The Effects Of Capital Inflows On Real Exchange Rate

(1)

ISTANBUL TECHNICAL UNIVERSITY  GRADUATE SCHOOL OF ARTS AND SOCIAL SCIENCES

M.A. THESIS

MAY 2017

THE EFFECTS OF CAPITAL INFLOWS ON REAL EXCHANGE RATE

Tuğçe YILDIZ

Department of Economics Economics Programme

Anabilim Dalı : Herhangi Mühendislik, Bilim Programı : Herhangi Program

(2)

(3)

MAY 2017

ISTANBUL TECHNICAL UNIVERSITY

GRADUATE SCHOOL OF ARTS AND SOCIAL SCIENCES

THE EFFECTS OF CAPITAL INFLOWS ON REAL EXCHANGE RATE

M.A. THESIS Tuğçe YILDIZ

(412141019)

Department of Economics Economics Programme

(4)

(5)

2 MAYIS 2017

İSTANBUL TEKNİK ÜNİVERSİTESİ  SOSYAL BİLİMLER ENSTİTÜSÜ

SERMAYE GİRİŞLERİNİN REEL DÖVİZ KURUNA ETKİLERİ

YÜKSEK LİSANS TEZİ Tuğçe Yıldız

(412141019)

İktisat Anabilim Dalı İktisat Programı

(6)

(7)

v

Thesis Advisor : Prof.Dr. Fuat ERDAL ... Ibn Haldun University

Jury Members : Prof. Dr. Bülent GÜLOĞLU ... Istanbul Technical University

Prof.Dr. Ekrem Tatoğlu ... Ibn Haldun University

Date of Submission : 5 May 2017 Date of Defense : 6 June 2017

Tuğçe Yıldız, a M.A. student of ITU Institute of Social Science student ID 412141019, successfully defended the thesis entitled “THE EFFECTS OF CAPITAL INFLOWS ON REAL EXCHANGE RATE” which she prepared after fulfilling the requirement specified in the associated legistations, before the jury whose signatures are below.

(8)

(9)

vii FOREWORD

First of all, I would like to thank Prof. Fuat Erdal for his support, patience and sensibility throughout my entire graduate study. Secondly, I would like to thank Prof. Bülent Güloğlu for his help and guidance. This work would not be completed without their advices and guidance.

Finally, I will never be able to find the correct words to express my gratitude to my family. I would like to thank my sister for her love. I thank my mother and father for their trust and encouragement. I am sure that the things would be much harder if I did not have their support and love. I owe them everything I have in my life.

(10)

(11)

ix TABLE OF CONTENT Page FOREWORD ... vii TABLE OF CONTENT ... ix ABBREVIATIONS ... xi

LIST OF TABLES ... xiii

LIST OF FIGURES ... xv

SUMMARY ... xvii

ÖZET……….. ... xix

1. INTRODUCTION ... 1

2. LITERATURE REVIEW ... 3

3. THE CHANGES IN EXCHANGE RATE POLICY IN TURKEY DURING THE LAST THIRTY YEARS ... 9

4. DATA ANALYSIS ... 13 5. EMPIRICAL ANALYSES ... 19 5.1 Model I ... 19 5.2 Model II ... 26 5.3 Model III ... 30 6. CONCLUSION ... 39 REFERENCES ... 41 APPENDICES ... 45 APPENDIX A ... 46 CURRICULUM VITAE ... 51

(12)

(13)

xi ABBREVIATIONS

IMF :International Monetary Fund

VAR :Vector Autoregression

PPP :Purchuasing Power Parity

LM :Lagrange Multiplier

AR :Autoregressive

CBRE :Central Bank of the Republic of Turkey

S.D :Standard Deviation

REELEXC :Reel Exchange Rate with Seasonality REELEXCSA :Reel Exchange Rate without Seasonality DIRECTINV :Direct Investment

PORTFOL :Portfolio Investment with Seasonality PORTFOLSA :Portfolio Investment without Seasonality OTHERIN :Other Investment with Seasonality OTHERINSA :Other Investment without Seasonality TCMB :Türkiye Cumhuriyet Merkez Bankası EXPORTSA :Export without Seasonality

IMPORTSA :Import without Seasonality REELINT :Reel Interest Rate

(14)

(15)

xiii LIST OF TABLES

Page

Table 4.1 : Variables with and without Trend. ... 17

Table 5.1.1 : Var Residual with Lags ... 19

Table 5.1.2 : Roots of the Characteristic Polynomial Endogeneous Variable… .... 20

Table 5.1.3 : Variance Decomposition…… ... 22

Table 5.2.2 : Roots of the Characteristic Polynomial Endogeneous Variables... 27

Table 5.2.3 : Variance Decomposition…… ... 28

Table 5.3.2 : Roots of the Characteristic Polynomial Endogeneous Variables... 32

(16)

(17)

xv LIST OF FIGURES

Page

Figure 4.1 : Real Exchange Rate (2003) ... 14

Figure 4.2 : Import with and without Seasonality…… ... 14

Figure 4.3 : Export with and without Seasonality…… ... 15

Figure 4.4 : Interest Rate with and without Seasonality…… ... 15

Figure 4.5 : Portfolio Investment with and without Seasonality……... 16

Figure 4.6 : Other Investment with and without Seasonality…… ... 16

Figure 5.1.1 : Inverse Roots of AR…… ... 21

Figure 5.1.2: Response to Real Exchange Rate…… ... 23

Figure 5.1.3: Response to Direct Investment.…… ... 23

Figure 5.1.4: Response to Portfolio Investment…… ... 24

Figure 5.1.5: Response to Other Investment.…… ... 25

Figure 5.1.6: Response to Interest Rate…… ... 25

Figure 5.2.1: Inverse Roots of AR…… ... 27

Figure 5.2.2: Response to Export…… ... 29

Figure 5.2.3: Response to Import.…… ... 29

Figure 5.2.4: Response to Interest Rate…… ... 30

Figure 5.3.1: Inverse Roots to AR…… ... 33

Figure 5.3.2: Response to Real Exchange Rate…… ... 34

Figure 5.3.3: Response to Direct Investment.…… ... 35

Figure 5.3.4: Response to Portfolio Investment…… ... 36

Figure 5.3.5: Response to Other Investment.…… ... 36

Figure 5.3.6: Response to Export…… ... 37

Figure 5.3.7: Response to Import.…… ... 37

(18)

(19)

xvii

THE EFFECTS OF CAPITAL INFLOWS ON REAL EXCHANGE RATE SUMMARY

This thesis aims to analyze the effects of capital inflows on real exchange rate. It is claimed that there is a strong relationship between capital flows and real exchange rate. International trade flows were seen as the primary determinant of exchange rates until 1970’s because of the strict restrictions on international capital flows. However, international capital flows have exceeded international trade flows in goods and services recently due to globalization so capital flows are expected to have more significant effects on real exchange rate than other variables. Therefore, we initially focus on the literature about the relationship between real exchange rate and capital flows as well as the other variables like exports and imports. Then, we mention about the changes in exchange rate and exchange rate policies during the last thirty years in Turkey.

