This dataset contains product-level classification maps that assign energy-using goods to a clean or fossil-based energy category and to one of five end-use sectors — transport, heating, residential, power, and industry. It was developed to support the indirect estimation of clean and dirty energy investment by sector, for which no direct data source exists. The underlying approach uses annual industrial production combined with trade flows as a proxy, applying the formula Investment = Production value − Exports + Imports. This reflects the principle that exported goods do not contribute to the domestic capital stock while imported goods do. When applied to production and trade data at the product level, the classification maps yield annual clean and dirty energy investment time series by sector for the EU and India. A key finding from the resulting series is that the energy transition is underway: clean investment is outpacing dirty investment across both advanced and emerging economies.
Products are classified as dirty if they require fossil fuels to operate (diesel, gasoline, natural gas, fuel oil, or LPG), and as clean if they require electricity or direct renewable energy such as solar thermal. Electricity-based capital goods are treated as clean, since only the upstream electricity generation needs to be emission-free for the capital good itself to contribute to a clean transition. Where a product's technical description is compatible with both fossil-fuel and electric operation, fractional clean and dirty shares are assigned. Classification was carried out using AI tools (mAI and ChatGPT) to generate product descriptions and typical energy use profiles, combined with human expert review.
This data set is part of the forthcoming WBG working paper "Capital Adjustment Costs and Nationally Determined
Contributions: How to Avoid Excess Transitions of Energy Capital?" authored by Anna-Maria Goeth, Leopold Zessner-Spitzenberg and Carolyn Fischer.