Statistical properties of volatility return intervals of Chinese stocks

Abstract

The statistical properties of the return intervals $τ_q$ between successive 1-min volatilities of 30 liquid Chinese stocks exceeding a certain threshold $q$ are carefully studied. The Kolmogorov-Smirnov (KS) test shows that 12 stocks exhibit scaling behaviors in the distributions of $τ_q$ for different thresholds $q$. Furthermore, the KS test and weighted KS test shows that the scaled return interval distributions of 6 stocks (out of the 12 stocks) can be nicely fitted by a stretched exponential function $f(τ/\barτ)\sim e^{- α(τ/\barτ)^γ}$ with $γ\approx0.31$ under the significance level of 5%, where $\barτ$ is the mean return interval. The investigation of the conditional probability distribution $P_q(τ| τ_0)$ and the mean conditional return interval $<τ| τ_0>$ demonstrates the existence of short-term correlation between successive return interval intervals. We further study the mean return interval $<τ| τ_0>$ after a cluster of $n$ intervals and the fluctuation $F(l)$ using detrended fluctuation analysis and find that long-term memory also exists in the volatility return intervals.