Forecasting Sectoral Volatility with ESG Data Using GARCH MIDAS and Hybrid Machine Learning in China

DOI

10.2991/978-94-6239-801-6_153How to use a DOI?

Keywords

ESG rating; volatility forecasting; GARCH-MIDAS; machine learning

Abstract

Environmental, Social, and Governance (ESG) ratings are increasingly recognized as potential drivers of stock volatility. However, the quarterly frequency of ESG ratings mismatches daily returns, posing a methodological challenge for forecasting. This study addresses this gap by systematically comparing nine volatility forecasting models across 31 Chinese A-share sectors from 2010 to 2025. The models include Generalized Autoregressive Conditional Heteroskedasticity (GARCH)‑type models, Long Short‑Term Memory (LSTM), eXtreme Gradient Boosting (XGBoost), and GARCH with Mixed Data Sampling (GARCH‑MIDAS) and GARCH‑MIDAS+LSTM hybrid models. The results reveal three main findings. GARCH-MIDAS+LSTM achieves the lowest forecast error (MSE=0.000814), with its MSE 52% lower than pure LSTM and 81% lower than GARCH(1,1). Mixed-frequency modeling via MIDAS substantially outperforms daily linear interpolation of quarterly ESG. Additionally, hybrid models do not universally beat traditional GARCH. In industries such as Power Equipment, GARCH(1,1) performs better, and adding GARCH conditional variance as LSTM input (GARCH‑LSTM) degrades performance. This study confirms the advantage of the MIDAS structure for low-frequency ESG data. It also verifies the counterintuitive finding that deep learning does not always outperform simple econometric models, providing methodological guidance for volatility forecasting in sustainable finance.

Copyright

© 2026 The Author(s)

Open Access

Open Access This chapter is licensed under the terms of the Creative Commons Attribution-NonCommercial 4.0 International License (http://creativecommons.org/licenses/by-nc/4.0/), which permits any noncommercial use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license and indicate if changes were made.

Download article (PDF)

TY  - CONF
AU  - Wenjing Cao
PY  - 2026
DA  - 2026/10/23
TI  - Forecasting Sectoral Volatility with ESG Data Using GARCH MIDAS and Hybrid Machine Learning in China
BT  - Proceedings of the 2026 2nd International Conference on Financial Innovation and Marketing Management (FIMM 2026)
PB  - Atlantis Press
SP  - 1456
EP  - 1465
SN  - 2352-5428
UR  - https://doi.org/10.2991/978-94-6239-801-6_153
DO  - 10.2991/978-94-6239-801-6_153
ID  - Cao2026
ER  -