The sub-prime crisis in the US has got everyone wondering about its implications for the global financial system. And, more than anyone else, central banks have been getting sweaty palms over what could happen to their charges, the banks.

What's worrying them is an age-old problem: If Bank A owes money to Bank B and is unable to repay and, as a result, Bank B gets into trouble as well, and if it owes money to Bank C, what will happen to banks C-Z?

That is, will there be what is euphemistically called "contiguous default" or more pithily, contagion? Will the bystander banks also take huge hits, jeopardising that holiest of holy grails for central banks -- the stability of the financial system?

The answer depends crucially on whether the default happens for completely idiosyncratic reasons to single banks or for 'better' systemic reasons that affects all banks. So central banks have been using counterfactual simulations -- rather as the military and the intelligence agencies do -- to get a set of alternative possibilities. They started research on this about 10 years ago, but don't appear very much better informed for it.

A recent paper by Christian Upper of the Bank of International Settlements examines those simulations. "On the whole," he says, "such simulations suggest that contagious defaults are unlikely, but cannot be fully ruled out, at least in some countries." Not very comforting to be told that it could happen, and it could happen to you.

The main problem, says Upper, is that these simulations are all based on very strong assumptions. Thus, they can help predict whether or not there will be contagion, but they tell us practically nothing about the ability of the systems to withstand massive stress. Nor do they tell us anything about what policy should do in a crisis "primarily due to their lack of behavioural foundations."

In short, unless you know how people -- ordinary ones, bankers, politicians, businessmen and so on -- will react, you really don't know what will happen. Even if one rules out idiosyncratic decision-making, you can never really predict what will happen unless you get the behavioural thing right. So, "while the models have improved considerably�there is still a long way to go until they become an integral part of the toolbox of any authority responsible for financial stability."

That said, it is also a fact that governments don't any longer let banks go under. Bailouts have become the order of the day and the bigger the bank, the more likely is a government and the central bank to step in.

Besides, ownership matters. For example, in India where 90 per cent of the banking system is still government-owned, a bank failure of the sort being discussed in this paper is almost impossible to envision. The taxpayer is always there to lend a helping hand. One government bank simply takes over another failing government bank.

But what will happen when the banking system becomes dominated by privately owned banks? Once again, there should not be a problem unless a whole lot of banks go under at the same time. But as long as the problem of contagion is contained and confined to three or four banks, one would imagine that any government worth its salt would be able to effect a rescue.

The real problem, therefore, at least in my view, is not contagion but moral hazard. The fact that Old Moneybags is standing by to help out, the temptation to take unwarranted risks will persist and, in certain situations, even increase.

Also, if one may be so rude as to ask: could central banks also suffer from moral hazard? Are they part of the problem or the solution?

BIS Working Paper No 234: Using counterfactual simulations to assess the danger of contagion in inter-bank markets, August 2007, http://www.bis.org/publ/work234.htm