Causal Links between Foreign Direct Investment and Trade in Turkey

This paper investigates the empirical evidence on the link between foreign direct investment and trade (export and import) in Turkey over the period from 1992:01 to 2008:04 by using the minimum LM unit root test for stationarity; Granger and Dolado-Lüthkepohl tests for causality. The test results based on the bi-variate VAR model indicate that there is no evidence of causality between foreign direct investment and trade in Turkey.

According to trade theory, whether FDI substitutes or complements trade depends on the motivation for FDI.If FDI is vertical where multinational firms geographically split stages of production, this is likely to stimulate trade.If FDI is horizontal where multinational firms produce final goods in multiple locations, this is likely to substitute for trade.If FDI is a substitute for trade, then FDI represents a diversion away from local production and exports to foreign production and affiliate sales.Conversely, if FDI and trade complements, then the local production benefits from investments abroad (Taylor and Wilson, 2006).In the neoclassical approach of trade theory, the paper of Robert Mundell (1957) is the first to focus on the relationship between capital movements and trade of commodities within the framework of the Heckscher-Ohlin-Samuelson (HOS) theory.The standard HOS theory of trade assumes homogeneous goods, constant returns to scale in production, identical, homothetic consumer preferences across countries and perfect competition in markets (Van Berkum, 2002).The HOS model suggests that international trade can substitute for international movement of factors of production including FDI and implies that international commodity trade involves an indirect exchange of factors between countries (Liu et al., 2001).Mundell (1957) argues that a tariff protection would generate a perfect substitution between FDI and trade.International trade and the international mobility of factors of production which includes FDI are substitutes rather than compliments for each other where there are barriers to trade.Trade impediments stimulate factor movements and that increased impediments to factor movements stimulate trade.The opposing view is that FDI and trade are complements (Helpman, 1984;Helpman and Krugman, 1985).In this view, the difference in factor endowments plays an important role in determining the direction in which trade and FDI flow.Helpman(1984) and Helpman and Krugman (1985) argue that the degree of specialization is a positive function of relative factor endowments.If there are substantial differences in factor endowments, the capital-abundant country tends to export services into the labour-abundant country in exchange for finished varieties of a differentiated good or a homogeneous good (Favara, 2007).Thus, FDI generates complementary trade flows from the labour-rich country.
As we can understand from the theorical consideration, it is very difficult to predict whether FDI and trade substitutes or complements.Existing empirical studies in the literature which use different data and estimation techniques gives mixed results.Blomstrom et al. (1988) examines the relationship between FDI and exports by using the United States (US) and Swedish firm-level data and find complementary relationship between FDI and export.Eaton and Tamura(1994) analyse the American and Japanese bilateral flows of FDI and trade over the period [1985][1986][1987][1988][1989][1990] and find that the relationship between FDI and trade is complementary.Pfaffermary(1996) finds by using Austrian cross-sectional industry level data that there is a complementarity relationship between FDI and exports.Fontagne and Pajot (1997) investigates the relationship between FDI and trade for France, Sweden, US, the European Union (EU) and Japan over the period 1984 and 1994.They find that FDI and trade mostly act as complements.Dunning(1998) argues that the relationship between FDI and trade are conditional on the kind of FDI and trade.Gopinath et al. (1999) investigate the relationship between FDI and trade in the US food industry over the period 1982-1994 by using foreign affilate sales, exports, affilate employment and FDI as endogenous variables and conclude that foreign sales and exports are substitutes.Pontes (2004) finds for high level of trade costs that FDI and trade act as complements and otherwise as substitutes.Mekki (2005) finds that FDI and trade are complemetary for the manufacturing sectors of economy but are substitutes for other sectors in Tunusia.The causal relationship between FDI and trade is as complex as substitution and complementary cases.Pfaffermary (1994) investigates the Granger causality relationship between Austrian outward FDI and exports and finds bi-directional causality among the variables.Jun and Singh (1996) find bi-directional causality between FDI inflows and exports for developing countries.Alguacil and Orts (2002) examine the impact of FDI on export in Spain by using quarterly data for the period 1970:Q1-1992:Q3 and find that there is a long-term Granger causality from FDI to exports.Alguacil et al. (2002) and Min (2001) investigate the relationship between FDI and export for Mexico and Malaysia, respectively, and find a positive causal relationship between FDI and exports.Liu et al. (2002) examine the relationship between growth, FDI inflows and trade for China and find bi-directional relationship between growth, FDI and exports but find only one way causality relationship to imports.Aizenman and Noy (2005) find bi-directional relationship between FDI flows and trade on the international level.

