Unemployment Effects of Greenfield and Brownfield Investments in Post-transition European Union Members

42020, XXIII, 2 Economics DOI: 10.15240/tul/001/2020-2-001 Introduction The foreign direct investment (FDI) infl ows have exhibited substantial increases with contri bution of relaxation of the impediments over the international fl ows of goods, services and capital mainly resulting from the accelerating liberalization and globalization as of 1980s. Consequently, international FDI infl ows reached USD 3.111 trillion in 2007, but then contracted due to economic crises and the increasing protectionism concerns in the recent years and became USD 1.95 trillion in 2017 (World Bank, 2019a). UNEMPLOYMENT EFFECTS OF GREENFIELD AND BROWNFIELD INVESTMENTS IN POST-TRANSITION EUROPEAN UNION MEMBERS Yilmaz Bayar1, Rita Remeikienė2, Jan Žufan3, Miloslav Novotný4 1 Usak University, Faculty of Economics and Administrative Sciences, Department of Economics, Turkey, ORCID: 0000-0002-6776-6524, [email protected]; 2 Mykolas Romeris University, Public Security Academy, Lithuania, ORCID: 0000-0002-3369-485X, [email protected]; 3 University College of Business in Prague, Czech Republic, ORCID: 0000-0001-9031-5791, [email protected]; 4 Brno University of Technology, Faculty of Civil Engineering, Institute of Building Structures, Czech Republic, ORCID: 0000-0001-8248-1466, novotny[email protected]. Abstract: International direct and portfolio investments have gone up considerably as of mid-1980s. The foreign direct investments with characteristic of long term horizon may affect the economic variables through know-how and technology transfer, physical capital expansion, and new job creation. However, foreign direct investments may have potential to negatively affect the domestic competitors with insuffi cient competitiveness in the industry. So, the economic effects of FDI infl ows have been one of the much-debated and studied issues in the international economics. This study investigates the unemployment effects of greenfi eld and brownfi eld investments in 11 posttransition EU members over 2003–2017 period through panel cointegration and causality tests. The article fi lls the gap in the literature, because the relevant empirical literature has generally researched the impact of total FDI fl ows on the unemployment/employment. The empirical fi ndings revealed that brownfi eld investments raised the unemployment in overall panel in the long run, but greenfi eld investments had no signifi cant impacts on the unemployment in overall panel in the long run. However, greenfi eld investments decreased the unemployment in Croatia, Hungary, and Slovenia, and raised the unemployment in Poland and Slovakia, while brownfi eld investments raised the unemployment only in Czechia. Consequently, it is not very reasonable to compare our fi ndings with the results of other studies using total FDI infl ows as the independent variable. But, it is generally consistent with theoretical and empirical expectations. Keywords: Greenfi eld investments, brownfi eld investments, unemployment, panel cointegration, causality analyses. JEL Classifi cation: C33, E24, F21, F23. APA Style Citation: Bayar, Y., Remeikienė, R., Žufan, J., & Novotný, M. (2020). Unemployment Effects of Greenfi eld and Brownfi eld Investments in Post-transition European Union Members. E&M Economics and Management, 23(2), 4–16. https://doi.org/10.15240/tul/001/2020-2-001 EM_2_2020.indd 4EM_2_2020.indd 4 1.6.2020 16:39:141.6.2020 16:39:14 5 2, XXIII, 2020 Economics The rapidly expanding FDI fl ows made the economic effects of FDI one of the muchdiscussed and studied topics in the international economics. On the one hand, the scholars have focused on the effect of FDI infl ows on the economic growth, unemployment, total taxes, technological development, environmental degradation (see, e.g., Lasbrey et al., 2018). On the other side, the determinants of FDI attraction have been investigated evenly considering its positive economic effects (see, e.g., Tocar, 2018). In this study, we researched the unemployment effects of greenfi eld and brownfi eld investments to fi ll the gap in the relevant literature, because the relevant empirical literature has generally researched