Economic freedom, foreign direct investment, and economic growth: The role of sub-components of freedom

Abstract

This study investigates the causality relationships among the economic freedom, foreign direct investment (FDI), and economic growth for top FDI attracting countries during 1995–2019. Apart from the previous studies, we examine these three sets of causal links simultaneously and use the panel Granger causality test of Kόnya [2006. “Exports and Growth: Granger Causality Analysis on OECD Countries with a Panel Data Approach.” Economic Modelling 23: 978–992], which considers heterogeneity and cross-sectional dependency across panel members. The findings provide weak evidence for the causal links between economic freedom, FDI, and economic growth for the overall score of economic freedom index. We also conduct causality tests for freedom vs. FDI, freedom vs. growth, and FDI vs. growth by using sub-components of the freedom index and reveal too many causality linkages among these variables. Thereby, we conclude that the direction of causality seems to be country and economic freedom indicator specific. These results have important implications for policymakers.

KEYWORDS:

• Foreign direct investment
• economic freedom
• economic growth
• causality
• panel data

JEL Classifications:

• F21
• O40
• C33

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1 Hereafter, we abbreviate foreign direct investment, economic freedom and economic growth as FDI, EF and EG, respectively.

2 Islam (1996) finds that the significance of EF indicator in growth models varies across different income groups. Azman-Saini, Baharumshah, and Law (2010) show that the impact of FDI on EG also changes according to the level of EF of the countries.

3 In our sample, FDI inflows (as a percentage of GDP) data consists of some negative values. Among the countries listed in Table 1, the negative FDI values in 18 countries are very close to zero and are only in a single observation. To be able to retain these negative values and include as many observations as possible, we follow Berger et al. (2010) and Busse, Königer, and Nunnenkamp (2010) and use the following logarithmic transformation: $\begin{array}{r}\text{y}=\text{ln}(\text{x}+\sqrt{\left({\text{x}}^2+1\right)}\end{array}$ By using this transformation, whereas the sign of x is unchanged, the values of x pass from a linear scale at small absolute values to a logarithmic scale at large values, but the sign of x is unchanged. Therefore, we choose to exclude Belgium from the analysis, since 25% of its observations included negative values and the Netherlands, which has a very high negative outlier value.

4 Considering that the sub-components of the EF index represent many potential macroeconomic determinants of FDI and EG, no additional variables are added to the estimated models.

5 In order to save space, we refer the study of Pesaran and Yamagata (2008) for the details of bias adjusted delta test.

6 To determine the optimal lag structure, we follow Kόnya (2006) and make the maximal lags to differ across variables, but to remain same across equations. We estimate the system by using 1 to 4 lags and then choose the combinations which minimize the Schwarz Bayesian information criterion. The results of the lag selection method are available upon request from the authors.

7 See Appendix for the detailed results for panel bootstrap causality tests of EF-FDI nexus. The other detailed results for EF-EG and FDI-EG linkages are available upon request.

8 One remark as pointed out by an anonymous referee is that some countries have experienced economic shocks during the sample period (e.g. 1998 Russian Crisis, 2008 Global Financial Crisis) or membership in Regional Trade Agreements such as South Asian Free Trade Area (SAFTA) in the case of India. Taking into account these structural shifts may also play a role on determining the causal linkages. However, in this study we preferred to focus on the heterogeneity and cross-sectional dependency across countries to understand the multilateral causality link among the variables.

9 We thank an anonymous referee for pointing out this issue.