Analysing electricity consumption in South Africa using volatility forecasting models

Abstract

The study explored the three-phase approach of SARIMA following the Box-Jenkins methodology, GARCH, and hybrid SARIMA-GARCH models. These volatility forecasting models were used to model electricity consumption in South Africa. The SARIMA (1, 1, 2)(0, 1, 1)12 model was found to be adequate to model South African electricity consumption. However, the series exhibits the presence of the ARCH effect, which led to modelling a GARCH model. The GARCH (1, 1) was modelled and confirmed to be a good fit. Furthermore, the study fitted the residual of the SARIMA model into the GARCH model to make it a hybrid SARIMA (1, 1, 2)(0, 1, 1)12-GARCH (1,1) model. It was observed that the hybrid model is the best fit model, with the smallest AIC value, and the diagnostic checking also confirmed that the model is adequate for forecasting electricity consumption. The results of the hybrid SARIMA (1, 1, 2)(0, 1, 1)12-GARCH (1, 1) model showed that the electricity consumption series has the highest volatility persistence value, and unconditional volatility for the series is finite.