The mechanism in the EU bank stress test was set up to fix the result – and it's too early to know if it worked, writes WOLFGANG MÜNCHAU

If you tried to test the safety of cars or children’s toys using the same method the European Union applied in its stress tests on banks, you would end up in jail.

How so?

Simply because the testing mechanism was calibrated to fix the result. The purpose of the exercise was to ensure that the only banks that failed it were those that would have to be restructured anyway.

At the same time, the supposedly clever idea was to demonstrate to the outside world that the rest of the banking system remained sound. The purpose of this cynical exercise was to pretend that the EU was solving a problem, when in fact it was not.

It is too early to judge whether the ploy worked. But from the informed reaction on Friday night, I suspect not. Expectations were not very high. But the EU undershot the lowest of them.

There were three fundamental problems with those tests – and each one would have invalidated them.

The first, and least serious of the three, is that the tests left out some important institutions, whose financial health is not entirely clear. One of those is KfW, the German state-owned institution that is legally not a bank but carries out bank-like functions – such as accumulating lots of toxic assets.

The second problem is the definition of the pass rate – a tier-one ratio of 6 per cent, which refers to various categories of capital as a percentage of a bank’s total assets.

The problem with this definition is that it does not tell us what we need to know. The reason we are interested in capital ratios is not because we are afraid that a bank may fall short of some legal requirement but that it could be insufficiently insured against an exogenous shock.

Tier one capital includes equity and retained earnings but also various types of hybrid debt instruments. For example, government support from Germany and Spain comes in the form of hybrid instruments, whereby the state does not become an owner of the bank.

Hybrid capital has some characteristics of equity but also some characteristics of a bond, including an entitlement to a guaranteed payment stream. As the question is how the system performs under stress, we are interested in the risk-absorbing elements of core capital – not some bureaucratic or legal definition.

The current definition of tier one capital is the reason why all the German Landesbanken have passed the tests. If one had used a narrower definition – equity and retained earnings only – the results would almost surely have been different.

The third problem is the most severe, and knocks the credibility of the entire exercise. There was no provision for the possibility of sovereign default.

Banks hold most of their bonds on their banking books – where they usually keep them until maturity – and a small minority on their trading books. The stress tests assumed some further loss only on the value of those bonds in the trading books. Yet the pricing of Greek bonds already implies a non-trivial probability of a default – and that would affect both books.

Certainly, the stress tests should be based on what one might call a plausible worst-case scenario, not one that represents the absolute worst.

Nobody is asking the bank supervisors to stress-test the impact of an alien attack. But sovereign default in the case of Greece is not such a far-fetched scenario – even if you believe it to be unlikely.

It is irresponsible for the stress testers to ignore that sort of event. That is like a car crash tester failing to consider the possibility of an oncoming vehicle.

In their briefing on Friday evening, officials from the Committee of European Banking Supervisors (CEBS) were reported to have made a strange statement. They calculated that the probability of the adverse stress scenario was 5 per cent.

But how can they know? If this estimate is based on some variant of normal distribution, as I suspect, then the 5 per cent threshold must surely include an assumption of a partial Greek default, as that is the probable inference from current market prices.

Default probabilities are admittedly a very complicated subject. An easy but very dirty method is to take the square root of the spread to some supposedly safe asset.

If the spread for Greece is, say, 900 basis points, that implies a 30 per cent chance of a 30 per cent default. If you really want to include stress scenarios that have only a 5 per cent probability of occurring, surely under current market prices you cannot ignore the possibility of a Greek government default.

The stress tests follow a pattern that has been evident since the outbreak of the acute phase of the financial crisis in September 2008.

The EU’s approach to the financial sector has been to apply patchwork fixes – a blanket bailout, some not very serious recapitalisation plans, plus loads of liquidity – rather than solve the problem.

A notable exception is Spain, where the situation is the most severe, and where a serious attempt is under way to address it.

But while in Madrid the stress tests are part of a political commitment to resolve the banking problems, that is not the case elsewhere.

A stress test without a resolution strategy - which is what is absent beyond Spain - is entirely pointless. – Copyright The Financial Times Limited 2010