*4.1. Descriptive Statistics and Correlations*

Table 5 reveals the summary statistics for the measures selected in the empirical analysis. The mean level of FDI inflows equaled 5.39% of GDP. Figure 1 shows the temporal evolution of the FDI inflows in CEECs and provides evidence that Slovenia (2.00%) and Lithuania (3.32%) registered the lowest mean values regarding FDI inflows, whereas Bulgaria (8.91%) and Hungary (11.45%) the highest average values. With reference to poverty, we notice that the mean share of individuals at risk of poverty or social exclusion is 29.39%, whereas the average income quintile share ratio is 5.06%. Moreover, a mean value of 35.55% of individuals out of CEECs are using the Internet. Concerning the average values of the variables regarding institutional quality, we notice a poor governance in the CEECs.

The average values of Worldwide Governance Indicators for every selected country are revealed in Figure 2. With reference to control of corruption estimate, the lowest mean values are registered in Romania (−0.21) and Bulgaria (−0.15), whereas the highest mean values in Slovenia (0.90) and Estonia (1.06). Government effectiveness estimate shows the bottom mean figures in Romania (−0.23) and Bulgaria (0.12), but the uppermost mean figures in Slovenia (1.014) and Estonia (1.015). Regarding political stability and absence of violence/terrorism estimate, the lowest mean values are reported in Romania (0.20) and Bulgaria (0.24), while the highest mean values are exhibited in the Czech Republic (0.98) and Slovenia (1.03). Regulatory quality estimate reveals the lowest average numbers in Romania (0.45) and Croatia (0.47), while the highest average numbers in the Czech Republic (1.13) and Estonia (1.43). Relating to the rule of law estimate, Bulgaria (−0.08) and Romania (0.004) shows the lowest mean values, whereas Slovenia (0.99) and Estonia (1.12) point out the highest mean values. In terms of voice and accountability estimate, Romania (0.42) and Bulgaria (0.51) displays the lowest average figures, but then again Slovenia (1.05) and Estonia (1.10) the highest mean figures.


**Table 5.** Descriptive statistics (raw data).

Source: Authors' computations. Notes: For the definition of variables, please see Table 4.

**Figure 1.** The evolution of FDI in CEECs. Source: Authors' work. Notes: For the definition of variables, please see Table 4.

In line with prior studies [2,4,5,46,55,72,73,112,115,116,118,119], the correlation coefficients of the selected measures are reported in Table 6. We ascertain high level of correlation amid institutional dimensions as in previous studies [2,4,13,23,28]. Therefore, to get rid of the multicollinearity issue, the highly correlated measures will be included in separate regression models.

**Figure 2.** Worldwide Governance Indicators (mean values) in CEECs. Source: Authors' work. Notes: For the definition of variables, please see Table 4.


**Table 6.** Correlation matrix.

Source: Authors' computations. Notes: Superscripts \*\*\*, \*\*, \*, and † indicate statistical significance at 0.1%, 1%, 5%, and 10% respectively. For the definition of variables, please see Table 4.

#### *4.2. Multivariate Analysis Outcomes*

The Hausman test was accomplished for all the estimated econometric models and most of the outcomes support that panel data fixed-effects estimators are consistent than panel data random effects, as in Kayalvizhi and Thenmozhi [28]. However, Zakharov [107] contended that fixed-effects regressions eliminate time-invariant factors typical for each country and comprise time dummies to catch the common time trend.

Table 7 reports the results of estimations concerning the impact of FDI, poverty, and income inequality on growth. The estimated coefficient of FDI is negative in (1), revealing a low statistical significance, while statistically insignificant in (3). However, in (2) and (4), the coefficient of FDI is also negative, but the coefficient of FDI\_SQ is positive, therefore showing a non-linear relationship with GDP per capita. The non-linear relationship between FDI and growth is in line with prior studies [37,66–68]. Furthermore, individuals at risk of poverty or social exclusion show a negative impact on economic growth, in estimations (1) and (2), since poverty act against human capital expansion through restraining the capacity of people to go on healthy and to contribute via talented workforce. On the contrary, estimations (3) and (4) show that income quintile share ratio positively influences economic growth, similar to Fawaz, Rahnama and Valcarcel [30].


**Table 7.** The outcomes of panel data regression models towards the influence of FDI, poverty, and income inequality on economic growth.

Source: Authors' computations. Notes: Superscripts \*\*\*, \*\*, \*, and † indicate statistical significance at 0.1%, 1%, 5%, and 10% respectively. Figures in brackets show t-statistic. FE denotes fixed-effects (within) regression. RE denotes Random-effects Generalized Least Squares (GLS) regression. For the definition of variables, please see Table 4.

