extraordinary position as an financial centre, this country was excluded from the estimation.
Including Singapore improves the results in fact.
3.1 Descriptive Statistics
This section discusses the sample’s descriptive statistics. Figures 1 to 3 are histograms of the
infrastructure variables used in the empirical section. Each histogram is divided into four groups.
Figure 1 shows air-departures per inhabitants at 1990. Group 1 contains the 16 countries (53
percentage of the sample size) with a level of less than 0.2 air-departures in 1990. Only two
countries have a ratio of 0.6 air-departures per capita or above. Figure 2 displays paved road
length. Whereby 16 countries are grouped in category 1, which represents paved roads per
square km below 0.05 km. Six countries belong to group 2 with a road length below 0.1 km per
square km. Only 27 percent of the countries provide a paved road length of more than 0.1. km
per square km. Figure 3 presents the ratio of international telephone circuits per inhabitant.
Group 1 contains 19 countries (67 percent of the sample size), which dispose of an international
telephone circuits ratio of less than 0.09. Group 3 and 4 consist of only three countries, which
have more than 0.17 international telephone circuits per capita. Accordingly, Figures 1 to 3
indicate considerable cross-country heterogeneity in the infrastructure variables. Table 4 reports
the summary statistics of all variables included in the regression analysis. as evidenced above,
there is a considerable standard deviation in the data on the level of infrastructure. The analysis
of the capital data reveals that the mean of all stock and flow positions is positive for all countries
in the sample.
3.2 Econometric Approach
For the estimation procedure ordinary least squares (OLS) is applied to specify the prediction
that there exists a link between the level of infrastructure and capital flows to a country.8 The
standard errors are corrected by the White procedure to adjust for the presence of heteroscedas-
ticity in the data. Overall, two concepts of external capital liabilities, namely gross liability flows
and stocks, are considered. Much of the benefits of asset trade arise from gross rather than net
positions. Results for net asset positions were also tested. Since most of the countries included
in this sample are net debtors, the relationship found is not very different to the one reported
for the gross liability data.
8An infrastructure index was also experimented with. The index was constructed out of a principal component
analysis. The first and second principal component were utilised to construct weights of the infrastructure
variables. However, the index did not significantly enter the regression analysis.