For decades a debate has been simmering over the advisability of deposit insurance. One side produces evidence that insuring deposits makes financial crises more likely.1 These critics of deposit insurance as the first line of defence against bank panics go on to argue that that the central bank, in its role as lender of last resort, can stem bank panics. Countering this is the view that, as a set of hard and fast rules, deposit insurance is more robust than discretionary central bank lending. In my view, the September 2007 bank run experienced by the British mortgage lender Northern Rock settles this debate once and for all – deposit insurance is essential to financial stability.
To understand this conclusion, we need to look carefully at experiences with central bank extensions of credit – discount lending – and at the varying experience with deposit insurance. Let’s start with the lender of last resort.
Lender of last resort
In 1873, Walter Bagehot suggested that, in order to prevent the failure of solvent but illiquid financial institutions, the central bank should lend freely on good collateral at a penalty rate.2 By lending freely, he meant providing liquidity on demand to any bank that asked. Good collateral would ensure that the borrowing bank was in fact solvent, and a high interest rate would penalize the bank for failing to manage its assets sufficiently cautiously. While such a system could work to stem financial contagion, it has a critical flaw. For Bagehot-style lending to work, central bank officials who approve the loan applications must be able to distinguish an illiquid from an insolvent institution. But since there are no operating financial markets and no prices for financial instruments during times of crisis, computing the market value of a bank’s asset is almost impossible. Because a bank will go to the central bank for a direct loan only after exhausting all opportunities to sell its assets and borrow from other banks without collateral, the need to seek a loan from the government draws its solvency into question.3
Deposit insurance operates in a way that contrasts sharply with the lender of last resort. A standard system has an explicit deposit limit that protects the bank’s liability holders – usually small depositors – from loss in the event that the bank fails. Guarantees are financed by an insurance fund that collects premiums from the banks. Logic and experience teach us both that insurers have to be national in scope and backed, implicitly if not explicitly, by the national government treasury’s taxing authority. Funds that are either private or provided by regional governments are simply incapable of credibly guaranteeing the deposits in the entire banking system of a country.
But as I suggested at the outset, deposit insurance has its problems. We know that insurance changes people’s behaviour. Protected depositors have no incentive to monitor their bankers’ behaviour. Knowing this, bankers take on more risk than they would normally, since they get the benefits while the government assumes the costs. In protecting depositors, then, the deposit insurance encourages creates moral hazard – something it has in common with the lender of last resort.
Which is better?
How can we figure out whether the lender of last resort or deposit insurance works better? A physical scientist faced with such a question would run a controlled experiment, drawing inferences from variation in experimental conditions. Monetary and financial policy-makers cannot do this – imagine a statement announcing a policy action beginning something like this: “Having achieved our stabilization objectives, we have decided to run an experiment that will help us with further management of the economic and financial system...”
There is an alternative to irresponsible policy experiments: figuring out which policies are likely to work best requires that we can look at the consequences of differences that occur on their own. Comparing the mid-September 2007 bank run experienced by UK mortgage lender Northern Rock with recent events in the United States provides us with just such a natural experiment.
The US example is typical of how the loss of depositors’ confidence, regardless of its source, can lead to a run. The Abacus Savings Bank serves large numbers of Chinese immigrants in New York, New Jersey and Pennsylvania. In April 2003 news spread through the Chinese-language media that one of the bank’s New York City managers had embezzled more than $1 million. Frightened depositors, unfamiliar with the safeguards in place at US banks, converged on three of the institution’s branches to withdraw their balances. Because Abacus Savings was financially sound, having recently concluded its annual government examination, it was able to meet all requested withdrawals during the course of the day. In the end, as a US Treasury official observed, the real danger was that depositors might be robbed carrying large quantities of cash away from the bank. Leaving their funds in the bank would have been safer. But rumour and a lack of familiarity with government-sponsored deposit insurance – Federal Deposit Insurance insured every depositor up to $100,000 – caused depositors to panic.4
Contrast this with the recent British experience, where deposit insurance covers 100% of the first ₤2000 and 90% of the next ₤33,000, and even then payouts can take months. Under these circumstances, the lender of last resort is an important component of the defense against runs.5
Central banks are extremely wary of taking on any sort of credit risk; in some cases there may be legal prohibitions against it. In lending operations, this translates into caution in the determining the acceptability of collateral. And here is where the problem occurs. In order to carry out their responsibility, central bankers must answer two important questions: (1) Is the borrower solvent? and (2) are the assets being brought as collateral of sufficient value?6
The Northern Rock case brings the weaknesses of this system into stark relief. The broad outlines of the case are as follows. Northern Rock is a mortgage lender that financed its long-term lending with funds raised in short-term money markets. When, starting in mid-August 2007, the commercial paper markets came under stress, Northern Rock started having trouble issuing sufficient liabilities to support the level of assets on its balance sheet.
