Case: 2008 and the Credit Channel

The mechanism, in the order it happened

The starting point was a funding structure that could run: a large share of the system’s assets financed with short-dated wholesale borrowing, much of it secured on collateral whose value was itself a market price. When the collateral’s price fell, lenders demanded more of it, and when the demand could not be met the funding was withdrawn. That is a run, conducted through collateral schedules rather than through queues at branches, and it is the same mechanism the bank balance sheet case examined from the asset side. The second link was the marks. Assets whose value was derived from a market that existed only while nobody was selling fell in price as forced sales began, which reduced the equity of every holder and invited more collateral calls. This is the amplification loop from the previous case generalized: the price is an input to the funding, and the funding is what forces the sales, so the loop has no natural stopping point until an outside party supplies either capital or a market. The third link was the credit channel, and it is the part that makes this a macro case rather than a financial one. Banks and dealers, whose balance sheets were impaired and whose funding was fragile, reduced lending — both because they had less capital and because they could not fund the assets. So the cost of credit to households and businesses rose even as the policy rate fell, which is the mechanism by which a financial shock becomes a recession: the spread, not the rate, is what the economy pays. The chain, with the observable at each step — Assets fall in price: watch the marks and the collateral haircuts being applied to them · Collateral calls and funding withdrawal: watch the spread on short-term wholesale borrowing against the policy rate ← · Forced sales depress prices further: watch the volume of sales into a thinning market · Balance sheets contract and lending falls: watch the quantity of credit issuance, not just its price · Demand falls and unemployment rises: the lagged consequence of the credit contraction The facilities of 2008 are often described as a bailout of institutions, and they were also a repair of the transmission mechanism: lending against collateral at a penalty rate, guaranteeing liabilities that could otherwise run, and buying assets to create a market where none existed. The distinction matters analytically because each one acts on a different link in the chain, and none of them works by changing the policy rate.

Why the response had to work on the spread

If the problem is that the spread has widened, a response that only moves the policy rate is fighting one term of a two-term sum. The first generation of the response therefore aimed at the funding: facilities that lent cash against collateral to institutions that could not roll their borrowing, which addressed the run directly. The second generation aimed at the guarantees: insuring money-market fund balances and guaranteeing bank debt, which removed the incentive to run by making the liability safe. The third aimed at the assets: purchases that gave sellers a bid where the private market had stopped, which broke the mark-to-market leg of the amplification loop. Each of those acts on a different link, and the sequencing is diagnostic: a central bank that lends against collateral is saying the problem is liquidity; one that guarantees liabilities is saying the problem is confidence in a class of funding; and one that buys assets is saying the problem is the market itself. Reading a facility announcement in that framework is how you tell a liquidity operation from a solvency one, and it is the fastest way to know how serious the authorities believe the situation to be. The macro consequence was that the policy rate went to zero and the economy still contracted, because the policy rate is one input to conditions and the others were hostile. That is the case for the unconventional tools that followed, and it is also the reason the framework in the transmission lesson has to be understood before the events: someone reasoning from the policy rate alone would have concluded that a five percent cut must work, and the spread was the answer. • Lending against collateral addresses a run; guaranteeing liabilities addresses confidence in a funding class. • Buying assets creates a market where the amplification loop depended on there being none. • The sequencing of facilities is a diagnosis of what the authorities think the problem is. • Rates at zero with a hostile spread is still a tightening. The same framework reads every subsequent episode: in March 2020 the run was in the Treasury market itself and the response was purchases at scale; in 2023 the funding fragility was in a specific class of deposit base. The channel is stable even when the instrument changes.

