Forecasting European Union CO2 emissions using autoregressive integrated moving average-autoregressive conditional heteroscedasticity models

Abstract

In the past few decades, there are lot of discussions around global warming and climate change primarily due to the increased CO2 emissions generated by the consumption of fossil fuels, such as oil and natural gas. After an enormous effort, the EU-28 managed to reduce CO2 emissions in 2014 by 25.7% comparing to 1990 (Kyoto Protocol). This effort should continue in the future so that the EU-28 achieve a 40% reduction on CO2 emissions by 2030. The current paper aims at investigating the optimum model to forecast CO2 emissions in the EU-28. To achieve this aim an autoregressive integrated moving average (ARIMA) (1,1,1)-autoregressive conditional heteroscedasticity (ARCH) (1) model was used, combined with the linear ARIMA model and the conditional variance of the ARCH model. The estimation of parameter optimisation of ARIMA(1,1,1)-ARCH(1) model was done with the maximum likelihood approach using the Marquardt (1963), and Berndt-Hall-Hall-Hausman algorithms and the three distributions (normal, t-student, generalized error), whereas for the estimation of the covariance coefficient the reversed matrix by Hessian was used. Finally, in order to forecast the ARIMA(1,1,1)-ARCH(1) model, a dynamic as well as a static process was applied. The results of the forecasting revealed that the static procedure provides a better forecast comparing to the dynamic one.