The data used in our analysis are taken from electronic data delivery system of the Central Bank of Turkey and the website of IMF. In this thesis, monthly data are used from 2003:01 – 2016:06. We implement the VAR method in our analysis and start our analysis by removing seasonality from our data in order to obtain more robust results. Three different models are estimated in the study. In the first model, we investigate the relationship between real exchange rate, foreign direct investments, portfolio investments, other investments and interest rate. We observe that the real exchange rate explains 91.25% of error prediction variance itself while foreign direct investment, portfolio investment, other investment and real exchange rate explain 5.33%, 2.03%, 1.21%, and 0.17% of error prediction variance, respectively. Furthermore, for our second model including real exchange rate, export, import and interest rate, the real exchange rate explains 99.80% of error prediction variance itself while export, import, and real interest rate explain 0.05%, 0.05% and 0.09% of error prediction variance respectively for the same period. Therefore, we clearly observe that capital inflows are

(20)

xviii

better in explaining the changes in real exchange rate. Moreover, we develop our third model by including all variables we use in both model 1 and model 2 and obtain consistent results.

(21)

xix

SERMAYE GİRİŞİNİN REEL DÖVİZ KURU ÜZERİNE ETKİLERİ ÖZET

Bu tezde, amacımız sermaye girişinin reel döviz kuru üzerine etkilerini analiz etmektir. Sermaye girişi ile reel döviz kuru arasında güçlü bir ilişki olduğu iddia edilir. 1970’lere kadar uluslararası ticaret akışı uluslararası para akışı üzerindeki kısıtlamalar yüzünden reel döviz kurunun ana belirleyici etkeni olarak kabul ediliyordu. Fakat, globalleşen dünyada uluslararası sermaye akışı uluslararası ticaret hacmini geçti ve sermaye akışının reel döviz kuru üzerindeki etkileri diğer etkenlerden daha önemli hale geldi. Literatür taramamızda reel döviz kuru ile sermaye girişi arasındaki ilişkiye odaklandık fakat ihracat, ithalat gibi diğer etkenlerle ilgili yapılan çalışmalarla da ilgilendik. İlk olarak son otuz yılda Türkiye’deki reel döviz kuru hareketlerinden ve döviz kuru politikalarındaki değişikliklerden bahsettik.

Analizlerimizde kullandığımız veriler Türkiye Cumhuriyeti Merkez Bankası ve IMF web sayfalarından alınmıştır. Veriler aylık olarak 2003:01 ve 2016:06 dönemlerini kapsayacak şekilde alınmıştır. Çalışmamızda VAR modellerini kullandık ve çalışmamıza daha kesin ve doğru sonuçlar elde edebilmek için verilerimizden mevsimsellik etkisini kaldırarak başladık. Verimizi analiz için hazır hale getirdikten sonra çalışmamızda bu veriyi kullanarak üç farklı modelle tahmin ettik. İlk modelimizde reel döviz kuru ile yabancı sermaye yatırımları, portfolyo yatırımlar, diğer yatırımlar ve faiz oranı arasındaki ilişkiyi inceledik. Elde ettiğimiz sonuçlar gösterdi ki ikinci periyottan itibaren tahmin öngörü hata varyansının %91.25’i reel döviz kurunun kendisi tarafından açıklanmaktadır, yabancı sermaye yatırımları, portfölyo yatırımlar, diğer yatırımlar ve faiz oranı, varyansın sırasıyla %5.3, %2.03, %1.21, ve %0.17 kısmını açıklamaktadır. Reel döviz kuru, ihracat, ithalat ve faiz oranını içeren ikinci modelde, sözkonusu değişkenlerin sırasıyla öngörü hata varyansının, %99.80, %0.05, %0.05, %1.21 ve %0.09’unu açıkladığı görülmüştür. Dolayısıyla sermaye girişinin reel döviz kuru üzerindeki etkisi çok daha önemli

(22)

xx

olduğunu gördük. Son olarak birinci ve ikinci modellerdeki değişkenlerin tamamını ekleyerek geliştirdiğimiz üçüncü modelimizin analizinden de tutarlı sonuçlar elde etmeyi başardık.

(23)

1 1. INTRODUCTION

In recent years, capital flows have become more influential on real exchange rates due to fast globalization. International trade flows were seen as the primary determinant of exchange rates until 1970’s because of the strict restrictions on international capital flows. However, international capital inflows have exceeded international trade flows in goods and services. Therefore, several alternative approaches for exchange rate determination are developed by economists during the last forty years.

Understanding the exchange rate behavior is important for determination of exchange rate policies. Implementation of different exchange rate policies affects countries’ economies considerably. Globalization leads to more appreciation of the exchange rates especially in developing countries. Therefore, their exchange rates become more sensitive to the external shocks. Furthermore, Turkey as a developing country has changed its exchange rate policies several times. There are several studies analyzing the behavior of the Turkish exchange rates.

The determinant factors of exchange rate are also affected by exchange rate itself. Therefore, there is a complex economic relationship that makes difficult to separate variables as endogenous and exogenous in simultaneous equations. VAR models are offered in order to remove the difference between endogenous and exogenous variables and all variables are considered as endogenous variables. Moreover, VAR models can be defined as time series predictive models that introduce every variable in the model with the other variables and the lag of all variables. In our study, we are also interested in VAR models.

In this thesis, our fundamental aim is to analyze the effects of capital inflows on real exchange rate in Turkey. It is claimed that there is a strong relationship between capital inflows and real exchange rate. However, the extent of the relationship can be estimated by developing different models.

(24)

2

Therefore, we generate three different models and compare them by using VAR models. The first model is our fundamental model considering capital inflow as determinant factor including foreign direct investments, portfolio investments and other investments. We expect that capital inflow is the most important factor in determining real exchange rate so we investigate the variance explained by capital inflow in our first model. Our second model considers imports, exports and interest rates. We do not expect that it is as powerful as our first model because the effects of imports and exports decrease by globalization. Thirdly, we analyze the contribution of all these factors to real exchange rate with our final model; in that we consider direct investment, portfolio investment, other investment, export, import and real interest rate as determinant factors and compare them.

This thesis includes six chapters and is structured as follows. In Chapter 2, we review the literature on determinants of exchange rates including import-export, interest rate and foreign investments. In Chapter 3, we make an overview for the changes in exchange rates and exchange rate policies during the last thirty years in Turkey. Chapter 4 presents the data used in the models. In Chapter 5, we present our three different VAR models for explaining real exchange rate in Turkey. Finally, conclusions are drawn and directions for the future work are offered in Chapter 6.

(25)

3 2. LITERATURE REVIEW

The focus of this study is on how capital flows affect the real exchange rates. In this chapter a brief literature survey is provided on the factors affecting the exchange rates. During our literature survey, we have seen that imports and exports, interest rate, relative prices and foreign investments are taken as the most common exchange rate determinant factors.

Throop (1993) claims that flexible price models based on purchasing power parity is not successful in explaining exchange rates but sticky price models with real interest rates is more successful. He determines that sticky price models of the real exchange rate might be disappointing in empirical studies due to the shocks in productivity growth, government budget deficits and the real price of oil. However, the models including these three factors along with the real interest rates are capable of explaining the eighty percent of the variation in exchange rates. Alper et al. (2007) have tested the uncovered interest parity condition in order to analyze exchange market in emerging markets after late 1980s. They conclude that deviations from the uncovered interest parity condition for different countries might be due to the choice of the base currency. Uncovered interest rate is also implemented for Turkey but it is found that uncovered interest rate parity does not hold for Turkey (Erdemlioglu, 2007). Hacker et al. (2009) investigate the relationship between the spot exchange rate and the interest rate differentials for several pairs of countries. They find that the relationship between the spot exchange rate and the nominal interest rate is negative in the short term and positive in the long term. Therefore, it is stated that the sticky price models are more appropriate in the short run while the flexible price models are used for the long run analysis. Thomas (2012) investigates the exchange rate determinants for the sub-Saharan countries with flexible exchange rate regimes. The changes in the EMBI spread and in the US Treasury bill rate significantly determine the exchange rates in countries with flexible exchange rate regimes but not in the closed countries. He concludes that domestic interest rates in these countries do not have direct relationship

(26)

4

with exchange rates. Aggarwal (2013) also studies the theory of uncovered interest rate parity but different from the previous literature, he finds that higher interest rate differentials lead to higher interest rate currency depreciating with time. He supports his ideas with a component GARCH model controlling for short term and long term volatility and positive coefficient estimations.

