Post by Sapphire Capital on Sept 23, 2009 8:48:36 GMT 4
Hedging Ideally with Realized Covariance
Jin-Huei Yeh
National Central University
Chih-Wei Huang
National Central University at Taiwan - Department of Finance
Chih-Chiang Hsu
National Central University at Taiwan - Department of Economics
July 15, 2009
Abstract:
In view of the recent documented hedging bias attributable to failing to accommodate volatility long memory, we suggest to use the simple, yet superior, realized variancecovariance (RVCOV) in dynamic hedging. For its incremental value from intradaily information, model-free and inherent long memory, RVCOV has been shown to be accurate without misspecification bias and easily generalized to high dimensions in financial applications. Examining the hedging performance for S&P500 spot position with futures contract across different parametric bivariate GARCH models and RVCOV approach shows that all methods can remove over 90% risks, though, RVCOV overwhelmingly dominates all parametric models for its almost perfect hedge. Further reality check using ex-ante one-step-ahead forecasts indicate that the realized optimal hedge ratio based on simple ARMA(1,1) RVCOV forecasts outperforms ideally.
papers.ssrn.com/sol3/Delivery.cfm/SSRN_ID1460584_code976403.pdf?abstractid=1460584&mirid=4
Jin-Huei Yeh
National Central University
Chih-Wei Huang
National Central University at Taiwan - Department of Finance
Chih-Chiang Hsu
National Central University at Taiwan - Department of Economics
July 15, 2009
Abstract:
In view of the recent documented hedging bias attributable to failing to accommodate volatility long memory, we suggest to use the simple, yet superior, realized variancecovariance (RVCOV) in dynamic hedging. For its incremental value from intradaily information, model-free and inherent long memory, RVCOV has been shown to be accurate without misspecification bias and easily generalized to high dimensions in financial applications. Examining the hedging performance for S&P500 spot position with futures contract across different parametric bivariate GARCH models and RVCOV approach shows that all methods can remove over 90% risks, though, RVCOV overwhelmingly dominates all parametric models for its almost perfect hedge. Further reality check using ex-ante one-step-ahead forecasts indicate that the realized optimal hedge ratio based on simple ARMA(1,1) RVCOV forecasts outperforms ideally.
papers.ssrn.com/sol3/Delivery.cfm/SSRN_ID1460584_code976403.pdf?abstractid=1460584&mirid=4