Global Rankings, Local Probabilities: Cross-Border Model Risk in Corporate Financial-Fragility Prediction — Reproducibility Data

Description

This dataset contains the derived reproducibility materials supporting the study “Global Rankings, Local Probabilities: Cross-Border Model Risk in Corporate Financial-Fragility Prediction.” The study evaluates the cross-border transportability, calibration, and decision validity of corporate financial-fragility prediction models between the United States and China. The workbook contains machine-readable derived results underlying the reported sample architecture, model-performance comparisons, zero-shot transfer tests, target-market recalibration, bootstrap inference, domain-shift diagnostics, rolling validation, robustness analyses, alternative outcomes, economic-loss analysis, conformal prediction, sector heterogeneity, descriptive statistics, and variable definitions. The underlying firm-level financial information was obtained from Refinitiv/LSEG and is subject to proprietary licensing restrictions. Accordingly, the licensed firm-level microdata and identifiable company-level observations are not redistributed in this repository. The deposited workbook contains derived analytical outputs intended to facilitate verification of the reported results, tables, and figures.

Steps to reproduce

The deposited workbook contains the derived analytical outputs used to construct and verify the tables and figures reported in the associated manuscript. Begin with the README and Reproduction Map worksheets. These identify the study design, temporal sample partitions, outcome construction, variable definitions, and links between workbook outputs and reported analyses. Model-performance sheets contain discrimination, calibration and classification metrics; inference sheets contain bootstrap comparisons; and the remaining worksheets document domain-shift diagnostics, robustness tests, temporal validation, decision-loss analysis, conformal prediction and alternative outcomes. Re-estimation from the original firm-level observations requires licensed access to the underlying Refinitiv/LSEG data and therefore cannot be performed using the public deposit alone.

Institutions

• Bangladesh Institute of Capital Market

  Dhaka Division, Dhaka

Categories

Licence