Global Investment Outlook 2008 — The Funding Break

2008 is filed under "credit crisis", and that description is accurate about the origin and misleading about the mechanism.

The losses on the underlying loans, large as they were, were not by themselves sufficient to bring down the institutions that failed. What brought them down was that they financed long-dated, hard-to-sell assets with short-dated money — borrowing that had to be renewed continuously, in some cases overnight — and that money stopped being renewed.

An institution in that position does not fail because its assets are worthless. It fails because it cannot find the cash to replace the funding that just left, and it cannot sell the assets quickly enough at anything like their carrying value. It is possible to be solvent on a mark-to-maturity basis and dead by Friday.

That distinction matters enormously, because the two problems have different warning signs, different remedies and different afterlives. A credit problem is resolved by absorbing losses. A funding problem is resolved by someone standing behind the funding — which is what central banks eventually did, and why the resolution arrived so abruptly once they did.

Three mechanisms compounded the basic mismatch, and each recurs in this archive:

Securitisation did not disperse risk as intended, because a large share of the securities ended up held by vehicles funded the same short-dated way. Risk was moved and not reduced.

Mark-to-market accounting created a spiral. A forced sale set a price, that price marked everyone else's holdings down, those marks triggered more calls, which forced more sales.

Counterparty uncertainty froze activity. Once nobody could tell which institutions were impaired, the rational response was to transact with none of them.

Why a funding run is different from a credit loss

The distinction is the most important thing in this report, and it is routinely collapsed.

A credit loss is a permanent reduction in the value of an asset. A borrower does not repay; the lender writes down. It is painful, it is measurable, and it is absorbed against capital. An institution with enough capital survives it.

A funding run is a withdrawal of the borrowing that finances the assets. Nothing about the assets need have changed. The institution must either replace the funding or sell assets, and in a stressed market both are unavailable at reasonable terms simultaneously.

Three properties make a run categorically different:

• Speed. Credit losses emerge over quarters as borrowers deteriorate. Funding disappears in days, sometimes hours, because the decision to withdraw is made by someone protecting themselves rather than by a credit process.
• Self-fulfilment. A rumour of trouble causes withdrawal, which causes trouble. Credit losses do not work this way — a loan does not default because people believe it will.
• Solvency is not the trigger. A run can kill an institution whose assets, held to maturity, would have paid in full. The question is not whether the assets are good but whether the holder can survive long enough to find out.

Solvency is a statement about assets and liabilities. Liquidity is a statement about timing. Institutions fail on timing far more often than on arithmetic, and 2008 is the largest demonstration of that on record.

The policy implication follows directly, and it explains why the eventual resolution was so abrupt. If the problem is a run, the remedy is a credible lender willing to provide the funding that fled. Once that existed at sufficient scale, the mechanism stopped working — not because the underlying credit had improved, but because the withdrawal no longer forced a sale.

The duration mismatch, stated precisely

Because this mechanism recurs throughout the archive, it is worth setting out in its general form rather than as a description of 2008.

The structure: hold assets with long or uncertain maturities. Finance them with liabilities that must be renewed frequently. Earn the difference between the long rate received and the short rate paid.

Why it is profitable: short-term borrowing is normally cheaper than long-term lending. The spread is the business model of every bank, and of a great deal else besides.

Why it is fragile: the profit depends on the short-term funding being continuously available. The institution is not being paid for credit risk alone. It is being paid for taking the risk that the funding stops — and that risk is invisible for long stretches, because the funding almost always rolls.

What makes the fragility acute:

• The shorter the funding, the more often the risk is taken. Overnight funding is renewed 250 times a year. Each renewal is a decision by someone else.
• The less liquid the asset, the worse the alternative. If funding stops and the asset cannot be sold quickly, there is no orderly exit.
• Leverage multiplies both. More assets per unit of capital means a smaller price move exhausts the buffer.

This is not a description of a mistake. It is a description of maturity transformation, which is a genuine economic function — savers want their money available, borrowers want it committed, and somebody must bridge that. The bridge is inherently fragile, and the fragility is the service being provided.

The recurrence is the point. The 2023 banking stress described in the Global Investment Outlook 2023 was the same structure with different assets: institutions holding long-dated, high-quality government bonds financed by deposits, which left when rates rose. The credit quality was impeccable and irrelevant. The mechanism does not care what the asset is.

Why securitisation transmitted rather than dispersed

Securitisation was justified on a sound argument that turned out to be conditional in a way few had examined.

The argument. Pooling many loans and selling claims against the pool distributes risk to many holders, each of whom takes a small, diversified exposure. Risk ends up spread across the system rather than concentrated in the originating bank. That is genuinely how it is supposed to work.

The condition the argument depended on: that the ultimate holders are diverse, and that they hold the securities in a way that lets them absorb losses without being forced to sell.

Why the condition failed:

• Many holders were themselves funded short. Off-balance-sheet vehicles bought long-dated securitised assets and financed them with short-dated paper — reproducing the original mismatch one step removed, and frequently with a support commitment back to the sponsoring bank.
• The risk returned to the sponsors. Where a bank had provided liquidity support to a vehicle it had sold assets to, the exposure came back when the vehicle could not fund itself. Risk that was sold contractually was retained economically.
• Diversification within the pool was overstated. Pooling reduces idiosyncratic risk — one borrower defaulting. It does nothing about common-factor risk — all borrowers being exposed to the same thing. When the common factor moved, the pool moved together, and the diversification that justified the structure was not there.
• Complexity obscured position. Tranched claims on pools of claims on pools made it genuinely difficult for a holder to determine what they owned, and impossible for a counterparty to determine what someone else owned.

