TY - JOUR
AU - Ramírez,, Josefina
AB - Abstract We hypothesize that generic frames influence what news people share on Facebook and Twitter through three different routes: emotions, motivations, and psychological engagement. Using a mixed-methods design, a content analysis of a representative sample of articles published in six Chilean outlets was combined with in-depth interviews with digital journalists. After controlling for issue, newsworthiness, informational utility, valence, and other confounds, results show that—across platforms—a morality frame increases news sharing, whereas a conflict frame decreases it. Emphasizing economic consequences also decreases sharing, but only on Facebook. Surprisingly, the human interest angle has no noticeable effects. These results show that news frames can have behavioral consequences, and confirm the existence of a gap between preferred frames of journalists and users. Sharing news over Facebook, Twitter, and other social media has become an everyday practice for online news users around the globe. As a consequence, professional news sites are increasingly dependent upon referrals from social media. In June 2015, Facebook surpassed Google as the main traffic driver for the 400 publisher sites tracked by the analytics firm Parsley (Boxer, 2015). News sharing on social media not only affects the media industry; it also has important democratic consequences. The dissemination of online news may influence users' interest in journalistic content, either through direct involvement in deciding what to share or through incidental exposure to information (Bode, 2016). The growing importance of news sharing has led to an explosion of research in various fields, including journalism (e.g., Trilling, Tolochko, & Burscher, 2017), health communication (e.g., Kim, 2015), computer sciences (e.g., Bandari, Asur, & Huberman, 2012), marketing (e.g., Berger & Milkman, 2012), and the neurosciences (e.g., Scholz et al., 2017). The literature shows that there is no “magic bullet” to predict what news will become viral. Rather, a complex combination of users' motivations, content characteristics, network structures, temporal patterns, and media systems affect what news catch on social media. Notwithstanding the complex set of influences, most of what we know stems from data-driven, rather than theory-driven, studies. In their comprehensive review, Kümpel, Karnowski, and Keyling (2015) found that less than half of the 97 analyzed articles employed a theory to explain news sharing. In most cases, this theory was diffusion of innovations, which does not really focus on news content or attributes. Relatedly, several studies have focused on a single news outlet and/or platform, which raises the question of how applicable are these findings to other outlets and platforms. Last, most of the empirical literature is based on data collected in the United States—a common trend in work on social media in general (for exceptions, see Boczkowski & Mitchelstein, 2013; Bright, 2016; García-Perdomo, Salaverría, Kilgo, & Harlow, 2017). To fill in some of the gaps, the current study examines the effects of content characteristics on users' diffusion of professional news on Facebook and Twitter in Chile—the country with the highest social media penetration rate in Latin America. It does so by drawing on one of the most popular theories of news content effects—framing—especially work on affective and behavioral consequences of framing (Lecheler, Schuck, & de Vreese, 2013; Semetko & Valkenburg, 2000). Several reasons justify our choice. From a theoretical perspective, the behavioral consequences of framing effects remain understudied. The bulk of the literature centers on the cognitive and attitudinal effects of journalistic news frames (de Vreese, 2012). Here, instead, we study the number of times users have actually shared news with their online networks. From a practical perspective, content is the only factor that is fully controllable by journalists and media organizations to influence what people share on social media. Because content factors can predict news virality by interacting with users' psychological needs (Cappella, Kim, & Albarracín, 2015), knowing which content factors matter for news sharing may help journalists and news media to prosper in the age of social media. Last, it is important to check whether the gaps in media and audience framing identified in the broadcast era (Neuman, Just, & Crigler, 1992) remain in the social media age. Why do people share news? In the current media environment, news sites structure their content to make it shareable on social media—as evidenced by the prominent display of social media toolbars. In her ethnography, Usher (2014) describes the efforts of The New York Times to make its content more spreadable: “Times content, when shared over social networks, talked about on Facebook, commented on, or the like, became spreadable content” (p. 190). Some journalists and editors also aim at producing viral news for gaining influence and promoting the personal brand (Hedman & Djerf-Pierre, 2013). Thus, maximizing the virality of stories on social media has become a desirable outcome for both media companies and the news professionals who work there (Khuntia, Sun, & Yim, 2016). This clarity of purpose, however, conflicts with the lack of clarity about what virality is and how to measure it. According to Alhabash and McAlister (2015), virality on social media comprises three dimensions: reach (e.g., volume of sharing online, forwarded messages, posts and reposts); evaluation (e.g., number of likes, favorites); and deliberation (e.g., comments). Applied to news content, this approach combines popularity with “contagious behavior” (Heimbach, Schiller, Strufe, & Hinz, 2015). However, clicking on a link is not the same as liking the link or sharing it. Thus, we define news virality on social media following Heimbach et al.’s (2015) definition: a property of content that enhances its likelihood to be shared by a multitude of users in different social media. In this sense, virality is equated with retransmission or diffusion—“the probability that [the message] be sent along” (Hansen, Arvidsson, Nielsen, Colleoni, & Etter, 2011, p. 2)—rather than sheer attention or positive evaluation garnered by the content. This definition narrows down the measurement of virality to specific metrics. On Facebook, it refers to the number of times users have shared a post or link produced by a news media organization by reposting it on their personal profiles or pages. On Twitter, it considers the number of times a message containing a link to a news story has been tweeted or retweeted. Some authors see a direct connection between basic human psychology and news sharing. As Hermida (2014) noted: “People are not hooked on YouTube, Twitter or Facebook but on each other. Tools and services come and go; what is constant is our human urge to share” (p. 1). Nevertheless, prior research has found that disseminating information on YouTube, Twitter, or Facebook serves specific purposes. Drawing from the uses and gratifications approach, studies have found that news sharing is motivated by information acquisition, impression management, self-expression, socializing, and entertainment. Which motivation is strongest, however, is not clear. Information-seeking motivations were found to be the strongest predictors in Lee, Ma, and Goh (2011), though the impression-management (e.g., seeking status) factor has also been identified as a major driver for spreading news online (Berger, 2014; Scholz et al., 2017). The literature on digital sharing also highlights the role played by emotions and emotional regulation (Heimbach & Hinz, 2016). In social psychology, Harber and Cohen (2005) advanced an emotional broadcaster theory (EBT) of news sharing, which posits that people have inner necessity to share experiences with others, and news stories—especially the emotionally arousing ones—are passed on to others to fulfill the need for sharing. EBT is consistent with research showing that content triggering high arousal emotions (e.g., awe, happiness, anger) is retransmitted more often (Berger & Milkman, 2012; Stieglitz & Dang-Xuan, 2013). Thus, social media users engage in “emotion-led sharing” (Bright, 2016, p. 346), passing on news that they feel more emotionally attached to. In sum, social media users have multiple motivations for news sharing. In the current study, however, we focus on news content attributes. The question, then, is why reviewing users' motivations is important for a content-based study. The answer lies in Shoemaker and Reese’s (1990) call to make more effort matching content characteristics to media effects. That is, to understand users' behaviors, it is important to focus on content factors as much as users' psychology. In fact, several of the gratifications sought or obtained that we described above suggest that specific content attributes may increase the virality of a news story. For instance, content attributes that match information-seeking gratifications, such as newsworthiness (e.g., relevance, deviation, timeliness) and utility (e.g., originality, practical value), are more likely to be shared (Bright, 2016; Kim, 2015; Trilling et al., 2017). Framing effects on sharing news Among the content attributes studied as predictors of virality, journalistic frames have received little scholarly attention. This is somewhat surprising, considering that framing is one the most popular areas of communication research. Although there are several definitions of what news frames are, here we conceptualize them as the dominant set of aspects and considerations emphasized by a news story, independent of the story’s specific topic—in framing-effects parlance, we study emphasis, instead of equivalent, frames (Druckman, 2001, p. 230). Thus, framing implies that news content is constructed through particular features that provide clues about the interpretation of the text and the news event itself, suggesting certain attributes, judgments, and decisions (Lecheler et al., 2013). These features include a wide array of presentation elements, such