Table 8 presents the estimation outcomes regarding the influence of FDI, education, and innovation on economic growth. The estimation results display the lack of any statistically significant relationship between FDI and growth in all the estimated models. On the other hand, the coefficients of pupils and students, in estimations (1) and (2), alongside the number of patent applications, in estimations (3) and (4), are positive and statistically significant. In fact, educational attainment would catch the absorptive capacity of the economy [46]. This implies that enhanced human capital formation, alongside innovation boost GDP per capita.

The outcomes of panel data regression models towards the impact of FDI, transport infrastructure, and information technology on economic growth are illustrated in Table 9. Analogous to the outcomes from Table 8, there is not found any statistically significant association between FDI and GDP per capita. As well, the coefficients of transport infrastructure in (1) and (2) and information technology in (3) and (4) shows a positive influence on economic growth. Nevertheless, better infrastructure rises the productivity of investment and thus spurs FDI flows [29].


**Table 8.** The outcomes of panel data regression models towards the influence of FDI, education, and innovation on economic growth.

Source: Authors' computations. Notes: Superscripts \*\*\*, \*\*, \*, and † indicate statistical significance at 0.1%, 1%, 5%, and 10% respectively. Figures in brackets show t-statistic. FE denotes fixed-effects (within) regression. RE denotes Random-effects Generalized Least Squares (GLS) regression. For the definition of variables, please see Table 4.


**Table 9.** The outcomes of panel data regression models towards the influence of FDI, transport infrastructure, and information technology on economic growth.


**Table 9.** *Cont*.

Source: Authors' computations. Notes: Superscripts \*\*\*, \*\*, \*, and † indicate statistical significance at 0.1%, 1%, 5%, and 10% respectively. Figures in brackets show t-statistic. FE denotes fixed-effects (within) regression. RE denotes Random-effects Generalized Least Squares (GLS) regression. For the definition of variables, please see Table 4.

Tables 10 and 11 show the estimation outcomes regarding the impact of FDI and institutional quality on economic growth. Similar to Table 7, the results of panel data fixed-effects regression models provide evidence for a non-linear link with economic growth. Furthermore, all the Worldwide Governance Indicators positively influence growth, except political stability and absence of violence/terrorism which is not statistically significant. In case of control of corruption, the "helping hand" viewpoint [18–20] is confirmed. Similar to Kato and Sato [108], the "grease the wheels" hypothesis is established. As well, consistent with Iamsiraroj [46], quality of institutions exerts a critical role in augmenting economic growth directly and through FDI inflows.

**Table 10.** The outcomes of panel data regression models towards the influence of FDI, control of corruption, government effectiveness, political stability and absence of violence/terrorism on economic growth.


Source: Authors' computations. Notes: Superscripts \*\*\*, \*\*, \*, and † indicate statistical significance at 0.1%, 1%, 5%, and 10% respectively. Figures in brackets show t-statistic. FE denotes fixed-effects (within) regression. RE denotes Random-effects Generalized Least Squares (GLS) regression. For the definition of variables, please see Table 4.

With reference to country-level control measures, general government final consumption expenditure negatively influences economic growth. Iamsiraroj [46] claimed that FDI may relocate domestic saving, whereas Awad and Ragab [42] reinforced for developing states where inland sources of capital are inadequate that private-sector investment may be "crowded out" by higher levels of government spending. Urban population negatively influences economic growth, similar to some extent to Alvarado, Iñiguez and Ponce [69], but contrary to Henderson [33] which claimed that urbanization accelerates growth. Domestic credit to private sector exerts a positive influence on growth similar to Feeny, Iamsiraroj and McGillivray [41]. Trade shows a positive effect on economic growth since openness empowers a more efficient manufacture of goods and services through moving production to nations that have an inclusive benefit [46]. As such, more openness to trade determines a higher FDI influxes, thus showing a positive influence on growth.


**Table 11.** The outcomes of panel data regression models towards the influence of FDI, regulatory quality, rule of law, voice, and accountability on economic growth.

Source: Authors' computations. Notes: Superscripts \*\*\*, \*\*, \*, and † indicate statistical significance at 0.1%, 1%, 5%, and 10% respectively. Figures in brackets show t-statistic. FE denotes fixed-effects (within) regression. RE denotes Random-effects Generalized Least Squares (GLS) regression. For the definition of variables, please see Table 4.

#### *4.3. Causality Analysis*

Tables 12 and 13 summarize the panel unit root tests, both for the variables in level and first difference, for types concerning individual intercept, as well as individual intercept and trend, to ensure more robustness. Since all the variables are highly statistically significant at first difference, we notice that all measures are integrated of order one I (1). Thus, we might expect there is a long-run connection between these variables together [109].

