Twitter and Politics: Identifying Turkish Opinion Leaders in New Social Media

Abstract

Online platforms now provide a valuable medium for political socialization and mobilization. Recent events such as the Gezi Park protests, the Occupy Wall Street, or the anti-government protests in Iran demonstrate how effective social media can be in shaping an individual's political attitudes and actions. Traditional public opinion research does not acknowledge this emerging data source to its fullest extent. In this study, findings from the “Identifying Policy Opinion Shapers and Trends in Turkey” project, which has been collecting and exploring Twitter data to define how the online political debates are shaped in Turkey, are presented. Having identified over ten million active Turkish Twitter users and produced a social network graph of these users, this study identifies public opinion leadership in the Turkish online discussion space. The findings suggest that who these opinion leaders are may not follow the conventional expectations, but these leaders employ various tactics in managing their online presence and disseminating their ideas. This research endeavor as well as the findings suggests that engaging in cross-disciplinary research with scholars from different backgrounds can advance Turkish studies, in terms of both content and methodology. Most importantly, such interdisciplinary research can render significant leverage toward making Turkey more globally salient for scholarly debates.

Disclosure statement

No potential conflict of interest was reported by the author.

Notes on Contributors

Osman Zeki Gökçe is a PhD student of Political Science and the research assistant for the International Studies program at Sabancı University. His research interests include Turkish foreign policy, international conflict, and quantitative research methods.

Emre Hatipoğlu is Assistant Professor at Faculty of Arts and Social Sciences, Sabancı University. His research interests lie at the nexus of domestic politics and international relations. His research has been published in journals such as Foreign Policy Analysis and International Studies Perspectives.

Gökhan Göktürk is an MA student in the Computer Sciences program at Sabancı University. His research interest focuses on data mining and large data analysis in online social networks.

Brooke Luetgert is Assistant Professor at Faculty of Arts and Social Sciences, Sabancı University, and a recipient of the Marie Curie Reintegration Grant. Dr Luetgert's research focuses on automated textual analysis and legislative compliance in the EU and Turkey. Her work has been published in journals such as British Journal of Political Science and European Union Politics.

Yücel Saygın is Associate Professor at Faculty of Engineering and Natural Sciences. He authored numerous articles on data mining, data privacy, and social media. Dr Saygın is the coordinator for the FP7 Project, MODAP, which focuses on data privacy and mining for mobile device users.

Notes

1. O'Connor et al., “From Tweets to Polls.”

2. Stimson, Public Opinion in America.

3. Hibbing and Theiss-Morse, Stealth Democracy.

4. Converse, “Nature of Belief Systems.”

5. Lippmann, Public Opinion. For a review of the massive body of research that engages these questions, see, e.g. Mutz, Sniderman, and Brody, Political Persuasion and Attitude Change; Kinder “Communication and Opinion”; and Druckman and Lupia “Preference Formation.”

6. Page et al., “What Moves Public Opinion?”

7. Song et al., “Identifying Opinion Leaders.”

8. Druckman, “Political Preference Formation.”

9. Our data are neither a full population nor a sample, but rather a “trimmed population.” Excluding dormant (passive) accounts reduced our number to around 10 million users.

10. Grusin, Premediation.

11. Note that, for some time, major media outlets such as CNN, have been supplementing conventional reporting and commentary with reader reporting from the incident locale.

12. Farhi, “The Twitter Explosion.”

13. Sakaki, Okazaki, and Matsuo, “Earthquake Shakes Twitter Users”; Singh and Jain, “Structural Analysis of the Emerging Event-Web”; Yardi and Boyd, “Dynamic Debates.”; Papacharissi and Oliveira, “Affective News and Networked Publics.”

14. Ahmad, “Is Twitter a Useful Tool for Journalists?”; Emmett, “Networking News.”

15. Turkey's recent row with Twitter which started amidst allegations of corruption against four ministers of the incumbent government can also be seen as an example of this “unaccountability” of Twitter as a news source.

