GDP/capita correlations are investigated in various time windows, for the time
interval 1990-2005. The target group of countries is the 25 EU members , 15 till
2004 plus the 10 countries which joined EU later on. The correlation
coefficients are averaged by turning them either into coefficients of
determination or Fisher z-scores respectively; a comparison between the two
methods is done; the first should be preferred. The mean correlation coefficients
are taken as the weights of a fully connected network having the countries as
nodes. A cluster structure of EU countries is derived from the statistically
relevant eigenvalues and eigenvectors of the adjacency matrix. The clustering
scheme is found to be in accordance with the results reported in the recent
economic literature. Finally, the properties of the weighted network are
investigated by means of some statistical physics tools, by considering the
statistical ensemble of the networks with fixed number of vertices. Particularly,
the concept of ‘’entropy’’, ‘’order parameter’’, ‘’susceptibility’’ and ‘’specific
heat’’ (!) are introduced through the moments of the probability distribution of
the weights at each one time particular realisation of the statistical ensemble.
This may be considered to yield some information about the structure, stability
and evolution of the EU country clusters in a macroeconomic sense. |