Türkiye için enerji talebi modellemesi ve tahmini

This thesis analyses the elasticities of electricity and natural gas demand in Turkey in terms of sub-sectors and aggregate levels by using five econometric approaches, namely Engle-Granger Two-Step, Fully Modified Ordinary Least Squares, Johansen Cointegration, Auto Regressive Distributed Lag Bounds Testing and Structural Time Series Modelling method. In addition, by using the elasticity results and several scenario assumptions, Turkey's possible future energy demands are forecasted for all sectors up to 2025. The results show that the price and income elasticities of industrial and aggregate electricity demand are lower than one in both short- and long-term. On the other hand, while the short-run residential income elasticities are also found as smaller than one, the long-run income elasticities are estimated as greater than one by conventional cointegration methods. In terms of natural gas demand, while the industrial own-price elasticities are smaller than one in the short- and long-run, the residential and electricity generation sectors price elasticities are greater than one in absolute value, especially in the long-term. Furthermore, the electricity generation sector's own-price elasticities are found to be positive. In the context of income, on the other hand, the highest elasticities are estimated for the residential sector. Moreover, the short-term income elasticities of the industrial sector are greater than that of the long-term, whereas the exact opposite of this situation is observed in the residential and electricity generation sectors. In this study, along with the elasticity estimations, the forecast results of Turkey's electricity and natural gas demand are also presented. The reference scenario's average estimations for the industrial, residential and aggregate electricity demand in 2025 are approximately 125 TWh, 97 TWh, and 345 TWh, respectively. Also, the average natural gas demands for the reference case are estimated as about 25 bcm, 22 bcm, 30 bcm and 90 bcm in terms of the industrial, residential, electricity generation sector and aggregate level, correspondingly. The main motivation of this thesis is to compare the performances of several econometric techniques in modelling and forecasting Turkey's energy demand. Considering the econometric methods used in this study, the Structural Time Series Modelling (STSM) method coupled with Underlying Energy Demand Trend (UEDT) concept has come into prominence, both methodologically and empirically. In addition, the prediction results of this approach are more stable than the conventional cointegration methods. In summary, the conventional cointegration methods and STSM/UEDT approach are used in this thesis for modelling and forecasting Turkey's energy demand in terms of electricity and natural gas. According to the findings of this research, in general, the STSM/UEDT method has given more consistent and significant results than other techniques. Consequently, since the STSM/UEDT approach uses a stochastic trend instead of the deterministic one, several factors that cannot be estimated by the conventional cointegration methods can be included in the model. In this way, the major structural changes in the energy demand trend can be determined. Therefore, for countries such as Turkey of which energy consumption changes over time, using the STSM/UEDT approach can provide advantages to observe the energy demand tendencies better.