金融时间序列 VaR.pdfVIP

Value at Risk Introduction Value at Risk (VaR) is a measure of financial (market) risk It is the amount by which a position could decline in a given period with a specified probability The following factors affect the VaR: The time horizon: ℓ The change in (mark-to-)market value of the position: ∆V (ℓ) The CDF of the change: F (x) ℓ The tail probability: p (0.01 is often used in practice) Data frequency A Formal Definition For a long position: p = Pr{∆V (ℓ) ≤ VaR} = F (VaR) ℓ For a short position: p = Pr{∆V (ℓ) ≥ VaR} = 1 − F (VaR) ℓ Recall that if F (·) is continuous, the pth quantile xp of ℓ F (x) is given by p = F (x ) ℓ ℓ p Thus, the VaR is the pth quantile of F (x) ℓ In general, xp = inf {x : F (x) ≥ p } ℓ Methods The VaR is usually calculated using log returns: VaR = Value × (VaR of log returns) Why use log returns? Simplicity: log returns ≈ percentage changes Alternatively, use VaR = Value × (exp{VaR of log returns} − 1) Methods available (refer to Tsay Chapter 7 for other methods based on extreme value theory): RiskMetrics Econometric modeling Empirical quantile Illustration Data used in illustrations [d-ibmln98.dat in %]: Daily log returns of IBM stock Span: July 3, 1962 to December 31, 1998 (Sample) size: 9190 points Position: long on $10 million Empirical Quantile Sample of log returns: {rt : t =