Prequential forecasting in the presence of structure breaks in natural gas spot markets

• Main Authors: Kannika Duangnate, James W. Mjelde
• Format: Journal
• Published: 2019
• Subjects: Economics, Econometrics and Finance; Mathematics; Social Sciences
• Online Access: https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85065698625&origin=inward; http://cmuir.cmu.ac.th/jspui/handle/6653943832/65572
• Institution: Chiang Mai University

© 2019, Springer-Verlag GmbH Germany, part of Springer Nature. The natural gas sector has undergone major regulatory and technological changes. These changes may induce structural changes in price relationships among natural gas markets. Tests for structural breaks suggest two potential structural breaks, around 2000 and 2009. Previous forecasting studies on natural gas prices/returns largely are point forecasts and focus on a single spot market; unlike those, this study undertakes simultaneous probabilistic forecasts of eight spot markets. Prequential forecasting analysis examines: (1) whether differences exist in the ability to probabilistically forecast returns among various natural gas markets and (2) how the presence of structural breaks in the natural gas sector influences the probability forecasts. The ability to forecast natural gas markets differs based on the different criteria. Disparities may be explained by each market’s role in price discovery, the alteration of the market’s participation, and whether the market is located in an excess supply or demand region. Irrespective of the models, Henry Hub and AECO returns appear to be easier to forecast, as they generally have the smaller root-mean-squared error, Brier score, and ranked probability score, while Dominion South and Chicago returns appear to be more difficult to forecast. Models using longer periods of data appear to forecast returns better than models using data starting after the breaks; the latter always produces the largest root-mean-squared error, Brier score, and ranked probability score.