Boardroom Centrality and Firm Performance

Two boards are linked when they share a director. Do that for every US public company and you get one enormous undirected graph — 115,411 directors, rebuilt each year. The paper scores each board’s position in it with the four standard centrality measures (degree, closeness, betweenness, eigenvector) plus an average of the four they call the N-Score, then sorts firms into quintiles. The most central minus the least central earns 4.68% a year, after controlling for industry, size, book-to-market and momentum, and it holds year by year and across industries.

The accompanying evidence says the return is not a fluke of the sort: central firms go on to post higher growth in return-on-assets, and analysts systematically under-forecast that growth. Both effects concentrate where a network would matter most — firms with high growth opportunities, and firms in trouble. The authors read this as prices under-reacting to a real economic benefit of director networks rather than as compensation for a hidden risk, though they are careful that the second reading is not excluded.

The transferable move is that the predictor is relational, not accounting. Nothing here is a property of the firm; it is a property of the firm’s position among other firms, assembled from disclosure that is public, free, and structurally discarded the moment you flatten a market into one row per company. Any dataset where entities share people, suppliers, auditors or counterparties has the same graph hiding inside it.