Politics and Big Data

Nowcasting and Forecasting Elections with Social Media

Andrea Ceron, Luigi Curini, Stefano Maria Iacus

• Publisher: Routledge
• Year: 2016
• Print ISBN: 9780367194550
• eBook ISBN: 9781317134138
• Topic: Politics & International Relations
• Subtopic: Political Campaigns & Elections
• Index: Politics & International Relations

The importance of social media as a way to monitor an electoral campaign is well established. Day-by-day, hour-by-hour evaluation of the evolution of online ideas and opinion allows observers and scholars to monitor trends and momentum in public opinion well before traditional polls. However, there are difficulties in recording and analyzing often brief, unverified comments while the unequal age, gender, social and racial representation among social media users can produce inaccurate forecasts of final polls. Reviewing the different techniques employed using social media to nowcast and forecast elections, this book assesses its achievements and limitations while presenting a new technique of "sentiment analysis" to improve upon them. The authors carry out a meta-analysis of the existing literature to show the conditions under which social media-based electoral forecasts prove most accurate while new case studies from France, the United States and Italy demonstrate how much more accurate "sentiment analysis" can prove.

1
Social media and electoral forecasts

An overview

Politics, society and social media: a flourishing relationship

The growing number of worldwide citizens with access to Internet, along with the striking growth in their use of social networks like Facebook and Twitter, paves the way to a new potential revolution. In the new “Big Data world”, citizens surf the Web, create their own account profiles and share information online, providing scholars with precious data concerning several areas of study that can potentially yield consequences in the real world (King 2014). Social media and social network sites (SNS) generate a large amount of data – particularly (but not only) textual data – which contain information on the tastes and the opinions of social media users. Thanks to the recent advances in text analysis, new quantitative approaches can be applied to these rich sources of data (Iacus 2014; King 2014). As a result, it becomes easier (with some caveats, as we discuss later) to profitably analyze the comments available on the Internet to enhance our understanding of public opinion, paying attention to elements that impinge on the formation of a climate of opinion as well as on the consequences of such opinions.

Not surprisingly, Internet penetration (in 2015, more than 40% of the world population had access to the Web¹ while SNS usage involved 30% of world population²), goes hand in hand with the interests of scholars. From the role of Web 1.0 (where Internet is merely one mass medium among the others) to that of Web 2.0, in which social media create a networked, unmediated and interactive online environment, the scientific literature has devoted increasing attention to such new avenues of information. So far, several studies have analyzed the relationship between Internet and social media on one side and politics and society on the other. In this regard, seven main streams of research have emerged. These macro-fields try to fulfill a comprehensive assessment of the links between the virtual world of Web and the real world.

The political impact of the Web

A first stream of research looks at the Web as an independent variable and investigates the effect of Internet usage and consumption of online news in order to assess whether, and to what extent, it affects attitudes and behavior of individual users with a particular emphasis on politics (Dimitrova et al. 2014; Stoycheff and Nisbet 2014; Valenzuela et al. 2009). On the one hand, some empirical analyses have attested that the Internet strengthens citizens’ demand for democracy (Norris 2011), commitment to democratic governance (Nisbet et al. 2012; Swigger 2013) and satisfaction toward democracy (Bailard 2012). Scholars have found a positive relationship between the use of e-government websites and trust in government (Tolbert and Mossberger 2006; Welch et al. 2005). Moreover, Internet and social media appear to wield positive effects on political knowledge, political participation and several other indicators of civic engagement (Anduiza et al. 2009; Bakker and De Vreese 2011; Boulianne 2009; Jennings and Zeitner 2003; Kaufhold et al. 2010; Kobayashi et al. 2006; Östman 2012; Tolbert and McNeal 2003) while Internet penetration seems even able to increase voter turnout (Miner 2012).

