Utilising big data to analyze intermedia agenda setting: A comparative analysis about Turkey

The agenda setting theory shaped by McCombs and Shaw in 1972 has been generally tested and validated for many times by elections. Inter-media agenda setting studies conducted have been mainly on how traditional media tools interact with one another. Twitter was included in their studies as a medium since it has been perceived as a specific alternative news source as well as the social media in previous years. The motivation for this dissertation was the restricted geographic area allocated for the inter-media agenda setting studies, the tests made with limited samplings and data sets, and the theories which couldn't be verified in some of the dissertations. Within the current big data ecosystem, the agenda setting power of Twitter can be re-confirmed by multi-national data via using computer aided content analysis methods. This dissertation aims to contribute to multi-national agenda setting studies which are detected to lack in the literature. Thus, the agenda setting effect of the media institution and Twitter and the direction of this effect, if any, will be identified. In this context, for 8 days between 19th and 26th October, 2018 twice a day, at 8.00-10.00 AM for morning data and at 6.00-8.00 PM for evening data, news on Turkey in the English websites of the five print media (Doğru Haber, Hürriyet, Sabah, Türkiye and Yeni Şafak Newspapers) in total, English tweets on Turkey from six neighboring countries of Turkey (Azerbaijan, Bulgaria, Armenia, Iran, Syria and Greece), Russia and Turkey, for control purposes, were used as samplings to measure the inter-media agenda effects. 2,045 news contents have been collected by the Knime program from the print media samplings. In order to measure the agenda setting impact of 39,000 tweets were collected from 8 countries via R Studio program. The tweets and news headlines cleaned out by the R Studio program were processed by the word frequency analysis; thus, codes were created which are classified under specific themes and a glossary (a code book) was formed. The content analysis via this glossary was applied to all media tools both day and night. Collected data was analyzed by using 78 cross-lagged panel correlation and Rozelle-Campbell baseline thresholds in total. When the collected cross-lagged panel correlation was evaluated via rigid approach, the inter-media agenda setting effect couldn't be determined clearly since the autocorrelations among the variables haven't been formed below the Rosella-Campbell baseline in any of the panels. As a consequence of all the analysis made, no inter-media agenda setting effect was detected on media institution agendas or on the agendas collected from the country tweets or between the media institution agendas and the agendas collected from country tweets. The reasons for the hypothesis not be verified are evaluated in full detail in the dissertation. There are many constraints such as high value of the autocorrelations among variables, the existence of the mainstream news, the data constraint of Twitter API and multiple sharing of a single tweet feature of Twitter, and as a sustained structure of Tweeter as an alternative medium may lead to the hypothesis not be verified. If moderate approach would be applied to analyze the data, it was possible to compare two different structures and attain meaningful results thereof. Consequently, it would be better to conduct inter-media agenda setting studies on Twitter in particular about the subjects highlighted by public opinion and to evaluate the agenda setting effects of Trend Topics (Trends for You) on Twitter. Additionally, it would be suggested to conduct further research based on the previous tweets which were ignored by the mainstream media first, and the tweets which were against the dominant discourse, and also the popular tweets posted by accidental journalists.