12.07.2015 Views

"Frontmatter". In: Analysis of Financial Time Series

"Frontmatter". In: Analysis of Financial Time Series

"Frontmatter". In: Analysis of Financial Time Series

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

90 CONDITIONAL HETEROSCEDASTIC MODELSforecast <strong>of</strong> σ 2 h+1 is σ 2 h (1) = α 0 + α 1 a 2 h +···+α ma 2 h+1−m .The 2-step ahead forecast isσ 2 h (2) = α 0 + α 1 σ 2 h (1) + α 2a 2 h +···+α ma 2 h+2−m ,and the l-step ahead forecast for σ 2 h+l isσ 2 h (l) = α 0 +where σh 2(l − i) = a2 h+l−iif l − i ≤ 0.m∑i=1α i σh 2 (l − i), (3.9)3.3.4 Examples<strong>In</strong> this subsection, we illustrate ARCH modeling by considering two examples.Example 3.1. We first apply the modeling procedure to build a simpleARCH model for the monthly log stock returns <strong>of</strong> <strong>In</strong>tel Corporation. The sampleACF and PACF <strong>of</strong> the squared returns in Figure 3.1 clearly show the existence <strong>of</strong>conditional heteroscedasticity. Thus, it is unnecessary to perform any statistical teststo confirm the need <strong>of</strong> ARCH modeling, and we proceed to identify the order <strong>of</strong> anARCH model. The sample PACF in the lower right panel <strong>of</strong> Figure 3.1 indicates thatan ARCH(3) model might be appropriate. Consequently, we specify the modelr t = µ + a t , a t = σ t ɛ t , σ 2t= α 0 + α 1 a 2 t−1 + α 2a 2 t−2 + α 3a 2 t−3for the monthly log returns <strong>of</strong> <strong>In</strong>tel stock. Assuming that ɛ t are iid standard normal,we obtain the fitted modelr t = 0.0196 + a t , σ 2t = 0.0090 + 0.2973a 2 t−1 + 0.0900a2 t−2 + 0.0626a2 t−3 ,where the standard errors <strong>of</strong> the parameters are 0.0062, 0.0013, 0.0887, 0.0645,and 0.0777, respectively. While the estimates meet the general requirement <strong>of</strong> anARCH(3) model, the estimates <strong>of</strong> α 2 and α 3 appear to be statistically nonsignificantat the 5% level. Therefore, the model can be simplified.Dropping the two nonsignificant parameters, we obtain the modelr t = 0.0213 + a t , σ 2t = 0.00998 + 0.4437a 2 t−1 , (3.10)where the standard errors <strong>of</strong> the parameters are 0.0062, 0.00124, and 0.0938, respectively.All the estimates are highly significant. Figure 3.4 shows the standardized

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!