1. Supply and Demand

On Thursday every week, the Energy Information Agency (Note that all the data presented in this report are taken from the EIA unless otherwise noted) releases national storage data for natural gas. This report is an important data piece for traders and economists who are looking for clues to the supply/demand disposition of natural gas. The inventory follows a similar pattern every year, whereby inventories build from late March to mid November, and then draw down quickly during the cold winter months (Figure 1).
Figure 1 Natural Gas Storage (Source EIA)

Note the cyclical nature of gas inventories. Since the spring and fall has neither excessive heat nor excessive cold, these periods are called the shoulder months and have the least demand for natural gas. Demand is elevated in the summer due to air conditioning, but inventories still grow in the summer, just at a slower pace. Since the expected number of degree days is different in every calendar week, the most meaningful comparisons of natural gas disposition are achieved by comparing the same week in different years. Additionally, other factors affecting demand (holidays and scheduled industrial furloughs) generally fall in the same calendar week and are neutralized by comparing year over year data.
While comparing the same week on a YOY basis can provide “first-order” estimation for changes in supply and demand (footnote 2: When speaking of supply and demand in this paper I do so informally; I intend to mean quantity supplied (excluding changes in inventory) and quantity demanded. Thus if I say “supply exceeds demand by 20 BCF” the reader can take it to mean that, ceteris paribus, the quantity supplied is 20 BCF greater than the quantity demanded, and that difference is made up by inventory levels) , it will only give an accurate depiction averaged over many weeks of data. This is because one finds that even in the same calendar week the weather can vary greatly from one year to the next. Therefore, a more sophisticated approach must account for changes in weather between the two years. This is done by constructing a regression using weather data as explanatory variables and inventory changes as the dependant variable. Once you know the predicted effects of changes in heating and cooling demand, you can then back out a decent estimate for the year over year change in supply and demand disposition.