Natural Gas: Predicting, Especially The Future

by: Moshe Ben-Reuven


A dynamic pricing model for NG is presented, based on the Hotelling (1934) approach for an exhaustible resource.

The model uses available US consumption, production, storage and proved reserves data published by the EIA. NG Henry Hub monthly price moves are correlated well for 2013-2014.

Our ad-hoc conclusion: NG prices are set to increase in the near term, quite appreciably.

In an earlier article ("Marcellus Shale: Through a Glass, Darkly"), we have carelessly predicted an average rise in the price of NG, based on an enhanced Hotelling model. Hotelling (1934), dealing with any exhaustible resource, originally projected forward an exponentially increasing asset value, driven by the current discount rate. We thought this should be further amplified, multiplied by the ratio of current consumption over remaining proven reserve. Of course, as the resource comes closer to near-exhaustion, the denominator will accelerate the increase in cost. The formulation provided a smooth exponential curve. Since the Henry Hub price is known to fluctuate quite a bit, could we do better?

Here, too, we are dealing with the physical world of balancing reserves, consumption, production and storage - over time. Natural gas travels in only two ways. As liquefied NG (LNG) by large shipping tankers, or in compressed gaseous form through pipelines. The market tends to be essentially confined and regional. In the US, external influence by inflows and outflows is still small and physically negligible. This near-hermetic system affords some simplicity, as cause and effect in an export-less and import-less environment may be clearly delineated up to a point.

1. Proved Reserves

To make absolutely certain, we are dealing with an exhaustible mineral resource. Whether we have already discovered (or accounted for) the ultimate extent of "Proved Reserves" or not makes little difference. How come? Because the rate of our usage relative to the amount of proved reserves indicates we have on the order of 10 years if we continued consuming 26 TSCF/year without any change. This, under the naïve assumption that 100% of the Proved Reserve would be indeed available for recovery, which is never the case. Skeptic? Ask Mr. Mark G. Papa, CEO of EOG Resources, Inc.

Regrettably, in a world where miracles are rare, we should forget about sudden new disruptive technologies to allow far more shale gas recovery, orders of magnitude better than hydro fracking. For one thing, E&P shale operators have invested not a red cent in R&D over the last 10 years; they have, however, invested billions of dollars in their respective books of hedges. A new technology needs some 25-30 years to mature for commercial implementation, so time is almost up. The service providers like Schlumberger, Halliburton, Weatherford, Baker-Hughes and National Oilwell Varco did invest in ways and means to improve hydro fracking. Although staged fracking and "short & fat" show improved yield, and multi-bore tightly-spaced well pads are saving on the drilling, one cannot expect an order of magnitude improvement in either yield or CAPEX.

The same goes for sudden, vast reservoir discoveries deep under the salt in Utah. Or elsewhere.

Fig. 1 shows the evolution of the US Proven Reserves of NG. Source: EIA-2014. The recent accelerated increase over the last 10 years or so is due to the implementation of horizontal drilling and hydro fracking in shale gas recovery. The 2013 drop shown is based on physical consumption. It is inferred that low prices will curtail development and discovery.

Note: TSCF is trillions of standard cubic feet. "Standard" means "standard temperature and pressure" STP, namely 0 degree C and 1 bar absolute.

Suppose we accept the EIA December 31, 2013 benchmark for Proven Reserve at some 322 TSCF as an ultimate reservoir (normally, less than 10% of proven reserves are recoverable), and we are using over 8% of it annually (current figure, 26 TSCF/y), then we should consider the effect of consumption on the remaining reserve.

2. NG Production, Withdrawals

There are four sources of NG in the US: gas wells, oil well gas, coal bed gas, and shale gas. Withdrawals are listed by the EIA, 2014. In Fig. 2, we see the proportions of each source recorded at the end of December 2013. Evidently, shale gas is the largest contributor, at 40% of the total of 2.56 TSCF/month withdrawn.

