Forecasting CO2 emissions: Does the choice of estimator matter?

Extant studies in the literature on carbon emissions have done so using numerous methodologies. However, the Environmental Kuznets Curve has remained the workhorse for modelling the link between development and emissions. This study sets out to test the predictability of the EKC hypothesis for CO2 emissions in the US and consequently offers to answer two key questions. First, does the choice of estimator matter for the predictability of EKC in forecasting CO2 emissions? Second, are the results sensitive to any of the following: measures of CO2 emission and output and multiple forecast periods? The results uphold the stance of the inverted U-shaped relationship postulated by the EKC hypothesis. Also, the choice of estimator matters for accurate forecast performance of EKC for CO2 measures. More importantly, any estimator that ignores the inherent statistical properties of the predictors such as endogeneity, conditional heteroscedasticity and persistence, among others, may produce less desirable forecasts than the time series models. This conclusion is valid regardless of the proxies for CO2 emissions and output.