The Effects of FDI on Greek Economy: An Empirical Analysis

This paper investigates the effect of Foreign Direct Investment (FDI) on economic growth in Greece, within a framework that also accounts unemployment rate, using annual data covering the period 1970 to 2017. Several econometric models are applied including the ARDL bound test approach for cointegration as well as ECM-ARDL model for causality. The results of the study confirm the existence of a long run relationship among the examined variables. The Granger causality results indicated a strong unidirectional causality between economic development and foreign direct investments with direction from economic development to foreign direct investments. Finally, the variance decomposition method and the impulse response functions are used to test the strength of causality between the variables. The results of the study offer new perspectives and insight for new policies for sustainable economic development, increasing investments and reducing unemployment.


Introduction
In recent years, there is a growing interest in the relationship between foreign direct investments, unemployment and economic growth. The economic crisis which started in 2008 has created serious concerns about high unemployment rates and negative growth. Today, despite the continued recovery in most European countries, there are still countries that are facing serious problems due to high unemployment rates. In 2012, unemployment in European Union reached 26 million people (AMECO, 2014).
There are several discussions about how foreign direct investments may be a possible solution in unemployment reduction and economic growth. Many economists believe that FDI enhances private investments, encourages the creation of new jobs, transfers knowledge and technological skills in the Regarding the studies that examine the impact of foreign direct investments on economic performance in a group of countries, Hsiao and Hsiao (2006) examined the relationship between FDI, exports and GDP for eight rapidly developing East and Southeast Asian economies using data covering the period [1986][1987][1988][1989][1990][1991][1992][1993][1994][1995][1996][1997][1998][1999][2000][2001][2002][2003][2004]. Their results showed the existence of a bidirectional causality relation between exports and GDP. In addition authors argued that FDI has unidirectional effects on GDP directly and indirectly through exports. Dritsakis and Stamatiou (2014) investigated the relationship between exports, FDI and economic growth in five Eurozone countries using data for the period 1970 to 2011. Their results revealed bidirectional causality relation between exports and economic growth. In addition authors argued that there is no causality between economic growth and FDI nor between FDI and exports, for the examined period. Agrawal (2015) examined the relationship between FDI and economic growth in the BRICS economies over the period 1980-2012. The results of the study revealed that there is a causal relationship between FDI and economic growth in the long run, with direction from FDI to economic growth. Dritsakis and Stamatiou (2017) examined the interactions between FDI, exports, unemployment and economic growth for thirteen new member states of European Union. Using annual data for the period 1995-2013 the argued that that there is bidirectional causal relation among exports and economic growth, in the long run. In addition they found that a unidirectional long term causal relationship between economic growth and unemployment exists, with direction from economic growth to unemployment.
Finally, the same authors (Dritsakis & Stamatiou, 2018) applied a similar study for the fifteen old EU members using data covering the period 1970-2015. Their results showed three bidirectional causalities between economic growth and exports, exports and FDI, and exports and unemployment and three unidirectional causalities running from FDI to economic growth, FDI to unemployment and from economic growth to unemployment.

Data
The variables that are used in this study are Gross Domestic Product ( converted to natural logarithms. The descriptive statistics for all variables are illustrated in Table 2.

Unit Root Tests
The literature proposes several methods for unit root tests. Since these methods may give different results, we selected Dickey-Fuller (ADF) (1979,1981), Phillips-Perron (P-P) (1988) and Elliott, Rothenberg and Stock (DF-GLS) (ERS) (1996). In all these tests, the null hypothesis is that the variable contains a unit root (i.e., it is not stationary).