Foreign Direct Investment and Trade in Turkey since 1980s
The liberalization of foreign trade in Turkey started after the structural reforms in 1980.A new way opened in front of Turkey in terms of financial liberalization applications and structural adjustment programs.Import substitution policies have been abandoned and export-led growth strategy has been adopted.Price controls have been lifted, import regime gradually liberalized and FDI and export encouraging policies were followed (Hasanov and Omay, 2008).The foreign exchange regime was liberalized, banks were allowed to accept foreign currency deposit from citizens and to engage in foreign transactions and new market institutions were established.After the full liberalization of the capital account and the recognition of full convertibility of the Lira in 1989, there was a massive inflow of short term capital into the economy (Balkan and Yeldan, 2001).The cumulative of FDI until 1980 was only $ 228 million.Since the mid-1980s, foreign investors have been taking increasingly prominent role in the Turkish economy as the recent liberal foreign investment and privatization policies began to show their results (Erdal and Tatoğlu, 2002).The distribution of cumulative authorized FDI by country of origin can be seen in Table 1.
[Insert Table 1 here] The table reports that European countries take the lead by accounting for over two-third of the total value of FDI.Following the European countries are the USA and Far Eastern countries with having shares of 12.0% and 6.4%, respectively.According to the report of August 1999, manufacturing sector accounts for 56.7 per cent of cumulative FDI authorizations with services constituting nearly 41 per cent.Agriculture and mining, however, take very small portion of FDI with both sectors together constituting 2.51 per cent of cumulative FDI authorizations.The integration process of the Turkish economy into the world economy gained further momentum following the Custom Union with the EU in 1996.Nearly 12 years after the Customs Union, EU has a stable share of around 50% in Turkey's foreign trade.The Custom Union is important for Turkey's production structure, since 85% of total imports are intermediary and investment goods.The biggest increase in imports from the EU is in consumer products.The share of consumer goods rose to 15.1% in 2007 from 7.3% in 1994, while the share of investment goods decreased to 21.4% from 29.3%.On the other hand, the share of investment goods in total exports from Turkey to the EU rose to 14% in 2007 from 2.9% in 1994, and the share of intermediate goods rose from 32.5% to 38.7%, while the share of consumer goods declined to 47.1% from 64.6% in the same period (İzmen and Yılmaz, 2009).However, liberalization brought important structural problems and deep crises to the Turkish economy.Integration into the European Customs Union affected the Turkish trade balance predominantly until as recently as 2006.The 1997 Asian financial crisis and the subsequent Russian crisis in 1998 affected the Turkish trade balance in a negative way because the export volume could not be increased while the import volume maintained a growing trend.And as the anti-inflationary stabilization program based on nominal exchange anchor appreciated the domestic currency, it also deteriorated the trade balance.The stabilization program was unsuccessful in attaining the ex-ante crawling-peg regime leading to the February 2001 Turkish economic crisis.This crisis resulted in a massive depreciation of domestic currency against hard currencies such as the US Dollar and Euro.These developments brought about a narrowing effect upon imports, and supported the exports in a positive way.However, having a stabilized economy and having attained a sustainable growth path, trade balance again began depreciating until 2006, although massive increases in exports can be provided by policymakers.As can be seen from the table, there is a prevaling reality that the import volume is more than export volume.Therefore, Turkey has been subject to a negative trade balance, balance of payment difficulties and necessary capital accumulation.The FDI performance of Turkey has significantly improved since 2005.Turkey was ranked the fifth among the emerging markets in terms of attracting FDI inflows in 2006.Total FDI inflows added up to US$ 52 billion during 2005-2006 while the total FDI stock reached US$ 115 billion at the end of 2007 (İzmen and Yılmaz, 2009).Despite the fact that, the amount of FDI inflow increases in Turkey, the state is subject to substantial amount of negative trade balance, current account deficit and budget deficit.Therefore, it is essential to facilitate more FDI inflow to solve economic problems of Turkey.

