Replication package for "Hydroelectric Generation and Wholesale Electricity Price Stabilization in Spain: Evidence from a Hydrological Instrument"

This dataset accompanies a study of how exogenous variation in water availability affects the wholesale electricity price in Spain. It contains 156 monthly observations for the Spanish peninsular power system from January 2012 to December 2024. The core hypothesis is that hydroelectric generation lowers wholesale prices through the merit order, but because reservoir operators respond dynamically to price signals, OLS estimates are biased. The dataset supports an instrumental-variables strategy in which the hydraulic producible — the electricity obtainable from observed precipitation, runoff and snowmelt, published by Red Eléctrica de España since 1920 — serves as a weather-driven, exogenous instrument for monthly hydroelectric generation. What the data show Wholesale prices range from 13.67 €/MWh (April 2024) to 283.30 €/MWh (March 2022, at the peak of the European energy crisis). Annual hydroelectric output varies by almost a factor of two, from 21,830 GWh in the drought year 2022 to 42,528 GWh in 2014. The producible tracks these fluctuations exogenously and predicts observed generation with a first-stage F-statistic of 87.5. The IV estimate is −0.665 €/MWh per GWh/day, 49 percent larger in absolute value than the OLS estimate, and robust across nine alternative specifications. The apparent amplification of the effect in dry years disappears once the 2021–22 gas crisis is excluded, indicating that the pass-through is stable across hydrological conditions once the gas-price regime is controlled. Data sources and construction Wholesale prices come from OMIE annual market reports. The hydraulic producible is from REE's historical monthly series. Hydroelectric, wind, solar and demand data combine REE's Informe del Sistema Eléctrico Español 2015 (2012–2015) with Ember Monthly Electricity Data (2016–2024). Natural gas TTF prices are from the IMF/FRED series PNGASEUUSDM converted to €/MWh. Monthly EUA prices for 2014–2024 are observed from public EU ETS data; 2012–2013 values are reconstructed from observed closing prices and cross-checked against Aleasoft and Agora Energiewende. Monthly mean temperature is from AEMET station data. Package contents Seven CSV data files, one Python replication script (replication.py), and a README with full documentation of sources, variable definitions, transformations and reconstruction procedures. Every table and figure in the paper is reproduced by running python code/replication.py from the package directory. A separate note in the README lists methodological caveats useful for reviewers, including the sensitivity of β(H) to alternative CO₂ series and the decomposition of the dry-year effect.本数据集配套一项关于水资源可获得性的外生变动如何影响西班牙批发电价的研究。数据集涵盖2012年1月至2024年12月期间西班牙半岛电力系统的156组月度观测样本。核心研究假说为：水力发电（hydroelectric generation）通过发电优先级排序（merit order）压低批发电价，但由于水库运营商会针对价格信号做出动态响应，普通最小二乘（Ordinary Least Squares, OLS）估计存在偏误。本数据集支持一项工具变量法（instrumental-variables strategy）研究策略：其中“可产出水力发电量（hydraulic producible）”——即通过实测降水、径流与融雪计算得出的可发电量，由西班牙国家电网公司（Red Eléctrica de España, REE）自1920年起发布——可作为月度水力发电量的天气驱动型外生工具变量。 数据表现 批发电价的区间为13.67 €/MWh（2024年4月）至283.30 €/MWh（2022年3月，即欧洲能源危机峰值时段）。年度水力发电量的波动幅度接近两倍：2022年干旱年的发电量为21830 GWh，2014年则为42528 GWh。可产出水力发电量可外生地追踪这些波动，且对实际发电量的第一阶段F统计量为87.5。工具变量估计结果为每GWh/日对应-0.665 €/MWh，其绝对值比普通最小二乘估计结果高出49%，且在九种不同的模型设定下均保持稳健。在排除2021-2022年天然气危机的影响后，干旱年份效应出现的放大现象消失，这表明在控制天然气价格环境后，电价的水力发电传导效应在不同水文条件下均保持稳定。 数据来源与构建 批发电价数据来自西班牙电力市场运营商（OMIE）年度市场报告。可产出水力发电量数据来自西班牙国家电网公司（REE）的历史月度序列。水力发电、风电、光伏与用电需求数据整合了西班牙国家电网公司2015年发布的《西班牙电力系统报告》（Informe del Sistema Eléctrico Español 2015，覆盖2012-2015年）与Ember月度电力数据（覆盖2016-2024年）。天然气TTF价格来自国际货币基金组织/圣路易斯联邦储备银行（IMF/FRED）序列PNGASEUUSDM，已转换为€/MWh计价。2014-2024年的月度欧盟碳排放权（EUA）价格来自欧盟碳排放交易体系（EU ETS）公开数据；2012-2013年的价格通过实测收盘价重构得到，并与Aleasoft和Agora Energiewende的公开数据交叉验证。月度平均气温数据来自西班牙国家气象局（AEMET）的站点观测数据。 数据包内容 本数据包包含7个CSV数据文件、1份Python复现脚本（replication.py），以及一份详细说明所有数据来源、变量定义、数据转换与重构流程的README文档。论文中的所有表格与图表均可通过在数据包目录下运行python code/replication.py复现。README中还附有一份单独的说明，列出了供审稿人参考的方法学注意事项，包括β(H)对不同碳排放序列的敏感性，以及干旱年份效应的分解过程。