as the presence (or absence) of specific keywords, phrases, stereotypes, images, news sources, metaphors, exemplars, and quotations (de Vreese, 2012). With the purpose of improving the replicability and generalizability of emphasis framing effects, scholars have advanced the analysis of generic frames (Matthes, 2009). These “transcend thematic limitations and can be identified in relation to different topics, some even over time and in different cultural contexts” (de Vreese, 2012, p. 368). Building on the seminal work of Neuman et al. (1992), Semetko and Valkenburg (2000), and others, in this study we focus on four generic frames: conflict, economic consequences, human interest, and morality. The conflict frame “emphasizes conflict between individuals, groups, or institutions as a means of capturing audience interest” (Semetko & Valkenburg, 2000, p. 95). One of the most frequently used frames in news coverage, conflict is common in political and public affairs coverage (Bartholomé, Lecheler, & de Vreese, 2015), in part because it facilitates the journalistic standard of balanced reporting (Neuman et al., 1992). The conflict frame tends to increase the perceived seriousness and news value of an event (Burscher, Odijk, Vliegenthart, De Rijke, & de Vreese, 2014). To the degree that informational utility drives news sharing on social media, the use of a conflict frame could increase the virality of an article. Furthermore, conflict stories tend to be negatively valenced, and prior research has found this type of information is more likely to be selected and attended than neutral or positively valenced stories (Zillmann, Chen, Knobloch, & Callison, 2004). In political communication, conflict has been found to be a mobilizing force (Schuck, Vliegenthart, & de Vreese, 2016). For instance, Weber (2014) found that online news featuring controversies motivated repeated commenting by the same set of users. The little empirical evidence on news sharing supports these expectations. In their study on predictors of the “shareworthiness” of news, Trilling et al. (2017, p. 38) found that articles with a conflict frame were shared on Twitter and Facebook 11% and 9% more, respectively. Conversely, conflict may also be associated with less virality. Fear of negative evaluation can make people more reluctant to transmit negative news, such as those involving conflicts (Rosen & Tesser, 1972). Because impression management and self-enhancement are major motivations for sharing on social media, positive stories tend to be shared more often than negative ones (Berger, 2014; Cappella et al., 2015)—and positive stories tend not to be framed as a conflict. On the other hand, there is a gulf between news selected online versus shared online (Gabielkov, Ramachandran, Chaintreau, & Legout, 2016). Thus, the use of a conflict angle may well increase attention to a news story, but may reduce its likelihood of being diffused on social platforms such as Facebook and Twitter. After all, most people do not want to be perceived as a Cassandra, always predicting bad news. Some research examining the behaviors of social media editors also shows that they prefer other frames (e.g., human interest) over the conflict frame when deciding what story link they post on their Twitter feeds (Wasike, 2013). In conclusion, it is not clear whether conflict-framed news are more or less viral than conflict-free news. To the degree that informational utility and negatively valenced news drive sharing, the use of a conflict frame could increase the virality of an article. To the degree that impression management, self-enhancement, and positive content determine what people share, conflict frames will be associated with less virality. We thus posit: RQ1: What is the relationship between the use of a conflict frame and the number of shares an article receives on social media? Framing news events in terms of their economic impact is typical in hard news content outlets, such as elite newspapers of record, because it signals high newsworthiness (Neuman et al., 1992). As defined by Semetko and Valkenburg (2000), the economic consequences frame “reports an event, problem, or issue in terms of the consequences it will have economically on an individual, group, institution, region, or country” (p. 96). Wasike (2013) found that despite its attributed newsworthiness, the economic angle is the second least used frame by social media editors. The trend among social media editors—he argued—is to personalize content and choose entertainment over other topics, which do not go together with the economic consequences frame. In addition, the literature on framing effects shows that the economic consequences frame is more cognitively demanding and complex than the human interest or conflict frames, in part because it demands using technical language (Neuman et al., 1992, p. 63). Furthermore, unless there is an economic downturn, stories framed in economic terms tend to employ less emotionally arousing language and make heavier use of abstract language and statistics (Bachmann, 2005). Prior studies show that statistical information is oftentimes less engaging and persuasive than other types of statements (Zillmann & Brosius, 2000). To the degree that entertaining content is a positive predictor of sharing, as suggested by research on the uses and gratifications of news sharing (Lee et al., 2011), it could be expected that statistical information and technical language that tends to accompany an economic frame is less entertaining and, thus, less viral. Because economic frames trigger less recall, are less arousing, and are less compatible with typical motivations for news sharing, we expect that: H1: The use of an economic consequences frame decreases the number of shares an article receives on social media. Another typical journalistic resource is to define a complex event through a specific situation or exemplar, in order to make it more accessible to the audience (Zillmann & Brosius, 2000). In the framing literature, the use of this particular angle is defined as the human interest frame. According to Semetko and Valkenburg (2000), it is a frame that “brings a human face or an emotional angle to the presentation of an event, issue, or problem […] refers to an effort to personalize the news, dramatize or ‘emotionalize’ the news, in order to capture and retain audience interest” (p. 95). Contrary to economic consequences, the human interest frame has been found to increase news virality. In their study of Dutch news media, Trilling et al. (2017) found that it increased Facebook shares by 33% compared to articles not employing this frame (on Twitter, the predicted effect was insignificant). In Wasike’s (2013) study about social media editors, the human interest angle was the most shared of the studied generic frames. And in their cross-national analysis, García-Perdomo et al. (2017) found that human interest triggered both Facebook and Twitter users to share and interact with news articles. The virality potential of this frame is consistent with research showing that it is effective in triggering emotional arousal. Jebril, de Vreese, Dalen, and Albæk (2013) found that the human interest frame was associated with learning from the news, and explained that this was especially true for those who are aroused by the personal, softer aspect of otherwise hard news. In addition to the arousal explanation, it could be argued that the human interest angle influences news sharing by increasing people’s psychological engagement with the news. For instance, in an experiment, Hong (2013) found that the human interest framing of medical news increased audiences' involvement in those stories. And, as we discussed above, recall and involvement can drive or manifest itself in sharing behavior. Lastly, this type of generic frame may interact with the impression management motivation for sharing and increase the diffusion of an article on social media. One of the drivers of news sharing is to communicate to others (or oneself) a specific identity (Berger, 2014). Compared to stories that lack a human face, the presence of the human interest frame may increase social media users' identification with the news event and the individuals mentioned in the story. Thus: H2: The use of a human interest frame increases the number of shares an article receives on social media. We now turn to the last generic frame studied here, the morality frame. This news angle puts the event or issue in the context of values, moral prescriptions, normative messages, and religious or cultural tenets (Semetko & Valkenburg, 2000, p. 96). Studies show that it is the least common of the generic frames in the news, in part because it clashes with journalists' idea of objectivity. Thus, it tends to show up in news content through the type of sources interviewed and selected quotes (Neuman et al., 1992). In a content analysis of three national Dutch daily newspapers between 1995 and 2011, the morality frame appeared in only 13% of the sampled news (Burscher et al., 2014). Morality was also among the most infrequent frames used in the stories shared by social media editors on their Twitter feeds (Wasike, 2013). Nevertheless, studies have found that audiences frequently use this frame to make sense of news (Druckman, 2001; de Vreese, 2012). After analyzing 29 in-depth interviews on how people process news, Neuman et al. (1992) concluded that “virtually all of the interviewees used moral and value-laden statements to talk about the five political issues” (p. 73). Prior studies on political psychology and framing lead us to expect that there is a strong connection between the morality angle and sharing. To the degree that intense emotions are activated by the use of a morality frame (e.g., outrage), these stories should be more viral because of their emotional arousal. That is, moral frames produce emotions that mobilize people into action. Morality framing can also lead to news sharing when these frames resonate with social media users' existing value predispositions. Prior work on media effects shows that messages that match audiences' values have a higher likelihood of