ARDL Cointegration Approach
We continue by testing the long run relationships between the examined variables using the ARDL approach (Autoregressive Distributed Lag) which developed by Pesaran et al. (2001).
This method has the following econometric advantages: 1) The bounds of ARDL approach are valid regardless of whether the variables are integrated I(0) or I(1).
2) The bounds of ARDL approach provide effective and consistent empirical evidence for small data samples.
3) The ARDL model is valid by taking a sufficient number of lags. The optimal lag length for the first difference of regressions is selected by the minimum value of Akaike (AIC), Schwarz (SIC) and Hannan-Quinn (HQC).
4) The ARDL method compared with other cointegration methods can distinguish and eliminate problems between dependents and independents variables such as the problem of autocorrelation and endogeneity.
5) Moreover, a dynamic error correction model can be derived from the ARDL method through a simple linear transformation. The dynamic ECM model integrates the short run dynamic with the long run equilibrium without losing any long run information.
The autoregressive distributed lag (ARDL) cointegration technique as a general vector autoregressive (VAR) model of order p: where Z t is a column vector composed of the three variables.
Thus, before we begin with the ARDL model we find the order of the VAR model, the lag length of the variables in the VAR model. Then, we use the minimum value of Final Prediction Error (FPE), Akaike Information Criterion (AIC), Schwarz Information Criterion (SIC), Hannan-Quinn Criterion (HQC), and Likelihood Ratio (LR) to find the optimal lag length of the variables.
The ARDL models that are used in this study are the following: where Δ denotes the first difference operator and ε 1t , ε 2t , ε 3t are error terms assumed to be independently and identically distributed.
Since the calculation of ARDL bounds is sensitive in the selection of the lag length, we select the optimal lag length from the first difference of the dependent variables by the minimum values of criteria Akaike, Schwarz and Hannan-Quinn in accordance with the following models.
where LUN t , LFDI t , and LGDP t are the dependent variables, α 1i , α 2i , and α 3i are the long terms and (p, q, c) are the optimal lag lengths of the ARDL model.  (2), (3) and (4)  the F-value exceeds the critical value of upper limit. Also, we accept the null hypothesis of no cointegration when the F-value is lower that the critical value of lower limit. Finally, the decision of cointegration is unclear when the F-value is between the lower and the upper limit (Pesaran et al.,

2001).
Then, we examine the long run relationships between the variables using the following equations: Moreover, a dynamic error correction model can be derived from the ARDL bounds test through a simple linear transformation. The dynamic unrestricted ECM integrates the short run dynamic with the long run equilibrium.
The dynamic unrestricted error correction model is expressed as follows: where ECM t-1 is the error correction term. The coefficient of error correction term (ECM t-1 ) should be negative and statistically significant. This coefficient indicates the speed of adjustment, how quickly the variables return to the long run equilibrium.

Stability of the Model
The existence of cointegration derived from equations (11), (12) and (13)

Granger Causality Analysis
After the long run relationship between variables, we examine the direction of causality using the ECM-ARDL model. The equations that are used to test Granger causality are the following: where i (i=1,…p) is the optimal lag length determined by the Akaike Information Criterion (AIC), ECM t-1 is the lagged residual obtained from the long run ARDL relationship presented in equations (8), (9) and (10), λ 1 , λ 2 , λ 3 are the adjustment coefficients, and u 1t , u 2t , u 3t are the disturbance terms assumed to be uncorrelated with zero means N(0,σ).

Variance Decomposition Method and Impulse Response Function
In order to obtain reliable estimations and further inferences on Granger causal relationships among the variables, we apply the Variance Decomposition Method (VDM) and the Impulse Response Functions (IRF) analysis.
The VDM allow us to evaluate the strength of causality beyond the selected sample period. This method measures the percentages of a variable's forecast error that is explained by another variable. In addition, IRF is used to determine the positive or negative responses of a variable to a one standard deviation shock of another variable, either in the short run or in the long run (Stamatiou & Dritsakis, 2019). This means that we can observe the direction, magnitude and persistence of foreign direct investments to variation in economic growth and unemployment rate.

Empirical Results
In the empirical analysis we use annual data concerning foreign direct investments inflows in Greece, unemployment rates and gross domestic product. We begin by testing the stationarity of three variables (FDI, UN and GDP).

Unit Root Results
Applying the unit root test of ADF by Fuller (1979, 1981), P-P by Philips and Perron (1988) and DF-GLS by Elliott et al. (1996) we present the results in Table 3. www.scholink.org/ojs/index.php/ijafs   As can be seen from Table 3, the results showed that FDI is stationary in levels in all the test that were applied which means that FDI is integrated I(0), while the other two variables are stationary in first differences which means that unemployment and GDP are integrated I(1). Therefore, we choose ARDL bounds test because there are variables with different integration order.

Cointegration Results
The process of cointegration applied to estimate the parameters of equations (2), (3) and (4)  The results of these criteria are presented in Table 4.  Note. *denotes the optimal lag selection.
The results of Table 5 show that there is one cointegrated vector (F-statistics seem to exceed upper critical bounds at 10%) confirming the existence of long run relationship among the series in equation   Table 6 presents the results of long run and short run relationship between the variables in our model. From the results of Table 6 we can see that in the long term equation of FDI an increase 1% of GDP will cause an increase 1.82% of FDI approximately, while a decrease in unemployment by 1% will cause an increase 0.31% of FDI. The ECM t-1 is negative and statistically significant which implies a long run relationship between the examined variables in the model. This means that in the short term the deviations from the long run equilibrium are adjusted by 91.4% every year. Finally, the diagnostics tests show that the error terms of the short and long run model are normally distributed and free of serial correlation, heteroskedasticity and ARCH problem. The Ramsey reset test suggests that functional form for the models is well specified.

