Part B

B. Collections of Information Employing Statistical Methods
B1. Description of the Survey Plan
Frames maintenance activities are conducted on a monthly and annual basis. Response to
all PSRS surveys is mandatory.
Monthly Frames Maintenance
The monthly frames maintenance procedures focus on examining several frequently
published industry periodicals that report changes in status (births, deaths, mergers and
changes in ownership) of petroleum facilities producing, transporting, importing, and/or
storing crude oil and petroleum products. These sources are augmented by articles in
newspapers, letters from respondents indicating changes in status, and information
received from survey systems operated by other offices. At the same time, the sample
frames for the weekly surveys are also updated. A sample control meeting is conducted
each month. This meeting focuses on changes in the current monthly data as it relates to
the weekly surveys, changes in the weekly surveys that impact the monthly surveys, and
changes in respondent reporting patterns. These meetings are conducted to assure a 90percent coverage of the total for each item collected and each geographic region for each
of the weekly surveys.
Annual Frames Maintenance
The annual frames maintenance is conducted to re-evaluate the consistency of frames
between the Forms EIA-810 and EIA-820.
2. Sampling Methodology and Estimation Procedures
The frame of respondents is considered as the universe for all PSRS surveys with the
exception of the following surveys:
EIA-800, “Weekly Refinery and Fractionator Report”
EIA-801, “Weekly Bulk Terminal Report”
EIA-802, “Weekly Product Pipeline Report”
EIA-803, “Weekly Crude Oil Stock Report”
EIA-804, “Weekly Imports Report”
EIA-805, “Weekly Terminal Blenders Report”
EIA-809, “Weekly Oxygenate Report.”
a. Sampling Frame
The EIA weekly reporting system, as part of the PSRS, was designed to collect data
similar to those collected monthly. The sample of companies that report weekly in the
WPSRS are selected from the universe of companies that report on the corresponding
monthly forms.

The sampling frame for Form EIA-800 “Weekly Refinery Report” includes refineries
reporting on Form EIA-810 “Monthly Refinery Report” as well as fractionators reporting
on Form EIA-816 “Monthly Natural Gas Liquids Report.” Monthly reports on Form
EIA-810 are required from operators of every operating and idle refinery located in the
50 States, District of Columbia, Virgin Islands, Puerto Rico, and other U.S. territories.
Monthly reports on Form EIA-816 are required from operators of every operating and
idle gas processing plant, fractionator, and butane isomerization plant located in the 50
States and the District of Columbia.
The EIA-801 sampling frame consists of all companies reporting ending stocks on the
EIA-815, “Monthly Terminal and Blender Report.” This includes every bulk terminal
operating company located in the 50 States, the District of Columbia, Puerto Rico, and
the Virgin Islands. A bulk terminal is primarily used for storage and/or marketing of
petroleum products and has a total bulk storage capacity of 50,000 barrels or more, and/or
receives petroleum products by tanker, barge, or pipeline. Bulk terminal facilities
associated with a product pipeline are included.
The EIA-802 sampling frame consists of all companies reporting on the EIA-812,
“Monthly Product Pipeline Report.” This includes all petroleum product pipeline
companies that transport refined petroleum products (including interstate, intrastate, and
intracompany pipeline movements) in the 50 States and the District of Columbia.
The EIA-803 sampling frame consists of all companies reporting on the EIA-813,
“Monthly Crude Oil Report.” This includes all companies that carry or store 1,000
barrels or more of crude oil. Included are gathering and trunk pipeline companies
(including interstate, intrastate, and intracompany pipelines), crude oil producers,
terminal operators, storers of crude oil (except refineries), and companies transporting
Alaskan crude oil by water in the 50 States and the District of Columbia.
The EIA-804 sampling frame consists of all companies reporting on the EIA-814,
“Monthly Imports Report.” This includes all companies, including subsidiary or
affiliated companies, that import crude oil or petroleum products (1) into the 50 States
and the District of Columbia, (2) into Puerto Rico, the Virgin Islands, Guam and other
U.S. possessions (Midway Islands, Wake Island, American Samoa, and Northern
Mariana Islands), and (3) from Puerto Rico, the Virgin Islands and other U.S. possessions
into the 50 States and the District of Columbia. Imports into Foreign Trade Zones
located in the 50 States and the District of Columbia are considered imports into the 50
States and the District of Columbia and must be reported.
The EIA-805 sampling frame consists of all companies reporting inputs and production
on the EIA-815, “Monthly Terminal Blender Report.” This includes all storage terminals
which produce finished motor gasoline through the blending of various motor gasoline
blending components, natural gas liquids, and oxygenates in the 50 States, the District of
Columbia, Puerto Rico, the Virgin Islands, Guam and other U.S. possessions.

