Justification

EIA-914 New Methodology.pdf

Monthly Natural Gas Production Report

Justification

OMB: 1905-0205

⚠️ Notice: This form may be outdated. More recent filings and information on OMB 1905-0205 can be found here:

Document [pdf]

Download: pdf | pdf

EIA-914 Monthly Gas Production Report
Methodology
Current as of April 2010
We implemented the methodology described here in April 2010 and applied it historically
to all of 2009 as well as to the current months. Fundamentally there are two parts to the
process: the sampling and the estimation. Both represent changes to previous
methodologies, and each is described separately below.

Sampling Methodology
The EIA-914 survey collects natural gas production volume information on a monthly
basis from a sample of well operators (companies). Production volumes are requested
specifically for Texas, Louisiana, Oklahoma, Wyoming, New Mexico, Federal Offshore
Gulf of Mexico and all Other States (except Alaska). Sampling occurs every month via a
monthly refresh of the sample. The latest available HPDI monthly data are used to select
companies to add to the sample group (HPDI is a commercial vendor of production data).
The sample group of companies changes by 1 or 2 every month. This keeps the sample
current and avoids a major change in the sample caused by less frequent updating. A
cutoff sample based on company production rates is used.

Data Preparation
The HPDI database is used for both the sampling and the estimation processes. HPDI
acquires well or lease level data from State agencies, places it in their own database
format, and sells it. A new HPDI database is acquired every month. HPDI data for five
of the smaller producing States is missing or inadequate. For Illinois, Indiana, Kentucky,
Pennsylvania, and Tennessee, annual production data from the EIA-23 survey (Annual
Survey of Domestic Oil and Gas Reserves) is used to supplement the HPDI database.
Hereafter, references to HPDI data include supplemental data from the EIA-23 survey for
these 5 States.
The monthly production data are split into two parts: the group of companies that
comprise last month’s sample and the group of non-sampled companies. The nonsampled group is sorted largest to smallest based on production from the Lower 48
States. A second sorting for Oklahoma only is also done (more on this below).
HPDI data for the most recent months are usually significantly incomplete. A 6-month
average natural gas production level is calculated by company for the most current 6month period where data are or are nearly complete. The non-sampled companies are
sorted by this 6-month average.

Cutoff Sample

The sampling process uses a cutoff criterion of 20 MMcf/d by company for Lower 48
production. For the State of Oklahoma a cutoff of 10 MMcf/d is used to increase the
coverage in Oklahoma (Oklahoma has an abundance of small companies, so a lower cut
off helps keep the percent coverage up in Oklahoma).

Adding and Dropping Companies
Each month, companies whose surveyed production rate falls below the sample cutoff
point are reviewed to determine whether they should remain in the sample. If a company
is below the cutoff (20 MMcf/d in the Lower 48 or 10 MMcf/d in Oklahoma) for 6
consecutive months it is contacted and asked why. If the decline in production is not
reversible or repairable in the near future, the company is dropped from the survey.
When adding a company, the sorted non-sampled data are used. The highest production
level in the non-sampled group of companies is about 20 MMcf/d. Data for the largest
companies in the non-sampled group can be reviewed quickly and any company in the
non-sampled group that produces more than 20 MMcf/d for 4 consecutive months
(previous to the last few months of incomplete data) is a candidate to be added to the
sample. Production is double checked and confirmed for these identified companies.
Once identified, a selected company is informed with a phone call, and a survey
information packet is sent. The same process is applied to Oklahoma production data
with a 10 MMcf/d cutoff.

Other Ways Companies are Added or Dropped
Mergers and acquisitions, or buying and selling properties can cause a company’s
production level to move above or below the sample cutoff value. An attempt is made to
accommodate the larger events in the sample as soon as possible after they occur. These
larger events usually appear in news reports, newsletters, press releases, industry trade
journals, etc. Minor events involving small companies or small volumes of production
are ignored if they involve only companies in the non-sampled group. Most smaller
mergers and property sales are unknown.

