34 smart data structures the use of smart data

3.4 "Smart" data structures
The use of "smart" data structures can significantly improve the cost-benefit tradeoff in
favor of a rigorous (i.e., full revaluation Monte Carlo) risk measurement system. A "smart" data
structure for storing financial transactions would have some or all of the following characteristics:
1. the financial instrument knows to which market risk factors its value is sensitive;
2. the financial instrument knows both a "more exact" and a "less exact" valuation method for
itself;
3. the financial instrument knows what error is introduced at different times by its different
valuation methods.
Exploiting such a "smart" data structure could significantly reduce the computational burden of
revaluing each instrument in a portfolio N times to calculate a Monte Carlo Value at Risk.
If a financial instrument knows to which market risk factors its value is sensitive, it can
avoid recalculating its value for some of the N Monte Carlo draws. If a DEM/USD currency swap is
asked to revalue itself for several draws in which all market risk factors that would affect its value
(presumably USD and DEM interest rates and the DEM/USD exchange rate) are identical, the value
will be the same for all such draws and the computational burden can be reduced accordingly.
Many derivatives are typically valued using numerical methods. A simple example would be
an American option, which could be priced exactly using a lattice as in Cox, Ross, and
Rubenstein (1979). For risk measurement purposes, if there are many American options in the
portfolio, it may be too time-consuming to value each American option on a lattice. Barone-Adesi and
Whaley (1987) give an approximate analytic valuation method for an American option. A "smart"
American option would know both valuation methods and would know how much error is introduced
by the approximation at different times. This last feature would allow the risk management
application to track how much uncertainty has been added to the estimate of firmwide Value at Risk
by using "less exact" valuation techniques, as well as to set up a threshold for approximation error that
would force the use of a "more exact" technique if a "less exact" technique gave a particularly bad
17
The information systems demands of risk management are driving thirty percent of UK banks to replace legacy
systems, according to a recent survey. "Risk Management is Driving Banks to Replace Legacy Systems," Risk
Management Operations , 16th December, 1996.

128
approximation for a certain set of changes in market risk factors. 18 Our impression is that firms have
begun to adopt "smart" techniques on an ad hoc basis, but their use has not yet become standard
practice.
3.5 Credit risk measurement
There is no single measurement concept for credit risk, unlike Value at Risk for market risk,
which has become widely accepted in the market. The current view of "best practice" for credit risk,
as expressed in the Group of Thirty's (1993) report, is to measure both current exposure and potential