The EIA-809 sampling frame consists of all companies reporting on the EIA-819,
“Monthly Oxygenate Report.” This includes all operators of facilities that produce fuel
ethanol located in the 50 States and the District of Columbia.
b. Sample Design
The sampling procedure used for all the sampled surveys is the cut-off method. In the
cut-off method, companies are ranked from largest to smallest on the basis of quantities
reported during some previous period. Companies are chosen for the sample beginning
with the largest and adding companies until the total sample covers approximately 90
percent of the total volumes for each item and each geographic region for which data may
be published.
To ensure 90-percent coverage of the total for each item collected and each geographic
region for each weekly survey, a sample control meeting is conducted each month. This
meeting focuses on changes in the current monthly frame due to sales, acquisitions,
mergers, new reporters, and reactivations. Companies are added or removed from the
weekly reporting samples based on the changes.
c. Imputation and Estimation Procedures
EIA-800 through EIA-809
After company reports have been checked and entered into the weekly database, values
are imputed for companies that have not responded, reported incomplete data, or reported
data that failed editing and could not be confirmed. The imputed values are calculated
using exponentially smoothed means of recent weekly reported values for this specific
company.
The equation for the exponential smoothing is:
Yt = α * yt + (1 - α) * Yt-1
where

Yt
yt
Yt-1
α

is the prediction for week t+1 (using data through week t),
is week t's reported value,
is the prediction for week t (using data through week t-1),
is a number between 0 and 1, chosen by survey/product/type

In the equation for exponential smoothing, the size of α controls the importance of last
week’s value relative to the aggregate of all weeks before that as represented by the
prediction for last week. For example, if α = 0.8, then last week’s value is much more
important in predicting this week’s value than all the previous week’s values are since the
weight of last week is 0.8 and the weight of the previous weeks collectively is 0.2. In
general, the α values for the expected means of the non-zero responses are low for
imports (last week is much less important than history) and much higher for production,
inputs and stocks.

The imputed values are treated like reported values in the estimation procedure, which
calculates ratio estimates of the weekly totals. First, the current week’s data for a given
product reported by companies in a geographic region are summed (weekly sum, Ws.)
Next, the most recent month’s data for the product reported by those same companies are
summed (monthly sum, Ms.). Finally, the most recent month’s data for the product as
reported by all companies, including adjustments
made in the monthly process, is summed (Mt). The current week’s ratio estimate for that
product for all companies, Wt, is given by:
Wt= (Mt / Ms) * Ws
The ratio (Mt / Ms) may be adjusted to account for very unusual events or industry
changes not yet reflected in the lagged monthly data. For example, the hurricanes in
September 2005 rendered the September data unrepresentative for purposes of applying
the ratio to the WPSR in December 2005.
This procedure is used directly to estimate total weekly inputs to refineries and
production. To estimate stocks of finished products, the preceding procedure is followed
separately for refineries, bulk terminals, and pipelines. Total estimates are performed by
summing over establishment types.
Weekly imports data are highly variable on a company-by-company basis or a week-toweek basis. Therefore, an exponentially smoothed ratio has been developed for weekly
imports. The estimate of total weekly imports is the product of the smoothed ratio and
the sum of the weekly reported values and imputed values.
For imports, the ratio is smoothed as follows:
Rt = α * rt + (1 - α) * Rt-1
where

is the smoothed ratio for week t+1 (using ratios through week t),
is week t's ratio of the most recent monthly total for all respondents
to the monthly total of respondents from the weekly sample,
Rt-1
is the smoothed ratio for week t (using ratios through week t-1),
α
is a number between 0 and 1, chosen by product but not by
PADD/Respondent ID.
Rt
rt

When Ms = 0, then rt is not defined for the week and the smoothed ratio is not updated,
that is, the previous smoothed ratio is used as the multiplier.
Forms EIA-810, 812, 813, 815, 816, and 819
In any survey, non-response can be a major concern because the effects can cause serious
bias in survey results. Non-response occurs whenever requested information is not
obtained from all units in a survey. Response rates for these surveys are generally 99 to
100 percent. Whenever survey responses are not received in time to be included in
published statistics, the data are imputed. Although imputing for missing data may not
eliminate the total error associated with non-response, it can serve to reduce the error.

