Migration Patterns of ESG Ratings

Abstract

This thesis considers the modelling of ESG rating migration based on both discrete and continuous-time observations. With data from MSCI ESG Research, we illustrate the difference between estimators based on discrete-time cohort methods and estimators based on continuous observations - with and without the assumption of time homogeneity. We apply a semiparametric regression technique, Cox’s time-varying proportional hazard model, to test for the effect of covariates in rating transitions. Duration, dependence on previous rating, continent and market cap size all show significant effects for some transition types, and especially, an increase in duration has a positive effect on rating probabilities. We also find signs of non-Markovian effects when looking at dependence on previous rating state.