The Exclusive Club of Alliances: von Neumann's Recipe for Stability ✛
Eight hundred participants, firms and workers, enter the market. Each holds a strict preference list in mind. The challenge: pair up so that no one wants a do-over. In mathematics, this is called a game theory with a ban on repeat alliances, and its roots trace back to the work of John von Neumann and Oskar Morgenstern. They described stable sets — a kind of exclusive club of alliances, where only special partner distributions are admitted. But what exactly belongs in this club remained a mystery for a long time.
Getting into the club isn't easy. It requires two qualities: internal stability (members don't overthrow each other) and external (any outsider is beaten by someone in the club). In theory speak, that's a von Neumann–Morgenstern stable set. Researchers wondered: can we describe this club without sifting through countless combinations? The answer turned out to be elegant: simply strike from the preference list all hopeless pairs. The first procedure, essentially optimization, relies on the lattice structure of the core — the set of distributions with no improving coalitions. By sequentially removing incompatible options, it preserves only those relationships that don't disrupt the dominance landscape. The second procedure, efficiency-oriented, filters out “breaker pairs” that block Pareto improvements. Combining the two cleanups yields the reduced profile P*—a collective behavior mirror of dominance ties.
Main result: the stable set is unique and exactly matches the core of the profile P*. In other words, the exclusive club is precisely all undominated outcomes in the trimmed game. The system of dominance ties works so that any outcome outside the club gets beaten by one of its members, and the path there takes at most three steps. The upshot: to find stable pairs, you don't need to enumerate subsets, just apply two simple procedures and take the core.
The discovery blurs the line between cooperative and structural views of stability. Dominance relations are now tied to topology. Beyond theory, a practical algorithm emerges for organ allocation, school choice, and Graphs. And in the future — extending the method to many-to-one markets and studying Dynamical Systems, where agents with decentralized decision making drift toward stability on their own. The three‑step estimate suggests such a process converges quickly. So eight hundred participants can rest easy: their club is predictable.
The decomposition lemma, normally a tool just for the core, is extended here for the first time to any internally stable set. This generalization is what allowed “slicing” the market into independent IS-cycles and building the trimmed profile.