|
?:abstract
|
-
The ever growing adoption of digital services, such as Twitter, by political actors, journalists, and the public - for example to advertise positions, comment on politics, or to gage public reactions - raises the question whether data collected on these services can be used to identify and analyze political phenomena. Researchers have used Twitter data to predict election results (Tumasjan et al., 2010), to predict the development of opinion polls (O'Connor et al., 2010), or to identify the political leaning of Twitter user (Barberá, 2014). In this paper we will add to this literature by assessing if Twitter messages can be used to identify the public's political agenda during the run of a campaign, or how topics prominent in messages posted by politically vocal Twitter users diverge from the public agenda identified in representative surveys. For this analysis, we focus on the dynamics of messages commenting on political topics during the campaign for the 2013 federal election in Germany. We will identify the most talked about topics in messages posted by users who had posted at least one message containing a politically relevant keyword during a three-month period preceding the election. We will compare these topics with those mentioned as pressing political topics in a representative survey in the context of the German Longitudinal Election Study (GLES) (Rattinger et al., 2014). This allows us to identify the relationship between political reality as measured in surveys and political reality mediated through Twitter. In addition, we are able to identify specific patterns in the dynamics of tweets commenting on political topics connected with different types of political topics, such as ongoing issue debates, political scandals, or in context of political media coverage. We base our analysis on a data set documenting every message posted by users who, during the run of the campaign, had at least once used one keyword from a list of predefined keywords covering mentions of political parties, prominent candidates, campaign related keywords, and important campaign related media events in various spelling variations. We thus collected all messages posted by 1,248,667 users. We call these users politically vocal Twitter users. In a first step, we used the data vendor Gnip to identify our set of relevant users. We then queried Twitter’s API to collect all messages posted by these users. To establish an agenda based on Twitter messages, we identified the most prominent keywords and hashtags in messages posted by politically vocal Twitter users. We then coded these prominent keywords and hashtags based on whether they referred to topics (political or otherwise) and then identified the most prominent topics referred to by keywords or hashtags. Our results show that measuring the public's political agenda through Twitter or representative surveys led to divergent results. Prominent topics on Twitter were connected with the campaign, and political television programs. Political issues were much less prominent. The most prominent political topics were Internet policy and human rights issues. In contrast, respondents of the GLES survey identified unemployment, the financial crisis, social benefits, and the justice of wealth distribution as the most pressing issues. These results show that Twitter data do not offer a true image of political reality. Instead, an image of political reality emerges that is mediated by characteristics of topics in the political discourse, and attention and interests of Twitter users. Twitter data thereby hold information on political reality but to unlock this potential researchers have to focus on the mediation process leading political phenomena to create traces in aggregates of Twitter messages posted by individual Twitter users.
(xsd:string)
|
![[This page as RDF]](https://data.gesis.org/gesiskg/static/rdf-icon.gif){kind=icon}