It would be remiss for me not to discuss the project I just completed for Econometrics, in which i analyzed the impact of foreign direct investment (FDI) on economic growth in Sub-Saharan Africa(SSA).
Using a panel data set primarily collated from the Penn World Tables and the UN World Investment Report, I constructed several models using other exogenous variables typical of the growth literature - education and health proxies, indicators of political stability and freedom, and so forth.
Estimation was carried out through the powers of SHAZAM as a pooled set (since the time dimension was only 5 years and the panel unbalanced due to missing data entries), and also through bootstrapping the panel, and both two and three-stage least squares.
Results revealed that the coefficient for FDI was subject to extreme variability between estimation methods, and its magnitude was determined almost wholly by outliers. Removing these from the set resulted in magnitudes of effectively zero (though still significant at 5%), as verified by bootstrapping. Such magnitudes were also witnessed for 2SLS and 3SLS, while retaining the influential observations.
As such, given the likely endogenity of investment (investment likely grows in response to growth and vice versa), the feedback relationship (and likely measurement errors when dealing with SSA) is most accurately modelled using the systems techniques.
In conclusion, foreign investment is not a major contribution towards SSA growth. What the countries do for themselves - as represented, for example, by the indicators of political stability - is much more effective than reliance on external forces when playing catchup in the standard of living game.