|Weekly Natural Gas Storage
EIA Home > Natural Gas > Natural Gas Analysis Publications > Weekly Natural Gas Storage EIA-AGA Comparison
|COMPARATIVE ASSESSMENT OF EIA and AGA WEEKLY STORAGE
This report was released on May 9, 2002, and reflects weekly storage estimates that were current as of that date.
Natural gas stocks are an important indicator of natural gas supply and demand conditions. Working gas in storage is one of the few timely indicators of the relative scarcity or abundance of gas in the market. Therefore, industry and market participants, as well as policymakers, watch stock movements closely.
The American Gas Association (AGA) began publishing weekly natural gas storage data in 1994. In October 2001, the AGA announced plans to discontinue its survey due to resource considerations. The Secretary of Energy announced in October 2001 that the Energy Information Administration (EIA) would undertake a weekly survey of natural gas storage operators and would begin publishing weekly regional and national estimates of natural gas in storage. The EIA already publishes monthly and annual natural gas storage data.
The EIA has collected survey data and produced weekly estimates since mid March. This report is intended to aid data users by examining differences between the EIA and AGA weekly surveys and comparing the results of the two surveys for the brief period of time in which they overlapped.
Comparison of AGA and EIA Weekly Storage Surveys and Procedures:
EIA's survey instrument, the form EIA-912, "Weekly Underground Natural Gas Storage Report," was developed to collect information on weekly inventories of working gas in underground storage facilities. As was true for the AGA survey, the purpose of EIA's survey is to compute and publish weekly estimates of working gas in storage for three regions of the United States, and the total for the United States. The AGA originally established a set of three regions based on similarities in the way companies tend to utilize their storage facilities in meeting the demands of their gas customers. The Consuming Region East includes many high gas-consuming states, has the largest proportion of storage capacity, and has the bulk of the nation's storage activity. The EIA adopted the same regional framework to maintain a useful comparability in the two series.
The two survey systems group the following states with storage facilities into three geographic regions (Figure 1).
Producing Region: Alabama, Arkansas, Kansas, Louisiana, Mississippi, New Mexico, Oklahoma, and Texas;
Consuming Region East: Illinois, Indiana, Iowa, Kentucky, Maryland, Michigan, Missouri, Nebraska, New York, Ohio, Pennsylvania, Tennessee, Virginia, West Virginia;
Consuming Region West: California, Colorado, Minnesota, Montana, Oregon, Utah, Washington, and Wyoming.
The two survey systems have a number of similarities in the primary characteristics and procedures (Table 1). The key differences between the two surveys involve: (1) survey sampling and reporting procedures, and (2) the methodology for estimating regional working gas volumes.
Survey Respondents and Procedures:
EIA-912 survey respondents were selected from the list of respondents to Form EIA-191, "Monthly Underground Natural Gas Storage Report." All operators of underground natural gas storage fields in the United States complete the Form EIA-191. To prepare the sampling frame for the EIA-912 for each region, reported volumes of working gas in storage as reported on the Form EIA-191 for the end of October 2001 were aggregated by storage operator and AGA region. A stratified sample of companies was selected from the list of operators to achieve a target standard error of the estimate of working gas in storage that was no greater than 5 percent of the estimate for each region. (Two operators in the Eastern Region and three in the Producing Region were removed from the frame for sampling purposes because they showed no variation in inventories for the preceding two years. They will be represented in estimation by constants. See the estimation section of this report.) (See the EIA report, Methodology for EIA Weekly Underground Natural Gas Storage Estimates, for further discussion of the sample selection process.) The respondents to the AGA survey were volunteers from among AGA members and other non-member storage operators.
Both EIA and AGA have measures of survey coverage statistics for their respective samples. The table below shows the sample survey coverage estimates for the two surveys. The overall coverage is higher for the EIA survey compared to that of the AGA. The coverage is highest in the Consuming Region East and lowest in the Consuming Region West for both EIA and AGA.
AGA and EIA survey data employ different computation methodologies. (For more detailed descriptions of these methodologies, see each organization's methodology report:
AGA: American Gas Storage Survey Procedures and Methodology 2001 Update
EIA: Methodology for EIA Weekly Underground Natural Gas Storage Estimates
From the description provided in AGA's Issue Brief 2001-03, one can derive the following AGA formula for estimated working gas volumes in a given region in a particular week:
AGA Regional Estimate =
n = number of respondent companies in the region who reported for the collection week
wwg= volumes of working gas in storage reported by respondent companies in the region for the collection week
xwg= maximum working gas held in storage by respondent companies in the region anytime in the period 1992-2001
XWG= maximum working gas held in storage by companies in the region anytime in the period 1992-2001
EIA Regional Estimate=
mwg= working gas in storage for operators in the sample in the region for the reference month
wwg= working gas in storage in the region for the collection week reported by sampled operators
K= constant volumes of working gas for operators in the region excluded from the sample frame because they showed no variation in inventories for the preceding two years
N= number of companies in the sample frame
n= number of operators in the sample in the region who reported for the collection week
The two methodologies embody different philosophies in approach: the AGA utilizes the concept of "percent full," while the EIA relies on scaling up reported information to account for companies who are not in the sample. However, a comparison of the two equations above shows that, except for the constant adjustment factor, the calculation methods of EIA and AGA are quite similar. Nevertheless, because the implementation depends heavily on the respective ratios by which the reported data are adjusted in the two methodologies, the results are likely to be different.
