312 VOLATILITY TRANSMISSION PATTERNS BETWEEN THE USA AND SPAINIt is important to note that this impulse response function examines how
fast asset prices can incorporate new information. This fact lets us test for
the speed of adjustment, analyze the dependence of volatilities across the
returns of the S&P500 and the IBEX35 and distinguish between negative and
positive shocks.
16.4 Empirical results
16.4.1 Model estimation
Table16.4displaystheestimatedVAR-asymmetricBEKKmodelofequations
(16.1) to (16.3). Panels A, B and C show the estimation results for the total
period, the pre-September 11 period and the post-September 11 periods
respectively. The lowp-values obtained for most of the parameters shows
that the model fits well the data. Table 16.5 shows the standardized residuals
analysis. It can be observed that the standardized residuals appear free from
serial correlation and heteroskedasticity.
As has been mentioned above, the parameters of Table 16.4 cannot
be interpreted individually. Instead, we have to focus on the non-linear
functions that form the intercept terms and the coefficients of the lagged
variance, covariance and error terms. Panels A, B and C of Table 16.6
Table 16.4 Estimation results for the VAR(1)-asymmetric BEKK
modelPanel A: total periodR1,t R2,tμ −0.000152 −0.000118
(0.64) (0.75)
R1,t− 1 0.000906 0.081108
(0.98) (0.06)
R2,t− 1 −0.008487 −0.037128
(0.79) (0.32)C=
−0.001149
( 0. 00 )
0. 000160
( 0. 29 )
−0.000469
( 0. 00 )
B=
0.949919
( 0. 00 )
0.031741
( 0. 00 )
−0.024743
( 0. 04 )
0.943548
( 0. 00 )
A=
−0.155282
( 0. 00 )0.222199
( 0. 00 )
−0.110431
( 0. 00 )
−0.059604
( 0. 09 )
G=
0.317651
( 0. 00 )
0.018773
( 0. 64 )
0.033911
( 0. 06 )
0.266847
( 0. 00 )
Continued