The fundamental determinants of financial integration in the European Union

19

financial integration caused by capital controls) vis-à-vis domestic factors in explaining closed interest
differentials. If financial integration is strong (coefficient β₁ is small) domestic factors will be a more
important explanation for closed interest differentials. The regressions also allow for the presence of country-
specific risk unexplained by the right-hand side of the regression. Including the lagged dependent variable
together with the domestic determinants x_i makes it possible to explain a substantial part of the intensity
of capital controls. The generalized country-specific effect in regressions D and E is interpreted slightly
differently than in regressions A and B. The domestic determinants x_i affect the magnitude of the coefficient
β₁ of the lagged dependent variable. Coefficient β₁ arguably captures existing capital controls. Ifwe compare
regression C with regressions D and E, the magnitude of coefficient β₁ declines with the country risk premium
that is explained by the other determinants x_i in the regression such as the perceived probability of debt
default (DEF) and the risk of loss of purchasing power (INF). The generalized country-specific effect is
assumed to pick up the other political risks. We expect countries with high political risk to show lower
country-specific effects ζ_i and, consequently, a lower generalized country-specific effect. A negative country-
specific effect may indicate that more capital export restrictions are expected. The existence of political
risk (unrelated to the economic fundamentals) puts a limit on the extent to which international investors
are willing to hold a particular asset. The results for Regression D are broadly in line with those of regression
A, B and C.

Regression E advances that closed interest differentials in EU countries have been largely determined by
realized inflation and government deficits together with the persistence in capital controls. So domestic
monetary and fiscal policy are dominant determinants of closed interest differentials. Remaining insignificant
country risk premia (see the country-specific effects of regression E) depend on a small number of
"fundamentals" indicating confidence in the country’s policies.

For ease of comparison, we have ranked the generalized country-specific effects in Table 2. The estimates
for ζ_i recurrently coincide with a priori ranking of EXR, ES, LEFT and SIGGOV in Appendix B. In Table
3 we report formal statistical tests for positive c.q. negative correlations between the generalized country-
specific effects from regressions B and E and the variables EXR, ES, LEFT and SIGGOV.

Table 3 - Test for positive c.q.negative correlation between the generalized country-specific effect
and EXR, ES, LEFT and SIGGOV (absolute t-statistics are indicated between brackets)^a

a We calculate the following t-statistic for the correlation coefficient r; t - r
1992, p. 112).

where n-2 are the degrees of freedom (Dougherty,

^*   Significantly different from zero at the 95 % level of confidence (one-tailed test).

^** Significantly different from zero at the 99 % level of confidence (one-tailed test).

The results for EXR and LEFT are ambiguous. An increase in EXR leads to both an increase in the
generalized country-specific effect (B) and a decline in the generalized country-specific effect (E). An increase
in LEFT leads to a decline in the generalized country-specific effect (B) and a decline in the generalized
country-specific effect. Both EXR and LEFT are insignificantly related to the country-specific effect. This
result of EXR contrasts with that of Alesina, Grilli and Milesi-Ferretti (1994), while the result for LEFT
corresponds with Alesina, Grilli and Milesi-Ferretti (1994). Apparently, other explanations at the righthand-
side of the regression dominate. An increase in ES leads to an increase in the country-specific effect (both
B and E), but again ES is insignificantly related to the generalized country-specific effect. This may be
due to the variable INF which is already included in the right-hand side of the regression equation. Finally,

<<   <   17   18   19   20   21   22   23   >   >>