Abstract:

Integrated volatility of asset return was estimated by means of maximum overlap discrete wavelet transform. The different wavelet functions were chosen to estimate the integrated volatility of Shanghai and Shenzhen 300 indices, and relative error statistics was calculated. The results show that integrated volatilities based on different wavelets had no significant difference. The estimated accuracy was improved with the increasing of sampling frequency. There was an obvious linear relationship between logarithmic scale and logarithmic volatility. The volatility decreasd gradually with the scale increasing.

Key words: high frequency data, maximum overlap discrete wavelet, volatility estimation, wavelet variance