Professional Documents
Culture Documents
Volatility
1 N 2 1/ 2
RW: = [ ei ]
N i 1
Daily returns. Annualized percentage: (T/t)1/2100% T = 252.
Moving window: t =
1
6. Volatility
2
SPY returns in 2009 - 2013
3
6. Volatility
GARCH(1, 1):
2(t) = + a2(t-1) + b2(t - 1) = /(1 b) +
a(2(t - 1) + b2(t - 2) + b22(t - 3) + ... ) => ARCH()
4
6. Volatility
6.2 Conditional heteroskedasticity (continued)
Unconditional expectation for GARCH(1,1): E[2(t)] = /(1 a - b)
Forecast for GARCH(1,1):
E[2(t + k)] = (a + b)k[2(t) - /(1 a - b)] + /(1 a - b)
2(t) = (1 ) i - 1 2(t - i)
i 1
, =1b
Negative price shocks often influence volatility more than the positive
shocks => exponential GARCH (EGARCH):
log [2(t)] = + log[2(t-1)] + z(t - 1) + (|z(t 1)| - 2/ )
where z(t) = (t)/(t), > 0, and < 0.
5
6. Volatility
6.3 Realized volatility (RV)
Intraday dynamics => Brownian motion: dX = (t)dt + (t)dW
T
(t )dt
2
Integrated volatility: IV =
0
N 1
RV = ( X (t ) X (t
i 1
i i 1 )) 2 t0 = 0 < t1 < < tN-1 = T
Pr( Z z) = exp[-z2/2] dz = 1 -
7
6. Volatility
8
6. Volatility
While VaR is an estimate of loss within given confidence level, ETL is an estimate of loss within
the remaining tail:
ETL = E[L | L > VaR]
For a sample of 100 P/L values and the confidence level of 95%, VaR is the sixth smallest
number in the sample while ETL is the average of the five smallest numbers within the sample.
9
6. Volatility
10