|
Data Collection Procedures
Every Monday, retail on-highway diesel prices are collected by telephone and fax from a sample of approximately 350 retail diesel outlets, including truck stops and
service stations. The data represent the price of ultra low sulfur diesel (ULSD) which contains less than 15 parts-per-million sulfur and low sulfur diesel (LSD)
which contains between 15 and 500 parts-per-million sulfur. EPA requires that all on-highway diesel sold be ULSD by December 1, 2010 (September 1, 2006 in California). ULSD was
projected to be available at many retail locations by October 15, 2006 (September 1, 2006 in California).
Historically, EIA collected the price of on-highway fuel without distinguishing
the sulfur level. In January 2007, the weekly on-highway diesel price survey
began collecting diesel prices for LSD and ULSD separately. On February 19,
2007 EIA began releasing the detailed sulfur price information that was
determined to be accurate. EIA will continue to monitor the accuracy of the LSD
and ULSD prices throughout the transition period and will discontinue release
of LSD prices when those prices are no longer considered to be accurate due to
the low volumes being sold.
The prices reported in this survey are subjected to automated edit checks during data collection and data
processing. Data flagged by the edits are verified with the respondent.
Imputation is used for companies that can not be contacted. The average survey
response rate for 2006 was 99.1%.
The price estimates for ULSD and LSD each week are obtained using simple averages
at the sampling cell level. For publication regions that constitute a combination of sampling cells, the volume
of on-highway diesel sold in that region as reported in the EIA-782 monthly
survey is used to weight the sampling cells and obtain publication level prices
for ULSD and LSD. The average price of all types of diesel is calculated by
weighting the average prices of ULSD and LSD according to the number of outlets
in each area selling each of the products as reported for that week in the
weekly on-highway diesel price survey.
The average prices are released by 5:00 P.M. Monday,
except on government holidays, in which case the data are released on Tuesday
(but still represent Monday's price). These data are made available through
EIA’s hotline (202-586-6966), EIA’s web page, and through EIA’s E-mail
notification, regular and wireless.
Statistical Methodology
The sample design for the weekly diesel price survey was a
two-phase design. The first phase constituted construction of a frame of 2,207 company-State
units (CSUs) from the combination of two sample cycles of the EIA-782A and
EIA-782B surveys that collected monthly petroleum products’ sales at the State
level. For sampling purposes, any combination of State and company where diesel was
sold through retail outlets as reported on the EIA-782 surveys defined a CSU,
the sampling unit.
For the second phase, a sub-sample of the 2,207 CSUs from
phase 1 was selected using probability proportional to size (PPS). The measure
of size for each of the two sample cycles separately was normalized using the
annual State sales’ volumes from the monthly survey divided by the unit's
probability of selection in the monthly survey. The two cycles’ normalized measures were then added to form one size.
Each CSU in the frame, therefore, had a size,
and the sum of the CSU sizes within each sampling cell equaled the allocation for
each cell. Within the second phase was a
second stage to identify and select the actual outlets reporting for the
company. This identification was done by
contacting the sampled companies and asking them to provide the names,
addresses and telephone numbers for truck-stops and retail outlets selling
diesel fuel for each sampled CSU. Additionally,
they were also asked to provide the proportion of diesel fuel sold through
truck-stops for each sampled CSU. This
proportion was used to preserve and control the number of truck stops and
retail stations selected for each CSU.
Sample allocations for each sampling cell were calculated
using the average standard errors across reporting periods for the previous
year of weekly diesel fuel survey prices for each of the sampling cells. The
sample size was determined for each cell by the formula: n'
= (e/t)2 n, where t was the
targeted standard error, n was the previous sample size for the cell, e was the
average of the previous sample's weekly standard errors, and n' was the new
sample cell allocation. In addition, a second allocation based on proportional
representation within the next larger aggregation cell to which the original
sampling cell would contribute was also obtained. For example, the PADD IB sampling cell
contributes to the PADD I cell. The maximum of the two allocations for each cell
was then designated as the cell allocation for the sampling cell. The sample was targeted at the time of
selection to yield standard errors of 1 cent for the U.S.,
the East Coast, the Midwest, and the Gulf Coast. The remaining areas were targeted at the 1.5 cent level. These targeted standard
errors were based on the price levels that were realized at the time of
selection which translated the errors to roughly a 1%/1.5% Coefficient of
Variation, respectively. This allocation
procedure yielded a targeted sample size of 350 truck stops and retail outlets
for the diesel fuel survey.
Further details of this design are contained in a published paper that can be found at: http://www.eia.doe.gov/pub/oil_gas/petroleum/data_publications/weekly_on_highway_diesel_prices/current/html/2cycasr.htm.
|