Research
Theses, projects and commentary, newest first. Abstracts open where there is one.
As of
Theses
-
The Fateful Adventures of the Good Soldiers: Loyalty, Desertion, and the Czechoslovak Legions
Abstract
Wartime desertion is typically studied in conscript armies, where staying may reflect fear rather than loyalty. This paper proposes to exploit the Czechoslovak Legions, a volunteer force formed from emigrants, prisoners of war, and Habsburg army deserters during World War I, as a cleaner setting for separating genuine commitment from coercion. The first part of the analysis examines what drove Czechoslovak soldiers to desert the Habsburg army and join the Legions, with soldiers’ nationalist identity and political views as the mechanisms of interest alongside those established in the existing literature. The second examines why some soldiers subsequently deserted the Legions themselves. Because enlistment was voluntary, this setting substantially reduces the initial coercion confound and allows a cleaner study of what sustained or weakened commitment. The study draws on two datasets totalling approximately 320,000 individual soldier records not previously used in economic research. More broadly, the findings could speak to the literature on organisational loyalty and identity, and the role of ideological motivation in sustaining commitment under high stakes.
The thesis is in progress; the project documentation describes the data and how it is built.
-
The Laffer Curve in the Czech Republic
Abstract
Taxation is a frequent and important topic of discussion in politics, economics and the public debate as well. This thesis analyses the overall tax system, the Personal Income Tax and the Corporate Income Tax from the Laffer Curve perspective with the objective of finding a revenue-maximising tax rate for each of the mentioned. Because of the non-stationarity of some data, two branches of models are estimated. One takes the stationarity as an assumption and finds the revenue-maximising rates at 28.6% for the overall tax system, 17.5% for the Personal Income Tax, and 27.5% for the Corporate Income Tax. The second branch finds linear relationships between the growth rates of the tax revenues and the growth rates of the tax rates, which are all stationary. Three methods are utilised for the estimation, Ordinary Least Squares, Two Stage Least Squares with Instrumental Variables, and the Generalised Method of Moments. The thesis uses quarterly data from 1996 Q1 to 2023 Q4. Comments on other determinants of tax revenues accompany the main findings. The research presented in this thesis provides the latest empirical evidence on the relationship between tax rates, macroeconomic factors, and their impact on tax revenues in the Czech Republic.
Term papers and projects
-
What Drives the Referees’ Performance? The Effect of Referee Coaches on Referees’ Performance
Finished as a term paper. I may return to it with stronger identification and more data.
Abstract
Referees in any sport can substantially influence match outcomes. Therefore, the natural aim of every sports association is to maximise referee quality. The current literature lacks empirical evidence on the effects of various training methods, especially coaching referees. I estimate the effect of cumulative previous observations by a referee coach on the referee’s received grade. Using a novel panel dataset with multiple controls and referee, coach, and referee-pair fixed effects, constructed from both publicly available and internal data from Czech Floorball, combined with one Regional Referee Committee dataset created from Referee Coach Reports, I find a statistically significant effect of previous observations with diminishing returns up to the fifth observation. The cumulative magnitude after the initial five observations improves (lowers) the grade by 0.24 grade points, reaching 30% of the Standard Deviation in the referee’s received grade. This represents a meaningful improvement in referee quality over a relatively short time, as it takes just five matches to enhance referees’ skills in the same way. In comparison, the similar effect could be achieved with three to four years of actual refereeing experience. Moreover, theoretical and practical seminars yield no statistically significant effect.
-
Written in Time: Unsupervised Discovery of Emotional and Stylistic Structure in the Project Gutenberg Corpus
Finished as a term paper. I may return to it with stronger identification and more data.
Abstract
This project explores whether machine learning and quantitative text analysis can identify emotional, stylistic, and thematic patterns in literature across time. Using the Project Gutenberg (n.d.) corpus, I construct a feature-level dataset of books based on emotion scores, stylometric measures, and semantic fields. The analysis first examines descriptive genre-level differences and then residualises the extracted features on decade, country, and genre fixed effects. Subsequently, I apply clustering and network analysis to map similarities between books and authors in the constructed feature space. Finally, I study temporal changes around selected historical shocks, with World War I serving as the headline case. The results suggest that literary texts contain measurable patterns connected to genre, period, and historical context. In particular, the WWI shock is associated with increases in expressed fear, anger, and conflict-related vocabulary, as well as shifts towards more conflict- and past-oriented clusters. These findings should not be interpreted as strictly causal, as the identification strategy remains limited. Rather, the project serves as an exploratory machine learning exercise showing how large literary corpora can be transformed into structured data and used to study literary change at scale.
Articles and commentary
-
Central and Eastern Europe Outlook: United in Diversity
-
Shifting Sands: The UAE Under an Oil Shock