Some key differences between the two methodologies are as follows:
2. AGA's sample represented companies who volunteered to participate. Changes to the sample occurred as participation patterns changed. The EIA-912 sample will be reviewed annually, and changes will be introduced based on formal statistical sampling considerations. Further, response to the EIA-912 survey is mandatory.
3. The AGA methodology did not provide a measure of the relative accuracy or level of confidence of its estimated regional volumes. EIA will calculate and publish a standard statistical measure (the "standard error of the estimate") of the probable range of variation due to sampling for each region. (The standard error applies only to that part of each region's estimate attributable to its noncertainty stratum.)
The EIA began collecting weekly underground storage data on March 18, 2002, requesting stocks as of 9:00 AM Friday, March 15. The overlap in EIA and AGA operations provides an opportunity to compare estimates directly. This assessment indicates that the two weekly data series are similar. Differences in the series and their possible causes are discussed below.
The EIA and AGA systems provided estimates for 7 weeks in common for the same 3 regions. The working gas estimates for the total U.S. indicate that the EIA values exceeded the AGA values by a slight amount-on average, 22 Bcf per week, which is a difference of 1.5 percent. This pattern is similar for the Consuming Region East, or East Region, where the average difference in volume is 13 Bcf, which is 1.7 percent. A comparison of stock estimates for the other 2 regions show larger differences. The average difference for the Producing Region is 65 Bcf, or 11.5 percent, while that for the Consuming Region West, or West Region, measures 56 Bcf, which is a relative difference of -25.4 percent (Figure 2).
The correlations between the EIA and AGA weekly storage series are quite high. The correlation coefficients are virtually 1.0 for both the total U.S. and the East Region. The lowest correlation value is 0.939 for the Producing Region, with the West Region at 0.967. These very high correlations indicate that the estimates in the EIA and AGA series tend to move closely.
The close movements between the 2 sets of storage estimates can also be seen in the changes between successive weeks recorded by the two weekly systems (Figure 3). The net changes in stocks for the total U.S. and the East Region are very close. In the Producing and West Regions, the direction of change for each series matches with one exception involving the Producing Region for the week ending April 5. In that week, the EIA series showed an increase of 4 Bcf for the Producing Region, while the AGA series had a slight decrease of 2 Bcf.
The high degree of similarity in the net change values based on the very high correlations and graphical comparisons is an important finding in light of the keen market interest in this weekly measure. While gas stocks have important implications for market conditions in the future, the net change in stocks is the primary volumetric indicator of very recent relative supply and demand conditions. It is those conditions that drive price movements and have implications for refinements to subsequent storage acquisition strategies.
Until EIA initiated its weekly survey, there had been no data series that is directly comparable to AGA's weekly data series. Prior to the EIA-912, EIA's storage data were collected only monthly via the EIA-191 survey. To compare EIA's monthly data series and AGA's weekly data series EIA analysts derived estimates for the end-of-month based on AGA estimates. These estimates were based on AGA weekly data and were produced by a linear interpolation technique whenever the end of a month did not coincide with the end of an AGA reporting week.
Based on the interpolated values for the AGA series, one finds that the EIA end-of-month data and the AGA-based end-of-month estimates for the U. S. as a whole over the period 1998-2001 differ on average by 4.4 percent (using EIA values as the base). The East Region and the Producing Region exhibited average percentage differences of 11.5 and 2.2 percent in the period from 1998-2001, respectively. The average difference for the West Region was -20.2 percent for the period. The comparison over time shows that prior differences between the series for the West Region were much smaller, with a larger difference developing in early 1998 (Figure 4). This large difference in the West Region is reflected also in the comparison of the two weekly series, which has been -20.2 percent. The specific causes for differences between the EIA and AGA estimates cannot be definitively established because confidentiality requirements preclude the disclosure of respondent information by either organization. Thus, a detailed comparative analysis of data is not possible.
The period of performance for the new EIA weekly underground storage series overlapped with the recently completed AGA series from mid March to the end of April. The two systems produced similar estimates of U.S. working gas storage levels with a high degree of correlation between them.
The differences between the regional and national stocks estimated by EIA and AGA are minimal for the U.S. as a whole and the East Region, where the bulk of storage activity is located. Some differences are present in the series for the West and Producing Regions. A key finding, however, is that the general pattern of weekly net stock changes are quite similar.
Factors behind differences between the EIA and AGA systems include sampling differences and non-sampling factors. Non-sampling factors include a slightly different formula for expansion of weekly sample reports and the fact that the EIA survey is collected on a mandatory basis.
Figure 2. Weekly Underground Natural Gas Storage Levels for Total U.S. and Three Regions. (Billion Cubic Feet)
Figure 3. Net Change in Weekly Underground Natural Gas Storage Levels for Total U.S. and Three Regions. (Billion Cubic Feet)
Figure 4. Comparison of Monthly Working Gas in Underground Storage Facilities for Total U.S. and Three Regions, 1994-2001