in columns (8) and (9). This implies that the geographical position matters. Countries with
increasing distance to the equator hold a lower stock of total liabilities. Column (10) in Table
5 includes all infrastructure variables and the other explanatory variables. 67 percent of the
cross-sectional variation in total liabilities is explained by the last regression equation. None
of the infrastructure variables is individually significant. Joint significance of the infrastructure
variables is tested for using a F-Statistic.10 The joint test for the three variables has a p-value
of 0.239. Thus, the infrastructure variables are jointly insignificant. Only trade openness keeps
its statistical significance and enters with a positive sign. The other regressors have much less
importance in the final specification.
The analysis of the relationship between the average stock of FDI liabilities and the level
of infrastructure is illustrated in Table 6.11 The sample size consists of 30 countries. The
specification for the columns (1) to (10) is identical to the one explained above. In the bivariate
analysis air-departures enter significantly although roads and international telephone circuits also
have a positive sign. A three percentage point increase in the level of air-departures as a ratio to
total population is associated with a 1 percentage point increase in the FDI stock of a country.
Overall, 48 percent of the variation in FDI is explained in the cross-section. Air-departures
remain individually significant when further controls are added and its positive point estimate
remains stable across columns (4), (7) and (10). Controlling for wealth and size in columns (4)
to (6) leaves the other infrastructure variables unchanged. GDP enters with a negative sign that
is only marginally significant in the specification for roads and international telephone circuits
in columns (5) and (6). Throughout columns (7) to (9) country size, trade openness and natural
resources have a strong negative and individually significant impact on the average stock of FDI
liabilities. This impact is stable for each of the infrastructure variables used. The implication is
that, other things being equal, less open countries have a higher stock of FDI liabilities. Countries
that are equipped with higher amounts of natural resources relative to merchandise exports have
less FDI liabilities. This emphasises, leaving other controls unchanged, that countries with a
diversified export structure are better candidates for receiving FDI. The dummy concerning
being landlocked has a negative and highly statistically significant sign throughout columns
(7) to (10). Landlocked countries receive less FDI investment, which implies that countries
with ports and harbours provide better opportunities for FDI. Column (10) shows the joint
impact of the infrastructure variables. Again, 67 percent of the variation in the cross-section is
10Since a linear restriction in a small sample is tested for the F statistic instead of the chi-square distribution
is utilised.
11 One can assume that parts of the FDI inflows in the 1990s are due to privatisation. The relationship between
the adopted privatisation schemes in many of the countries and FDI stocks and flows were analysed. However, a
significant relationship for the countries could not be found.