ImpactOfFossilFuelPrice-WACC

A compact research dataset for studying how fossil fuel price forecasts and inflation forecasts relate to the weighted average cost of capital (WACC) for energy transition projects.

The release packages cleaned regional WACC panels, forecast covariates, robustness metrics, price-shock elasticities, transmission diagnostics, and Croissant-compatible metadata.

At A Glance

Field	Value
Domain	Climate finance, energy transition, capital cost modeling
Geography	China, Europe, Middle East
Main horizon	Monthly panel from 2026-01 to 2035-12
Core panel	360 rows, 15 columns
Method context	LSTM-GPR price forecasting, inflation forecasting, tabular WACC sensitivity models
License	CC BY 4.0
Reproducible code	https://github.com/Global-Nomad-Nexus/ImpactOfFossilFuelPrice-WACC

Files

File	Rows	Description
wacc_cleaned_panel.csv	360	Combined cleaned WACC panel across all three regions.
cleaned_china_wacc.csv	120	China WACC panel with fossil fuel prices and inflation covariates.
cleaned_europe_wacc.csv	120	Europe WACC panel with fossil fuel prices and inflation covariates.
cleaned_middle_east_wacc.csv	120	Middle East WACC panel with fossil fuel prices and inflation covariates.
oil_price_forecast_to_2035.csv	133	Monthly oil price forecasts with confidence intervals.
inflation_forecasts_2026_2035.csv	10	Annual inflation forecasts for the three regions.
stage3_holdout_metrics.csv	810	Train/test WACC sensitivity metrics using 2026-2032 for training and 2033-2035 for testing.
stage3_rolling_metrics.csv	1,620	Rolling-origin one-year-ahead WACC sensitivity metrics.
stage3_price_elasticities.csv	81	Log-price elasticities and 10 percent price-shock WACC deltas.
transmission_granger_diagnostics.csv	26	Granger-style diagnostics for fuel-price-to-macro transmission.
stage1_existing_accuracy_metrics.csv	4	Existing Stage 1 oil forecast comparison metrics.
stage1_strict_oil_baseline_metrics.csv	5	Fixed-origin oil-price baseline metrics trained only through 2020.
stage1_strict_oil_baseline_forecasts.csv	47	Fixed-origin oil-price baseline forecasts through 2024.
input_cleaning_report.csv	3	Audit report for regional inflation corrections and horizon alignment.
croissant.json	-	Croissant-compatible dataset metadata.

Core Variables

Variable	Meaning
Region	Geographic region for the WACC panel or diagnostic output.
Month	Monthly timestamp in the cleaned WACC panel.
Onshore, Offshore, Nuclear, CCS, Coal, Gas, Solar, Hydro, Biomass	Project or technology WACC columns.
Coal Price, Gas Price, Oil Price	Fossil fuel price forecast covariates used in the WACC sensitivity workflow.
inflation rate	Region-specific inflation forecast covariate.
R2, MAPE, RMSE, MAE	Model evaluation metrics for forecasting and WACC sensitivity experiments.
Elasticity_LogPrice	Estimated log-linear sensitivity of WACC to fossil fuel price changes.

Quick Start

from datasets import load_dataset

panel = load_dataset(
    "global-nomad-nexus/ImpactOfFossilFuelPrice-WACC",
    "cleaned_panel",
)["train"]

print(panel[0])

To load a specific CSV directly:

from datasets import load_dataset

holdout = load_dataset(
    "global-nomad-nexus/ImpactOfFossilFuelPrice-WACC",
    data_files="stage3_holdout_metrics.csv",
)["train"]

Provenance

• Stage 1 uses World Bank commodity data for historical coal, oil, and natural gas prices.
• Stage 2 uses public macroeconomic indicators for inflation forecasting.
• Stage 3 uses WACC projections derived from Calcaterra et al. and cleaned regional panels for China, Europe, and the Middle East.
• The cleaned regional WACC horizon is aligned to 2026-01 through 2035-12.

Reproducibility

The executable workflow and source files are maintained on GitHub:

https://github.com/Global-Nomad-Nexus/ImpactOfFossilFuelPrice-WACC

Run the workflow in this order:

1. code/Price_Prediction.ipynb
2. code/Prediction_Inflation.ipynb
3. code/Impact_FossilFuel.ipynb
4. python code/revision_experiments.py

Interpretation Notes

The Stage 3 outputs should be interpreted as predictive sensitivity and stress-scenario evidence. They are not formal causal estimates. For manuscript revision, prefer the holdout, rolling-origin, elasticity, and transmission-diagnostic outputs over training-set fit summaries.

Citation

@misc{ou2026fossilfuelwacc,
  title = {Fossil Fuel Prices and Capital Cost: A Machine Learning Driven Study on Energy Transition},
  author = {Ou, Shilin and Zhang, Luyao and Huang, Ming-Chun},
  year = {2026},
  howpublished = {Hugging Face Dataset},
  url = {https://huggingface.co/datasets/global-nomad-nexus/ImpactOfFossilFuelPrice-WACC}
}