Structural Time Series Modelling and Forecasting of Energy Demand: The Case of Uzbekistan

Abstract

This study investigates the energy (considering electricity) demand drivers and future estimates through time series data in Uzbekistan. The research focuses on both aggregated (total electricity demand) and disaggregated (residential, industrial and agricultural) levels. Initially, the paper briefly outlines the global and Uzbek energy market with specific considerations such as energy efficiency, renewable energy development, legislative developments and practical issues happening in the sector. Then based on the econometric techniques developed by renowned researchers such as Hunt and Harvey, the research estimates energy demand models through the sectors. Moreover, the study investigates the forecasting estimations based on out-of-sample method considering the differentiated values of the dependent variables. The findings show that the GDP has much effect on estimating the demand models through all sectors. In most of the models, GDP shows statistically significant values, but in some exceptions gives wrong signs. On average, the GDP affects the electricity consumption around 0.5-1%. Consumer price index effect is not seen as considerable across models. Also, annual average temperature does not have any relationship with electricity demand at all levels. Similar to GDP, lagged values of electricity demand including per capita use or per GDP USD show significant values. Alternative sources such as coal and gas have some effects at residential level, but not at other sectors. The effect of population does not give plausible results. Constant terms are found to be significant, but inconsistent values. On the other hand, forecasting estimates show that residential electricity demand will have stable growth both in short- and long-run. Other sectors including aggregated level will be affected by variances in the short-run meaning that they will fluctuate for short period, then reach their steady rate (stable growth). Overall, demand models have met the conditions based on the previous papers. They point out that not all the time factors have the significance level or expected signs. Besides, forecasting estimations were affected by the seasonality and low stationarity rates with high variance, but still give consistent results. In conclusion, some policy recommendations are followed by the results not only considering the derived results, but current state of the energy sector stated through chapters.