‹ Weekly Natural Gas Storage Report

Discussion of the Sample Reselection for the Weekly Natural Gas Storage Report

Released: August 16, 2018

Overview

Consistent with standard statistical practice, the U.S. Energy Information Administration (EIA) has selected and tested a new sample for the Weekly Natural Gas Storage Report (WNGSR). The transition to estimates based on the new sample will occur in two steps. First, using the new sample, EIA will provide historical revisions to WNGSR estimates for the most recent eight weeks and publish them on Monday, September 10, 2018, outside of the regular publication schedule. Then, on Thursday, September 13, 2018, EIA will begin publishing estimates using the new sample in the WNGSR during the regular publication cycle.

EIA samples are reselected periodically as a standard practice using an established methodology. Although some of the companies selected for the weekly sample will be different in this new sample, the established sampling and estimation methodologies for the WNGSR have not changed.

This report provides information about estimates based on the new sample, compares these estimates with published values for working gas stocks based on the old (2015) sample, and details the process for transitioning to the publication of estimates based on the new (2018) sample.

Highlights of this report include:

  • Using available data since February 2018, differences between the two sets of estimates of working gas stocks based on the new (2018) sample and the old (2015) sample are small and not statistically significant at the 5% level of significance.
  • For regions other than the Pacific region, the operators common to both the old (2015) and the new (2018) samples account for at least 93% of the total volume for the sample in each region, based on the measure of size used to select the new (2018) sample.
  • Because of the relatively small number of operators (eight) in the Pacific region, estimates of working gas stocks in that region will be based on weekly reports from all of them, i.e., a census data collection, for the first time.
  • To help our data users prepare for the transition to this new sample, EIA will revise the estimates for/backcast eight prior weeks using the weighted average of estimates produced from the new sample and the old sample before we begin publishing WNGSR estimates solely based on that new sample. EIA will publish these revised/backcasted weekly estimates on September 10, 2018, covering the eight weeks of July 13, 2018 through August 31, 2018. However, during the transition period, published working gas estimates in the WNGSR will continue to be based solely on the old (2015) sample. Revised/backcasted transitional estimates will use a weighted average combination of estimates produced from the old (2015) and new (2018) sample and will be released as revisions to the Weekly Natural Gas Storage Report historical database on Monday, September 10, 2018. At that time, EIA will also publish revisions to estimated measures of sampling variability for the transition period.
  • Working gas estimates published on Thursday, September 13, 2018, for the week ending September 7, 2018, will be based solely on data from the new (2018) sample.

Sample Reselection

EIA survey samples are reselected periodically as a standard practice. The validity of the results from a sample depends on the completeness and accuracy of the frame from which the sample is drawn. The frame for the EIA-912 survey, the Weekly Underground Natural Gas Storage Report, is based on companies reporting data on Form EIA-191, Monthly Underground Natural Gas Storage Report. The Form EIA-191 is completed by all operators of underground natural gas storage fields in the United States. Data are provided for storage volumes in each storage field and reservoir. The sample for the EIA-912 survey is prepared by aggregating volumes of working gas in storage by storage operator and by region.

Over time, the size and number of the storage fields operated by a given company change, and operators enter or exit the industry. As these changes occur, eventually the initial samples drawn from that frame may no longer be fully representative of the current industry. This situation is addressed most often by selecting a new sample from a more current frame, which better reflects the current industry. Selection of a new sample does not necessarily require or reflect a change in sampling or estimation methodology.

The sampling procedure for the WNGSR uses a stratified sample design with a probability proportionate to size (PPS) sample selection and a measure of size equal to the average between an operator’s most recent October and March monthly working gas stocks at the time of the sample selection. The six strata are based on geographic region and field type (i.e., salt or nonsalt): East, Midwest, Mountain, Pacific, South Central-Salt, and South Central-Nonsalt.

Reported volumes of working gas in storage as reported on the Form EIA-191 are aggregated by storage operator and by stratum to prepare the sampling frame for the EIA-912 for each region. For each stratum, the data are divided into two groups. The first--the certainty group—consists of the larger operators in the region as well as all operators with storage fields in more than one stratum. All operators in the certainty group are selected for the sample. The second—the noncertainty group—consists of all other operators, who were selected for the sample with a probability proportional to their size.

Old (2015) sample

In April 2015, EIA implemented a sample design for WNGSR based on a new five-region format for the Lower 48 states, replacing the three-region format that was previously in use. Data collection for the new five-region sample began with reports for the week ending April 10, 2015. EIA conducted internal testing and validation of the new sample between April and November 2015 before publishing estimates on the five-region basis. The methodological changes were announced and documented in August 2015. In addition, EIA published a data history from 2010 to 2015 to help data users get accustomed to the new data series.

