Specification Uncertainty and Model Averaging: Do Institutions Matter for FDI?

Abstract

The role of Foreign Direct Investment (FDI) in economic development is a well-researched phenomenon. However, extant research fails to arrive at a consensus on the key drivers of FDI and the literature is particularly eclectic on the role of natural resources and institutions and how they influence FDI. We argue that this lack of consensus could be on account of model uncertainty as empirical studies often tend to be selective on the variables included in the final model on which inference is drawn. Drawing on more recent literature on model uncertainty and the determinants of FDI, we demonstrate the benefits of Bayesian Model Averaging (BMA) in reducing the impact of model uncertainty on both economic and statistical inference. We use BMA to analyze the determinants of FDI into the Middle East North Africa (MENA) region and our results indicate that neither natural resources (oil and gas) nor institutional quality seem to be important once model uncertainty is accounted for. Our fundamental point, focusing on model uncertainty, applies to the broader literature on international business and strategy and suggests that researchers may need to look beyond traditional model selection methods before drawing substantial inference from their research.