Table 14 summarizes the results of panel co-integration examination among the variables using the Pedroni statistics. For the first model, one out of seven Pedroni tests rejects the null hypothesis of no co-integration by means of both the panel and group forms of the Phillips–Perron and ADF tests. Onward, four statistics confirm the co-integration for the second model, three statistics for the third model, two statistics for the fourth model and the sixth models, while six statistics for the fifth model.



**Table**

, *11*, 5421

*Sustainability* **2019**

PP are computed using an asymptotic Chi-square distribution.

variables, please see Table 4.

 Probabilities

 for the LLC, Breitung, and IPS tests are computed assuming asymptotic normality. For the definition of



respectively. LLC reveals Levin, Lin and Chu t\* stat. Breitung reveals Breitung t-stat. IPS reveals Im, Pesaran, and Shin W-stat. ADF reveals Augmented Dickey–Fuller Fisher Chi-square.PP reveals Phillips–Perron Fisher Chi-square. LLC and Breitung assumes common unit root process. IPS, ADF, and PP assumes individual unit root process. Probabilities for ADF andPP are computed using an asymptotic Chi-square distribution. Probabilities for the LLC, Breitung, and IPS tests are computed assuming asymptotic normality. For the definitionvariables, please see Table 4.

 of


The outcome of Pedroni (Engle Granger-based) panel co-integration

 test.

**Table 14.**



248

length. For the definition of variables, please see Table 4.

Furthermore, to confirm the strength of the outcomes, we apply the Kao test for panel co-integration which is a residuals-based test. Table 15 displays the results of the Kao panel co-integration test. It is obvious that ADF (t-Statistic) is significant, except the sixth model. As such, null hypothesis, specifically there is no co-integration is rejected [109]. Thus, there is a long association between selected measures.



Source: Authors' computations. Notes: Superscripts \*\*\*, \*\*, \* and † indicate statistical significance at 0.1%, 1%, 5% and 10% respectively. Schwarz Info Criterion was selected for lag length. For the definition of variables, please see Table 4.

Table 16 reports the results of the Johansen panel co-integration test. For the Johansen panel co-integration test, the suppositions of co-integration tests permit for individual effects, but no individual linear trends in vector autoregression [73]. The null hypothesis of no co-integration is rejected, thus being confirmed the incidence of a long-run association between the variables.


**Table 16.** The outcome of Fisher (combined Johansen) panel co-integration test.

Source: Authors' computations. Notes: Superscripts \*\*\*, \*\*, \* and † indicate statistical significance at 0.1%, 1%, 5% and 10% respectively. Probabilities are computed using asymptotic Chi-square distribution. For the definition of variables, please see Table 4.

In as much as the variables are cointegrated, the long-run associations are further estimated via FMOLS and DOLS. The outcomes of the FMOLS and DOLS estimated co-integration connections are displayed in Table 17. The FMOLS outcomes of estimations (1), (3), (4) and (6) provide support for an undesirable influence of FDI on growth, contrary to Pegkas [1], Freckleton, Wright and Craigwell [39]. As well, the DOLS results confirm the negative impact merely in the first estimation. With reference to the institutional quality variables, the outcomes are similar to those reported in Tables 10 and 11, but voice and accountability estimate loses its statistical significance.


**Table 17.** Panel data examination of long-run output elasticities.

Source: Authors' computations. Notes: Superscripts \*\*\*, \*\*, \* and † indicate statistical significance at 0.1%, 1%, 5% and 10% respectively. Panel method: Pooled estimation. Heterogeneous variances. Schwarz lag and lead method in case of DOLS estimation. Figures in brackets show t-statistic. For the definition of variables, please see Table 4.

The outcomes of PVECM Granger causalities are reported in Table 18. Therefore, the following causal relationships are identified for each model:




Source: Authors' computations. Notes: Superscripts \*\*\*, \*\*, \* and † indicate statistical significance at 0.1%, 1%, 5% and 10% respectively. ECT reveals the coefficient of the error-correction term. The number of appropriate lags is two according to Schwarz information criterion. For the definition of variables, please see Table 4.

The significance of the ECT is valuable for discussing long-run causality. Therefore, in the long term, we acknowledge a bidirectional causal relationship between FDI and economic growth in all the PVECM models similar to Chan, Hou, Li and Mountain [3], Mahmoodi and Mahmoodi [70], as well as a one-way causal relationship running from each institutional quality indicator to economic growth and FDI.

