2 nd International Annual Meeting of Sosyoekonomi Society October 28-29, 2016, Amsterdam / The Netherlands 232 The Determinants of Health Expenditures in Turkey: An ARDL Bounds Testing Approach Tugay GUNEL Hacettepe University tugaygunel@gmail.com Emin Efecan AKTAS Hacettepe University efecanaktas@yahoo.com Bahadir Sazak DOGAN Hacettepe University bhd44@hotmail.com Abstract Health care expenditure is a major health policy concern worldwide. To assist decision-makers in finding best policies to manage health care cost, it is important to understand the factors that affect health care expenditure. In thıs paper, determinants of health expenditure in Turkey have analyzed within the period from 1975 to 2013. Firstly, Augmented Dickey Fuller test has employed to check whether the variables are stationary or not. Then Autoregressive distributed lag (ARDL) bound testing approach has employed to investigate the short-run dynamics and long-run relationship between health expenditure and its determinants. According to results, the major determinants of health expenditure in Turkey is unemployment rate. The coefficient sign of unemployment rate is positive and statistically significant in both long run and short run. Hence, an increase in unemployment rate will lead to increase health expenditure. The other important factor that increase health expenditure at the long run is the previous year health expenditure which is positively related to the health expenditure and statistifically significant with the very low probability (0,0075). Therefore, a unit increase in the previous year expenditure will lead to increase the health expenditure by 0,47 units. The coefficient of Error Correction Model, indicating the speed of adjustment to restore equilibrium in the dynamic model and how quickly variables converge to equilibrium, is equalto (-0.52) and imply that more than 0.52 percent of disequilibrium in the previous year corrected in the current year. Keywords : Health Expenditure, Autoregressive Distributed Lag, Cointegration, Error Correction Method. JEL Classification Codes : C22, H51, I18. Methodology and Data Data Analysis: Data for the study was obtained from OECD and Turkstat. Data are annual and cover the period 1975-2013. Table 1 Basic Descriptive Statistics of Variables HE LIFEX POP15 POP65 UNEMP DOCT Mean 63.44103 67.19487 32.60769 5.656923 8.587179 1.068974 Median 2.800000 69.00000 32.60000 5.440000 8.400000 1.100000 Maximum 5.800000 76.60000 40.40000 7.830000 13.10000 1.760000 Minimum 1.500000 56.10000 24.70000 4.200000 6.500000 0.540000 Std. Dev. 1.424686 6.333867 5.138420 1.217292 1.383394 0.389540 Jarque-Bera 4.860544 3.607617 3.302681 4.194174 8.685752 2.831415 Probability 0.088013 0.164671 0.191793 0.122814 0.012999 0.242754 Observations 39 39 39 39 39 39 According to Jarque-Bera statistic, all series are normally distributed except UNEMP. Because probability of UNEMP is 0,012999, less than 5 percent, meaning that UNEMP is not normally