Nonstationary Semiparametric ARCH Models

• 영문명: Nonstationary Semiparametric ARCH Models
• 발행기관: 한국계량경제학회
• 저자명: Heejoon Han Shen Zhang
• 간행물 정보: 『한국계량경제학회 학술대회 논문집』2009년 하계학술대회, 1~35쪽, 전체 35쪽
• 주제분류: 경제경영 > 경제학
• 파일형태: PDF
• 발행일자: 2009.08.30

영문 초록

We consider nonstationary semiparametric ARCH models. One model combines the nonparametric ARCH model with an integrated or a near-integrated covariate. The other model combines the nonparametric cointegrating model with the parametric ARCH(1) model. These models can generate the commonly observed long memory property in volatility, and these models allow that the unconditional variance of return series is not constant. Additionally we investigate the nonparametric cointegrating volatility model, and we show that the kernel estimate of this model is consistent and the asymptotic distribution is mixed normal. For the monthly return series of the S&P 500 index of the sample period 1919-2008, the nonstationary semiparametric ARCH models give a better explanation of the squared stock return series and provide a better outof-sample forecast than a stationary nonparametric ARCH model and two parametric models, GARCH(1,1) and ARCH-NNH, when we use the squared return series for the estimation of their nonparametric components. For the sample period of 1928-2008, if we use the realized volatility instead of the squared return series for the estimation of nonparametric components of the models, the nonstationary semiparametric ARCH models give a better explanation and forecast of the realized volatility. We also …nd out that, if we use realized volatility for the estimation of nonparametric components of the nonparametric or semiparametric models, the nonparametric or semiparametric models perform in general better than the case when we use squared return for the estimation.

키워드

Kernel estimation ARCH semiparametric nonparametric cointegrating regression nonstationarity volatility persistence long memory property