The coldest temperatures of the new millennium have prompted record-setting volatility in the natural gas sector as storage levels have dwindled to 10-year lows, more than 30% below historical averages. The commodity reached a 4-year high of nearly 6.50/MMBTU on February 24 before staging a sharp 25% correction over the next 5-days, the largest 1-week swan dive in the last 16 years. As the end of the winter withdrawal season approaches, it is appropriate to assess where supplies will finish the season, to what extent they will recover during the summer, and what the impact on prices will be.
Historically, the natural gas withdrawal season runs from the first week of November through the final week of March, meaning that around four storage reports remain in the 2014 season. As of the week ending February 28, the EIA reports that there are 1196 BCF of natural gas currently in storage, 758 BCF or 39% below the five-year mean and 908 BCF or 43% below last year's levels. If the withdrawal season were to have ended last week, the 1196 BCF in storage would already be the sixth lowest season-ending level in the last 15 years dating back to 2000.
However, my storage model indicates that we still have four withdrawals to go. For the week of March 1-7 to be released this Thursday, I am projecting a significantly above-average storage withdrawal of -183 BCF thanks to record-setting cold across the Midwest, Great Lakes, and East Coast that saw temperatures drop as cold as -30F last week. A -183 BCF withdrawal would be a huge 88 BCF greater than the five-year mean withdrawal of just -95 BCF and -38 BCF greater than last year's already bullish -145 BCF draw. If the projection verifies, it would be the largest withdrawal for the first week of March in the last 15 years, blowing away the previous record by 38 BCF.
As milder conditions replaced the arctic cold late in the week, natural gas demand has been in a steady decline since last Friday. With above-average temperatures expected to persist across much of the country through Wednesday, I expect several days of single-digit daily withdrawals through the middle of this week. My projection for the week of March 8-14 shows withdrawals slumping nearly 120 BCF week-over-week to just -60 BCF, which would nonetheless still be 30 BCF greater than the five-year mean.
For the final two weeks and a half weeks of March, computer models suggest that perhaps a second surge of cold air will infiltrate the nation late next week-along with the possibility of a major late-winter Nor'easter for Nee England--and persist into the third week of March followed by generally seasonal temperatures during the final days of the month. For the final two weeks of the withdrawal season, my model projects that -55 BCF will be withdrawn during week ending March 21 (48 BCF greater than average) and -15 BCF will be withdrawn the week ending March 28 (-7 BCF greater than average), before net injections begin the following week. Figure 1 below shows these projections overlaid against the observed storage numbers reported by the EIA. Note: These numbers fluctuate daily as updated forecast data and pipeline numbers are integrated into the model and the most recent numbers may differ slightly from those displayed here. Projections are updated twice daily and can be viewed on my website, with the link available in my profile.
Figure 1: Natural Gas Storage Projections For the next 5 weeks
Should these projections verify, we will conclude the withdrawal season with 883 BCF of natural gas in storage. This would be 931 BCF or 50.8% below the five-year mean average season-ending storage total of 1814 BCF and 790 BCF or 46.9% shy of last year's season ending 1673 BCF. Based on the table of the 10 lowest season-ending storage levels of the last 15 years dating back to 2000 shown below in Table 1, 2014 will finish with the third smallest volume of natural gas in storage behind only 2003 and 2001, although this year's departure from 5-year averages will be the largest on record, by a nearly 400 BCF margin.
|Rank||Year||Week Ending||Storage (BCF)||Departure||Price ($/MMBTU)|
|3.||2014||March 28||883 (PROJ)||-931||???|
Table 1: Top 10 smallest withdrawal season-ending storage levels 2000-2014
Should these next four weeks play out as projected, what is an appropriate price for natural gas, given the huge storage deficit? When season-ending prices are plotted against season-ending storage departures over the last 15 years, the following scatterplot is generated in Figure 2 below:
Figure 2:Fair Price Calculation using end of season prices and storage deficits, 2000-2013
If the trend-line is extended to the ~-930 BCF projected deficit for the current season, a fair price of $6.14/MMBTU is calculated. However, besides the dismal fit of the trend line (R-squared of 0.035), there are several problems with calculating a fair price using this technique. First, roughly half of the years, from 2000-2007, were in the "pre-fracking" era of natural gas, during which time the supply/demand picture was very different than it is today, with generally tighter supplies, but more consistent production that was less vulnerable to freeze-offs and other disruptions. There was also concern, particularly in the early part of the millennium, that domestic supplies were going to soon be exhausted, requiring more expensive foreign imports, which helped to keep prices inflated, even in the face of higher storage levels.
A stronger method to calculate a fair price is to compare weekly storage departures from average and prices regardless of season over the last twenty-four months from February 2012 to February 2014. The 24-month period provides sufficient data points while being short enough to ensure that pricing fundamentals have largely remained the same. This data is shown below in Figure 3, along with a single point plotting the most current price of natural gas ($4.62/MMBTU) against the end-of-season storage projection.
Figure 3: Fair Price Calculation using weekly storage deficits and prices, 2012-2014
With an R-squared value of 0.91, there is a much stronger correlation between storage levels and price. Nonetheless, this model still calculates a fair price of $5.42/MMBTU, a significant 14.8% discount versus the current price, suggesting that natural gas is undervalued at current levels.
What could explain this discount to the fair price predicted by end-of-season storage levels? This apparent undervaluation suggests one of two things. Either investors expect the forecast over the next four weeks to warm dramatically and storage levels to bottom out significantly above my current projection, or investors are looking beyond the 4-5 week timeframe and expect storage levels to recover rapidly during the spring shoulder season and summer cooling season. While I would like to think that I am the only one capable of constructing a forecast model for natural gas storage, I realize that any respectable commodities firm will have its own proprietary model along with a fleet of in-house meteorologists to back it up. Therefore, I believe it to be relatively unlikely that natural gas will experience a sudden spike as major investors suddenly realize that natural gas is going to finish the season with under 1000 BCF in storage.
To evaluate the second proposed rationalization of natural gas' present undervaluation, we turn to history. Let's assume natural gas bears expect natural gas storage levels to recover quickly as the heating season ends and temperatures moderate. This leads to the question: despite storage levels that are likely to be 930 BCF or more below the five-year average, is it feasible to return to average storage levels by the start of the next withdrawal season in November, thereby justifying the present discount versus the calculated end-of-season fair price? Unfortunately for natural gas bulls, history tells us that the answer is yes, given the proper conditions.
While not ideal due to the previously discussed evolution in natural gas fundamentals over the last decade, the years 2001 and 2003 will have to serve as historical analogs, seeing as they are the only years with remotely similar end-of-season storage levels and departures from average. Storage departures for the injection seasons from April through October of 2001 and 2003 are plotted below in Figure 4.
Figure 4: Swift recovery in Natural Gas storage during summer 2001 & 2003
Clearly, storage deficits can be recouped very quickly. In 2001, it took only 2 months and storage was above-average by early June while in 2003, it took until September to break back above average, although a larger deficit (-570 BCF vs. -348 BCF) was overcome. 2001 saw 10 >100 BCF storage builds during the injection season while 2003 saw 8, the top 2 years in the last 15 years, with the distant third place, 2009, seeing just 5.
Taking a closer look at 2003, which I see as the better analog year, total natural gas consumption during the injection season totaled 9.307 Trillion Cubic Feet (TCF), the lowest injection season demand of the entire decade and down steeply from the previous year when 10.1 TCF was consumed. In order to correct for demand growth since 2003, the injection season fractional demand compared to the calendar year was calculated as well and was determined to be just 41.7%, the lowest fractional injection season demand in the last 15 years and a full percentage point lower than the second lowest year.
Looking at temperature patterns, the summer of 2003 was comparatively cooler than more recent summers. It was only the 34th warmest summer in 109 years of data, during a decade which saw 6 different top-5 all time hottest summers recorded, as records were set and then re-set. 2003 had the fourth coolest April-September period in the 12 years between 2002 and 2013 for which records are available. The month of June was particularly cool. Not only was it the coolest June between 2002 and 2013, but it was the 6th coldest in the last 109 years. June 2003 saw injections of 125 BCF, 114 BCF, 127 BCF, and 97 BCF with the storage deficit collapsing from -427 BCF at the start of the month to -302 by month's end, meaning that supply/demand averaged 4.5 BCF/day loose to the 5-year average during the month.
Overall, particularly cool temperatures dominated the densely populated East Coast and Ohio Valley throughout the summer, the major cooling demand centers of the nation, as shown in the map below in Figure 5.
Figure 5: April-September 2003 Temperatures Anomalies
Overall, during the April through October injection season, natural gas supply/demand was 2.75 BCF/day loose to historical averages, with the largest differential occurring during June as discussed above. Cool temperatures therefore likely played a large part in the quick normalization of storage levels by autumn. Given that total industrial demand, which is less dependent on temperatures, fell to 3.36 Trillion Cubic Feet during the injection season from 3.47 TCF and 3.65 TCF in 2001 and 2002, respectively, the slump in demand cannot be entirely blamed on temperature. It is likely that a certain degree of demand destruction also occurred due to high prices and low supplies via fuel switching. Natural gas production was unchanged from a year earlier and does not appear to have played a significant role.