the impact of total FDI fl ows on the unemployment/employment. But however, FDI can be implemented in two different ways such as greenfi eld investment or brownfi eld investment. The greenfi eld investment is the FDI type in which direct investors make a new plant, distribution plant or shop in the host country. But the direct investors make investment in or take over an operating company in case of brownfi eld investment (Galeza & Chan, 2015). Therefore, the interaction mechanisms among greenfi eld investment, brownfi eld investment, and unemployment may differ depending on the FDI type. The effect of FDI infl ows on the unemployment depends on job creating capacity of FDI infl ows. In this regard, greenfi eld investments are theoretically expected to decrease the unemployment, because the greenfi eld investment includes building the new plants, distribution plants and facilities and in turn creating new jobs. However, greenfi eld investments also can raise the unemployment in case the similar domestic fi rms cannot compete with the foreign competitors in terms of technological level, know-how, and production scale and terminate their activities. However, the greenfi eld investments are quite likely expected to decrease the unemployment. On the other side the effect of brownfi eld investments, in other words mergers and acquisitions on the unemployment depend on the direct investor’s behavior. Hence, the direct investor can make a contribution to the fi rm enlargement and in turn create new jobs. On the contrary, the direct investors can raise the unemployment by technology and know-how transfer, and productivity improvements. Consequently, net effect of brownfi eld investment on the unemployment is highly unclear. The EU transition economies went through a process of institutional and economic transformation as of late 1980s and then respectively integrated with global economy and the EU (European Union). The FDI infl ows to the EU transition economies raised considerably during 2003–2007 period especially together with EU membership/membership negotiations and reached about USD 153.4 billion in 2007, but then experienced signifi cant contractions due to the recent economic crises and was about USD 27.4 billion in 2017 (World Bank, 2019a). The main aim of the article is to investigate the unemployment effects of greenfi eld and brownfi eld investments in EU transition economies. The article contributes to the relevant literature twofold. First, the nearly all the empirical studies examining the economic effects of FDI infl ows have generally used the variable of total FDI infl ows in the analyses. So our paper will be one of the early studies investigating the unemployment effects of two main types of FDI infl ows separately, in other words greenfi eld investment and brownfi eld investments. Secondly, the empirical studies generally have employed regression analysis, cointegration and causality analyses disregarding structural breaks, although FDI infl ows and the main macroeconomic variables have been infl uenced seriously by the crisis periods. The Westerlund and Edgerton (2008) cointegration test used in the empirical analysis regards not only the structural break in the study period, but also cross-sectional dependence, and heterogeneity. The remainder of the paper is structured as follows. The next section provides a brief conceptual background of the subject with the literature review. Data and methods used in examining the unemployment effects of greenfi eld and brownfi eld investments are described in Section 2. Section 3 provides and discusses the results, while last part of the paper concludes with direction for further research. 