Title: Week 10: VaR and GARCH model
1Week 10 VaR and GARCH model
2Estimation of VaR with Pareto tail
- Disadvantages (parametric and nonparametric
method) - For parametric method the assumption of normal
distribution is always not true. - b. For nonparametric estimation, it is usually
possible only for large a, but not for small a.
So, one may expect to resort to - use nonparametric regression for large ato
estimate small one.
3Extreme Value Theory (EVT)
- Assume rt i.i.d with distribution F(x), then
the CDF of r(1) , denoted by Fn,1(x) is given by
- Fn,1(x)1-1-F(x)n.
- In practice F(x) is unknown and then Fn,1(x)
is unknown. The EVT is concerned with finding two
sequence an) and bn such that - (r(1) an)/ bn
- converges to a non-degenerated distribution as
n goes to infinity. Under the independent
assumption, the limit distribution is given by - F(x)1-exp-(1kx)1/k, when k?0
- and 1-exp-exp(x), when k0.
4Three types for the above EVT distribution
- Type I k0, the Gumbel family with CDF F(x)
1-exp-exp(x), x?R. - b. Type II klt0, the Frechet family with CDF
- F(x) 1-exp-(1kx)1/k, if xlt-1/k and
- 1,
otherwise. - c. Type III kgt0, the Weibull family with CDF
- F(x) 1-exp-(1kx)1/k, if xgt-1/k and
- 1,
otherwise.
5Pareto tails
- In risk management, we are mainly interested in
the Frechet family, which includes stable and
student-t distribution. - We know that when rt i.i.d with tail
distribution P(r1gtx)x-ßL(x), i.e., rt has
Pareto tails, then Xt converges to a stable
distribution with tail index ß. Here the tail
index ß is always unknown. To evaluate the VaR of
small a, we estimate the tail index ß first and
apply the nonparametric method to draw the value
of VaR for largea0, and then use the VaR(a0) to
estimate VaR(a) . (how?) This is the so-called
semi-parametric method.
6How to estimate the tail index ß?
- MLE when the whole distribution of rt is known.
- Linear regression suppose rt have a Pareto
left tail, for xgt0, P(r1 lt-x) x-ßL(x), then - log(k/n)log P(r1 ltr(k) ) -ßlog (-r(k))
log L(-r(k)). - Hill estimator based on MLE, one can get that
- n/ß log(r1/c) log(r2/c). log(rn/c), when
- P(r1 lt-x) 1-(x/c)-ß, xgtc. When one only uses
the data in the tail to compute the tail index,
then - ß n(c)/?rigtc log(ri/c)
- and let H1/ß, then
-
7The properties of Hill estimator
8VaR for a derivative
- Suppose that instead of a stock, one owns a
derivative whose value depends on the stock. One
can estimate a VaR for this derivative by - VaR for derivative LX VaR for asset,
- where L(Delta pt-1asset/ pt-1option),
- and Deltad C(s, T,t, K, s,r)/dS.
9Volatility modeling
- Volatility is important in options trading for
example the price of a European call option, the
well known Black-Scholes option pricing formula
states that the price is C(S0) S0?(d1) Ke rT
?(d2), - where
- d1 (??T) 1log (S0/K) (r?2/2)T,
- d2 (??T) 1log (S0/K) (r?2/2)T d1
??T.
10- In VaR, let Rt be the daily asset log-return and
St be the daily closing price, then - Rt1log(St1/ St) .
- Suppose the return is normal distributed with
mean zero, then one can write it as
11- The variance as measure by square return, exhibit
strong autocorrelation, so that if the recent
period was one of high variance, then tomorrow is
tend to have high variance. To capture this
phenomenon, the easiest way is to use
12(No Transcript)
13(No Transcript)
14(No Transcript)
15The advantages of Riskmetrics
- It is reasonable from the observed return that
recent returns matter tomorrows variance than
distance returns. - It is simple only one parameter is contained in
the model. - Relative little data need to be stored to
calculate tomorrows variance.
16Shortcoming of Riskmetrics
- It ignores the fact that the long-run average
variance tends to relative stable over time.
