Markov-Modulated ESG Rating Dynamics under Exogenous Environmental Regimes

ESG ratings are increasingly used to assess firms’ sustainability profiles and to support investment, risk-management, and disclosure decisions. This paper investigates whether ESG rating dynamics depend on firm-level carbon emission intensity. We propose an exogenous-regime Markov-modulated model in which ESG rating transition probabilities vary across regimes defined by a carbon-emission-intensity index. The index is based on total CO2 emissions normalized by revenue, which controls for firm size and allows environmental conditions to be compared across companies operating at different scales. This index is discretized through a likelihood-based change-point procedure, and the resulting regime process is modeled as a Markov chain. We apply the model to annual Refinitiv ESG rating categories for 100 firms over 2015-2024. The empirical results show that ESG ratings are highly persistent, but their transition dynamics differ across environmental regimes. The regime-dependent modeling is statistically supported and improves one-step-ahead forecasting accuracy relative to the homogeneous Markov-chain benchmark, with the two-threshold model providing the best overall performance.