This paper contributes to extend the Environmental Kuznets Curve (EKC) literature by introducing a new estimation technique able to deal with the issue of slope heterogeneity, functional form issues, the role of unobserved time-dependent factors and the presence of outliers in the data. A robust hidden Markov model is used to estimate an Environmental Kuznets Curve for a sample of 28 Organisation of Economic Co-operation and Development countries over the period 1968 to 2006. Dependent mixture models provide more flexibility than single-distributions do by assuming the existence of a latent process that follows a finite state first-order Markov chain and affects the distribution of the response variable. The main results indicate that the relationship between emissions and income is country-specific and changes over time. By using robust estimation, five homogeneous groups of countries are identified and country-membership transition dynamics are shown. Copyright © 2013 John Wiley & Sons, Ltd.