Purchasing power parity is another determinant factor for exchange rates. Froot and Rogoff (1995) study purchasing power parity with three different stages. It is stated that stationary alternatives are more favorable than random walk specifications with single variant. However, non-stationarity hypotheses cannot be rejected by multivariate tests. Mohapatra and Biswas investigate whether real shocks cause a shift in equilibrium real exchange rate. Then, it is used to explain the failure of purchasing power parity and suggested that PPP can also be regarded as a long run relationship between exchange rate and price levels. Muscatelli and Spinelli use a data set from 19th century to prove the purchasing power parity and apply VAR models indicating productivity differentials and divergences in fiscal policies lead to deviations from PPP. Recently, better implementations of purchasing power parity both in the long run and in the short run are stated thanks to larger data sets and nonlinear econometric methods. Therefore, the deviations between real exchange rates and PPP become smaller and confidence in long run PPP increases (Taylor and Taylor, 2004). Bergvall (2004) investigates the factors determining the real exchange rate. He combines supply and demand factors with purchasing power parity by developing an intertemporal optimizing model. Bergin et al. (2016) study the relationship between purchasing power parity and real exchange rate by examining the countries joining the euro currency. It is found that the adjustment of real exchange rates with respect to PPP is faster after abandon of national exchange rates for euro zone.

Since liberalization increases international capital flows, monetary based exchange rate determination approaches become more popular. Wilson (2009) reviews the monetary approximations in the history in his review paper and interested in the fiscal variables and regime changes as determinant factors of exchange rates. The monetary approaches are tested empirically by either new econometric methods by using the

(27)

5

older asset based methods or time series methods by using newer models (Chinn, 2012).

VAR approach is one of the most widely used methods for determinants of exchange rate. Koray and Lastrapes (1989) use VAR method to analyze the relationship between import and exchange rate volatility. They conclude that the relationship is weak between these two variables but a siginificant relationship exists under the flexible exchange regime instead of the fixed exchange regime. Another study by Bhuiyan (2008) uses a Bayesian structural VAR model for Canada to determine the effects of monetary shocks on the exchange rate. It is concluded that there exists considerable effects of monetary policy not only on exchange rate but also on interest rate. The VAR technique is appropriate for various models of exchange rate determinants. The studies applying VAR model are also important for us since we also use the VAR model to investigate the effects of capital flows on exchange rates.

Direct foreign investment is one of the monetary determinants of real exchange rate. As we stated, monetary determinants become more important on exchange rate due to globalization. Edwards (1988) makes one of the oldest studies in the literature by generating a dynamic model of exchange rate behavior with monetary determinants for developing countries. He focuses especially on the devaluations and payment crisis for real exchange rate behavior and lists the fundamentals affecting the equilibrium real exchange rate. Charfi (2013) investigates the relationship between capital flows and real exchange rates by considering Tunisia as an example. It is said that liberalization creates capital inflows for developing countries generally and an appreciation in the real exchange rate. In order to determine the effect of capital inflows on real exchange rate, a VAR model is used and an impulse response function is examined. Khan and Abbas (2015) test whether money supply and bond of a country affect its exchange rate. They use auto regressive distributed lag model to examine the data from 2001 to 2010 and prove that there is a strong relationship between these variables in the long term. Combos et al. (2011) search the effect of capital flows and the exchange regime on the real exchange rate. They make a comprehensive study including forty-two developing countries. It is proven that the flow leads to increase

(28)

6

in real exchange rate among inflows portfolio investment has the greatest impact on the real exchange rate while private flows have considerably less impact. It is also stated that flexible exchange rate is suitable to lessen the appreciation due to capital inflow. Goldberg (2009) studies the relationship between exchange rates and foreign direct investment that is also our fundamental aim in this study. She states that the share of investments abroad increases if exchange volatility is greater but it does not imply that domestic investments decrease. Thus, it is concluded that exchange rate volatility can increase internationalization without harming domestic investments. Combes et al. (2012) use panel co integration methods to analyze the effects of capital inflows on exchange rates in developing countries. They also indicate an appreciation in exchange rate with increasing capital inflow similar to the other studies in the literature. Moreover, they warn the countries with great capital inflow for the possibility of destabilizing macroeconomic management. It is said that especially short term flows like portfolio investments is required a higher attention avoiding a real appreciation.

Kim (1986) uses portfolio approach in Korea for exchange rate determination. He also states that the volatility of exchange rates depend mostly on financial market prices rather than exports and imports after 1970’s. He finds that account balance affects exchange rate inside the borders of portfolio management. He matches his simulation results with theoretical results to prove great impact of portfolio approach as an exchange rate determinant. Goldberg and Klein (1998) indicate the relationship between direct foreign investment and real exchange rate and the relationship between direct investment and trade for Southeast Asia and Latin American countries. Athukorala and Rajapatirana (2003) carry out a comparative study for Asian and Latin American countries in order to understand the relationship between real exchange rate and capital inflows. They also find that capital inflow creates appreciation in real exchange rate but the appreciation is greater for Latin American countries than Asian countries. They explain this by the composition of capital inflow and different response of real exchange rate due to fiscal contraction and nominal exchange rate adjustment. Jongwanich and Kohpaiboon (2010) analyze the relationship between capital flows and real exchange rate by dividing capital flows into three groups as foreign direct investment, portfolio investment and other investment. It is an important

(29)

7

study for us because we have generally seen the studies that consider capital inflow as a whole but our main interest in this study is the impact of foreign direct investment on real exchange rate. They find that portfolio investment and other investment causes a faster appreciation in real exchange rate than foreign direct investment However, the magnitude of impacts of these factors is not different and it is better to consider increasing importance of foreign direct investment activities.

Because our main interest is the impact of capital inflows on real exchange rate in Turkey, we also cover the literature on exchange rate behaviors in Turkey. Ucer et al. (1997) study the nexus among fiscal policy, capital inflows and real exchange rate in Turkey after 1980s. They generate an autoregressive model including related government spending, capital inflows, interest rate differentials and components of real exchange rate. They state that positive shocks on capital inflows result in an appreciation in real exchange rate in Turkey also. In addition, positive shocks on interest rate differentials lead to capital inflow so indirectly causes an appreciation in real exchange rate. Balkan et al. (2002) study short term capital inflows in Turkey after 1989 by excluding domestic portfolio investments. They apply time series econometrics for several factors affecting capital inflows for Turkey like exchange rate, real interest rate and real wages. Gumus et al. (2013) examine the relationship between foreign portfolio investments and macroeconomic factors in Turkey by using various methods including VAR, VAR Granger Causality Tests, Impulse Responses and Variance Decomposition. We also choose VAR techniques for our main analysis among these methods. Their results obtained from different tests prove that foreign portfolio investment affects exchange rate considerably. Karpuz and Kızıltan (2014) investigate whether there is a relationship between real exchange rate and short term investments in Turkey. They find that the relationship between these two variables is bidirectional by considering Akaike and Schwarz information criteria.