being remembered and, thus, influence the importance and evaluations of the issues covered in the news (Schemer, Wirth, & Matthes, 2012). So long as news stories laden with values with which people identify or perceive as important are more influential on people’s preferences and behavior, morality frames have a higher virality potential (Brady, Wills, Jost, Tucker, & Van Bavela, 2017). Relatedly, prior research has found that news sharing on social media is contingent upon users' attitudes, such that people tend to share content that reflect their viewpoints (Arendt, Steindl, & Kümpel, 2016). To the degree that sharing allows people to tell others what users think and believe, news with moral frames that resonate with users will be shared more often. It is even possible that these different mechanisms of influence work in tandem—a morality frame increases emotional arousal while simultaneously impacting cognitive processing and selectivity. Furthermore, with the parallel rise of so-called “culture wars” and partisan news outlets such as Fox News in the United States, the use of morality framing could be on the rise as well (Sobieraj & Berry, 2011). If this is the case, the use of a morality frame will have higher chances of resonating with news users, thus increasing the recall and involvement with the content. And, as argued, involvement may increase sharing. H3: The use of a morality frame increases the number of shares an article receives on social media. To sum up, the literature suggests three main routes of influence of generic frames on news sharing: emotions, motivations, and psychological involvement. Some frames (e.g., human interest) may increase arousal, while others (e.g., economic consequences) may decrease it. At the same time, frames may prime specific gratifications, such as impression management and self-enhancement. Lastly, generic frames are related in varying degrees to news recall and involvement, which in turn can increase the likelihood of sharing in social media. Sharing across platforms So far, the effects of generic frames have been discussed in terms of shares in social media, though specific platforms may be associated with unique uses and effects. This is because platforms may differ on their affordances, which refer to the properties of platforms that emerge from users' interactions with them. According to prior research (Ellison & Vitak, 2015; Treem & Leonardi, 2012), among the typical affordances are visibility (e.g., ease of locating information), persistence (e.g., endurance of information after being shared), editability of content, replicability (e.g., sharing), associations (e.g., connections between users and content), and scalability (e.g., potential of reaching large audiences). Facebook and Twitter share many of these affordances—their content persists over time, can be edited or deleted, and is easily replicable (Vraga & Bode, 2017). Nevertheless, there are important differences, especially in terms of associations and scalability. Facebook’s network of personal profiles is rather symmetrical. In order to connect with someone, both parties have to approve the relationship. Twitter, in contrast, allows asymmetrical connections; reciprocal approval is not needed to follow and contact other users. As a consequence, the intended audience of what users share in each platform is different. On Facebook, it could generally mean people with whom the user has (or has had) some deeper level of intimacy or personal knowledge. On Twitter, it could well include people with similar interests but with little or no level of personal intimacy (Valenzuela, Correa, & Gil de Zúñiga, 2017). Relatedly, Facebook has a larger population than Twitter, which means that users' offline networks are more likely to be represented online on Facebook than on Twitter. This, in turn, may affect the motivations for sharing particular content as well as the degree of visibility of the content being shared. Thus, it is likely that some differences will arise in terms of what content triggers sharing behavior across platforms. Because we do not know of published research connecting affordances to virality in specific ways, we ask: RQ2: What are the differences, if at all, across Facebook and Twitter in the relationships between the use of frames and the number of shares an article receives? Methods This study relied on a mixed-methods approach that combines a content analysis of a representative sample of stories from Chilean news sites with in-depth interviews of digital journalists. Following Creswell and Plano Clark’s (2007) typology, we used an explanatory sequential design, in which a first stage of quantitative data collection and analysis is followed by a qualitative analysis. The quantitative data will be used to test whether generic news frames predict news sharing in Facebook and Twitter and, if so, what is the size and sign of these relationships. The in-depth interviews, in turn, will explore how journalists perceive the relationship between the use of generic frames and virality on Facebook and Twitter. Quantitative content analysis Data We collected a dataset of stories published in 2014 by six Chilean news websites. These included the sites of the two leading national newspapers (Emol.cl and LaTercera.com), an evening newspaper from Chile’s capital city (LaSegunda.com), the two most popular radio news networks (BioBioChile.cl and Cooperativa.cl), and the public TV broadcaster’s main news program (24Horas.cl). In addition to representing a variety of outlets and audiences, all of them compete in the general news market and are within the most visited news sites in Chile (Díaz & Mellado, 2017). To obtain the dataset, we used Dixit (http://dixit.io), a platform developed by the Chilean journalist Tomás Pollak. Dixit records the URL addresses of each article produced by the six websites described above, counts the number of unique Twitter accounts that have shared via tweets or retweets content linking to each of these URLs, and visualizes articles by time, topic, source, and number of shares. This information is updated every 10 minutes, on a daily basis. The database includes every single URL produced by each site that has been linked at least once on Twitter, excluding those that have been published but never shared (for a similar approach, see Khuntia et al., 2016). The sampling frame considered a year’s worth of news (i.e., all stories published between 1 January and 31 December 2014). Following extant recommendations (Hester & Dougall, 2007), we produced 10 constructed weeks, for a total of 70 days of news and an initial sample of 39,085 articles. After removing duplicates, the total was reduced to 37,843. Considering the available human and time resources, we had to select 3,500 articles for coding. To maintain a representative sample, we ordered the records for the 37,843 articles in descending order of shares, divided the list in five strata based on the number of shares, and selected 700 articles within each strata in random fashion. From these 3,500 articles, 91 had to be discarded due to broken links to the full-text version, for a final sample of 3,409 articles. Coding Seven undergraduate journalism students, all blind to this study’s hypotheses and trained by the authors between March and May 2015, conducted manual coding. A random sample of 150 news stories was taken from the sampling frame (but excluded from the final sample analyzed here) as reliability data. Using the ReCal3 software (Freelon, 2010), intercoder reliability averaged 0.85 using Krippendorff’s α and 95.1% using percentage agreement (see Appendix S1 for reliability estimates and descriptives of each variable, Supporting Information). Variables We obtained Twitter shares from the Dixit database, queried on January 2015, whereas we obtained Facebook shares using the application’s programming interface (API Graph) on January 2016. To gauge the presence of generic frames, we relied on the operationalization performed by Burscher et al. (2014) of Semetko and Valkenburg’s (2000) conceptualization of conflict, economic consequences, human interest, and morality. This process employs a set of 10 questions to which coders are required to answer yes or no after reading the full text of each news article, such as: “Does the news item reflect disagreement between parties, individuals, groups or countries?” (conflict); “Is there a reference to economic consequences of pursuing a specific course of action?” (economic); “Does the item employ adjectives or personal vignettes that generate feelings of outrage, empathy or caring?” (human interest); and “Does the item contain any moral or ethical message?” (morality) (for exact question wording in English, see Table 1 in Burscher et al., 2014, p. 197). Subsequently, responses to the indicator questions were aggregated such that a frame was considered to be present when at least one of its indicators was coded positively. A principal component analysis confirmed a four-factor solution (see Appendix S1). Table 1 Predicting the Number of Shares on Twitter and Facebook (N = 3,409) . . Facebook Shares IRR . Twitter Shares IRR . Frames Conflict 0.55*** 0.83* Economic consequences 0.62** 0.93 Human interest 1.07 1.00 Morality 1.75*** 1.54*** Issue (reference: social issues) Politics 0.45*** .76+ Economy 1.96+ 0.80 Crime 1.12 0.91 International 1.28 1.23 Sports 0.33*** 0.53*** Science/technology 2.07* 1.89*** Disasters/accidents 1.24 1.39+ Odd news 2.29** 1.51** Animals 3.22*** 3.16*** Celebrities/entertainment 0.88 0.89 Newsworthiness Deviance 2.24*** 1.64*** Relevance 1.35* 1.25** Proximity 1.97*** 1.03 Exclusiveness Original news 4.29*** 2.05*** Utility Practical utility 1.24 1.14 Tone % Positive words (logged) 0.71 0.76 % Negative words (logged) 2.96*** 1.25 Complexity % Cognitive mechanisms words 3.06** 1.79+ Format (reference: inverted pyramid) Listicle 1.92* 1.70* Visual only 1.75** 1.43* Feature 2.83*** 1.41* Interview (Q&A) 0.53 0.85 Repeated updates 1.99 2.33*** Multimedia Audio embedded 0.60* 0.61** Video embedded 1.09 1.00 Embedded social media quote 2.17** 2.11*** Embedded image/photo 0.98 1.80*** Length Word count (logged) 6.61* 8.08*** Nagelkerke R2 0.30 0.26 . . Facebook Shares IRR . Twitter Shares IRR . Frames Conflict 0.55*** 0.83* Economic consequences 0.62** 0.93 Human interest 1.07 1.00 Morality 1.75*** 1.54*** Issue (reference: social issues) Politics 0.45*** .76+ Economy 1.96+ 0.80 Crime 1.12 0.91 International 1.28 1.23 Sports 0.33*** 0.53*** Science/technology 2.07* 1.89*** Disasters/accidents 1.24 1.39+ Odd news 2.29** 1.51** Animals 3.22*** 3.16*** Celebrities/entertainment 0.88 0.89 Newsworthiness Deviance 2.24*** 1.64*** Relevance 1.35* 1.25** Proximity 1.97*** 1.03 Exclusiveness Original news 4.29*** 2.05*** Utility Practical utility 1.24 1.14 Tone % Positive words (logged) 0.71 0.76 % Negative words (logged) 2.96*** 1.25 Complexity % Cognitive mechanisms words 3.06** 1.79+ Format (reference: inverted pyramid) Listicle 1.92* 1.70* Visual only 1.75** 1.43* Feature 2.83*** 1.41* Interview (Q&A) 0.53 0.85 Repeated updates 1.99 2.33*** Multimedia Audio embedded 0.60* 0.61** Video embedded 1.09 1.00 Embedded social media quote 2.17** 2.11*** Embedded image/photo 0.98 1.80*** Length Word count (logged) 6.61* 8.08*** Nagelkerke R2 0.30 0.26 Notes: Cells report incidence rate ratios (IRR) from negative binomial regressions (truncated at 0 for Facebook shares and at 1 for Twitter shares). IRR < 1 = negative effect; IRR = 1 = positive effect. The effects of news sites and temporal variables were included but not shown here due to space constraints. To facilitate interpretation and comparison with the other variables, tone and complexity measures were standardized to a 0 (minimum) to 1 (maximum) range prior to model estimation. Significance was assessed with robust standard errors. Nonparametric (Spearman’s rank-order) correlations between key variables: ρFB,conflict = −0.02; ρFB,economic = −0.01; ρFB,human = 0.05; ρFB,morality = 0.07; ρTW,conflict = −0.05; ρTW,economic = −0.04; ρTW,human = 0.05; ρTW,morality = 0.07. + p < .10. * p < .05. ** p < .01. *** p < .001 (two-tailed). Open in new tab Table 1 Predicting the Number of Shares on Twitter and Facebook (N = 3,409) . . Facebook Shares IRR . Twitter Shares IRR . Frames Conflict 0.55*** 0.83* Economic consequences 0.62** 0.93 Human interest 1.07 1.00 Morality 1.75*** 1.54*** Issue (reference: social issues) Politics 0.45*** .76+ Economy 1.96+ 0.80 Crime 1.12 0.91 International 1.28 1.23 Sports 0.33*** 0.53*** Science/technology 2.07* 1.89*** Disasters/accidents 1.24 1.39+ Odd news 2.29** 1.51** Animals 3.22*** 3.16*** Celebrities/entertainment 0.88 0.89 Newsworthiness Deviance 2.24*** 1.64*** Relevance 1.35* 1.25** Proximity 1.97*** 1.03 Exclusiveness Original news 4.29*** 2.05*** Utility Practical utility 1.24 1.14 Tone % Positive words (logged) 0.71 0.76 % Negative words (logged) 2.96*** 1.25 Complexity % Cognitive mechanisms words 3.06** 1.79+ Format (reference: inverted pyramid) Listicle 1.92* 1.70* Visual only 1.75** 1.43* Feature 2.83*** 1.41* Interview (Q&A) 0.53 0.85 Repeated updates 1.99 2.33*** Multimedia Audio embedded 0.60* 0.61** Video embedded 1.09 1.00 Embedded social media quote 2.17** 2.11*** Embedded image/photo 0.98 1.80*** Length Word count (logged) 6.61* 8.08*** Nagelkerke R2 0.30 0.26 . . Facebook Shares IRR . Twitter Shares IRR . Frames Conflict 0.55*** 0.83* Economic consequences 0.62** 0.93 Human interest 1.07 1.00 Morality 1.75*** 1.54*** Issue (reference: social issues) Politics 0.45*** .76+ Economy 1.96+ 0.80 Crime 1.12 0.91 International 1.28 1.23 Sports 0.33*** 0.53*** Science/technology 2.07* 1.89*** Disasters/accidents 1.24 1.39+ Odd news 2.29** 1.51** Animals 3.22*** 3.16*** Celebrities/entertainment 0.88 0.89 Newsworthiness Deviance 2.24*** 1.64*** Relevance 1.35* 1.25** Proximity 1.97*** 1.03 Exclusiveness Original news 4.29*** 2.05*** Utility Practical utility 1.24 1.14 Tone % Positive words (logged) 0.71 0.76 % Negative words (logged) 2.96*** 1.25 Complexity % Cognitive mechanisms words 3.06** 1.79+ Format (reference: inverted pyramid) Listicle 1.92* 1.70* Visual only 1.75** 1.43* Feature 2.83*** 1.41* Interview (Q&A) 0.53 0.85 Repeated updates 1.99 2.33*** Multimedia Audio embedded 0.60* 0.61** Video embedded 1.09 1.00 Embedded social media quote 2.17** 2.11*** Embedded image/photo 0.98 1.80*** Length Word count (logged) 6.61* 8.08*** Nagelkerke R2 0.30 0.26 Notes: Cells report incidence rate ratios (IRR) from negative binomial regressions (truncated at 0 for Facebook shares and at 1 for Twitter shares). IRR < 1 = negative effect; IRR = 1 = positive effect. The effects of news sites and temporal variables were included but not shown here due to space constraints. To facilitate interpretation and comparison with the other variables, tone and complexity measures were standardized to a 0 (minimum) to 1 (maximum) range prior to model estimation. Significance was assessed with robust standard errors. Nonparametric (Spearman’s rank-order) correlations between key variables: ρFB,conflict = −0.02; ρFB,economic = −0.01; ρFB,human = 0.05; ρFB,morality = 0.07; ρTW,conflict = −0.05; ρTW,economic = −0.04; ρTW,human = 0.05; ρTW,morality = 0.07. + p < .10. * p < .05. ** p < .01. *** p < .001 (two-tailed). Open in new tab To take into account any stable difference in sharing practices across websites not captured with our variables, we included dummy variables for each site. We also produced dummies for hour of publication and number of months between publication and queries of the sharing data (day of week was already controlled for with the use of constructed week sampling). Considering that some news topics are more viral than others, we controlled for the dominant topic of the story: politics, economy, crime, social issues, international, sports, science and technology, disasters and accidents, odd news, animals, and celebrities/entertainment. We dummy coded these categories as well, with social issues as reference category because it was the closest to the mean number of Facebook and Twitter shares in the sample. Based on Shoemaker’s (1996) bio-cultural theory of news, we measured relevance and deviance. Proximity was gauged as geographical and cultural proximity, coded such that higher scores mean more proximal events to the average Chilean reader. Because exclusiveness has been found to be correlated to news sharing (Trilling et al., 2017), coders were instructed to annotate an article’s byline to determine whether the article is original to the site or not (e.g., authored by a news agency). The story’s practical value was gauged by asking coders to answer whether the news contains practically useful information or provides a direct appeal to the reader to change his/her behavior (Berger & Milkman, 2012). To measure the valence of news, we relied on the Linguistic Inquiry and Word Count (LIWC) 2007 dictionary developed by Pennebaker, Booth, and Francis (2007), and validated for Latin American Spanish by Ramírez-Esparza, Pennebaker, Garcia, and Suriá (2007). This version of LIWC counts 642 words and word stems for positive emotion (e.g., happy, nice, sweet), and 742 words and word stems for negative emotion (e.g., hurt, ugly, nasty). Separate measures of the percentage of positive and negative words used in the full-text articles were produced. For the analysis, both measures were log-transformed because of their skewed distribution. We also measured content complexity (Heimbach & Hinz, 2016; Kim, 2015) with the proportion of words identified by LIWC that fall into the cognitive mechanisms category (e.g., cause, know, ought). The last set of control variables taps into format and presentation. We instructed coders to decide whether the article’s written style was inverted pyramid, feature, list (i.e., “listicle”), interview (i.e., Q&A style), repeated updates, or visual (e.g., stand-alone video, photograph, or infographic). Following Khuntia et al. (2016), the use of multimedia elements was coded using indicators for embedded audio, image, or video files, as well as embedded social media content (e.g., tweets). Last, story length was measured with LIWC’s word count, also log transformed (Kim, 2015). Qualitative in-depth interviews Participants The authors conducted 10 separate, semistructured in-depth interviews with journalists and editors handling the social media accounts of different Chilean news media by the authors in Santiago, Chile, between March and May 2016. In the selection process, we considered two criteria: (a) to include personnel from the six websites that were content analyzed; (b) to interview the websites' main competitors in case these were not included already. In the end, we interviewed 10 journalists: Half of them produced content for their website’s social platforms, and half supervised their sites' social media operations. The organizations represented in the interviews are: Emol.cl, LaTercera.com, LaSegunda.com, BioBioChile.cl, Cooperativa.cl, 13.cl, Publimetro.cl, ADNRadio.cl, and ElMostrador.cl. Procedure Eight interviews were conducted in person at the media organizations, one at the university where the authors work, and another over Skype. They recorded all interviews and subsequently the second and third authors transcribed them. The interviews lasted, on average, a little over an hour and followed a semistructured questionnaire inquiring about the organization of social media activities within the company, the role played by different content characteristics in sharing—including, of course, generic frames— and the metrics employed to assess the impact of social media. After going over the script, we presented to each participant the results of the content analysis, including the relationships between framing and news sharing. This way, we gathered from them possible explanations of the framing effects identified in the quantitative analysis. The transcripts were then textually analyzed, searching for common themes. Prior to conducting this second phase of the study, all materials and consent forms were approved by the Ethics Committee of the authors' university. To protect respondent confidentiality, the names of the participants were anonymized. Results The average number of news shares is significantly higher for Facebook (M = 115, standard