1. Literature Review The signifi cant increases in international FDI fl ows have led the researchers to investigate the economic effects of FDI fl ows such as FDI impact on economic growth, unemployment, EM_2_2020.indd 5EM_2_2020.indd 5 1.6.2020 16:39:141.6.2020 16:39:14 62020, XXIII, 2 Economics technological development, competitiveness, fi nancial sector development, tax revenues, and environmental degradation. In this research, we focused on the unemployment effects of FDI infl ows considering the gap in the relevant empirical literature. The relevant empirical literature was summarized in Tab. 1 regarding the extensive number of studies about FDIemployment/unemployment nexus. The relevant empirical literature revealed that the impact of FDI infl ows on the unemployment has stayed inconclusive in keeping with theoretical considerations. In the relevant empirical literature Çiftçioğlu et al. (2007), Balcerzak and Zurek (2011), Carp (2012), Strat et al. (2015), Zdravković et al. (2017), Dritsakis and Stamatiou (2018) and Ali et al. (2018) researched the nexus FDIunemployment for the country/countries from our sample, and the studies also reached mixed fi ndings. For example, Çiftçioğlu et al. (2007) revealed that FDI raised the unemployment, while Balcerzak and Zurek (2011), Carp (2012) and Dritsakis and Stamatiou (2018) discovered that FDI decreased the unemployment, but Zdravković et al. (2017) revealed no signifi cant effects of FDI on the unemployment. The links between unemployment and other macroeconomic indicators has also been studied by Sasongo and Huruta (2019). Furthermore, nearly all the empirical studies investigating the economic effects of FDI infl ows have used total FDI infl ows in the econometric analyses without making any separation between greenfi eld and brownfi eld investments, although both FDI types may have different economic implications depending on the industry attracting the FDI fl ows. The study researches the effect of greenfi eld and brownfi eld investments on the unemployment regarding the gap in the literature and the relevant theoretical considerations. Study Country/Country group and period Method Impact of FDI on unemployment Seyf (2000) France, Germany, Spain, UK, 1994 Regression No signifi cant effects Chang (2005) Taiwan, 1981–2003 VAR analysis No signifi cant effects Craigwell (2006) 20 Caribbean countries, 1990–2000 Panel data analysis Negative Çiftçioğlu et al. (2007) 9 Central and Eastern European countries, 1995–2003 Regression analysis Positive Jayaraman and Singh (2007) Fiji, 1970–2003 Cointegration and causality analyses Negative Lin and Wang (2008) 52 industrialized/ developing countries, 2000–2004 Regression analysis Negative in developing countries; No signifi cant effects in industrialized countries Rizvi and Nishat (2009) Pakistan, India and China, 1985–2008 Panel data analysis No signifi cant impact Subramaniam (2009) Malaysia, 1975–2004 Cointegration analysis No signifi cant effects Aktar et al. (2009) Turkey, 2001–2007 VAR analysis No signifi cant effects Karlsson et al. (2009) China, 1998–2004 Time series analysis Negative Pinn et al. (2011) Malaysia,1970–2007 ARDL cointegration and causality tests One-way causality from FDI to employment Palát (2011) Japan, 1983–2009 Regression analysis Negative Tab. 1: Literature summary – Part 1 EM_2_2020.indd 6EM_2_2020.indd 6 1.6.2020 16:39:151.6.2020 16:39:15 7 2, XXIII, 2020 Economics Study Country/Country group and period Method Impact of FDI on unemployment Balcerzak and Zurek (2011) Poland, 1995–2009 VAR analysis Negative Carp (2012) Romania, 1991–2010 Regression Negative Yayli and Deger (2012) 27 developing countries, 1991–2008 Causality analysis One-way causality from FDI to employment Shaari et al. (2012) Malaysia, 1980–2010 Regression Negative Mehra (2013) India, 1970–2007 Regression Negative Mucuk and Demirsel (2013) 7 develoing countries, 1981–2009 Panel data analysis Positive in Turkey and Argentina; Negative in Thailand; No signifi cant effects in Colombia, Chile, Philippines, and Uruguay Habib and Sarwar (2013) Pakistan,1970–2011 Cointegration analysis Negative Zeb et al. (2014) Pakistan, 1995–2011 Regression Negative Jaouadi (2014) Saudi Arabia,1991–2012 Cointegration analysis Positive Bayar (2014) Turkey, 2000–2013 ARDL cointegration Positive Schmerer (2014) 19 OECD countries, 1980–2003 Regression analysis Negative Stamatiou and Dritsakis (2014) Greece, 1970–2012 ARDL cointegration No signifi cant effects Kurtovic et al. (2015) 6 Western Balkan countries, 1998–2012 Pedroni and Fisher-Johansson cointegration tests and Granger causality test Negative Djambaska and Lozanoska (2015) Macedonia, 1999–2013 Regression No signifi cant effects Strat et al. (2015) 13 last EU member countries, 1991–2012 Causality analysis