Revisiting the multifractality in stock returns and its modeling implications

In this paper, we investigate the multifractality of Chinese and the U.S. stock markets using a multifractal detrending moving average algorithm. The results show that stock returns in both markets are multifractal at a similar extent. We detect the source of multifractality and find that long-range correlations are one of the major sources of multifractality in the US market but not in the Chinese market. Fat-tailed distribution plays a crucial role in multifractality of both markets. As an innovation, we quantify the effect of extreme events on multifractality and find the strong evidence of their contribution to multifractality. Furthermore, we investigate the usefulness of popular ARFIMA-GARCH models with skew-t distribution in capturing multifractality. Our results indicate that these models can capture only a fraction of multifractality. More complex models do not necessarily perform better than simple GARCH models in describing multifractality in stock returns.