projects

Research projects by Johannes Bleher in econometrics, data science, finance, and computational social science.

Selected current and recent projects. Short descriptions emphasize the problem, method, and current output.

economics

Administrative Reform Options in Baden-Wuerttemberg

Ranks Baden-Wuerttemberg county reform options with a reproducible cost-frontier screen, producing a cautious prioritization tool for institutional review.

Hydrological Shocks and Fuel-Supply Resilience

Combines river gauges, AIS tanker data, and fuel prices to measure waterway-shock pass-through and supply-chain resilience.

Local Implementation-Capacity Frictions in Baden-Wuerttemberg

Builds a county-year index of local implementation-capacity frictions using official regional, fiscal, labor-market, and infrastructure data.

econometrics

Dynamic Generator Inversion for Observable Conditional Distributions

Builds a Doi-Peliti-style generator framework and empirical-characteristic-function estimator for observable conditional distributions.

financial econometrics

Systemic Tail-Risk Contributions in European Banking

Estimates hierarchical Bayesian CoVaR for European banks, producing a broad bank-level ranking of systemic tail-risk contributions.

Expected-Shortfall Envelopes and Hansen-Jagannathan Kernels

Uses regularized conditional quantiles to compare downside-risk envelopes and admissible Hansen-Jagannathan kernels across asset universes.

Using Exact Simulation of Order Book Dynamics for Bayesian Estimates of Structural Parameters

Uses exact simulation of limit-order-book dynamics to support Bayesian estimation of structural market-microstructure parameters.

statistical methodology

Algorithmic Evidence Processes

Formalizes inference from repeated randomized algorithm runs using sequential evidence processes, producing diagnostics for stability and reproducibility.

Bayes-Factor-Guided Post-Double Selection with Bootstrapped Multiple Imputation

Aggregates variable-selection evidence across bootstrap and multiple-imputation runs with Bayes-factor-style stopping and inclusion rules.

Fast Non-parametric Test on the Equivalence of Multivariate Empirical Distributions

Develops a fast non-parametric test for equivalence of multivariate empirical distributions in high-dimensional settings.

data science

FAIR Data Science

Book project on reproducible, findable, accessible, interoperable, and reusable data-science workflows for empirical research.

computational social science

Political Sentiment in German Newspapers over Time -- Using LLMs to Model Political Sentiment

Uses LLM-based text measurement to track political sentiment in German newspapers over time.

computational finance

Using Symbolic Regression to Describe Order Book Dynamics

Uses symbolic regression to discover interpretable equations for order-book dynamics from high-frequency market data.

computational chemistry

Optimizing Catalysts - Machine Learning-Driven Discovery from Crystal Structure Data

Applies machine learning to crystal-structure data to screen catalyst candidates and accelerate materials discovery.