2. Why The EIA reports aren’t bearish anymore

First two weeks of July 2009 (i.e. the last two inventory reports)
Average cooling degree days for the past two weeks=62
Average inventory fill for the past two weeks=82.5
First two weeks of July (2006-2008)
Average cooling degree days for the first two weeks= 75
Average inventory fill=85.5
Look at the difference in cooling degree days between this year and the average over the past 3 years (62 versus 75). How big of a difference could that make in cooling demand?
That big of a difference in CDDs is good for about 25 BCF less use per week,
(Footnote 3
I ran a least squares regression using the difference between changes in inventory in two consecutive years as the dependant variable and changes in heating degree and cooling degree days as the independent (or explanatory) variables. I use terms for the differences in the square of the CDDs and HDDs as well to allow for a non-linear relationship (for example, if 10% of the population uses an air conditioner at 80° F, but 50% use an air conditioner at 90° F, then a linear specification won’t give us as accurate of a specification). The equation generated by the least squares regressions to explain inventory changes based on differences in degree days is the following:
d(dINV)= -2+0.97*d(HDD)+0.79*d(CDD)+0.0013*d(HDD^2)+0.0096*d(CDD^2)
where:
dINV=change in EIA natural gas inventory from one week to the next
d(dINV)= difference in dINV for a given week, between year x and year (x+1).
d(HDD)=difference in Heating Degree Days for a given week, between year x and year (x+1).
d(CDD)=difference in Cooling Degree Days for a given week, between year x and year (x+1).
D(HDD^2)= difference in the square of the HDD for given week, between year x and year (x+1).
D(CDD^2)= difference in the square of the HDD for given week, between year x and year (x+1).

The fact that the constant term at the beginning of the equation is -2 instead of 0 reflects the fact that we have had a YOY inventory build of approximately 100 BCF per year over the past three years. In the long run this constant term will equal 0! end footnote)


so based on simple weather YOY comparisons, we would have predicted fills in the 110 range over the past two weeks. Instead we got 82.5, a fact that signifies less gas was put in storage than what is predicted based on the regression.
Based on the past two weeks of data, and assuming the regression is roughly accurate, demand is outstripping supply by 25 BCF/week or almost 4 BCF a day! Normally this level of discrepancy would create an absolute panic in the market. However, because the stock of natural gas is currently 450 BCF above average, the market reaction was muted. Still, the price has moved up 10% since before the release of the July 3rd inventory report.
One way to think of the current situation is to parse the bearish or bullish reality in terms of stock versus flow.
The stock:
Stock of natural gas is 450 BCF above normal. VERY BEARISH
1st derivative (The flow):
After accounting for changes in weather and other exogenous factors, I estimate that demand has outstripped supply by 25 BCF/ week for the past two weeks. Over the past 4 weeks, demand outstripped supply by 21 BCF/week. While four weeks is a small sample size, this level of discrepancy is statistically significant. Future inventory reports will continue to confirm, deny, or accelerate this apparent disparity between supply and demand. BULLISH

2nd derivative (change in the flow):
In 2009Q1, the supply demand disparity was the opposite: supply outstripped demand by 40 BCF/week. Thus, in one quarter, the market went from a 40 BCF/week surplus to a 25 BCF/week deficiency. Obviously, this implies the second derivative of the stock (change in the change of the stock) is negative and steeply so. Why this is and whether it will continue will be the focus of my discussion in sections 3 and 4. BULLISH

In summary, if a trader were to focus only on the stock of natural gas, they will see a 450 BCF glut. Examples of natural gas prices getting driven into the ground during the fall shoulder months abound (and under much less extreme storage gluts then the current one). However, if the trader considers the flow (1st derivative) and change in flow (2nd derivative) a different perspective emerges. Assuming the previous calculations are accurate we are currently running 25 BCF/week below expectations, at which rate the entire storage glut would be gone in 18 weeks. If the 2nd derivative (change in flow) is negative, then the glut could disappear even faster. How can we estimate which effect will dominate prices in the near future? Well, this is not the first time the market has been in this scenario: we had a natural gas glut in both 1999 and 2002, and history can be our guide. In section 5, we look for possible rhymes in these previous instances to divine the probable outcome of our current situation.