Article ID | Journal | Published Year | Pages | File Type |
---|---|---|---|---|
480781 | European Journal of Operational Research | 2011 | 11 Pages |
Abstract
The subprime crisis has reminded us that effective stress tests should not only combine subjective scenarios with historical data, but also be probabilistic. In this paper, we combine three hypothetical shocks, of varying degrees, with more than six years of daily data on USD-INR and Euro-INR. Our objective is to compare six simulation-based stress models for foreign exchange positions. We find that while volatility-weighted historical simulation is the best model for volatility persistence, jump diffusion based Monte Carlo simulation is better at capturing correlation breakdown. Loss estimates from very fat-tailed distributions are not sensitive to the severity of stress scenarios.
Related Topics
Physical Sciences and Engineering
Computer Science
Computer Science (General)
Authors
Sanjay Basu,