One-way causality from FDI to unemployment in 4 countries; One-way causality from unemployment to FDI in 3 countries; No signifi cant causality in 6 countries Haddad (2016) Jordan, 1998–2015 Regression analysis Negative Irpan et al. (2016) Malaysia,1980–2012 ARDL cointegration Negative Chella and Phiri (2017) South Africa, 1970–2014 ARDL cointegration No signifi cant effects Nikoloski (2017) Macedonia, 2009–2015 Regression analysis Negative Tab. 1: Literature summary – Part 2 EM_2_2020.indd 7EM_2_2020.indd 7 1.6.2020 16:39:151.6.2020 16:39:15 82020, XXIII, 2 Economics 2. Aim, Data and Econometric Methodology The article’s empirical aim is to investigate the unemployment effects of greenfi eld and brownfi eld investments in 11 EU transition economies during the 2003–2017 period by panel cointegration and causality analyses. 2.1 Main Statistical Variables The dependent variable of unemployment as percent of total labor force was extracted from World Bank (2019a). On the other side, greenfi eld and brownfi eld investments were provided from the database of UNCTAD and included in the model as a percent of GDP. All the variables were annual. The presence of greenfi eld and brownfi eld investments’ data led us to determine the study period as 2003–2017. The sample of the econometric analysis consisted of Bulgaria, Croatia, Czechia, Estonia, Hungary, Latvia, Lithuania, Poland, Romania, Slovakia and Slovenia. The econometric analyses were implemented through the software of Stata 14.0 and Gauss 10.0. The main characteristics of the dataset were shown in Tab. 3. The average employment was about 9.5% in the sample, but varied considerably from country to the country. The average greenfi eld investment was about 4.4% of GDP in the sample and the average brownfi eld investment was about 0.55% of GDP in the sample. However, all greenfi eld and brownfi eld investments also changed signifi cantly among the countries. Study Country/Country group and period Method Impact of FDI on unemployment Bayar and Şaşmaz (2017) 21 emerging economies, 1994–2014 Cointegration analysis Positive Yildirim and Yildirim (2017) Turkey, 2005–2016 VAR analysis Negative Zdravković et. al (2017) 17 transition economies, 2000–2014 Panel cointegration No signifi cant effects Onanuga and Onanuga (2018) 23 emerging economies, 1991–2016 Regression analysis Negative Mohamed (2018) Sudan, 1990–2016 VAR and causality analyses No signifi cant effects Dritsakis and Stamatiou (2018) 15 EU members, 1970–2015 Causality analysis Negative Nguyen (2019) 5 central Asian countries, 1997–2016 Cointegration analysis Negative Source: own based on the literature review Tab. 1: Literature summary – Part 3 Variables Description Source UNEMP Unemployment, total (% of total labor force) World Bank (2019a) GFDI Greenfi eld investments (% of GDP) UNCTAD (2019) BFDI Brownfi eld investments (% of GDP) UNCTAD (2019) Source: own based on the literature review Tab. 2: Data description EM_2_2020.indd 8EM_2_2020.indd 8 1.6.2020 16:39:151.6.2020 16:39:15 9 2, XXIII, 2020 Economics 2.2 Econometric Methodology Westerlund and Edgerton (2008) cointegration test was derived from the unit root test of Schmidt and Phillips (1992), Ahn (1993), and Amsler and Lee (1995) rests on Lagrange Multiplier (LM). The test takes notice of not only cross-sectional dependence and structural break, but also heteroscedasticity and serial correlation. The cointegration Formula is expressed as following: (1) (2) where i = 1, 2,…, N indicates the crosssections and t = 1, 2,…, T indicates the time dimension of the dataset. Di,t dummy variable in (1) numbered Formula is defi ned in (3) numbered Formula. Furthermore, αi and βi respectively denotes the constant and slope coeffi cients before the structural breaks, and γi and denotes the ones after structural breaks. Finally, wi,t represent the error term. (3) zi,t error term is derived the following Formulas allowing the cross-sectional dependence. Ft and Fj,t shows the common vector with k dimesions, λi represents the the compatible vector of factor loadings. Ft is stationary under the assumption of pj < 1 for all the js. Thereby (1) numbered Formula is cointegrated under the condition of Øi < 0. (4) (5) (6) is calculated as the following