17GARCH model
18The unconditional kurtosis of ARCH(1)
- Suppose the innovations are normal, then
E(at4 Ft-1)3 E(at2 Ft-1) 2 - 3(a0a1at-12 )2,
- it follows that
- Eat4 3a0 2 ( 1 a1 ) / ( 1 -a1 ) ( 1 -3a1
2 ) - and
- Eat4 /(Eat2 )2 3 ( 1 -a1 2 )/ ( 1 -3a1 2 )gt3.
- This shows that the tail distribution of at is
heavier than that of a normal distribution.
19(No Transcript)
20How to build an ARCH model
- Build an econometric model (e.g. an ARMA model)
for the return series to remove any linear
dependence in data abd use the residual series of
the model to test for ARCH effects - Specify the ARCH order and perform estimation
AIC(p)log(s)p2)2p/n - Model checking Ljung-Box statistics.
- Qn(n2) ?k1 h (?) k2/(n-k).
21GARCH(1, 1) model
22Note that s2?/(1-a-ß)
23(No Transcript)
24(No Transcript)
25The forecast of variance of k-day cumulative
return
26If the returns have zero autocorrelation, then
the variance of the K-day returns is
- For RiskMetrics model, it is just Kst12. But for
a GARCH model, we have
27If the returns have zero autocorrelation and
st1lts, then Var forecast of GARCHgt RM
28GARCH(p, q)
- A process Rt is called a GARCH(p, q) model if
Rtstet, where
29Between GARCH and ARMA model
- Let etRt2-st2, then Rt2 follows an ARMA models.
This can be seen by - Rt2? ?i1 maxp, q ( aißi) Rt-i2et-
- ?j1 qßj et-j.
- This also explains why simple GARCH models,
such as GARCH(1, 1) may provide a parsimonious
representation for some complex autodependence
structure of Rt2.
30Some properties
- Theorem A The necessary and sufficient condition
for a GARCH(p, q) being a unique strictly
stationary process with finite variance is ?j1
paj ?j1 qßj lt1. Further, ERt0, Cov(Rt, Rt-k)0
for kgt0 and - Var(Rt)?/(1- ?j1 paj ?j1 qßj ).
- In addition, if
- E(et4)1/2 ?j1 paj /(1- ?j1 qßj )lt1,
- then Rt has fourth moment.
31- Theorem B Under ?j1 paj ?j1 qßj lt1, Rt2
is a causal and invertible ARMA(maxp, q, q)
process and exhibits heavier tails than those of
et in the sense of kurtosis.
32Some related model
- GARCH-M model when the conditional standard
deviation is a regressive variable, we called
this model as GARCH-in-mean (GARCH-M) model,
i.e., - Yt aXt b st at,
- Where at is a GARCH model.
- For example, when Y is a return, it may depends
on the variability, higher variability will lead
to higher returns.
33- The leverage effect model negative return
increases variance by more than a positive return
of the same magnitude. - Model A let It1, if day ts return is negative
and zero, otherwise and define
34Model B E-GARCH model
35Weekend effect
- It is always known that days that followed a
weekend or a holiday have higher variance than
average day. We can try the following model - st12?ßst2ast2 Zt2?ITt1,
- where ITt1 takes value 1 if day t1 is a
Monday, for example.
36More general EGARCH
37IGARCH model
- A GARCH(p, q) process is called an I-GARCH
process if -
38(No Transcript)
39How to estimate the parameters in a GARCH model
40- Quasi-Maximum Likelihood Estimation (QMLE)
41Whittles estimator
- By Theorem B, we see that Rt2 can be written as
- Rt2 c0?j cj Rt-j 2 et, where cjgt0.
- If Var(et) is finite, then the spectral density
of the process Rt2 is - g(?) Var(et) 1- ?j cj exp(ij ?)-2 /2p.
- And the Whittles estimator is given by
minimizing - ?j1T-1 IT(?j)/ g(?j), where
- IT(.) is the periodogram of Rt2 , ?j2jp/ T.