Examining the impact of Exports and Exchange rate on Economic Growth of Turkey

Purpose This study intends to analyze the long-run and short-run relationships along with the identication of causal links between exports, economic growth, and exchange rate in Turkey. Data/Design: This study uses auto-regressive distributed lags (ARDL) and Granger causality over time series monthly data from the year 2010–2018. The results indicate that exports are signicantly positively related to economic growth while the exchange rate is found to be negatively related to economic growth. Findings: Moreover, ndings from the test of Granger causality indicate that a unidirectional causal association is found from exports to foreign direct investment and economic growth and from economic growth to foreign direct investment. The Granger causality results indicate that an increase in exports accelerates the economic growth of Turkey and a change in growth rate and exchange rate leads to a change in foreign direct investment.


Introduction
Economic growth is affected by various factors and has been widely studied in the literature. Among others, one of the main factors is the country's exports (Medina-Smith, 2000). Export provides employment opportunities and makes contributions to the country's balance of payments. Exports encourage local producers to be more e cient in the global market by using better and advanced production techniques. It helps in reducing the impacts of external shocks to the economy and promotes intra-industry trade among economies to integrate with the global market (Abou-Stait, 2005).
Exports impact on the economy can be identi ed via various channels. Country's real output increases by an increase in its exports. Besides, increase in country's exports encourage domestic rms to be specialized in the production of export goods and eventually leads an increase in the level of productivity. Moreover, export-oriented sectors employ more skilled and e cient labors . Trade allows a country to attain economies of scale in the global market by improving its production techniques as trade provides access to more advanced production techniques which will increase country's growth rate (Giles & Williams, 2000). Export-oriented country's real GDP growth and its rate will be higher than less export-oriented country (Dadora, 1991). Therefore, trade is considered a major factor in accelerating economic growth (Fajana, 1979).
On the other hand, among others, FDI (foreign direct investment) is also an economic growth's major indicator, particularly in economies having low saving rates. FDI not only provides capital but also the know-how of technology and management required for the rms in restructuration (Borensztein et al. 1998; Chao and Yu, 1994). FDI plays a crucial role in the country's development.
Many economies are gaining growth by attracting FDI and by promoting better investment environment to the foreign investors.
Foreign exchange policies have huge impact on economic growth and a signi cant divide is there between economists and policy-makers while discussing exchange rate and economic growth relationship. On a view of common laymen and politicians, lower exchange rate may accelerate economic growth. Whereas, Economists suggest that in long run, relative price of two currencies can be considered as an essential driver of economic growth and exchange rate may be treated as an endogenous variable while considering its contribution to economic growth.
Turkey is considered as an upper-middle income country, located between Europe and Asia, having a population about 80.7 million (WDI, 2017), and a candidate of EU membership. Turkey sustained its annual growth at 4-5%, despite its political and economic problems in the last decade and about 3-4% annual growth rate is forecasted for the next coming years (WorldBank, 2016). Turkey, as an emerging economy, is expected to be FDI-attractive market but because of its investment environment, it lagged behind other emerging economies (Euduardo and Martin, 2014; Erkilek, 2003). On average Turkey's performance is worse as compared to other emerging economies. From 2002 to 2007, Turkey's FDI in ows raised from 0.5-3.4% of GDP but later again because of global nancial crises, it reduced to 1.5% of GDP. However, EU accession process and new reforms in the Turkish economy (which includes liberalization and privatization) created a more suitable environment in attracting more foreign portfolio investment in the country. The causal relationships between exports, economic growth, and FDI in ows remain controversial. Thus, considered to be a crucial matter among researcher and economists.
Analyzing these relationships are more important particularly in emerging economies having and suffering more from economic problems (Ericsson and Irandoust, 2001;Zhang, 2001). This is the pioneer study analyzing the relationship between exports, exchange rate and economic growth of Turkey, where industrial production is used as a proxy for economic growth in order to avoid the income identity problem between exports and economic growth. This study intends to analyze the longrun and short-run relationships along with the identi cation of causal links between exports, foreign direct investment, economic growth, and exchange rate in Turkey by using the methodologies of auto-regressive distributed lags (ARDL) and Granger causality over time series monthly data from the year 2010-2018.