in case of cross-sectional dependence: (7) (8) Lastly, the standardized test statistics of Westerlund and Edgerton (2008) cointegration test are calculated as in (9–10) Formulas. is the OLS estimation of in (8) numbered Formula and is the estimated standard error. Further, is the estimated long term variance of is the estimated standard error of . The refusal of the null hypothesis suggesting the cointegration relationship among the variables showed the existence of the cointegrating relationship among the variables. (9) (10) (11) (12) The slope coeffi cients of the cointegration Formula was estimated by AMG (augmented mean group) estimator of Eberhardt and Bond (2009). The AMG estimator calculates both cross-sectional coeffi cients and panel coeffi cients and also provides more reliable results than CCE (Common Corelated Effects) estimator of Pesaran (2006) does, because it estimates the panel cointegration coeffi cient Variables Mean Std. deviation Minimum Maximum UNEMP 9.578691 3.706703 2.89 19.482 GFDI 4.390562 5.323532 0.0189589 45.1763 BFDI 0.5580028 1.206754 -0.8189105 8.607166 Source: own Tab. 3: Main characteristics of the dataset EM_2_2020.indd 9EM_2_2020.indd 9 1.6.2020 16:39:151.6.2020 16:39:15 10 2020, XXIII, 2 Economics by weighting the arithmetic average of the cross-sectional cointegration coeffi cients. Furthermore, the AMG estimator takes notice of the dynamic effects and common factors in the series and also yields the effi cient results for the unbalanced panels. The estimator also can be used in case endogeneity problem (see Eberhardt & Bond, 2009; Eberhardt & Teali, 2011 for detailed information about the estimator). Lastly, the causal interaction among greenfi eld investment, brownfi eld investment, and unemployment is tested with Dumitrescu and Hurlin (2012) causality test, a modifi ed version of traditional Granger causality test for heterogeneous models and also yields robust results in case of cross-sectional dependence. 3. Empirical Analysis In the applied section of the paper, fi rst crosssectional dependence was tested with LM CD test of Pesaran (2004) taking notice of dataset’s time and cross-section dimensions and the test consequences were shown in Tab. 4. The null hypothesis suggesting the cross-sectional independence was denied at 1% signifi cance level. So the tests pointed out the presence of cross-sectional dependence among three series. Secondly, slope coeffi cients’ homogeneity was tested with adjusted delta tilde test of Pesaran and Yamagata (2008) and test consequences were shown in Tab. 5. The null hypothesis suggesting the existence of Test Test statistic P-value LM (Breusch & Pagan, 1980) 281.7 0.0000 LM adj. (Pesaran et al., 2008) 41.66 0.0000 LM CD 15.82 0.0000 Source: own based on cross-sectional dependence tests Variables Constant Constant + trend UNEMP -1.316 (0.094)* 0.070 (0.528) D(UNEMP) -1.262 (0.004)*** -0.308 (0.079)* GFDI -2.438 (0.007)*** -0.042 (0.483) D(FDI) -4.093 (0.000)*** -3.144 (0.001)*** BFDI -1.173 (0.120) -0.520 (0.302) D(FDI) -3.339 (0.000)*** -1.554 (0.050)** Source: own based on panel unit root test Note: Optimum lag length was specifi ed as 1 taking notice of Schwarz and Hannan-Quinn information criterion. ***, **, * indicated that it is respectively signifi cant at 1%, 5%, and 10%. Tests Test statistic P-value Δ ῀-1.391 0.918 Δ ῀ adj. -1.606 0.946 Source: own based on homogeneity tests Tab. 4: Results of cross-sectional dependency tests Tab. 6: Results of CIPS unit root test Tab. 5: Results of homogeneity tests EM_2_2020.indd 10EM_2_2020.indd 10 1.6.2020 16:39:161.6.2020 16:39:16 11 2, XXIII, 2020 Economics homogeneity was accepted in the light of p-values of both tests. So the slope coeffi cients of the cointegration Formula were homogenous. The presence of unit root in the series was examined with Pesaran (2007) CIPS (Crosssectionally augmented IPS (Im, Pesaran, & Shin, 2003)) unit root test taking notice of cross-sectional dependence and the test consequences were shown in Tab. 6. The test consequences revealed that all the series were I(1). The long run unemployment effect of greenfi eld and brownfi eld investments were tested by Westerlund and Edgerton (2008) cointegration test considering the crises in