Stability Breeds Instability: The Minsky Model as a Positioning Tool

Hyman Minsky spent a career arguing something that sounds like a paradox and is a mechanism.

Stability is destabilising. A long calm period does not reduce the probability of a crisis. It manufactures one, and it does so through the ordinary, prudent behaviour of participants who are each behaving sensibly.

Kindleberger turned it into a history. Every mania he catalogued follows the same five beats, over four centuries, across every asset that has ever existed.

The five beats

Note what is missing from that sequence: villains, stupidity, and irrational exuberance. Minsky’s participants are all rational. That is the point, and it is why the cycle cannot be regulated away.

The mechanism, in one table

Here is how a prudent risk model builds a crash.

A trader, a fund, or a bank targets a constant amount of risk. Say a two percent daily Value-at-Risk. That is responsible. It is what a risk committee exists to enforce.

Leverage is then whatever it needs to be to reach that risk target. And leverage is inversely proportional to volatility, because when the market moves less, you must hold more to lose the same amount.

Read the third column. Measured risk is constant. The risk committee is satisfied at every row. Nothing has been violated.

Read the fourth. The same shock, the same seven percent adverse move, produces a loss six times larger at the bottom of the table than at the top.

Nobody made a mistake. Every participant obeyed a rule designed to protect him. The rule was calibrated on a quantity that was, by the time it mattered, meaningless.

Why the shock is not proportional

The table’s crucial assumption is that the shock does not care about recent volatility. Seven percent is seven percent, whether the market has been drifting or convulsing.

This is true, and it is why the paradox exists. Shocks are caused by events. A rate decision, a default, an invasion, an earnings revision. None of those consult last quarter’s realised volatility before choosing their magnitude.

A risk model that scales exposure by recent volatility is therefore making a bet: that tomorrow’s surprises will be proportional to yesterday’s noise. In quiet regimes that bet is systematically wrong, and it is wrong in the direction that produces maximum leverage at the moment of maximum fragility.

Where you meet this personally

The quiet market that tempts you to size up. Ranges tighten, ATR compresses, your stops get closer, and your position size, computed correctly from the stop, grows. This is right, and it is the same mechanism. If the range breaks on news, your correctly-sized position is several times larger than it was a month ago.

The strategy that has not lost in months. Every quiet month lowers your estimate of the strategy’s volatility, raises your confidence, and raises your size. Your risk of ruin has been increasing throughout the period in which your evidence for the strategy has been improving.

The funded account after a good run. Profits build a buffer, the buffer invites larger size, and the larger size arrives just as your measured drawdown risk looks lowest.

The carry trade, in any form. Small, steady gains, financed. It works until it does not, and the unwind is fast because everyone financed the same thing for the same reason at the same time.

Distress, and what it looks like

Stage four is the tradeable one, and it is the only stage that gives a signal before the loss.

Distress is not a decline. It is the failure to advance, on good news, at the top. The asset that no longer rises when it should is telling you that the marginal buyer has already bought, and that the position is now held by people whose only reason for holding is that it has been going up.

Kindleberger’s markers, translated for a chart: volume rising on down days and falling on up days. New highs made by fewer and fewer names. Volatility rising while direction stalls. Credit spreads widening while equity ignores it.

None of these predict a date. All of them describe a market in which the seller has more urgency than the buyer, and that condition precedes revulsion in every episode in Kindleberger’s four hundred years.

What to do with this

Treat falling volatility as a warning, not a green light. When realised volatility drops below its long-run average, your risk model will invite you to hold more. Decline the invitation. Cap leverage in absolute terms, not in volatility-adjusted terms.

Size the shock, not the noise. Ask what a seven percent adverse gap does to your book, tonight, with no chance to exit. That number is your real risk, and it is the only one that will matter on the day it matters.

Notice euphoria in yourself first. The tell is not the price. It is the sentence “this has been working so well,” which is a statement about the past being used as a claim about the distribution.

Do not wait for the top. Distress precedes revulsion by weeks or months, and the top precedes nothing, because it is only visible afterwards. You are not trying to sell the high. You are trying not to be levered four times when the shock arrives.

The four-hundred-year conclusion

Kindleberger’s history is not a warning about greed. It is something colder: a demonstration that the crash is assembled, brick by brick, out of prudent decisions made by intelligent people using risk models that work.

Which is why it will happen again, and why the participants will once more be surprised, and why their surprise will be sincere. The models will have reported two percent risk, every day, right up until the day they reported a twenty-eight percent loss.

Stability is not the absence of risk. It is the period during which risk is being accumulated, invisibly, by people who have been told to measure it in the one unit that has stopped working.