Welcome! I'm a Visiting Assistant Teaching Professor in the Department of Economics at Florida International University. My primary research interests are in Applied Microeconomics, with focus on Organizational Economics, Labor Economics, and Industrial Organization. My methodological interests include Game Theory, Market Design, Causal Inference, and Machine Learning in Economics.

You can find my CV here.

Research | Teaching

Research

The Art of Waiting [Job Market Paper] (with Vasundhara Mallick)

Revision resubmitted to Journal of Law, Economics, and Organization

Abstract: This paper studies delegated project choice without commitment: a principal and an agent have conflicting preferences over which project to implement, and the agent is privately informed about the availability of projects. We consider a dynamic setting in which, until a project is selected, the agent can propose a project, and the principal can accept or reject the proposed project. Importantly, the principal cannot commit to his responses. In this setting, the agent has an incentive to hold back on proposing projects that the principal favors so that the principal approves a project favored by the agent. Nevertheless, the principal achieves his commitment payoff in an equilibrium of the game in the frequent-offer limit. This high payoff equilibrium showcases the art of waiting and contrasts with Coasian logic: by giving proposer power to the agent, the principal makes it credible to reject his dispreferred projects until later in the game giving the agent an incentive to propose principal-preferred projects earlier on. We apply these results to the economics of organization and merger analysis. In particular, these results suggest a pure strategic gain from giving workers the initiative to pursue their desired projects, and to solicit their ideas "bottom-up" rather than issuing "top-down" commands.

Agenda-Setting with Legislative Precommitments

Abstract: I consider an agenda setting environment where the voters commit upfront to the reforms they are willing to pass, and the agenda setter chooses from among the passable reforms or the status quo. I first characterize the outcomes that emerge in subgame perfect equilibria of this game with majoritarian voting rules. Motivated by the weak predictive power of subgame perfection for this game, I consider a refinement to account for the possibility of coalitional deviations. Compared to the predictions of standard models, the agenda setter's power is significantly reduced in this game, especially in the presence of coalitions and with simple majority rule.

Gender Gaps in Skill Disclosure: The Role of Confidence (with Berk Idem - Working Paper)

Abstract: This paper examines the determinants of skill disclosure behavior using the Burning Glass Institute profiles dataset. We document that reported skills are significant determinants of income even after controlling for education and job experience. However, there exists a substantial gender gap: men report on average 1.4 more skills than women with similar backgrounds. To investigate whether confidence differences drive this gap, we develop a DistilBERT-based language model to assign self-confidence measures to profile text. We find that men's confidence scores are 26% higher than women's on average. Using causal machine learning methods, we show that the variation in confidence explains over half of the observed gender gap in skill reporting. These findings suggest that gender differences in self-promotion, rather than differences in actual skills, contribute significantly to disclosure patterns that may perpetuate income inequality.

Semantic Similarity Metrics for Token Classification Evaluation (with Berk Idem - Working Paper)

(Draft Available Upon Request)

Abstract: Token classification models are essential tools for extracting structured economic data from unstructured text, yet their evaluation remains challenging. Standard exact-match metrics penalize models for economically irrelevant parsing differences (e.g., "machine learning" vs. "Machine Learning"), leading to misleading performance assessments. This paper proposes a semantic similarity-based evaluation framework that uses pretrained embeddings to match predicted entities to ground truth labels via cosine similarity rather than exact string matching, inspired by Earth Mover's Distance. Human-annotated validation on skill extraction tasks shows the metric better captures model quality for economic applications, distinguishing between substantive errors and superficial differences.

Teaching

Florida International University (2023-Present)

Teaching 4+4 course load with three preparations per semester. Course formats include large lectures (120 students), mid-sized sections, and upper-level seminars in both in-person (75% of courses) and online (25% of courses) formats.

Pennsylvania State University (2017-2023)

As Instructor:

As Teaching Assistant:

Bilkent University (2015-2017)

As Teaching Assistant: