Christoffer Ramsden presents his master's thesis

Capturing nonlinearities of financial assets using interpolation methods
in risk calculations
	
Abstract

This Master's Thesis studies how to perform efficient and accurate
interpolation of prices for nonlinear financial assets when data is
limited and data transfer expensive. Assets are priced using a separate
derivatives pricing system and transferred to the risk system on a grid.
Risk calculations are made using simulated scenarios and an asset price
is needed for each scenario. A full pricing scheme such as Monte Carlo
might need tens of thousands of paths to calculate one price. With tens
of thousands of scenarios in the risk calculations this would not be
applicable. Here interpolation, also in this setting referred to
as re-pricing, provides a cheap and effective alternative.
Several interpolation methods are empirically tested on a range of
derivative securities with different properties. The analysis is also
done on an aggregated level in order to get a picture of how
interpolation errors effect risk numbers such as Value at Risk and
Expected Shortfall for portfolios. Finally, suggestions on how to place
grid points are given by implementing grid construction algorithms that
produce an adaptive grid depending on the properties of the security
repriced.