×

Improving ARMA-GARCH forecasts for high frequency data with regime-switching ARMA-GARCH. (English) Zbl 1334.62183

Summary: In the literature, we propose a new regime switching autoregressive model for financial time series that improves ARMA-GARCH performance for high frequency data. Although the hybrid model of autoregressive moving average and generalized autoregressive conditional heteroskedasticity (ARMA-GARCH model) has been widely used to characterize and model observed financial time series in stock markets at daily and lower frequencies (e.g., weekly, monthly), few studies occurs on models which can characterize stock prices at very high sampling frequencies. Aim to improve ARMA-GARCH performance for high frequency data, We attempt to incorporate autoregressive HMM driven models into GARCH style models to generate better backtesting results than existing models. We estimate and test ARMA-GARCH, ARIMA-IGARCH, and FARIMA-FIGARCH first and find that the white noise series is not i.i.d. Gaussian. Thus we apply the autoregressive HMM driven model to the white noise series and test it’s forecasting effect. The results indicate that standard ARMA-GARCH and our autoregressive-HMM-noises model can both performs good in daily S&P 500 log returns, while autoregressive-HMM-noise model can do better in high frequency data.

MSC:

62P05 Applications of statistics to actuarial sciences and financial mathematics
62M05 Markov processes: estimation; hidden Markov models
62M10 Time series, auto-correlation, regression, etc. in statistics (GARCH)
91G70 Statistical methods; risk measures
PDFBibTeX XMLCite