Relatively small sectoral productivity shocks could lead to sizable macroeconomic variability. Whereas most contributions in the literature analyze the issue of aggregate sensitivity using simple general equilibrium models, a novel approach is proposed in this paper, based on stochastic simulations with a global CGE model. We estimate the statistical distribution of the real GDP in 109 countries, assuming that the productivities of the industrial value added composites are identically and independently distributed random variables. We subsequently undertake a series of regressions in which the standard error of the GDP is expressed as a function of variables measuring the "granularity" of the economy, the distribution of input-output trade flows, and the degree of foreign trade openness.We find that the variability of the GDP, induced by sectoral shocks, is basically determined by the degree of industrial concentration as counted by the Herfindhal index of industrial value added. The degree of centrality in inter-industrial connectivity, measured by the standard deviation of second order degrees, is mildly significant, but it is also correlated with the industrial concentration index. After controlling for the correlation effect, we find that connectivity turns out to be statistically significant, although less so than granularity.