3.3 Results
Following the procedure introduced above the results are discussed in the following subsections.
Regression results for the cross-section analysis on capital stocks are presented in Tables 5 to 7
whereas Table 8 exhibits capital flow data. Since the focal point of interest lies in comparing
the influence of the same set of determinants across categories, the same set of specification as
explained below for Table 5 is adopted in each case.
3.3.1 1990-95 Cross-Section Analysis of the Stock Data
For the analysis of the stock of total liabilities, 29 countries are included in the sample as South
Africa is excluded due to missing observations. Table 5 contains the regression results of the
total liability stock relative to GDP as the dependent variable. In columns (1) to (3) the basic
bivariate relationship between the average total liability stocks and the infrastructure variables
are shown. A positive effect for measures on the level of roads, air-departures as well as the level
of international telephone circuits is established. However, only air-departures and telephone
circuits enter significantly into the basic specification. Those variables are able to explain 20 and
42 percent of the cross-country variations in total liabilities respectively. A one percentage point
improvement in the level of air-departures in 1990 is associated with an increase in the stock of
total liabilities of 0.802 percentage points. The effect is even stronger for international telephone
circuits. Here, a one percentage point improvement in the level of international telephone circuits
in 1990 leads to a rise in total liabilities by 2.982 percentage points. To allow for the cross-
sectional variations in size and wealth, GDP and GDP per capita are included as general control
variables in the regressions in columns (4) to (10). The other regressors are jointly added in
columns (7) to (10). When controlling for wealth and size of the countries in columns (4) to (6),
only international telephone circuits maintain a positive and individually significant relationship
with the average total liability stock. Interestingly, throughout columns (4) to (6), total GDP
enters negatively and is statistically significant in the specification. Thus, smaller countries hold
more liabilities relative to GDP. Columns (7) to (9) include all explanatory variables for each
of the infrastructure variables. Again, only international telephone circuits reveal a significantly
positive relationship with the total liability stock. The size of the country loses significance as
the remaining explanatory variables are added. Instead, trade openness now enters significantly
and its point estimate is positive and stable across columns (7) to (9). One explanation is that
trading countries represent a better credit risk and receive more liabilities. On average, a ten
percentage point increase in trade openness is associated with an 0.15 percentage point increase
in total liabilities. Latitude enters the specification individually significantly and negatively