Applying the same Fair Price calculation discussed earlier to 2003, natural gas at the end of the winter withdrawal season priced at $5.69/MMBTU was trading at a similar 17.8% discount (versus the present 15% discount) to a fair price of $6.42/MMBTU. What was the impact on price during the 8-month injection season as storage levels normalized? Natural gas prices during the summer of 2003 are shown below in Figure 6.
Figure 6: Natural Gas prices trend lower during the summer of 2003 after storage deficits normalize to near historical averages
Prices climbed during the first part of the summer, ultimately peaking near $6.40/MMBTU - the season-ending fair price - in early June, even as the storage deficit had been slashed in half. Prices then slumped the remainder of the season as demand fell in the face of cool weather in June and the tropics remained quiet, ultimately entering November under $4.00/MMBTU, down 24% from April prices and 38% from the summer high, and back within about 3% of its fair price at the end of the injection season. 2001 was even worse with prices slumping 43% from the beginning of April through the end of October.
What conclusions can be drawn from this historical discussion? It is first important to note that a much larger storage deficit must be overcome this year, -930 BCF vs. -570 BCF in 2003 and -348 BCF in 2001. If we are to eat up this storage deficit over the next 8 months, natural gas supply/demand must be roughly 3.75 BCF/day loose to historical levels from April through October. The summer of 2003, despite below-average temperatures and expensive gas, averaged just 2.75 BCF/day loose to averages. If 2013 followed a similar pattern, we would enter the 2014 withdrawal season still roughly -300 BCF below the five-year mean. Using the same fair price calculation as before, such a storage deficit would yield a fair price of $4.40/MMBTU, which is only about 5% below the last price of the November 2014 contract, suggesting relatively limited downside from current levels.
Last Thursday, NOAA issued an El Nino watch, stating that there was a 50% chance of El Nino conditions developing during the summer of 2014. The most recent El Nino year was 2009 which actually saw comparable temperatures to 2003 (which was not an El Nino Year), checking in at the 62nd warmest summer in the last 115 years (versus the 75th warmest summer in 2003), with the coolest temperatures over the Great Plains and Ohio Valley. This suggests that it is reasonable to expect temperatures this summer to be at or below-average across much of the nation, favoring larger-than-average storage builds.
I expect lower demand and higher supplies will aggressively eat into the storage deficit once the shoulder season gets into gear next month as rebounding natural gas production adds supply and switching to more cost-effective fuels such as coal blunts demand. (Further discussion regarding the specifics of rising natural gas production and gas-to-coal switching are beyond the scope of this post, and will be addressed in detail in later articles.) The springs and summers of 2001 and 2003 proved that it is feasible to erase a large deficit in a single season, with concomitant declines in natural gas prices. It is for this reason that investors are likely valuing the commodity nearly 20% below its "fair price." However, neither of these variables is sufficient to normalize storage levels independent of the others and it will take a perfect storm of circumstances to wipe out the entire 900 BCF deficit in a single season. El Nino conditions must develop as expected suppressing summertime heat across major demand centers, there must be some degree of demand destruction due to gas-to-coal switching due to cheaper coal, and production must again exceed record levels. Due to the historically low starting storage levels, any disruption in this higher production + gas switching + El Nino thesis could lead to investors rapidly re-evaluating their positions leading to above-average volatility and price spikes. Particularly, without the bearish impact of below-average summer temperatures, I expect it will be difficult to completely abolish our current storage deficit by the end of the injection season, regardless of other factors. For now, I am largely on the sidelines with just 25% of my holdings in UNG and UGAZ (of which 75% is UNG and 25% is UGAZ), with the remainder of my portfolio in cash or unrelated equities and funds. I will continue to hold this position with intent to sell above $5.00/MMBTU or close the position upon further evaluation. Should natural gas drop below $4.40/MMBTU or so, I will consider adding to my position on the basis that lower prices will blunt anticipated gas-to-coal switching and surges in supply, leading to dangerously low supplies entering next winter's withdrawal season. Likewise, on a spike above $5.00/MMBTU I will consider adding some short positions via KOLD or DGAZ, on the theory that the reverse will take place.
Disclosure: I am long UGAZ, UNG. I wrote this article myself, and it expresses my own opinions. I am not receiving compensation for it. I have no business relationship with any company whose stock is mentioned in this article. I have no plans to initiate additional positions in the next 72 hours