Applied Statistics

Language and Legislation: How the Branding of Legislation Spreads Throughout the Chambers

(with Mark Hill and Jaclyn Kaslovsky)

Brands matter. For example, the phrase “death panels,” which was used to described a minor portion of the Affordable Care Act, was used by Republicans to easily mar- ket their anti-reform position. The brand caught on, and debate over death panels nearly derailed President Obama’s effort at reforming the health insurance system in the United States. While incredibly effective, little is known about the sources of these legislative brands and how they spread into the legislative discourse. In order to measure the diffusion of branding in Congressional debate, we combine a network theory of legislative power with an automated text approach. Using floor speeches scraped from the Congressional Record, we plan to generate a network of text-reuse across legislators to measure the creation and evolution of a specific brand/terminology for a given piece of legislation across Congresses. We hypothesize that the branding starts with party leaders, and proceeds to diffuse through the party hierarchy. We conceptualize “branding” to mean the repeated use of, or plagiarism, of language in Congressional speeches. This analysis will provide further insight into power dynamics within the party system, giving us a window into how representatives choose to market their policies to constituents and other leaders.

(with Albert Rivero)

Precedent is perhaps one of the cornerstones of the US litigation system. While existing studies measure this concept through a specific case’s position in a case citation network, these types of measures fail to incorporate all relevant information into measuring the precedential value of a case. We combine insights from the literature on network analysis and natural language processing to generate new measures of the precedential value of US Supreme Court cases.

Text Analysis of International Law

(with Eric Arias)

An increasing amount of international law manifests itself on paper. In this project, we seek to characterize the evolution of international law within and across various regimes (eg, trade, investment, and human rights) and the consequences that they may have on state behavior. To do this, we utilize a number of techniques from the field of text analysis and natural language processing.


  • Identification in Regression Discontinuity Designs with Aggregated Outcomes
  • Strengthening Weak Instruments: A Machine Learning Approach
  • Estimating Political Preferences from Historical Text