Applications of state space models in finance
Applications of state space models in finance
Applications of state space models in finance
Create successful ePaper yourself
Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.
6.2 Model<strong>in</strong>g conditional betas 91<br />
Regard<strong>in</strong>g asymmetric effects, symmetric GARCH <strong>models</strong> with t-distributed <strong>in</strong>novations<br />
are superior to the alternative GJR-counterparts. These results are <strong>in</strong> l<strong>in</strong>e with Bollerslev<br />
et al. (1992, ¢ 3.3). They argue that leverage effects are <strong>of</strong>ten attributable to a few<br />
outliers, which can be better captured by employ<strong>in</strong>g a fat-tailed distribution than by<br />
explicitly allow<strong>in</strong>g for asymmetric effects. The <strong>in</strong>clusion <strong>of</strong> a constant term generally<br />
improves the values <strong>of</strong> the correspond<strong>in</strong>g log-likelihood without lead<strong>in</strong>g to higher BIC.<br />
These results suggest to model ARCH-based conditional betas by a bivariate GARCH<br />
framework, <strong>in</strong> which univariate t-GARCH(1,1) <strong>models</strong> with nonzero constants <strong>in</strong> the<br />
mean equation are fitted to the excess returns <strong>of</strong> each sector and the overall market.<br />
The estimation results are summarized <strong>in</strong> Table 6.4. The correspond<strong>in</strong>g beta series are<br />
denoted as β tG<br />
i,t .<br />
Table 6.4: Parameter estimates for t-GARCH(1,1) <strong>models</strong>.<br />
This table reports the estimated parameters for the t-GARCH(1,1) <strong>models</strong> for the eighteen<br />
DJ STOXX sectors and the DJ Stoxx Broad as market <strong>in</strong>dex; *** means significance at the<br />
1% level (**: 5%, *: 10%).<br />
Sector µ × 10 2 ω × 10 4<br />
γ δ DF a<br />
JB b<br />
Q(12) c<br />
Broad 0.25 *** 0.16 ** 0.13 *** 0.84 *** 9.29 *** 127.12 *** 16.96<br />
Automobiles 0.17* 0.25** 0.10*** 0.88*** 7.03*** 223.82*** 17.46<br />
Banks 0.26*** 0.12** 0.12*** 0.87*** 7.88*** 245.82*** 13.14<br />
Basics 0.22*** 0.05 0.06*** 0.94*** 6.63*** 580.94*** 22.99**<br />
Chemicals 0.22*** 0.27** 0.14*** 0.82*** 9.48*** 44.56*** 9.70<br />
Construction 0.23*** 0.24** 0.11*** 0.85*** 8.26*** 279.77*** 18.71*<br />
F<strong>in</strong>ancials 0.25*** 0.16* 0.14*** 0.84*** 6.27*** 485.41*** 37.67***<br />
Food 0.20*** 0.23** 0.11*** 0.84*** 7.79*** 144.74*** 12.76<br />
Healthcare 0.27*** 0.23* 0.10*** 0.86*** 10.54*** 27.956*** 13.70<br />
Industrials 0.26*** 0.16** 0.14*** 0.84*** 8.26*** 139.28*** 34.59***<br />
Insurance 0.18** 0.14** 0.09*** 0.90*** 6.01*** 648.31*** 14.16<br />
Media 0.23*** 0.18** 0.10*** 0.88*** 7.76*** 88.42*** 10.92<br />
Oil & Gas 0.22*** 0.20 0.08*** 0.89*** 9.43*** 29.39*** 9.38<br />
Personal 0.23*** 0.51** 0.16*** 0.76*** 12.10*** 69.55*** 18.74*<br />
Retail 0.24*** 0.67** 0.14*** 0.78*** 6.17*** 14986.00*** 10.97<br />
Technology 0.29*** 0.14** 0.11*** 0.89*** 8.70*** 137.26*** 17.75<br />
Telecom 0.29*** 0.14* 0.09*** 0.90*** 15.42*** 8.94** 19.20*<br />
Travel 0.17** 0.54** 0.15*** 0.75*** 8.43*** 90.72*** 6.26<br />
Utilities 0.24*** 0.38** 0.15*** 0.76*** 7.76*** 90.24*** 19.19*<br />
a DF denotes the number <strong>of</strong> degrees <strong>of</strong> freedom <strong>of</strong> the Student-t distribution, which has<br />
been estimated along with the other parameters <strong>of</strong> the t-GARCH(1,1) <strong>models</strong>.<br />
b JB is the Jarque-Bera statistic for test<strong>in</strong>g normality. The relevant critical values at the<br />
95% (99%) level are 5.99 (9.21).<br />
c Q(12) is the test statistic <strong>of</strong> the Ljung-Box portmanteau test for the null hypothesis <strong>of</strong> no<br />
autocorrelation <strong>in</strong> the errors up to order 12. The critical values at the 95% (99%) level are<br />
21.03 (26.22).