How can machine learning drive innovation in responsible finance?
Andrés Alonso-Robisco, José Manuel Carbó and José Manuel Marqués focus on climate finance and probe a corpus of articles to understand where and how ML is used in academic literature in their article Machine learning methods in climate finance: a systematic review.
They uncover seven research domains in which ML adds significant value and draw 5 main conclusions regarding its current use:
- ML covers most climate finance topics, including Socially responsible investing, Climate financing, Green financing, Impact investing, Carbon financing, Energy financing, Governance of sustainable financing and investing.
- ML was originally used in physical risk research on topics like natural hazard forecasting, and is now making its way to market-related topics like ESG scoring and compliance.
- ML is more common in mature topics like Biodiversity and Energy economics but begins speeding up emerging field such as climate data and ESG investing.
- Economic and Finance journals pay attention to topics related to CSR and Transition, but lag behind on Physical risk compared to publications in other knowledge domain.
- Artificial neural networks do not always lead, and more easily linear models like Ridge, Elasticnet and Random Forest tend to be preferred in some areas where explainability is at stake.
Even though it speeds up research in specific fields, ML still needs high-quality data and sometime large amounts of energy to be applied properly. Regulators should put an emphasis on technologies to collect, standardise and distribute sustainability data to facilitate its use.