Page 4/23
The remaining study is organized in the following manner: the review of the literature, data and methodology part, ndings and discussions and then the conclusion of the study.

Literature Review
The relationship between the variables including FDI, export and economic growth has widely been studied in the literature ( Consequently, the FDI-led economic growth hypothesis could not be accepted for the case of Greece. The links between FDI, GDP, Export, and Employment are examined by Kersan-Skabic and Zubin (2009) for the economy of Croatia. The results show a negative relationship between FDI and employment while no relationship is found between export and GDP growth.
The study of Miankhel et al. (2009) investigated the nexus of export-FDI-economic growth, by employing VECM for the economies including Chile, Pakistan, Malaysia, Mexico, Thailand, and India. Their ndings indicate that the Malaysian economy is having a bi-directional causality link between FDI and economic development while the export-led growth hypothesis was accepted for all the studied economies. Besides, causality from export to growth and FDI was found in only Latin American economies.
For sixty-six economies, Makki and Somwaru (2004) examine the relationship between the variables; FDI, trade and economic growth. Human capital, FDI, domestic investment and trade was found to be important determinants of economic growth. Moreover, a strong relationship was found between trade and FDI in promoting economic growth. Hansen and Rand (2006) examined the link between GDP and FDI in ows in thirty-one emerging economies by employing Granger causality. They explored that FDI is positively associated with economic growth through adoption of new & advanced technology and through knowledge-transfer. Their results indicate bi-directional causality link between GDP and FDI in ows. Besides, previous literature focused on the nexus of FDI-exports-growth. However, this study also includes exchange rate while discussing the relationship between FDI, exports and economic growth by employing the ARDL Bound testing approach and Granger causality test to nd out the causal relationships between these variables for the Turkish economy over the period of 2010-2018.

Data And Methodology
Monthly data regarding the variable in ows of foreign direct investment (FDI) (Million USD), exports (EXP), exchange rate (EXR) and industrial production (IND) are taken from the Central Bank of Turkey (CBT). All the variables including dependent and independent variables used in this study are in real form; EXP is constant at 2010 USD prices, EXR is constant at 2003 USD prices and IND is constant at 2015 USD prices. The variable FDI is measured by real FDI adjusted by CPI and then was converted into log form. The Data description and source, descriptive statistics of the variables, Graphs, correlation table, which are included in the empirical model, are presented in the table below.
This study intends to analyze the in uence of exchange rate (EXR) and exports (EXP) on the economic growth (IND) in Turkish economy over the period of 2010 to 2018 by employing the ARDL Bound testing and also intends to nd out the causal links between these variables by using Granger causality test. The foreign direct investment (FDI) has been taken as control variable. Industrial production has been used as a proxy for real economic growth in Turkey.
Here, we lay our discussion about the relevant methods utilized for conducting of this paper. Which include the unit root tests and bounds test for cointegration and causality within the ARDL modelling approach. This model has been developed by Pesaran et al. (2001); and can be applied without considering the order of integration of the variables (whether regresses are purely I (0), purely I (1) or mutually co-integrated). This is particularly linked with the ECM models that are called VECMs.
In ARDL model normally rst step is determining the order of integration of each variable because it uses the variables at the level which they are stationary Shabbir and Muhammad (2019). For testing the stationary of the series, the paper uses the Augmented Dickey Fuller (ADF) unit root testing procedure (Dickey and Fuller, 1979). In the ADF test, we want to determine the size of the coe cient δ 2 in the following equation: The ADF regression tests for the existence of unit root of Z t , in all model variables at time t. The variable ΔZ t−1 shows the rst differences with n lags and nal εt is the variable that modi es the errors of autocorrelation. The coe cients, δ 0 , δ 1 , δ 2 , and βi are estimated ones. The null and alternative hypothesis for the presence of unit root in variable Z t is as follow: H 0 : δ 2 = 0, H 1 : δ 2 < 0 3.1 Testing for Granger causality in the Bounds test approach In this study we carry out the Bounds test to evaluate the existence of causality between exports, exchange rate, and economic growth in Turkey. The model for the relationship between export, exchange rate, and economic growth carried out in this paper is de ned as: INDt = f (EXPt, EXRt, FDIt) in which INDt, EXRt, and FDIt represent net export, exchange rate, and foreign direct investment. The following equations are the error correction models estimated under the ARDL Bounds testing methodology: EXR t = α 31 + δ 32 IND t−1 + δ 33 EXP t−1 + δ 34 EXR t−1 + δ 35 LFDI t−1 + Σ φ 3i IND t−i + Σ β 3i EXP t−i + Σ ψ 3i EXR t−i + Σϒ 3i LFDI t−i + ε t (4) LFDI t = α 41 + δ 42 IND t−1 + δ 43 EXP t−1 + δ 44 EXR t−1 + δ 45 LFDI t−1 + Σ φ 4i IND t−i + Σ β 4i EXP t−i + Σ ψ 4i EXR t−i + Σϒ 4i LFDI t−i + ε t (5) Where α 11 , α 21 , α 31 , and α 41 are constants for the four equations. We can test for cointegration among INDt, EXPt, EXRt, and LFDIt using the Bounds test approach. For Eq. After demonstration of cointegration then there has to be at least a uni-directional causality. Granger causality existence implies the presence of cointegration between variables which provides information regarding long run and short run Granger causality. For empirical aims, the error correction representation can be derived from the VECM Granger method provided as follow: In which (1 − L) is the difference operator, ECM t−1 is the lagged error correction term which is developed from long run co-integrating relationship while η1t, η2t and η3t are white noise serially independent random error terms.
If the rst difference of variables shows a signi cant relation it is an evidence for the direction of short run causality; whereas, long run causality is represented through a signi cant t-statistic concerning to the error correction term (ECM t−1 ). Error Correction model speci cation is a combination of short run equations and long run representation:  implies that an increase in exchange rate causes economic growth to fall while an increase in exports in a country accelerates its economic growth. However, the relationship between LFDI and IND is not signi cant as shown in Tables 8 and 9 as well. The rst step, done in the empirical analysis is to nd out the stationary of the dependent and explanatory variables by applying the unit root test known as ADF (ADF, 1979), whose results are presented in table 4.  After the unit root test, optimal lag length is found by considering ve various criteria and presented in Table 5 which shows that optimal length is 1 out of 3 lags. Note: (*) shows the lag order selected by different criteria.

