Garman Unveils 'VAR Delta' Methodology

A new buzzword for risk managers was born last week when options theorist Mark Garman unveiled his latest concept: VAR delta.

As its name suggests, VAR delta is a technique for estimating the effect of a small number of trades on the value-at-risk figure associated with a portfolio of instruments.

Garman, who is also president of Californian analytics house Financial Engineering Associates, describes VAR delta as a "technological breakthrough" that will move value-at-risk calculations into the real-time front office arena.

FEA is marketing its new system as the key enhancement in the latest version 1.3 of its VARworks library of C-coded value-at-risk routines.

The latest release of Outlook, FEA's spreadsheet add-in package, will also include the new technique, which has been trademarked by the vendor as VaRdelta.

Value-at-risk measures the worst-case loss in a portfolio's value over a specified time period, given a specified confidence interval.

The technique has been a popular concept in market risk management for some time now, shooting to prominence when JP Morgan launched its RiskMetrics VAR data service (Derivatives Engineering & Technology, October 17, 1994).

Simple appeal

Part of VAR's appeal lies in its simplicity - the bedazzling and complex risks of a diversified financial portfolio are elegantly summarised in a single, intuitively meaningful number.

However, this simplicity comes at a price. VAR figures are time-consuming to produce. As a result, most institutions deploy the technique on their books overnight, to produce daily risk management reports.

"Existing VAR software is usually run on a slow, once-a-day cycle, meaning that VAR does not really provide substantial feedback to the trading process," says an FEA spokesperson.

"By contrast, VaRdelta technology provides for real-time VAR trading limits, or the quick ranking of dozens of candidate trades for risk-improving qualities, thereby yielding virtually instantaneous VAR feedback," the spokesperson adds.

Speeding up VAR

Many top-tier banks and securities houses are toying with the idea of implementing VAR as a real-time process on their trading desks.

Morgan Stanley announced plans to move in this direction earlier this year (RMO, February 12). But Morgan risk officials conceded that full-blown real-time VAR would require immense computing resources - thus making the technique unlikely to appear on traders' desks in the near future.

FEA's approach is to avoid recalculating VAR altogether. Instead, the VAR delta technique is used to estimate the impact of a single trade, or group of trades, on the overall risk figure.

Garman says this incremental VAR assessment method is most accurate when used with relatively large portfolios.

"What you want is that the trade you're analysing is relatively small compared to the book as a whole," he says.

He adds that VAR delta is a first-order sensitivity measure analogous to the delta measure used by options traders.

But Garman cites a problem in applying such approximation techniques to value-at-risk. "VAR can only be computed forwards," he says.

By boiling down a complex portfolio into a single figure, "you lose a lot of information which you can't get back," he adds.

FEA's solution to this is to use an intermediate step in the calculation. First, the portfolio's risks are simplified into what Garman describes as a "cashflow space representation."

VAR delta is then computed as a "vector derivative" of this cashflow representation, using standard vector calculus techniques.

FEA is currently applying for patents for its new methodology, though it intends to "maintain software prices within the moderate range," says Garman.

"FEA's mission has always been to package financial engineering knowledge and provide it as reasonably priced software," he adds.

Beta concept

VAR delta remains as a concept at present - Garman describes last week's revelations as "both a product announcement and a theory announcement."

He adds that he will be liaising closely with FEA clients to see how the first deployments of incremental VAR assessment work out in practice.

But Garman is confident that the new technique will prove to be a practical shortcut to the intricacies of VAR computation. "I don't know why it hasn't been invented before," he says.