Sie sind auf Seite 1von 3

Threshold GARCH Model

The parameterization of Threshold GARCH (TGARCH) model in Splus is as follows:


m m s
t2 = 0 + i a2ti + i Sti a2ti + 2
X X X
i ti , (1)
i=1 i=1 i=1

where (
1 if ati < 0
Sti =
0 if ati 0.
From the model, depending on whether ati is above or below the threshold value of zero,
a2ti can have different effects on the conditional variance t2 . When ati is negative, the
(conditional) impact of a2ti on t2 is i a2ti ; when ati is negative, the impact is (i + i )a2ti .
One would expect that i to be positive so that bad news shown in negative ati to have larger
impacts on the volatility of the returns. In practice, threshold values different from zero can
be used as one would expect that only large shocks attract investors attention.
Example. Consider the daily log returns of Yhoo stock from 1997 to 2003.

> fit.yhoo=garch(yhoo~1,~tgarch(1,1),cond.dist=ged)

> summary(fit.yhoo)

Call:
garch(formula.mean = yhoo ~ 1, formula.var = ~ tgarch(1, 1),
cond.dist = "ged")

Mean Equation: yhoo ~ 1

Conditional Variance Equation: ~ tgarch(1, 1)

Conditional Distribution: ged


with estimated parameter 1.420862 and standard error 0.05603733
--------------------------------------------------------------
Estimated Coefficients:
--------------------------------------------------------------
Value Std.Error t value Pr(>|t|)
C 1.224e-03 9.823e-04 1.246 1.065e-01
A 3.882e-05 1.464e-05 2.652 4.034e-03
ARCH(1) 4.959e-02 1.185e-02 4.184 1.502e-05
GARCH(1) 9.101e-01 1.477e-02 61.634 0.000e+00
GAMMA(1) 5.752e-02 2.015e-02 2.854 2.182e-03
--------------------------------------------------------------
AIC(6) = -5772.658

1
BIC(6) = -5739.816

Normality Test:
--------------------------------------------------------------
Jarque-Bera P-value Shapiro-Wilk P-value
300.2 0 0.9799 1.112e-09

Ljung-Box test for standardized residuals:


--------------------------------------------------------------
Statistic P-value Chi^2-d.f.
16.86 0.1549 12

Ljung-Box test for squared standardized residuals:


--------------------------------------------------------------
Statistic P-value Chi^2-d.f.
9.076 0.6964 12

Lagrange multiplier test:


--------------------------------------------------------------
Lag 1 Lag 2 Lag 3 Lag 4 Lag 5 Lag 6 Lag 7 Lag 8 Lag 9 Lag 10
-0.4595 -0.2222 0.306 -0.1319 0.2746 -1.029 -1.506 -0.6847 1.303 -0.9179

Lag 11 Lag 12 C
-1.034 -0.1627 1.252

TR^2 P-value F-stat P-value


9.686 0.6435 0.8855 0.6664

Comparison with EGARCH model

> fit.yhoo=garch(yhoo~1,~egarch(1,1),leverage=T,cond.dist=ged)

> summary(fit.yhoo)

Call:
garch(formula.mean = yhoo ~ 1, formula.var = ~ egarch(1, 1), leverage = T,
cond.dist = "ged")

Mean Equation: yhoo ~ 1

Conditional Variance Equation: ~ egarch(1, 1)

Conditional Distribution: ged


with estimated parameter 1.445849 and standard error 0.05991048

2
--------------------------------------------------------------
Estimated Coefficients:
--------------------------------------------------------------
Value Std.Error t value Pr(>|t|)
C 0.0012 0.001004 1.195 1.161e-01
A -0.4108 0.089946 -4.567 2.650e-06
ARCH(1) 0.2037 0.030807 6.613 2.496e-11
GARCH(1) 0.9583 0.012394 77.320 0.000e+00
LEV(1) -0.2430 0.078605 -3.091 1.014e-03
--------------------------------------------------------------
AIC(6) = -5779.968
BIC(6) = -5747.126

Normality Test:
--------------------------------------------------------------
Jarque-Bera P-value Shapiro-Wilk P-value
217.1 0 0.9817 2.928e-06

Ljung-Box test for standardized residuals:


--------------------------------------------------------------
Statistic P-value Chi^2-d.f.
15.37 0.2217 12

Ljung-Box test for squared standardized residuals:


--------------------------------------------------------------
Statistic P-value Chi^2-d.f.
7.736 0.8054 12

Lagrange multiplier test:


--------------------------------------------------------------
Lag 1 Lag 2 Lag 3 Lag 4 Lag 5 Lag 6 Lag 7 Lag 8 Lag 9 Lag 10
-0.7664 -0.2433 0.2679 -0.1959 0.5049 -1.004 -1.468 -0.599 1.29 -0.5052

Lag 11 Lag 12 C
-1.032 -0.1693 0.7717

TR^2 P-value F-stat P-value


8.42 0.7515 0.7691 0.7821

Das könnte Ihnen auch gefallen