This post was written by Dennis Coyne and does not reflect the opinion of Ron Patterson.
I have considered an alternative way of estimating Texas oil (C+C) output using the Drilling info data provided in the EIA’s 914 monthly production reports.
The Texas estimate is a weak part of the EIA’s estimate for US C+C output. In the chart above I show the EIA’s most recent monthly estimate from the Petroleum supply monthly and compare with an alternative estimate that substitute’s my best estimate for EIA’s TX C+C estimate. The slope of the trend line needs to be multiplied by 366 to give the decline at an annual rate, for the EIA estimate it is 528 kb/d per year, and for the alternative estimate it is 364 kb/d per year.
The decline of 222 kb/d in April 2016 was due to a 66 kb/d decline in North Dakota, a 58 kb/d decline in Federal Offshore, a 47 kb/d decline in Texas, and a 22 kb/d decline in Alaska, the total for these three states plus offshore accounted for 87% of the decline in US output. Those same areas produced about 71% of US C+C in April 2016 based on EIA estimates.
New Texas C+C Estimate
I have downloaded EIA 914 data from June 2015, May 2016 and June 2016 only and have used those sets of data to estimate Texas oil output. Two methods were used and the average result of the methods is my “corrected” estimate. This corrected estimate was combined with the average of Dean’s corrected estimates from March and April 2016 (most recent two months), which I call the “2 mo” estimate. The combined estimate is my “best” estimate for TX C+C. Further details on the estimation method are provided in an Appendix for those interested.
In the Chart above Dean’s estimate and the EIA estimate are shown for comparison and my corrected estimate and the 2 mo average estimate using Dean’s correction factors from March 2016 and April 2016 are shown with dashed lines. I expect the final Texas C+C values will be somewhere between these two estimates and the “best” estimate is simply the average of those two estimates. More data will be needed to refine these estimates further. The slope of the linear trend must be multiplied by 366 to get the annual rate of decline, which is 158 kb/d per year or 4.6% per year for the best estimate. If we only consider the past 12 months, the decline rate is 2.7% per year.
Permian and Eagle Ford Output
Many experts believe that Texas output must be decreasing rapidly, my non-expert opinion is that Texas output has been declining relatively slowly over the past 12 months.
One confirmation of this thesis is found by considering RRC data for the Eagle Ford and Permian basin.
I estimate the percentage of statewide Texas C+C from the Permian basin and Eagle Ford using RRC data. I then multiply the percentage of total output by the EIA estimate and by my “best” estimate to find the Eagle Ford (EF) and Permian estimate of C+C. Those charts are below.
Some have argued that the fall in the number of oil well completions in Texas must have led to a fall in output.
I agree in principle, but the question is how much has output fallen and how have the number of oil completions changed in the Permian and EF. The chart below shows the number of oil completions in the Permian (districts 7C, 8 and 8A) and the Eagle Ford (districts 1, 2, and 3). Note that in March 2016 about 89% of Texas output was from the Permian and Eagle Ford and in Jan 2015 86% of Texas C+C output was from the Permian and EF (output was between these two levels for all months in between).
Chart below shows oil completions in the two plays.
Looking at the chart above we would expect that oil output should have been relatively flat in the EF through April 2016 and that Permian output should have fallen dramatically from Sept to Dec 2015.
Instead we see EF output falling and Permian output rising. How can this be?
I propose that the dramatic decrease in completions in the Permian was mostly vertical well completions. If the level of Permian oil completions fell from roughly 650 oil completions to about 450 oil completions and 200 of the 650 completions had been vertical completions and at 450 oil well completions there were zero vertical completions, this would be similar to a drop of 50 horizontal completions.
This is because on average in the Permian basin, a horizontal completion produces about 4 times as much oil as a vertical completion. This explains why there would be less of a drop in output, but not an increase in output.
An increase in output would be explained by a drop from 300 vertical wells and 350 horizontal wells in the case when there were 650 completions to zero vertical wells and 450 horizontal wells. The 300 vertical wells would have produced about as much as 75 average horizontal wells, so the net for 450H wells would be similar to an increase of 25 horizontal wells compared to the 300V + 350H well scenario.
Also worth considering is that the flatter output curve for the Eagle Ford using the “best” estimate rather than the EIA estimate matches better with the completion history for oil wells in the Eagle Ford.
Appendix- New correction method using Drilling Info data.
What follows will be of little interest to those who do not care for mathematics, and those who are adept will find the presentation appalling. It is intended for people with little math(s) background, basic algebra would be plenty, in fact only arithmetic might be enough.
Correction Method 1
The first method is similar to the method Dean introduced, but uses Drilling Info Data instead of RRC data, the primary difference is that the Drilling info data includes “pending file” data that is not included in the freely available RRC PDQ database. The EIA claims this data is within 0.5% of the final value within 5 months. A snippet of the spreadsheet is shown below.
The correction factors (T, T-1, … , T-4, T-5) are added to the June 2016 Drilling info data (DI Jun16) data column to get a correction1 estimate. T is added to Mar16, T-1 to Feb16, etc. This only gives us an estimate through March 2016, for April 2016 we assume the difference between the EIA 914 estimate and the corrected estimate in March will be the same in April. The EIA 914 survey estimate is 2958 kb/d in March 2016 and the corrected estimate is 3056+356=3412 kb/d, the difference is 3412-2958=454 kb/d. The April 2016 EIA 914 data point is 2901 kb/d and our corrected estimate is 2901+454=3355 kb/d. Chart below shows the “corrected 1” estimate as well as the EIA estimate, Dean’s estimate and the June 2016 Drilling Info data (DIJun16).
Correction Method 2
The second correction method uses Drilling info data from June 2015 and compares with the average of the May and June 2016 drilling info data. This was done because the difference between the May and June 2016 data fluctuates from positive to negative differences over time so that it is not clear which is the better estimate from June 2013 to June 2015 so the average was used. Chart below shows data and initial correction factors.
In this case the initial correction factor is the difference between May and Jun 2016 average and the June 2015 data. These initial correction factors are applied to both the May and June 2016 data to see how they compare, in this case 15 correction factors are used (instead of only 6 correction factors in method 1). Chart below shows correctedJun16 and correctedMay16 and the difference between the corrected estimates on the right axis.
The differences are small, between -20 kb/d and +20 kb/d except for the final month where the difference becomes quite large (135 kb/d). The correction factor for month T is adjusted to account for this anomaly by increasing it by 135 kb/d.
The “new” correction factors are applied to the May and June 2016 data in the chart below with the difference shown on the right axis.
The “newJun16” and “newMay16” corrected data agree well and similar to in method 1 we assume the difference between the March 2016 corrected data and EIA 914 survey data (462 kb/d in this case) remains the same for April 2016, the April 2016 corrected estimate is 3363 kb/d.
The average of the newMay16 and newJun16 corrected estimates plus the April 2016 estimate above are the corrected 2 estimate. The method 1 and method 2 corrected estimates are then averaged to find an “average” estimate as shown in the chart below.
I rename the “average” line in the chart above to “corrected” and this is combined with the March 2016 and April 2016 average correction factors using Dean’s method (RRC data only) which is labelled “2 mo” on the chart below. The average of the 2 mo estimate and the new “corrected” estimate using the DI (drilling info) data is the “best” estimate. Dean’s estimate and the EIA estimate are included for comparison. The slope of the trend line for the best estimate from May 2015 to April 2016 is 92 kb/d per year about 2.7%/year. (Slope needs to multiplied by 366 to get decline in kb/d per year, average output for past 12 months for best estimate is 3441 kb/d.)