November 15, 2017 | By Jack Weixel
Residential and commercial natural gas demand (res/com) is both the largest demand segment in the Lower 48 gas complex and the most dependent on weather. Winter res/com demand can swing upwards of 70.0 billion cubic feet per day (Bcf/d) on a given winter day, while summer res/com demand often resides below 7.0 Bcf/d. A cold winter brings elevated levels of res/com demand, which depletes storage inventory rapidly and can have a dramatic impact on price. A warm winter has the opposite effect, leaving an ample surplus of natural gas underground once heating degree days (HDDs) begin to dissipate at the end of March.
In this edition of Get the Point, we’ll address this demand elephant in the room: What is res/com winter demand historically and how will temperature influence the res/com demand forecast for winter 2017/18?
Source: PointLogic Energy Supply & Demand
Over the past five years, res/com demand has seen record highs and record lows. Looking at the graph above, one might think a pattern has emerged in which the U.S. experiences two high-demand winters followed immediately by two low-demand winters. However, if we go back 10 years, no distinct pattern emerges. One can only predict res/com demand as accurately as one can predict the weather. With the rise of unconventional production and the accompanying changes and constraints to the pipeline grid, predicting regional weather and demand is becoming more important than before – where it’s cold can be as relevant as how low temperatures dip.
During winter 2013-14, the Polar Vortex was heavily concentrated in the Upper Midwest, drawing hard and long on storage, and pushing regional prices to over $20.00 per MMBtu. Last winter, when Henry Hub was short supply, a warmer than normal winter in the Southeast allowed Henry Hub to avoid prices north of $4.00 per MMBtu. Overall in winter 2013/14, the incredibly cold months of January and February saw average res/com demand spike to 42.7 Bcf/d on the season, or 5.8 Bcf/d above the five-year average. On Jan. 7, 2014, res/com demand nearly hit 75.0 Bcf/d, and the cold persisted, with res/com clocking in at 73.6 Bcf/d three weeks later on Jan. 28, 2014. That winter saw six days above 65.0 Bcf/d of res/com, a feat still unmatched to this day.
Over the course of 151 days of winter 2013-14, incremental res/com demand led to 876 Bcf of additional gas withdrawn from storage compared to the five-year average. The following winter 2014/15 saw impressive weather as well, with three days above 65.0 Bcf/d and 40.9 Bcf/d of average res/com demand.
Then came the past two winters of moderate temperatures across the U.S. Heating degree days (HDDs) in the U.S. fell by about 440 on average in winter 2015/16 and winter 2016/17 compared to the ten-year average, with res/com demand bottoming out below 33.0 Bcf/d, on average, for both seasons.
What is Normal?
Looking at average res/com demand over the past 10 winter seasons spanning winter 2007/08 through winter 2017/18, PointLogic data shows an average of 36.2 Bcf/d; this is lower than the five-year average of 36.9 Bcf/d. However, as explained above, an average is not necessarily indicative of any future performance, though it can be useful to review past winter’s deviation. One way to look forward is to assume normal weather over the course of any given winter, and then as winter develops track overages or underages against that normal.
Because res/com demand tracks so well with weather, we can calculate a normal weather scenario – the only question that remains is what constitutes a normal weather time frame. At PointLogic, we use eight years, which we believe reflects better on more recent trends in demand than a ten-year span, but levels out any anomalies seen over the course of a five-year span. The ten-year average of population-weighted HDDs for the lower 48 is 3,056 from November to March, while the eight-year average is slightly lower at 3,015. On a monthly basis, the graphic below plots how an eight-year weather normalized res/com demand season compares to the past four winters.
Source: PointLogic Energy Supply & Demand and Two Season Balanced Forecast
Clearly, there are some large gaps between what is normal and what has occurred over the past four winter seasons. Our eight-year normal forecast for res/com predicts 35.9 Bcf/d of demand, which is 3.7 Bcf/d greater than last winter but 6.8 Bcf/d below winter 2013/14. This value is reflected in PointLogic’s Two Season Supply and Demand Balanced Forecast available to clients in our Markets Module.
Looking back to last winter 2016/17 (yellow line), the market was fooled by forecasts calling for a “normal” winter from some of the most respected weather outfits in the business. All indications through the month of December 2016 were supportive of this assertion, and then temperatures rose and the demand wheels fell off in the months of January and February. In both the winters of 2013/14 and 2014/15 (blue lines) November through February temperatures and corresponding res/com demand levels were supportive of a very cold winter the whole season through.
What the Weather Might Do
For short-term weather forecasts of 1-15 days, PointLogic uses Statweather as our official weather provider. Short-term weather forecasts are typically more accurate than longer-term forecasts, and we use forecast weather data for 330 weather stations across the Lower 48 U.S. to estimate 1-15 day demand levels across three sectors of demand (power, industrial and res/com) in six different regions of the U.S.
