Why a captured company can't un-capture itself: the charter reason
Institutional capture feels irreversible - a captured company stays captured even after the faction that captured it shrinks. That intuition is right and old: it is the same bistable, hysteretic mathematics as a magnet, as QWERTY lock-in, as stuck unemployment. A tidy toy model reproduces it - but the model is textbook physics, its 'irreversibility' is a parameter choice, and its mechanism (shareholders tipping into consensus) is not how corporate control actually works. The real reason a captured firm stays captured is charter architecture - staggered boards, poison pills, coupled ownership - set before the fight and hard to undo by vote.
The intuition is right, and old. A captured company - one flipped by a faction of committed shareholders or activists - tends to stay captured even after that faction shrinks. Removing the cause does not restore the prior state. This is hysteresis, and it is not new: it is the same bistable mathematics James Ewing named for magnets in 1885, that Paul David called QWERTY lock-in in 1985, and that economists imported for stuck unemployment. Some transitions are one-way.
A toy model reproduces it - and shows exactly why it is textbook. We modeled a board/shareholder bloc as a mean-field Glauber (kinetic Ising) system: persuadable agents, a coupling J, a field favoring good governance, and a committed faction pinned at the capture position. We measured the stake that captures a well-run firm (f_up) and the stake below which a captured firm recovers (f_down):
| ownership coupling | capture stake (f_up) | recovery edge (f_down) | hysteresis |
|---|---|---|---|
| weak (J=1.2) | 14% | 14% | 0% - reversible |
| J=2.0 | 22% | 0% | 22% |
| J=3.0 | 28% | 0% | 28% |
| J=4.0 | 32% | 0% | 32% |
A hysteresis loop opens above a critical coupling and widens - which is exactly what a mean-field Ising model in a field must do (a first-order transition with a metastable region and a spinodal). The full loop for competing committed groups - two stable states bounded by two spinodal lines meeting at a cusp - was already published (Xie et al., 2012); the forward threshold f_up is a committed-minority tipping threshold of the same kind measured empirically (Xie et al., 2011, ~10%; Centola et al., 2018, ~25% - different numbers, same mechanism). And the “irreversibility” (f_down=0) is not a governance law - it is the assumption that the ownership coupling is stronger than the fundamentals field. Strengthen that field in the model and the loop closes; the hysteresis widths (0/22/28/32%) are a coordinate readout of the parameters, not measured facts about firms.
The mechanism is wrong for real firms
Here is the honest problem, and it is the interesting part. Corporate control is not shareholders tipping into a consensus. It is vote-counting against a legal threshold, gated by charter architecture. In 2010 Air Products won three of Airgas's nine board seats and the shareholder argument - and still could not take control, because Airgas had a staggered board plus a poison pill; the Delaware Court of Chancery upheld the pill (In re Airgas, 2011) and Air Products walked away. No hostile bidder in the modern record has taken a staggered board by voting out incumbents: it requires winning director majorities at two consecutive annual meetings. The irreversibility is written into the charter before the contest - which is why founders lock in staggered and dual-class structures at the IPO, when shareholders can still amend them, and why activists strike before the defenses harden. That is a discrete legal state, not a tipping point in belief.
The real stickiness is documented - and it is not a fitted loop
The governance literature already has the substance. Corporate ownership is path-dependent (Bebchuk & Roe, 1999); managerial entrenchment is powerful, asymmetric, and value-destroying (the entrenchment index, widely used in empirical governance research); regulatory capture (Stigler, 1971) is the industry-side analogue; and ownership is now highly coupled - the “Big Three” index funds rose from ~5% to ~20% of the S&P 500 and are projected toward ~40% of the vote (Bebchuk & Hirst, 2019), a bloc that overwhelmingly backs incumbents. But this literature measures persistence and asymmetric switching costs - not a two-sided capture-vs-recovery threshold. Our hysteresis is modeled; the recovery threshold has, as far as we can find, never been measured in real firms.
The falsifierThe claim that would have been worth publishing is not “a captured firm stays captured” (textbook, and known since Ewing) but a governance-specific mechanism physics lacks - e.g. an endogenous field, where firm performance feeds back on the entrenchment incentive - or a model built on the actual legal control state rather than opinion tipping. We have neither yet. Absent that, this is a mean-field Ising model in a governance costume.
Honest scope - what changed. An earlier version of this post framed the hysteresis loop as a novel, additive finding. A full audit found the core is textbook statistical mechanics, the full loop was already published (Xie 2012), the numbers are parameter-dependent, and the model's mechanism does not match how corporate control actually works. What stands, and is worth saying plainly: institutional lock-in is the same bistable mathematics across magnets, technologies, economies and opinions; and in real firms the one-way door is charter architecture, not a tipping point. Runnable model.
FAQ
Can a captured company simply reverse course? Often not - but the real reason is charter architecture (staggered boards, poison pills) plus coupled ownership that backs incumbents, not a tipping point in shareholder opinion. Air Products won three of the nine Airgas board seats in 2010 and still could not take control.
Is “governance hysteresis” a new finding? No. Hysteresis / bistability above a critical coupling is textbook physics (Ewing named it in 1885), and the full loop for competing committed groups was published by Xie et al. in 2012. The contribution here is only a governance-flavored illustration, and the honest correction that the corporate mechanism is legal, not opinion-based.
How big is the governance hysteresis? The model's numbers (0/22/28/32%) are a coordinate readout of its parameters, not measured firm quantities. f_down going to 0 is simply a weak fundamentals field relative to a strong ownership coupling; strengthen the field and the loop closes. Real recovery thresholds have not been measured.
What actually locks capture in? Defenses set into the charter before the contest (staggered boards, poison pills, dual-class shares) that shareholders later cannot amend, plus concentrated/coupled ownership (the Big Three, proxy advisors) that reliably backs incumbents. Prevent it early; afterward it is a legal problem, not a persuasion problem.