Results And Discussion
As presented in Table 4, all the variables except LFDI are stationary at rst difference where the LFDI is stationary at its level. Therefore, the ARDL test (developed by Pesaran et al., 2001) is used to nd out cointegration showing the long-run correlation between the dependent and independent variables. The results of the ARDL Wald test and long-run coe cients are presented in Table 6, 7 and 8.  In Table 6, the F-statistic value exceeds upper and lower critical bounds at 1%, 5% and 10%, illustrated in Table 7, con rming the co-integration and the long-run association between the dependent and explanatory variables.
The outcomes postulate that exports are statistically signi cantly positively related to economic growth. Turkish economy will grow at 0.59% by a one percent increase in its exports. The exchange rate is found to have a negative relationship with economic growth which indicates that economic growth will reduce by 0.44% by a one percent increase in the exchange rate. Moreover, foreign direct investment is found to have a positive but insigni cant relationship with economic growth.

Error Correction Model (Investigating Short-run Relationship)
After analyzing the long run relationship between the variables, discussed in the model, short run relationships are investigated between FDI, exports, economic growth and exchange rate by employing ECM-ARDL method, which shows that if economy is hit by any shock, then how much correction would be made in each period towards the long run equilibrium, depicted by the term ECT in table 8. Error correction model speci cation is a combination of short run equation and long run representation. The error correction term (ECT) is found highly statistically signi cant and negative, con rming the presence of cointegration and the relationship between dependent and independent variables. The ECM coe cient or the error term (-0.824151) indicates the correction speed (if the economy hit by any shock) towards the long-run equilibrium of the economy, implying that the speed of correction is 0.82% over the period of one quarter. Moreover, in short run, EXP variable is found to have statistically signi cant and positive relationship with economic growth at a signi cant level of 5% as well as FDI variable at 1%. The results of short run dynamics by utilizing the error correction methodology are shown in Table 8.

