Long-Memory Modeling and Forecasting of High-Carbon Intensity Rating Exchange-Traded Funds (ETFs)

This paper offers statistical insights into the predictability of returns and volatility for exchange-traded funds (ETFs) with low to high-carbon intensities, utilizing three configurations of long-memory models: autoregressive fractionally integrated moving average(ARFIMA) combined with generalized autoregressive conditional heteroskedasticity (GARCH), ARFIMA integrated with fractionally integrated GARCH, and ARFIMA paired with hyperbolic GARCH. The findings reveal that high-carbon intensity ETFs generally yield higher positive returns and exhibit decreased volatility than their low-carbon intensity counterparts. Additionally, the study identifies volatility clustering, where lagged conditional variances exert a greater influence than significant lagged mean returns. The analysis also demonstrates the presence of positive long-term dependence in the time series of several high- and low-carbon intensity ETFs, indicating that forecasting using fractionally integrated models is feasible. However, the results suggest no definitive differences in the characteristics of highand low-carbon intensity ETFs concerning short-term, intermediate-term, and long-term memory processes, as some ETF datasets yielded insignificant findings. Notably, the papers observe anti-persistent characteristics, which caution investors against holding these ETFs for extended periods or relying heavily on current trends for decision-making.