Please use this identifier to cite or link to this item:
Title: Network of firms: an analysis of the relevance of integrated ownership in market concentration
Authors: Rotundo, Giulia
D'Arcangelis, Anna Maria
Keywords: ownership; control; operations research; complex networks
Issue Date: 2009
Source: G. Rotundo, A. M. D’Arcangelis, “Network of firms: an analysis of the relevance of integrated ownership in market concentration”. In: Science and Technology for Humanity (TIC-STH), 2009 IEEE Toronto International Conference, Toronto, Canada (26-27 Sept. 2009) pp. 685 – 690. ISBN: 978-1-4244-3877-8, INSPEC Accession Number: 11229830, DOI: 10.1109/TIC-STH.2009.5444411.
This paper aims to provide an analysis the structure of ownership and control of firms whose
shares are traded on the Italian Stock Market. The work is relevant for adding knowledge on the
diversification of risk and ultimately to work for the development of risk indicators beyond the
standard approaches most based on the time series analysis of raw price time series. The dataset
reports the shareholdings of 247 companies traded on the Italian stock market. Data, adjourned at
May 2008, were retrieved through the AIDA database, integrated by the Bureau van Dijk
databases BANKSCOPE, ISIS, and cross-validated through CONSOB and MEDIOBANCA
reports. Therefore, this dataset allows to consider a sampling larger than the one examined in [2],
even considering the difference in the companies traded in the Italian stock market due to the
different sampling date. Fig. 1 shows the most connected nodes of the shareholding network.
We are most interested in understanding the role of portfolio diversification and portfolio size in
the structure of ownership and control.
Methods used in the present analysis involve statistical analyses most proper of the field of
complex networks [3] and graph flow analysis typical of operations research approach [1, 4].
We start our analysis building a network from data. Each company corresponds to a node and a
link from node i to node j exists if i owns shares of j. Therefore, we obtain a directed graph and
the direction of our links is the opposite of the ones of [2], but the same used in [4].
Therefore, in our network construction, the number of links exiting from a node, kout, measures
portfolio diversification. The number of links entering in a node, kin, shows the number of
shareholders, but this data is biased due to the sample, like it happens in [2].
We are most interested in outlining the difference between ownership, control, and the overlap
between ownership and control paths and portfolio diversification. Portfolio diversification is
measured through mere statistical analysis, assortativity, and hierarchical paths estimate, in
accord with the measure introduced in [5]. We both estimate the probability distribution and the
relationship between kout of network nodes. We remark the absence of power laws, although a
comparison with the results of [2] allows to detect the tendency to lower the diversification.
We consider also the correlation of kout between different nodes. Having detected a low positive
degree of assortativity (0.17), we deepen the analysis by estimating the percentage of
hierarchical paths. Two nodes i and j are in a hierarchical path if an “up” path exists from node i
through nodes with higher kout, followed by a “down” path where nodes on the path have a
decreasing kout. Therefore, a hierarchical path exists if node i is in the portfolio of a larger
investor, in which also the smaller j is investing.
Capitalization is introduced and the analysis is carried on the portfolio size as well. This allows
to measure the overlap between portfolio diversification and capitalization, and to answer to
questions on the possibility to diversify portfolios for companies having a smaller capitalization.
Given the distribution of nodes, we determine its distance form the maximum (minimum)
hierarchical ones, and provide ranges for possible scenarios.
We compare this data with the results of the analysis of ownership/control, based on techniques
of operational research [1, 4]. The method shown in [4] deals with directed acyclic graph (DAG),
whilst [1] considers connected components. The entire network becomes DAG only whether
links corresponding to more 10% share ownerships are considered. Hence, cycles cannot be
eliminated without cutting relevant information. The results emphasize that companies traded on
the Italian stock market most use direct control.
ISBN: 978-1-4244-3877-8
Appears in Collections:DEIM - Archivio della produzione scientifica

Files in This Item:
File Description SizeFormat
R18.doc29 kBMicrosoft WordView/Open
Show full item record

Page view(s)

Last Week
Last month
checked on Dec 5, 2020


checked on Dec 5, 2020

Google ScholarTM



Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.