Take the insurance fund against future bank failures proposed by Germany and France. A standard criticism, as with the more straightforward provision of liquidity or capital support, is that this will encourage bankers to go beserk. Yet few bankers go to work and prepare to take big risks on the basis that if they pull it off they make a killing and if they fail, they will be bailed out.

The central bankers' traditional view is that the chief executive of any bank subject to a bail-out should be ignominiously despatched. In the current crisis, that message has been muddied. Many have been fired, but some remain in place. Even so, few chief executives will want to risk not only losing a job, but forfeiting a hard-won reputation along with the value in any equity and stock options they may hold.

The reason why some of them did bet the bank was that they made errors of judgment about the nature and scale of the risks they were taking in a very competitive environment. Capital market pressure and poorly designed incentive schemes were pushing them towards grabbing short-term profits, while taking on tail risk - low probability events that are catastrophic when they happen.

Did Dick Fuld at Lehman Brothers know he was betting the bank until it was manifestly crumbling? I doubt it. And it seems entirely plausible that executives at Lehman were complacent about the capital position because, as so often happens with creative accounting, they were fooled by their own trickery with the numbers.

The way moral hazard chiefly works is by reducing the discipline that creditors bring to bear. Because so many holders of bank liabilities expect to be bailed out, they do not do enough homework on risk. That lack of discipline is then compounded by excessive reliance on all too fallible rating agencies. And because creditors underprice the risk taking of institutions that are too big or too interconnected to fail, systemically important outfits enjoy a lower cost of funds than they should.

For bank management, that last point creates a morally hazardous incentive to increase the size of the bank until it is safely past the threshold of systemic importance. Not only do markets recognise the value to banks of this status. Credit rating agencies take into account the likelihood of bail-outs in their judgements about default risk.

Pathetically inadequate

In a recent paper, academics Elijah Brewer and Julapa Jagtiani estimated that banks were willing to pay an additional premium for mergers that put them over the generally acknowledged too-big-to-fail threshold of $100bn. For eight merger deals of this kind between 1991-2004 they estimate that at least $14bn in added premiums were paid to do that.

As Andy Haldane of the Bank of England pointed out in a compelling speech last week, there is evidence of diseconomies of scale above the $100bn mark, as well as of diseconomies of scope in banking. He argues that the maximum efficient scale for banks could be relatively modest, which would support the case for regulatory action to shrink the 145 banks around the world with assets over $100bn in 2008.

To return to the Franco-German insurance fund, Mr Haldane's thoughts are equally germane. He estimates the present value of output losses from the financial crisis at between $60,000bn (£39,450bn) and $200,000bn for the world economy. A systemic levy to recoup these costs if a crisis occurred every 20 years would exceed $1,500bn a year.

The proposed €1.2bn (£1.1bn) annual levy for German banks thus looks pathetically inadequate. Mr Haldane concludes that fully internalising the output costs of financial crises would risk putting banks on the same trajectory as the dinosaurs, with the levy playing the role of the meteorite.