Diagnostic Test
In order to check the robustness and accuracy of the model, some diagnostic tests are applied including; serial correlation, heteroscedasticity and normal distribution. In all diagnostic tests, the P-value is more than 5%, which implies that there is no serial correlation in the residual, no heteroscedasticity and the model is normally distributed. The results of the diagnostic tests are presented in table 9.

Stability Test
In order to check the stability of the model at 95% con dence interval, test of Cumulative Sum (CUSUM) is applied which shows that the statistics-plots are within the 5% signi cance level of critical bounds (see Figure 8). Therefore, con rming the stability of the economic growth's determinants over the period from 2010 to 2018.

Granger Causality (GC) Test
After nding the co-integration and long-run relationship between the dependent and explanatory variables, the test of Granger causality is applied to nd out the causal links between FDI, exports, exchange rate and economic growth of Turkey. The ndings from the GC test are presented in Table 10. According to the Granger causality test, a causal link (uni-directional) has been found from exchange rate to export as well as a causal association (uni-directional) is found from industrial production (a proxy for economic growth) to exports and foreign direct investment. Moreover, a causal link (uni-directional) is investigated from exchange rate towards industrial production. In addition, a causal link (bi-directional) from foreign direct investment to export and from export toward foreign direct investment. Granger causality results indicate that a change in industrial production of Turkey leads to a change in its export and foreign portfolio investment. Furthermore, a change in exchange rate causes change in exports. Additionally, bi-directional relation between foreign direct investment and export implies their causal relations.

Impulse Response
The test of the impulse response lies between 95% percent con dence interval which is illustrated in the gure 9.
The given graphs above we can observe that when there is a one standard deviation shock from IND, the industrial production is decreasing rapidly and from period 2 it is increasing and again declining till period 4 and afterward becomes stable. As well as exports and foreign direct investment are declining rapidly and get negative but exports become relatively stable after period 4 and remains in positive region; however, FDI becomes stable but remains in the negative regions after period 3. Last but not least, the exchange rate in the earlier stages has a relatively stable response in the negative part but declines gradually after second period; while, beyond the period 5 becomes stable but still in the negative region, which indicates that shocks in IND will have a negative impact on IND, EXP, EXR, and LFDI. We can observe that a shock in industrial production accounts for 100% deviation in the level of IND in the short-run and it decreases as the focusing horizon increases in table 11. Whereas, a shock in export counts for 3.7% variation in the level of IND in the short-run while it increases till period 5 and declines as the focusing horizon increases. In addition, a shock in exchange rate counts for 2.8% variation in the level of IND in the short-run and it increases as the focusing horizon increases. Moreover, a shock in foreign direct investment counts for 11.88% variation in the level of IND and it increases in the short-run and decreases beyond period 5 as the focusing horizon increases.

Conclusion
This study intends to analyze the long-run and short-run relationships along with the identi cation of causal links between exports, economic growth, and exchange rate in Turkey by using the methodologies of Auto-Regressive Distributed Lag (ARDL) and Granger causality over time series monthly data from the year 2010-2018. The outcomes indicate that exports are positively related while the exchange rate variable is negatively related to economic growth. According to the Granger causality test, a causal link (uni-directional) has been found from exchange rate to export as well as a causal association (unidirectional) is found from industrial production (a proxy for economic growth) to exports and foreign direct investment. Moreover, a causal link (uni-directional) is investigated from exchange rate towards industrial production. In addition, a causal link (bi-directional) from foreign direct investment to export and from export toward foreign direct investment. Granger causality results indicate that a change in industrial production of Turkey leads to a change in its export and foreign portfolio investment. Furthermore, a change in exchange rate causes change in exports. Additionally, bi-directional relation between foreign direct investment and export implies their causal relations.
In the GC test, economic growth and exchange rate are found as factors of FDI while export is found as a factor of economic growth; economic growth and exchange rate attracts FDI in ows in Turkey. Moreover, there is an indirect effect from exports to FDI; exports affect economic growth and economic growth affects FDI. The overall ndings suggest that exports should be promoted along with the liberalinvestment economic policies to boost the overall economic growth in Turkey.

Availability of data and material
The data is available on request Competing interests Figure 8 Cumulative sum of recursive residuals Figure 9 Impulse Response