Variance Reduction & Martingale Tests Presentation
Document ID: 2007-371
Published on: 31st May 2007
Author: Steven Morrison, Colin Holmes
ESG Users will be familiar with the idea that valuations carried out using ESG scenarios are subject to sampling error. Sampling error is particularly evident in valuation of simple assets for which we know the “right” answer, in particular when performing martingale (or “1=1”) tests.
Several techniques have been proposed for reducing the size of sampling error, and improving the performance in satisfying validation tests. For example, scenarios can be re-weighted so that martingale tests are satisfied exactly. In this presentation we will review this and other related techniques, demonstrate how they can be implemented and discuss their pros and cons. We will also highlight some potential pitfalls with using these techniques in general, and highlight where they should be used with care.
