Abstract
We analyze daily log-returns data for a set of 1200 stocks, taken from US stock markets, over a period of 2481 trading days (January 1996–November 2005). We estimate the degree of non-stationarity in daily market volatility employing a polynomial fit, used as a detrending function. We find that the autocorrelation function of absolute detrended log-returns departs strongly from the corresponding original data autocorrelation function, while the observed leverage effect depends only weakly on trends. Such effect is shown to occur when both skewness and long-time memory are simultaneously present. A fractional derivative random walk model is discussed yielding a quantitative agreement with the empirical results.

This publication has 11 references indexed in Scilit: