Determinants of Foreign Direct Investment: A Bayesian Model Averaging Approach From 55 Countries

Abstract

Purpose: This study aims to investigate the determinants of Foreign Direct Investment (FDI) for the 55 countries from 4 different continents and 3 different development levels. Method: A panel dataset is constructed over the period of 1995-2019 with 17 candidate independent variables and Bayesian Model Averaging (BMA) is employed for each continent and for each development level to reveal the determinants of FDI. Results: Results showed that globalization is the primary factor to attract FDI regardless of the development level and continent. But there are also other factors that affect FDI which differ based on the development level and geographic location of the country. FDI in transition economies benefits from qualified labor force which negatively affects developed countries and has no effect on developing countries. Corruption is only effective in developing countries. Secondary education is effective at European and Asian countries, tertiary education is effective on American countries. This study contributes the literature by using a BMA method in a panel data context. BMA uses weighted average of all the possible models which gives more balanced results than a single model. Applying BMA to the countries in different continents and development levels separately allows to make inferences based on region and wealth. Originality: This study used a fairly large dataset on terms of selected time period, number of countries and selected variables which most of the studies lack of. Also, the advantage of applying BMA to a panel dataset, is obvious as both controls the cross section variation and also considers a weighted average of all possible linear regression results rather than giving only a single linear regression model result. Thus applying BMA on a panel dataset give a better insight as to which variables should be considered as determinants of FDI.