Natural gas plays a crucial role in the global economy as one of the most important energy commodities. Its price volatility affects not only producers and distributors but also industries that depend on gas as a raw material. The global natural gas market has recently faced severe instability due to geopolitical factors, particularly the Russia–Ukraine conflict, which disrupted gas exports and triggered extreme price fluctuations. These developments underscore the importance of accurate forecasting methods to help governments, industries, and investors make informed decisions.
This study aims to improve the accuracy of natural gas price prediction by applying a Pulse Function Intervention Analysis approach combined with the Autoregressive Integrated Moving Average (ARIMA) model. While previous studies using ARIMA or ARIMA–GARCH models have provided reliable forecasts under normal market conditions, they often fail to capture sudden and temporary shocks caused by extraordinary events. The pulse function method, introduced by Box and Tiao (1975), is specifically designed to detect short-term interventions or “shocks” in time series data—making it suitable for analyzing the price effects of the Russia–Ukraine war.
The researchers used weekly closing prices of natural gas on the New York Mercantile Exchange (NYMEX) from January 2020 to October 2022, comprising 147 data points. Data were divided into training (January 2020–August 2022) and testing (August–October 2022) sets. Through a sequence of statistical analyses—including stationarity testing, parameter estimation, and diagnostic checking—the best-fitting model was identified as Probabilistic ARIMA (0,2,1). When integrated with the pulse function intervention, the model parameters were determined as b = 0, r = 2, and s = 0, indicating an immediate and short-lived impact of external shocks. The model achieved a Mean Absolute Percentage Error (MAPE) of 12.26%, classifying it as a good forecasting model. The comparison between predicted and actual data showed consistent fluctuation patterns, validating the model’s accuracy and stability.
The results demonstrate that pulse function intervention analysis can effectively capture short-term volatility and improve predictive performance compared to conventional time series models. In conclusion, this research contributes to developing more responsive forecasting tools for energy markets. Its findings can assist policymakers in designing adaptive energy strategies, guide companies in optimizing production and trading, and help investors better assess risks. However, the authors note that further studies with larger datasets are needed to enhance the generalizability of the model under different market conditions.
Author: Sediono, Drs., M.Si.
Details of research can be viewed at https://scholar.unair.ac.id/en/publications/prediction-of-natural-gas-prices-on-the-new-york-mercantile-excha/
