1. Is Electricity Consumption Stationary in Turkey? Eyup DOGAN Department of Economics, Abdullah Gul University, TURKEY eyup.dogan@agu.edu.tr

2. Introduction The stochastic behaviors of energy variables are worth knowing due to several reasons. Smyth (2013) and, Smyth and Narayan (2014) address several known limitations in the literature. Such as, many studies focus on: 1.Multi-country panel data 2.United States of America 3.National Data

3. Literature Review Only five studies use sub-national data : Apergis and Payne, 2010; Apergis et al., 2010a, Apergis et al., 2010b; Aslan, 2011. Many studies use national data: Narayan and Smyth (2007), Chen and Lee (2007), Narayan et al. (2008), Mishra et al. (2009), Apergis and Tsoumas (2011), Ozturk and Aslan (2011), Lean and Smyth (2013), Shahbaz et al. (2014), Yilanci and Tunali (2014).

4. Contribution For the first time in the literature an attempt is made to investigate integration properties of a type of energy consumption by sectors at sub-national level. In addition, it focuses on a country other than the USA. This research study aims at investigating unit root properties of electricity consumption for 12 regions of Turkey by four sectors in addition to total electricity consumption. Turkey is the sixth largest consumer of electricity, a candidate country for European Union, and one of the G-20 members.

5. Methods

6. Findings DFGLS reports that we can reject the null hypothesis of unit root in the cases of agriculture in TR1, TR4 and TR12, industrial in TR5 and TR12, business in TR6, residential in TR9. The second unit root test PP reveals that agriculture sector in TR1, TR4 and TR12, industrial sector in TR5 and TR12, business sector in TR6, residential sector in TR7 do not contain unit root and thus electricity consumption in these sectors are stationary process. According to the outcome of the Zivot-Andrews unit root tests with one structural break, agriculture in TR1, TR4 and TR7, industrial in TR5, TR8, TR9, TR10 and TR11, business in TR4, TR7 and TR9, total electricity consumption in TR11 are found to be integrated of order zero.

7. Conclusion Electricity consumption is trend stationary in 7 cases according to DFGLS and PP. It is stationary process in 12 cases when we take into account the structural changes by applying ZA. Thus, shocks to electricity consumption have transitory effects in those cases; namely, agriculture in region 1, region 4 and region 7, industrial in region 5, region 8, region 9, region 10 and region 11, business in region 4, region 7 and region 9, total electricity consumption in region 11. Regarding policy implications, policies to decrease or stimulate the use of electricity have a long-run impact on electricity consumption in 80% of cases in Turkey given that 48 cases are non-stationary process. On the other hand, the past behavior of electricity consumption can be used to predict the future behavior of that in only 12 cases.