(30)

(31)

9

3. THE CHANGES IN EXCHANGE RATE POLICY IN TURKEY DURING THE LAST THIRTY YEARS

The exchange rate policy in Turkey has changed significantly in recent years. Before 1980’s, exchange rate in Turkey was mainly determined by the government based on economic conditions so fixed exchange rate regime was enforced. Devaluation was the most important tool among stabilization measures implemented during this time interval. Before 1980’s, devaluation in Turkey was implemented many times. These devaluations were implemented due to economic conditions. As an example, the inflation was above 15% and the foreign trade deficit was continuously increasing in year 1958 so an additional tax that was equal to 6.22 lira for every dollar purchased externally was enforced to lessen the effects of difference between domestic and foreign prices. In addition, one dollar increased to nine liras in 1960 by a devaluation that is equal to 220%. In 1974, balance of payments got worse due to the increase in oil prices, problems in global market and the increase in the foreign trade deficit. Thus, firstly 30% devaluation was done, and then 88% devaluation came true before 1979. However, the devaluations becoming in the long time intervals caused liras to overvalue (Özçam, 2004).

The exchange rate policies for the time periods between 1980 and 1989 and after 1989 are different from the fixed exchange rate policy before 1980s. The fixed exchange rate regime cannot remedy the problems in the economy and foreign payment deficits. Therefore, January 24 decisions were announced targeting to decrease the pressure on the inflation by strict monetary policy and to encourage exports by flexible exchange rate policy. These decisions yielded several results. Between 1981 and 1989, liras depreciated against dollar and daily exchange rate was announced. In 1988, banks, private financial institutions, and other institutions authorized to make exchange transaction were participated in sessions on setting exchange rate. Foreign trade was liberalized by removing restrictions on import and encouraging export. As a result of short term capital inflows in Turkey during 1989 – 1990, lira appreciated against other national currencies (Barışık and Demircioğlu, 2006).

(32)

10

Flexible exchange rate regime controlled by the Central Bank became valid between 1989 and 1999. In 1991, liras depreciated due to the short term capital outflows caused by Gulf war but the fluctuations on the exchange rate were prevented in 1992. At the end of 1993, pressure on exchange rates began because of excess Lira liquidity and increase in foreign exchange demand. The attempts of the Central Bank in order to decrease excess Lira liquidity by increasing interest rate failed and Liras devaluated against dollar at a rate of 13.6% in 1994. A great financial crisis happened in 1994 due to huge current account deficit with high inflation. The demand for foreign exchange increased after the big loss in Istanbul stock exchange market up to 20%. The Central Bank anted dollar to the market in order to decrease the demand for foreign exchange but it failed and the foreign exchange reserve diminished (Arat, 2003). The exchange rate regime made was changed due to the loss in foreign exchange reserves and fixed exchange regime was enforced again. During the time with fixed exchange policy, Liras depreciated and import increased considerably with economic decisions. In addition, economic shrinkage became 6% after the crisis, the interest rate increased excessively and index of wholesale prices became 150%. Turkey cannot take foreign loan because its credit rating was decreased (Gaytancıoğlu, 2010). April 5 decisions were taken due to these economic problems and Liras depreciated at the rate of 70% in the following days. Even if these devaluations had positive effects on foreign exchange deficit and current account deficit, the effects decreased by time. Even if flexible exchange regime was implemented 1990s, the Central Bank applied policies to stabilize the value of Liras after 1995. The central Bank determined the nominal devaluation based on estimated inflation so it both controlled domestic assets by decreasing the government loans and increased foreign exchange reserves. The economic problems increased due to the effects of Southeast Asia and Russia crisis and the increased public expenditures to cover the costs of two earthquakes in 1999.

The standby with IMF was signed in December 1999 targeting to decrease chronic inflation. The monetary and exchange policy enforced with this agreement was based on a moving anchor system in that the monthly value of an exchange basket consisting of 1 dollar and 1.5 German marks was increased by inflation rate. The Central Bank intervened the exchange market by selling in order to perform the anticipated increase in the exchange basket. The performance of standby with IMF became better than the

(33)

11

expectations and the economy went better. An anchor was determined during 2000 in which exchange rate could fluctuate in case the exchange rate went beyond the borders, the Central Bank intervened the exchange market. Therefore, exchange rate regime was enforced based on estimated inflation rate that had been previously implemented regarding target inflation rate. An economic crisis happened that started in November 2000 and continued during 2001 due to lack of liquidity and some structural problems in the system (Ertekin, 2003).

The policy based on a moving anchor system was abandoned due to the financial crisis in 2001 and a flexible exchange rate was implemented. Some precautions preventing short term debts of public banks were taken and strong capital structure of private banks was aimed. The Central Bank could not interrupt the exchange market with flexible exchange rate regime except extreme fluctuations. In addition, implicit inflation was implemented between 2002 and 2005 and explicit inflation was enforced after 2006. Foreign exchange buying and selling tenders were carried out in 2002. The aim of these tenders was not determination of exchange rate level but prevention of high volatility and reserve accumulation (TCMB, 2005). The inflation rate was decreased from 60% in 2001 to 10% in 2005 through explicit inflation regime. The explicit inflation regime began to be implemented after the confidence of monetary policy ensured by implicit inflation regime in 2005. The inflation rate calculated based on price index of consumer goods for three years was announced in explicit inflation policy. The inflation rate was tried to keep within a specific interval with the monetary policy considering determined targets (TCMB, 2006). The decrease in capital inflows caused depreciation in Lira by 30% in 2006. The increase in food prices due to several reasons also caused to increase inflation in the same year. The Central Bank interrupted tenders of foreign exchange buying and selling and made liquidity arrangements. These precautions stopped the increase in the exchange rate in the second half on 2006 and ensured steady decline. Even if the inflation rate decreased in the second half of 2006, it became 9.6% for 2006 while targeting to 5%. In the beginning of 2007, positivity in economics continued but the inflation rate again increased in August due to global price trend. In addition, the panic in global market due to risks in Mortgage market caused exchange rate depreciation in Turkey like other developing countries (TCMB, 2009)

(34)

12

A global crisis outbreak happened in 2009 that originated from financial sector but also affected reel sector. The crisis started in Mortgage market in the United States in the second half of 2007. Then, it affected firstly the European market and the economies of developing countries later. Turkey felt the effects of crisis considerably because most of Turkey’s export volume was dependent on European countries. Lira appreciated considerably compared to previous periods due to parity movements during which European country increased protective measures against imports and rapid capital movements. The Central Bank could stabilize exchange rates with instantaneous interventions by using limited foreign exchange reserve. The Central Bank enforced various politic tools compatible with price stability from November, 2010 in order to prevent financial risks caused by short term capital inflows, exchange fluctuations, and credit growth by developed countries. Compulsory reserve and other macro urgency measures were taken in the frame of credit policy. Weekly repo interest rate was implemented in the frame of interest policy and interest corridor and other funding methods were carried out in the frame of liquidity policy. It was aimed to decrease the adverse effects of short term capital inflows and to support to financial stabilization with these tools (Degerli and Fendioglu, 2013).

