Model uncertainty and variable selection: an application to the modelization of fdi determinants in europe
- Autores: Laura Montolio Breva
- Directores de la Tesis: Cecilio Tamarit Escalona (dir. tes.), María Amparo Camarero Olivas (codir. tes.)
- Lectura: En la Universitat de València ( España ) en 2019
- Idioma: español
- Tribunal Calificador de la Tesis: Antonio Montañés Bernal (presid.), María Engracia Rochina Barrachina (secret.), Felicitas Nowak Lehmann (voc.)
- Programa de doctorado: Programa de Doctorado en Economía Internacional y Turismo por la Universitat de València (Estudi General)
- Materias:
- Texto completo no disponible (Saber más ...)
- Resumen
The last decades have seen an increasing interest in FDI, and a growing debate about its modelization in terms of the variables considered as its drivers, the model specification and the estimation methods of the FDI gravity model. The reason for this is the uncertainty surrounding both theories and empirical approaches to FDI.
This doctoral dissertation aims to contribute to the literature by investigating the driving forces of MNEs activities to and from European countries, both at the regional and national level, tackling the variable selection and model uncertainty problems faced in the modelization of FDI. We focus in the European Union because of the growth of intra-EU investments triggered by the Single Market. Particularly, we focus in the cases of Spain and Germany, as they represent one of the largest FDI recipient and investor countries within the EU, respectively.
The thesis is structured into three chapters. Chapter 2 focuses on the modelization of FDI from a regional perspective by investigating the determinants of FDI activity in Spanish regions for the period 2004–2013. We implement an exploratory factor analysis to avoid the collinearity problem that arise when considering as regressors an ad hoc set of FDI determinants put forward by the literature. Then, we estimate an extended gravity model by using the Poisson pseudo-maximum-likelihood estimator with country-origin fixed effects. Our findings revealed that FDI locational strategies in the Spanish regions are determined significantly by the Competitiveness and agglomeration factor, and to a lesser extent, by the productive capacity and economic potential factors. Thus, the empirical analysis revealed that at the regional level FDI motivations are not market-seeking but efficiency-seeking. We also find that FDI to Spanish regions is highly determined by the degree of industrial specialization and the geographical location of the regions. Finally, we confirm that FDI stock data is more suitable than flows to approximate the long-run allocation patterns of FDI.
The following chapters focus on the modelization of FDI at the national level by analyzing the long-run determinants of German outward FDI. Chapter 3 deals with the variable selection problem by adopting a Bayesian model averaging (BMA) approach. The analysis covers the period 1996-2012 and is conducted for 59 recipient countries disaggregated by country-groups to avoid the so-called aggregation bias. Our results show that determinants linked to market-seeking or HFDI are found to be relatively more important in developed countries; whereas covariates related to VFDI prevail in developing countries. Within developing countries, our findings revealed that both HFDI and VFDI strategies coexist together with institutional factors in Latin American and Asian countries. Nevertheless, determinants associated with HFDI appear to be dominant for explaining FDI into Asian countries; whereas those related to VFDI play a larger role in Latin American countries. As regards the European Union, while the majority of FDI is market-seeking in “core” countries, vertical motivations are dominant in the “periphery”. Furthermore, our findings are compatible with complex integration strategies of MNEs where vertical determinants and institutional variables are gaining prominence together with the leading role currently played by Germany in the global value chains (GVC) network.
Lastly, Chapter 4 goes more into detail and addresses the uncertainty in the econometric specification of the FDI gravity model by comparing several alternative Generalized Linear Model (GLM) estimators: Poisson Pseudo Maximum Likelihood(PPML), Gamma Pseudo Maximum Likelihood (GPML), Negative Binomial Pseudo Maximum Likelihood (NBPML) and Gaussian-GLM. We follow a model selection approach based on several goodness-of-fit statistics and graphical techniques and found that NBPML is the best performing estimator for our data set, followed by GPML. The analysis conducted provides a comprehensive empirical evidence of the determinants of German outward FDI across country-groups.
Overall, the findings of this doctoral dissertation allow us to draw some policy implications about the factors that should be emphasized to attract FDI either at the regional or national level.
