Reviewed by
Simone Righi
"Lendület" Research Center for Educational and Network Studies (RECENS), Hungarian Academy of Sciences
James Grattfelder's book develops a set of analytical and numerical tools which investigate the economic system as a large scale network. The starting point is that companies' economic decisions are made by the shareholders that control them. Therefore, the precondition to understand companies' behaviour is to investigate their ownership structure which control depends on. By studying massive amounts of ownership data, Grattfelder discovers that, unknown to most economists, the global economy system forms a network characterized by a very large connected component. This book is the Ph.D. thesis of the author, realized at ETH in Zurich. The main methodological innovation and empirical findings have been already published in peer-reviewed journals, but here they find a more comprehensive systematization. Probably every network scientist is familiar with some of its main results through the ample publicity that Grattfelder's work had in popular media. Nevertheless, this manuscript is very good reading for anyone interested in going beyond the flagship results to familiarize with their scientific derivation.
Effectively investigating large amount of ownership data as a network requires a series of specific measures. Specifically, one needs to move from the concepts of direct ownership, value and control to their network-based counterparties. This is extensively documented in Chapter 2, where the reader is also guided through the critical development of the flow measure "control" as obtained from a static measure "ownership". Chapter 3 then applies the methodology investigating the nature of ownership network (and consequently the flow of control) in a large number of companies quoted in 48 different national stock markets. While this chapter touches upon some interesting facts, it is in Chapter 4 that the power of the analysis of the economic system as a network thoroughly unfolds. Here, the focus shifts to the global network of corporate control that includes the world's most important transnational companies. Finally, Chapter 5 presents a modelling framework that aims to reproduce the most important empirical results of the previous chapters, while uncovering the driving forces behind the formation of the network topology observed.
The social simulation community will find this book stimulating both for the interesting stylized facts and the innovative methodological implications. This will allow social simulation scholars to derive inputs for models and new measures to analyze the structural characteristics of economic networks produced by models. Let me now briefly discuss more in detail the most interesting elements of this book and why they are relevant for social simulation.
The reader searching for a methodological reference and new ideas on how to study economic networks will find in this book an invaluable source of information. The mathematical derivation of the new network measures used by the author, as well as all the necessary steps to pass from data to network analysis, is explained in detail (although some knowledge of linear algebra is required). It is worth noting that the implications of this method go beyond the study of economic networks as they are applicable to the extraction of information from complex weighted networks in general (provided that some non topological variable can be associated to its nodes and edges). Among the different measures introduced, probably the most useful, for social simulators, is the "network control" of a node. This quantity integrates both the intrinsic value and the "integrated control" of an actor, defined as the economic value that she/he controls while considering the network of firms in which it has direct and indirect shares. This requires to understand companies and shareholders in their "network environment" in order to correctly estimate their value and their motivations, as well as to interpret their decisions. In this sense, agent-based models of heterogeneous agents are essential to help practitioners and managers to grasp the economic consequences of their position in the global economic network.
By applying this methodology, Grattfelder extracts the national backbone of control for different countries, which is in author’s words "the core subnetwork where most of the value of the stock market resides". He uncovers a large degree of heterogeneity with respect to the patterns of ownership and control that reflects the different social, legislative and institutional settings of the nations studied. While ownership patterns have already been explored in the literature, the size of the international comparison, the quality of the data available and the novelty of the approach makes this an exceptional result in itself. Furthermore, the network approach allows the author to expose the existence of an inverse relationship between the degree of dispersion of local ownership in the backbone and the degree of concentration of control observed at the aggregate level. Surprisingly, the more shareholding is diffuse the more aggregate control is restricted in the hands of a very few agents. The economic micro mechanisms that lead to this outcome and the influence of institutional and legislative settings are currently still unclear and a considerable modelling effort will be required to fully explore this problem. Given the complexity of the problem, agent-based simulation will be pivotal for this purpose.
The main finding of this book (widely popularized on the Internet) is that the corporations' network of control has a bow-tie structure both at a national and transnational level. A relatively small core is nested inside this structure. This is formed by a "tightly-knit group of corporations that cumulatively hold the majority of each other". Within the core, power-holders own a very large part of the control in the network (while accounting for a relatively small part of the revenues).
Grattfelder proposes a theoretical mechanism that reproduces these findings. His model allows for the co-evolution of network topology and a network-based fitness measure in the context of a preferential attachment model. While both the empirical results and the theoretical modelling are general and very interesting, this is just a first step towards understanding the emerging structure and the dynamical evolution of ownership network. For instance, the model assumes that individuals are interested to get attached to nodes with high network fitness. On the other hand, very little is known about the individual motivations that lead agents to make this choice. To understand this, conceptual and simulation work is still needed in the future.
Even considering the scientific advancements proposed in this book, the practical impact of Grattfelder's work is potentially the most important aspect of his contribution. This methodological approach to the study of economic networks is a must read for the anti-trust decision makers of the world and for financial regulators. Indeed, the high density of the control's network backbone exposes the world economy to systemic risks that need to be dealt with, for example with new rules on cross-ownerships among corporations. While unwinding the network of inter-relation among power-holders would be more bad than good for the economic system, a new - systemic - approach to the problem could advance our understanding of economic networks. The lesson here is that decision makers should consider companies not as isolated entities but as institutions with a position in a network and with, potentially wide, indirect influence (e.g. through control or counter-party risk) on many other system’s components.
To sum up, this book is of paramount interest for anyone concerned with social applications of complex networks and, in particular, for social simulators who want to base their models on new network measures and new stylized facts to better understand socioeconomic phenomena. While in my opinion a comprehensive economic analysis of these findings would have required a more dedicated effort, this books can stimulate the growth of innovative agent-based simulations targeted to disentangle social and economic implications of network structures in the economy.
Return to Contents
of this issue