deviation [SD] = 881) than Twitter (M = 62, SD = 130). This is to be expected, for at the time of the study the ratio of Facebook users to Twitter users in Chile was almost 5 to 1. Still, the distribution of shares within each platform is similar in terms of exhibiting very skewed distributions (see Figure 1). For instance, more than half of the stories have 15 shares or less in Facebook and Twitter. In contrast, only 20% have more than 100 shares in each platform. Their long-tail distribution notwithstanding, sharing patterns are not identical; using a nonparametric measure of rank correlation revealed a moderate association of the dependent variables (Spearman’s ρ = 0.61, p < .001). Furthermore, Facebook shares show a bigger spread than Twitter shares, which is similar to Trilling et al.’s (2017) finding using Dutch data. This means that sharing in Facebook is significantly more concentrated on a relatively fewer number of articles than on Twitter. We will come back to this finding in response to RQ3. Figure 1 Open in new tabDownload slide Distribution of shares on Twitter and Facebook (log–log plots). Figure 1 Open in new tabDownload slide Distribution of shares on Twitter and Facebook (log–log plots). Turning now to generic frames, the most prevalent ones are human interest (44% of the total sample) and conflict (33%). Less common is the morality frame (22%), while economic consequences is the least prevalent of all (16%). Interestingly, only 29% of articles do not employ any of the measured generic frames, whereas 40% used one, 22% used two, and less than 10% used three or all four frames.1 To test the relationship between these frames and news sharing on Facebook and Twitter, we estimated multivariate negative binomial regressions, which are suitable for count variables with overdispersion (truncated at 0 for Facebook, and at 1 for Twitter).2 To facilitate substantive interpretation, we report incidence rate ratios as well as average marginal effects. The results are detailed in Table 1 and in Figure 2. Figure 2 Open in new tabDownload slide Predicted change in Twitter and Facebook shares (average marginal effects). Note: Estimated counts are based on models reported in Table 1. Figure 2 Open in new tabDownload slide Predicted change in Twitter and Facebook shares (average marginal effects). Note: Estimated counts are based on models reported in Table 1. Conflict In response to RQ1, we found that—above and beyond the influence of control variables—the presence of conflict reduces the probabilities that the article will be shared. More specifically, news that are framed as a conflict are shared 45% (Facebook) and 17% (Twitter) less than news without the conflict frame. This means that conflict costs, on average, 70 less Facebook shares and 10 less Twitter shares. This is quite a substantial effect, considering that nearly 85% of the sample has less than 70 Facebook shares and 40% has less than 10 Twitter shares. When presented with the negative effect of conflict, all interviewees said it was in line with their expectations. When asked why, two main explanations emerged. First, that social media users are conflict-avoidant, in that they shy away from sharing content that may trigger confrontations with their own personal networks. This explanation was advanced by three interviewees. One of the editors at 13.cl explained that conflict may well trigger “clicks” (i.e., attention) but not “retweets” (i.e., shares): “People like conflict, but they don't want to become engaged in a conflict, they want to see one (…) But watching is one thing, sharing is another.” A second explanation, raised by two participants, was related to the issues in the news that employ the conflict frame. Thus, an editor at ElMostrador.cl argued that governmental and political affairs exaggerate conflict, and that these topics are not especially entertaining for most users. Her assessment is supported by the content analysis data, which show that 69% of political news employ a conflict frame, compared to 29% in news about other issues. Thus, the interview data are consistent with two mechanisms by which the conflict frame may reduce news sharing: It goes against socializing motivations (i.e., seeking social harmony) and is not emotionally arousing. Economic consequences The presence of an economic frame does not make a statistically significant difference on Twitter, but it does matter on Facebook, for stories with this frame receive 0.62 times as many shares on this platform as articles without it. Put another way, holding all else constant, stories with an economic angle have on average 53 less Facebook shares than stories without this angle, which is in line with H1. Thus, using the same comparison as before, the mere presence of an economic consequences frame could put an otherwise average story from the top 20% to the bottom 80% of the distribution of Facebook shares. When queried about these results, approximately half of our interviewees mentioned that the negative effects of this frame on Facebook shares relative to Twitter shares could be attributed to different user profiles across platforms. In the words of the social media editor at Publimetro.cl: “Twitter has the most informed audience, it’s where you have the early adopters, [elitist] people who care more about the economy. (…) Facebook has a more parochial audience, [people] who don't care about news all the time.” In other words, Facebook users are perceived as less cognitively skilled. To the degree that the economic frame employs technical or abstract interpretation of events—as argued earlier—people with lower skills may have a harder time understanding it and, consequently, sharing it. Nevertheless, two of the interviewees agreed that it is hard to untangle the effects of the economic frame from the issue of the economy, which—at the time of the study—was shared more often than politics, sports, and other well-covered issues in the news. The content analytic data lend some support to this claim, for the correlation between articles about the economy and articles using an economic consequences frame is positive and strong (r = 0.40, p < 0.001). Still, three interviewees claimed that the economic frame may increase sharing in stories that have practical utility—“service stories,” as described by a digital news strategist of ADNRadio.cl. For instance, articles about how to find unclaimed monies held in government agencies or banks tend to employ an economic frame, though its effect on virality is overridden by practical utility. In fact, the content analysis data show that 25% of stories with practical value employ an economic frame, whereas 15% of stories without practical value use it. Thus, both quantitative and qualitative analyses suggest that the economic consequences frame is negatively correlated with sharing in general, though its effect may vary contingent upon other content factors being present in the news. Human interest Estimates of the regression models suggest that the human interest frame is not a significant predictor of news sharing. From a purely statistical assessment, it would suggest that human interest has a negligible role in explaining news sharing. However, the data from the interviews suggest a completely different picture. All interviewees mentioned as examples of typical viral news those that have a human interest angle, especially on Facebook. When asked why they expected the human interest frame to trigger shares, all agreed that it was related to what we refer to as emotional arousal—a key driver of virality, as explained earlier. Two of them also advanced the explanation that human interest stories are shared because users like to share inspirational content by way of life examples, and one connected this result to communicating to others a specific identity: “Facebook is pure empathy. (…) They like to share life examples [or] feel represented, that’s the main thing” (journalist at Emol.com). Part of the divergence between the perceptions and the reality of the role played by human interest on sharing may be explained by the journalists' difficulty in distinguishing among popular stories in terms of attention versus diffusion. As the digital strategist of ADNRadio.cl noted: “There’s a bit of cynicism on all this. Because, if you ask me, what gets more clicks (…), it’s clearly human interest stories.” Still, several of the exemplar stories mentioned by the interviewees that went viral applying a human interest frame were ones that could be regarded as emotional, especially those causing feelings of awe (e.g., a story with the headline “The girls who never grow older”). While the content analysis data do not allow us to test this possibility directly—a proper test would require measuring users' emotional arousal—we nevertheless tested the effects of an interaction between the human interest indicator and LIWC’s affective variables. In both platforms, the interactive term was a positive, statistically significant predictor of sharing. Thus, although this study did not support H2, there is still a possibility that a human interest angle impacts news sharing when it triggers emotional arousal. Morality In stark contrast to the human interest frame, adopting a morality frame increases the likelihood of sharing the article on both Facebook and Twitter quite substantially. The statistical analysis suggests that news with a morality frame is shared on Facebook 75% more often than news without a morality frame. This translates into an average point estimate of 85 Facebook shares, which is more than nearly 85% of the articles in our sample received on this platform. Likewise, on Twitter, news with a morality angle are shared 54% more. This association translates into 27 additional tweets and retweets, more Twitter shares than 60% of the final sample. Thus, H3 received strong support. The morality frame is present in several viral articles related to social and cultural issues, such as a new law prohibiting discrimination against sexual minorities, the promotion of animal rights, and a government agency’s approval of medical marijuana use. A common theme in the responses of