Data
To investigate the relationship between foreign direct investment and trade in Turkey, we use the ratio of foreign direct investment to GDP (FDI/GDP), the ratio of export to GDP (EX/GDP) and the ratio of import to GDP (IM/GDP).The quarterly data over the 1992:01-2008:04 period are obtained from the IMF's International Financial Statistics database.All the variables are measured in Million dollars.

Methodology and Empirical Results
This section analyses the empirical relationship between foreign direct investment and trade in Turkey.On one hand, we investigate the relationship between FDI/GDP and EX/GDP.On the other hand, we investigate the relationship between FDI/GDP and IM/GDP.As a first step of our empirical analysis, we test for the order of integration of the series by using Augmented Dickey Fuller (ADF) and Philips Perron (PP) unit root tests.The test results are reported in Table 3. [Insert Table 3 here] The results for the order of integration of the variables do not seem to be clear cut.The variables are found stationary after first differencing according to ADF test but found stationary in levels according to PP test.These findings show that the variables are inconclusive either ) 0 ( I or ) 1 ( I depending on the specific unit root test procedures.Perron (1989) argues that the presence of structural change in ADF test can reduce the power of the test and that the changes in the deterministic components of the time series can lead to biased results.Therefore, in order to capture the effect of any possible structural breaks, we also use the minimum LM unit root test suggested by Lee and Strazicich (2003).This test is not affected by the breaks in examining the stationarity properties and allows for two endogenously determined breaks in the intercept and trend.

Minimum LM unit root test with two structural breaks
The data-generating process (DGP) for the minimum LM unit root test suggested by Lee and Strazicich (2003) Here, TB 1 and TB 2 denote the break dates.Model CC includes two breaks in the intercept and the trend, and is described by The LM unit root test statistic is obtained from the following regression: where t S  is a detrended series such that t and 1 y and 1 Z are the first observations of t Y and t Z , respectively. is the difference operator.The lagged terms , are inserted to correct for serial correlation in equation 2. The number of augmentation terms , is determined by following a "general to specific" procedure (starting with max k=4 for quarterly data) described in Strazicich et al. (2004).The unit root hypothesis is tested via the t-ratio of  , this statistic being denoted as  .The null hypothesis of a unit root is tested against the alternative hypothesis of trend stationarity.Structural break ( ) TB is determined by selecting all possible break points for the minimum t-statistic as follows: where . The critical values are tabulated in Lee and Strazicich (2003) for the two breaks case.
The results of the minimum LM unit root test with two structural breaks can be seen in Table 4.
[Insert ratio, the break points coincide at 1997:03 and 2001:02.These breaks can be attributed to the effects of Asian financial crisis and February 2001 Turkish economic crisis.The breaks for the ratio of GDP IM / occur at 1996:02 and 1998:03, the years of European Customs Union and Russian crisis, respectively.Engel and Granger (1987) and Yoo (2005) point out that if there is no evidence for cointegration among the variables, the specification of the Granger causality test will be a Vector Autoregression (VAR).If evidence for cointegration is found, then one needs to augment the Granger type causality test with a one period lagged error correction term.In a bi-variate case, we can represent the two variable VAR system by the following equations:

Granger Causality Test
where 1t y and 2t y denote the two stationary time series and 1t  and 2t  are assumed to be serially uncorrelated with zero mean and finite covariance matrix.k is the maximum lag order.When the null hypothesis is not rejected, it implies that 1t y does not Granger cause 2t y .
Given that the series are stationary in levels, it will be infeasible to consider a cointegration analysis between the variables.Therefore, the natural course is to estimate the VAR model in our analysis.Since the ratio of where t y  is detrended stationary series (Altınay and Karagol, 2004).By using detrended series, the VAR model is specifically constructed to examine the Granger causality relationship between FDI/GDP and EX/GDP and between FDI/GDP and IM/GDP.The lag structure of the VAR model for FDI/GDP and EX/GDP relationship is chosen by the Akaike Information Criterion (AIC), Schwartz Information Criteria (SC), LR test statistic, Final Prediction Error (FPE) and Hannan-Quinn Information Criterion (HQ), setting the maximum lag at 1.The optimal lag length of the VAR model for FDI/GDP and IM/GDP relationship is chosen by using the AIC, LR and FPE criteria, setting the maximum lag at four.The results of the Granger causality test are shown in Table 5.
[Insert Table 5 here] The test statistics show that there is no evidence of Granger causality between FDI/GDP and EX/GDP and between FDI/GDP and IM/GDP.To test causality relationship among the variables, we also use Modified Wald (MWALD) test developed by Dolado and Lüthkepohl (1996).This test procedure is a simple and intuitive method that overcomes the difficulties associated with standard Granger causality test that variables may or may not be integrated-cointegrated.

