Economist questions banks’ approach to VAR analysis

Bank risk managers must use more intra-day price data – also referred to as high-frequency data – to improve their value-at-risk analyses, according to Richard Olsen, an economist and founder of Zurich-based hedge fund and risk services company Olsen.

The variation in the price of different assets throughout the day can vary by 70%, claimed Olsen, much greater than the 10% to 15% he said he and his colleagues expected when they performed tests to help market high-frequency data services several years ago.

To see the effect of intra-day asset price and volatility variations on risk analysis, Olsen suggested risk managers run two VAR tests within a day on a static portfolio. He said they would note a significant difference in values.

Olsen said the importance of using intra-day data for risk analysis was made clear in 1997 with the publication of papers written by Tim Bollerslev, the inventor of the general autoregressive conditional heteroscedasticity (Garch) model. The Garch model has became widely used in financial risk analysis since its introduction in the late 1980s because of its recognition that financial asset return distributions are non-normal and exhibit volatility that is not constant through time. According to Olsen, the work by Bollerslev showed improvements in Garch model accuracy of 30% through use of high-frequency data.

Despite Olsen’s work in recent years on high-frequency finance and the renown of concerned finance academics such as Bollerslev, Olsen said the majority of dozens of banks he has consulted over the years have done little to integrate high-frequency data with risk management. This is due in part to the perception among bank staff of exorbitantly high costs for migrating existing bank risk systems to incorporate high-frequency data, said Olsen. He added that this perception was wrong and says securing appropriate data services for a bank’s most important instruments and re-engineering bank risk engines costs between $500,000 and $1 million.

Olsen will be speaking on the need for high-frequency data at banks at a conference on data standards hosted by the Bank for International Settlements in two weeks.