Historical monthly withdrawals (i.e., production) from 2006 through end of 2013 are shown in Fig. 3. The four sources and the totals are depicted. Withdrawals from gas wells, oil wells, and coal beds are in steady decline, while shale gas is the only source increasing. Note that a significant percentage of the total recovered gas is used for recovery, production, and separation of non-hydrocarbon components. Thus, for December 2013, from a total of 2.56 TSCF/month produced, only 2.18 TSCF were actually brought to market. We also note that although shale gas withdrawals are still rising, the production gradient is tapering off from 2011 and on.

Of course, not all gas comes to the same distribution point. Henry Hub is quite distinct from north Appalachia, and their prices are quite different. We will nevertheless sum up all of the sources as we answer the question of production over time. Note that although production of each source goes through fluctuations in time, the amplitudes are small relative to the respective mean value from each source. We focus here on the total Marketed Production curve, showing quasi-linear rise correlated in Fig. 3. Marketed production has increased on average from 1.65 to 2.17 TSCF/month over a period of 7 years to 2013, a 4.5% average annual increase.

3. NG Consumption, Demand

There are four major consumption sectors: residential, commercial, industrial, and power generation. Additional, smaller depletion goes for the production and distribution process, and a fraction of a percent to vehicle use in the form of compressed NG or CNG. Figure 4 shows a 2013 snapshot of the division among the sectors, based on the EIA, 2014 monthly consumption update. Total consumption in 2013 was 26 TSCF.

Figure 5 shows historical annual consumption per sector, 1997-2013 (EIA annual). The only sector in significant Y/Y growth is electric power generation, affecting total consumption. Even this sector is showing a sharp decline in 2012/2013. Industrial usage, following a long decline, is also in recovery from 2009, which, combined with power generation, causes the total consumption (top curve) to be stronger on average.

It should be mentioned that the 2012/2013 decline in NG use for electric power is "compensated" by a significant increase in usage of coal. Figure 5a shows historical data on electric power generation by source (EIA, 2014). The total electric power generation from all sources remains roughly constant at 4 trillion kWh. However, between 2012/2013, NG-based generation declined by 112 billion kWh, while coal-based generation surged by 72 billion kWh.

Remarkably, this was accomplished while NG was going through a very low price period, touching on $2/MBTU. So much for "clean air" policies, or "Green House Gas reduction": To generate 72 billion kWh/year, one must burn 33.2 million tonnes/year of bituminous coal at 30% efficiency. To reduce NG generated electric power by 112 billion kWh, at typical 50% efficiency, one uses 16.2 million tonnes (0.7 TSCF) less NG per same year. Thanks, Peabody Energy, but we can't breathe.

Back to Fig. 5. Whereas annual consumption figures seem rather smooth, not so the corresponding monthly values. Intra-year, or monthly, consumption figures vary significantly and quite sharply. Here we used EIA, 2014 monthly consumption data. Figure 6 shows the monthly consumption curves per sector and the sum of all four sectors.

We note that total consumption per month fluctuates with near-perfect 12-month periodicity, and a peak-to-peak amplitude of 1-1.5 TSCF. The sharpest saw-tooth contribution is from spiking of the synchronized residential, commercial, and industrial heating during winter months. The lesser spikes are due to electric generation peaking during hot summer months. It is also worthy of note that the 3 last winter peaks of total consumption, 2011 through 2013, have consistently risen, from 2.55 to 2.99 TSCF/month, exponentially. The last peak rise was 12.4%, which is significant. We likewise note that the summer's mini-peaks of consumption in the same 2011-2013 period were also above the averages of the preceding 3-year interval, 2009-2011.

4. NG Storage

The last parameter in the current NG puzzle is storage. Very clearly, storage is not a huge reservoir which acts as an alternative to production in case of major disruptions. It is a relatively small reservoir set, intended as buffer between slowly varying steady production and sharp seasonal variations in consumption, as we have observed. Figure 7 shows weekly reported (EIA, 2014) storage levels in billions of SCF, or BSCF.

Storage has three physical locations reported: consuming region east, consuming region west (far smaller), and producing region. The sum of all 3 appears in Fig. 7. We note again a near-perfect 12-month periodicity of the level of storage vs. time, and that peak storage levels are between 3.5 and 4 TSCF. In other words, the maximal amount stored is of the same order of magnitude as monthly consumption. We also note that the spring minima of the last 3 years go lower and lower. These bottoms occur at the end of March each year, and the most recent minimum in March 2014 has slipped below 1 TSCF. This will be discussed further.