MWALD Approach to Causality
The Dolado and Lüthkepohl (1996) causality test uses the Augmented VAR model and does not require pretesting for cointegration properties.The Dolado-Lüthkepohl test can be directly applied to the series in levels, regardless of whether the series are integrated, stationary or stationary around a trend.To apply this test, the first step is to select the number of optimal lags for the variables in the VAR systems.In the second step of the causality analysis, we estimate VAR models with (k+1) lags and apply standard Wald test to the first k VAR coefficient.The Wald statistic will be asymtotically distributed as a chi-square, with degrees of freedom equal to the number of "zero restrictions" irrespective of whether the variables are , non-cointegrated or cointegrated of an arbitrary order.By using the FDI/GDP, EX/GDP and IM/GDP series in levels, we find the optimal lag length as 4 for both FDI/GDP -EX/GDP and FDI/GDP -IM/GDP relationships according to AIC, SC, LR, FPE and HQ criterations.Following Dolado and Lüthkepohl (1996), we estimate VAR model by adding only one extra lag to the true lag order (4+1=5) and perform a Wald test on the coefficient of the first 4 lags.Table 6 gives the results from the Dolado and Lüthkepohl (1996) causality test.
[Insert Table 6 here] As can be appreciated from the table, there is no evidence of a causal relationship between FDI/GDP -EX/GDP and FDI/GDP -IM/GDP.These results show that there is no causal relationship between foreign direct investment and trade in Turkey.As reported before, despite the fact that, the amount of FDI inflow increases in Turkey, the state is subject to some economic problems and in order to solve these problems, it is essential to facilitate more FDI inflow.Thus, Turkish economy should motive to improve some conditions, such as educated labor, developed financial system and political instability, before having foreign investment.

Conclusion
This paper examines the causal relationship between FDI and trade in Turkey by using quarterly time series data for the period 1992:01 to 2008:04.The ADF and PP unit root tests are performed to determine whether the series are stationary in levels or require first differencing.The findings show that these unit root test procedures give inconclusive results.The series are found stationary after first differencing according to ADF test but found stationary in levels according to PP test.Following Perron(1989), we take into account possible structural breaks in the data to avoid from the biased results.The minimum LM unit root test of Lee and Strazicich (2003) which allows for two endogenously determined breaks in the intercept and trend is used in the analysis.The findings indicate that the series are stationary in the levels with two structural breaks.To test causality relationship among the variables, Granger causality and Dolado and Lüthkepohl (1996) causality tests based on VAR models are applied.These test results imply that there is no causal relationship between FDI and trade in Turkey.According to these results, it can be said that Turkish economy should motive to improve some conditions, such as educated labor, developed financial system and political instability, before having foreign investment.
parentheses indicate the number of lags in the augmented term of the ADF regression and are determined by using AIC information criteria.The number of truncation lags for PP test is chosen based on the Newey-West method.The unit root tests include a constant and time trend.a, b represent the significance at 1% and 5% levels, respectively.
Table 2 reports the foreign trade indicators over the period 1980-2008.
[Insert Table 2 here] As explained in the Section 3, FDI performance of Turkey has significantly improved since 2005 and total FDI stock reached US$ 115 billion at the end of 2007.According to the results for GDP EX / Table 4 here] Based on the minimum LM unit root test results, we find that all the series are stationary in levels.Estimated break points are shown in Table 4 under columns 1 TB and 2 TB .The second break in the intercept ( 2t D ) and the first break in the trend ( 1t T ) are significant for the

Table 2 .
Foreign Trade Indicators 1980-2008 ($ Billion) Source: State Planning Organization of Turkey, Undersecretariat of Trade & Treasury of Turkey.

Table 3 .
The results of ADF and PP unit root tests