Tsallis entropy for cross-shareholding network configurations

TL;DR

This paper introduces a novel application of Tsallis entropy to analyze the dependence between diversification and integration in cross-shareholding networks, providing insights into market structure and policy implications.

Contribution

It develops a Tsallis entropy-based framework combined with copulas to model and analyze the joint distribution of diversification and integration in corporate networks.

Findings

Tsallis entropy reveals market polarization and fairness dynamics.

Copula-based dependence modeling captures market reaction to shocks.

Empirical validation on Italian companies supports the theoretical approach.

Abstract

In this work, we develop the Tsallis entropy approach for examining the cross-shareholding network of companies traded on the Italian stock market. In such a network, the nodes represent the companies, and the links represent the ownership. Within this context, we introduce the out-degree of the nodes -- which represents the diversification -- and the in-degree of them -- capturing the integration. Diversification and integration allow a clear description of the industrial structure formed by the considered companies. The stochastic dependence of diversification and integration is modelled through copulas. We argue that copulas are well suited for modelling the joint distribution. The analysis of the stochastic dependence between integration and diversification by means of the Tsallis entropy gives a crucial information on the reaction of the market structure to the external shocks, - on the basis of some relevant cases of dependence between the considered variables. In this respect, the considered entropy framework provides insights on the relationship between in-degree and out-degree dependence structure and market polarisation or fairness. Moreover, the interpretation of the results in the light of the Tsallis entropy parameter gives relevant suggestions for policymakers who aim at shaping the industrial context for having high polarisation or fair joint distribution of diversification and integration. Furthermore, a discussion of possible parametrisations of the in-degree and out-degree marginal distribution, -- by means of power laws or exponential functions, -- is also carried out. An empirical experiment on a large dataset of Italian companies validates the theoretical framework.