The surprising breakdown of the relationship between natural gas production and gas-directed rig count, which has been manifest since the beginning of this year, remains one of the industry's most intensely debated and still unresolved conundrums. The U.S. Lower 48 gas rig count has been on a steady, steep decline for twelve months in a row, sliding from 927 rigs as of 10/14/2011 to 415 rigs as of 10/19/2012, according to Baker Hughes' survey. In defiance, the U.S. dry gas production has remained virtually unchanged since last October, as the Energy Information Administration (EIA) data shows (latest available for the month of July). Given the hyperbolic natural declines in production from shale and tight gas wells, the phenomenon seems very counterintuitive.
There is no shortage of explanations offered as to why the U.S. natural gas supply has so far managed to stay immune to the dramatic cuts in gas-directed drilling. Among various factors, the rig data itself comes into question as a possible source of this divergence.
Traditionally, the industry and investors have relied upon two weekly rig surveys, one of which is provided by Baker Hughes (BHI) and the other by Smith Bits, a subsidiary of Schlumberger (SLB). My review of the Baker Hughes data suggests that the process and methodology used by the survey may in certain instances produce data that is not comparable to prior periods. These effects were particularly noticeable at the end of 2011, the time when the gas-directed rig count started showing fast attrition. While such data inconsistencies in the gas-directed rig count appear to be significant over the past eighteen months, their effect is not large enough, in my estimate, to explain the lack of supply response. Other, more impactful factors, which I will discuss in Part II of this note, need to be taken into consideration.
How Reliable Is The Rig Count Data?
To illustrate some of the issues that investors and analysts face when analyzing rig counts, I turn to Baker Hughes' rig data for the Mississippian Lime play. Baker Hughes shows that during the week ending 11/04/2011 the oil-directed rig count in the play increased by 27 rigs to 43 rigs while the gas-directed rig count dropped by 27 rigs to zero. During that week, the play's rig mix by drilling objective went from approximately 70% gas/30% oil to essentially 100% oil. Of course, this does not mean that all the rigs drilling for natural gas moved overnight to an oily sweet spot. The drilling objectives were simply reclassified from gas to oil, triggering the adjustment in Baker Hughes' rig data.
In this case, the reclassification has operating logic behind it. SandRigde Energy (SD), the most active driller in the Mississippian Lime play, shows in its most recent presentation (October 2, 2012, page 14) that 49% of its production from the play during the second quarter of 2012 was oil. The remaining 51% were split between gas and NGLs. The classification of the drilling objectives as "oil" is fairly straightforward in this instance (even though the 49% oil yield is an average metric, which includes both the wells with high oil content in the oilier parts of the play and the wells with high natural gas content - which could qualify as gas wells for rig count purposes - in the gassier areas).
Under Baker Hughes' survey methodology, a reclassification of a producing area by the operator impacts only the wells in progress (and, very likely, future wells in the same area and same geologic formation), but it does not change the categorization of the wells already drilled. As a result, reclassifications create inconsistencies in the rig count data that may complicate historical comparisons. In the Mississippian Lime example, the drop in the gas-directed rig count due to the reclassification, if taken out of its context, would have the appearance of the gas drilling coming to a halt and would certainly fail to explain the continued steady growth of the gas production from the play. The graph below shows how the Mississippian Lime gas-directed rig count looks on a stand-alone basis - a striking difference from the graph above where a reclassification is more obvious and explains the rig count dynamics.
The Mississippian Lime reclassification does not appear to be an exception (although it is the largest single anomaly that I could identify in my review of the data for almost two thousand rigs over an eighteen-month period). In the Cana Woodford Shale play, during the six weeks from 12/09/2011 through 1/13/2012, Baker Hughes' oil-directed rig count increased from 4 rigs to 19 rigs while the gas-directed rig count declined from 48 rigs to 35 rigs. The rig mix by drilling objective changed from mostly gas to two-thirds gas, one-third oil.
While operators in the Cana Woodford play have been expanding drilling in the oil window, it is implausible that the entire shift happened within just one month. It is more likely that certain areas within the play were re-categorized from gas to oil.
Similarly, by looking at Baker Hughes' weekly rig data for the Eagle Ford Shale play and applying certain screening criteria to single out anomalies, I could identify at least three instances during the last eighteen months when a significant increase in the oil-directed rig count coincided with a similar decline on the gas side.
|Week||Change in Gas Rigs||Change in Oil Rigs|
It is not possible to ascertain the nature of each "kink" in the rig counts without knowing specific rig-by-rig reporting circumstances. Theoretically, it is possible that all of the rigs in question moved from the Eagle Ford's gas window to the condensate or oil window. However, a reclassification of certain areas of the play from gas to oil appears to be a more likely explanation for at least a portion of these data.
