Various new methodologies attempt to make liquidity risk transparent by putting a number against it - for example, paper 2000-E-14 at this link, which also reviews the recent literature. These new methodologies try to mend some of the simplifying assumptions of the standard value-at-risk approach to market risk management. VAR assumes that the liquidation of a position will not itself have an impact on the market, that positions can be liquidated in a short period of time, and that the bid-offer spread will remain steady.

But, with a few exceptions, the new methodologies say more about the liquidity risk of exiting large positions in normal markets, and the problem of optimising exit strategies in such markets, than about the liquidity risks of abnormal markets. The lesson of 1998, when market turbulence helped bring down Long-Term Capital Management, was that the liquidity that matters is the kind - almost impossible to measure - that persists in a market when it is under extreme stress.

Since 1998, the threat that's really scared banks and their regulators is that of a simultaneous increase in illiquidity and market price volatility that feeds credit concerns and leads to a breakdown in the normal relationships between markets. As various reports and background paperson the problem have attested over the last couple of years, no single measure can capture this kind of stress liquidity risk. Aside from market depth, tightness and resilience, market liquidity in a crisis depends on the nature and diversity of market participants, their reasons for entering the market, and the reaction of their credit departments to deteriorating conditions.

A BIS paper on stress-testing practices in large financial institutions pointed out in March 2000 that one way to anticipate market fragilities like this might be to aggregate stress testing results from leading market participants. But the report also noted the many practical challenges of attempting to do this, and said that the information would not help banks allocate risk capital as it would not indicate how likely such severe situations were.

It's proved difficult to obtain good data on the historical cost of liquidity deficits during market because the data that's most interesting - panic-stricken sales at low prices on the over-the-counter markets - is the most difficult to track down.

Debate continues on whether VAR itself might make liquidity crises in stressed markets worse. In a recent prize-winning essay, Avinash Persaud of State Street bank claimed that quantitative market risk models, such as value at risk, tell institutions to exit from volatile markets at the same time, and so exacerbate market shocks - though not everyone in the market risk community agrees.

Meanwhile, the long-term drift in the banking industry away from funding based on customer core deposits, and towards money market funding sources of various kinds, means that liquidity risks of various kinds are on an upward trend. This, along with the shifting liquidity structure of some critical financial markets, such as the government debt market in the US, means there's only one certainty. The market liquidity crises of the early years of the 21st millennium - and they will happen - won't be a re-run of 1998, or of any other historical example.