Instability Tests
The ECM of equation (3) is selected to implement Brown et al. (1975) stability tests. The graphs of these tests are shown in the next figures (Figures 1-6).   As can be seen from the above figures, the graphs of statistical CUSUMQ and Recursive Residuals are not within the critical values at 5% significance level, both in the short and in the long run models. This means that all the coefficients in ECM are not stable. Table 7 reports the results on the direction of long and short run causality. Notes. *, ** and *** show significant at 1%, 5% and 10% levels respectively. Δ denotes the first difference operator. Recursive Residuals ± 2 S.E.

Causality Results
Published by SCHOLINK INC.
From results of Table 7 we see that there is a short run, a long run and a strong unidirectional causality relation between economic development and foreign direct investments with direction from economic development to FDI. The knowledge about the direction of causality helps policy makers to develop a proper economic policy.

Variance Decomposition and Impulse Response Analysis Results
To further explore the dynamic interactions between the foreign direct investments, unemployment and economic growth we proceed with the Variance Decomposition Method (VDM) and Impulse Response Function (IRF) techniques. The results of VDM are provided on the following table. The empirical results reveal that the most significant shocks effect of FDI (96.19%) is contributed by its own innovative shocks. The contribution of GDP to FDI is minimal and is 0.23%. In addition, a standard deviation shock stemming in unemployment attributes FDI by 3.56%.
Also, a contribution of 73.22% exists in GDP by shocks arising by its own innovative shocks.
Furthermore, a quite large portion of GDP is explained by innovative shocks stemming by UN (i.e., 22.74%) and the rest is being explained by FDI (i.e., 4.03%).
Finally, the contribution of FDI and GPD to UN is 7.03% and 14.92% respectively and the rest is being explained by its own standard innovative shocks.

Figure 7. Impulse Response Function
From the above figure we see that FDI is found to be significantly responsive to its own shock in the first 4 years. In the long run, these effects tend to zero. Besides, shocks in UN and GDP seem to have a slight effect (minimal) on FDI during the examined period.
Shocks in FDI cause an increase on UN over the first 5 years followed by a decrease for the remaining period. In addition, shocks in GDP cause a decrease in UN for the first three years followed by a steady increase for the rest 7 years. UN is significantly and positively responsive to its own shocks in the first 3 years, whereas there is a negative impact over the remaining 7 years.
Finally, shocks in FDI and UN cause a decrease on GDP over the first 5 and 3 years respectively, followed by a an increase for the remaining period. GDP is significantly and positively responsive to its own shocks in the first 3 years, whereas there is a negative impact over the remaining 7 years.

Conclusions and Policy Implications
The main objective of all the governments is the connection of growth and investments. However, the connection between FDI and unemployment is not easy to be determined by policy makers. Some economists argue that FDI inflows have a positive impact in the labor market only for the skilled workforce. This means that in the long term the quality of the work force is being improved. Some other argue that green investments in high tech industries tend to have a long term improvement in the economy of a country. So, this type of FDI inflows should be the priority of governments' policy and especially in Greece with the abundant natural wealth. In this paper we investigate the relationships between FDI, growth and unemployment in Greece over the period 1970-2012. In the empirical investigation we use ARDL approach as developed by Pesaran et al. (2001) and the ECM-ARDL model to find the casual relationships between the examined variables. In addition, for the test of the dynamic causal relationship we used the variance decomposition approach in combination with the impulse response functions.
The results of the study show that in the long term an increase 1% of growth will cause an increase 1.82% of FDI approximately, while a decrease in unemployment by 1% will cause an increase 0.31% of FDI. Finally, the causality results show both in the short and in the long run a strong unidirectional causality relationship with direction from economic development to FDI.
Based on variance decomposition method and impulse response functions we find that variations in economic growth respond more to shocks in foreign direct investments and unemployment. Economic growth seems to have negative response to shocks in FDI and unemployment rate. This implies that any policies, either active or passive, related with the labour market as well as the investment policy framework should be noted by the government in order to enhance economic growth.
The analysis of equations of FDI in the short run and in the long run shows that an increase of FDI will increase growth and will reduce unemployment. Therefore, the Greek government should immediately implement policies to attract foreign direct investments and foreign capital. The attraction of these funds in Greece is closely linked to the public debt. The reduction of debt for Greece (either by reducing interest rates, or with longer repayment, or with a haircut) will be the trigger for new capital inflows which will increase FDI, will boost economic development and will help in reducing unemployment.