most interviewees regarding why they thought the morality frame sparks shares relates to users' impression-management motivations. An editor at LaTercera.com explained that social media users project an identity and want to cause a good impression on others by supporting, for instance, stories on social causes that employ a morality frame: “People [on social media] like projecting their values [and] are politically correct. They like to project an identity. It’s well received that one supports these moral causes or that, this thing—justice—is good.” Another explanation advanced by few respondents was that morally framed news are easier to understand, which harkens back to the discussion of the effect of cognitive skills raised earlier. An interviewee provided the example of news regarding an educational reform to install free tuition in all state-funded Chilean universities. When framed in terms of economic consequences, content about the legal reform became complex and not widely shared. When framed in terms of equality and the right for education, however, the opposite was true. Importantly, in several of the interviews a connection between the human interest and morality frames was highlighted. For instance, at the time, the most viral news story in LaTercera.com was a Sunday feature about the death of an 11-year-old child living in the state-run National Service for Minors' foster care program. As its editor noted, this story went viral because it was morally outrageous that an abused, vulnerable child under state protection died without officials knowing about her health problems. Likewise, an article about a retired school teacher, who was homeless and begging on the streets in a town in northern Chile despite having worked for 35 years in the public school system, was in the 98th percentile of shares in both Facebook and Twitter. This possibility was formally tested by re-estimating the regression models, allowing for a morality by human interest interaction. The results showed a statistically significant positive effect for Twitter shares only, which bolsters the point we made earlier about the contingent effects of the human interest frame. Facebook versus Twitter In response to RQ2, we reported earlier a strong—but not perfect—correlation between Facebook and Twitter shares. Consequently, the effects of generic frames across platforms should be similar, though not identical. This is precisely what we found. The morality frame is associated with a significant increase in both Facebook and Twitter shares, while the opposite is true for conflict. In neither platform is human interest a significant predictor of sharing. The only sizeable difference lies in the economic frame, which is a negative predictor of Facebook shares only. Still, the regressions suggest that generic frames are more influential on Facebook than on Twitter. When extending this comparison to the control variables, we observe that the same topics trigger shares in both platforms. Thus, typical viral topics such as science, animals, and odd news (e.g., “Teen gets 232 teeth removed in India”), and articles written in social media-friendly formats like the listicle, all increase sharing in both Facebook and Twitter. Newsworthy events, such as those with high levels of social relevance and deviance, are also strong predictors of news virality, as are originality, length, and use of multimedia. Discussion Based on the literature on framing effects and social media users' motivations, this study proposed a series of research questions and hypotheses regarding the impact of four generic frames on sharing behavior. To do so, a representative sample of articles shared during a year’s worth of coverage in six Chilean news sites was analyzed to quantify the effects of four frames: conflict, economic consequences, human interest, and morality. In addition, 10 digital journalists and social media editors were interviewed with the aim of seeking possible explanations of the quantitative findings. Both quantitative and qualitative results suggest that the morality frame—a relatively infrequent news angle—is a powerful driver of news sharing, both on Facebook and Twitter. In contrast, the conflict and the economic consequences frames—typical content angles of “hard news” journalism—reduce rather dramatically the number of shares of news stories, especially on Facebook. The evidence for the human interest frame, however, is not as clear-cut, in part because its effect on sharing seems contingent upon the presence of other frames. What is the theoretical explanation of these results? Both the literature review and the interviews suggest three possible mechanisms of influence of generic frames on news sharing. First, frames seem to be an important vehicle for producing emotional arousal, which has been demonstrated to be a potent determinant of news sharing. This is particularly likely with the morality and human interest frames, which interviewees associated with affective responses, whereas the economic and conflict frames, in turn, were associated with less intense feelings. These findings extend to the domain of behavioral responses prior research showing that emotional responses can operate as mediators of framing effects on people’s attitudes and opinions (e.g., Lecheler et al., 2013). So far, behavioral effects have received little attention in framing theory (de Vreese, 2012). Yet, there is no reason why the affective route of framing effects to attitudes cannot be extended to behaviors such as sharing news online. Recent work on which emotional dimension (i.e., arousal or valence) or specific feelings (e.g., hope, anger) drive news virality could be integrated with framing effects to have a more complete understanding of the affective mechanism (e.g., Heimbach & Hinz, 2016). Second, frames can impact people’s likelihood of sharing news on their social media accounts by priming specific motivations and gratifications, such as status-seeking, socializing, entertainment, and so forth. Several interviewees confirmed these expectations. For instance, they mentioned that users are less likely to share conflict-framed news because they generally are conflict-avoidant and, thus, motivated to share what promotes social harmony. The popularity of the morality frame, in turn, was explained by some participants from an impression-management motivation, with users trying to communicate a specific identity to others. These results are consistent with another strand of framing theory, which connects the effects of message frames to specific motivations (e.g., Yan, Dillard, & Shen, 2012). Thus, in addition to—or, perhaps, in combination with—producing specific emotional responses, frames may activate specific motivations related to news sharing. To the degree that gratifications sought from news sharing involve a more conscious process than feelings, the findings of this study would suggest that framing effects on news virality operate through a cognitive—not just emotional—route, as well. The cognitive processing of frames is a well-traversed route in framing theory (Druckman, 2001; Lecheler et al., 2013), and the current study could be extended by exploring the mediating and moderating influences played by specific cognitive processes, such as motivations. Lastly, the unique effects of specific generic frames on sharing news may be explained by differences in attention, recall, and comprehension, that is, psychological involvement with news. For instance, in the interviews, the economic consequences frame was perceived as more abstract and less engaging than the morality frame. To the degree that involvement is a prerequisite for sharing, it is not surprising that the economic frame depresses shares while the morality frame increases them. Nevertheless, we have to be cautious in drawing inferences about which specific frames do or do not promote news involvement, as extant research is rather contradictory (cf., Jebril et al., 2013; Valkenburg, Semetko, & de Vreese, 1999). Thus, a possible future research theme emerging from the current study is the need to study in more detail the nature and significance of the mediators of framing effects on news sharing. The study also found that sharing patterns were largely similar, though not identical, across platforms. We had argued that the platforms have similar affordances, except in terms of scalability and social connections. This could explain why the distribution of Facebook shares is spikier than Twitter’s: Facebook’s algorithm, unlike Twitter, shows each news link to a small number of users, and then only shows it to more users if it generates enough shares, likes, and comments. On the other hand, several interviewees argued that Twitter users, relative to Facebook users, have higher cognitive skills. We cannot test the validity of this assumption, but if true, it may explain the weaker framing effects on Twitter relative to Facebook. The results of this study have important theoretical, practical, and normative lessons. A disjuncture between the frames used by journalists and audiences was characteristic of the broadcast era (Neuman et al., 1992). Our content analysis shows that this is still the case in this postbroadcast, social media era, as suggested by the negative effect of conflict—the most popular frame among news professionals—and the positive effect of morality—a relatively infrequent journalistic frame—on Facebook and Twitter shares. Our results also resonate with the larger “news gap” between online news producers and readers identified by Boczkowski and Mitchelstein (2013). This, of course, need not be the case in the near future; journalists may decide to close this gap by publishing only news that is perceived to perform well in social media, which may increase user engagement at the cost of producing a less informed citizenry. Khuntia et al. (2016) raised the ethical concern of media firms being tempted to “do whatever it takes” (p. 69) to promote user engagement. If news content containing conflict is less sharable, should news professionals stop producing conflict-oriented news even when these may be the most important to