(35)

13 4. DATA ANALYSIS

The data we used in our analysis is taken from electronic data delivery system of the Central Bank of Turkey and the website of IMF. The real exchange rate was calculated with three different methods in the website of the Central Bank that are based on labour costs, consumer price index and producer price index. In this thesis, monthly data are used from 2003:01 – 2016:06. Real interest rate was calculated by subtracting inflation rate from nominal interest rate that is the policy interest rate of the Central Bank. The policy interest rate and the inflation rate were taken from the website of IMF. Imports, exports, capital inflows as direct investment liabilities C.9, portfolio investment liabilities C.11 and other investment liabilities C.13 were taken from electronic data delivery system of the Central Bank. We use monthly data in our analysis. In our models, real exchange rate was taken as dependent variable and export, import, real interest rate, foreign direct investments, portfolio investments and other investments were regarded as independent variables. Other investment means all other financial transactions except for direct, portfolio investment, financials derivatives and reserve assets. It includes :

✓ Currency and deposit, ✓ Loans,

✓ Insurance, pension and standardized guarantee schemes ✓ Trade credits and advances

✓ Other account receivables. (TCMB)

As it can be seen from the graph in Figure 4.1, there was a break in real exchange rate during 2008. The year of 2008 was a critical year for Turkey due to a global crisis started in the United States and the adverse effects of political turbulences in Turkey on the economy. We also add a dummy variable into our three models.

(36)

14

Figure 4.1: Real Exchange Rate (2003).

As it can be observed from Figure 4.2, Figure 4.3, Figure 4.4, Figure 4.5 and Figure 4.6, real interest rate, portfolio investment, other investment, export and import include seasonality while foreign direct investment does not include seasonality. We continue our data analysis with new data series after removing seasonality.

Figure 4.2: Import with and without Seasonality 0 0,5 1 1,5 2 2,5 3 3,5 O ca.03 A ğu .03 Mar .04 Eki .04 May.05 Ara. 05 Te m.06 Şu b .07 Eyl .07 N is .08 Kas .08 H az .09 O ca.10 A ğu .10 Mar .11 Eki .11 May.12 Ara. 12 Te m.13 Şu b .14 Eyl .14 N is .15 Kas .15 Haz .16

(37)

15

Figure 4.3: Export with and without Seasonality

(38)

16

Figure 4.5: Portfolio Investment with and without Seasonality

(39)

17

Simultaneous equations are more favourable than univariate equations due to the complexity of economic relationships. It is observed that macro-economic variables affect one another in economics. Therefore, it is difficult to separate variables as endogenous and exogenous in simultaneous equations. Vector Autoregressive (VAR) models are generated by Sims that can cope with these problems. VAR model removes the difference between endogenous and exogenous variables and all variables are regarded as endogenous variables. Therefore, it is considered that every variable can affect every other variable and can be affected by every other variable. VAR models also defined as time series predictive models that introduce every variable in the model with the other variables and the lag of all variables. The presence of lag values in VAR models makes powerful future estimations possible (Sims, 1980). Therefore, we use VAR model in our study.

Because we implement VAR models, we firstly need to test whether our data are stationary. We use augmented Dickey Fuller (ADF) test developed by Dickey and Fuller by using unit root tests in order to check stationary of our data.

ADF test is based on the assumption of independent and constant variance of shocks. Thus, it assumes that there is no correlation between shocks. The results of ADF tests for our variables with trend and without trend are given in the Appendix. We also summarize the p values for our all variables with and without trend in Table 4.1.

Table 4.1: Variables with and without Trend

Variables ADF test (without trend) ADF test (with trend)

reelexc 0.9809 0.7759 d(reelexc) 0.0000 0.0000 reelintsa 0.0018 0.2173 directinv 0.1972 0.1911 d(directinv) 0.0000 0.0000 portfolsa 0.0000 0.0000 otherinvsa 0.0044 0.0000 export 0.8947 0.4335 d(export) 0.0000 0.0000 importsa 0.8473 0.7011 d(importsa) 0.0000 0.0000

(40)

18

Our results show that the null hypothesis of non – stationary was rejected at 5% level of significance. We observe that the p values of real exchange rate, foreign direct investment, export and import are greater than 0.05. Therefore, we conclude that they are non-stationary. In order to make our data stationary, we take the difference for these variables and we show that their differences represented as a function of d () are stationary with p values smaller than 0.05.

(41)

19 5. EMPIRICAL ANALYSES

5.1 Model I

We initially estimate a Vector Autoregressive (VAR) model as follows:

Reel Exchange Rate = (Foreign Direct Investment, Portfolio Investment, Other Investment, Real Interest Rate, Dummy2008 )

We firstly apply the LM test in order to see whether there is an autocorrelation problem or not among our variables. We make LM test for different lags and we observe best results for lag 4. The results of LM test for lag 4 are presented in Table 5.1.1.

Table 5.1.1: Var Residuals with Lags

VAR Residual Serial Correlation LM Tests

Null Hypothesis: no serial correlation at lag order h Sample: 2003M01 2016M12

Included observations: 157

Lags LM-Stat Prob

1 16.09410 0.9120 2 20.77864 0.7049 3 38.34058 0.0428 4 14.64379 0.9493 5 20.69179 0.7097 6 10.31747 0.9957 7 14.30934 0.9561 8 32.95836 0.1321 9 32.42747 0.1460 10 32.62231 0.1408 11 35.34629 0.0821 12 22.30263 0.6182

The null hypothesis in LM test is that there is no serial correlation at lag. We take our significance level as 0.05. When we check our probability values in Figure 5.1, they all are greater than 0.05. This means that there is no correlation at lag so we do not have a correlation problem that is there is no similarity between observations as a function of the time lag between them.

(42)

20

Moreover, we carry out r root test to control whether our model are stable or not. The results are presented in Table 5.1.2.

Table 5.1.2: Roots of Characteristic Polynomial Endogenous Variables

Roots of Characteristic Polynomial

Endogenous variables: D(REELEXCSA) D(DIRECTINV) PORTFOLSA OTHERINSA REELINTSA

Exogenous variables: C DUMMY1 Lag specification: 1 4 Root Modulus 0.965644 0.965644 -0.643338 + 0.592580i 0.874663 -0.643338 - 0.592580i 0.874663 0.259443 + 0.733041i 0.777599 0.259443 - 0.733041i 0.777599 0.062248 + 0.731932i 0.734574 0.062248 - 0.731932i 0.734574 0.670838 - 0.156646i 0.688884 0.670838 + 0.156646i 0.688884 0.685631 0.685631 -0.308222 + 0.554343i 0.634269 -0.308222 - 0.554343i 0.634269 -0.609132 0.609132 0.062082 + 0.529395i 0.533023 0.062082 - 0.529395i 0.533023 -0.481005 0.481005 0.320928 - 0.254482i 0.409580 0.320928 + 0.254482i 0.409580 -0.281665 + 0.197608i 0.344070 -0.281665 - 0.197608i 0.344070

When we check modulus values in Table 5.1.2, it can be observed that there is no modulus value that is beyond the reference value that is equal to one. This shows that our VAR model 1 is stable.

We can also confirm the stability of our model by observing the graph of AR characteristic inverse root in Figure 5.1.1 The roots within AR root circle prove that there is no problem.

