CARIFORUM and UK EPA Study

Implicit in the balanced panel approach is the assumption that the errors for different cross sectional units are uncorrelated. We cannot generally conclude that this is the case for the cross sectional units in the regression; see the cross-section dependence test—we reject the null of no cross-sectional independence (correlation). The size of the coefficients might offer less information than the general association of the variables. We suspect that common regional shocks more than the presence of multicollinearity in the model specification could be responsible for cross-sectional interdependence (see also Pesaran (2004) and Hoyos and Sarafidis, 2006).

Residual Cross-Section Dependence Test Null hypothesis: No cross-section dependence (correlation) in residuals Equation: Untitled Periods included: 11 Cross-sections included: 14 Total panel observations: 154 Note: non-zero cross-section means detected in data Cross-section means were removed during computation of correlations

Test

Statistic

d.f.

Prob.

Breusch-Pagan LM Pesaran scaled LM

198.5529 7.972355 7.232049

91 0.0000

0.0000 0.0000

Pesaran CD

We disaggregate the index to isolate the contributions of human capital (HK) to the growth of the average volume of trade. Human capital and GDP growth are weakly significant (below the 95% significance level) but openness and liquidity remain very strong at conventional level. The existence of cross-sectional dependence is persistent; see the results below.

Page 227 of 241