporary Psychology, 37, 1309-1310.
Bayer, M.; Sommer, W., & Schacht, A. (2012). Font size matters—Emotion and attention in cortical responses to written words. PLoS ONE, 7(5), e36042.
Berger, J., Milkman, K.L. (2012). What Makes Online Content Viral? Journal of Marketing
Research, 49 (2), 192-205.
Cho,M., Schweickart, T. &Haase, A. (2014). Public engagement with nonprofit organizations on Facebook. Public relations review, 40 (3), 565-577.
Coulter, K. & Roggeveen, A. (2012). “Like it or Not” Consumer Responses to Word‐ of‐Mouth Communication in On‐line Social Networks. Management Research
Review, 35(9), 878‐899.
Daft, R.L. & Lengel, R.H. (1986). Organizational information requirements, media richness and structural design, Management Science, 32 (5), 554-71.
Das, E. , Kerkhof, D.P., & Kuiper, J. (2008). Improving the Effectiveness of Fundraising Messages: The Impact of Charity Goal Attainment, Message Framing, and Evidence on Persuasion. Journal of Applied Communication Research, 36(2), 161-175, DOI: 10.1080/00909880801922854.
Davis, J. J. (1995). The effects of message framing on response to environmental communication. Journalism and Mass Communication Quarterly, 72, 185-299.
Detweiler, J. B., Bedell, B. T., Salovey, P., Pronin, E., & Rothman, A. J. (1999). Message framing and sunscreen use: Gain-framed messages motivate beach-goers. Health
Psychology, 18, 189-196.
Ditto, P. H., Lopez & David F. (1992) Motivated skepticism: Use of differential decision criteria for preferred and nonpreferred conclusions. Journal of Personality and Social
Psychology, 63(4), 568-584.
Dobele, A., Lindgreen, A., Beverland, M., Vanhamme, J. & van Wijk, R. (2007). Why Pass On Viral Messages? Because They Connect Emotionally. Business Horizons, 50 (4), 291-304.
Field, A. (2012). Discovering Statistics. Analysis of Covariance (ANCOVA).
www.discoveringstatistics.com. Retrieved from:
http://www.statisticshell.com/docs/ancova.pdf.
Flandez (2011, August 21). Most Charities Still Do Not Raise Much Money Via Social Media, The Chronicle of Philanthrophy. Retrieved from:
https://philanthropy.com/article/Big-Charities-Gear-Up-to-Use/157949.
Forgas, J.P. (2006). Affective influences on interpersonal behavior: Towards understanding the role of affect in everyday interactions. In J.P. Forgas (ed.), Affect in Social
Thinking and Behavior. New York: Psychology Press, 269–290.
Green, M. (2006). Narratives and cancer communication. Journal of Communication, 56, 163-183.
Green, M. C., & Brock, T. C. (2000). The role of transportation in the persuasiveness of public narratives. Journal of Personality and Social Psychology, 79, 701-721.
Green, M. C., Brock, T. C., & Kaufman, G. F. (2004). Understanding media enjoyment: The role of transportation into narrative worlds. Communication Theory, 14, 311-327. Hankinson, P. (2000). Brand orientation in charity organisations: qualitative research into key
charity sectors. International Journal of Nonprofit and Voluntary Sector Marketing, 5 (3), 207-219.
Heath, C. (1996). Do people prefer to pass along good news or bad news? Valence and relevance of news as a predictor of transmission propensity. Organizational Behavior
Hennig‐Thurau, T., Gwinner, K., Walsh, G. & Gremler, D. (2004). Electronic Word‐ of‐Mouth via Consumer‐opinion platforms: What Motivates Consumers to Articulate Themselves on the Internet? Journal of Interactive Marketing, 18, 38‐ 52.
Herr, P. M., Kardes, F. R., & Kim, J. (1991). Effects of word-of-mouth and product
attributeinformation on persuasion: An accessibility-diagnosticity perspective. Journal
of Consumer Research, 17, 454-462.
Hollander, G. (2013, August 29). UK Newspapers Ranked by Total Readership (Print and Online). Press Gazette. Retriever from: http://www.pressgazette.co.uk/uk-newspapers-ranked-total-readership-print-and-online.