Potential Sources of Errors
Unknown, deficient reporting of, or improperly handled mergers and property sales are
likely the largest cause of sample errors. These events are continuous and make the
sample calibration data (HPDI data) a very dynamic data set. The company production in
the historical HPDI data set must be merged to match the reported sample data every
month. The unknown or missed events are usually small and probably do not contribute
large errors, but it is still possible to miss a larger event. The historical sample data used
to make estimates can be missing production or have too much production and therefore
adversely affect the production estimates because of name changes, multiple name
spellings, companies that report under multiple names in the HPDI data set, the lag
between the time of a merger and the time of its appearance in the HPDI data set, past

multiple mergers, and the potential to improperly assign EIA operator codes. It is
extremely difficult to account for all of the mergers and property sales.

Estimation Methodology
The Simple Ratio Method (SR) is used for all the individual States in the Monthly Gross
Natural Gas Production Report. The SR method allows the use of the most current
historical data available to determine a straightforward ratio. The short lag times mean
that any changes in the sample over the shorter lag will be as small as possible and can
normally be neglected.
For the Other States, the ratio of the EIA-895 survey (Annual Quantity and Value of
Natural Gas Production Report) to the EIA-914 survey annual volumes in the previous
calendar year is applied to the current monthly EIA-914 volumes to calculate the
estimate. This is the same process used previously for the Other States group.

Simple Ratio Method
The SR method uses a ratio of the total production to the current sample’s production at
some point in history. This ratio is then applied to the current reported sample volume to
estimate the current total production volume. The ratio is a 6-month average ratio
calculated at some lag time that varies by State. The time frame for the 6-month average
ratio calculation is moved ahead one month every month so that the lag time is a constant
over time. Lag times vary from 6 to 18 months for the different States. Lags are
necessary because the HPDI data are incomplete in current months. Some States require
a longer lag than others to get back to a time when the data are complete to calculate the
6-month average ratio. Currently a 6-month lag is used for Wyoming, New Mexico, and
Louisiana, 9 months for Texas, 12 months for the Federal Gulf of Mexico, and 18 months
for Oklahoma. At these lag times the reported production in HPDI should be less than
0.5% different than the final reported production. Recent changes in HPDI’s data
collection in Oklahoma may allow a shorter lag time in the future. The equations are as
follows:
 iL5 TPi
 
i  L SP
i
Avg SRiL  
 6










TPesti  Si  Avg SRi L
where:
Avg SR =
TP
=

Simple Ratio, 6-month average
Total Production, from HPDI

L
TPesti
Si
i

=
=
=
=

Sample Production, current sampled group of companies historical
production, from HPDI
Lag time in months
Total Production estimate for the current month
Sampled production for the current month
Current or estimation month.

Potential Sources of Errors
This method is a significant improvement over previous methods in making use of
information that is as current as possible. Even so, the historical data still have some lag
and so some potential for error remains. For example, the rapid development of the
Haynesville shale in Louisiana caused a change in the State production trend that, in turn,
may cause the simple ratio method to over estimate. Also, a sample affected by
improperly handled mergers or property sales, as described above, can adversely affect
the production estimates. EIA continues to look for better ways to handle the mergers
and property sales.

Supporting Documents
A brief comparison with the previous method is contained in Attachment 1. A full
description of the previous methodology used prior to April 2010 can be found in
Attachment 2. A document describing all of the methods considered and their testing can
be found [here]. The new methodology was applied to 2009 and the difference from the
previous estimates is described in this [document]. Lastly, the paper submitted by ICF
International, the company that performed the outside review of the 914 program, is
found [here].

Attachment 1
Brief Comparison with the Previous Methodology
The fundamental change in the new process concerns timing; both the currency of the
historical data used and the frequency of updating are improved in the new methodology
for both the sampling and estimation processes. The previous estimation methodology
had 2 main deficiencies also related to timing:



It depended on historical data that was too far in the past to adequately represent
the current situation. Because the previous methods were anchored to data far in
the past, they performed poorly when the industry changed rapidly.
Annual updating in January could cause a disconnect between December and
January. Sometimes the reported change in the production estimate from
December to January was caused by the change in the process rather than an
actual change in the data.