5. What About the Money?

How does all this combine on the same time-axis? Fig. 8 has all relevant monthly variations depicted on the same timeline. We introduce here a monthly storage depletion variable, Dp(t) = St(t-1) - St(t), obtained by subtracting the current month storage from that of the trailing month. A net usage for the current month, i.e. depletion, would be a positive increment in TSCF/month. Conversely, a net increase in the storage would be a negative number. The monthly storage depletion variable, Dp(t), is the yellow curve in Fig. 8, which shows all variations in TSCF/month.

The second variable plotted in Fig. 8 is the net of monthly marketed production minus consumption, N(t) = M(t) - C(t); this is the red line, with winter heating consumption spikes shown as sharp negative bottoms, while the periods of net summer production are the positive intervals in red, with the slight summer dips due to electric generation for cooling. What is very clear from Fig. 8 is that charging and discharging of the storage system is nearly perfectly synchronized with the net production/consumption. Note the near-equality of the areas between the opposing red and yellow curves and the t-axis. During winter, storage is used to supplant the high demand. In summer, storage is charged with the excess production. The red and yellow lines form near-perfect mirror images on the 2 sides of the horizontal t-axis. The 12-month periodicity of the entire system is also highly evident.

To measure the ultimate gain/loss of the system, we finally combined the monthly storage depletion, the production, and the consumption. What emerges is the Deficit line, in green. When Deficit is negative, like during the years up to March 2011, it means the system is in gain, producing more than is consumed. After that date, the Deficit line crosses over to positive, which means we are consuming more than is produced for the market. The Deficit line has some small-amplitude monthly fluctuations, and is increasing linearly on average: Every month, the mean Deficit increases by 3.9 billion SCF/month (BSCF/month). If we consider for a moment storage as reservoir, at its lowest drawn state (March 28, 2014, at 822 BSCF), then one may see about 5% annual depletion.

The increasing Deficit observation above is also supported by the continuous increases in peak consumption rate and peak storage withdrawal rates in TSCF/month. It is evident from Fig. 3 that should shale gas production decline in the coming months, as expected, the aforesaid Deficit would increase considerably.

Deficit, however, is far less frightening than the specter of punching through zero storage (hard bottom) during a winter peak in consumption. We are not so far from it. Looking again at Fig. 7, during the last 3 years, the storage bottom has decreased by about 800 BSCF each year. The last bottom-record, as mentioned above, was 822 BSCF in last March 28.

5.1 The Dynamic NG Price Model

Following the enhanced Hotelling formulation of the earlier article, it seems natural to use actual data for the trailing month to predict the current month price. Of course, we must start at a date where both the current Proved Reserve and consumption rate are known. We start with the consumption rate, as published in recent EIA monthly update. This is shown in Fig. 9. We have also projected the trailing 12-month data forward three full cycles, effectively extrapolating consumption data in BSCF/Month to 2017.

Now we assume that the Proved Reserve is a virtual reservoir, which is drawn down by the amount consumed each month. Proved Reserves data is normally published annually, and may be corrected downward or upward. Since the annual rate of withdrawals (marketed production) and consumption roughly agree, we have seen that on a monthly basis (see Figs. 3 and 6 above) their rates differ substantially. We elected to use the consumption dynamics, which are directly incorporated in all acts of buy and sell (regardless of whether physical or virtual) as the variable representing demand and affecting price. Using the consumption data and the available proved reserves data, we may then plot the remaining reserve as PR(t) - C(t) for the current month, t. This is plotted vs. time in Fig. 10, using our composite extrapolated C(t) of Fig. 9.

The two points over-plotted are actual recent Proved Reserves updates by EIA. We note that although 2 orders of magnitude separate the monthly consumption C(t) and PR(t) initially, the effect over several years is pronounced, and PR(t) declines nearly linearly.