Statweather also provides its clients with a 90-day forecast, which can be viewed as a useful barometer against normal weather assumptions when looking forward over the course of any season. In its most recent 90-day forecast (issued on Nov. 6, 2017), Statweather temperature data and our modeled demand data from this forecast indicate a slightly cooler trend to the beginning of winter than the past 8-year average, the results of which are detailed below.
Source: PointLogic Energy and Statweather (derived)
In the Statweather 90-day forecast res/com demand in November will shift upwards at the end of the month to 26.9 Bcf/d, or 1.8 Bcf/d greater than the normal November forecast. The largest delta is seen in the month of December, when the 90-day forecast predicts 43.8 Bcf/d of demand compared to 37.9 Bcf/d of normal res/com demand. In sum, using the three months of forecasted weather data available in the Statweather forecast, winter 2017/18 res/com demand would come in at 37.7 Bcf/d, or 1.8 Bcf/d greater than the normal weather forecast assuming February and March res/com comes in at normal levels.
Impact on End of Season Storage Levels
As detailed in PointLogic’s Get the Point Winter 2017-18: Reaching for Higher Heights, we expect total demand for winter 2017/18, including LNG exports and exports to Mexico, to reach 97.1 Bcf/d. Of this amount, 90.0 Bcf/d is made up of domestic demand line items — namely power burn, industrial demand, fuel & pipe loss and res/com.
Source: PointLogic Energy Supply & Demand
Again, a “normal” winter would yield 35.9 Bcf/d of res/com demand. While the most significant delta compared to past winters resides in the export market, with a record 7.1 Bcf/d of LNG and Mexican exports expected, the weather will still be a major determinant of whether this winter is a chart topper or a disappointment from a total demand perspective.
Should Statweather’s 90-day forecast hold true through January, an additional 1.8 Bcf/d of res/com demand could be added to the 97.1 Bcf/d total demand number, bringing total demand to 98.9 Bcf/d. This would be a full 10.1 Bcf/d above last winter demand levels and 4.2 Bcf/d over the winter 2013/14 average. Either way, total demand in the natural gas market is expected to be stout and the impact on storage withdrawals will be significant.
In the graphic below, we show winter on winter deltas for each element of supply and demand. In Scenario 1, we assume normal winter demand, while in Scenario 2 we assume Statweather 90-day winter demand levels.
Source PointLogic Energy
With production expected to grow to 76.2 Bcf/d this winter, or 5.5 Bcf/d greater than last winter, it’s clear the market is primed to meet increased export and weather incensed demand levels. However, with 97.1 Bcf/d of demand or 98.9 Bcf/d of demand, the market will still end up short under either scenario. In Scenario 1, the market is short about 0.8 Bcf/d, which equates to about 127 Bcf more gas withdrawn than last year. Under the Statweather scenario, the market is short by 2.6 Bcf/d (or 1.8 Bcf/d greater) and 399 Bcf more gas is withdrawn. The impact on storage levels is palpable.
Source: PointLogc Energy and EIA
With last week’s Energy Information Administration (EIA) calculation of a 15 Bcf injection into storage for the week ending Nov. 3, prorated inventories as of Oct. 31 are estimated by EIA at 3,784 Bcf – about 58 Bcf lower than the five-year average. This is our starting point for the winter withdrawal season (regardless of whether the next few weeks are net injections). Last winter’s withdrawals totaled 1,935 Bcf, so therefore under Scenario 1, the market would withdraw about 2,062 Bcf (127 Bcf more) and end-of-season inventories would come in at about 1,722 Bcf. Under Scenario 2, which shows a slightly colder start to winter, withdrawals would increase to 2,334 Bcf (399 Bcf greater), and end-of-season inventories would fall to about 1,450 Bcf. For context, the Lower 48 U.S. withdrew 2,958 Bcf of storage gas over winter 2013/14.
Scenario 1 would put end-of-winter inventories about 66 Bcf lower than the five-year average, or about where they stand as of the end of October 2017. Thus, the pace of withdrawals this winter would largely mimic the five-year average withdrawal rate.
Under Scenario 2, end-of-winter inventories would fall to about 336 Bcf below the five-year average. Under this scenario, it’s difficult the gauge the exact response from the supply side. However, it’s reasonable to think that the price response from a market under siege both by weather and export demand would be positive. Other components of demand would ebb and flow as well – export demand could weaken; industrial demand to heat factory floor space could increase; and power demand could decrease due to price or, conversely, increase due to an increase in generation demand for residences with electric baseboard or radiant heating systems.
It’s reasonable to assume that dry gas plays the likes of the Haynesville and the Marcellus in Northeast Pennsylvania would be motivated by rising prices to fill storage to levels that approach the five-year average. There is also an interesting wrinkle surrounding the price of oil and whether associated gas production could increase to fill the storage void left under a cold winter with increased export demand. In either case, the timing of the production response is also crucial. Should production come in too late, the market could find itself off balance, with inventories in a deeper trough below the five-year average for a longer period of time over the back half of winter and leading into summer 2018.
Stay tuned to PointLogic to follow the delicate balance between supply and demand this winter and to see how the res/com elephant in the room reacts to changes in the weather.