(43)

21 -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5

Inverse Roots of AR Characteristic Polynomial

Figure 5.1.1: Inverse Roots of AR

We also implement variance decomposition in order to understand the effects of every random shock on prediction error variance for future periods. Thus, variance decomposition gives information about system dynamics. One can determine the most effective variable on a macroeconomic magnitude by using variance decomposition. Beside, variance decomposition determines the source of variance that can be itself of the magnitude and other variables. The results are given in Table 5.1.3.

When we check the results in this table, it is seen that 100% of error prediction variance during period 1 is explained by itself. In period 2, the real exchange rate explains 91.25% of error prediction variance itself while foreign direct investment, portfolio investment, other investment and real exchange rate explain 5.33%, 2.03%, 1.21%, and 0.17% of error prediction variance respectively. When we consider all periods, we see that the percentage of error prediction variance explained by itself decreases with time. Furthermore, the percentage of error prediction variance explained by foreign direct investment, portfolio investment, other investment and real interest rate increases with time and converges to 8%, 7%, 4% and 2%. Therefore, foreign direct investment became the most significant variable explaining error variance.

(44)

22

Table 5.1.3: Variance Decomposition

Period S.E. D(REEXCSA) D(DIRECT) PORTSA OTHERSA REINTSA 1 0.045988 100.0000 0.000000 0.000000 0.000000 0.000000 2 0.050150 91.25696 5.336860 2.026192 1.214628 0.165365 3 0.051479 89.00121 5.266449 3.761413 1.743948 0.226982 4 0.052934 85.53275 7.134502 3.557703 2.926071 0.848972 5 0.054297 81.29397 7.030376 6.781371 3.588040 1.306245 6 0.054619 80.44910 7.164550 6.971585 4.045302 1.369466 7 0.054953 79.68458 7.690884 7.274924 3.996211 1.353401 8 0.055055 79.45997 7.781500 7.275877 4.066255 1.416401 9 0.055245 78.91323 7.985234 7.479044 4.042732 1.579757 10 0.055369 78.56103 8.013127 7.561983 4.150154 1.713702

We finally make impulse response analysis. Impulse response functions indicate the effects of shocks with one standard deviation on present and future values of endogenous variables. It can be determined whether a variable can be used as a policy tool with impulse response functions.

We present the response of real exchange rate to one standard deviation shocks of real exchange rate, foreign direct investment, portfolio investment, other investment and real interest rate. In the graphs, the horizontal axis represents the time periods aftershocks while the vertical axis represents responses of the variables included by the model. It can be seen that all results are within lower and upper bounds.

After one standard deviation shock to real exchange rate, real exchange rate results in a positive response in Figure 5.1.2. In second period, the response becomes 1.3% based on our results and then disappears after five periods.

(45)

23

Figure 5.1.2: Response to Real Exchange Rate

After one standard deviation shock to foreign direct investment, real exchange rate results in a positive response in Figure 5.1.3. In second period, the response starts to decrease and fluctuates in following periods. It disappears after five periods. As seen as graph, increases in direct investment appreciate TL and depreciate dollars. Therefore, export declines and import increases. According to theory, increase in imports leads to a decrease in amount of dollar, so dollar appreciates while TL depreciates.

(46)

24

After one standard deviation shock to portfolio investment, real exchange rate results in a negative response and become positive after period three by increasing for four periods in Figure 5.1.4. Its effects disappears after four periods. The effect of the portfolio investment can be seen in the short run. As seen in the graphs, when portfolio investment increases, TL appreciates and reel exchange rate depreciates.

Figure 5.1.4: Response to Portfolio Investment

Thirdly, one standard deviation shock to other investment causes real exchange rate to respond negatively and moves back to equilibrium after seven periods. This shock is actually statistically insignificant in Figure 5.1.5. Like portfolio investment, other investments show effect in the short run. Increase in other investment leads to depreciatiation of the reel exchange rate and so import increases. Demand of dollars rises and reel exchange rate appreciates.

(47)

25

Figure 5.1.5: Response to Other Investment

After one standard deviation shock to real interest rate, real exchange rate results in a negative response in Figure 5.1.6. The effect of the shock reduces after the first period. Increase in reel interest rate causes appreciation in TL and depreciation in US dollars.

(48)

26 5.2 Model II

We estimate a new model using the trade flows variable as follows:

Reel Exchange Rate = (Exports, Imports, Real Interest Rate, Dummy2008)

We again apply the LM test in order to see whether there is an autocorrelation problem or not between variables. We make LM test for different lags and we observe best results for lag 2. The results of LM test for lag 4 are presented in Table 5.2.1.

We take our significance level as 0.05 and our null hypothesis in LM test is that there is no serial correlation at lag. When we check our probability values in Table 5.2.1, we observe the values for lags smaller than 10 are greater than 0.05 and the probabilities for small lags are important for us. Therefore, we conclude that there is no correlation at lag so we do not have a correlation problem that is there is no similarity between observations as a function of the time lag between them.

Table 5.2.1: VAR Residuals with Lags

Lags LM-Stat Prob

1 19.39957 0.2485 2 24.36590 0.0818 3 15.12202 0.5157 4 24.61445 0.0769 5 5.569959 0.9921 6 19.01489 0.2679 7 18.68963 0.2851 8 8.724858 0.9243 9 14.94612 0.5286 10 27.14546 0.0399 11 23.71394 0.0959 12 52.00713 0.0000

Then, we check r roots in order to control stability of our model. The results are presented in Table 5.2.2. When we check modulus values in Table 5.2.2, we again

(49)

27

observe no modulus that is greater than the reference value one. This shows that our VAR model 2 is stable like our model 1.

Endogenous variables: D(REELEXCSA) D(EXPORTSA) D(IMPORTSA) REELINTSA Exogenous variables: C DUMMY1

Lag specification: 1 2 Root Modulus 0.957592 0.957592 0.188155 - 0.518609i 0.551686 0.188155 + 0.518609i 0.551686 -0.357441 - 0.394702i 0.532497 -0.357441 + 0.394702i 0.532497 -0.322867 - 0.289021i 0.433332 -0.322867 + 0.289021i 0.433332 -0.020983 0.020983

We also confirm the stability of our model by observing the graph of AR characteristic inverse root in Figure 5.2.1. Because all roots are inside the AR root, we do not have a problem.

(50)

28

When we check the results in Table 5.2.3, it is seen that 100% of error prediction variance during period 1 is explained by itself. In period 2, the real exchange rate explains 99.80% of error prediction variance itself while export, import, and real interest rate explain 0.05%, 0.05%, 1.21%, and 0.09% of error prediction variance respectively. Therefore, we can conclude that our model 1 independent variables explain more variance than our model 2 independent variables. When we also consider all periods, we see that the percentage of error prediction variance explained by itself decreases with time and the percentage of error prediction variance explained by export, import, and real interest rate increases with time. They converge to 1%, 0.34%, and 2%. It is apparent that real interest rate is more important variable. It is also be proved that our model 1 including foreign direct investment explains greater variance than model 2.