Other studies, however, have underlined a partially different and less optimistic story. Scholars have, in fact, also found a null or negative impact of the Internet on democratic regimes, showing that the Web does not promote either political knowledge and awareness (Groshek and Dimitrova 2011; Kaufhold et al. 2010; Scheufele and Nisbet 2002) or political participation (Quintelier and Vissers 2008), and can even be associated with lower democratic satisfaction (Norris 2011). For instance Avery (2009) did not find differences in trust in government between citizens exposed to online campaign news compared to others; Kaye and Johnson (2002) noticed that the use of the Web for information-seeking purposes is uncorrelated with trust in government; analogously, McNeal et al. (2008) argued that looking for information on institutional websites does not have a statistically significant effect on political trust. On top of that, Im et al. (2014) attested that citizens who spend more time on the Web display a lower degree of trust (although such negative effect is moderated when surfing governmental websites). Falck et al. (2012) and Campante et al. (2013) even witnessed a negative effect of Internet penetration on turnout. What is more, online communities can produce undesirable consequences for the democratic polity as long as they radicalize (rather than moderate) the positions of their users (Alvarez and Hall 2010; Hilbert 2009; Hindman 2009), becoming a source of ideological lock-ins (Sunstein 2001).

Despite these controversial findings on the effects of Web usage, a meta-analysis of Internet studies analyzing 38 different works has shown that Internet is overall beneficial for democracy, even though this positive effect holds only when the Web is expressly used to gather news and retrieve information (Boulianne 2009).

Results of recent cross-sectional analyses of Eurobarometer data related to 27 countries provided more detailed insights that partially confirm such findings (Ceron 2015; Ceron and Memoli 2016). These studies have revealed that Internet usage, per se, has no effect on different measures of democratic support. However, the consumption of online news can make the difference, even though this effect is positive when users consume news from online traditional media, while consumption of news from social media has a negative effect on the satisfaction with democracy (Ceron and Memoli 2016) and on trust in political institutions (Ceron 2015).

Finally, other studies have considered the Web as a “medium” and evaluated how, by providing citizens with additional information, the Web can impinge on the retrospective judgment about the performance of the government (Bailard 2012), and becomes a source to discover government malfeasance and electoral frauds (Reuter and Szakonyi 2015).

Old versus new media

A second stream of literature compares the new media (SNS) with traditional media in order to evaluate whether these latter media (both offline as well as online) keep their agenda-setting power even after the advent of new media or, conversely, if the Web has the potential to exert such “first-level agenda-setting”, affecting the attention devoted to certain policy issues by mass media (Hindman 2005; Hosch-Dayican et al. 2013; Jungherr 2014; Meraz 2009; Neuman et al. 2014; Papacharissi and de Fatima Oliveira 2012; Parmelee 2013; Sayre et al. 2010; Vargo 2011; Wallsten 2007).

By reducing the transaction costs typical of old media technologies, SNS leaves room to a new bottom-up style of communication (Benkler 2006), providing egalitarian access to the production and the consumption of news that may no longer be elite biased (Hermida et al. 2014; Woodly 2008) that can potentially break the dominance of traditional media (Lewis 2012; Meraz 2011; Meraz and Papacharissi 2012).

In addition to some anecdotal evidence on their agenda-setting power (for a review: Neuman et al. 2014; Wallsten 2007), scholars have been investigating the impact of blogs and interactive social network sites on the media agenda (McCombs 2005; Meraz 2009) to assess whether they have replaced traditional mass media as a source of first-level agenda-setting, holding the ability to influence the attention devoted to a certain issue of the policy agenda.

In fact, some scholars have provided evidence in this respect (Hindman 2005; Lewis 2009; Lloyd et al. 2006). Woodly (2008) argued that blogs affect the selection and presentation of news stories and set the agenda for political elites like journalists and politicians. Similarly, by analyzing the most popular viral video during the 2008 US presidential electoral campaign, Wallsten (2010) reported that bloggers played a crucial role in attracting media coverage. With respect to SNS, Parmelee (2013) indicated that political tweets play a role in agenda building. Analogously, Conway et al. (2015) investigated agenda-setting effects in seven topics related to the 2012 US presidential primary, showing that in six of them Twitter posts by politicians influenced newspaper coverage. Jiang (2014) found that the Chinese SNS Weibo tends to impact the agenda of the state-controlled television CCTV.

