Applications of Latent Dirichlet Allocation (LDA) in political science rarely leverage information from the full posterior topic mixture distributions underlying the generation of documents in a corpus. Rather much of the extant literature employing this approach simply focuses on the maximal probability topic for every document. We present an approach to help solve this problem, and validate our method via an exploration of Congressional rhetoric and its relationship to party power. Congressional rhetoric — the content of floor speeches — is a clear manifestation of well-studied concepts such as party organization and agenda setting, yet remains largely understudied. Based on the text of speeches from the 104th through 109th Congresses, we use LDA to model speech content, and create a Euclidean space for speeches in Congress via a Principal Components Analysis (PCA). We then predict 1) the amount of money given to a candidate by their party committee and 2) prestige committee appointment based on how much their oor speeches resemble those of the rest of their party. Consistent with findings in American politics, we show that speaking like the rest of one’s party is associated with receiving more party goods.

I present an approach to analyze agenda control on and before the floor in the House and Senate based on estimates of how members would have voted on bills that were killed before the floor (pre-floor bills). These votes are estimated using quantitative representations of a bill's policy content derived from its text with doc2vec, a class of vector embedding models uncommon to the political science literature. I find evidence of strong negative and positive agenda control, the latter of which has received less attention due to measurement limitations. I also find, contrary to recent findings but in keeping with our institutional understanding of Congress, that the majority party in the House has more control over the agenda than the majority party in the Senate. I close by discussing how this approach can lead to a fuller understanding of agenda control across legislatures, propose a standard for measuring agenda control, and discuss how agenda control can help us more fully understand relationships within and between parties.

While members of Congress need the party’s brand to be effective, party and member interests sometimes diverge. I present an account of policymaking that explains how parties react to these situations, attempting to extract “Yea” votes necessary to pass bills from members who want to vote “Nay,”: Public statements act as a signal of voting intention that helps hold members to their intended vote. Using a novel dataset of public statements made by members of Congress, I describe this process using recent salient health care bills: the Affordable Care Act and the America’s Health Care Act. When parties convince reluctant members to vote with the party, those members are more likely to make a public statement in favor of the bill. Further, I show that these statements are precommitment devices to voting with the party, separating this purpose from credit claiming or otherwise explaining their decisions to their constituents.

Estimating the Trump Bump. With Hans Hassell and Michael Heseltine

While some research has analyzed whether campaigning for candidates affects important outcomes like turnout, campaign finance patterns, and voting, recent changes in the political sphere have yet to be studied. Particularly, does President Trump's Twitter usage affect Congressional elections, and is any effect separable from actually visiting with candidates to campaign for them?

Candidate Emergence in the 2018 Midterm Elections. With David Rohde and Katelyn Mehling

In this project, we build on recent trends in Congressional elections and, looking forward to the 2018 midterms, investigate candidate emergence. We analyze how trends in candidate emergence have changed in recent years, and how these changes can lead to different electoral outcomes. We focus particularly on the rise of female, minority, and amateur candidates on the national stage.

Civility of Elite Discourse in the United States. With Spencer Dorsey and Michael Heseltine

Pundits and public opinion both suggest that divisive rhetoric in public discourse is on the rise. But how prevalent are these trends among elected officials, and who is driving these changes, members of Congress or the public? We study these questions using a novel measure of political divisiveness derived from supervised machine learning methods applied to tweets and Facebook posts.

Giving the Minority Credit: Restrained Majority Party Agenda Control in the House of Representatives. with James Curry

Recent research has found that the minority party is all but shut out of the agenda-setting process in Congress. At the same time, about three-quarters of bills pass with minority party support and the prevalence and success of minority-introduced bills has increased substantially in recent decades. So what is the role of the minority party in Congress? And more broadly, what is the difference between agenda-setting and policymaking?

Legislative Voting Behavior and Theories of Party Operations. With John Aldrich and Jacob Montgomery

Recent research has suggested that partisan polarization in Congress is at least partly an artifact of the scaling methods used on votes to derive ideology scores, and that Congress actually operates in a much higher-dimension space than we previously thought. However, there has been no application of this finding to test how well theories of party behavior and operation fit with roll call data. We test between these theories using both simulated and real roll call data and changes to common scaling techniques that allows for higher-dimensional behavior.