Abstract
The study aims to examine the FDI-growth nexus in EU countries for the period 1996–2018. Empirical literature has suggested ambiguous results regarding the beneficial effects of FDI on host economies. In this study panel ARDL approach is employed using the PMG estimator, as the most suitable for our dataset. DFE and MG estimators are also implemented. The results indicate strong evidence that there is a positive long-run relationship between FDI and growth for the original EU members. However, regarding the new EU members, only the extended model provides statistical evidence of positive long-run relationship between FDI and growth. The positive and statistically significant long-run relationship between FDI and economic growth for all countries underlines the need for policy makers to formulate policies to attract FDI. Furthermore, the results of the study show that policy makers of EU countries characterized by high levels of corruption should focus on anti-corruption policies, as corruption seems to hinder the positive effects of both FDI and other determinants of economic growth.
Notes
1 Data taken from https://unctadstat.unctad.org/
3 Due to data availability when the study was conducted, namely in 2021.
4 United Kingdom left Europeran Union on January 31, 2020 (https://ec.europa.eu/info/relations-united-kingdom_en)
5 The same adjustment equation was used by Tsanana, Chapsa, and Katrakilidis (Citation2016) and Tsitouras et al. (Citation2017).
6 World Bank national accounts data, and OECD National Accounts data files.
8 World Bank national accounts data, and OECD National Accounts data files.
9 All the variables included in the model are expressed in logarithms and are described in .
10 Based on the studies of Johansen (Citation1988) and Fisher (Citation1932).
11 Pesaran and Smith (Citation1995), Pesaran, Shin, and Smith (Citation1997) and Pesaran and Shin (Citation1999) introduce ARDL model in error correction form as a cointegration test. Asteriou et al. (Citation2021) indicate that the application of panel ARDL models do not require pretesting with the implementation of cointegration tests.
12 Given the size of the two samples, the model that is selected is of the form ARDL (1, 1, 1, 1, 1, 1). See also Mongo, Belaïd, and Ramdani (Citation2021); Murindahabi et al. (Citation2019); Demetriades and Law Citation2006.