New (2018) sample

In October 2017, EIA selected a new sample, in accordance with the established WNGSR sample methodology. There were 85 operators selected for the new (2018) sample from a sampling frame of 134 operators in the Lower 48 states, compared with 79 operators that were selected for the old (2015) sample. The net effect of this sample reselection was that 17 operators were added and 11 operators were dropped.

Operators in the 2018 sample account for 97% of the total measure of size (average of October 2016 and March 2017 values of working gas storage in the Natural Gas Monthly) in the Lower 48 states and range from 96% in the Midwest region to 100% in the Pacific region. The table below shows the sample coverage rates in terms of the measure of size by region.

Table 1. Sample coverage rates for the new (2018) sample
Region Sample Coverage Rate
East 97%
Midwest 96%
Mountain 97%
Pacific 100%
South Central 97%
South Central (Salt) 97%
South Central (Nonsalt) 98%
Source: U.S. Energy Information Administration

The methodology for selecting the new sample (2018) is similar to the process used to draw the selected operators in the old (2015) sample. However, three new features are included with the implementation of the new (2018) sample, and more details on the first two features are given in the subsections that follow:

  • The new (2018) sample will survey all eight operators in the Pacific region each week, increasing coverage in that region to 100%.
  • Data collection from the South Central region operators will explicitly separate salt and nonsalt volumes in the survey submissions.
  • In the WNGSR, EIA will perform independent rounding for all published weekly estimates of working gas stocks. Currently, weekly estimates of working gas stocks for the Lower 48 states are derived by summing the corresponding weekly estimates by region.

Census of the Pacific Region

Because of the small number of operators, with differing behavior, in the Pacific region, EIA has opted to survey all of them (census) with the implementation of the new sample rather than conduct a sample as is done in the other regions. Only eight operators exist in the entire region, which ranges from Southern California to the Pacific Northwest, including merchant operators and utilities.

The small number of operators in this region limits the number of available selections that can occur. The established sampling methodology uses a size cutoff to determine the largest operators selected for the sample with certainty, and the remaining companies are sampled with probability proportional to size. Under the established sampling methodology, four operators would have been included in the sample with certainty because they exceeded the size cutoff for the region. Two of the remaining four operators would have been included as noncertainty companies that were selected with probability proportional to size sampling. To reduce sampling error, and to remain consistent with the established sampling methodology, EIA opted to collect data from all eight operators in the Pacific region.

Although a census survey may be subject to other sources of error, such as coverage error, measurement error, or processing error, selecting a full census in the Pacific region means that there will be no sampling error in working gas estimates in the region. As a result, estimated measures of sampling variability for estimates in the Pacific region will equal zero.

Separate Data Collection for Salt and Nonsalt Facilities in the South Central Region

The estimates from the 2018 sample reflect a change in data collection practices for a few operators reporting working gas in the South Central region on the Form EIA-912, Weekly Underground Natural Gas Storage Report. Previously, for most respondents operating both salt and nonsalt fields in the South Central region, EIA internally classified the entire inventory reported based on the field type having the preponderance of the volume. The most recent version of the Form EIA-912, approved by the U.S. Office of Management and Budget in late 2017, requires storage operators in the South Central region to separately report the volumes of working gas held in salt facilities and nonsalt facilities. By February 2018, all survey respondents to the Form EIA-912 had transitioned to the newest version of the form. Implementation of the new sample allows EIA to more accurately allocate storage between salt and nonsalt fields. As a result, the weekly estimates in the salt and nonsalt regions will reflect the new (2018) sample selection and the effect of the new data collection practices.

A Comparison of Results from the Old (2015) and New (2018) Samples

The EIA-912 system has collected data from both samples since the week ending February 2, 2018, to allow for comparison and assessment of the results before implementation of the 2018 sample on September 13, 2018. EIA has conducted a comparison of WNGSR published data using the 2015 sample with results from the new 2018 sample using the monthly data from the EIA-191 survey. As a census survey, the monthly EIA-191 data generally are considered more accurate because they include finalized data adjustments from some respondents and do not include sampling error.

A comparison of national results by month shows comparable performance between the weekly estimates under the old (2015) and new (2018) samples compared with monthly data in each month (Table 2). The average difference of the weekly estimate using the new sample compared with the monthly volumes was only -4.75 Bcf (billion cubic feet), or -0.31%, and only slightly smaller than the difference from the old (2015) sample over the period analyzed. These differences are within the target relative standard error of no more than 5% of the working gas estimate.