The Simple Ratio method and the monthly refreshing of the sample addressed these 2
deficiencies. Where the previous estimation method depended on data that was 2 to 7
years old, the new SR method now uses historical data that is only 6 to 18 months old and
is updated monthly instead of annually. The monthly refreshing of the sample with
monthly data that is as current as possible is an improvement over annual updating with
data that is 2 years old. Rather than a change in the sample of 25 operators in January
(roughly 10 percent), monthly updating yields a change of 1 or 2 operators each month
and keeps the sample up to date. A full description of the previous methodology follows
in Attachment 2.

Attachment 2
Form EIA-914
Monthly Natural Gas
Production Report
Background and Methodology
September 2005

Contents
Background
Introduction, Purpose of Survey, Description of Survey, Sampling Frame, Form and
Instructions, Response Rates
Methodology
Gross Production Estimation for the Six Areas, Gross Production Estimation for “Other
States,” Imputation, Editing and Data Review, Revision Policy, Analysis Plan

Background
Introduction
Starting with the January 2005 report month, EIA began collecting monthly natural gas
production information from well operators using a new survey, Form EIA-914,
“Monthly Natural Gas Production Report.” This report describes the background,
purpose, technical methodology and initial results of the survey. Although data from this
survey are being collected and posted on the EIA Website, the new data series has not
replaced natural gas production data series currently being published by EIA (in the
Natural Gas Monthly, the Natural Gas Navigator, the Monthly Energy Review and other
EIA publications). When monthly natural gas production volumes based on the EIA-914
data are considered reliable, they will replace the current data series and become the
official EIA natural gas monthly production data series. This is expected to occur by the
end of the 2005 report year (March 2006).
Currently EIA publishes estimates of natural gas production based on data supplied by or
collected from individual State agencies and the Minerals Management Service. Because
these production estimates were not considered sufficiently timely or accurate to meet
customer needs (to understand and resolve natural gas supply issues) EIA obtained
approval from the Office of Management and Budget (OMB) to implement the new
survey, EIA-914, “Monthly Natural Gas Production Report,” which collects production
data directly from well operators.
Purpose of Survey
The purpose of the EIA-914 survey is to collect more reliable and timely monthly natural
gas production information for the Lower 48 States and six States or regions (Texas,
Oklahoma, Louisiana, Wyoming, New Mexico and the Federal Offshore Gulf of
Mexico). The goal is to provide accurate information not more than 60 days after the

close of a report month. Current EIA monthly natural gas production estimates generally
aren’t available until about 120 days after the close of a report month, and even these
estimates do not always accurately depict the levels of production or directions of monthto-month changes. These estimates are generated using a variety of different data sources
and procedures, which are described in How EIA Estimates Natural Gas Production.
Description of Survey
The EIA-914 survey collects natural gas production volume information on a monthly
basis from a sample of well operators. Production volumes are requested specifically for
Texas, Louisiana, Oklahoma, Wyoming, New Mexico, Federal Offshore Gulf of Mexico
and all Other States (except Alaska). Two volumes are requested from respondents:
(1) “gross withdrawals (wet),” which is full-bore well stream gas minus lease
condensate, oil, and water; and
(2) “natural gas lease production,” (sometimes referred to as “sales production” or
“gas available for sales,”) which indicates the net amount of produced gas that
leaves the lease to go to natural gas processing plants or directly to end-users. 1
The two volumes reported on the EIA-914 are illustrated in the diagram in Figure 1.

Natural gas lease production does not include gas used as fuel on the lease, but the quantity “marketed
production” currently being published by EIA does.

W e lls
( G a s , O il, C o a lb e d
M e th a n e )

F ig u r e 1 . N a tu r a l G a s
P r o d u c tio n S tr e a m

F u ll W e lls tr e a m s

C r u d e O il

L e a s e S e p a ra to rs

L e a s e C o n d e n s a te
W a te r

G ro s s

W ith d r a w a ls
(W e t)

R e p r e s s u r in g a n d R e in je c tio n

L e a s e F a c ilitie s
E x c lu d in g N a tu r a l G a s
P la n ts

V e n te d a n d F la r e d
Fuel U sed on Lease
N o n h y d ro c a rb o n s R e m o v e d o n L e a s e

