Enquête sur les déterminants de lavolatilité des taux de change en turquie;exemple de la période 2003 – 2021

The subject of the study is to examine the effects of Brent oil prices in US dollars, foreign direct investments in Turkey, headline inflation volatility calculated over the consumer price index in Turkey, and the foreign trade volume of Turkey variables on exchange rate volatility in Turkey between 2003 and 2021. Volatility as a concept is the movement of a numerical value within a certain time period. Volatility, which is used in the financial markets to mean the return of an asset or the volatility of its price, is an important indicator of the risk of an invested financial instrument. In addition to being one of the investment tools, foreign currencies are an important tool for carrying out transactions related to all economic activities with foreign countries. In today's world, with the developments in technology, the economies of the countries; production processes have become highly dependent on each other in terms of foreign trade activities and capital movements. In this respect, the current value and the expected value of foreign currencies used as payment instruments have an important place in international economic activities. Under the floating exchange rate regime, the volatility that occurs in the exchange rates whose value is determined by the market supply and demand in the free market is effective in the decision-making stages of economic actors in many respects. The negative consequences that may occur due to the increase in exchange rate volatility are as follows: the cost determination of the producers using the tradable goods as inputs, the pricing of the goods or services produced by the producers, the welfare level of the consumers as a result of the fluctuations in the prices of the final goods planned to be imported and the decision-making process of the economy management to reach the targets. Knowing the short-term and long-term determinants of exchange rate volatility, as well as determining whether these determinants increase or decrease the exchange rate volatility, ensure that the expectations of economic actors in the country's economy are formed in a healthy way. Basket exchange rate data, which is the average of the US Dollar / Turkish Lira and Euro / Turkish Lira exchange rates, was used in the study. The volatility of the basket rate was modeled with the Generalized Autoregressive Conditional Heteroskedasticity Model (GARCH) using monthly data for the period January 2003 to February 2021. Unlike the Autoregressive Conditional Heteroskedasticity (ARCH) model, the variance parameter, which represents the volatility of the relevant variable in the GARCH model, is a function of its own historical values in addition to the past values of the square of the regression error term. The appropriateness of modeling a series with an ARCH-based method depends on whether there is an ARCH effect in the series. The ARCH-LM test was used to determine whether the basket currency data had an ARCH effect. In the ARCH-LM test, which has a chi-square distribution, the null hypothesis is rejected when the calculated probability coefficient is less than 0.05, against the H_0 hypothesis that there is no ARCH effect in the series, and it is concluded that there is an ARCH effect in the series. If the probability coefficient is greater than 0.05, the null hypothesis that there is no ARCH effect in the series cannot be rejected. As a result of the test, it was concluded that the basket exchange rate data contains autoregressive conditional variable variance (ARCH). Exogenous variables were chosen to examine their effects on exchange rate volatility modeled by GARCH (1,1). Since Turkey is an oil importing country as an exogenous variable, the Brent oil price was chosen as the reference value for the oil price. Inflation volatility has been chosen as another exogenous variable, which is effective on the exchange rate in line with the dynamic purchasing power parity model and is thought to affect the exchange rate volatility due to the dollarization tendency of savers in Turkey due to the fluctuations in inflation in the past. As another external variable, foreign direct investments were chosen because they are an important source of foreign currency input for countries. Finally, foreign trade volume data was chosen as an exogenous variable to represent foreign trade activities, which is the main element of foreign exchange use. In order to obtain the inflation volatility variable, the consumer price index data was first tested with the ARCH-LM test and then modeled with the GARCH (1,1). Intrinsic and extrinsic variables were tested with stationarity tests. Augmented Dickey Fuller (ADF) and Perron structural break unit root tests were used as the stationarity test of the series. The H_0 hypothesis in the ADF test means that the series contains a unit root and is not stationary. In an autoregressive equation in which the first difference of the variable tested whether it contains a unit root or not is the dependent variable, the ADF test tests whether the coefficient of the lagged value of the variable is equal to zero. If the calculated probability coefficient is less than 0.05, the null hypothesis that the series contains a unit root can be rejected. In the opposite case, the series is not stationary and it must be made stationary by taking the difference and similar methods in order to be estimated correctly. Another reason why a series contains a unit root is due to structural breaks. Since the Perron structural break unit root test takes this into account, it was applied after the ADF test. As a result of both the ADF test and the Perron structural break stationarity test, it was concluded that all the data were stationary when first difference was used. ARDL was preferred in the estimation made to examine the effects of exogenous variables on exchange rate volatility. because this method allows the direction of both short-term and long-term relationships between variables to be analyzed, and also allows for the variables that are stationary in the first difference to be included in the analysis. The Distributed Autoregressive Model (ARDL) is preferred. After the model was estimated with ARDL, whether the regression assumptions were met was tested with various tests. It