Another stream of literature suggests that social media and traditional media are actually sharing the agenda-setting function but neither is the dominant actor in this process. For instance Meraz (2009) focused on the inter-media agenda-setting dynamics between traditional elite newsroom blogs and top independent political blogs, suggesting that traditional media still retain a strong agenda-setting power, but this power is no longer universal since independent blog platforms redistribute it between mainstream media and citizen media: “traditional media agenda setting is now just one force among many competing influences” (Meraz 2009: 701) and social media seem to enhance citizens’ influence in setting the agenda of media. Analogously, Wallsten (2007: 567) found evidence of a “complex, bidirectional relationship between mainstream media coverage and blog discussion rather than a unidirectional media or blog agenda-setting effect” while YouTube too was found to follow and lead mainstream media salience (Sayre et al. 2010). Along the same vein, Neuman et al. (2014) analyzed Granger causality in a variety of issue areas but found mutual and reciprocal causality between traditional media and social media. Finally, Jungherr (2014) found that Twitter messages follow the same logic of traditional news media in some cases, even though not everything that receives media attention reaches a similar level of attention online.

A wide number of studies, however, have retained that mainstream media still affect the salience of issues discussed online. Traditional media seem to set the agenda of Internet-fueled communication tools, such as bulletin boards and chat rooms (Roberts et al. 2002), weblogs (Lee 2007) and online information seeking (Scharkow and Vogelgesang 2011; Weeks and Southwell 2010). Scharkow and Vogelgesang (2011) resorted to Google Insights for Search to measure the public agenda. They highlighted the potential of such data, suggesting that can even be used in “forecasting tomorrow’s news from today’s recipients’ information seeking” (Scharkow and Vogelgesang 2011: 111). Along this vein, another work has explored the relationship between mainstream media and online search activity, showing that media coverage influences Google Trends’ public salience of a particular topic (Weeks and Southwell 2010). Focusing on Twitter, Vargo (2011) analyzed the reciprocal influence between new and old media, showing that traditional media affect public conversations on a large scale while Twitter has only a limited effect. Analogously, Vargo et al. (2015) analyzed the mortgage and housing crisis agendas and the BP oil spill in the United States, showing that traditional media are still in control of the agenda, even though agenda-setting effects on Twitter are not equal in regards to issues and events.

In this regard, Ceron et al. (2016) recently analyzed the Italian political debate on Twitter and newspapers, focusing on two heated political debates that took place between 2012 and 2014 on issues that were particularly salient for voters, media and the political elites: the reform of public funding of parties, enacted between April and July 2012 after the eruption of several corruption scandals, and the debate over the policy of austerity, which played an important role in view of the 2014 European elections. Their results confirmed that online news media keep their first-level agenda-setting power. The attention devoted to an issue by online media outlets influences the SNS salience of that issue and drives Twitter users to discuss it.

However, there is a difference between being able to affect what the public thinks about (first-level agenda-setting) and being able to influence how the public thinks about that issue (second-level agenda-setting). In other words, the (possible) enhanced attention built by online media outlets does not imply that they exert a second-level agenda-setting in influencing Internet users (Meraz 2011). For example going back to the previous quoted study, it has found a citizen-elite gap in the degree of anti-politics/anti-austerity sentiment expressed on Twitter, which is remarkably higher if compared to the level of negativity observed in the frame of online news stories (Ceron et al. 2016).

Collective action and public policy

A third stream of literature links social media with collective action and public policy. Under the idea that the Web can decrease the cost of political mobilization, scholars have examined the role of social media in promoting radical protests (Bastos et al. 2015; Calderaro and Kavada 2013; LeFebvre and Armstrong 2016; Segerberg and Bennett 2011) or uprisings (Cottle 2011; Hermida et al. 2014, Howard and Hussain 2011; Meraz and Papacharissi 2012; Shirky 2011; Tufekci and Wilson 2012).

For instance Howard and Hussain (2011) argued that the new digital information technology played a major role in the Arab Spring and social media have changed the tactics of democratization movements. Indeed, with respect to the Egyptian case, Tufekci and Wilson (2012) showed that people learned about the protests primarily through interpersonal communication using Facebook (along with phone contact or face-to-face conversation) and social media use increased a citizen’s probability to attend the protests. Analogously, Benn...

Table of contents

• Cover
• Title
• Copyright
• Contents
• List of figures
• List of tables
• Introduction
• 1 Social media electoral forecasts: an overview
• 2 From noise to signal in sentiment and opinion analysis
• 3 Nowcasting and forecasting electoral campaigns: evidence from France, the United States and Italy
• 4 Leaders, promises and negative campaigning: digging into an electoral campaign through social media
• 5 Social media and electoral forecasts: sources of bias and meta-analysis
• 6 Conclusion: “To predict or not to predict?” Future avenues of social media research within and beyond electoral forecasts
• Postscript
• Index