The natural move at this point was to seek funds from the Bank of England. But lending requires that the answer to the two questions about solvency and collateral quality are both “yes”. Were they for Northern Rock? I have no idea. Some combination of people in the Bank of England and the UK Financial Services Authority may have known, but I wonder. Since Northern Rock is rumored to have had exposure to American sub-prime mortgages, securities for which prices were nearly impossible to come by, it is no exaggeration to suggest that no-one was in a position to accurately evaluate solvency. As for the value of the collateral, again it was likely very difficult to tell.
Problem with last resort lending
So, here’s the problem: discount lending requires discretionary evaluations based on incomplete information during a crisis. Deposit insurance is a set of pre-announced rules. The lesson I take away from this is that if you want to stop bank runs – and I think we all do – rules are better.
This all leads us to thinking more carefully about how to design deposit insurance. Here, we have quite a bit of experience. As is always the case, the details matter and not all schemes are created equal. A successful deposit-insurance system – one that insulates a commercial bank’s retail customers from financial crisis – has a number of essential elements. Prime among them is the ability of supervisors to close preemptively an institution prior to insolvency. This is what, in the United States, is called ‘prompt corrective action,” and it is part of the detailed regulatory and supervisory apparatus that must accompany deposit insurance.
In addition to this, there is a need for quick resolution that leaves depositors unaffected. Furthermore, since deposit insurance is about keeping depositors from withdrawing their balances, there must be a mechanism whereby institutions can be closed in a way that depositors do not notice. At its peak, during the clean-up of the US savings and loan crisis, American authorities were closing depository institutions at a rate in excess of 2 per working day – and they were doing it without any disruption to individuals’ access to their deposit balances.
A well-designed rules-based deposit insurance scheme is the first step
Returning to my conclusion, I will reiterate that the current episode makes clear that a well-designed rules-based deposit insurance scheme should be the first step in protecting the banking system from future financial crises.
1 This is the conclusion reached by Asli Demirgüç-Kunt and Edward Kane in their summary of international research on this issue. See their paper “Deposit insurance around the globe: where does it work?" Journal of Economic Perspectives, Spring 2002, volume 16, no. 2, 175–95.
2 The original source is Walter Bagehot Lombard Streeet: A Description of the Money Market. London: Henry S. Kin & Co., 1873.
3 Another flaw in the Bagehot framework is that banks appear to attach a stigma to discount borrowing. For example, in over one-third of the days between 9 August and 21 November 2007 there were federal funds transactions reported at rates in excess of the discount lending rate. In one case, on 25 October 2007 when the lending rate was set at 5.25 percent, the Federal Reserve Bank of New York reports an intra-day high 15 percent.
4 See James Barron, “Chinatown Bank Endures Run as Fear Trumps Reassurances,” New York Times, 23 April 2003.
5 For an exhaustive description of deposit insurance systems in the EU see Robert A. Eisenbeis and George G. Kaufman, “Cross-Border Banking: Challenges for Deposit Insurance and Financial Stability in the European Union,” Federal Reserve Bank of Atlanta Working Paper 2006-15, October 2006.
6 As an episode twenty years ago demonstrates, the Federal Reserve turns out to have substantial discretion in answering these questions. On 20 November 1985, a software error prevented the Bank of New York from keeping track of its Treasury bond trades. For 90 minutes transactions poured in, and the bank accumulated and paid for U.S. Treasury bonds, notes, and bills. Importantly, BONY promised to make payments without actually having the funds. But when the time came to deliver the securities and collect from the buyers, BONY employees could not tell who the buyers and sellers were, or what quantities and prices they had agreed to—the information had been erased. By the end of the day, the Bank of New York had bought and failed to deliver so many securities that it was committed to paying out $23 billion that it did not have. The Federal Reserve stepped in and made an overnight loan equal to that amount, taking virtually the entire bank – buildings, furniture and all – as collateral. See the discussion in Stephen G. Cecchetti Money, Banking and Financial Markets, 2nd Edition. Boston: McGraw-Hill, Irwin, 2008.