Kissler, J., Herbert, C., Peyk, P., & Junghöfer, M. (2007). Buzzwords: Early cortical
responses to emotional words during reading. Psychological Science, 18 (6), 475–480. Kopfman, J. E., Smith, S. W., Ah Yun, J. K., & Hodges, A. (1998). Affective and cognitive
reactions to narrative versus statistical evidence organ donation messages. Journal of
Applied Communication Research, 26, 279-300.
Levin, I. P., & Gaeth, G. J. (1988). How consumers are affected by the framing of attribute information before and after consuming the product. Journal of Consumer Research, 15, 374 – 378.
Lin T. M.Y., Wu, H., Liao C. & Liu, T. (2006). Why are some e-mails forwarded and others not? Internet Research, 16, 81-93.
Lovejoy, K. & Saxton, G.D. (2012). Information, Community, and Action: How Nonprofit Organizations Use Social Media . Journal of Computer-Mediated Communication, 17, 337–353.
Markus, M.L. (1994). Electronic mail as the medium of managerial choice. Organization
Science, 5 (4), 502-27.
Martin, B. A. S. (1995). Is the glass half-empty or half-full? How the framing of advertising messages affects consumer responses. New Zealand Journal of Business, 17, 95–105.
Morgan, S. E., & Miller, J. (2002). Communicating about gifts of life: The effect of knowledge, attitudes, and altruism on behavior and behavioral intentions regarding organ donation. Journal of Applied Communication Research, 30, 163-178.
Morman, M. T. (2000). The influence of fear appeals, message design, and masculinity on men’s motivation to perform the testicular self-exam. Journal of Applied
Communication Research, 28, 91-116.
Ngwenyama, O.K. & Lee, A.S. (1997). Communication richness in electronic mail: critical social theory and the contextuality of meaning. MIS Quarterly, June, 145-67. Peters, K.; Kashima, Y.; & Clark, A. (2009) Talking about others: Emotionality and the
dissemination of social information. European Journal of Social Psychology, 39(2), 207–222
Reinard, J. C. (1988). The empirical study of the persuasive effects of evidence: The status after fifty years of research. Human Communication Research, 15, 359.
Reynolds, R. A., & Reynolds, J. L. (2002). Evidence. In J. E. Dillard & M. Pfau (Eds.), The
persuasion.
Rimé, B. (2009) Emotion elicits the social sharing of emotion: Theory and empirical review.
Emotion Review, 1, 60–85.
Rimer, B. K., & Kreuter, M. W. (2006). Advancing tailored health communication: A
persuasion and message effects perspective. Journal of Communication, 56, 184-201. Rook, K. S. (1987). Effects of case history versus abstract information on health attitudes and
behaviors. Journal of Applied Psychology, 17, 533– 553.
Rothman, A. J., Bartels, R. D., Wlaschin, J., & Salovey, P. (2006). The strategic use of gain-and lossframed messages to promote healthy behavior: How theory can inform practice. Journal of Communication, 56, 202-220.
Saxton, G.D & Waters, R.D (2014). What do Stakeholders Like on Facebook? Examining Public Reactions to Nonprofit Organizations’ Informational, Promotional, and Community-Building Messages. Journal of Public Relations Research, 26 (3), 280-299.
Seta, C.E., Hayes, N.S. & Seta, J.J. (1994). Mood, memory, and vigilance: the influence of distraction on recall and impression formation. Personality and Social Psychology
Bulletin, 20, 170-7.
Small, D. & Verrochi, N. M (2009). The Face of Need: Facial Emotion Expression on Charity. Advertisements. Journal of Marketing Research, 777 (46), 777–787. Waters, R.D. (2007). Nonprofit organizations' use of the internet: A content analysis of
communication trends on the internet sites of the philanthropy 400. Nonprofit
Management and Leadership, 18, 59–76.
Footnotes
1
It should be noted from now on that the same effect that is argued/assumed for the message is also valid for the post. When it is discussed that message will receive more ‘likes’, shares or comments, it also means that post will receive the same. As post is constituted from the message, aspects of the message will also influence the number of ‘likes’, shares,
comments of the post.