#### **5. Conclusions and Policy Implications**

Foreign direct investment, a crucial factor of globalization, is regarded a noteworthy engine of productivity heightening, technical progress, and job creation. However, reduced institutional quality may discourage FDI inflows. Using panel data for 11 Central and Eastern European countries from 2003 to 2016, current paper examined at first glance the impact of FDI on economic growth, also covering institutional quality measures, as well as several sustainable development goals established in the 2030 Agenda for Sustainable Development [48]. The quantitative outcomes by means of panel data regression models lead us to conclude that a non-linear association occurred between FDI and GDP per capita. In terms of institutional quality, the whole Worldwide Governance Indicators, except political stability and absence of violence/terrorism, exhibited a positive influence on economic growth. Moreover, poverty, income distribution, education, innovation, transport infrastructure and, information technology appeared as significant drivers of growth. The estimations of FMOLS and DOLS confirmed the findings to a partial extent. Furthermore, the panel vector error-correction model Granger causalities provided evidence for a short-run unidirectional causal link running from FDI to growth and a long-run two-way causal connection between FDI and economic growth. Therefore, in

the long run, unidirectional causal associations running from each institutional quality indicator to economic growth and FDI were established.

The outcomes may have some policy implications. Therefore, the policymakers out of the CEECs should attempt to attract higher FDI inflows to boost economic growth. Overall, for the development of an appropriate setting to attract FDI, favorable economic, political, social and cultural conditions are more than required [7], alongside macroeconomic steadiness and the decrease of the market deformations [1]. For instance, the labor force in such countries is quite trained, but modern management and worldwide manufacturing knowledge is regularly missing. Therefore, investors with the required capitals, managerial skills, marketing chain, and industry strength are crucial [35]. First, there should be shaped an economic setting that ensure proper rewards to corporations which would bring ideas from overseas and put them to use with inland funds [81]. As such, financial stimulus in form of labor subventions, energy and property rental discounts, allowance on customs duties for main machinery/raw resources, export credit conveniences or tariff shield are required [22]. Secondly, a sound institutional setting should be implemented. The "helping hand" viewpoint may be explained via the early stages of democracy in the explored region. As such, in line with Harms and Ursprung [101], Mengistu and Adhikary [100] and Sabir, Rafique and Abbas [102], the democracy in the CEECs should be deepened in order to spur the inward flow of FDI [18]. Good governance may create suitable environments for investors, respectively lower cost of doing business, lesser uncertainty and upper productivity expectation [32]. A suitable stage of institutional progress can support synergies between FDI and resident companies, thus enhancing productivity spillovers. Also, capital accumulation may be increased by means of complementarities between foreign and internal investment [38]. Moreover, in line with Jude and Levieuge [38], a better institutional quality may entice high-quality technology. Thirdly, the minority shareholders and creditor rights should be enhanced via capital market directives and banking sector reorganizations, aiming to rise equity or bond release. Thus, a better extension of domestic credit to enterprises should be allowed in the benefit of wealth creation [35]. Similarly, strengthening of trade will also help in appealing FDI. Not least, the policies concerning sustainable development goals should be promoted. In this vein, greenfield and brownfield investments may lessen unemployment, while increasing incomes, public and private consumption, capital accumulation and productivity [89]. As well, infrastructure should be improved to guarantee that FDI which is acquired manifest an increased gain to the entire economy. Accordingly, improved transport systems and modern information technology ease the spread of novel products and technologies [47].

This paper has contributed to the prevailing literature by adding evidence for the case of CEECs on the nexus between FDI and economic growth. Future research can be done to explore the drivers of FDI inflows in CEECs.

**Author Contributions:** Conceptualization, ¸S.C.G., L.N.S. and O.S.H.; Data curation, ¸S.C.G., L.N.S. and O.S.H.; Formal analysis, ¸S.C.G., L.N.S. and O.S.H.; Funding acquisition, ¸S.C.G., L.N.S. and O.S.H.; Investigation, ¸S.C.G., L.N.S. and O.S.H.; Methodology, ¸S.C.G., L.N.S. and O.S.H.; Project administration, ¸S.C.G., L.N.S. and O.S.H.; Resources, ¸S.C.G., L.N.S. and O.S.H.; Software, ¸S.C.G., L.N.S. and O.S.H.; Supervision, ¸S.C.G., L.N.S. and O.S.H.; Validation, ¸S.C.G., L.N.S. and O.S.H.; Visualization, ¸S.C.G., L.N.S. and O.S.H.; Writing—Original draft, ¸S.C.G., L.N.S. and O.S.H.; Writing—Review and editing, ¸S.C.G., L.N.S. and O.S.H.

**Funding:** This research received no external funding.

**Conflicts of Interest:** The authors declare no conflict of interest.