Period S.E. D(REEXCSA) D(EXPSA) D(IMPSA) REINTSA

1 0.049062 100.0000 0.000000 0.000000 0.000000 2 0.052140 99.80693 0.052338 0.046740 0.093989 3 0.053295 98.62303 0.839878 0.284761 0.252333 4 0.054218 97.99495 0.951260 0.275495 0.778292 5 0.054345 97.58080 0.985992 0.347285 1.085927 6 0.054420 97.41741 0.991327 0.346331 1.244930 7 0.054464 97.27230 0.994080 0.346027 1.387589 8 0.054512 97.12028 0.993333 0.346673 1.539717 9 0.054563 96.96521 0.996843 0.346423 1.691527 10 0.054606 96.82342 1.000549 0.347585 1.828447

We also present the response of real exchange rate to one standard deviation shocks of real exchange rate, export, import and real interest rate. In the graphs, the horizontal axis represents the time periods aftershocks while the vertical axis represents responses of the variables included by the model. It can be seen that all results are within lower and upper bounds.

After one standard deviation shock to export, real exchange rate results in a negative response and increases until period 3 in Figure 5.2.2. The effect of the shock is quite minor and disappears after five periods. In the short run, increase in exports causes to increase demand for Turkish liras, so TL appreciates and reel exchange rate

(51)

29

depreciates. In long run, reel exchange rate appreciates because Turkey is an import-based economy and intermediate good of exports actually are imported.

Figure 5.2.2: Response to Exports

After one standard deviation shock to imports, real exchange rate results in a negative response similar to export case and increases until period three as it can be seen form the graph in Figure 5.2.3. The effects of the shock disappear after five periods. In long run, a rise in imports appreciates the reel exchange rate.

(52)

30

After one standard deviation shock to real interest rate in Figure 5.2.4, real exchange rate results in a positive response and decreases for a short time period. In long run, a rise in reel interest rate appreciates Turkish liras.

Figure 5.2.4: Response to Interest Rate

5.3 Model III

Finally, we estimate our last model which include both trade flow and capital flow variables as follows:

Real Exchange Rate = (Direct Investment, Portfolio Investment, Other Investment, Exports, Imports, Real Interest Rate, Dummy2008)

We again check if there is an autocorrelation problem or not by using LM test. We obtain best results for lag 4 among LM tests with different lags. The results for lag 4 are presented in Table 5.3.1.

(53)

31

Table 5.3.1: VAR Residuals with Lags

Lags LM-Stat Prob

1 47.75922 0.5235 2 57.25413 0.1956 3 58.27240 0.1711 4 48.02199 0.5127 5 65.42560 0.0583 6 35.14654 0.9318 7 41.76962 0.7585 8 54.89326 0.2612 9 53.59250 0.3026 10 49.97840 0.4343 11 52.03433 0.3566 12 63.30111 0.0823

Then, we check r roots in order to control stability of our model. The results are presented in Table 5.3.2. When we check modulus values in Figure 26, we again observe no modulus that is greater than one. This shows that our VAR model 3 is stable.

(54)

32

Endogenous variables: D(REELEXCSA) D(DIRECTINV) PORTFOLSA OTHERINSA D(EXPORTSA) D(IMPORTSA) REELINTSA

Exogenous variables: C DUMMY1 Lag specification: 1 4 Root Modulus 0.966138 0.966138 -0.628405 + 0.589851i 0.861869 -0.628405 - 0.589851i 0.861869 0.134345 + 0.769967i 0.781599 0.134345 - 0.769967i 0.781599 -0.768837 - 0.113562i 0.777178 -0.768837 + 0.113562i 0.777178 -0.443054 - 0.631524i 0.771440 -0.443054 + 0.631524i 0.771440 0.239501 + 0.731115i 0.769344 0.239501 - 0.731115i 0.769344 0.768414 0.768414 0.651543 - 0.176891i 0.675129 0.651543 + 0.176891i 0.675129 -0.189522 - 0.634742i 0.662432 -0.189522 + 0.634742i 0.662432 0.292739 - 0.551845i 0.624683 0.292739 + 0.551845i 0.624683 -0.442459 + 0.330416i 0.552218 -0.442459 - 0.330416i 0.552218 -0.521665 0.521665 0.026016 + 0.464431i 0.465159 0.026016 - 0.464431i 0.465159 0.392783 0.392783 0.279747 + 0.186717i 0.336335 0.279747 - 0.186717i 0.336335 -0.274266 0.274266 -0.151934 0.151934

We also confirm the stability of our model by observing the graph of AR characteristic inverse root in Figure 5.3.1. Because all roots are inside the AR root, we do not have a problem.

(55)

33

Figure 5.3.1: Inverse Roots of AR

We also implement variance decomposition and the results are given in Table 5.3.3. When we check the results in the table, it is seen that 100% of error prediction variance during period 1 is explained by itself. In period 2, the real exchange rate explains 90.22% of error prediction variance itself while foreign direct investment, portfolio investment, other investment, export, import, and real interest rate explain 5.41%, 2.47%, 0.98%, 0.05%, 0.43 and 0.43% of error prediction variance respectively. Therefore, we can conclude that our model 3 is not significantly better than model 1. When we also consider all periods, we see that the percentage of error prediction variance explained by itself decreases with time and the percentage of error prediction variance explained by foreign direct investment, portfolio investment, other investment, export, import, and real interest rate increases with time. They converge to 8%, 7.5%, 4.5%, 3%, 1% and 1.5%. It is apparent that foreign direct investment is very important among other variables.

(56)

34

Period S.E. D(REEXSA) D(DIREC) PORTSA OTHERSA D(EXPSA) D(IMPSA) REINTSA 1 0.045873 100.0000 0.000000 0.000000 0.000000 0.000000 0.000000 0.000000 2 0.050341 90.22825 5.414022 2.467696 0.975992 0.052559 0.433124 0.428360 3 0.051864 87.63458 5.281215 4.441706 1.297278 0.284488 0.586127 0.474605 4 0.053933 82.46906 7.342888 4.153060 2.441389 2.174154 0.657363 0.762091 5 0.055473 77.97348 7.253232 6.760684 3.094080 2.937033 0.638244 1.343250 6 0.056114 76.38751 7.254455 6.945517 4.447737 2.893644 0.667391 1.403746 7 0.056448 75.68015 7.830193 7.132426 4.395934 2.909322 0.664743 1.387237 8 0.056657 75.14524 7.907028 7.087778 4.533237 2.939049 1.007125 1.380543 9 0.056829 74.70148 8.030955 7.351286 4.506377 2.922876 1.035974 1.451052 10 0.056994 74.27573 8.015030 7.411698 4.644305 2.992756 1.092258 1.568224

We end our analysis for model 3 with impulse response analysis. We present the response of real exchange rate to one standard deviation shocks of foreign direct investment, portfolio investment, other investment, export, import, real interest rate, and real exchange rate.

After one standard deviation shock to real exchange rate in Figure 5.3.2, real exchange rate results in a positive response and then become negative in period three. The effects of the shock disappear after period five.

(57)

35

After one standard deviation shock to foreign direct investment in Figure 5.3.3, real exchange rate results in a positive response. In second period, the response starts to decrease and fluctuates in following periods. It disappears after five periods. When direct investment rises, TL appreciates and dollar depreciates in the long run. Dollar appreciation leads to lessen exports and rise imports. Thus, dollar appreciates and TL is depreciates.

Figure 5.3.3: Response to Direct Investment

After one standard deviation shock to portfolio investment in Figure 5.3.4, real exchange rate results in a negative response and become positive after period three by increasing during four periods. Increase in portfolio investment causes TL appreciation and reel exchange rate depreciation. Therefore, imports increase and exports diminish. Real exchange rate appreciates.