N a tu ra l G a s L e a s e
P r o d u c tio n

Sampling Frame
The EIA-914 cut-off sample is selected from the sample of approximately 1,500
operators selected to respond to Form EIA-23, “Annual Survey of Domestic Oil and Gas
Reserves.” The list of 1,500 operators contains the top producing gas companies in the
United States. The EIA-23 sampling frame contains approximately 15,000 potential
respondents, which are drawn from a master list of nearly 21,000 well operators
maintained by EIA. Natural gas producers comprise a highly skewed and volatile
industry, with a small number accounting for the majority of the natural gas production in
the United States. There are approximately 280 respondents to the EIA-914. This number
can change over time as companies merge, buy and sell properties, or go out of business.
The initial cut-off sample was chosen to yield at least 85 percent coverage for each
surveyed area and 90 percent coverage for the Lower 48 States.

Form and Instructions
The EIA-914 survey form is one page with three pages of instructions. It is similar in
format to other Oil and Gas surveys. The form was designed based on direct input from
potential respondents during pre-survey design visits and the quality and usability of the
survey instrument were tested during cognitive testing visits conducted by survey
methodologists in EIA’s Statistics and Methods Group. Based on results from these tests,
the form includes some imbedded instructions, intended to help respondents complete the
form without the need to refer to instructions located elsewhere.
The form and instructions are e-mailed to respondents (formats include PDF and XLS)
and they are also available on EIA’s Website. The respondents can return forms by fax,
e-mail, secure file transfer or conventional mail. For the months of January through
March of 2005, respondents were expected to report within 60 days after the close of the
reporting month. Beginning with the April 2005 report month, respondents are expected
to provide EIA with their data 40 days after the close of the report month. Respondents
are encouraged to provide reasonable estimates if necessary to meet the deadline and to
report zeros when there is no production to report. This reduction in turnaround time is
necessary to produce more timely estimates of natural gas production for EIA customers.
Response Rates
Response rates for the EIA-914 survey have been excellent. The production weighted
response rate for each month’s estimate (January through June 2005) is 100% for all
areas except Oklahoma, which has been 99.8% every month since January. This has
resulted in a response rate of 99.9% for the Lower 48 States each month. Figure 2 shows
the response rates for March, April, and May. March was the last month the respondents
were given 60 days to respond. At the 60 day due date for March data, the production
weighted response rate was 56.9 percent. However, just three days later the response rate
rose to 98.5 percent. April was the first month respondents were given 40 days to
respond, and the production weighted response rate was 15.9 percent at the 40 day due
date for April data. However, the response rate for May data at the 40 day due date was
85.8 percent, and high response rates have been the rule ever since.
EIA has encouraged respondents to submit their best estimates in order to meet due dates,
believing that a respondent’s estimate of their production level is more reliable than
EIA’s imputed value for their production level. Respondents are required to submit data
revisions if the revised data and the data originally provided differ by more than four
percent. However, revisions are encouraged for differences less than four percent.

Figure 2. Production Weighted Response Rates - March, April, and May 2005
100
Present due date, 40 days

May
85.8%

May
99.3%

March
99.4%

April
98.5%

Percent

80
70

(March) Production Weighted
Response Rate

(April) Production Weighted
Response Rate

(May) Production Weighted
Response Rate

March
56.9%

40
Jan-March due date, 60 days

30
20

April
15.9%

10
0
0

Days Measured From End Of Production Month

Methodology
This section describes the data estimation methodology used to estimate total production
from respondent data, as well as data imputation and editing techniques, and data revision
policy.
Gross Production Estimation for the Six Areas (Texas, Louisiana, Oklahoma, Wyoming,
New Mexico, and Federal Gulf of Mexico)
A preliminary estimate of the final Total Gross Production Rate for each area is based on
data provided by a cut-off sample of all operators for the data month. The cut-off sample
was selected based on data for 2003. The preliminary total estimate will be made for
each month in 2005 by collecting gross production data from the sampled operators for
the data month, dividing by the number of days in a month to obtain an estimate for the
gross production rate in billion cubic feet per day, and multiplying by an inflation
factor, f t .
The value of f t can be determined using the classical Ratio Estimate Method for any area
and time period for which the historical data are essentially complete, (Brewer,
“Combining Survey Sampling Inferences: The Weighing of Basu's Elephants,” Arnold:
London, 2002).
[1]

