Open Information: A Defining Perspective on Web Datasets for Carbon Pricing

Battle Count: 68

The paper provides a conceptual and statistical framework for integrating web-based alternative data (Open Information) into pricing models, which is directly relevant to quantitative trading. The GARCH-X formulation with external regressors and VAR models are standard tools in quant finance. However, the practical trading signal is weak (small effect sizes, low R-squared), the improvement in forecasting is marginal (~0.01% RMSE), and no explicit trading strategy or backtest is provided. The value is more in the conceptual framework for alternative data integration than in a directly deployable trading signal. The carbon market focus narrows applicability.

The econometric models (VAR, GARCH-X, fGARCH, EWMA, iGARCH) are well-established and available in standard packages (R, Python). However, the pipeline involves: (1) processing ~100GB GDELT data with CAMEO code filtering, (2) log transformations and regression to extract uncorrelated components, (3) randomized bootstrap imputation for missing values, (4) rolling window estimation with 253 windows, (5) multiple model variants with AIC/BIC comparison, and (6) leave-one-out out-of-sample forecasting. The data engineering component is substantial, and the multi-step methodology requires careful implementation to avoid data leakage.

Data sources are identified (GDELT Events Database, ICAP for EUA prices), methodology is described with equations, and dataset configuration is provided (date range, size ~100GB). However, no code repository is provided, the bootstrap imputation algorithm is randomized (results may vary), and the GDELT CAMEO code filtering logic, while described, would require careful replication. The leave-one-out window configuration starting at index 2501 is specified.