Somewhere around the third day of a credit scare, a bank treasurer at an institution that made no bad loans, holds no troubled assets, and lent nothing to the firms in the headlines starts to notice something uncomfortable. The overnight rate she pays to roll her funding has ticked up. Not by a lot. Just enough to wonder.
That is the contagion mechanism, operating exactly as designed, which is to say: not designed at all.
The plumbing nobody thinks about until it leaks
Interbank lending markets are, at their simplest, the system by which banks lend spare reserves to each other overnight, or for a few days, or a few weeks, to manage the gaps between what flows in and what flows out on any given day. A large retail bank might collect more deposits on a Tuesday than it can immediately deploy; a mid-sized lender might have the opposite problem. The interbank market stitches these mismatches together, and the rate at which this happens, the overnight rate in a given currency, is the rate central banks spend most of their energy trying to influence.
The critical structural feature is that this market runs almost entirely on confidence and counterparty reputation, not on collateral. Unsecured interbank lending, the kind that dominated wholesale funding before the 2008 crisis and still constitutes a significant portion of short-term bank funding globally, involves one institution deciding that another is good for the money by tomorrow morning. No collateral changes hands. The lender is simply trusting the borrower's balance sheet.
This creates a system that is spectacularly efficient in calm conditions and spectacularly fragile in stressed ones. The fragility doesn't require any direct exposure between the two banks. It requires only that they share the same market.
How stress travels without a wire
Consider two fictional but structurally plausible banks: Aldermoor Bank, a mid-sized lender whose loan book is concentrated in commercial real estate in one region, and Crestfield Savings, a retail-focused institution two countries away whose loans are almost entirely residential mortgages. They have never lent to each other, never held each other's bonds, never shared a meaningful counterparty. By any direct-exposure analysis, they are strangers.
Now suppose Aldermoor's commercial real estate portfolio starts showing strain. Rumours circulate, then a credible analyst report, then a regulator's comment. Aldermoor's overnight borrowing rate rises as lenders in the interbank market demand a premium for the perceived risk. Aldermoor starts paying, say, forty basis points above the benchmark to roll its overnight book.
Crestfield watches this and faces a structural problem that has nothing to do with its own portfolio. Other banks in the market, having just absorbed a reminder that counterparty risk is real and information is imperfect, revise their general willingness to lend unsecured. They cannot easily distinguish which balance sheets are pristine and which are concealing problems. The rational response, under uncertainty, is to demand more from everyone, or to shorten maturities, or both. Crestfield's funding cost rises by fifteen basis points. It has done nothing wrong. It is simply downstream of the same information environment.
This is not a theoretical edge case. It is the normal transmission mechanism, and it has a name in the academic literature: information contagion, as distinct from the direct balance-sheet contagion that comes from shared exposures. The two reinforce each other in a real crisis, but information contagion can travel alone. It spreads the way a rumour does through a crowded room, touching people who never met the source.
The roll-over cliff and why maturity matters
The structural vulnerability deepens when you account for maturity transformation. Banks routinely borrow short and lend long. A bank might fund a five-year corporate loan with a sequence of ninety-day interbank borrowings, rolled over twenty times. This works beautifully when each rollover is routine. It becomes a cliff edge when lenders in the interbank market decide to shorten their own lending maturities or exit certain counterparties entirely.
A bank that previously rolled ninety-day funding might find, after a stress event elsewhere in the system, that the market will only commit to overnight. The bank's underlying asset hasn't changed. The loan is performing. But the funding structure has compressed from ninety days to one, meaning the treasury must now return to the market every single morning to stay solvent. The operational cost is real, the psychological pressure on treasury staff is real, and the probability of a bad morning, one where the market simply isn't there at the right price, has multiplied by ninety.
This is the roll-over cliff. It killed Northern Rock not because its mortgage book was catastrophically bad by the standards of that era, but because its funding model depended on a wholesale market that evaporated. The mortgages were still there. The lenders weren't. That is worth sitting with for a moment, because the instinct is always to hunt for the rotten asset. Sometimes the asset is fine. The architecture is the problem.
What people consistently get wrong about this
The most common misconception is that stress transmission requires a visible chain of exposure, a to b to c, like dominoes arranged in a tidy line. Regulators spent decades mapping direct counterparty networks on exactly this assumption. The network maps were useful, but they missed the ambient channel: the shared price of confidence.
When one institution's distress raises the general risk premium in an interbank market, it taxes every participant in that market simultaneously, regardless of their individual health. A bank with a perfectly clean balance sheet and a funding model that requires daily market access is, by this logic, more vulnerable to this ambient effect than a poorly run bank that happens to be deposit-funded and doesn't touch the interbank market at all. Risk, in this framing, is partly a function of architecture. Asset quality alone does not save you.
There is a related error in how people interpret central bank interventions. When a central bank floods the system with liquidity, it is not primarily rescuing the banks with bad assets. It is restoring the ambient confidence level so that the interbank market can function again, which means it is also rescuing Crestfield Savings, the bank that did nothing wrong but built its funding model on a foundation that required other people to keep showing up every morning.
Ask yourself this: if your institution's solvency depends on forty counterparties independently deciding, each night, that you are trustworthy, what exactly are you managing? Not a balance sheet. A reputation in a market that can reprice you in hours.
The banks that came through 2008 most comfortably were not uniformly the ones with the best loan books. Several had mediocre ones. What they shared was a funding structure that didn't require daily permission from a nervous market to stay alive. That is a quieter kind of resilience, harder to read on a balance sheet, and it tends to get almost no attention in the good years. History suggests that is precisely when it deserves the most.