Tt  f t S t ,

where,
Tt = Total Gross Production Rate (bcf/day) in data month at time t (middle of a month
in 2005),
S t = Gross Production Rate (bcf/day) reported by sampled operators in data month at
time t
(middle of a month in 2005), and
f t = Inflation Factor used to estimate Total Gross Natural Gas Production Rate at time
t.
From [1], the inflation factor is
[2]

ft 

Tt
St

The ratio estimator, typically used for estimation with a cut-off sample, assumes that the
sample coverage remains constant over time.
[3]

T 
Tˆ R t   xx  S t  f xx S t ,
 S xx 

where
Tˆ R t = Standard Ratio Estimator for Total Gross Natural Gas Production Rate (bcf/day)
at time t
Txx = Total Gross Production Rate (bcf/day) in calibration year xx

S xx = Aggregate Gross Production Rate (bcf/day) reported by sampled operators during
calibration year xx . The sample is selected to achieve a specified coverage rate during
year xx .
S t = Aggregate Gross Production Rate (bcf/day) reported by sampled operators at time
t , and
t = Time of current survey month, measured as number of months from the middle of
calibration year. For example, if the calibration year is 2003, then t =18.5 for January
2005 and the middle of calibration year is t = 0.
IHS, Inc. is the source for historical monthly production data used to calibrate the EIA914 gas production estimation method. These data are for gross production as defined and
collected by the States and Mineral Management Service of the U.S. Department of
Interior and are available at the company operator level. These data were very close to
complete (final) for 2003 when this methodology was calibrated.
During the development of the estimation methodology, it was observed that the
population of natural gas operators was very dynamic. For example, companies selected
in a cut-off sample in 2001 that had 87 percent coverage had less than 85 percent
coverage by December 2003 (Figure 3). Correspondingly, the share of total production

that the non-sampled operators in a given calibration year represent increased over time.
As an example of how volumetric shares from non-sampled operators change over time,
the smallest operators (i.e., the ones that produce the bottom 10 percent of total volume in
the calibration year) may be considered as representative of at least some portion of the
non-sample group. These operators added 1.2 percent of total production in 2 years while
operators that were not producing in the calibration year had 0.7 percent of the
production after 2 years, as shown in Figure 4. While these operators may not all be
outside the sample, these results are considered indicative of trends that are expected to
prevail for non-sample operators.
Figure 3. Top 87% of Texas Operator's Production
For Calibration Years 1997 - 2003
88

Percent of Total Production

85
1997
1998
1999
2000
2001
2002
2003

80
Jan97

Jul97

Jan98

Jul98

Jan99

Jul99

Jan00

Jul00

Jan01

Jul01

Jan02

Jul02

Jan03

Jul03

Jan04

Jul04

Figure 4. Texas Smallest 10% (5473 Operators) Split into Producing and NonProducing Groups, Plotted as the Change from the Calibration Year Average
1.4

1.1

4,158 Small Operators had 10% of Texas
Total Production in 2000.

0.9

The 4,158 Operators increased
their share of the total to 11.2%
through 2002 adding 1.2% of
growth in the Total Production

Percent

0.7

Producing, > 0 mcf/d in 2000 (normalized)

0.5

Non-Producing, = 0 mcf/d in 2000
0.2

0.0
139 Operators started
production
in 2001, adding almost 0.7%
of the total in 2 years

-0.2

-0.5
Jan-00

Jul-00

Jan-01

Jul-01

Jan-02

Jul-02

To account for the observed decline in coverage over time, the “Adjusted Ratio Method”
was developed, which is a modification of the Ratio Estimate Method in [3], as follows:
A general linear regression model for f t is
[4]

f t = b0

 b1 (t )   t ,

where V  t    2 . If sample coverage were constant over time, then bˆ1  0 and fˆt =
T
f xx = xx = bˆ0 .
S xx
Equation [4] can be expressed as
[5]

ft 

Txx
+ bxx  2 (t )   t
S xx

The mean square error (MSE) allows direct comparisons between different
methodologies for natural gas production estimates. After Battaglia, “Mean Square
Error,” AMP Journal of Technology, v.5, June 1996, the MSE is defined in Equation [6]
as:

