Gold-Silver & Geopolitical Risk

Description

This dataset contains the compiled, structured, and feature-engineered daily time-series data used to model the non-linear transmission channels and risk spillovers between precious metals (gold and silver) and geopolitical risk dimensions. The data supports a triple-target XGBoost machine learning architecture integrated with SHAP (Shapley Additive exPlanations) to evaluate system connectedness across different conditional distributions (Mean, Median Normal, and Extreme Crisis states). The final synchronized dataset ready for machine learning estimation consists of 10,324 rows.

Steps to reproduce

Data Collection and SourcesFinancial Market Data: Daily price series for Gold and Silver futures extracted to compute log returns and rolling Total Connectedness Indices ($TCI$) via time-varying parameter vector autoregressions (TVP-VAR).Geopolitical Risk (GPR) Data: Daily indices obtained from the Geopolitical Risk Index (constructed by Dario Caldara and Matteo Iacoviello), including:GPRD: The baseline composite Geopolitical Risk Index.GPRD_ACT: The sub-index capturing actual, physical geopolitical events (e.g., outbreaks of war, military acts, terrorist attacks).GPRD_THREAT: The sub-index capturing forward-looking geopolitical threats (e.g., rhetorical escalations, official war warnings, media-induced risks).File Structure & Variable DefinitionsThe provided dataset includes the raw variables, their first-order differences (denoted by the prefix D_), and their immediate historical lags (denoted by the suffix _lag1).

Institutions

• Soran University

  Erbil, Soran

Categories

Licence