|
Abstract:
The recent growth of derivative securities - options, swaps
and similar instruments - has inspired substantial work on inference
for time-dependent data. However, there is a lack of connection between
the statistical methods and the presumed application, namely, the regulation,
pricing and trading of derivative securities. This activity is mostly
carried out without reference to historical data.
This talk reviews the background for this phenomenon, and explores
recent results on how to incorporate statistical analyses into valuations.
Most typically, a (confidence, credible, or prediction) set is created
on the basis of statistics, and then trading strategies can be found
that remain solvent if the set covers the parameter or the outcome.
This has the advantage of separating the statistical and the financial
engineering aspects of the problem. The approach is particularly useful
for setting regulatory bounds and monitoring mechanisms, and for creating
exit strategies which are less damaging than wholesale liquidation.
It also sets constraints on derivatives prices.
We also discuss downsides and alternatives to this approach.
|
©
2003 Purdue University