Modelling volatility persistence under stochasticity assumptions: evidence from common and alternative investments

Long-range memory estimation is a functional statistical mechanics technique to assess predictability in time series. In our work, we scrutinize the long-range dependence structure of volatility in a plethora of alternative and common investments via the use of a fractionally integrated conditional heteroscedastic model. Particularly, we evaluate the fractional persistence parameter of the temporal dynamics for the volatility series of Islamic, socially responsible and common investment indices related to the world major international stock markets. Long-range memory in volatility is measured under different types of market randomness, namely distributional specifications wherein the stochastic error components follow Normal and non-Normal densities. Our empirical results show strong evidence of long-range dependence in the volatility of alternative and common investments under all stochasticity assumptions. Furthermore, we show that the randomness profile has no effect upon the variability of long-range memory. We indicate differences in the statistical significance of the long-range memory across the investigated markets as well as in the degree of persistence. Our findings yield serious implications in terms of quantitative portfolio management and optimization.