Spring 2026 Projects

This project presents a hypothetical research question exploring how immigrants who experience occupational status loss after migrating to the United States navigate changes in identity, belonging, and workplace relationships when transitioning into low-status or gig work. Using an inductive qualitative research approach, the project would build theory from observations, one-on-one interviews, and focus groups to better understand how people rebuild community, recognition, and purpose during these transitions. The project focuses on how shifts in professional status can reshape identity, relationships, and feelings of belonging.

Final Presentation Slides

Incentive structures are used across many contexts in an attempt to guide human behavior toward desired outcomes. External incentives can fundamentally alter the way individuals perceive and respond to a given situation, sometimes even unintentionally encouraging the very behavior they were meant to discourage. Drawing from empirical studies that span sectors such as education, real estate, and labor, this project examines how incentives designed to achieve a certain outcome can instead produce unintended distortions in behaviors. We will analyze the potential causes of these behavioral distortions and determine the characteristics of effective and ineffective incentive structures. 

Final Presentation Slides

We begin by studying the theoretical foundations of causal inference, following a textbook written by Prof. Dylan Small about observational studies. Topics include accounting for biases using sensitivity analysis, as well as adapting to study design methods such as matching and stratification. Later in the semester, we will explore applications of these methods by engaging with field-specific papers.

Final Presentation Slides

Although traditional statistical inference addresses the unilateral problem of testing a single null hypothesis, many modern problems are multifaceted, with multiple variables, outcomes, or datasets being simultaneous objects of study. In this program, we will track the evolution of statistical methods developed to address problems arising from multiplicity---including settings such as global null testing, family-wise error rate control, and false discovery rate control. As we study these methods, we will also place them in context. What real-world problems motivated these settings and procedures? What practical and theoretical improvements did these methods provide over their predecessors? What issues continue to plague statisticians? By doing so, we will unlock a deeper understanding of the current state of large-scale statistical inference.

Final Presentation Slides

This presentation studies algorithmic pricing in peer-to-peer marketplaces, with a focus on how platforms set prices, coordinate market participants, and move toward marketplace equilibrium. We begin with a literature review on algorithmic and dynamic pricing, drawing on prior work related to platform pricing in marketplaces such as Uber and Airbnb. We then frame peer-to-peer platforms as markets with multiple interacting agents, where pricing algorithms must balance supply and demand while accounting for factors such as timing, location, and market conditions.

To ground these ideas, we examine case studies from Airbnb, Uber, and more, highlighting similarities and differences in how pricing operates across short-term rentals and ride-sharing platforms. The empirical portion of the project centers on an Uber/Lyft dataset, where we investigate which observable factors are most useful for explaining ride prices. We first apply linear regression using a small set of basic features and show that this approach has limited predictive power. We then improve the model by incorporating additional variables motivated by prior research and economic intuition, including engineered features such as bad-weather indicators and nonlinear temperature effects.

Finally, we compare the performance of linear regression, decision trees, random forests, and XGBoost using metrics such as R^2 and feature importance. While optimized linear models remain interpretable, tree-based methods allow us to start from a broader feature set and capture more complex nonlinear relationships. Overall, the project offers an exploratory view of how machine learning can help reveal the key components of platform pricing algorithms and why feature importance matters for understanding, not just predicting, marketplace prices.

Final Presentation Slides

A prominent result in recent urban economics holds that expanding luxury housing supply is the most effective policy for improving affordability across all market segments. The mechanism is filtering: new high-end units attract wealthy within-city households who vacate lower-tier homes, cascading vacancy and price relief downward. This paper argues that the key assumption sustaining this result — that city demand is determined solely by existing residents — is unlikely to hold in globally-connected housing markets. When new luxury supply attracts external demand from domestic migrants or foreign capital, outside buyers absorb the new units without vacating existing stock, blocking the filtering chain at its source. I formalise this through a model with tier-specific external absorption rates and propose a framework for estimating these rates from transaction-level data. The framework produces a city-specific filterability profile that identifies at which tier the filtering chain remains intact and at which it breaks down. The goal is a practical policy tool: by gauging out-of-town demand at each level of the housing market, local governments can make targeted supply-side decisions that direct new construction toward tiers where filtering most effectively benefits existing residents.

Over the course of the semester, this project will examine the institutional, legal, and political-economic factors shaping the ongoing development of Central Bank Digital Currencies (CBDCs). We begin by understanding CBDCs in the US political system, asking questions about the incentives for a country to develop a CBDC. From there, the project plans to explore how institutional structures and legal frameworks influence CBDC policy design and implementation. 

This project will take a comparative case study approach, focusing on countries that have piloted or implemented CBDCs to understand how their regulatory environments and relationships may differ or look similar to the US. The goal with this project is to situate CBDC development with broader questions of how institutional differences may produce divergent or similar policy incentives in a rapidly digitizing economy. 

Final Presentation Slides

Differential privacy is a fundamental building block of the modern Internet - from Covid-19 data reports to X-ray medical imaging. Throughout the DRP, we have investigated an intersection between Differential privacy and Conformal prediction, a statistical method for obtaining valid prediction sets under the minimal assumption of exchangeability of data. Specifically, we have perused the "Private Prediction Sets" paper, focusing on the proofs necessary to establish the validity of the procedure. Furthermore, we have applied the methodology to a financial dataset to examine the validity of the method proposed in the paper.

Final Presentation Slides

This project investigates why environmental regulations so often fail to produce the outcomes they promise. It argues that distorted information caused by misaligned incentives within the regulatory system is an important constraint. I describe empirical research that shows how firms, auditors, and government officials each face incentives to distort the signals regulators rely on, resulting in enforcement that operates on systematically inaccurate data. Firm-hired auditors falsify pollution readings, officials concentrate enforcement where they are politically monitored, and consumers bribe emissions inspectors to pass failing vehicles. Rather than improving environmental quality, the result is enforcement that tracks what is observable or rewarded, leaving environmental outcomes largely unchanged or even worse off. These findings suggest that improving information integrity, rather than simply increasing enforcement, is essential to effective environmental regulation.

Final Presentation Slides

This independent study examines how firms operate as political actors in the international

arena—especially under conditions of conflict, war, and national-security regulation. We will

connect normative debates in business ethics (complicity, political responsibility, “business

and peace”) with political economy, management, and legal scholarship on policy risk,

investment strategy, and national security review of corporate transactions.

Over the course of the semester, we hope to cover introductory topics in the field of ranking problems as well as their applications. This study will begin with an overview of topics including preference representations using tournaments, aggregations of those preferences and their limitations as explained by Arrow's Impossibility Theorem. We will study how existence of Condorcet Cycles in tournaments make ranking problems difficult. We will spend the rest of the semester examining its applications to aligning LLMs based on human feedback, which includes modern techniques like RLHF and NLHF. We plan to read through a collection of papers starting with earlier works of Kendall, Moran, and Moon, on paired comparisons and rankings, culminating in state-of-the-art LLM-aligning works such as "Statistical Impossibility and Possibility of Aligning LLMs with Human Preferences: From Condorcet Paradox to Nash Equilibrium."