What the case adds to a portfolio process

Three things. First, a position in a levered financial institution is a position in its funding structure as much as its assets — and the two cannot be separated in a stress event, which is why the funding mix belongs in the valuation rather than in a risk footnote. Second, the instruments that benefit are the ones whose cash flows are fixed and whose counterparties are strong: a widening spread is compensated for in the return of credit that survives, and the price of that compensation is the correlated risk the credit lesson described. Third, and more usefully in real time, the observables are specific and public. The spread on short-term wholesale funding against the policy rate, the haircut applied to the collateral being financed, the volume of credit issuance, and the facility announcements themselves. None of those requires a model, and all of them move before the employment data that confirm the recession. That is the practical transfer from the case: a list of series that describe the transmission channel directly, rather than the outcome it produces. The lesson that generalizes beyond 2008 is therefore about the direction of the diagnosis. When the policy rate is lowered and conditions do not ease, the answer is in the spread and in the quantities, not in the level of the rate. And when a central bank announces a facility, the instrument it chooses tells you which link in the chain it believes is broken — which is more informative than the size of the facility, and much more informative than the commentary about it. • A levered financial position is a position in its funding structure. • Credit that survives a widening is compensated; the cost is correlated risk. • Watch the funding spread, the haircut, credit issuance and the facilities themselves. • The instrument a facility uses diagnoses which link in the chain is broken. This is the same discipline the markets rung applies to order flow: read the mechanism rather than the headline, because the headline is a consequence of the mechanism and arrives later.

The run that happened outside the banks

The mechanism in this lesson is the spread, and the machinery that produced it sat largely outside the regulated banking system — which is why the event was both fast and, to supervisors watching bank balance sheets, largely invisible until it was under way. The structure is worth understanding in its own right, because it repeats. The model was simple. Loans were originated, packaged and sold; the buyers funded themselves with very short-term wholesale borrowing — commercial paper backed by those assets, and repo — rather than with deposits. A balance sheet that borrows short and holds long is a bank in everything but name. What it lacked was the furniture that makes a bank survivable: deposit insurance to make the funding sticky, capital requirements to absorb loss, and access to a lender of last resort as a backstop. The consequence was a run with a different shape from the one the textbooks describe. Deposit funding withdraws gradually, because depositors queue and because insurance removes the urgency for most of them. Wholesale funding does not reprice on the way out — it simply stops at the rollover date, and the rollover date is often the same day or the next week. Rollover risk is a cliff rather than a slope, which is why the sequence took days instead of weeks, and why the entities that failed did not first become expensive to fund; they first became impossible to fund. Three amplifiers turned a funding problem into a system event. Mark-to-market accounting transmitted falling prices into capital, which forced sales, which pushed prices further — the same spiral shape as a margin call. The money-market funds held the paper as cash equivalents, and when one large issuer’s notes became worthless and a fund’s net asset value fell below a dollar, that broke a belief rather than a price: investors who had treated those funds as cash withdrew from all of them, which cut off another source of short-term funding at once. And the monoline insurers and the rating models had certified the structure of the collateral, so the collateral’s failure invalidated a large part of the system’s risk assessment in one step. The policy response follows from the diagnosis. Because the problem was the availability of funding rather than its price, the facilities lent directly against collateral to institutions the ordinary discount window did not reach, and the guarantee programmes restored a belief rather than adjusted a rate. That is why the episode reads as a liquidity crisis with a solvency component rather than the other way around. The transferable reading is a question rather than a ratio. When a funding spread widens, ask which entity must roll over and cannot. That identifies the marginal borrower, which is the thing that breaks first, and it is more useful than the level of any index — the credit indices were late, and the funding data was not. • A balance sheet borrowing short to hold long is a bank without the supporting furniture. • Wholesale funding stops at the rollover date rather than repricing, so the run is a cliff. • Mark-to-market, the money-market funds and the rating models were the amplifiers. • Ask which entity must roll and cannot; that identifies what breaks first. The observables to watch were the funding spreads and the commercial-paper volumes, not the equity indices. A portfolio process that reads funding data gets its warning from the place the stress originates rather than from the place it arrives.

Sources

• The credit channel of monetary policyBernanke & Gertler (1995), “Inside the Black Box”; Bernanke (1983) on the financial accelerator
• The run on repo and the shadow banking systemGorton & Metrick (2012); Financial Crisis Inquiry Report (2011)
• The policy response and its facilitiesFederal Reserve and FDIC documentation of the 2008–2009 facilities; standard accounts of the period