The classic bank run looks like this:
Queues of retail customers outside a bank, waiting to withdraw their money. But actually this sort of bank run is very rare. The run on the UK's Northern Rock in 2007 was the only retail run in the financial crisis and the first in the UK for 140 years.
What is more frequent - and far more damaging - is a wholesale run. The Northern Rock retail run was preceded by a wholesale run that forced the Rock to go to the Bank of England for emergency liquidity assistance (ELA). And the wholesale run continued for the next five months despite ELA and government guarantees, only ending when Northern Rock was nationalised.
Wholesale runs occur when investors pull their funds all at once. "Investors" in this case means institutional investors and high net worth individuals. There were two major wholesale runs in the financial crisis, one in 2007 after BNP Paribas forced the closure of the Bear Sterns hedge funds by announcing that it could not value mortgage-backed securities used as collateral, and the other - much larger and potentially devastating - a run on tri-party and bilateral repo when Reserve Primary money market fund "broke the buck" after the fall of Lehman. It was this second run that very nearly brought down the entire financial system and forced central banks around the world to provide liquidity support not only to commercial banks but to investment banks and other "shadow" institutions.
The main risk with a classic retail run is that banks will simply run out of physical cash. The job of the lender of last resort, therefore, is to ensure that they don't. And if necessary, of course, banks experiencing classic retail runs can simply shut their doors. Unfortunately that is far more difficult with a wholesale run. Wholesale runs do not involve withdrawal of funds in physical cash. They typically involve three phases:
Phase 1: electronic transfer of funds from deposit accounts
Phase 2: increasing haircuts on collateral posted against secured borrowing
Phase 3: refusal to roll over short-term (overnight) wholesale borrowings.
These three phases may be concurrent or sequential. But they have different effects.
- Electronic transfers of funds can happen at any time and require funds to settle. If the institution being run upon is a commercial bank, the lender of last resort is responsible for ensuring that funds to settle are available. But if the institutions being run upon are investment banks or other non-banks, the commercial banks through which they settle transactions will need central bank liquidity support. Investment and shadow banks have no direct access to payments systems, so are forced to settle payments through commercial banks - and because of this they have deposit (transaction) accounts with commercial banks. So in a run on investment banking, the volume and value of payment transactions through commercial banks vastly increases, often in a rather unbalanced way, and there may be a high incidence of fails as the demand for funds exceeds the credit available in the transaction accounts. In effect, a run on investment and shadow banks causes a run on the commercial banks that support them. Clearly, the lender of last resort can prevent this run becoming a crisis - but there are limits to central bank support, which can be breached in a major bank run, as I shall explain shortly.
- Increasing haircuts on collateral cause increased demand for cash. The haircut is the amount by which the value of the collateral exceeds the amount lent in, say, a repo. If the value of the collateral remains the same (this is a VERY big "if" in a bank run, as I shall explain shortly), then an increase in the haircut means that the amount lent is reduced. If the loan already exists, or (more likely) is rolled over, that means that the borrowing institution has to find additional cash to plug the gap left by the reduction in the loan amount.
- If lenders refuse to roll over existing lending, or raise the interest rates on short-term borrowing to unaffordable levels, institutions that are dependent on short-term wholesale borrowings suddenly find that they cannot fund their commitments. If they are denied central bank liquidity support, they are forced to sell assets to raise cash. If they cannot do this then they fail in a disorderly manner. This is the most damaging phase of a wholesale bank run. This is what brought down Northern Rock in 2007. This is what brought down Lehman in 2008, followed by many other banks. This is what very nearly killed the entire financial system.
Some people think that because a bank run always takes place via commercial banks even if the institutions actually run upon are shadow banks, it is only commercial banks that need liquidity support: shadow banks can be allowed to fail without commercial banks and their retail customers being adversely affected. That would be true if a bank run only involved deposit account withdrawals - phase 1. But when phases 2 and, especially, 3 occur, that separation doesn't work. A run on shadow banks threatens the solvency of commercial banks. And that is because of the effect of a bank run on the ASSET side of the balance sheet.
Institutions being run upon have to find cash. Lots of it, and fast. If they cannot borrow, then they have to raise money by selling assets. Distressed asset sales by a number of institutions at the same time force down the market price of those assets. So a widespread run on shadow banks that cannot obtain liquidity support from central banks causes asset values to collapse. All asset classes are affected, including "safe" assets used by commercial banks as collateral for central bank funding. This is why the Fed provided liquidity support to shadow banks. It was supporting asset prices. And it continued to do so with QE and other unconventional interventions.
Collapsing asset values are disastrous not only for the financial sector but for the economy as a whole. If there was one thing that Lehman and its aftermath demonstrated, it was that the idea that investment banks can be "allowed to fail" if the result will be asset value collapse is sheer fantasy. Central banks have no choice but to support any financial institution caught up in a major wholesale bank run, whatever its status.
This is not to say that an insolvent investment bank should be bailed out by taxpayers. How we deal with imprudent financial firms that have got into difficulty is an entirely different matter from how we deal with a financial panic. I would be the last person on earth to suggest that governments should recapitalise or otherwise rescue failing investment banks. Indeed I am not keen on governments bailing out any sort of bank: government backstop without control encourages bad management and poor asset quality, which are the two biggest causes of bank failure. But denial of liquidity support to some institutions in a widespread bank run would be disastrous.
To my mind the Fed got it right in the crisis. It threw money at everything, regardless of its status, and by so doing, prevented a catastrophe. But the direction of discussion since, particularly in Europe, has been counterproductive. Cherry-picking which institutions qualify for central bank support is never going to make the financial system more robust. So my recipe would be: provide central bank support to everything and taxpayer support to nothing. And take steps to ensure that investment banks and commercial banks alike maintain good asset quality and defensive balance sheet structures.
Grieving for Glass-Steagall - Coppola Comment
Why do we never learn? - Coppola Comment
Reducing the systemic risk in shadow maturity transformation - Krieger (NY Fed)
Repo runs: evidence from the tri-party repo market - Copeland, Martin & Walker (NY Fed) (pdf)