Assets' returns can be efficiently clustered in regimes, that are suitably defined non‐overlapping intervals creating a partition of the real numbers. This paper explores the relationship between the transition probabilities from one regime to another in assets' returns and the assets' MSCI Environmental, Social and Governance (ESG) scores. We apply the proposed methodology to the relevant empirical instance of the assets in the STOXX® Global 1800 Index. We consider three regimes—low, medium and high, on the basis of the variation range of the considered returns. Regimes are endogenous, in that their identification comes out from an entropy‐based optimization problem over the possible ranges of variation of the returns. We specifically investigate the possible linear relationship between transition probabilities among regimes and the ESG scores for different geographic regions, namely, America, Europe and Asia Pacific. The reference empirical period is the quadrennium 2018–2021. Results suggest that assets that are low ranked in ESG tend to remain in the low state of returns, if they are in the low state, while they tend to switch from higher to lower return states when the initial state is higher. On the other hand, assets that are highly ranked in the ESG dimensions, are likely to switch from a lower to a higher return state, when they are in a lower state or to remain in the same state when they are in a higher state. Results are more evident for America and Asia Pacific regions rather than Europe where regulation on ESG integration is at a more developed stage with respect to the other regions.