Political Science PhD candidate focusing on American institutions


Dissertation Project

My dissertation is an empirical study of legislative subsidy, comprised of three papers in which I explore 1) how staffers’ ideology impacts their evaluation of sources of subsidy, 2) ideological coordination among subsidy providers, and 3) actual patterns of subsidy by applying text-as-data and network analysis to a large original corpus of think tank white papers, CRS reports and Committee reports.

Biasing Their Bosses: Staff Ideology, Motivated Reasoning, and the Distortion of Information in Congress. (Job Market Paper) [working paper][supplementary information] [apsa 2019 slides]

The central representational duties of Congress require acquiring and assessing widely dispersed information. In this paper, I present a theory of congressional information processing in which staffers act as information gate-keepers whose own ideological preferences shape the picture of the world they present to their bosses. I present the first systematic empirical test of three competing perspectives on congressional staffer behavior: staffers as faithful agents, staffers as independent agents and staffers as motivated reasoners. I adjudicate between these perspectives using original survey data from the 2017 Congressional Capacity Survey and an experiment embedded in the 2019 Congressional Capacity Survey. I find strong evidence that 1) rather than simply selecting sources that are attitudinally aligned with their bosses, staffers' own attitudes shape how they evaluate and use information, 2) staffers trust and use attitudinally aligned information sources at far higher rates than attitude incongruent sources, 3) this relationship is more pronounced among more ideologically extreme staffers, 4) there is considerable asymmetry in the relationship between ideological extremism and evaluations of internal sources for conservatives and liberals, 5) at least some of these effects appear to be driven by cognitive biases rather than strategic action intended to advance staffer's positions. Together, these results show substantial support for the proposition that staffers act as largely independent agents, exercising considerable leeway to present a biased selection of information to their bosses.

Coalitions and Coordination in Washington Think Tanks: Board interlock among Washington D.C.-based policy research and planning organizations [extended abstract]

This paper tests whether think tanks organize into ideological communities, which, despite the recent theoretical prominence of extended party networks, has not yet been tested. I use IRS 990 data from over 270 Washington D.C.-based policy research institutes and analysis organizations from 2008-2015, to construct the largest board interlock network of policy organizations to date. Taking interlock as a proxy measure for a strong coordination tie between two organizations, I use this network to evaluate several different motivations that think tanks may have for coordinating with each other, including ideological signaling, accruing status, and accessing funding. While this work is ongoing, preliminary analysis suggests a substantial degree of homophily in the policy-planning interlock network, with groups more likely to be connected to their allies, and to other organizations in the same issue areas. In addition to evaluating think tank coordination, this study characterizes an important set of sources of subsidy that staffers draw from as they inform their bosses.

Information Use in Congressional Committees: inferring the flow of policy ideas from white-papers to congressional committee reports

Despite the prominence of the legislative subsidy theory, there is very little direct empirical evidence of this sort of subsidy occurring. The evidence that exists tends to be either circumstantial, anecdotal, or focused on a narrow set of activity (e.g., model legislation). In this paper I find direct evidence of subsidy by using a large, original corpus that I have compiled from more than 100,000 white papers produced by think tanks and research organizations, over 40,000 CRS reports, and 14,000 committee reports from Congress. I use natural language processing tools to detect instances where language from research or policy papers produced by think tanks or the CRS was reused in congressional committee reports. To find instances of language reuse, I have a modified reuse detection algorithm to be more flexible in catching instances of paraphrasing or text adaptation. I then use the set of observed language reuse to estimate a latent diffusion network that allows me to test whether committees rely on sources that are aligned with their chairs and ranking members. These hypotheses are tested using exponential random graph models on the inferred latent diffusion network. I make use of involuntary committee departures to gain causal leverage on this question, testing the changes in the composition of the subsidies that a committee makes use of as a result of the changes in committee leadership.