Up to 2007 the UK formally had only £2,000 of full deposit insurance (and, for reasons we shall see below, that was almost all the relevant insurance we had).  Deposit insurance was then extended enormously, at the time of the Northern Rock crisis, to cover all deposits made in any bank up to September 2007, later revised down to the first £50,000 and then subsequently revised up to £85,000 (€100,000) under an EU deposit insurance directive.  I argued vigorously against the extension of deposit insurance, along with the bailing out of banks and the sparing of bank bondholders.  I was dismissed, told that no significant bank in a Western economy had imposed losses on bank bondholders since the early 1980s and none on depositors for even longer, and that it was inconceivable that depositors or bondholders should take losses now.

It is simply not true that it is inconceivable to do without government insurance of bank bondholders and depositors.  The UK had no deposit insurance until 1979.  It was introduced only because of an earlier EU deposit insurance directive.  Many other countries have no deposit insurance, including developed countries such as New Zealand (which will have no deposit insurance under the Open Bank Resolution framework).  In Cyprus this week we have seen that depositors of over €100,000 are taking a haircut of 30-40% and bondholders (including senior bondholders) are participating.  The official policy of the Basel Committee on banking, the European Commission and the UK government has been to use "bail-ins" (debt-equity swaps) of bondholders to recapitalise banks in future crises rather than taxpayer recapitalisations.  I have argued that that will be insufficient, as if depositors are not exposed to risk of loss then bondholder loans will be converted into deposits.  Yesterday, Eurozone chairman Jeroen Dijsselbloem said the principle of bail-ins of large depositors (above the deposit insurance threshold) and bondholders would be applied across the Eurozone, according to the Cypriot model (though, comically and without anyone believing him, he later recanted).

That is not precisely how I would do it.  A deposit insurance threshold of €100,000 is extremely high, and has the consequence that larger depositors take much larger haircuts (around twice as big given current deposit structures – presumably more than twice as big in future as large depositors spread their deposits around) than they would if large deposits were not treated as "senior" to smaller deposits (as they are being treated in Cyprus).  If Cypriot deposits had been uninsured and all shared equally in losses, haircuts would have been of the order of 15-20%.  This is typical in developed economy bank failures (it was, for example, the sort of losses depositors experienced in the bank failures in the US in the 1930s – famously the Bank of the United States paid 92.5 cents in the dollar), and it's worth dwelling on a moment.

I have quizzed quite educated people how much they would expect depositors to recover in an uninsured bank failure.  Folk say things like "9%" or "20%", even though typical corporate insolvencies mean losses for creditors of around 50%.  They imagine that bank depositors lose much more, almost all their money, perhaps because of the common language of banks "not having enough money" to cover all withdrawals if they happened at once.  But the "not enough money" is a liquidity point, not a solvency point – functioning banks have more than enough assets to cover deposits.

Let's imagine a bank that had no bonds, but funded itself entirely from equity and deposits (Cypriot banks were almost like this).  Older banks used to carry larger buffers, but a modern bank may only have 3-5% more in assets than its total liabilities (deposits are liabilities).  So a bank will go bust if it makes of order 3-5% losses on its book of loans (the loans it makes are assets).  It's very hard for a bank to make very large losses, absent serious fraud, because if it starts to experience any losses, eating into its buffer, that is liable to trigger a bank run and the self-liquidation of the bank.  Many commentators are very critical of bank runs, but in fact they are an extremely valuable and integral element of banking, providing an excellent protection mechanism for depositors.  Because uninsured banks self-liquidate very easily, they must be run very prudently, as there will be little chance for dodgy deals to be made right later.  They are also rather unlikely to make large losses, as they are most unlikely to be able to continue trading once insolvent (they often self-liquidate even before becoming insolvent at all).

On the other hand, banks may adopted a certain investment strategy that goes bad all at once, despite regulation and the discipline of bank runs.  And regulators may fail.  Let's suppose things havegone really badly wrong, so the bank loses five times as much as would be needed to make it insolvent.  That is, instead of 3% losses, it makes 15%.  Let us also assume that when the bank is liquidated there are significant "costs of bankruptcy".  In a typical business, the costs of bankruptcy are around 20% of value, but that varies from business to business.  In many businesses bankruptcy costs arise because much of the value of the business is connected to intangible assets that are worthless in resale.  For example, a pharmaceuticals company may have invested in a pipeline of research that is simply abandoned in bankruptcy, or there may be investment in a brand that is tarnished by bankruptcy.

But the vast majority of the assets of a bank do not depend on the bank being a going concern.  They are the loans the bank makes, and thus depend on how likely borrowers are to repay.  It is possible that there are some modest losses on selling these to another institution for management and run-off – about 10% is not unreasonable.  So there we have it: the bank loses something like 15%, in a disaster scenario, plus has costs of bankrupcty of around 10%, meaning 25% losses all up on assets.  If there is a 3% equity buffer to begin with, depositors recover 103% minute 25% equals 78% – i.e. the around 80% recovery I mentioned.

I have stepped you through this fairly slowly so you can grasp how limited the exposure bank depositors in developed economies actually have is.  Where bank depositors have lost more than about 20% – as in some Latin American bank failures - that is often because of fraud or absurdly amateurish investment mistakes.  In passing, I note that since bank deposit recovery rates should be expected to be around 80% or more, even absent deposit insurance, when the UK government offered 90% insurance on sums between £2,000 and around £35,000 that was precious little insurance at all.

Many commentators believe that deposit insurance is necessary to prevent bank runs.  That is wrong in two ways.  First of all, it is not desirable to prevent bank runs.  Bank runs are an essential and healthy part of what banks are, provided that they do not spill over into disorder.  If you think a company in which you have shares is going to do badly, you will be able to sell your shares quickly and get out what you can.  Why should you not do the same with a bank if it is going badly?  Second, people do not run on banks mainly because they have a rational fear of losing much of their money.  UK depositors have lost about 20% since 2007, via inflation and depreciation of the pound.  That has not caused them to withdraw money en masse.

Depositors run on banks because they fear losing access to their money.  They keep money in sight deposits in banks precisely because they might need it for liquidity purposes – they might need to be able to spend it quickly, so it's no good if it gets frozen for months or years in a bank resolution.  Hence deposit insurance does not deter bank runs (insofar as it does deter them) because it guarantees the value of capital.  It deters them because it comforts depositors about their liquidity – they see deposit insurance as an implicit promise that the government will keep the bank running, come what may.  And they are right to see things that way, for when there is deposit insurance governments bail out banks, keeping them running.  But of course if governments are known to be going to bail out banks, then banks take more risks and bailouts become more common and larger.  That is why, for example, in New Zealand the authorities have argued against deposit insurance in the new Open Bank Resolution framework, saying:

"The New Zealand Government has looked hard at deposit insurance schemes and concluded that they blunt the incentives for investors and banks to properly manage risks, and may even increase the chance of bank failure…Deposit insurance is widely used in Europe, including Cyprus, but hasn’t prevented banking failures, as we saw during the Global Financial Crisis."

Deposit insurance is unnecessary, does not in itself prevent bank runs (which is in turn an undesirable objective), damages the incentives for depositors to take sound investment decisions when deciding which bank to lend their money to, and makes bank bailouts more likely.  Why would anyone want to be in favour of that?

If you want your money safe, don't lend it to a fractional reserve bank and demand deposit insurance.  Instead, put your money in a savings bank (like National Savings) or a storage deposit (if I ever manage to persuade the authorities to introduce them).  Then savings can be saving and investments investment.