Great Britain balancing-market price signals, renewable share and forecast-error dataset, 2022–2026
Description
This dataset supports the article “Renewable abundance and the stability of balancing-market price signals in Great Britain.” The article asks whether balancing prices in a high-renewables electricity system continue to provide a stable merit-order signal, or whether renewable abundance increasingly changes the persistence, forecast-error sensitivity and tail-risk structure of price formation. The archive contains data and code for an empirical study of Great Britain’s system buy price using half-hourly observations from 26 April 2022 to 20 March 2026. The dependent variable is the system buy price, measured in GBP/MWh. Key explanatory variables include renewable share, renewable actual generation, renewable forecast generation, signed and absolute renewable forecast error, over-forecast and under-forecast components, indicated margin, actual demand, renewable ramping, wind output, solar output, gas-price controls and selected balancing-market controls. The repository is organised into raw, intermediate, processed and output layers. The raw layer contains source/provenance files used in the original data construction. The intermediate layer contains Power Query and patched workbook outputs used in the dataset-building process. The processed layer contains the final analysis dataset used to generate the submitted results. The output layer contains the tables and figure corresponding to the article and online appendix. The included outputs cover variable-construction audits, baseline OLS estimates with classical, HC3 and HAC inference, coefficient magnitudes in intuitive units, gas-control specifications, balancing-control specifications, signed forecast-error models, wind and solar decomposition models, known-break and regime-interaction tests, rolling 90-day coefficients, dynamic autoregressive specifications, event-logit coefficients and analytical marginal effects for tail-price states. These outputs are provided as CSV files to make the results transparent and reusable. The package also includes a reproducibility README, requirements file, data availability statement, citation metadata, licence file, manifest and table/figure mapping document. The scripts use relative paths and are intended to reproduce or check the submitted outputs from the processed dataset.
Files
Steps to reproduce
1. Download and unzip BMRS_submission_data_code_package_FINAL.zip. 2. Create a Python environment using the packages listed in requirements.txt. 3. Use data/processed/analysis_dataset_with_rebuilt_forecast_variables_2022_2026.csv.gz as the final analysis dataset. 4. Run the scripts in scripts/ using the supplied relative-path structure. 5. Compare generated outputs with the CSV files in outputs/tables/ and the figure in outputs/figures/. 6. See docs/table_figure_output_mapping.md for the correspondence between manuscript/appendix outputs and repository files.