Modelling and Forecasting Energy Market Volatility Using GARCH and Machine Learning Approach - Seulki Chung Abstract

Aug 3, 2026

Upload by Liam

This paper provides a comparative study of univariate and multivariate GARCH-family models and machine learning methods for modeling and forecasting volatility in major energy commodities, including crude oil, gasoline, heating oil, and natural gas. A combined dataset of financial, macroeconomic, and environmental variables is used to evaluate out-of-sample predictive performance. Volatility persistence and transmission are examined jointly through the SHAP method, showing that machine learning generally forecasts better, while GARCH tends to overpredict and machine learning underpredict volatility. Spillovers are strongest from crude oil to gasoline and heating oil, whereas natural gas transmission is weaker.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages30

Reading Time1 h 16 min

AccessFree to download

Introduction

Energy market volatility and forecasting motivation

GARCH-family models for conditional variance

Limitations of GARCH and motivation for machine learning

arXiv :2405 . 19849v1 [ econ .EM] 30 May 2024 Modelling and forecasting energy market volatility using GARCH and machine learning approach Seulki Chunga aGSEFM, Department of Empirical Economics, Technische Universit¨at Darmstadt, Karolinenpl.5, Darmstadt, 64289, Germany Abstract This paper presents a comparative analysis of univariate and multivariate GARCH-family models and machine learning algorithms in modeling and forecasting the volatility of major energy commodities: crude oil, gasoline, heating oil, and natural gas. It uses a comprehensive dataset incorporating financial, macroeconomic, and environmental variables to assess predictive performance and discusses volatility persistence and transmission across these commodities. Aspects of volatility persistence and transmission, traditionally examined by GARCH-class models, are jointly explored using the SHAP (Shapley Additive exPlanations) method. The findings reveal that machine learning models demonstrate superior out-of-sample forecasting performance compared to traditional GARCH models. Machine learning models tend to underpredict, while GARCH models tend to overpredict energy market volatility, suggesting a hybrid use of both types of models. There is volatility transmission from crude oil to the gasoline and heating oil markets. The volatility transmission in the natural gas market is less prevalent. Keywords: Energy markets, Volatility, Forecasting, Univariate GARCH, Multivariate GARCH, Machine learning, SHAP JEL: C32, C45, C52, C53, Q41, Q43, Q47 March 26, 2024 1. Introduction Energy serves as a fundamental building block for driving economic development. Fluctuations in its pricing significantly influence the flow and allocation of resources within the energy market, yielding substantial economic impact. (Kaufmann and Connelly (2020); Hamilton (1983)) . This inherent volatility in energy markets, depicted by fluctuations in the prices of energy commodities such as crude oil, has implications for economic stability, investment strategies, and policymaking decisions (Pindyck (1999)) . High volatility can discourage investment in fixed capital due to price uncertainty and encourage firms to protect their assets against price risk. Additionally, high volatility can lead to greater demand for storage, resulting in higher spot prices and convenience yield. Understanding changes in volatility can help explain shifts in other economic variables (Henriques and Sadorsky (2011); Karali and Ramirez (2014); Pindyck (2004a)) . Therefore, accurately forecasting this volatility is important as it carries direct implications for hedging and derivatives trading. Furthermore, time-varying volatility is subject to heteroskedasticity in the data, resulting in biased standard error estimates and invalid statistical inference (Efimova and Serletis (2014)) . However, energy market volatility is determined by various factors, including geopolitical events, supply-demand imbalances, regulatory changes, and macroeconomic factors, and is thus difficult to model and predict. Historically, the volatility of energy prices, particularly crude oil, has been an important point of academic interest due to its wide-ranging implications on economic activity, and the literature on energy market volatility has evolved substantially over the past few decades. Traditional econometric models, especially the Generalized Autoregressive Conditional Heteroskedasticity (GARCH) model introduced by Engle (1982) and further developed by Bollerslev (1986), have been extensively employed to model the conditional variance characteristic of time series data. Their variants such asthe Exponential GARCH (EGARCH) introduced by Nelson (1991) and the Glosten-JagannathanRunkle GARCH (GJR-GARCH) proposed by Glosten et al. (1993), have also been used to capture the asymmetric effects of shocks on market volatility, a feature particularly pronounced in energy markets (Sadorsky (1999); Reboredo (201

FAQs

Which models are compared for forecasting energy market volatility?

The study compares univariate and multivariate GARCH-family models with machine learning algorithms for volatility forecasting across several energy commodities.

How are volatility persistence and transmission analyzed in this work?

The paper jointly examines persistence and transmission using the SHAP (Shapley Additive exPlanations) method.

What do the results indicate about prediction accuracy and volatility spillovers?

Machine learning shows superior out-of-sample forecasting performance; GARCH overpredicts while machine learning underpredicts. Volatility transmission is evident from crude oil to gasoline and heating oil, but is less prevalent for natural gas.

Aug 3, 2026

Upload by Liam