InvestESG: A multi-agent reinforcement learning benchmark for studying climate investment as a social dilemma

Xiaoxuan Hou,Jiayi Yuan,Joel Z. Leibo,Natasha Jaques

2024-12-01

Abstract:InvestESG is a novel multi-agent reinforcement learning (MARL) benchmark designed to study the impact of Environmental, Social, and Governance (ESG) disclosure mandates on corporate climate investments. Supported by both PyTorch and JAX implementation, the benchmark models an intertemporal social dilemma where companies balance short-term profit losses from climate mitigation efforts and long-term benefits from reducing climate risk, while ESG-conscious investors attempt to influence corporate behavior through their investment decisions, in a scalable and hardware-accelerated manner. Companies allocate capital across mitigation, greenwashing, and resilience, with varying strategies influencing climate outcomes and investor preferences. Our experiments show that without ESG-conscious investors with sufficient capital, corporate mitigation efforts remain limited under the disclosure mandate. However, when a critical mass of investors prioritizes ESG, corporate cooperation increases, which in turn reduces climate risks and enhances long-term financial stability. Additionally, providing more information about global climate risks encourages companies to invest more in mitigation, even without investor involvement. Our findings align with empirical research using real-world data, highlighting MARL's potential to inform policy by providing insights into large-scale socio-economic challenges through efficient testing of alternative policy and market designs.

Machine Learning,Computers and Society,Multiagent Systems,General Economics

What problem does this paper attempt to address?

This paper attempts to study the impact of environmental, social, and governance (ESG) disclosure policies on corporate and investor behavior, especially in climate - change investment, by constructing a multi - agent reinforcement learning (MARL) benchmark - InvestESG. Specifically, the paper aims to explore the following key questions:

1. **Can ESG - conscious investors motivate companies to make emission - reduction efforts?** And how is this motivating effect influenced by the level of investors' ESG consciousness?
2. **Is there a phenomenon of strategy divergence among different types of investors?** That is, only some companies participate in emission reduction and attract investment from ESG - conscious investors, while other companies and their investors choose to free - ride?
3. **Will companies tend to engage in greenwashing activities?** In other words, will companies choose to improve their ESG scores by low - cost means without actually reducing emissions?
4. **What measures can enhance the effectiveness of ESG disclosure policies?**

To answer these questions, the author designed the InvestESG simulation environment, which includes two types of agents: companies and investors. Companies need to decide how to allocate capital for emission reduction, greenwashing, or enhancing climate resilience in each cycle; while investors select companies in their investment portfolios according to their financial return preferences and the degree of attention to ESG factors. In this way, the paper attempts to simulate a complex multi - agent system to study the long - term interaction between companies and investors under different policy settings and its impact on climate - change risks.