The pathways to probability

In the old days of the British diplomatic service, ambassadors could expect a knighthood if the Queen visited their embassy. But there was a condition attached - the Queen's visit had to proceed smoothly.

In the unlikely event that the Queen were to visit the Committee of European Insurance & Occupational Pensions Supervisors (CEIOPS) in Frankfurt, a knighthood would be unlikely for John Tiner, the UK's ambassadorial equivalent at this institution. For the Queen's ears would be ringing with the complaints of CEIOPS chairman Henrik Bjerre-Nielsen.

Last month, Bjerre-Nielsen told a European Insurance Association conference that he had "procedural concerns" about the UK Treasury's increasingly bold interventions in the Solvency II process. "Regulators should not be subject to political instructions," Bjerre-Nielsen stormed, not completely unreasonably. But the Treasury's dogged officials are taking Bjerre-Nielsen's tirades like water off a duck's back. With EU Commissioner Charlie McCreevy keen to chalk up a success for the EU's internal market directorate, they know that once something is on the table in Brussels, it doesn't get taken off again.

At a closed-door meeting between the European Insurance & Occupational Pensions Committee (EIOPC) and the Commission in Brussels on 30 November, finance ministry officials debated calibration of Solvency II. The debate hinges upon the arcane but crucial question of risk measures: should the fundamental risk measure used in the directive be value-at-risk (VaR) or tail VaR. Up to now, one-year VaR at the 99.5% confidence level has been the standard. Put in plain English, it says that companies should hold solvency capital that covers the maximum loss likely to occur in 199 out of 200 years. But VaR is open to criticism.

For many years, experts have complained that it is 'incoherent' - in other words, VaRs from different companies or business units cannot be reliably added together. That not only makes it difficult for regulators to understand the risk of multiline or multinational businesses, but it also makes it difficult for chief risk officers to allocate capital.

Moreover, the second quantitative impact study (QIS2) highlighted the onerous capital charges for asset classes such as equities if insurers were forced to hold capital to cover stock market swings over a one-year horizon. Tail VaR attempts to make up for these flaws.

A tail VaR target would require companies to hold sufficient capital to cover the expected loss, given that a rare event occurred. This sounds like VaR but it isn't. Put simply, VaR says where the extreme risk zone begins, but tail VaR says what happens once you are inside the zone.

The critical question for the EIOPC is how you move from one measure to the other. Clearly, moving from the boundary of a zone to its middle means that the boundary can be set at less extreme levels. It is often suggested that 99% tail VaR is equivalent to 99.5% VaR. But research in Switzerland - a leader in these matters - suggests that 99.6% is more accurate.

Getting the answer right will not be easy. Moving from VaR to tail VaR involves more modelling of rare events and their joint dependence. And the issue is also proving critical to UK occupational pensions (currently ringfenced from Solvency II) where VaR and tail VaR have become relevant to the takeover of steelmaker Corus. Indian magnate Ratan Tata's intentions could play a big role in this balance of probabilities, and perhaps one day might even bring him an honorary knighthood.