Now, we can complete our dynamic pricing model in the manner outlined above. The projected (forward month) price is calculated as:

P(t+1) = Po * exp(a * t) *A * C(t) / [ PR(t) - C(t) ]

Where Po is the Henry Hub price in $/MBTU at time t=0 (the selected reference time, which in our case is $ 4.49 on 12/31/2010); "a" is the prevailing interest rate at t=0, taken as the yield on the 10-Y Treasury Note; "t" is time in months, and "A" is an initial normalization factor, defined:

A = [ PR(0)/C(0) -1 ].

It is quite obvious that the initial point at P(1)=Po is a miss; however, the idea is to get sufficiently far from the initial conditions. This model contains the actual dynamics of intra-year (monthly) variations, and is compared with the earlier Mean (annual) projection in Figure 11. The Mean formulation is percent GDP-driven, which is held constant; The Mean looks like a reasonable correlation, if an average GDP=1.2% is used. The Dynamic part, P(t+1), shows the periodic behavior typical of consumption, C(t), as well as the amplification due to drawing closer and closer to the reserve bottom.

In Fig. 11, regarding the red P(t+1) curve: aside from matching the initial price to the trailing month HH value, there is absolutely no adjustment done. It should be interesting now to compare the NG price predictions above, with the available monthly closings of Henry Hub, as published by EIA, 2014. This is done in Fig. 12.

Evidently, the initial 2 years of pricing prediction in Fig. 12 are nearly exactly out of phase with HH, as peaks of P(t+1) match bottoms of HH in the same period. Nevertheless, the periodicity in both P and HH are strikingly similar. The prediction improves significantly as we approach 2013/2014. It appears that the market recognizes, perhaps implicitly, the physical drawdown of an exhaustible resource, causing prices to converge on this Hotelling mechanism. In the meantime, our prediction remains a conjecture, although based on the underlying physical argument, that price is driven by demand, and also, when close to depletion, by the amount of stock remaining. Yet, the good fit with HH in 2014 may be just a coincidence, or some funny stochastic process. And having matched the recent-past performance of HH, there is absolutely no guarantee that the future shall be likewise matched.

6. Conclusions

We have made several inconvenient observations.

1. There is a persistent annual deficit in our balance of production and consumption, which may contribute to a consistent reduction in the minimal (bottom) storage over the last 3 years and forward. There are severe winters ahead. According to Dr. Hathaway of NASA, we are currently at the peak of sunspot cycle 24, and from hence to 2020, sunspot activity will decline to near zero. This means an increased likelihood of severe winters. Each of the last three resulted in NG consumption spikes exponentially higher. Closeness between storage bottom and such peak winter consumption may cause a sharp NG pricing spike considerably above what is predicted above.

2. NG shortages are likely. Unfortunately, shale gas operators have turned to more (yet, now far less) lucrative shale liquids production and away from gas. This was due to miserably low pricing for most Marcellus operators, which are away from the major pipeline infrastructure. As a full 40% of the NG in the US is from shale, Houston, we may have a problem. Most operators are actively using hedges to alleviate the pain of dropping NG prices, essentially (taking a page from the master, Aubrey McClendon) betting against their own commodity. This inertia will be difficult to overcome, should demand rise appreciably. Also noteworthy, no R&D ergo no vision for a future.

3. Think LNG imports. Instead of getting gung-ho about conquering the world (now, minus China) with some cheap energy exports, the US should take a hard look at LNG imports on a massive scale. A reverse Cheniere? Yes, and most likely 10 additional import terminals. It is quite right to complete building the 11 or so Australian export LNG terminals. The US should be one big customer. Of course, not at JCC prices.

4. NG prices are set to increase significantly over the coming months. As projected in Fig. 12, $6.58, $5.71, and $5.37 per MBTU at the end of February, March, and April, respectively.

This is the place to point out that trading futures is not for the poor in spirit, (for to them belongs the kingdom of heaven.) It takes balls, experience, clever strategy, and deep pockets to enter the futures market. Please be warned. To the expert, caveat emptor. And to all, good luck.

Disclosure: The author has no positions in any stocks mentioned, and no plans to initiate any positions within the next 72 hours.

The author wrote this article themselves, and it expresses their own opinions. The author is not receiving compensation for it (other than from Seeking Alpha). The author has no business relationship with any company whose stock is mentioned in this article.