Co-taught by Professors Kenneth Shepsle and Jeffry Frieden, the Research Workshop in Political Economy (Government 3007) is a year-long graduate seminar that aims to encourage cross-disciplinary research and excellence in graduate training. Political economy is a research tradition that explores how institutions affect political and economic outcomes. The workshop emphasizes the development of dissertation proposals and is a place where graduate students can present their research to an audience of committed and informed peers. It is open to graduate students in the Departments of Government and Economics, and the Program in Political Economy and Government. The workshop holds both internal and public seminars and meetings. At the internal meetings, approximately twelve per semester, graduate students and faculty present their own work to one another. At the public meetings, up to two per semester, leading scholars are invited to Harvard to present their work. Although the workshop is by invitation only, affiliates of the Weatherhead Center are encouraged to attend the public meetings.
Christopher Lucas will present his paper “What’s a Picture Worth? A Method for Jointly Analyzing Text and Images”. Jeremy Bowles will be the discussant.
Pamela Ban will present her paper “The Search for Consensus: Committee Polarization and Voting Behavior in Congress”. Ken Shepsle will be the discussant.
Abstract for "What’s a Picture Worth? A Method for Jointly Analyzing Text and Images"
Visual imagery can frame political news and affect attentiveness to and selection of media. However, researchers largely ignore these images because they are difficult to analyze in an automated way. To resolve this, I develop a new method - the speech and computer image neural network (SCI-NET) - that permits the joint analysis of textual and visual information. Within this framework, I model the network as a gaussian process to generate measures of predictive uncertainty and provide new methods for interpreting neural networks, both important developments for social scientific applications of neural networks. I then illustrate the method with two applications. First, using a new dataset of online articles posted by local newspapers throughout the United States, I learn the distribution of non-fatal officer involved shootings in the United States, which was previously not known. Second, I train the model to learn about the imagery within these news articles about non-fatal shootings, and examine differences by race and media market.
Abstract for “The Search for Consensus: Committee Polarization and Voting Behavior in Congress”
Polarization in Congress has reached record highs in recent times. In this polarized environment, do committee members cooperate to capture gains from trade, or do committees fall victim to partisanship? I examine the rates of committee consensus in floor voting and the conditions under which committees unify in their voting in the presence of internal disagreements. Using a new text dataset of committee hearings and speeches in the 108th-112th Congresses, I measure intra-committee polarization on bills under consideration in the committee stage and empirically investigate the extent to which internal polarization affects consensus in committee voting. My preliminary results suggest that committee integration -- the degree to which committee members cooperate by heading off or resolving the conflicts that arise -- is strong even in a highly polarized Congress or after earmark reform.