(58)

36

Figure 5.3.4: Response to Portfolio Investment

One standard deviation shock to other investment in Figure 5.3.5 causes real exchange rate to respond negatively and moves back to equilibrium after eight periods. This shock is actually statistically insignificant. Like portfolio investment, a rise in other investments result in reel exchange rate depreciation.

Figure 5.3.5: Response to Other Investment

After one standard deviation shock to export in Figure 5.3.6, real exchange rate results in a negative response and increases until period three. The effects of the

(59)

37

shock disappear after five periods. When export increases, TL gains value and dollar depreciates. The decline in value of dollar leads to rise import and so the US dollar appreciates.

Figure 5.3.6: Response to Export

After one standard deviation shock to import, real exchange rate results in a negative response similar to export case and increases until period 3 as it can be seen form the graph in Figure 5.3.7. When import rises, the reel exchange rate increases in the long run.

(60)

38

After one standard deviation shock to real interest rate in Figure 5.3.8, real exchange rate results in very small response. An increase in reel interest rate encourages capital inflows so reel exchange rate depreciates.

(61)

39 6. CONCLUSION

In this thesis, we investigate the effects of capital inflows on real exchange rate and compare these effects with the effects of other factors like imports and exports.

In the literature, we observe that imports and exports are considered as the main determinant factors of the real exchange rate, but capital flows have become more important in explaining the changes in the real exchange rate recently because liberalization increases international capital flows. Throop (1993) claims that flexible price models is not successful in explaining exchange rates so he advocates that sticky price models with real interest rate is more successful. Furthermore, we make our analysis with real exchange rates for better comparison during our study. Jongwanich and Kohpaiboon (2010) analyze the relationship between capital flows and real exchange rate by regarding capital flows into three groups foreign direct investment, portfolio investment and other investment different from others who consider capital inflow as a whole. We also add capital inflows into our models by dividing them into three groups as foreign direct investment, portfolio investment and other investment.

In order to examine the relationship between real exchange rate and its determinants, we use the monthly data taken from electronic data delivery system by the Central Bank of Turkey and the website of IMF from 2003:1 – 2016:6.

In our study, we carry out our analysis by using VAR approach that is one of the most widely used method for determinants of exchange rate in the literature because it allows to solve simultaneous equations without considering variables as endogenous and exogenous. Khan and Abbas (2015) use auto regressive distributed lag model to examine the data that proves the strong relationship between these

(62)

40

variables. We estimate three different models explaining real exchange rate. Our model 1 including capital flows in the forms of foreign direct investment, portfolio investment and other investment is better in explaining variance in real exchange rate than our model 2 including exports and imports. This is an expected result because capital flows has become more important than trade flows due to globalization. The findings of the third model which covers all trade and financial variables also support this result.

Meanwhile, among the variables we used in model 1, foreign direct investment is the most explanatory factor for the real exchange rate. However, there are several studies in the literature, finding that portfolio investments and other investments are more significant on the changes in the real exchange rate movements.

This study can also be extended by adding other variables and investigating their effects on real exchange rate and also by implementing other estimation methods.

(63)

41 REFERENCES

Aggarwal, S. (2013) “The Uncovered Interest Rate Parity Puzzle in the Foreign

Exchange Market” New York University Press 101(2).

Alper, C. E., Ardic, O. P., and Fendoglu, S. (2007) “The Economics of Uncovered

Interest Parity Condition for Emerging Markets: A Survey” Munich Personal RePEc Archive.

Arat, Kürşad (2003) "Türkiye'de optimum döviz kuru rejimi seçimi ve döviz

kurlarından fiyatlara geçiş etkisinin incelenmesi”, The Central Bank of Turkey, Expert Thesis.

Athukorala, P. C., and Rajapatirana, S. (2003) “Capital inflows and the real

exchange rate: a comparative study of Asia and Latin America” The World Economy, 26(4), 613-637.

Balkan, E., Biçer, G., and Yeldan, E. (2002) Patterns of financial capital flows and

accumulation in the post-1990 Turkish economy. mimeo, METU, Ankara.

Barışık, S., and Demircioğlu, E. (2012). “Türkiye’de Döviz Kuru Rejimi,

Konvertibilete, İhracat-İthalat İlişkisi (1980-2001)” Uluslararası Yönetim İktisat ve İşletme Dergisi, 2(3), 71-84.

Bergin, P. R., Glick, R., and Wu, J. L. (2017) “Conditional PPP and Real Exchange

Rate Convergence in the Euro Area” Journal of International Money and Finance, 73, 78-92.

Bergvall, A. (2004) “What determines real exchange rates? The Nordic countries” The

Scandinavian Journal of Economics, 106(2), 315-337.

Bhuiyan, R. (2008) “Monetary transmission mechanism in a small open economy: a

Bayesian structural VAR approach” (No. 1183). Queen's Economics Department Working Paper.

Charfi, M. F. (2013) “Capital flows, real exchange rates, and capital controls: What

is the scope of liberalization for Tunisia?” Panoeconomicus, 60(4), 515-540.

Chinn, M. D. (2012) “Macro approaches to foreign exchange determination”. J.

(64)

42

Combes, J. L., Plane, P., and Kinda, T. (2011) “Capital flows and their impact on

the real effective exchange rate”, IMF Working Paper, No:11/9

Combes, J. L., Kinda, T., & Plane, P. (2012) “Capital flows, exchange rate

flexibility, and the real exchange rate” Journal of Macroeconomics, 34(4), 1034-1043. Değerli, A. and Fendoğlu, S. (2013) “Reserve Option Mechanism as a Stabilizing Policy Tool: Evidence from Exchange Rate Expectations,” CBRT Working Paper No. 13/28.

Edwards, S. (1988) “Real and monetary determinants of real exchange rate behavior:

Theory and evidence from developing countries” Journal of Development Economics, 29(3), 311-341.

Erdemlioglu, D. M. (2007) “A New Test of Uncovered Interest Rate Parity: Evidence

from Turkey”, MPRA Paper No. 10787

Ertekin Murat (2003) “Döviz Kuru Rejimleri ve Türkiye’nin Dış Ticareti”, Dış Ticaret Müsteşarlığı, Ekonomik Araştırmalar ve Değerlendirme Genel Müdürlüğü, Ankara, 1-17,

Froot, K. A., and Rogoff, K. (1995) “Perspectives on PPP and Long-Run Real

Exchange Rates” Handbook of international economics, 3, 1647-1688.

Gaytancıoğlu, S. (2010), “Rasyonel Beklentiler Teorisi Çerçevesinde Türkiye’de 1994 Krizi”, Ekonomi Bilimleri Dergisi, 2(1), 139-46.

Goldberg, L. S., & Klein, M. W. (1997) “Foreign direct investment, trade and real

exchange rate linkages in Southeast Asia and Latin America” (Vol. 6344). National Bureau of Economic Research.

Goldberg, L. S. (2009) “Exchange rates and foreign direct investment” The Princeton

Encyclopedia of the World Economy, 1, 393-396.

Gumus, G. K., Duru, A., and Gungor, B. (2013) “The relationship between foreign

portfolio investment and macroeconomic variables” European Scientific Journal, ESJ, 9(34).

Jongwanich, J., and Kohpaiboon, A. (2013) “Capital flows and real exchange rates

in emerging Asian countries” Journal of Asian Economics, 24, 138-146.

Karpuz, F., and Kızıltan, A. (2014) “TÜrkiye’de Kısa Vadeli Yabancı Yatırımlar ile