16. Ettema, “New Media and New Mechanisms”; McNair, “Journalism in the 21st Century—Evolution, Not Extinction.”

17. Yardi and Boyd, “Dynamic Debates.”

18. Tumasjan et al., “Predicting Elections with Twitter.”

19. Jansen et al., “Twitter Power”; Hamdy, “Arab Media.”

20. Ifukor, “‘Elections’ or ‘Selections’?”

21. Bertot, Jaeger, and Grimes, “Using ICTs.”

22. Jewitt and Dahlberg, “Commentaries.”

23. Ediger et al., “Massive Social Network Analysis”; Papacharissi and Oliveira, “Affective News and Networked Publics.”

24. Yang and Leskovec, “Patterns of Temporal Variation.”

25. Data discussed in this section were collected over the month of December, 2014. While Twitter networks continuously change, no a-priori reason exists to challenge the validity of our findings over short to medium term.

26. The number of connections per Turkish tweeting user was not distributed uniformly, but rather followed a skewed distribution. That is, a few number of users had many connections while a very large portion of users had very little number of connections.

27. Wasserman and Faust, Social Network Analysis.

28. Since practically a user can follow all others on Twitter if she wants to, out-degree centrality, that is how many nodes (accounts) a user follows, does not render a meaningful measurement of centrality. Therefore, we will confine our analysis to in-degree centrality that is how many users follow a given node, only. Note that in other network analysis settings, out-degree centrality may render significant information. For instance, in research on epidemics, to whom an individual reaches out to may have substantial implications.

29. Hanneman and Riddle, Introduction to Social Network Methods.

30. For an example on how Gül's tweet was echoed in a major news channel, see CNN Turk, March 21, 2014, http://www.cnnturk.com/video/turkiye/twitterin-kapatilmasina-gulden-tepki, retrieved on April 10, 2014, for an example on how Gül's tweet was reported in a Turkish daily newspaper, see Sabah Daily, March 22, 2014, http://www.sabah.com.tr/Gundem/2014/03/22/fazla-uzun-surmez, retrieved on April 10, 2014.

31. The number of calculations to compile a list of eigenvector scores is approximately the cube of the number of nodes in the graph. While calculating this figure for small networks (i.e. networks with less than one thousand nodes) is a relatively easy task for a regular computer, our very large network rendered conventional computational methods useless since such methods required about 10²¹ calculations. Instead, the I-POST team utilized advanced mathematical methods and network sampling techniques to approximate eigenvector centrality values for each node.

32. Cheliotis, “Social Network Analysis (SNA).”

33. Studies on Turkey indeed suggest that the Turkish society, and the electorate in particular, are organized along clusters with economic (Keyder, State and Class in Turkey), cultural (Güneş-Ayata and Ayata, “Ethnic and Religious Bases of Voting”; Kalaycıoğlu and Çarkoğlu, Turkish Democracy Today), and geographic (Çarkoğlu and Avcı, “An Analysis of the Electorate”) dimensions. While the I-POST Project employs automated content analysis techniques to identify such clusters on Twitter, this endeavor falls outside of the scope of this study.

34. Friedkin, “Theoretical Foundations for Centrality Measures”; Wasserman and Faust, Social Network Analysis; Rothenberg et al., “Choosing a Centrality Measure”; Kolaczyk, Statistical Analysis of Network Data.

35. These three lists seem to be mostly comprised news broadcasters' accounts, popular figures such as musicians, actors, and actresses, as well as popular content providers tweeting daily content on horoscopes, cartoons, jokes, etc.

36. Email lists were the golden standard for online discussion before the invention of social media platforms.

37. Liu and Zhang, “A Survey of Opinion Mining.”

38. Turney and Littman, “Measuring Praise and Criticism.”

39. Esuli and Sebastiani, “Sentiwordnet.”

40. Wloka et al., “Schlussbericht zum Forschungsprojekt.”

41. See Merton, Social Theory and Social Structure; Weimann, The Influentials.

42. Moody, “Social Network Analysis Lecture Notes.”