was concluded that the model met the regression assumptions with the Breusch – Pagan – Godfrey variance test, the Breusch – Godfrey serial correlation tests and the Ramsey RESET test. The H_0 hypothesis in the Breusch – Pagan – Godfrey test means that the regression error terms do not have varying variance. In the Breusch – Pagan – Godfrey test, a new regression is estimated, with the square of the error term of the estimated regression equation being the dependent variable and the explanatory variables of the regression being the explanatory variables. Then, model is tested in terms of whether the explanatory variables are all statistically equal to zero or not using the F test. The H_0 hypothesis in the Breusch–Godfrey serial correlation test means that there is no serial correlation between the regression error terms. A new regression is estimated in which the error terms of the regression estimated in the test in question are dependent, the lagged values of the error terms and the independent variables of the regression are independent variables. In the new predicted regression, whether all the coefficients are equal to zero is tested with the Chi-Square test. If the calculated probability coefficient is below 0.05, the null hypothesis is rejected and it is concluded that the regression error terms have a serial correlation. H_0 hypothesis in the Ramsey RESET test means that there is no error in the mathematical form of the model. In the test in question, a new regression is estimated in which the predicted value of the dependent variable in the regression is the dependent variable, and the exponent of the dependent variable as the determined number of exponents as the independent variables. In the new predicted regression, it is tested with the F test whether the coefficients of the independent variables are all equal to zero. If the calculated probability coefficient is below 0.05, the null hypothesis is rejected. The null hypothesis was not rejected in any of the tests, and the regression conditions were met. CUSUM and CUSUMQ tests have tested whether the variable coefficients estimated by the distributed lag autoregressive model (ARDL) meet the stability conditions. According to the CUSUM test, the estimated coefficients satisfy the stability condition when the successive error terms of the model are between the critical values at the 5% level during the considered time period. According to the CUSUMQ test, when the squares of the consecutive error terms of the model are among the critical values at the 5% level during the period under consideration, the estimated coefficients meet the stability condition. In the study, it was concluded that the regression error terms estimated by the ARDL model were found to be between the critical values at the 5% level and provided the stability condition according to the CUSUM test. After the model estimation, the F bounds test was applied to test whether there is a long-term relationship between all exogenous variables and exchange rate volatility. Despite the H_0 hypothesis that there is no long-term relationship between the variables with the F bounds test, it was concluded that there is a long-term relationship between the variables as a result of the probability coefficient being calculated as less than 0.05. In order to determine the direction of the effects of exogenous variables on exchange rate volatility in the long run while other variables are constant, the long-term results of the model estimated with ARDL were examined. In the model, where it was concluded that the variables were in a significant relationship together in the long run with the F bound test, the statistical significance of the variable coefficients was examined one by one this time, and they were not statistically significant: while the other variables were constant, there was no statistically significant effect on the exchange rate volatility of each variable in the long run. It was concluded that the H_0 hypothesis that there was no significant effect could not be rejected. In order to determine whether there is a relationship between exogenous variables and exchange rate volatility in the short term and if so, the error correction model was established from the results obtained by the estimation using the ARDL method. With the estimated error correction model, it was concluded that the value of the exchange rate volatility one period ago and the value of the foreign trade volume the previous period were effective in decreasing the exchange rate volatility in the short term, while the value of the exchange rate volatility two periods ago and the inflation volatility in the current period were increasing. By means of the error correction coefficient calculated by estimating the error correction model, it is concluded that a long-term deviation in the model returns to the balance by 33 percent in the next period, in other words, the effect of the deviation almost disappears after three periods. It provides the necessary conditions for the error correction coefficient to be valid; negative value, between 0 and 1 in absolute value and statistically significant. The results obtained from the study conducted in Turkey between 2003 and 2021 show that in order to keep the exchange rate volatility low in the short term, monetary policies that reduce the volatility of inflation and stabilize the inflation should be followed by the Central Bank. In addition, it is seen that foreign trade policies, which will be implemented in a way to decrease the foreign trade volume, will increase the exchange rate volatility in the short term. In the long-term results, exchange rate volatility is a result of all exogenous variables in the model; It is seen that oil prices, inflation volatility, foreign trade volume and foreign direct investments all affect exchange rate volatility. However, it is seen that none of these variables alone have an effect on increasing or decreasing exchange rate volatility while the others are constant, and exchange rate volatility is the total result of all these variables. It is concluded that exchange rate volatility depends on variables that policy makers can affect in the long run, such as the volatility in the inflation rate, as well as variables that cannot be influenced by decision makers, such as oil prices determined in international markets.