Portrait comparison of binary and weighted Skill Relatedness Networks

Abstract: In this paper we compare Skill-Relatedness Networks (SRNs) for selected countries, that is to say statistically significant inter-industrial interactions representing latent skills exchanges derived from observed labor flows, a kind of industry spaces. Using data from Argentina (ARG), Germany (DEU) and Sweden (SWE), we compare their SRNs utilizing an information-theoretic method that permits to compare networks of "non-aligned" nodes, which is the case of interest. For each SRN we extract its portrait, a fingerprint of structural measures of the distributions of their shortest paths, and calculate their pairwise divergences. This allows us also to contrast differences in structural (binary) connectivity with differences in the information provided by the (weighted) skill relatedness indicator (SR). We find that, in the case of ARG, structural connectivity is very different from their counterpart in DEU and SWE, but through the glass of SR the distances analyzed are all substantially smaller and more alike. These results qualify the role of the SR indicator as revealing some hidden dimension different from connectivity alone, providing empirical support to the suggestion that industry spaces may differ across countries.