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

Research by: John Francis Diaz, Florian Gerth & Michael Young

Executive Summary

This study investigates the long-memory properties and predictability of returns and volatility in high-carbon and low-carbon intensity Exchange-Traded Funds (ETFs). Against the backdrop of increasing investor focus on climate-related risks and sustainable finance, the research addresses a significant gap in the literature by examining how carbon intensity influences the temporal dynamics of ETF performance. Utilizing advanced fractional integration models—ARFIMA-GARCH, ARFIMA-FIGARCH, and ARFIMA-HYGARCH—the analysis provides a nuanced understanding of return persistence and volatility clustering in these ESG-themed assets.

The findings reveal distinct behavioral patterns between high-carbon and low-carbon ETFs. On average, high-carbon intensity ETFs delivered positive returns (2.61%) with lower volatility (variance of 6.96), while low-carbon ETFs exhibited negative average returns (-17.13%) and higher volatility (variance of 14.40). This challenges the common assumption that high-carbon assets are inherently riskier, suggesting a more complex risk-return trade-off.

Empirical results indicate the presence of long-memory processes in the conditional mean returns of several high-carbon ETFs (e.g., CHIU, XLU, VPU), implying predictability that contradicts the weak-form efficient market hypothesis. Low-carbon ETFs, particularly those with inverse or leveraged structures like TZA and CHAD, displayed stronger evidence of volatility persistence and anti-persistent (mean-reverting) behavior. This anti-persistence suggests that these ETFs may not be suitable for long-term holding strategies based on current trends.

Model performance evaluation showed that ARFIMA-HYGARCH generally provided the best fit for low-carbon ETFs, effectively capturing the hyperbolic decay in volatility memory. In contrast, ARFIMA-GARCH models were often more suitable for high-carbon ETFs. The presence of volatility clustering was evident in both groups, with lagged variances exerting a stronger influence than lagged returns, highlighting the importance of modeling volatility dynamics for accurate forecasting.

The study offers practical implications for investors and policymakers. Portfolio managers can leverage the identified predictability for active strategies, especially with volatile low-carbon ETFs. The findings also caution that product structure (e.g., leverage) may influence memory properties as much as ESG ratings, suggesting that policymakers should scrutinize the stability and design of ESG-aligned investment vehicles.

Theoretical contributions include pioneering the application of combined long-memory models to carbon-intensive ETFs and challenging simplistic narratives about ESG performance. Limitations include a relatively short sample period impacted by the COVID-19 pandemic and reliance on a single carbon scoring source.

Future research should extend the analysis to longer timeframes, incorporate additional fractional volatility models, and explore other asset classes or geographic markets to enhance the generalizability of the findings. This study underscores the value of integrating sophisticated time-series modeling with sustainable finance, providing a robust framework for understanding and forecasting the behavior of carbon-sensitive investments.

Keywords:  high-carbon intensity and low-carbon intensity ETFs, long-memory models, anti-persistent properties

To cite this article: Diaz, J. F., Gerth, F., & Young, M. (2025). Long-memory modeling and forecasting of high-carbon intensity rating exchange-traded funds (ETFs). FinTech and Sustainable Innovation. https://doi.org/10.47852/bonviewFSI52025050

To access the article: https://doi.org/10.47852/bonviewFSI52025050

About the Journal