Optimizing Impact Investment Portfolios with Reinforcement Learning: A Data-Driven Framework for Balancing Financial Returns and SDG Alignment

Abstract

Impact investors face a complex multi-objective optimization challenge: balancing financial returns with sustainability outcomes, particularly alignment with the UN Sustainable Development Goals (SDGs). Traditional portfolio optimization methods fall short in dynamically integrating real-time sustainability metrics and adapting to changing market conditions. This paper introduces a novel reinforcement learning (RL) framework designed to optimize impact investment portfolios by simultaneously maximizing risk-adjusted financial returns and SDG alignment. We formulate the portfolio management task as a Markov decision process, incorporating both financial indicators and sustainability metrics into the state space, and propose a dual-objective reward function that allows investors to specify their preferred trade-off between financial and impact goals. Using a Deep Deterministic Policy Gradient algorithm, our RL agent learns optimal allocation strategies through interaction with a simulated market environment. Empirical results demonstrate that the proposed framework significantly outperforms traditional methods, achieving an 80.8% higher Sharpe ratio (1.32 vs. 0.73 for mean-variance optimization (MVO)), 87.1% SDG alignment, and a 34.2% reduction in maximum drawdown (–12.3% vs. –18.7% for MVO). The framework also maintains an average environmental, social, and governance score of 82.4 and reduces carbon intensity by 27.6%. The study contributes a scalable, data-driven approach to sustainable finance, enabling more responsive and responsible investment strategies without compromising financial performance.

Received: 27 August 2025 | Revised: 10 October 2025 | Accepted: 3 November 2025

Conflicts of Interest

The authors declare that they have no conflicts of interest to this work.

Data Availability Statement

The data that support this work are available upon reasonable request to the corresponding author.

Author Contribution Statement

Sanjay Agal: Conceptualization, Methodology, Formal analysis, Investigation, Writing – original draft, Writing – review & editing, Visualization, Supervision, Project administration, Funding acquisition. Krishna Raulji: Conceptualization, Methodology, Formal analysis, Investigation, Writing – original draft, Writing – review & editing, Visualization, Supervision, Project administration, Funding acquisition. Kishori Shekokar: Writing – review & editing. Nikunj Bhavsar: Software, Validation, Resources, Data curation.