Replication data for "Risk Contagion in Firm-Bank Credit Networks: How Credit Strategies Shape Systemic Risk through Risk Clout and Vulnerability"

Description

This dataset supports independent replication of the empirical analysis in the manuscript submitted to Economic Modelling (ECMODE-D-25-03516). It contains the full code-and-data pipeline for constructing firm-bank credit network backbones (2013–2022), measuring systemic risk via DebtRank (Risk Clout and Risk Vulnerability), computing credit-strategy indicators (diversification and niche overlap), extracting network topology controls, estimating SUR models (Table 1), and generating main-text figures (Fig. 1–5) plus bounded CDS co-evolution supplementary outputs (Appendix D). Contents include: (i) CSMAR-derived raw loan Excel files and processed weighted adjacency matrices; (ii) intermediate CSV outputs for backbone detection, DebtRank, credit strategy, and topology; (iii) the final bank-year panel (05_CreditNet_Driver_Bank_2013_2022.csv); (iv) R and Python scripts organized in Steps 0–7 folders; (v) master_run.R for end-to-end orchestration; (vi) Nature-style PNG figures in 06_Plotting/; and (vii) README.pdf and DATASET_README.txt with replication instructions. Raw loan files derived from CSMAR are included with permission from the data provider. Users must comply with CSMAR terms for any further redistribution. Bank and firm identifiers use coded IDs (BNK_*, STK*); bank names appear in GBK-encoded Chinese in the panel file.

Steps to reproduce

1. Install R (>= 4.0). For Modes A/B also install Python (>= 3.8) with pandas, numpy, and networkx. From the Finance_Data_File root, install R packages with: Rscript master_run.R --install-packages --check-only Package list: README.md §4.2. Mode C needs only systemfit, dplyr, ggplot2, ggdensity, and cowplot. 2. Unzip the archive and open a terminal in the Finance_Data_File root (the folder that contains master_run.R and README.md). 3. Pre-flight validation: Rscript master_run.R --check-only 4. Fast replication (Table 1, Appendix Table C.1, Fig. 1–5; ~5–8 min): Rscript master_run.R C This fits the three SUR models, checks Table C.1 coefficients against the manuscript, and writes Nature-style Fig. 1–5 to 06_Plotting/06_Output_SN/. 5. Code-pipeline replication excluding the null model (~3–6 h): Rscript master_run.R (Mode B: Steps 1, 2, 4, 5, 6, 7. Step 3 is skipped by default.) To reuse shipped backbone and DebtRank outputs and still refresh Steps 4–7: Rscript master_run.R B --skip-step1 --skip-step2 6. Optional Appendix D (bounded CDS figures; not invoked by master_run.R): Rscript 06_Plotting/06_Plotting_CDS_2022_Bounded.R See README.md (primary guide) and DATASET_README.txt for folder layout, Mode A (includes Step 3), and troubleshooting.

Institutions

• Shanghai University

  Shanghai, Shanghai

Categories

Licence