A joint model for temperature and natural gas with an application to the US market

Natural gas spot prices and temperatures have been studied in detail in the literature as separate processes. We propose a simple joint model that, in spite of its parsimony, describes accurately many stylized facts of the two time series: in particular we show the role played by a time-delay parameter in order to take into account the impact of temperature forecast in cross-dependency. We discuss in detail a stepwise procedure in order to calibrate model parameters, describing the elementary estimation techniques involved and the statistical accuracy achieved. In the analysis, we focus on the benchmark market in the USA (Henry Hub) and the temperatures in the Northeast and Midwest regions; we observe a negative, statistically significant, gas-temperature correlation in the cold season.