National industry cluster templates and the structure of industry output dynamics: a stochastic geometry approach

Cluster analysis has been widely used in an Input-Output framework, with the main objective of uncover the structure of production, in order to better identify which sectors are strongly connected with each other and choose the key sectors of a national or regional economy. There are many empirical studies determining potential clusters from interindustry flows directly, or from their corresponding technical (demand) or market (supply) coefficients, most of them applying multivariate statistical techniques. In this paper, after identifying clusters this way, and since it may be expected that strongly (interindustry) connected sectors share a similar growth and development path, the structure of sectoral dynamics is uncovered, by means of a stochastic geometry technique based on the correlations of industry outputs in a given period of time. An application is made, using Portuguese input-output data, and the results do not clearly support this expectation.