the study period and the presence of cross-sectional dependence and the test consequences were shown in Tab. 7. The null hypothesis suggesting no cointegration relationship among the series was rejected in all the three models. So a signifi cant cointegration relationship among greenfi eld investment, brownfi eld investment, and unemployment in the light of test results. The dates of structural breaks determined endogenously by the test revealed the recent crises. The cointegration coeffi cients was estimated by AMG estimator taking notice of cross-sectional dependence and the test consequences were shown in Tab. 8. The test consequences revealed that brownfi eld investments positively affected the unemployment in overall panel, but greenfi eld investments had no signifi cant effects on the unemployment in overall panel. However, the individual coeffi cients revealed that greenfi eld investments decreased the unemployment in Croatia, Hungary, and Slovenia, but raised the unemployment only in Slovakia. On the other side, brownfi eld investments raised the unemployment only in Czechia. All the EU transition countries attracted much more greenfi eld investments than brownfi eld investments during the study period. However, greenfi eld investments had a decreasing effect on the unemployment only in Croatia, Hungary, Model Zφ(N) P-value Zτ(N) P-value No shift -2.230 0.013 -2.806 0.003 Level shift 0.173 0.049 -1.816 0.035 Regime shift 1.121 0.024 0.040 0.016 Country Level shift Regime shift Bulgaria 2015 2015 Croatia 2015 2015 Czechia 2005 2014 Estonia 2004 2004 Hungary 2014 2014 Latvia 2009 2009 Lithuania 2011 2011 Poland 2011 2012 Romania 2012 2012 Slovakia 2012 2012 Slovenia 2012 2014 Source: own based on panel cointegration test Note: The information criterion by Bai and Ng (2004) was utilized in specifi cation of the common factors and determined 5 as maximum. Westerlund and Edgerton (2008) followed the method suggested by Bai and Perron (1998) for determination of structural breaks. Tab. 7: Results of cointegration test EM_2_2020.indd 11EM_2_2020.indd 11 1.6.2020 16:39:161.6.2020 16:39:16 12 2020, XXIII, 2 Economics and Slovenia, but raised the unemployment in Poland and Slovakia and no signifi cant effects in the rest of the countries. However, brownfi eld investments raised the unemployment in the panel considering the panel homogeneity. The brownfi eld investments generally have been realized as mergers and acquisitions in the EU and this process inevitably raised the unemployment. We evaluate that full labor mobility in the EU had infl uence on the effect of both greenfi eld and brownfi eld investments on the unemployment. On the other side, all the empirical studies conducted for the country/countries in our sample used total FDI infl ows and did not research the effect of greenfi eld and brownfi eld investments on the unemployment separately. In this regard, it is not very much meaningful to compare our fi ndings with the results of other studies. However, the aforementioned empirical studies also have stayed inconclusive about the employment/unemployment effect of the FDI infl ows in the countries. For example, Çiftçioğlu et al. (2007) revealed that FDI raised the unemployment, Balcerzak and Zurek (2011), Countries Coeffi cients GFDI BFDI Bulgaria 0.0069215 0.0446915 Croatia -1.47446*** -0.3603813 Czechia 4.133843 0.456585** Estonia -0.0798053 0.8744853 Hungary -0.480004* 1.168243 Latvia -0.1367851 3.289604 Lithuania 0.127508 -0.2796247 Poland 2.348943*** -0.9850148 Romania 0.0370108 0.1674633 Slovakia 0.2818848*** 1.898311 Slovenia -0.6352004** 1.061451 Panel 0.3754415 0.6668921* Source: own based on cointegration coeffi cients’ estimation Note: ***, **, * indicated that it is respectively signifi cant at 1%, 5%, and 10%. Tab. 8: Results of cointegration coeffi cients’ estimation Null Hypothesis W-Stat. Zbar-Stat. Prob. DGFDI ↛DUNEMP 2.03690 -0.59174 0.5540 DUNEMP ↛DGFDI 5.24624 1.89692 0.0578 DBFDI ↛DUNEMP 3.26906 0.36373 0.7161 DUNEMP ↛DBFDI 2.27400 -0.40788 0.6834 DBFDI ↛DGFDI 8.69262 4.56939 5.E-06 DGFDI ↛DBFDI 5.38393 2.00369 0.0451 Source: own based on causality test Tab. 9: Results of causality test EM_2_2020.indd 12EM_2_2020.indd 12 1.6.2020 16:39:161.6.2020 16:39:16