The stock position is the relevant state variable on the macroeconomic level. Flows arise to
close the gap between the actual and desired stock position. Financial flows and stocks can be
divided into FDI, portfolio equity and debt. Thus, to get a precise picture, the stock and flow
variables are split into their sub-components. All liability measures are calculated as a share of
GDP and then used as a dependent variable y-i in the regression analysis. This ratio therefore
expresses the stocks or flows relative to GDP. The variable y∣ is calculated as an average over
the period 1990-95.
For the OLS application, the level of infrastructure in the countries is captured by ^i,∣evel for
each infrastructure variable separately and jointly by the vector Xi,∣eve∣.9 The latter provides
the set of infrastructure variables explained above: air-departures, roads as well as international
telephone circuits. Air-departures and telephone circuits are expressed in per capita terms while
roads are calculated as a ratio to square km of the countries. Since the time period 1990-95 is
observed, the level of infrastructure relates to the stock of infrastructure in 1990.
In the first step, yi and ^i,∣eve∣ are regressed in a bivariate form. Subsequently, a set of
other variables Zi is added. Zi includes regressors such as natural resources, latitude and trade
openness, which could also be potential determinants for capital inflows to countries. Zi also
includes variables which control for heterogeneity between the cross section of countries. Country
differences are adjusted by using measures of country size (GDP) and wealth (GDP per capita).
These two variables are in logs. Accordingly, the basic specification of the regression analysis
can be expressed as follows:
yi = α + β¾∣θvθ∣ + γ 0Zi + Ui
(1)
In the last step we use the complete set of infrastructure variables Xi,∣eve∣ is utilised. Addi-
tionally, a set of dummy variables, di, is used in order to account for unobservable factors such
as being landlocked. As a result, the estimated model then takes the form:
yi = α + β0A^i,∣eve∣ + γ0Zi + δdi + Ui
(2)
9In previous drafts effects of other infrastructure variables, such as railway length, energy generating power,
number of telephones and telephone mainlines as well as air transport (airfreight and persons carried) were tested
for. The last two can be seen as substitutes for international telephone circuits and air-departures. The results
obtained were similar to the ones we report here. The former variables were not included due to the fact that
they did not add any further explanatory power to the model. To measure the effect of human capital on the
attraction of capital inflows a variable, which is calculated by the percentage of secondary schooling degrees in
the work force, was defined. This variable was not statistically significant.