Modeling Bank Systemic Risk of Emerging Markets under Geopolitical Shocks: Empirical Evidence from BRICS Countries

Referee: [ABM Simulations] ABM Simulations section: the headline claims of near-total collapse under correlated geopolitical shocks rest on ABM rules whose behavioral parameters (herding, propagation speed, correlation structure) are described only at high level with three free parameters; no calibration to historical BRICS crises (2008, 2014, 2020), no sensitivity analysis, and no comparison to observed outcomes are reported, leaving open the possibility that the collapse result follows from the chosen rules rather than from the data.

Authors: We acknowledge that the current manuscript presents the ABM behavioral parameters at a high level without explicit calibration or sensitivity checks. In the revised version, we will expand this section to calibrate the herding, propagation speed, and correlation parameters directly against observed BRICS bank data from the 2008 global financial crisis, 2014 geopolitical and currency shocks, and 2020 pandemic period, using metrics such as changes in interbank lending and failure rates. We will also report a full sensitivity analysis across plausible ranges of the three free parameters and compare simulated systemic damage metrics to historical outcomes. These additions will demonstrate that the near-total collapse result under correlated shocks is robust and empirically supported rather than rule-driven. revision: yes

Referee: [Results] Results section: the reported simulation outcomes (largest-bank failures vs. correlated shocks) are presented without baseline models, error bars, or robustness tables; this undermines the comparative claim that geopolitical shocks dominate largest-bank failures, as it is impossible to judge whether the ordering is robust to reasonable variation in ABM parameters.

Authors: We agree that the results section requires additional comparative baselines and quantitative robustness measures to support the ordering of damage from largest-bank failures versus correlated shocks. In the revision, we will add baseline simulations (including static network contagion and random-failure models without TGNN inputs), report all key metrics with error bars obtained from 1,000 stochastic Monte Carlo runs, and include dedicated robustness tables showing how the comparative damage rankings hold across variations in ABM parameters. This will allow readers to assess the stability of the finding that correlated geopolitical shocks produce greater systemic damage. revision: yes

Referee: [Methodology] TGNN and DTW sections: the anomaly thresholds and DTW alignment parameters are listed among the free parameters, yet no cross-validation, out-of-sample performance metrics, or ablation on how detected anomalies affect downstream ABM inputs are supplied; this makes the pipeline's intermediate steps non-reproducible and the final stability conclusions difficult to evaluate.

Authors: We recognize that the absence of validation for the DTW and TGNN components limits reproducibility and evaluation of their influence on ABM results. In the revised manuscript, we will add a dedicated validation subsection detailing cross-validation procedures for DTW alignment parameters and TGNN anomaly thresholds, along with out-of-sample performance metrics (e.g., precision, recall, and F1 scores) on held-out years (2021–2024). We will also include ablation studies that quantify the impact on downstream ABM stability metrics when TGNN-detected anomalies are removed from the input networks. These changes will make the pipeline steps transparent and clarify their contribution to the final conclusions. revision: yes