An explainable machine learning analysis of retail renewable premiums

Résumé fourni par la source

Estimating the supplier’s premium on retail renewable energy is persistently challenging. 14 U.S. states and several international jurisdictions have implemented retail restructuring (i.e., ”retail choice”) through which consumers can shop for electricity supply from marketers. These marketers often offer plans with varying compositions of renewable energy supplied by renewable energy certificates (RECs). This study quantifies the renewable energy premium on residential electricity offers using comprehensive data from Ohio’s retail choice market (2014–2023). We develop an algorithm that matches renewable and non-renewable offers by the same supplier on the same day, with virtually identical offer details (e.g., maturity, monthly fee, termination fee) over 120,000 matched contracts. Using a gradient boosting model with SHAP (Shapley Additive Explanations), we characterize how observable contract features and wholesale market conditions (e.g., forward hub prices) on the renewable energy premium of renewable offers. On average, renewable contracts include a markup of 7.67% (price difference of 0.642 cents/kWh). However, results also indicate that longer contract maturities and higher early termination fees are associated with lower renewable energy premiums, suggesting that renewable premiums vary systematically across contract structures. Importantly, we find robust empirical evidence for the presence of a trigger price: when forward wholesale energy prices are below $47.9/MWh, marketers obtain renewable energy premiums through higher marginal prices. Above this trigger, marketers no longer obtain a higher premium on renewable offers and instead manipulate maturities and early termination fees. Policy implications for market design and societal environmental objectives are presented.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.