Table 2. Published Monthly Data for Working Gas in Underground Storage in the Lower 48 States Compared With Interpolated Estimates from the Weekly EIA-912 Survey Using the Old (2015) and New (2018) Samples (Billion cubic feet) (Lower 48)
Date Storage Volumes Differences from
Published Monthly Data (Bcf)
Differences from
Published Monthly Data %
Published Monthly 2015 Sample 2018 Sample 2015 Sample 2018 Sample 2015 Sample 2018 Sample
Feb-18 1,638 1,641 1,648 3 10 0.18% 0.61%
Mar-18 1,358 1,351 1,355 -7 -3 -0.52% -0.22%
Apr-18 1,396 1,381 1,380 -15 -16 -1.07% -1.15%
May-18 1,813 1,804 1,803 -9 -10 -0.50% -0.55%
Average n/a n/a n/a -7 -4.75 -0.45% -0.31%
Note: The 2015 sample results are based on published weekly estimates. The 2018 sample estimates reflect the effect of all data submissions, including unpublished revisions of less than 4 Bcf. Differences are calculated as weekly volumes minus monthly.
Source: U.S. Energy Information Administration. Published monthly data in the Differences Between Monthly and Weekly Working Gas in Storage, August 2018; 2015 and 2018 sample results are linearly interpolated values based on weekly volumes from the EIA-912 estimation system.

The new (2018) sample resulted in generally comparable differences from the monthly data at the regional and national levels (Table 3). For the Lower 48 states, the average absolute value of differences from the monthly data increased just slightly with the new (2018) sample for the period analyzed, while also performing better in some regions. However, both the new and old samples performed very well compared with the monthly census estimates published in the Natural Gas Monthly. In addition, the maximum absolute differences from the new (2018) sample are comparable.

Table 3. Summary Statistics Comparing Published Monthly Natural Gas in Storage Data with Interpolated Estimates from the Weekly EIA-912 Survey System using the Old (2015) and New (2018) Samples, by region, February 2018 to May 2018
Average Absolute Differences (Billion Cubic Feet)
East Region Midwest Region Mountain Region Pacific Region South Central Region Total Lower 48
2015 Sample 1.50 4.75 1.50 1.75 4.50 8.50
2018 Sample 3.00 4.75 0.75 0.75 5.50 9.75
Average Absolute Value of Differences (percentage)
East Region Midwest Region Mountain Region Pacific Region South Central Region Total Lower 48
2015 Sample 0.51% 1.56% 1.50% 0.91% 0.68% 0.55%
2018 Sample 1.02% 1.56% 0.75% 0.39% 0.83% 0.63%
Note: The 2015 sample results are based on published weekly estimates. The 2018 sample estimates reflect the effect of all data submissions, including unpublished revisions of less than 4 Bcf. Differences are calculated as weekly volumes minus monthly volumes.
Source: U.S. Energy Information Administration. Published monthly data in the Differences Between Monthly and Weekly Working Gas in Storage, August 2018; 2015 and 2018 sample results are linearly interpolated values based on weekly volumes from the EIA-912 estimation system.

Working Gas Stocks. The EIA-912 system provided weekly estimates based on the two samples for the period from February 2, 2018, through August 3, 2018. For the Lower 48 states, the average difference between the two samples is less than 1 Bcf (Table 4), with a maximum of 8 Bcf.

Table 4. Summary Statistics Comparing Weekly Estimates of Natural Gas in Storage Based on the Old (2015) and New (2018) Samples, by region, February 2, 2018 to August 3, 2018 (Billion cubic feet)
East Region Midwest Region Mountain Region Pacific Region South Central Region Total Lower 48
Average Differences -1.70 -3.85 -0.48 1.37 0.44 -0.81
Average Differences % -0.47% -1.03% -0.43% 0.65% 0.06% -0.05%
Note: The 2015 sample results are based on published weekly estimates. The 2018 sample estimates reflect the effect of all data submissions, including unpublished revisions of less than 4 Bcf. Differences are calculated as volumes based on 2018 sample minus volumes based on 2015 sample.
Source: U.S. Energy Information Administration.

The transition to the new (2018) sample

The new sample will be phased in by revising prior published data over eight weeks to transition from the old (2015) sample to the new (2018) sample. Working gas values will be calculated as a weighted average of the estimates from the current sample and the new sample during the transition period. The published stock values will be calculated as a weighted average of the estimates based on both samples, with each ensuing week using a decreasing weight on the estimates from the old sample and an increasing weight on the estimates from the new sample, as shown in Table 5.

Transition period estimates based on this weighted average combination will be released as revisions to the (WNGSR) historical database on Monday, September 10. Revisions to estimated measures of sampling variability during the transition period also will be published at this time. Until these revisions are officially published on September 10, working gas stock estimates will be based solely on the old (2015) sample. Revisions to the historical data also will be noted in the revisions history spreadsheet on the EIA website. The first WNGSR report based entirely on the new sample (2018) will be for the week ending September 7, 2018, which will be published on September 13, 2018.

Table 5. Weights for combination of old (2015) and new (2018) samples
Week Ending Weight on Estimated
Stocks for 2015 Sample
Weight on Estimated
Stocks for 2018 Sample
7/13/2018 7/8 1/8
7/20/2018 3/4 1/4
7/27/2018 5/8 3/8
8/03/2018 1/2 1/2
8/10/2018 3/8 5/8
8/17/2018 1/4 3/4
8/24/2018 1/8 7/8
8/31/2018 0 1