Michael Weylandt proposes framework to forecast natural gas prices


Jul 30, 2020
Michael Weylandt proposes framework to forecast natural gas pricesBy Shawn Hutchins

Michael Weylandt is lead author on a paper selected for the JSM Best Student Paper Award by the Business and Economic Statistics Section (B&E) of the American Statistical Association (ASA). 

As part of the award, Weylandt, a fifth-year doctoral student in statistics and researcher in the Center for Computational Finance and Economic Systems (CoFES) at Rice University, will present the paper virtually as part of the Joint Statistical Meetings (JSM), August 2 - 6, 2020.

“Natural gas prices, like all financial markets, are largely driven by supply and demand. But unlike stocks and bonds, natural gas pricing also reflects transportation and storage dynamics that are part of our nation's highly integrated gas network,” Weylandt says.

This natural gas network contains hubs, such as the Henry Hub in Louisiana, that connect large numbers of sellers and buyers. Henry Hub, one of the most important transaction and distribution junctions for natural gas, interconnects with nine interstate and four intrastate pipelines. The hub has been used to set gas price standards across the U.S. and in markets worldwide.

Weylandt explains, “Natural gas spot pricing problems exist when there is not enough intra-day market data communicated between the highly liquid markets that allow for efficient and robust price discovery, and more thinly traded hubs, which are highly illiquid and limit the effectiveness of standard risk management techniques.”

In the paper titled, “Multivariate Modeling of Natural Gas Spot Trading Hubs Incorporating Futures Market Realized Volatility,” Weylandt and co-authors Rice alumnus Yu Han and Katherine Ensor, the Noah G. Harding Professor of Statistics and director of CoFES, demonstrate a multi-variable modeling tool that uses high-frequency intraday data from thickly-traded hubs to improve natural gas volatility estimation and risk management at thinly traded hubs.

The model combines open, high, low, and closing price information from Henry Hub futures markets with end of day spot prices from across the country. By combining these two sources of data in a Bayesian framework, the Rice scientists are able to incorporate temporal and spatial aspects of natural gas markets in a single unified model.

“I look forward to presenting the usefulness of our modeling technique at the JSM virtual meetings. The market for natural gas is among the most important commodities in the U.S.,” added Weylandt.

Beginning in 2015 natural gas-fired generation exceeded coal in the U.S. on an annual basis, and it is forecasted to continue to make up a larger proportion of the U.S.'s electricity generation mixture. This increase in usage has been primarily driven by the development of fracking technology.

After defending his thesis this summer under Professor Ensor’s supervision, Weylandt will begin his postdoctoral fellowship at the University of Florida. Working with Professor George Michailidis, he will conduct research in computational statistics, high-dimensional and multivariate machine learning, and graphical models with applications in finance and neuroscience. Weylandt received a B.S.E. in operations research and financial engineering from Princeton University in 2012.