Presented at QuantMinds International | May 2018
This presentation entitled “AD-on-LSMC for MVA and CVA Greeks: Simplifications and Efficiencies ” was given by Andrew McClelland of Numerix at QuantMinds International in Lisbon on May 17th.
The presentation addressed:
- Pathwise methods for CVA Greeks and future value Greeks for MVA both require differentiation of continuation values over model parameters
- When using LSMC, algorithmic differentiation can be applied in reverse mode (CVA) and forward mode (MVA) for efficiency
- Careful inspection of the regression algorithm reveals specific dependencies, namely those of coefficients on contemporaneous states, which vanish as the sample size grows large
- Ignoring these dependencies dramatically simplifies propagation and yields significant reductions in the computational expense with minimal impact on accuracy
- A set of illustrative examples involving Bermudan swaptions
Numerix Presenter Bio:
Andrew McClelland, Ph.D., Director, Quantitative Research, Numerix
Andrew McClelland's work at Numerix focuses on counterparty credit risk issues including valuation adjustments and counterparty exposure production for structured products. He also works on numerical methods for efficient production of risk profiles under real-world measures.
Dr.McClelland received his Ph.D. in finance at the Queensland University of Technology in financial econometrics. His research involved markets exhibiting crash feedback, option pricing, and parameter estimation using particle filtering methods. His work has been published in the Journal of Banking and Finance, the Journal of Econometrics, and the Journal of Business and Economic Statistics.
Complete the form to the right to download this slide deck from Andrew McClelland's May 2018 QuantMinds International presentation.