Securitisation moved risk. It did not reduce it, because the buyers were financed the same way as the sellers. Moving a fragility into a less visible place is not the same as removing it.

The general lesson recurs in this archive. The Private Credit Report 2024 asks whether risk that left the banking system truly left it, or merely changed form and became harder to observe. That is the same question, thirty years of financial innovation later.

The mark-to-market spiral

A feedback loop operated in 2008 that turned a price decline into a mechanism, and it is worth setting out because it explains the speed.

The sequence:

1. An institution needs cash and sells an asset into a thin market.
2. The sale establishes a price — a low one, because it was forced.
3. Every other holder of that asset marks to that price. Accounting requires it.
4. Those marks reduce reported capital and trigger margin calls or breach limits.
5. Those calls force further sales, into a market that is now thinner still.
6. Return to step 2.

Nothing in that loop requires any deterioration in the underlying assets. The loans in the pool may be performing exactly as before. The loop is driven entirely by the interaction of forced selling and mark-to-market accounting.

Why the loop is hard to interrupt from inside: every participant is behaving rationally. Selling before the price falls further is correct for the individual seller. Marking to observed prices is correct accounting and required. Issuing a margin call against a falling collateral value is correct risk management. The aggregate outcome is bad and no single participant is making a mistake — which is the same coordination structure the Global Investment Outlook 2021 describes in a very different market.

The tension the episode exposed, which is genuinely unresolved:

• Mark-to-market is honest. It reports what the asset would fetch, which is the relevant number if you may have to sell.
• In a forced-sale market it is also misleading, because the observed price reflects the seller's urgency rather than the asset's value.
• Suspending it is worse. An accounting regime that lets institutions carry assets at whatever they wish destroys the information that lets counterparties distinguish the sound from the impaired — which is precisely what froze markets in the first place.

There is no clean answer here, and the archive's later reports on appraisal-based valuation in private markets — the Private Equity Report 2015 and the Private Credit Report 2016 — are the same trade-off approached from the opposite direction.

Counterparty uncertainty and the freeze

The most consequential effect of 2008 was not any individual failure but the general refusal to transact, and its cause was informational.

The problem. Losses were distributed across the system in ways that were genuinely unknown. Complex securities, off-balance-sheet vehicles and derivative exposures meant an institution frequently could not determine its own position quickly, let alone anyone else's.

The rational response to that uncertainty is not to price the risk. It is to withdraw. If you cannot distinguish a sound counterparty from an impaired one, and the cost of being wrong is severe, you deal with neither.

The consequences compounded:

• Interbank lending contracted even between institutions that were individually sound, because soundness could not be demonstrated.
• Collateral requirements rose, which consumed liquidity precisely when it was scarce.
• Anyone dependent on wholesale funding was exposed regardless of asset quality.
• The freeze spread to unrelated markets, because institutions raised cash wherever they could rather than where it was economically sensible to sell.

That last effect explains the otherwise puzzling correlation of 2008, where assets with no connection to the underlying problem fell together. They were not being repriced on fundamentals. They were being sold by people who needed cash — a positioning event of the kind the Global Investment Outlook 2018 describes in its purest and largest form.

The remedy was informational as much as financial. Measures that credibly identified which institutions were sound — capital assessments with published results and a backstop for those that fell short — restored transactions in a way that liquidity alone had not. Uncertainty was the binding constraint, and resolving uncertainty was the fix.

What an allocator could act on

Ask how a position is funded, not only what it holds. The 2008 failures were not distinguishable by asset quality. They were distinguishable by liability structure — how short the funding was and how quickly it could leave. That question is answerable from disclosure and is asked far less often than the asset question.

Treat liquidity as a property of conditions, not of an asset. An instrument that trades continuously in normal markets is not therefore liquid; its liquidity depends on a counterparty existing when you need one. The Global Investment Outlook 2020 describes the same discovery in March of that year.

Assume correlations converge in a funding crisis. Assets with no economic connection fall together when holders are selling whatever they can. Diversification computed on fundamental relationships overstates the protection available in exactly the scenario where protection matters.

Distinguish moved risk from reduced risk. When a structure claims to disperse risk, ask who the ultimate holders are and how they are funded. If they are funded the same way as the seller, the fragility has been relocated rather than removed.

Watch the common factor, not the pool size. Pooling removes idiosyncratic risk and does nothing about a shared exposure. A diversified portfolio of assets that all depend on one variable is one position.

Expect the resolution to be abrupt when the mechanism is a run. A credit problem grinds through slowly as losses are absorbed. A funding problem stops the moment a credible backstop exists, because the withdrawal stops being rational. That asymmetry makes the timing of recovery genuinely different from the timing of decline.

What 2008 established

• The funding-versus-credit distinction, and the fact that institutions fail on timing far more often than on solvency arithmetic.
• The duration mismatch as a general mechanism — not a mistake but the fragility inherent in maturity transformation, and one that recurs in 2023 with pristine assets.
• Securitisation was shown to move rather than reduce risk, because the ultimate holders shared the sellers' funding structure.
• The mark-to-market spiral demonstrated that a price decline can become self-propelling with no deterioration in the underlying assets.
• Counterparty uncertainty was shown to bind harder than losses, and resolving the uncertainty was what restored transactions.