Schranz, Marc (2025). Essays in Applied Econometrics. (Thesis). Universität Bern, Bern
|
Text
25schranz_m.pdf - Thesis Available under License Creative Commons: Attribution (CC-BY 4.0). Download (1MB) | Preview |
Abstract
As economic data becomes more abundant and diverse, and new technologies grow more powerful and capable, economic analysis increasingly relies on the application of innovative methodologies to enhance our understanding of complex questions across various economic topics. In the following three chapters, I contribute to these developments by implementing innovative approaches in the fields of policy evaluation, income mobility, and monetary policy analysis, employing advanced statistical techniques, distributional modeling, and Natural Language Processing. All three studies emphasize the importance of moving beyond traditional aggregate metrics to capture more granular distributional dynamics and to improve upon past measures. The first chapter introduces an extension to the synthetic control method, which was originally proposed by Abadie and Gardeazabal (2003) and Abadie, Diamond, et al. (2010), to evaluate effects across the distribution. Traditional synthetic control methods focus on average treatment effects, even when the treated and control units comprise a sizable number of individual entities. This chapter proposes a distributional synthetic control utilizing information that is available on a finer granularity and capturing heterogeneous effects across different thresholds of the cumulative distribution function. The second chapter, which is joint work with Jonas Meier, proposes a new estimator of the conditional distribution of multivariate outcomes given covariates. The estimator builds on the Local Gaussian Representation from Chernozhukov, Fernández-Val, and Luo (2023) and employs distribution regression and a conditional copula with a copula parameter that is local in the value of the outcome. The proposed method allows for flexible, semi-parametric control of covariates, enabling the analysis of multivariate counterfactual distributions. The third chapter is joint work with Alexandra Piller and Larissa Schwaller. The chapter explores the increasingly critical role of central bank communication in monetary policy using state-of-the-art Natural Language Processing techniques. In recent years, high-frequency monetary policy surprise series have been used as external instruments to identify monetary policy effects. This chapter improves upon these surprise series by employing a Natural Language Processing model that is based on transformers, an architecture introduced in a groundbreaking paper by Vaswani et al. (2017). The model is trained to isolate the component of the surprises driven solely by central bank communication. This further required the creation of a text dataset comprising statements and speeches issued by the Federal Reserve Board through web scraping. In conclusion, the three studies collectively advance our understanding of complex economic debates through innovative analytical approaches. From distributional synthetic control methods and copula regression to natural language process-driven monetary policy analysis, the three chapters demonstrate the potential of these novel analytical tools in addressing challenging economic questions. By leveraging these advanced methodologies, the research bridges gaps in existing literature and offers insights for policymakers, academics, and practitioners.
| Item Type: | Thesis |
|---|---|
| Dissertation Type: | Cumulative |
| Date of Defense: | 22 May 2025 |
| Subjects: | 300 Social sciences, sociology & anthropology > 330 Economics |
| Institute / Center: | 03 Faculty of Business, Economics and Social Sciences > Department of Economics > Institute of Economics |
| Depositing User: | Hammer Igor |
| Date Deposited: | 15 Jul 2025 16:37 |
| Last Modified: | 16 Jul 2025 12:07 |
| URI: | https://boristheses.unibe.ch/id/eprint/6382 |
Actions (login required)
 |
View Item |