We employ an ARDL bounds testing approach to cointegration and Unrestricted Error Correction Models (UECMs) to estimate the relationship between income and CO2 emissions per capita in 21 Latin American Countries (LACs) over 1960-2017. Using time series we estimate six different specifications of the model to take into account the independent effect on CO2 emissions per capita of different factors considered as drivers of different dynamics of CO2 emissions along the development path. This approach allows to address two concerns. First, the estimation of the model controlling for different variables serves to assess if the EKC hypothesis is supported by evidence in any of the LACs considered and to evaluate if this evidence is robust to different model specifications. Second, the inclusion of control variables accounting for the effect on CO2 emissions is directed at increasing our understanding of CO2 emissions drivers in different countries. The EKC hypothesis effectively describes the long term income-emissions relationship only in a minority of LACs and, in many cases, the effect on CO2 emissions of different factors depends on the individual country experience and on the type and quantity of environmental policies adopted. Overall, these results call for increased environmental action in the region.