image

Yuki Atsusaka

Hello! I am a Ph.D. Candidate (political science) and M.A. student (statistics) at Rice University in Houston, Texas. Thank you for visiting my website!

Broadly speaking, I am studying how institutional engineering affects people’s political representation in American politics. Currently, I am working on studying how redistricting, ranked-choice voting implementation, and Vote-by-Mail elections have impacts on minority descriptive representation, racially polarized voting patterns, and voter turnout. Additionally, I am also interested in developing statistical and mathematical tools that can advance our research in related areas (e.g., racial and ethnic politics, minority representation, American political behavior, election reforms). I am currently working on exploring how rank data analysis (as part of computational social science) and quantitatively predictive logical models can advance political science research and legal debates about minority representation.

I draw many ideas and tools from Comparative Politics and Statistics to study American Politics. I like learning about new and practical approaches to solve social science problems and thinking about how we could do better. I would love to network and love to exchange emails if you are interested in similar topics and methodologies. My Curriculum Vitae is available here. I am on Twitter too!

Working Papers

– “A Logical Model for Predicting Minority Representation: Application to Redistricting and Voting Rights Cases” Resubmitted [Logical models + representation]
-“Bias-Corrected Crosswise Estimators for Sensitive Inquiries” (with Randy Stevenson) [poster summary] Revise and Resubmit [Survey methodology]
– “Why Do Vote-by-Mail Elections Boost Voter Turnout?” (with Andrew Menger and Robert Stein) Under Major Revision [Causal inference + election reform]
– “Does Ranked-Choice Voting Reduce Racially Polarized Voting?” (with Theo Landsman) [Rank data + election reform]
– “A Logical Model Approach to Detect Racially Polarized Voting” [Logical models + representation]
– “Race and the Demand for Legislative Responsiveness” (with Matthew Hayes) [Rank data + representation]
– “A Framework for Rank Data Analysis in Political Science” [Rank data]
– “Statistical Methods for Partially Ranked Data with Strategic Rank Concentration” [Rank data]
– “A Pitfall and Solution in Applied Ecological Inference” [Ecological inference]
– “Causal Inference for Ranked Outcomes” (with Michelle Torres) [Rank data + causal inference]

Dissertation

In my dissertation, I work on three essays which intend to advance political science research by providing new statistical methods for analyzing rank and ecological data. The contribution of my work is two-fold. First, it introduces two entirely new fields of methodologies to political science — rank data analysis and optimal transport — which enable researchers to work on many problems in the discipline that they were not able to solve or even think of due to the lack of available tools. Second, it provides extensive empirical illustrations of the proposed methods to facilitate both substantive and methodological research in the new areas. Substantively, my project examines the consequences of an emerging electoral reform — switching from plurality rule to ranked-choice voting — on voters’ preferences and representation in the U.S.. My dissertation is supported by the Electoral Reform Research Group (New America) .

In “Statistical Methods for Partially Ranked Data with Strategic Rank Concentration,” I propose a new methodological framework for analyzing partially ranked data with strategic rank concentration. The proposed approach enables researchers to explain and predict the nature and variation of partial rankings within and across rank data. I also provide new graphical representations of partial rankings, a hypothesis test for strategic rank concentration, statistical models for rank aggregation and clustering in the presence of plumping, a strategy to include individual and aggregate covariates, and a practical guideline for facilitating substantive interpretation. I illustrate the proposed methodology by applying it to data from over 200 ranked-choice voting (RCV) elections in the U.S., attempting to resolve a controversy over the nature, size, and trend of partially ranked ballots in RCV elections in American politics.

In “A Pitfall and Solution in Applied Ecological Inference,” I propose a novel class of ecological inference models for learning about individual-level properties when only aggregated data is available based on optimal transport. Existing methods fail to provide a guideline for and safeguard against common problems which emerge in applied ecological inference such as serious counting errors in ecological data and solutions are left to researchers’ discretion. To resolve this problem, I introduce an ecological inference model based on optimal transport and discuss how to use it along with existing methods to improve ecological inference. Unlike other approaches, the proposed model allows ecological data to contain counting errors in both marginals and also accommodates and learns from spatially correlated errors, which often appear in the construction of ecological data. I derive a fast algorithm for achieving exact inference on model parameters based on the Dykstra-Bregman method.

In “Does Ranked-Choice Voting Reduce Racially Polarized Voting?” (with Theo Landsman), I offer an empirical evidence for the moderation effect of the RCV implementation on the degree of racially polarized voting. To provide a valid causal effect of switching from first-past-the-post (FPTP) to RCV elections on racial polarization, I offer a novel measurement of racially polarized voting that is comparable across the electoral systems based on rank clustering methods and ecological inference. Moreover, I collect precinct returns and cast vote records from Bay Area mayoral elections from 1994 to 2020. With the measurement of racial polarization, I estimate the causal effect of the RCV implementation by dealing with structural missing values (no elections) in the collected time-series cross-sectional data.

Software

  1. cWise: A (Cross)Wise Method to Analyze Sensitive Survey Questions (forthcoming)
  2. p.rank: Alternative Method to Analyze Partially Ranked Data (under preparation)
  3. optEI: Optimal (Transport) Approach to Ecological Inference (under preparation)
  4. rankid: Causal Identification for Ranked Outcomes (under preparation)
  5. logicalmodel: Computing Quantitative Predictions of Logical Models (under preparation)

Teaching

As an instructor, I have taught the Math Prep (aka Math Prefresher) for the first-year graduate students in 2018-2019, Social Analysis and Simulation in R for the second-year graduate students in 2019, and ecological inference in Advanced Political Methodology in 2019. In addition, I have worked (will be working) as a teaching assistant for Causal Inference (Michelle Torres, Fall 2019), Advanced Maximum Likelihood Estimation (Randy Stevenson, Spring 2019), and Machine Learning/Computational Social Science (Michelle Torres, Fall 2020). I am also an organizer of the Methodology Research Group in the Department of Political Science.

Background

I am a Fulbright scholar since 2016 and a founder of SHIPS (Seminar for Hiphop Studies), a study and social group on Black culture in Kyoto, Japan. Before moving to Houston, I studied African American History and Culture and Hiphop in an M.A. program (American Studies) at Doshisha University in Kyoto, Japan. I still got the vision like a line between two dots ÷ !


I will be presenting my work at the PolMeth Online on July 14th. Looking forward to seeing you there!