This thesis contains three chapters that explore the nature of political communication and public opinion formation by analyzing social media data. In terms of content, I test how major Supreme Court cases affect public opinion and discussions of policy issues (chapter two), the differences between partisan social networks and their implications for information dissemination and coordination (chapter three), and how elite message flows affect discussions about guns. politics (chapter four).
POLICY CHANGE AND PUBLIC OPINION: MEASURING SHIFTING POLITICAL SENTIMENT WITH SOCIAL
MEDIA DATA
Introduction
By implementing machine learning algorithms to extract sentiment measures from a large collection of tweets before and after the Supreme Court ruling, I analyze a fine-grained dataset that enables new insights into the short-term dynamics between Supreme Court rulings and public opinion. opinion. Studying this relationship is crucial to discovering whether or not the judiciary is able to sway public opinion toward their views, or rather acts as a catalyst to further divide the public.
The Supreme Court and Public Opinion
Without the ability to enforce its decisions, a number of scholars have argued that public opinion constrains the Supreme Court (Hall, 2014). Given the empirical support for the structural response hypothesis, I predict that, overall, the Supreme Court will polarize public opinion.
Twitter Data and Sentiment Scoring
The population of U.S. Twitter users is not a representative sample of the adult population in the United States, and research on the demographic composition of Twitter users shows that densely populated U.S. counties are often overrepresented (Mislove et al., 2011). probably younger and richer (Barberà and Rivero, 2015), and generally shows a liberal and pro-democratic bias compared to the country as a whole (Mitchell and Hitlin, 2013). To test the largest number of messages in the shortest time, the validation set corresponds to the top 3,000 most retweeted messages, a set that represents 39.6% of the data.
Aggregate Shift in Public Opinion
Impact of Policy Change
The treated coefficient is negative and significant in the baseline models, which I expect since the treated states contain more conservative citizens. This interpretation requires the belief that the untreated states are a good counterfactual to what might have happened in the treated states.
Conclusion
Public Resistance and Public Attitudes: Is Political Progress on Gay Rights Counterproductive?” In: American Journal of Political Science 60.3, p. Sentiment Analysis on Twitter: The Good, the Bad and the OMG!” In: Proceedings of the Fifth International Conference on Weblogs and Social Media, Barcelona, Catalonia, Spain, July.
THE PARTY STRUCTURE: EXAMINING HETEROGENEOUS PARTY NETWORKS
Twitter Data and Networks in Political Science Parties-As-Networks
Early work in the 'parties as networks' literature involved the construction of local partisan networks through interviews and observations. For example, Twitter data allows previously undetectable members of the party electorate to be included in the party structure.
Data and Methods Twitter DataTwitter Data
This turned each user's tweets into a list of separate words, ignoring the original order of the words in the sentence. I also refer to components, a subgraph of a network where any two nodes can be reached along a path.
Polarization in Networks
Greater polarization in the retweet network makes intuitive sense, as Twitter users are far more likely to retweet the message of a user they agree with. It is thus possible that a subset of connections in said network correspond to users arguing with users on the other side of the ideological spectrum.
Differences Between Democratic and Republican Networks
Furthermore, in the said network, there are almost 20 percent more Democrats who are isolated from the network. Overall, there appears to be a large number of distinct conversational networks in the Democratic mention network.
Elite and Non-Elite Party Networks
In Table 3.6, it is clear that the top hashtags used by Republican Twitter users consistently represent a larger percentage of the total number of hashtags used. These networks show all the same trends as Table 3.4, indicating a denser republican network structure. 24.
Conclusion
25I specifically take the log of the number of users in each community as the relative weight. This general pattern makes sense because as I reduce the size of the network, I reduce the total possible number of edges in the network (the denominator of the edge density statistics).
Core Analysis
In: Journal of Statistical Mechanics: Theory and Experiment2008.10. url: https://github.com/fchollet/. 2008).The Paty decision: Presidential nominations before and after reform. Paper presented at the annual meeting of the American Political Science Association, Washington, DC.
WORDS AND WEAPONS: ANALYZING REACTIONS TO GUN VIOLENCE WITH A SOCIAL MEDIA PANEL
Introduction
Beyond simply testing theories of public opinion in a new setting, using social media data allows for methodological innovations that can better evaluate key components of the RAS model, yielding a richer and more robust understanding of American public opinion. With a direct estimate of each user's incoming message flow and the ability to observe exactly when a particular user engages with a problem topic, social media data allows me to draw better inferences with the RAS model than traditional cross-sectional survey methods.
