Main content start

Rebecca Weiss

The Impact of Partisan Identity Salience on Self-Reported Political Behaviors
2012–13 Survey Lab Project

Though it is commonly accepted that elite polarization is a real phenomenon, meaning policy position extremism among politicians is occurring and increasing over time (Hetherington), similar issue-based polarization amongst the mass electorate has not been demonstrated (Fiorina). Most evidence indicates the opposite; that American partisans have remained rather stable with respect to the strength of their policy positions, measured as self-placement on an issue in a 1-dimensional space.  Recent research has provided evidence that observed polarization is actually an result of the strength of partisan identity, and that growing animus between partisans is largely a result of social identity activation.  In my research, I focus on how the manipulation of political identity increases or decrease the self-report of positions on key issues and on reported media consumption in order to identify the effect of identity salience on self-reported political behaviors.

Distortions from Reality: A Computational Approach to the Study of Issue Salience and Coverage in News
2012 CSS Fellowship

The news media are an essential institution for a well-functioning democracy, informing the people about national issues so that they can make informed political decisions. Therefore, understanding the process by which news content is generated is highly important to society, particularly given the accusation that news content is intentionally distorted from reality in order to lead the people to interpret issues and events in a politically strategic way. However, an alternate account for news content can be found in the market model of news, where a news outlet’s economic considerations predominantly drive editorial decisions in generating content. These accounts both presume that the following qualities indicate distortion in the news: 1) proportion of coverage and 2) elements of language style. I intend to use computational methods and tools devised for investigating large-scale text corpora in order to validate or reject these theories. A computational approach towards evaluating distortion in news content is superior to previous attempts to study this issue, which relied upon the use of small samples of news data. This is because studying news data of an appropriate size for this question would be infeasible for humans, a shortcoming that computational methods lack. Furthermore, advances in natural language processing, both at the sentence- and document-level of analysis, allow for the quantitative study of coverage and style in news content.