produce as democracy’s “watchdog”? Such a scenario could also increase knowledge and participation gaps among voters as well (Bright, 2016). While addressing these concerns is beyond the scope of this study, our results suggest the need for further work on the ethics of journalism in the era of “spreadable content” (Usher, 2014, p. 190). As in any study, there are limitations that merit some attention. The quantitative findings stem from a correlational, not an experimental, study, which limits our ability to identity causal effects of generic frames. The sample does not include stories that were published but never shared on Twitter. While the interview data helped us interpret the content analysis data, they also suggested conflicting evidence regarding the results of the content analysis, as was the case with the virality potential of the human interest frame. Last, the evidence is based on a specific national context and timeframe, which may not necessarily generalize to other countries. Limitations notwithstanding, this study contributes to both framing and news sharing research. It shows that news frames shape social media users' behaviors by determining what content they decide to share with their networks. The results suggest various mechanisms by which news angles influence this diffusion process, including emotional arousal, involvement, and the priming of individual and social motivations. Furthermore, it demonstrates the utility of news framing as a theory capable of explaining sharing, which to date has been mostly studied based on newsworthiness and diffusion of innovations. Future work should expand these findings by including other social media behaviors, such as reading, commenting, and “liking” online news. Other communication theories closely related to framing, such as agenda setting and priming, and alternative methods, such as experimental work, could also be applied to further corroborate and expand our findings. Last, because our research is based in a single country, there is a strong need for conducting replications and extensions in other news media systems to further understand news sharing. Acknowledgments The authors received funding from Chile’s National Commission for Scientific and Technological Research (CONICYT) through grants Fondecyt Iniciación/11140897 and CIGIDEN/Fondap/15110017. The authors acknowledge Katherine Páez for her assistance in programming, and the undergraduate assistants Natalia Alvarado, Mariabelén Briones, Franco Jaramillo, Javier Miranda, Juan Manuel Ojeda, Javier Olivares, and Valentina Proust for their help in the content analysis. Supporting Information Additional supporting information may be found in the online version of this article: Appendix S1. Reliability estimates and descriptives of each variable, factor loadings for four-factor model of news frames, and robustness checks. Notes 1 " To check whether news stories with more than one generic frame behave differently in terms of the sharing patterns reported in the main text, we also tested for all two-way, three-way, and four-way interactions among frames. Of the 22 interactive terms, only three were statistically significant, and all referred to Twitter shares: human interest × morality (positive effect), conflict × human interest (positive effect), conflict × economic × morality (negative effect). Because testing for such a large number of interactions is bound to produce—due to chance—a few statistically significant relationships, the study focuses mostly on main effects (though the human interest × morality and conflict × human interest effects are discussed below). 2 " The results are robust to using the sample truncation point and alternative regression models (Poisson, ordinary least squares with log-transformed Facebook and Twitter shares, and ordinal regressions with Facebook and Twitter shares recoded into the five strata used in the sampling process). References Alhabash , S. , & McAlister , A. R. ( 2015 ). Redefining virality in less broad strokes: Predicting viral behavioral intentions from motivations and uses of Facebook and Twitter . New Media & Society , 17 , 1317 – 1339 . https://doi.org/10.1177/1461444814523726 Google Scholar Crossref Search ADS WorldCat Arendt , F. , Steindl , N., & Kümpel , A. S. ( 2016 ). Implicit and explicit attitudes as predictors of gatekeeping, selective exposure, and news sharing: Testing a general model of media-related selection . Journal of Communication , 66 , 717 – 740 . https://doi.org/10.1111/jcom.12256 Google Scholar Crossref Search ADS WorldCat Bachmann , I. ( 2005 ). Estudio en la prensa económica: La contaminación del discurso periodístico [Study on the economic press: The contamination of journalistic discourse] . Cuadernos.info , 0 ( 18 ), 122 – 131 . https://doi.org/10.7764/cdi.18.152 Google Scholar Crossref Search ADS WorldCat Bartholomé , G. , Lecheler , S., & de Vreese , C. ( 2015 ). Manufacturing conflict? How journalists intervene in the conflict frame building process . The International Journal of Press/Politics , 20 , 438 – 457 . https://doi.org/10.1177/1940161215595514 Google Scholar Crossref Search ADS WorldCat Bandari , R. , Asur , S., & Huberman , B. S. ( 2012 ). The pulse of news in social media: Forecasting popularity. In Proceedings of the Sixth International AAAI Conference on Weblogs and Social Media (pp. 26–33). Dublin, Ireland: The AAAI Press. Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Berger , J. ( 2014 ). Word of mouth and interpersonal communication: A review and directions for future research . Journal of Consumer Psychology , 24 , 586 – 607 . https://doi.org/10.1016/j.jcps.2014.05.002 Google Scholar Crossref Search ADS WorldCat Berger , J. , & Milkman , K. L. ( 2012 ). What makes online content viral? Journal of Marketing Research , 49 , 192 – 205 . https://doi.org/10.1509/jmr.10.0353 Google Scholar Crossref Search ADS WorldCat Boczkowski , P. J. , & Mitchelstein , E. ( 2013 ). The news gap: When the information preferences of the media and the public diverge . Cambridge, MA : MIT Press . Google Scholar Crossref Search ADS Google Scholar Google Preview WorldCat COPAC Bode , L. ( 2016 ). Political news in the news feed: Learning politics from social media . Mass Communication and Society , 19 , 24–48. https://doi.org/10.1080/15205436.2015.1045149 OpenURL Placeholder Text WorldCat Boxer , B. ( 2015 ). Facebook is changing the news game—Are publishers ready? Forbes. Retrieved from: http://bit.ly/Facebook_is_changing_the_news_game Brady , W. J. , Wills , J. A., Jost , J. T., Tucker , J. A., & Van Bavela , J. J. ( 2017 ). Emotion shapes the diffusion of moralized content in social networks . Proceedings of the National Academy of Sciences of the United States of America , 114 , 7313 – 7318 . https://doi.org/10.1073/pnas.1618923114 Google Scholar Crossref Search ADS PubMed WorldCat Bright , J. ( 2016 ). The social news gap: How news reading and news sharing diverge . Journal of Communication , 66 , 343 – 365 . https://doi.org/10.1111/jcom.12232 Google Scholar Crossref Search ADS WorldCat Burscher , B. , Odijk , D., Vliegenthart , R., De Rijke , M., & de Vreese , C. H. ( 2014 ). Teaching the computer to code frames in news: Comparing two supervised machine learning approaches to frame analysis . Communication Methods and Measures , 8 , 190 – 206 . https://doi.org/10.1080/19312458.2014.937527 Google Scholar Crossref Search ADS WorldCat Cappella , J. N. , Kim , H. S., & Albarracín , D. ( 2015 ). Selection and transmission processes for information in the emerging media environment: Psychological motives and message characteristics . Media Psychology , 18 , 396 – 424 . https://doi.org/10.1080/15213269.2014.941112 Google Scholar Crossref Search ADS PubMed WorldCat Creswell , J. , & Plano Clark , V. ( 2007 ). Designing and conducting mixed methods research . Thousand Oaks, CA : Sage . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC de Vreese , C. H. ( 2012 ). New avenues for framing research . American Behavioral Scientist , 56 , 365 – 375 . https://doi.org/10.1177/0002764211426331 Google Scholar Crossref Search ADS WorldCat Díaz , M. , & Mellado , C. ( 2017 ). Agenda and use of sources in headlines and lead stories in Chilean media. A study of the press, online, radio and television news . Cuadernos.info , 0 ( 40 ), 107 – 121 . https://doi.org/10.7764/cdi.40.1106 Google Scholar Crossref Search ADS WorldCat Druckman , J. N. ( 2001 ). The implications of framing effects for citizen competence . Political Behavior , 23 , 225 – 256 . https://doi.org/10.1023/A:1015006907312. Google Scholar Crossref Search ADS WorldCat Ellison , N. B. , & Vitak , J. ( 2015 ). Social media affordances and their relationship to social capital processes . In S. Sundar (Ed.), The handbook of psychology of communication technology . (pp. 205 – 237 ). Boston, MA : Wiley-Blackwell . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Freelon , D. ( 2010 ). ReCal: Intercoder reliability calculation as a web service . International Journal of Internet Science , 5 ( 1 ), 20 – 33 . OpenURL Placeholder Text WorldCat Gabielkov , M. , Ramachandran , A., Chaintreau , A., & Legout , A. ( 2016 ). Social clicks: What and who gets read on Twitter? Proceedings of the 2016 ACM SIGMETRICS International Conference on Measurement and Modeling of Computer Science. (pp. 179–192). https://doi.org/10.1145/2896377.2901462 García-Perdomo , V. , Salaverría , R., Kilgo , D. K., & Harlow , S. ( 2017 ). To share or not to share: The influence of news values and topics on popular social media content in the United States, Brazil, and Argentina . Journalism Studies . Advance online publication. https://doi.org/10.1080/1461670X.2016.1265896 OpenURL Placeholder Text WorldCat Hansen , L. K. , Arvidsson , A., Nielsen , F. Å., Colleoni , E., & Etter , M. ( 2011 ). Good friends, bad news-affect and virality in Twitter . In J. J. Park, L. T. Yang, & C. Lee (Eds.), Future information technology (Vol. 185, Pt. 1 , pp. 34 – 43 ). Berlin, Germany : Springer . Google Scholar Crossref Search ADS Google Scholar Google Preview WorldCat COPAC Harber , K. D. , & Cohen , D. J. ( 2005 ). The emotional broadcaster theory of social sharing . Journal of Language and Social Psychology , 24 , 382 – 400 . https://doi.org/10.1177/0261927X05281426 Google Scholar Crossref Search ADS WorldCat Hedman , U. , & Djerf-Pierre , M. ( 2013 ). The social journalist . Digital Journalism , 1 ( 3 ), 1 – 18 . https://doi.org/10.1080/21670811.2013.776804 Google Scholar Crossref Search ADS WorldCat Heimbach , I. , & Hinz , O. ( 2016 ). The impact of content sentiment and emotionality on content virality . International Journal of Research in Marketing , 33 , 695 – 701 . https://doi.org/10.1016/j.ijresmar.2016.02.004 Google Scholar Crossref Search ADS WorldCat Heimbach , I. , Schiller , B., Strufe , T., & Hinz , O. ( 2015 ). Content virality on online social networks: Empirical evidence from Twitter, Facebook, and Google+ on German news websites. Proceedings of the 26th ACM Conference on Hypertext & Social Media—HT '15 (pp. 39–47). https://doi.org/10.1145/2700171.2791032 Hermida , A. ( 2014 ). Tell everyone: Why we share and why it matters . Toronto, ON : Doubleday . