Multi-period Value-at-Risk Forecasting

Please enable javascript in your browser.

Rezaei, Mohammad Hosein | 2013

491 Viewed

Type of Document: M.Sc. Thesis
Language: Farsi
Document No: 44028 (44)
University: Sharif University of Technology
Department: Management and Economics
Advisor(s): Barakchian, Mahdi
Abstract:
Multi-period Value-at-Risk (VaR) forecasting is important for risk management. Financial institutions especially commercial banks according to Basel Committee regulations for capital adequacy in “Internal Approach”, should forecast their VaR for multi-period horizons (longer than one day). The most conventional approach for forecasting multi-period VaR is scaling one-day forecasted VaR that is called “square root of time rule”; therefore most works in this area have been focused on forecasting one-day VaR. In this study, we review performance of a wide range of different methods in forecasting multi-period VaR. The results of our study show that historical simulation performs weakly in accounting for the shocks and volatility of returns for long periods. Instead, the models based on MIDAS method, both parametric MIDAS and non-parametric MIDAS (QR-MIDAS) rapidly adjust the VaR with new information in the market and account for big shocks including the recent crisis 2007. The results also show that the square root of time rule mostly underestimates VaR and is associated with considerable losses
Keywords:
Backtesting ; Multiperiod Value-at-Risk Forecasting ; MIDAS Regression ; Quantile Regression ; Multiperiod Volatility Forecasting

No TOC