[6]

1 m ˆ
MSE   (Tt  Tt ) 2 ,
m m 1

where
Tˆt = Estimated production rate for month at time t ,

Tt = Production rate for month at time t ,
m = number of months since t.
The root mean square error as a percent (RMSEP) is defined as:

[7]

 1 m
(Tˆt  Tt ) 2


m m 1
RMSEP = 

Tt





 (100)




A test close to the actual task of estimating monthly 2005 production calibrated to 2003
production would be to use 2001 as the calibration year, estimate total monthly natural
gas production rates for 2003 and compare the results to the estimates made using the
Adjusted Ratio Estimate Method. The Ratio Estimate Method would have estimates
given by:
[8]

T
Tˆt R  fˆt R ,01 S t = 01 S t
S 01

where t is measured from the middle of 2001 and for January 2003, t = 18.5 and for
December 2003, t = 29.5. The Ratio Estimate Method errors are shown in Figure 5
(lower curve). The January error is negative and the errors get more negative during the
year.
This result is consistent with the Texas monthly production data shown previously in
Figure 3 from a sample of operators that had 87 percent of the production in each of the
calibration years from 1997 through 2003. The operator sample that averaged 87 percent
of the production in 2001 had only about 84.5 percent of the production in December
2003.
Using the Adjusted Ratio Estimate Method, the adjusted estimator of the Total Gross
A
Production Rate, Tˆt , for months in 2005 and calibration year 2003 follows:
[10]

T

A
Tˆt  fˆt S t   03  bˆ05 t  S t ,
 S 03


where
Tˆt At = Adjusted Ratio Estimator for Total Gross Natural Gas Production Rate (bcf/day) at
time, t which accounts for declining coverage of sample.
fˆ = Inflation Factor used to estimate Total Gross Natural Gas Production Rate at time,
t

t.
The errors using the Adjusted Ratio Estimate method are shown in the upper curve in
Figure 5.
Figure 5. Comparison of Errors of the Ratio Estimate Method and
Method, Texas 2001 Calibration
4.0

3.0
2.0

Linear Model - 2001
Production Estimates
Average Error = 0.11%
Minimum Error = -0.09%
Maximum Error = 0.39%
Average Absolute Error = 0.14%
MSE = 0.03

Linear Model - 2003
Production Estimates
Average Error = 0.05%
Minimum Error = -0.46%
Maximum Error = 0.78%
Average Absolute Error = 0.26%
MSE = 0.12

1.0
Pe
rce
nt 0.0
Err
or
-1.0

Ratio Method
Linear Model

-2.0
-3.0

-4.0
-5.0
Jan-01

Ratio Estimate Method - 2001
Production Estimates
Average Error = 0.11%
Minimum Error = -0.30%
Maximum Error = 0.66%
Average Absolute Error = 0.28%
MSE = 0.12
Apr-01

Jul-01

Oct-01

Jan-02

Ratio Estimate Method - 2003
Production Estimates
Average Error = -1.78%
Minimum Error = -2.69%
Maximum Error = -0.65%
Average Absolute Error = 1.78%
MSE = 3.51

Apr-02

Jul-02

Oct-02

Jan-03

Maximum Error = 2.69%

Apr-03

Jul-03

Oct-03

If t =0, the time associated with the 2003 average annual production rates T03 and S 03 ,
then Equation [10] reduces to the standard ratio estimator of Equation [3]. However,
unlike Equation [3], Equation [10] incorporates an inflation factor that increases as the
time from the sample selection increases.
As noted above, data from the 2003 calibration year were used for selecting the sample
for 2005 because they were considered the most recent complete data available. The
estimated bˆ05 , reflects the monthly decline in coverage from the calibration year, 2003, to
the survey month in 2005.