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Hester , J. B. , & Dougall , E. ( 2007 ). The efficiency of constructed week sampling for content analysis of online news . Journalism & Mass Communication Quarterly , 84 , 811 – 824 . https://doi.org/10.1177/107769900708400410 Google Scholar Crossref Search ADS WorldCat Hong , H. ( 2013 ). The effects of human interest framing in television news coverage of medical advances . Health Communication , 28 , 452 – 460 . https://doi.org/10.1080/10410236.2012.693013 Google Scholar Crossref Search ADS PubMed WorldCat Jebril , N. , de Vreese , C. H., Dalen , A., & Albæk , E. ( 2013 ). The effects of human interest and conflict news frames on the dynamics of political knowledge gains: Evidence from a cross-national study . Scandinavian Political Studies , 36 , 201 – 226 . https://doi.org/10.1111/1467-9477.12003 Google Scholar Crossref Search ADS WorldCat Khuntia , J. , Sun , H., & Yim , D. ( 2016 ). Sharing news through social networks . International Journal on Media Management , 18 , 59 – 74 . https://doi.org/10.1080/14241277.2016.1185429 Google Scholar Crossref Search ADS WorldCat Kim , H. S. ( 2015 ). Attracting views and going viral: How message features and news-sharing channels affect health news diffusion . Journal of Communication , 65 , 512–534. https://doi.org/10.1111/jcom.12160 OpenURL Placeholder Text WorldCat Kümpel , A. S. , Karnowski , V., & Keyling , T. ( 2015 ). News sharing in social media: A review of current research on news sharing users, content, and networks . Social Media + Society , 1 , 1 – 14 . https://doi.org/10.1177/2056305115610141 OpenURL Placeholder Text WorldCat Lecheler , S. , Schuck , A. R. T., & de Vreese , C. H. ( 2013 ). Dealing with feeling: Positive and negative discrete emotions as mediators of news framing effects . Communications: The European Journal of Communication , 38 , 189 – 209 . https://doi.org/10.1515/commun-2013-0011 OpenURL Placeholder Text WorldCat Lee , C. S. , Ma , L., & Goh , D. H.-L. ( 2011 ). Why do people share news in social media? In N. Zhong, V. Callaghan, A. Ghorbani, & B. Hu (Eds.), Active media technology: Lecture notes in computer science (Vol. 6890 , pp. 129 – 140 ). Berlin, Germany : Springer . Google Scholar Crossref Search ADS Google Scholar Google Preview WorldCat COPAC Matthes , J. ( 2009 ). What’s in a frame? A content analysis of media framing studies in the world’s leading communication journals, 1990-2005 . Journalism & Mass Communication Quarterly , 86 , 349 – 367 . https://doi.org/10.1177/107769900908600206 Google Scholar Crossref Search ADS WorldCat Neuman , W. R. , Just , M. R., & Crigler , A. N. ( 1992 ). Common knowledge: News and the construction of political meaning . Chicago, IL : University of Chicago Press . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Pennebaker , J. W. , Booth , R. J., & Francis , M. E. ( 2007 ). Linguistic inquiry and word count: A computer-based text analysis program [Computer software]. Austin, TX : LIWC.net. Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Ramírez-Esparza , N. , Pennebaker , J. W., Garcia , F. A., & Suriá , R. ( 2007 ). La psicología del uso de las palabras: Un programa de computadora que analiza textos en español [The psychology of word use: A computer program that analyzes texts in Spanish] . Revista Mexicana de Psicología , 24 , 85 – 99 . OpenURL Placeholder Text WorldCat Rosen , S. , & Tesser , A. ( 1972 ). Fear of negative evaluation and the reluctance to transmit bad news . Journal of Communication , 22 , 124 – 141 . https://doi.org/10.1111/j.1460-2466.1972.tb00140.x Google Scholar Crossref Search ADS WorldCat Schemer , C. , Wirth , W., & Matthes , J. ( 2012 ). Value resonance and value framing effects on voting intentions in direct-democratic campaigns . American Behavioral Scientist , 56 , 334 – 352 . https://doi.org/10.1177/0002764211426329 Google Scholar Crossref Search ADS WorldCat Scholz , C. , Baek , E. C., O'Donnell , M. B., Kim , H. S., Cappella , J. N., & Falk , E. B. ( 2017 ). A neural model of valuation and information virality . Proceedings of the National Academy of Sciences of the United States of America , 114 , 2881 – 2886 . https://doi.org/10.1073/pnas.1615259114 Google Scholar Crossref Search ADS PubMed WorldCat Schuck , A. R. , Vliegenthart , R., & de Vreese , C. H. ( 2016 ). Who’s afraid of conflict? The mobilizing effect of conflict framing in campaign news . British Journal of Political Science , 46 , 177 – 194 . https://doi.org/10.1017/S0007123413000525 Google Scholar Crossref Search ADS WorldCat Semetko , H. A. , & Valkenburg , P. M. ( 2000 ). Framing European politics: A content analysis of press and television news . Journal of Communication , 50 ( 2 ), 93 – 109 . https://doi.org/10.1111/j.1460-2466.2000.tb02843.x Google Scholar Crossref Search ADS WorldCat Shoemaker , P. J. ( 1996 ). Hardwired for news: Using biological and cultural evolution to explain the surveillance function . Journal of Communication , 46 ( 3 ), 32 – 47 . https://doi.org/10.1111/j.1460-2466.1996.tb01487.x Google Scholar Crossref Search ADS WorldCat Shoemaker , P. J. , & Reese , S. D. ( 1990 ). Exposure to what? Integrating media content and effects studies . Journalism Quarterly , 67 , 649 – 652 . https://doi.org/10.1177/107769909006700418 Google Scholar Crossref Search ADS WorldCat Sobieraj , S. , & Berry , J. M. ( 2011 ). From incivility to outrage: Political discourse in blogs, talk radio, and cable news . Political Communication , 28 , 19 – 41 . https://doi.org/10.1080/10584609.2010.542360 Google Scholar Crossref Search ADS WorldCat Stieglitz , S. , & Dang-Xuan , L. ( 2013 ). Emotions and information diffusion in social media: Sentiment of microblogs and sharing behavior . Journal of Management Information Systems , 29 , 217 – 248 . https://doi.org/10.2753/MIS0742-1222290408 Google Scholar Crossref Search ADS WorldCat Treem , J. W. , & Leonardi , P. M. ( 2012 ). Social media use in organizations: Exploring the affordances of visibility, editability, persistence, and association . Communication Yearbook , 36 , 143 – 189 . https://doi.org/10.1080/23808985.2013.11679130 OpenURL Placeholder Text WorldCat Trilling , D. , Tolochko , P., & Burscher , B. ( 2017 ). From newsworthiness to shareworthiness: How to predict news sharing based on article characteristics . Journalism & Mass Communication Quarterly , 94 , 38 – 60 . https://doi.org/10.1177/1077699016654682 Google Scholar Crossref Search ADS WorldCat Usher , N. ( 2014 ). Making news at The New York Times . Ann Arbor, MI : University of Michigan Press . Google Scholar Crossref Search ADS Google Scholar Google Preview WorldCat COPAC Valenzuela , S. , Correa , T., & Gil de Zúñiga , H. ( 2017 ). Ties, likes and tweets: Using strong and weak ties to explain differences in protest participation across Facebook and Twitter use . Political Communication . Advance online publication. https://doi.org/10.1080/10584609.2017.1334726 OpenURL Placeholder Text WorldCat Valkenburg , P. M. , Semetko , H. A., & de Vreese , C. H. ( 1999 ). The effects of news frames on readers' thoughts and recall . Communication Research , 26 , 550 – 569 . https://doi.org/10.1177/009365099026005002 Google Scholar Crossref Search ADS WorldCat Vraga , E. K. , & Bode , L. ( 2017 ). I do not believe you: how providing a source corrects health misperceptions across social media platforms. Information, Communication & Society. Advance online publication. https://doi.org/10.1080/1369118X.2017.1313883 Wasike , B. S. ( 2013 ). Framing news in 140 characters: How social media editors frame the news and interact with audiences via Twitter . Global Media Journal , 6 ( 1 ), 5 – 23 . OpenURL Placeholder Text WorldCat Weber , P. ( 2014 ). Discussions in the comments section: Factors influencing participation and interactivity in online newspapers' reader comments . New Media & Society , 16 , 941 – 957 . https://doi.org/10.1177/1461444813495165 Google Scholar Crossref Search ADS WorldCat Yan , C. , Dillard , J. P., & Shen , F. ( 2012 ). Emotion, motivation, and the persuasive effects of message framing . Journal of Communication , 62 , 682 – 700 . https://doi.org/10.1111/j.1460-2466.2012.01655.x Google Scholar Crossref Search ADS WorldCat Zillmann , D. , & Brosius , H.-B. ( 2000 ). Exemplification in communication: The influence of case reports on the perception of issues . Mahwah, NJ : Lawrence Erlbaum . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Zillmann , D. , Chen , L., Knobloch , S., & Callison , C. ( 2004 ). Effects of lead framing on selective exposure to Internet news reports . Communication Research , 31 , 58 – 81 . https://doi.org/10.1177/0093650203260201 Google Scholar Crossref Search ADS WorldCat © 2017 International Communication Association
TI - Behavioral Effects of Framing on Social Media Users: How Conflict, Economic, Human Interest, and Morality Frames Drive News Sharing
JF - Journal of Communication
DO - 10.1111/jcom.12325
DA - 2017-10-01
UR - https://www.deepdyve.com/lp/oxford-university-press/behavioral-effects-of-framing-on-social-media-users-how-conflict-19ijzYPxbR
SP - 803
VL - 67
IS - 5
DP - DeepDyve
ER -