Theories of Mass Opinion Formation and Activation
But there is scant evidence that after a mass shooting, elite messages lead to a change of heart. Instead, I seek to find partisan actors who “activate”—citizens who discuss gun control as an important policy in the wake of mass shootings.
Data and Methods
After a mass shooting, a partisan who receives more messages from elites in the same party is more likely to tweet about gun control. The messages of the opposing party's elites are less likely to be accepted and will not be accompanied by more gun control partisans sending messages.
Testing The RAS Model
This indicates that the number of elite gun policy messages a user receives is positively correlated with the likelihood that the user sends their own gun policy message. In the case of the Las Vegas shooting, going from no elite gun control tweets to ten elite tweets (median) increases the probability that a user will send their own gun control tweet on a given day excluding retweets) and in the case of Parkland, increases the probability with the exclusion retweets).
Expanding the Notion of Elite
Moreover, I find evidence that Democratic elite messages at the one-hour lag explain trends in the Republican message stream, which is not what I would expect in the RAS model. Thus, I may be underestimating verified elite actors' impact in driving conversations in the following analysis.
Conclusion
Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas Vegas. Sutherland, Sutherland, Sutherland, Sutherland, Sutherland, Sutherland, Sutherland dežela Sutherland Sutherland Sutherland Sutherland Sutherland Sutherland Sutherland Sutherland Sutherland Sutherland, Sutherland, Sutherland, Sutherland, Sutherland, Sutherland, Sutherland dežela Sutherland Sutherland Sutherland Sutherland Sutherland Sutherland Sutherland Sutherland Sutherland Sutherland, Sutherland, Sutherland, Sutherland, Sutherland, Sutherland, Sutherland dežela Sutherland Sutherland Sutherland Sutherland Sutherland Sutherland Sutherland Sutherland Sutherland Sutherland, Sutherland, Sutherland, Sutherland, Sutherland, Sutherland, Sutherland dežela Sutherland Sutherland Sutherland Sutherland Sutherland Sutherland Sutherland Sutherland Sutherland Sutherland, Sutherland, Sutherland, Sutherland, Sutherland, Sutherland, Sutherland dežela Sutherland Sutherland Sutherland Sutherland Sutherland Sutherland Sutherland Sutherland Sutherland Sutherland, Sutherland, Sutherland, Sutherland, Sutherland, Sutherland, Sutherland dežela Sutherland Sutherland Sutherland Sutherland Sutherland Sutherland Sutherland Sutherland Sutherland Sutherland, Sutherland, Sutherland, Sutherland, Sutherland, Sutherland, Sutherland dežela Sutherland Sutherland Sutherland Sutherland Sutherland Sutherland Sutherland Sutherland Sutherland Sutherland, Sutherland, Sutherland, Sutherland, Sutherland, Sutherland, Sutherland dežela Sutherland Sutherland Sutherland Sutherland Sutherland Sutherland Sutherland Sutherland Sutherland Sutherland, Sutherland, Sutherland, Sutherland, Sutherland, Sutherland, Sutherland dežela Sutherland Sutherland Sutherland Sutherland Sutherland Sutherland Sutherland Sutherland Sutherland Sutherland, Sutherland, Sutherland, Sutherland, Sutherland, Sutherland, Sutherland dežela Sutherland Sutherland Sutherland Sutherland Sutherland Sutherland Sutherland Sutherland Sutherland Sutherland, Sutherland, Sutherland, Sutherland, Sutherland, Sutherland, Sutherland dežela Sutherland Sutherland Sutherland Sutherland Sutherland Sutherland Sutherland Sutherland Sutherland Sutherland, Sutherland, Sutherland, Sutherland, Sutherland, Sutherland, Sutherland dežela Sutherland Sutherland Sutherland Sutherland Sutherland Sutherland Sutherland Sutherland Sutherland Sutherland Sutherland Sutherland Sutherland Sutherland Sutherland Sutherland Sutherland Sutherland Sutherland Sutherland Sutherland Sutherland Sutherland Sutherland Sutherland Sutherland Sutherland Sutherland Sutherland Sutherland Sutherland Sutherland Sutherland Sutherland Sutherland Sutherland Sutherland Sutherland Sutherland Sutherland Sutherland Sutherland Sutherland Sutherland Sutherland Sutherland Sutherland.
CONCLUSION