The linear model of Equation [5] was used in the least squares estimating procedure with
T
T
T
the ratios, 99 , 00 , and 01 for calibration years, 1999, 2000, and 2001, and the
S 99 S 00
S 01
corresponding monthly historical production data, S t , for t =18.5 to t =29.5 for those
three calibration years (monthly data from 2001 through 2003). The estimated
coefficient, bˆ05 , in Equation [10] has a subscript, 05, to reflect the fact that it will be used
to estimate the gross natural gas production rate for the EIA-914 during 2005.
Figure 6 shows the error plots for the Ratio Estimate Method and the Adjusted Ratio
Estimate Method.
Figure 6. Texas Percent Errors for 42
Adjusted Ratio Method - 3 Calibration Years - 2 Year Out
4.0

3.0

2.0

Adjusted Ratio Method
Average Error = -0.23%
Minimum Error = -1.07%
Maximum Error = 0.40%
Average Absolute Error = 0.30%
MSE = 0.15

Ratio Estimate Method
Average Error = -1.39%
Minimum Error = -2.69%
Maximum Error = -0.54%
Average Absolute Error = 1.39%
MSE = 2.18

Pe 1.0
rce
nt
Err
or 0.0
s
-1.0

-2.0
Adjusted Ratio Method

-3.0

Ratio Method
-4.0
Jan-01

Jul-01

Jan-02

Jul-02

Jan-03

Jul-03

Jan-04

Jul-04

Once a year, a new sample will be selected for the next calibration year based on the
current EIA-23 survey data and new model parameters will be estimated. The sample for
use in 2006 will be based on calibration year 2004 and the parameters will be estimated
using the three calibration years, 2000, 2001, and 2002 and the corresponding monthly
historical production data, S t , through 2004. Table 2 shows the sample coverage for
each region.

Table 2. Initial Sample Coverage and Parameter Estimates
State

S 03

FG
LA
NM
OK
TX
WY
OT

T03

T03
.9756
.8847
.9240
.8434
.8681
.9694
.8604

S 03

 b0

1.02398
1.12890
1.07929
1.18485
1.15151
1.03117
1.16225

b1
0.00085
0.00288
0.00057
0.00039
0.00067
0.00059
n/a

Gross Production Estimation for “Other States”
A Ratio Estimate Method and an Adjusted Ratio Estimate Method (that accounts for
declining coverage) are also used to obtain gross production estimates for Other States.
The ratio estimator for gross natural gas production for Other States follows the same
form as for the six areas.
As above with calibration year 2003, the standard ratio estimator for gross natural gas
production for Other States is given by:
[11]

T 
Tˆ RO t   03  S t  f 03 S t ,
 S 03 

T03

is estimated using data from the EIA-23 survey, “Annual
S 03
Survey of Domestic Oil and Gas Reserves,” on total gas production (wet after lease
separation) because company level data on gross production were not available from IHS,
Inc. The company level historical production data by month were not available for nine
of the States that compose Other States. These States account for about 9 percent of the
gross production in Other States.
However, the ratio

It is assumed that the percentage of wet gas produced by the sampled operators based on
the EIA-23 survey is a good approximation of their percentage of gross gas produced.
The equation for the ratio estimator for calibration year 2003, which accounts for the
decline in sample coverage over time in Other States, is given by:
[12]

T 
AO
Tˆt  fˆt S t   03 Oˆ t S t
 S 03 



Ot is based on the results of the Adjusted Ratio Estimate Method described above for
Louisiana, New Mexico, Oklahoma, Texas and Wyoming.

[13]

Tˆt ,ALA  Tˆt ,ANM  Tˆt ,AOK  Tˆt ,ATX  Tˆt ,AWY
ˆ
Ot  R
Tˆt , LA  Tˆt ,RNM  Tˆt ,ROK  Tˆt ,RTX  Tˆt ,RWY



Only these 5 States were used to calculate Ot , because the Federal Gulf of Mexico did
not correlate as well historically with the Other States and had dramatically different
changes in production between 2003 and January 2005. These rates and percent changes
are shown below in
Table 3.

Table 3. Percent Change 2003 Rate to January 2005 Rate
R