Electricity demand analysis and forecasting: A panel cointegration approach

This article analyzes the demand for electricity and provides out-of-sample forecasting at the sectoral level using a panel cointegration approach. The econometric model permits cross-sectional heterogeneity within a dynamic framework that incorporates information on relevant income and prices of domestic and foreign goods. Both the short-run dynamics and the long-run slope coefficients are allowed to vary across cross-sections. Also, the testing for unit roots and cointegration in panels allows for heterogeneous fixed effects and deterministic trends. Using Egyptian data, it is shown that the empirical model produces reliable ex-post forecasts near the end of the full sample period. These pseudo forecasts are representative of what one would expect if the forecasting relationship is stationary. The long-run parameter estimates are then used to conduct ex-ante forecasting under plausible assumptions for policy making.