2
The variance ratio is the largest variance divided by the smallest ratio, and it should be less than about two. Variance can be calculated by squaring the SDs (Field, 2012, p.6). Thus variance ratio = SD2largest/SD2smallest . In the table 4 you can find the values of largest and
smallest standard deviations of the variables, for which Levene’s test was significant, as well as the values of variance ratios.
Table 4. Variance Ratios and Largest and Smallest Standard Deviations of Variables with
Significant Levene’s Tests
Variable SDlargest SDsmallest Variance Ratio
Evidence Shares Lg10 .77 .74 1.09 Frame Likes Lg10 .59 .44 1.78 Frame Shares Lg10 .77 .72 1.15 Em. Types Shares Lg10 .76 .61 1.58 (Info) Frame Likes Lg10 .64 .48 1.73 (Info) Em. Presence Comments Lg10 .61 .49 1.54 (Info) Em. Types Shares Lg10 .76 .65 1.35 (Action) Em. Types Shares Lg10 .77 .28 8
3
Emotional messages are the messages that contain any types of emotions (both positive and negative). Namely, when a message contains any questions, statements or words which aim to trigger any of the following emotions: fear, sadness, frustration, worriedness, miserableness, angriness, happiness, cheerfulness, gladness, warmness, vigorousness or enthusiasm, it is considered as an emotional message (Lin et al., 2006).
4
It should be noted that face expressions only of refugees were taken into account (not simply any people’s facial emotions). Refugees are the ones whom the studied (sampled) organisations are aiming to help, and the same is valid for Small’s and Verrochi’s (2009) study.
Appendix
Detailed Results for ‘Information’, ‘Community-Building’ and ‘Action’ Function Messages
There was a statistically significant difference in the number of
‘likes’/shares/comments based on message evidence, only for messages with ‘Information’ as dominant function F (6,116) = 4.81, p < .0005, Wilk’s Λ = 0.91, partial η2 =
.045. Statistically significant moderate effect of message evidence was found only on the number of shares F (2,310) = 7.77, p = .001, partial η2 = .048 for ‘Information’ –dominant function messages. Post-hoc test indicated the significant difference found between statistical and narrative evidence messages (Mdifference = .40, p < .0005) and between messages with no
evidence and messages with statistical evidence (Mdifference = -.238, p = .045) regarding shares.
Messages with ‘Information’ function as a dominant one and with statistical evidence got more shares (M = 1.10, SD = 0.76) than messages with narrative evidence (M = 0.75, SD = 0.78) and than messages with no evidence (M = 0.98, SD = 0.73), which on their turn still scored more on shares than messages with narrative evidence
It should be noted that the assumption of equal variances in the population was not violated for shares (Levene’s F (2,315) = 2.08, p = .126) and comments (Levene’s F (2,315) = 0.18, p = .832), but was violated for ‘likes’ (Levene’s F (2,315) = 3.77, p = 0.24). However, the variance ratio for ‘likes’’s data was less than two, indicating that there was equal variance in the population (Field, 2012, p. 6)2.
There was a statistically significant difference in the number of
‘Information’ as a dominant function, but not for messages with dominant function ‘Community-building’ (please see table 5).
Table 5
Results of Multivariate Tests for the Effect of Message Frame for Messages with Different Dominant Functions
Dominant Function
F Hypothesis df
Error df p Partial eta squared Information 10.19 6 616 <0005 .09 Community-building .70 3 32 .56 .061 Action 4.08 6 158 .0001 .13
Message frame had a statistically significant effect on the number of shares and ‘likes’, and no significant effect on the number of comments for messages with ‘Information’ as dominant function, and the same effect for messages with ‘Action’ as a dominant function. Please see table 6.
Table 6
Results of the Tests of Between-Subjects Effects of Message Frame for Messages with ‘Information’ and ‘Action’ as Dominant Functions
Dominant Function F df Error df p Partial eta squared Information Shares Lg10 7.60 2 310 .001 .047 Likes Lg10 11.18 2 310 <0005 .067 Comments Lg10 2.95 2 310 .054 .019 Action Shares Lg10 2.19 2 81 .006 .118 Likes Lg10 0.95 2 81 .001 .155 Comments Lg10 2.27 2 81 .115 .052
