I woke up to this story on Boing Boing:
Cory Doctorow | Densely-linked cluster of 147 companies control 40% of world's total wealth
Normally I would not step into the abyss of economic ignorance which is Bng Bng, but I am rested, and feeling strong, and I have actually written a paper on network analysis in business, so I thought I would flip through the paper Doctorow references (Vitali, Glattfelder & Battiston: The network of global corporate control [PDF]).
I will say this once, and clearly: What the authors definitely DO NOT claim is that 147 companies control 40% of the world's total wealth. What they claim is that those companies control 40% of the gross operating revenue of the transnational corporations in their data set. Revenue is not wealth, and the wealth of some large firms is not the wealth of the entire world.
Furthermore, as best I can tell from their numbers, that 40% figure assume a threshold model of control, so that if Company A owns 51% of Company B, and B owns 51% of C, A "controls" all of the revenue of A, B, and C. This threshold model (and to a lesser extent, the softmax model they mention) is naturally going to lead to a concentration. No way around it. I will leave it to the business governance experts as to whether this is a good assumption or not, but my inclination is that it entirely ignores agency problems, which can not be a good thing. (Good enough for the statistical and network analysis methods Vitalli et al. wanted to demonstrate; not good in terms of accurately understanding the world economy.)
PS: "A relevant additional fact at this point is that 3/4 of the core are financial intermediaries." Is it not the purpose of financial firms to invest in other firms? It sounds kind of scary to hear Merrill Lynch controls 1% of the revenue of the network, but then you realize "oh that just means my 401k is invested in large multinational firms; what the hell else do you expect Merrill to do?" Corporations aren't these alien things; they're us.
PPS – Edited to add – Naked Capitalism addresses the faulty assumptions in this paper regarding ownership here. This is a good example of why physicists, mathematicians and computer scientists who do this sort of network analysis, machine learning, systems modeling or statistics on data sets in other fields really, really ought to find a co-author with good domain knowledge. The authors here do some interesting work calculating influence* on cyclical graphs. Too bad it's muddied up with dubious business interpretations.
* I mean "influence" in the limited, value-neutral sense with which it is used in network analysis, not in the broader, negative sense of potentially nefarious control over another person.