The data reported in the previous month are used as imputed values for missing data.
Data are not imputed for the EIA-814 and 817 because these data series, by respondent,
are highly variable.
d. Macro Editing
EIA-800 through 809
After the flagged respondent data have been resolved, preliminary tables are produced
and used to identify anomalies. These tables show U.S. and PADD estimates for the
current week and the prior 3
weeks and also show year-ago data for the same week along with 4 week average.
Anomalies result in further review of respondent data which in turn may result in
additional flagged data and imputation.
EIA-810, 812 through 819
After the flagged respondent data have been resolved, preliminary tables are produced
and used to identify anomalies. These tables show U.S. and PADD estimates for the
current month and the prior 4 years. Anomalies result in further review of respondent
data which in turn may result in additional flagged data and imputation. In addition, other
adjustments are made to aggregate data from time to time. For example, unusual
industry conditions, including fuel transitions, business practice shifts, or hurricane
dislocations, may generate reporting anomalies and require adjustments. Measurement
error and frame deficiencies may occasionally result in inconsistencies when individual
respondent data are aggregated to publication levels and require adjustment. Monthly
supply data are reviewed throughout the year and some estimates may be replaced with
newly available or resubmitted respondent data in the Petroleum Supply Annual (PSA).
e. Data Accuracy
The reliability of data is subject to two types of possible errors, non-sampling errors and
sampling errors. Sampling errors occur because observations are made only on a sample,
not on the entire population. Non-sampling errors can be attributed to many sources in
the collection and processing of data such as, response coverage; a difference in
interpretations of definitions or questions; mistakes in recording or coding the data from
respondents; and other errors of collection and estimation. The accuracy of survey results
is determined by the joint effects of sampling and non-sampling errors.
For the monthly surveys EIA-810, 812, 813, 815, 816, 817, 819 and the annual survey
EIA-820, there is no sampling error because the surveys are based upon a complete
census of the frame. However, response error is the major factor affecting the accuracy
of data. Response error, or reporting error, is the difference between the true value and
the value reported on a survey form.
To aid in detecting and minimizing reporting errors, automated editing procedures are

used to check current data. These checks include verifying the current data for
consistency with past data, for internal consistency (e.g. totals equal sum of parts),
examining orders of magnitude, and cell position. Data elements that fail edit criteria are
flagged.
Monthly and weekly data are compared on a regular basis to rectify discrepancies in data.
In addition, a comparison of PSRS data with sources outside of the Petroleum Division is
performed each year. The results of this effort have been published in the feature article,
“Accuracy of Petroleum Supply Data, 2007” located on the Internet.

B3. Methods to Maximize Response Rate
To maximize response rates, forms are designed to be easily completed, and instructions
are written to be clear, concise, and easily understood. Forms and instructions are made
available on our website. Survey nonrespondents are contacted by telephone to discuss
the requirement to file and any problems or questions that are delaying filing. Follow-up
letters regarding the failure to file may be emailed to respondents. Specific schedules are
followed for telephone calls and letters to nonrespondents for the various surveys. Every
effort is made to assist respondents in completing the survey and submitting them in a
timely manner. The response rate for weekly surveys averages above 97 percent. The
response rate for monthly surveys is 99 to 100 percent.
4. Test Procedures
The petroleum surveys are established continuing surveys. Modifications to all of the
existing forms were made by the EIA staff in conjunction with discussions with industry
representatives and consultations through the Federal Register notice discussed earlier.
These actions served as a test of the availability of data and the clarity of instructions of
the survey forms, as well as the proposed modifications.
5. Questions
Questions regarding the Petroleum Supply Reporting System may be directed to Sylvia
Norris of the Department of Energy, Energy Information Administration, at (202) 5866106. Questions regarding the EIA Forms Clearance process should be directed to Grace
Sutherland at (202) 586-6264.