A lot more difficult to detect are "creeping" reclassifications when just a few rigs are reclassified at a time, or when a rig finishes drilling a well that is categorized as a gas well and moves to a nearby location, which has the same geological characteristics but this time is designated by the operator as an oil well.
How is the determination between drilling for oil and gas made?
In the case of Baker Hughes' survey, it is the E&P company that makes the call. According to Baker Hughes:
The determination is made by the operating company when the rig permit is issued by the state's permitting authority. The operating company will drill appraisal well(s) to determine the hydrocarbon target. Based on the results, the operator makes a judgment call on how to classify the well. For example, if a well is producing - on a Btu basis - 50% gas; 20% NGLs and 30% oil, it could either be listed as a gas well (gas is the largest component), or an oil well (which is driving the economics). This judgment is solely up to the operator.
For liquids-rich plays in early delineation stage, the uncertainty with regard to the drilling objective - oil versus gas - is normal. The operator often has limited ability to predict a priori the well's EUR and production mix. In addition, the hydrocarbon content may vary noticeably within the play depending on location (e.g., position within the oil, condensate or dry gas windows for which the limits yet need to be determined). As the delineation of a new play progresses and production history is accumulated, it is natural to expect instances when acreage is reclassified for "geological" reasons. The Mississippi Lime play is perhaps one of such examples.
What appears to be a lot more arbitrary, is the operator's ability, under Baker Hughes' surveying method, to designate the drilling objective as "gas" or "oil" based on what "is driving the economics." As a result, the same "combo" well (where all the three components -- gas, NGLs and oil -- are present in the production mix) can be classified as an oil well or gas well, depending on the relative price of the commodities or simply the operator's judgment. This practice is prone with potential inconsistencies in the data as different operators can have different - and changing - internal policies and agendas with regard to how drilling objectives are designated. This approach also introduces a pro-cyclical bias in the oil rig and gas rig counts and makes certain historical comparisons not fully meaningful.
In 2011 and 2012, as natural gas prices plummeted and natural gas producers rushed to demonstrate to investors and analysts the radical shift in capital allocation towards oil and liquids and away from natural gas, the definition of what represented an "oil well" was possibly stretched in more than a few cases by applying the economics driver criterion. Not an insignificant number of wells that are being drilled today "for oil" are essentially similar to the wells that were classified as being drilled "for gas" just eighteen months ago. As visible from many recent E&P company presentations, the "optics" of the oil-versus-gas rig count has become a highly sensitive point for operators. As a result, the possibility of a systemic behavioral bias by survey respondents, which is driven by the commodity cycle, becomes a real issue in rig data interpretation and historical comparisons.
It would be incorrect to conclude that Baker Hughes' surveying methodology is inconsistent. Quite the opposite, the survey consistently relies on operators to provide a real-time assessment of what is the expected production mix from each well being drilled and what drives the operator's decision to commit capital. The approach bears an obvious trade-off. Clearly, operators understand the geology of their properties best, and it is inconceivable that a third party would be able to produce more accurate forecasts with regard to the expected production mix for the over two thousand new wells that are spud every month in the U.S. alone. On the other hand, the loose economic driver criterion, when applied by hundreds of operators, may not produce truly consistent oil rig and gas rig count results throughout the commodity cycle for difficult-to-classify wells.
A relatively small percentage of rigs is impacted by reclassifications
I must emphasize that well classification difficulties arise mostly in the "combo" plays. Based on my review, the rig counts for dry gas and oil areas appear to have been consistent. The two graphs below show rig counts in the Marcellus Shale and Barnett Shale. Both plays have "rich" and "super-rich" gas areas; the Barnett also has a significant "combo" play [familiar from EOG Resources' (EOG) presentations] in which the production stream is almost equally split between oil, NGLs and dry gas on a Btu basis. One would expect that the presence of liquids-rich production stream puts these two shale plays "at risk" for possible reclassifications. However, analyzing Baker Hughes' data for these two plays for the past eighteen months, I could not detect any anomalies that would suggest acreage reclassifications from gas to oil.
The examples and discussion above illustrate that it may be precarious to rely on a simplistic historical correlation between the rig counts and production to derive supply forecasts. To be a meaningful tool in future production estimation, the rig count requires a granular, play-by-play analysis and careful interpretation.
Also, as I will discuss in Part II of this note, it would be incorrect to blame the rig data inconsistencies alone for the failure of the gas rig count as the leading indicator of gas production volumes. I estimate that the combined inconsistency in rig classification between today's data and similar data eighteen months ago in the Baker Hughes survey is less than 150 rigs (and possibly less than 100 rigs). There are several other factors that are in fact more impactful and need to be taken into account.
Part II of this note presents quantitative conclusions of the analysis and offers an explanation of how the "supply vs. rig count" conundrum can be largely explained.
This discussion is fundamentally relevant for natural gas (UNG) and the natural gas producer stocks. My natural gas producer index includes: