When natural gas prices are subject to periodic decoupling from oil prices, for instance due to peak-load pricing, conventional linear models of price dynamics such as the Vector Error Correction Model (VECM) can lead to erroneous inferences about cointegration relationships, price adjustments and relative values. We propose the use of regime-switching models to address these issues. Our regime switching model uses price data to infer whether pricing is oil-driven (integrated) or gas-specific (decoupled). We find that UK natural gas (ICE) and oil (Brent) are cointegrated for the majority of the sample considered (1997-2014). Gas prices tend to decouple during fall and early winter, when they increase relative to oil consistent with heating demand for natural gas creating gas-specific pricing. Using the model to infer relative values when evidence favors integrated markets, we find that the industry 10-1 rule-of-thumb holds, meaning that the value of one barrel of oil is 10 times the value of one MMbtu of natural gas.
oil natural gas peak load pricing regime switching relative value